No title

Journal of Semantics 22: 1–2 doi:10.1093/jos/ffh028

A Note from the Parting Editor

Peter Bosch University of Osnabru¨ck

A Note from the New Editor When the Journal of Semantics was launched in 1982, Peter Bosch was among the founding editors, and he became managing editor three years later. Without Peter, the Journal wouldn’t have existed, let alone have become one Journal of Semantics, Vol. 22, No. 1, Ó Oxford University Press 2005; all rights reserved

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

As some readers may have noticed the Journal of Semantics has a new Managing Editor, a new team of Associate Editors, and there are some changes on the Editorial Board as well. I am leaving the post of Managing Editor at the end of the calendar year 2004, and I want to thank everyone who has helped, over the past twenty years, to make the Journal one of the more respectable places for publications in natural language semantics: Its authors, reviewers, members of the editorial board, and, of course our publisher and subscribers. It has been these people who have made the journal what it is. The reason I am leaving as Managing Editor is simply that I want to find more time for my own research, while, at the same time, the number of submissions to the Journal, and thus the workload of the Managing Editor, keeps increasing. Having been in this office for twenty years now, since 1985, I feel that I have done my fair share of service to the community and that it would not only be good for my research if I stopped looking after the Journal, but that it would also be good for the Journal to see a change in this position. Let me add that I very much enjoyed managing the Journal and that I’m very happy that I can hand it over to Bart Geurts and his team, all of whom are excellent semanticists, with a very good feeling also for interdisciplinary developments, and very fit to run the Journal of Semantics. I want to congratulate Bart Geurts and his team on their new responsibility and wish them all the best for the future of the Journal. Finally, and perhaps also as an excuse for the new editors, I must confess that nearly all of the contributions that are appearing in Volume 22 (2005) were already reviewed and accepted for publication during my period of office. So all complaints regarding the contents of this volume can still be directed to me, every other business is from now with the new Managing Editor.

2 Editorial Note of the principal venues in the field of natural language semantics. It is understandable, therefore, that I’m taking over his job with a fair amount of trepidation. Fortunately, I will be supported by a formidable new team of associate editors, and maybe the eight of us together will just manage to do what Peter did almost single-handedly. Bart Geurts University of Nijmegen

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Journal of Semantics 22: 3–51 doi:10.1093/jos/ffh016

There and Back Again: A Semantic Analysis SIGRID BECK University of Connecticut and Universita¨t Potsdam

Abstract

1 INTRODUCTION Sentences like (1) are known to be ambiguous—the two readings possible for (1) are paraphrased in (1#). (1) Bilbo opened the door again. (1#) a. Bilbo opened the door, and that had happened before. b. Bilbo opened the door, and the door had been open before.

(repetitive) (restitutive)

Analyses of the ambiguity of (1) have alternatively been structural (the ambiguity arises because two different syntactic structures are possible for (1): McCawley 1968; Stechow 1995, 1996, among others), or lexical (the ambiguity arises because again has two different meanings: see, for example, Fabricius-Hansen 1983, 2001; Kamp & Rossdeutscher 1994). In this paper, the availability of restitutive readings is tested across languages. An important aspect of the cross-linguistic survey is that I test two different types of predicates. One type is illustrated by (1) above: lexical accomplishment predicates. Another type of predicate is the combination of an activity verb with a directional PP, which I call a goal PP construction (following Beck & Snyder 2001a,b). An example is given in (2a).
The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This paper presents a cross-linguistic survey of the interpretations that decomposition adverbs like again permit. The survey distinguishes two different types of predicates that are combined with again: lexical accomplishments like ‘open the door’, on the one hand, and combinations of a motion verb with a directional Prepositional Phrase (PP) on the other (for example, ‘walk to the summit’). There is systematic crosslinguistic variation with the latter type of predicate: a language only permits a restitutive reading of again with such a predicate if the language has resultative constructions. I argue that in these languages the PP can function as a result phrase. In languages without resultatives, the PP cannot be a result phrase, and a restitutive reading is impossible. The data support an analysis of restitutive again that is sensitive to the presence of a result phrase in the syntax.

4 There and Back Again: A Semantic Analysis (2) a. Bilbo walked to the hall. b. Bilbo walked to the hall in ten minutes. It has been observed (in particular by Aske 1989) that there is crosslinguistic semantic variation with this type of predicate. In English, walk to the hall behaves like an accomplishment predicate, as modifiability with a temporal delimiter in (2b) shows. On the other hand, this is impossible for example in Spanish: (3) Bilbo anduvo hasta la caverna Bilbo walked up-to the hall

(*en diez minutos). (in ten minutes)

(4) (4#)

(G)

Bilbo walked to the hall again. a. Bilbo walked to the hall, and he had done that before. b. Bilbo walked to the hall, and he had been there before.

(repetitive) (restitutive)

Sentences like (4), in which a goal PP construction is combined with again, are ambiguous between repetitive (4#a) and restitutive (4#b) in a language only if that language permits resultative constructions like (5). If a language does not permit resultatives, then the analogue of (4) only has the repetitive reading (4#a).

(5) Thorin hammered the metal flat. Restitutive readings with lexical accomplishment predicates do not follow the same cross-linguistic pattern. They are available in languages that do not permit restitutive readings with goal PP constructions. We will see that these results have consequences for the semantic analysis of goal PP constructions on the one hand, and for the analysis of again on the other. More specifically, I propose that there is parametric variation in the semantics of goal PP constructions, to the effect that they express accomplishments in languages that have resultatives, but nonaccomplishment change of state predicates in languages that do not have resultatives. Moreover, I argue that our cross-linguistic findings support a theory of again according to which the repetitive/restitutive ambiguity is structural.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In English, the combination of a goal PP construction with again permits a restitutive reading (cf. (4)). This is not the case in Spanish and in many other languages. It turns out that there is systematic crosslinguistic variation in this respect. Generalization (G) summarizes the most important empirical result.

Sigrid Beck 5

The structure of the paper is as follows. In section 2, I present the two competing analyses of again (structural and lexcial). In section 3, I introduce our test case, goal PP constructions, and show that there is systematic cross-linguistic variation in their interpretation, correlated with the availbility of resultative constructions. Section 4 presents the cross-linguistic data on restitutive again. I provide an analysis of those data in section 5, and explain why they favour the structural ambiguity analsysis. Section 6 concludes the paper. 2 ANALYSES OF RESTITUTIVE AGAIN

2.1 A structural analysis (Stechow 1995, 1996) This subsection introduces Stechow’s (1995, 1996) analysis of the repetitive/restitutive ambiguity, which is the analysis we will eventually adopt and support. I first discuss Stechow’s analysis for syntactically complex predicates1 like resultatives, which is where it receives its intuitive support from (section 2.1.1.). Section 2.1.2. extends the analysis to lexical accomplishment predicates like open in (1). In section 1 I use the term ‘complex predicate’ for predicates that consist of more than one predicative expression. For the time being, these are resultative constructions. My intention is to use the term in the sense of Snyder (2001)—see section 3. The reader will note that we follow Stechow (1995) in assuming structures with a small clause for resultatives. There has been considerable debate in the syntactic literature as to whether a small clause analysis (e.g. Stowell 1983; den Dikken 1995) or a complex predicate analysis (e.g. Larson 1988) is more appropriate. I do not attempt to contribute to this debate and assume Stechow’s small clause analysis for semantic convenience.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This section introduces representatives of the two types of analysis of the repetitive/restitutive ambiguity. One kind of analysis views the ambiguity as purely structural. In these analyses the meaning of again is constant. Modern representatives are Stechow (1995, 1996) and Rapp & Stechow (1999); they are anticipated in Generative Semantics (McCawley 1968). Alternatively, again has been analysed to be ambiguous between a repetitive reading and a ‘restitutive’ reading (e.g. Fabricius-Hansen 1983, 1995, 2001; Kamp & Rossdeutscher 1994; Ja¨ger & Blutner 2000) (Dowty’s 1979 analysis in the framework of Montague Grammar shares aspects of both; see Stechow 1995, 1996 for discussion of Dowty’s proposal). I take Stechow (1995, 1996) and Fabricius-Hansen (1995, 2001) to be appropriate representatives of the two main theoretical approaches in this ongoing debate. They will be discussed in turn. The two types of analysis lead to different expectations regarding the prerequisites for availability of a restitutive reading. We will work out these differences to make use of later, for the analysis of our cross-linguistic data.

6 There and Back Again: A Semantic Analysis 2.1.3. I report findings by Rapp & Stechow (1999) that indicate that a lexical parameter for the adverb is involved in determining availability of restitutive readings.

(6) (6#)

Sally hammered the metal flat again. a. Sally hammered the metal flat, and that had happened before. b. Sally hammered the metal flat, and the metal had been flat before.

(repetitive) (restitutive)

Before we can explain this ambiguity, we need to develop a semantic analysis of resultatives. We will follow Stechow (1995), who suggests that a special interpretation priniciple is at work that combines the result phrase (the Adjective Phrase (AP) flat) with the verb (here: hammer), by introducing a causal element into the interpretation. We should note that the same basic intuition, that the causal component in this construction is not the meaning of any lexical item in the construction, and that this component is what permits us to combine the verb with the result phrase, has been expressed prior to Stechow in different theoretical frameworks (see, in particular, Goldberg 1995; Jackendoff 1990; Levin & Rappaport Hovav 1995). I concretely report and use Stechow’s analysis, which relies on the same theoretical background as this paper. Stechow (1995) proposes that (7a) has the Logic Form (LF) in (7b). The verb hammer combines first with a small clause and then with an object Noun Phrase (NP). The object NP ‘the metal’ raises to semantically bind the empty pronominal (PRO) subject of the small clause.2 (7) a. Sally hammered the metal flat. b. [[the metal] [1[VP Sally [V# t1 [V# hammered [SC PRO1 flat] ]]]] 2 Logical Forms are presented in the notation of Heim & Kratzer (1998). The only unusual aspect of these LFs is the adjunction of an index (1 in the example) to the sister of the moved phrase. This is interpreted as predicate abstraction, semantically binding all variables with that index (in the example, t1 and PRO1).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

2.1.1 Complex predicates. The basic idea behind the structural analysis of the repetitive/restitutive ambiguity is that again always indicates repetition. What varies is just what is repeated: the whole event described, or just its result state. Example (6) with a resultative construction provides a good illustration of this approach. It can mean either (6#a) or (6#b). What is repeated in (6#a) is the whole event of Sally hammering the metal flat; in (6#b), just the result state of the metal being flat is repeated.

Sigrid Beck 7

(8)

Principle (R) (Stechow (1995)): If a ¼ [Vc SCb] and b# is of type Æi, tæ and c# is of type Æe, . . . Æe, Æi, tæææ (an n-place predicate), then a# ¼ kx1 . . . kxnke.c#e(x1) . . . (xn) & $e#[BECOMEe#(b#) & CAUSE(e#)(e)]

I follow Stechow (1995, 1996) in assuming the standard semantics for CAUSE from Lewis (1973) and from Dowty (1979) for BECOME—informal versions adapted to our framework are given in (9) and (10) (see Lewis, Dowty and Stechow for discussion and more precise definitions). (9)

[[BECOME]] (P)(e) ¼ 1 iff

(10) [[CAUSE]] (e#)(e) ¼ 1 iff

e is the smallest event such that P is not true of the prestate of e but P is true of the result state of e e# occurred, e occurred and if e hadn’t occurred then e’ wouldn’t have occurred

Let’s apply the principle to our example structure in (7b); the relevant substructure is (11). The result of applying principle (R) is a two-place predicate, which will be true of two individuals and an event iff the event is a hammering of the first individual by the second and there is another event which is a becoming flat of x1 (the subject of the small clause) caused by the first event. The interpretation of the whole structure in (7b) 3 I represent the meanings of natural language expressions as translations into a standard kcategorial language. The implicit event argument of predicates will often be written as a subscript for readability. When our formal language contains non-standard expressions (like CAUSE and BECOME), a meaning for those will be provided.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The LF in (7b) is not interpretable by standard mechanisms of composition. The transitive verb hammer looks for an argument of type Æeæ, and finds instead a propositional category (the small clause). Stechow proposes that a special principle of interpretation is responsible for the combination of the verb and the small clause. A variant of his principle is given in (8). I have made some formal changes in the principle which come from my use of a Davidsonian event semantics (Davidson (1967)); I use Æiæ as the semantic type of events and I will call properties of events, Æi, tæ, propositions. The basic intuition behind the principle is the same as Stechow’s: the resultative is true iff an event of the kind denoted by the verb causes a becoming of the small clause proposition. Thus the principle combines the verb with the small clause by inserting a CAUSE BECOME component that glues them together.3

8 There and Back Again: A Semantic Analysis will add binding of the PRO subject of the small clause and filling of the individual argument slots by the usual interpretation mechanisms. (11) [V# hammered [SCPRO1 flat] ] / kxkyke.hammere(x)(y) & $e#[BECOMEe#(ke$.flate$(x1)) & CAUSE(e#)(e)] (12) [[the metal] [1[VP Sally [V# t1 [V# hammered [SCPRO1 flat] ]]]] / ke.hammere(the_metal)(S) & $e#[BECOMEe#(ke$.flate$(the_ metal)) & CAUSE(e#)(e)] (12#) Sally’s hammering the metal caused it to become flat.

4 Note that principle (R) cannot be freely available as an interpretation principle throughout the grammar, even in those languages that by assumption have it. (i) below, for example, cannot have a resultative meaning of the kind described. I hypothesize that application of principle (R) is limited to ‘morphological’ domains within syntax, as the tie to compounding, found by Snyder (2001) and reported in section 3, suggests. In this case, the node dominating hammer and the Small Clause in (7b) would more appropriately labeled V0.

(i)

a. Otto snored that his roomate left. b. Otto’s snoring caused his roomate to leave.

5 An anonymous reviewer observes that under this analysis, one might expect that (ia) be able to mean (ib). I concur that this is a caveat of this analysis of resultatives, as it stands. It is noted in Beck & Johnson (2004) that there is no good theory of the distribution of empty pronominal elements in such constructions. I can only reiterate the need for such a theory here.

(i)

a. Peter drank [PRO under the table]. b. Peter drank himself under the table.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Thus the resultative will be true of an event e iff e is an event of Sally hammering the metal which causes another event that is the metal becoming flat. This is an intuitively appropriate description of the semantics of the resultative construction. Note that the principle adds a CAUSE BECOME component to the semantics; according to Dowty (1979) this is what characterizes accomplishments predicates. I generally use the terms ‘activity’ and ‘accomplishment’ in the sense of Vendler (1967) and Dowty (1979)—see Dowty for various criteria for these aspectual categories of predicates. Principle (R) thus turns the event type denoted by the verb in the construction into an accomplishment predicate when the result phrase is added. Stechow (1995), conjectures that this interpretation principle may be available in some languages, but not others. This could account for the variation in the acceptability of resultatives across languages. I will come back to this point in section 3 when we talk about cross-linguistic variation in the availability of resultative constructions.4,5 Before we return to the repetitive/restitutive ambiguity, a comment on the notion of causation. Lewis (1973) distinguishes between

Sigrid Beck 9

(13)

(14) (14#) (15) (15#)

a. [VP [the metal] [1[VP [VP Sally [V# t1 [V# hammered [SCPRO1 flat] ]] again]]] b. [VP [the metal] [1[VP Sally [V# t1 [V# hammered [SC [SCPRO1 flat] again]]]]] ke$.againe$(ke.hammere(t_m)(S) & $e#[BECOMEe#(ke*.flate* (t_m)) & CAUSE(e#)(e)]) Once more, Sally’s hammering the metal caused it to become flat. ke.hammere(t_m)(S) & $e#[BECOMEe#(ke$.againe$(ke*.flate* (t_m)) & CAUSE(e#)(e)]) Sally’s hammering the metal caused it to become once more flat.

I assume that again makes (roughly) the semantic contribution in (16). It denotes a relation between a predicate of events and an event. It presupposes that there was a preceding event of which the predicate is true; it asserts that the predicate is true of the event (see Stechow 1995, 1996; Fabricius-Hansen 2001; Klein 2001 and references in these papers for discussion and further considerations; the semantics in (16) will serve for our purposes). (16)

[[again]](PÆi, tæ)(e)

¼ 1 iff P(e) & $e#[e# < e & P(e#)] ¼ 0 iff ; P(e) & $e#[e# < e & P(e#)] undefined otherwise.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

immediate causation and causal explanation (where we say that / caused u if there is a series of causal connections leading from / to u). It is important to note that our data require an interpretation of CAUSE as immediate causation. For example, (7a), would not be true if Sally’s hammering so startled the elephant Benjamin that he sat down unexpectedly, thereby flattening the metal. Thus CAUSE cannot be interpreted as causal explanation. This holds quite generally for resultatives (as noted recently for example by Bittner (1999)). This aspect of the interpretation of resultatives is lost in our paraphrases (since the English verb cause does permit an interpretation as causal explanation), but must be taken to constrain this application of principle (R) as well as all subsequent cases interpreted with its help. The above analysis of resultatives allows us to capture both readings of (6) straightforwardly. The resultative LF in (7b) contains two propositional categories that could be modified by again: the entire VP or just the small clause ‘PRO flat’. The two LFs are given in (13). Using principle (R) these structures can straightforwardly be interpreted as (14) and (15) respectively (‘t_m’ stands for the referent of ‘the metal’).

10 There and Back Again: A Semantic Analysis 2.1.2 Lexical accomplishments. Interestingly, a repetitive/restitutive ambiguity is also possible with predicates that do not provide us with a result state in such an obvious fashion. An example is (17). (17) Sally opened the door again. (17#) a. Sally opened the door, and that had happened before. b. Sally opened the door, and the door had been open before.

(18)

a. b. c. d.

Sally opened the door. openTV ¼ openAdj + BECOME + CAUSE [VP Sally [ØV [SC openAdj [the door]]]] [VP Sally [[Ø + openAdj]V [SC t [the door]]]]

The verb, while phonologically empty, is semantically non-vacuous: it contributes the CAUSE BECOME component. Its semantics is given in (19). Thus the structure in (18c) is interpreted in (20a) and the sentence (18a) means (roughly paraphrased) (20b)—an intuitively appropriate description of its truth conditions. Note that the small clause expresses the result state of the opening event, the door being open. (19)

[ØV] / kpkxke.$P[Pe(x) & $e#[BECOMEe#(p) & CAUSE(e#)(e)]

(20)

a. ke.$P[Pe(Sally) & $e#[BECOMEe#(ke*.opene*(the_door)) & CAUSE(e#)(e)]] b. There was an action of Sally’s that caused the door to become open.

Back to restitutive again: now of course there obviously are two appropriate adjunction sites for again: the whole VP or the small clause. Two possible LFs for (17) are given in (21). Compositional interpretation of these structures yields (22) and (23). Thus we derive the repetitive/restitutive ambiguity of (17).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The lexical accomplishment predicate open allows a restitutive reading, even though there seems to be no phrase in the syntax that would independently express a result state and that could be modified by again. Stechow proposes a solution to this problem in terms of decomposition (resurrecting ideas from generative semantics). Thus the transitive verb open comes apart into the adjective open and a CAUSE BECOME component. Importantly, Stechow assumes that this decomposition is reflected in the syntax. (18a) underlyingly has the structure in (18c), where open is decomposed into a phonologically empty verbal head and the adjective open. In overt syntax the adjective incorporates into the verbal head and appears as the transitive verb open (cf. (18d)).

Sigrid Beck 11

(21) (22)

(23)

a. [VP [VP Sally [ØV [SC openAdj [the door]]]] again] b. [VP Sally [ØV [SC [SC openAdj [the door]] again]]] a. ke$.againe$(ke.$P[Pe(S) & $e#[BECOMEe#(ke*.opene*(t_d)) & CAUSE(e#)(e)]]) b. Once more, there was an action of Sally’s that caused the door to become open. a. ke.$P[Pe(S) & $e#[BECOMEe#(ke$.againe$(ke*.opene*(t_d))) & CAUSE(e#)(e)])] b. There was an action of Sally’s that caused the door to become once more open.

2.1.3 Finding result states: A parameter. It is worth pointing out that few adverbs have access to result states in the way that again does. Adverbs like never or repeatedly cannot have readings that parallel the restitutive reading of again: (24)

(25)

a. Gandalf never/repeatedly opened the door. b. Gandalf did something that caused the door to come to be never/repeatedly open a. Gandalf painted the door red repeatedly. b. Gandalf ’s painting the door caused it to come to be red repeatedly.

In (25), for example, imagine a paint that for a couple of weeks changes its colour according to the temperature. Even so, (25a) can only mean that Gandalf repeatedly applied paint. The adverb repeatedly cannot find the result state in a complex predicate or the result state in a decomposition structure. In fact, the only adverbs that I found in the literature reported as being able to do so are (translations of ) almost and again (hence the term decomposition adverb used for them, e.g. in

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The structural ambiguity theory assumes only the repetitive lexical entry for again in (16) above. This implies that in the case of restitutive readings, again modfies a result state predicate, and syntactically adjoins to a constituent that expresses that result state. The attractive feature of this theory is that there is just one again: again expresses repetition. The rest follows the pattern of regular syntactic ambiguity. On the negative side, the theory forces us to assume that in the case of lexical accomplishments, lexical decomposition in the syntax provides the necessary result state denoting constituent. We will see that the competitor of this theory, to be discussed in section 2.2, avoids this consequence. But first, we have one more thing to say about adverbs like again.

12 There and Back Again: A Semantic Analysis Stechow 1995, 1996). Example (26) describes the relevant (result state modifying) reading for an example with almost. Interestingly, it seems to be a lexical property of a given adverb whether or not it has access to the result state in a decomposition structure. German fast, the equivalent to almost, does not, as observed by Rapp & Stechow (1999). A relevant example is given in (27)—the reading in which fast would modify the result state ‘the door is closed’ is indeed impossible. (26)

There are also adverbs that are synonymous with again that still do not permit restitutive readings. An example noted by Rapp & Stechow (1999) is the German adverb erneut (‘anew’),6 which in contrast to the otherwise equivalent wieder does not permit a restitutive reading in (28). (28)

Maria hat die Tuer wieder geo¨ffnet. (repetitive/restitutive) Maria hat die Tuer erneut geo¨ffnet. (repetitive/ *restitutive) Maria has the door again opened ‘Maria opened the door again.’

Rapp & Stechow conclude that it is a lexical property of a given adverb whether or not it can look inside a decomposition structure. They capture this with their Visibility Parameter7 for adverbs (a D-adverb is a decomposition adverb; the formulation in (29) presupposes that the question of visibility arises only for a subset of adverbs in the first place). 6 A similar effect seems to exist in Hebrew, where informants get a restitutive reading more easily with me-xadaS (lit.: anew) than with od pa’am (lit.: one more time) or shuv. In Serbian/Croatian, ponovo (lit.: anew) seems to allow a restitutive reading more easily than opet, at least for some speakers. In Spanish, on the other hand, both otra vez (lit.: one more time) and de nuevo (lit.: anew) seem to allow restitutive readings, and similarly for Kannada thirga and punaha. It does not seem possible to infer from the morphological make-up of the adverb whether behaves like a decomposition adverb. 7 The term ‘parameter’ is their choice; it should be clear that this is not a Universal Grammar parameter in the usual sense that is set for the grammar of a language. This ‘lexical parameter’ is set in the lexicon for each adverb. Rapp & Stechow offer more empirical support for their proposal than I report here, specifically interaction of fast and tense in German.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. The dwarves almost closed the door. b. The dwarves did something that caused the door to become almost closed. (27) a. . . . weil Ottilie die Tuer fast schloss. since Ottilie the door almost closed ‘since Ottilie almost closed the door.’ b. # Ottilie did something that caused the door to become almost closed.

Sigrid Beck 13

(29)

The Visibility Parameter for decomposition adverbs (Rapp & Stechow (1999)): A D-adverb can/cannot attach to a phrase with a phonetically empty head.

Here is how visibility applies to (28): The AP that we get from decomposition is not accessible to erneut because it does not have a phonetically overt head. The boldface category in (30) is the only one visible to erneut. On the other hand, wieder can access both overt and phonetically empty result states, i.e. both the AP and the VP in (30). Compare Rapp & Stechow for more discussion. [ Maria [VP [Ø + openAdj]V [AP t [the door] ]]

There is a complication to this picture: Rapp & Stechow are not concerned with complex predicates like resultatives, and in particular do not investigate the question how visible the result state of a complex predicate is compared to a result state expressed by a syntactically independent phrase on the one hand, and the result state in a decomposition structure on the other. It turns out that the distinction matters. The adverb fast, for example, is able to access result states expressed by the result phrase in a complex predicate construction, like the example in (31) with a resultative. This is not true of erneut. (32) provides a parallel example in which the restitutive reading is impossible. Thus, there is evidence that visibility decreases from ordinary syntactic phrases to complex predicate result states, and further to decomposition result states.8 (31)

(32)

a. . . . weil Ottilie den Tisch fast sauber gewischt hat. since Ottilie the table almost clean wiped has ‘since Ottilie wiped the table almost clean.’ b. Ottilie’s wiping the table caused the table to become almost clean. a. . . . weil Ottilie den Tisch erneut sauber gewischt hat. since Ottilie the table again clean wiped has ‘since Ottilie wiped the table clean again.’

8 It is interesting that many of the most natural examples of restitutive again in English involve result states that are, to some extent, visible. Examples are given below (result states marked with up and back, respectively; both from: The Hobbit, or, There and Back Again, by J.R.R. Tolkien, Unwin Paperbacks, Fourth Edition, 1981, pages 97 and 233 respectively). Context makes it clear that the intended reading is restitutive. I have not investigated this further.

(i)

‘ ‘‘I must see if I can’t find a more or less decent giant to block it up again,’’ said Gandalf, ‘‘or soon there will be no getting over the mountains at all.’’ ’ (ii) ‘Now if you wish, like the dwarves, to hear news of Smaug, you must go back again to the evening when he smashed the door and flew off in rage, two days before.’

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(30)

14 There and Back Again: A Semantic Analysis b. # Ottilie’s wiping the table caused the table to become once more clean. I conclude that visibility comes in three stages: all adverbs can access fullfledged syntactic phrases; some adverbs can in addition find the result state in a complex predicate construction (again, almost, fast, wieder, but not erneut); a subset of those can even look inside a decomposition structure (again, almost, wieder, but neither erneut nor fast). I suggest to revise the visibility parameter as in (33). The setting of the parameter is marked in the lexical entry of an adverb. We will see below that visibility phenomena like the ones discussed here exist cross-linguistically. The visibility parameter for adverbs: An adverb can modify (i) only independent syntactic phrases (ii) any phrase with a phonetically overt head (iii) any phrase The default setting is (i).

Let us summarize the predictions made by the structural analysis: A restitutive reading with again is expected to be possible for a predicate only if that predicate is internally complex, i.e., involves more than one property of events. On the restitutive reading, again semantically modifies a result state, and syntactically attaches to the constituent that denotes that result state. Typical examples for such predicates are accomplishment predicates (also achievements, but we will not be concerned with achievement predicates in this paper). Additional restrictions on the availability of a restitutive reading may arise from the visibility parameter. Pace visibility, the theory makes restitutive again a detector of the internal syntactic and semantic complexity of a predicate.

2.2 A lexical ambiguity analysis (Fabricius-Hansen 1983, 1995, 2001) It is now time to introduce the alternative theory of the repetitive/ restitutive ambiguity, which locates the source of the ambiguity not in syntax, but in a lexical ambiguity of again. I will take Fabricius-Hansen (1995, 2001) as the representative of this type of analysis.9 The basic 9 I should stress that my presentation of this work is much simplified in respects that are not crucial to the point I want to make, and therefore in many ways does not do it justice. This concerns, for example, the description in terms of lexcial ambiguity: Fabricius-Hansen’s actual analysis is in terms of polysemy. Other simplifications concern the role of the presupposition associated with again. See, in particular, Fabricius-Hansen (2001) for interesting discussion.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(33)

Sigrid Beck 15

idea here is that besides an again that expresses repetition, there is a second meaning of again which expresses reversal of direction and will lead to what we call the restitutive reading. This is illustrated with (34): (34)

(The temperature was falling all morning.) Now it is rising again.

(35)

a. [[again1]](PÆi, tæ)(e) ¼ 1 iff P(e) & $e#[e#< e & P(e#)] ¼ 0 iff ; P(e) & $e#[e#< e & P(e#)] undefined otherwise. b. [[again2]](PÆi, tæ)(e) ¼ 1 iff P(e) & $e#[e# < e & Pc(e#) & resPc(e#) ¼ preP(e)] ¼ 0 iff ; P(e) & $e#[e# < e & Pc(e#) & resPc(e#) ¼ preP(e)] undefined otherwise.

Counterdirectional again applies to a predicate of events P and an event; it presupposes that there is a preceding event of which the counterdirectional predicate Pc of P is true, and the result state resPc of which is the starting point, or prestate preP, for the new event. Applied to our temperature example, the relevant concepts are given in (36). The counterdirectional predicate to a rise is a fall, and the result state of a falling event is an event of the temperature being low (perhaps a local minimum on the temperature curve). Given that these are the concepts involved, the restitutive or counterdirectional reading of our example amounts to (37). (36)

(37)

predicate P: ke.risee(the_temp) counterdirectional predicate Pc: ke.falle(the_temp) ks.lows(the_temp) possible prestates of P, preP: possible result states of P, resP: ks.highs(the_temp) a. The temperature is rising again. b. ke.risee(the_temp) & $e#[e# < e & falle#(the_temp) & resPc(e#) ¼ preP(e)]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The adverb again can be used in this example without there having to be a preceding rising of the temperature. Instead, it seems to be licensed by virtue of the fact that there is a preceding falling of the temperature. Fabricius-Hansen calls this the counterdirectional again. Thus besides the repetitive again that we are already familiar with, repeated in (35a), there is also counterdirectional again given in (35b) (both are adapted to our framework and simplified in ways that do not matter for the points made here; see Fabricius-Hansen for discussion).

16 There and Back Again: A Semantic Analysis Thus (34) is true of an event if that event is a rising of the temperature, and there is a preceding falling of the temperature the result of which is the starting point for the rise. This seems a good description of the relevant reading of (34). Compared to the structural analysis of restitutive again, this analysis relies on conceptual (as opposed to syntactic) prerequisites: the availability of the concepts of a counterdirectional predicate, a result state of an event and a prestate of an event. (38b) and (39) apply this analysis to our original example (38a); the analysis is straightforward.10 (38)

Let’s compare this analysis to the structural ambiguity theory. An attractive feature of the lexical theory is that we do not need to structurally decompose lexical accomplishment predicates. The syntactic structure from which we derive an event description is irrelevant on this analysis. All that matters is that the event described is reversible, and thus there is a counterdirectional predicate conceptually available. The price we pay for this simplicity is that we have to assume a second lexical entry for again. It looks as if the lexical analysis were somewhat more tolerant regarding which predicates give rise to a restitutive/counterdirectional reading. Fabricius-Hansen notes that atelic change of state predicates like the predicates fall and rise that motivated our discussion (called degree achievements by Dowty 1979) permit a counterdirectional reading. It seems that it suffices that a predicate express a change of state (telic or atelic) for restitutive/counterdirectional again to be possible. The structural analysis made it appear that only accomplishments and achievements (telic transition events in Fabricius-Hansen’s terms), which have a clear result state, permit restitutive readings. The question arises what the structural theory could say about the temperature example (and other predicates of this kind). Stechow (1996) provides an analysis of the predicate rise along the lines illustrated below (see Stechow for more details). The sentence (40a) is receives the analysis in 10 Actually it doesn’t have to be Sally who closed the door—somebody else could have done it, or the wind could have blown it shut. This will be irrelvant to anything we say below. I will not pursue the question of which participants of an event have to be carried over to the counterdirectional predicate under this analysis.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. Sally opened the door again. b. ke.opene(the_door)(S) & $e#[e# < e & closee#(the_door)(S) & resPc(e#) ¼ preP(e)] (39) predicate P: ke.opene(the_door)(S) counterdirectional predicate Pc: ke.closee(the_door)(S) possible prestates of P, preP: ks.closeds(the_door) possible result states of P, resP: ks.opens(the_door)

Sigrid Beck 17

(40c) and is true of an event e if at the end of e, the temperature is higher than it was at the beginning of e. Again in (41a) can then be interpreted in the scope of BECOME, leading to (41b). (40)

Thus the sentence has a reading on which it is true of an event e if at the end of e, the temperature is higher than it was at the beginning of e; it presupposes that once before, the temperature was higher than at the beginning of e. Hence the example will be true in a situation in which the temperature rises after a preceding fall, as desired. For the structural analysis, a restitutive reading invariably requires that a result state denoting constitutent be available as the adjunction site of again. This makes decomposition in the syntax necessary—a step that may seem fairly natural for the case of accomplishment predicates like open, but perhaps less natural for rise. On the other hand, Stechow builds his structural theory on the basis of a set of data that indicate that the prerequisites for restitutive again are indeed syntactic, and not purely conceptual. These facts are illustrated by the word order effect in German (42a) vs. (42b): (42)

a. because because b. weil because

Ottilie

die Tu¨r wieder o¨ffnete. (restitutive, repetitive) Ottilie the door again opened Ottilie wieder die Tu¨r o¨ffnete. (repetitive only) Ottilie again the door opened

‘because Ottilie opened the door again.’ A restitutive reading is possible if wieder follows the direct object, but not if it precedes it. Stechow’s analysis is, in a nutshell, that the direct object in German obligatorily moves to a fairly high position in the syntax (SpecAgrO, for Stechow). If wieder precedes the object, then

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(41)

a. The temperature rose. b. The temperature became higher. c. ke.BECOMEe(ke#.MORE(kd.d-highpre(e)(the_temp)) (kd.d-highe#(the_temp)))) d. MORE (D)(D#) ¼ 1 iff max(D#) > max(D) e. Let S be a set ordered by £. Then max(S) ¼ is[s 2 S & "s#2 S[s £ s#]] a. The temperature rose again. b. ke.BECOMEe (ke#.againe#(ke$.MORE(kd.d-highpre(e)(the_temp))(kd.dhighe$(the_temp)))))

18 There and Back Again: A Semantic Analysis wieder is higher in the structure than the CAUSE BECOME component and must take scope over it (resulting in the repetitive reading, the only reading of (42b)). If wieder follows the direct object, it could be either above or below the CAUSE BECOME component, and both readings are possible (resulting in an ambiguous example like (42a)). The structures for (42a,b) are sketched in (43). The word order effect shows that conceptual availability of a counterdirectional predicate is not sufficient for the restitutive reading to be available. Availability of a restituive reading depends on syntactic factors. This lends strong support to a structural theory of restitutive again. a. [ Ottilie [AgrOP [die Tu¨r] [VP (wieder) [VP ØV [SC (wieder) [SC offen]]]]] b. [ Ottilie [AgrOP wieder [AgrOP [die Tu¨r] [VP ØV [SC offen]]]]]

It seems fair to say, then, that given the current state of the debate around restitutive again, both analyses have their attractions, and for both this comes at a cost. The issue of restitutive again is a much discussed problem, and a lot more arguments in favour of one or the other type of analysis could be reported. This is not what I want to do here (compare instead Stechow 1995, 1996; Fabricius-Hansen 2001; Ja¨ger & Blutner 2000; Klein 2001 and references therein). The goal of this paper is to bring forth a new type of evidence that this debate has to take into consideration. For this purpose, what is crucial are the prerequisites that each theory has to assume for the availability of the restitutive reading. Under the structural theory, those prerequisistes are lexical (setting of the visibility parameter for the adverb) and structural (a result state denoting constituent for again to adjoin to). For the lexical theory, the prerequisites are lexical (availability of counterdirectional again) and conceptual (a change of state predicate that counterdirectional again can modify). Our cross-linguistic data differentiate between the two theories on the basis of those prerequisites. The next section introduces the cross-linguistic test case.

3 GOAL PP CONSTRUCTIONS Remember that our crucial empirical finding concerns combinations of an activity predicate and a PP expressing a goal, like (4), which we call goal PP constructions (henceforth: goal-PPCs). Our generalization can be described as in (G#). In this section, we will gather evidence that the semantics of goal-PPCs is subject to cross-linguistic variation, and provides a promising test case for the two theories of restitutive again

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(43)

Sigrid Beck 19

3.1 The complex predicate parameter It is well-known that the availability of resultative constructions like (44a) varies across languages (see Snyder 2001 and references therein, e.g. Green 1973; Levin & Rapoport 1988; Aske 1989; Talmy 1991). For example, the Spanish translation of (44a) in (44b) is ungrammatical. (44) a. Mary beat the metal flat. b. Mary golpeo´ el metal (*plano). Mary beat the metal (*flat). In Synder (2001; see also Snyder 1995) it is argued that this contrast is the result of the setting of a genuine parameter of grammar. Besides governing availability of resultatives, that parameter governs the availability of a whole set of other constructions which he calls complex predicate constructions. Complex predicate constructions in English include putlocatives, double-object constructions, and verb-particle constructions like (45). Crucial evidence for Snyder’s position comes from child language acquisition. The constructions mentioned are all acquired by English learning children at approximately the same time. His hypothesis (simplified for our purposes) is summarized in (46). (45) Bilbo picked up the Arkenstone. (46) Complex Predicate Parameter (Snyder 2001): One grammatical parameter is responsible for the availability of complex predicate constructions (resultatives, verb-particle constructions and others). Snyder further observes that the availability of complex predicate constructions is dependent on compounding: Complex predicate constructions are possible in a language only if root compounding is productive. Once more, acquisitional data support this conclusion: children

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

that I introduced. More precisely, we will see that the semantics of goal-PPCs varies systematically with the availability of resultatives. (4) Bilbo walked to the hall again. (G#) A goal PP construction in combination with again has a restitutive reading only in languages that permit resultatives. We will first look at cross-linguistic variation in the availability of resultatives and introduce Snyder’s (2001) position that this is the result of the setting of a genuine grammatical parameter. Then, resultatives will be tied to goal-PPCs, and we will see that variation in the interpretation of goal-PPCs is correlated with the proposed parameter (as argued by Beck & Snyder 2001a).

20 There and Back Again: A Semantic Analysis

3.2 Goal PP constructions across languages Beck & Snyder (2001a) propose that the Complex Predicate Parameter is involved in yet another point of cross-linguistic variation: goalPPCs. In English, the verb walk is an activity predicate, as shown for instance by the ungrammaticality of modifying walk with a temporal in-phrase. It is well-known that when we add a PP expressing the goal of the walk, modification with a temporal in-phrase becomes grammatical—indicating that the predicate ‘walk to the summit’ is an accomplishment predicate rather than an activity. This has already been observed by Dowty (1979), for example. Importantly, this is a systematic phenomenon; some further illustration is provided in (48). Higginbotham (2000) notes that examples like (48d) are ambiguous. (47)

a. * b. (48) a. b. c. d.

Paul walked in an hour. Paul walked to the summit in an hour. Paul swam to the island (in 10 minutes). The baby crawled under the bed/behind the sofa (in 10 minutes). Sally hopped up the stairs/across the meadow (in 10 minutes). I walked around the fountain in an hour/for an hour.

It is this type of predicate that we call a goal-PPC: an activity verb combined with a goal phrase. Aske (1989) (drawing on work by Talmy 1985; compare also Levin and Rappaport Hovav 1995 and references therein) explores cross-linguistic variation in such constructions. He notes that Spanish permits the analogue of English ‘walk to the summit’ as a grammatical string; however, modification with a temporal in-PP is

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

tend to learn complex predicate constructions shortly after they have acquired compounding, and crucially, no child in Snyder’s study acquired a complex predicate construction before they had acquired root compounding. Snyder proposes that complex predicate constructions involve a morphological compound at some level of representation. See Snyder (2001) for discussion and details. I suggest, following Beck & Snyder (2001b), that availability of principle (R) from section 2 is part of the Complex Predicate Parameter. If a language lacks (R), then resultative structures like (44) will be uninterpretable. This can account for their unacceptability in that language, as hypothesized in section 2. For the purposes of this paper, I will in fact regard principle (R) as a central part of the Complex Predicate Parameter.

Sigrid Beck 21

(49) (50)

Pablo camino hasta la cima (* en una hora). Pablo walked up-to the summit (* in one hour) a. John swam under the bridge. b. Juan nado debajo del puente (*en una hora). Juan swam under the bridge (in an hour)

Beck & Snyder (2001a) investigate how general Aske‘s phenomenon is cross-linguistically. In the table in (51) below I report their results. The grammaticality of modifying a goal-PPC with a temporal in-PP was tested in languages that allow the plain goal-PPC as a grammatical string (last column in (51)).12 All relevant data are reported in Appendix A. Before we proceed, a comment is in order on the translations we chose for the English goal-PPC. Our criteria for regarding a particular translation of ‘Suresh walked to the village’ as the analogue of the English sentence in that language are (i)–(iii). (i) the truth conditions of the translation imply (a)–(c): (a)

Suresh arrived at the village (i.e. he didn’t walk towards the village).

11 Other researchers have, like Aske, made a connection between result phrases in resultative constructions and goal PPs combined with verbs of motion: Goldberg (1995), Jackendoff (1990) and Levin & Rappaport Hovav (1995). 12 Interestingly, this is not always the case. In Basque, for example, the direct translations of goal PP constructions are ungrammatical. A paraphrase with ‘go by foot’ has to be used for walk plus goal PP, for instance. So Basque does not allow goal PP constructions at all.

(i)* Suresh etxe-ra ibili zen. Suresh home-to walked Aux Suresh walked home. (ii) Suresh oin-ez herrira joan Suresh foot-by village-to go Suresh went to the village by foot.

zen. Aux

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

ungrammatical. This indicates that there is a subtle semantic difference between English ‘walk to the summit’ and Spanish ‘caminar hasta la cima’. The English predicate behaves like an accomplishment predicate while the Spanish predicate does not. Aske concludes that Spanish does not permit what he calls telic path phrases, in which the PP is a result state predicate. He relates this to the absence of resultatives in Spanish. Note that this is very much in line with, and in fact anticipates, our idea of relating goal-PPCs to the Complex Predicate Parameter and resultatives.11 Aske further observes that a verb of motion combined with a purely locative PP gives rise to a directional/locative ambiguity in English, but can only be locative in Spanish—clearly a phenomenon related to the effect with directional ‘to’.

22 There and Back Again: A Semantic Analysis (b)

Suresh was away from the village at some point (i.e. he didn’t walk within the village). (c) The sentence describes an event of walking by Suresh.

(ii) the surface form of the goal-PPC is as close to English ‘V PP’ as possible in that language (e.g. no adverbial paraphrases). (iii) The verb by itself is an activity predicate by the criteria of Vendler/ Dowty (e.g. it does not mean ‘arrive’).

(51) language English German Japanese Korean Mandarin French Hebrew Hindi/Urdu Spanish

(R)-parameter + + (+) + + – – – –

goal PP + temporal in PP ok ok ok ok ok * * * *

The results obtained matched perfectly with the setting of the Complex Predicate Parameter (abbreviated as (R)-parameter, in the second column in (51)).13 The setting reported for the parameter is based on the criteria in Snyder (2001), specifically availability of resultatives. We differ from Snyder’s assessment in one case: Japanese is reported to have the positive setting of the (R) parameter in Snyder (2001). However, there is considerable variation across speakers concerning in how far 13 The informed reader will notice that the distinction made by the (R)-parameter cuts across another parameter suggested in connection with verbs of motion: Talmy’s (1985, 1991) typological distinction between satellite-framed v. verb-framed languages. In contrast to Talmy, I am not concerned with lexicalization patterns for motion verbs, so the empirical domain is quite different, and the (R)-parameter can indeed be expected to cut across Talmy’s distinction.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

As the reader will see, a certain amount of variation in the translations is unavoidable. This concerns for example the choice of the preposition. Not all languages offer the same possibilities for path-related prepositions. Our choice was governed by the truth conditional requirements under (i). Some further comments on other minor variations are made in Appendix A for specific languages. Where we could not in good conscience regard any translation as the analogue of an English goal-PPC, we excluded the language from consideration (e.g. Basque, Russian). Finally, it should be stressed that the predicate ‘walk to the village’ is to be regarded as a representative of goal-PPCs in general.

Sigrid Beck 23

resultatives are grammatical (cf. Washio 1997). Levin and Rappaport Hovav (1995) mention similar uncertainty with respect to goal-PPCs in Japanese. The bracketed + indicates this reservation. The languages in (51) give us a clear correspondence between the setting of the (R)parameter and modifyability of a goal-PPC with a temporal delimiter. This shows that (49) is not an isolated fact about Spanish, but rather evidence of a much more general phenomenon. There must be a subtle semantic difference in the interpretation of goal-PPCs in (+R) languages versus (
R) languages. Beck & Snyder (2001a) drew the following conclusion: (52)

We should note here that Beck & Snyder’s (2001a) hypothesis is not based exclusively on cross-linguistic evidence. In addition to the data that we have presented, there is evidence from child language acquisition pointing in the same direction. Beck & Snyder observe that the claim that goal-PPCs are a complex predicate construction in English implies that, like the other complex predicate constructions, goal-PPCs should be acquired shortly after compounding is, and, importantly, children should not master goal-PPCs before they have mastered compounding. Beck & Snyder compare the ages of first clear use of compounding and goal-PPCs in ten children from Childes (MacWhinney & Snow 1985, 1990; MacWhinney 2000). Their predictions are born out. The ages of acquisition for the two constructions were significantly correlated (p ¼ 0.0115), thus further supporting (52). I have found two more languages that confirm this hypothesis, but also one language which clearly has the negative setting of the (R)parameter and nonetheless allows modification with a temporal delimiter—Kannada, a Dravidian language spoken in Southern India: (53) language

(R)-parameter

goal PP + temporal in PP

Hungarian Khmer

+ +

ok ok

Kannada

–

ok

I cannot offer an explanation for this fact. In view of the overall good correlation between the setting of the complex predicate parameter and

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Beck & Snyder (2001a): Goal PP constructions are complex predicates in (+R) languages. They denote accomplishments in (+R) languages, but not in (
R) languages.

24 There and Back Again: A Semantic Analysis

(54)

a. The dwarves opened the door for 10 minutes. b. The dwarves opened the door, and the door stayed open for 10 minutes.

English and German goal-PPCs permit this reading (data in (55), reading paraphrased in (55#)), but Spanish and Hebrew goal-PPCs do not (cf. (56)); my informants actually judged the examples deviant. In all languages, I use as a translation of ‘for ten minutes’ a phrase that gives rise to the reading in (54b) when combined with a lexical accomplishment predicate. The data support Beck & Snyder’s (2001a) 14

An obvious solution would be that the phrase used as a temporal delimiter in Kannada, eradu ganteyalli, is not semantically identical to English ‘in two hours’. It is notoriously difficult to determine such semantic equivalence across languages; I do not at present have sufficient evidence about Kannada to draw a conclusion. Grolla (2003) analyses the (
R) language Brazilian Portuguese, another apparent problem case. She observes that data using the preposition ate´ look like a counter-example to (52), cf. (ib). However, ate´ is the only preposition that shows this behaviour. Other prepositions lead to the expected pattern, cf. (ii). She suggests that the lexical semantics of ate´ is to be blamed for its exceptional behaviour. Brazilian Portuguese shows us the importance of investigating the systematicity of the phenomenon rather than isolated examples. Grolla comes to the conclusion that, once ate´ is excluded from goal-PPCs, Brazilian Portuguese fits the pattern of a (
R) language described here (in particular impossibility of a temporal delimiter and unavailability of a restitutive reading with goal-PPCs), and thus confirms my claims. (i)

a.

A Creuza martelou o metal (*plano). the Creuza hammered the metal (flat) ‘Creuza hammered the metal flat.’ b. A Virgulina andou ate´ a vila em uma hora. the Virgulina walked to the village in an hour ‘Virgulina walked to the village in an hour.’ (ii) a. O Joao nadou debaixo da ponte (*em meia hora). the Joao swam under the bridge (in half hour) ‘John swam under the bridge in half an hour.’ b. O bebe engatinhou atra´s da cerca (*em meia hora). the baby crawled behind the fence (in half hour) ‘The baby crawled behind the fence in half an hour.’

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

the interpretation of goal-PPCs I will maintain the hypothesis in (52) and leave Kannada as a puzzle.14 Dowty (1979: 56–58) proposes several other tests besides modification with a temporal delimiter like ‘in an hour’ that distinguish accomplishments from activities. It would be interesting to see how the goal PP predicates in (
R) languages fare with regard to these other criteria. Unfortunately, applicability of Dowty’s tests varies considerably from language to language, making this a task beyond the scope of this paper. As a first step, I have started to investigate the availability of a result state modifying reading with a phrase indicating temporal duration (Dowty’s test VIII). Only accomplishments permit such a reading. An English example and the relevant reading are in (54).

Sigrid Beck 25

conclusion that goal-PPCs are not accomplishments in (
R) languages. Further investigation must be left for future research. (55)

The questions that the data in this section raise are: in what way exactly does the semantics of goal-PPCs differ in (+R) versus (
R) languages, how does this difference arise, and how is it related to the availability of resultatives. These questions are taken up in section 5. At this point, we conclude that goal-PPCs are a type of predicate that can plausibly distinguish between the predictions of the lexical analysis of again and the structural analysis. It is clear from our truth conditional requirements that in all languages the predicate implies a change of state and should be reversible. However, the connection with principle (R) suggests that their internal composition as well as the resulting semantics are different in (+R) languages versus (
R) languages. This makes them a promising candidate for research on the cross-linguistic behaviour of again. 4 RESTITUTIVE AGAIN CROSSLINGUISTICALLY I will now, in subsection 4.1, present the cross-linguistic data on restitutive again, which lead to the generalization in (G#). We can rephrase (G#) as (G$), given the discussion in the last section. In section 4.2. we interpret our results and detail their implications for linguistic theory.15 A goal PP construction in combination with again has a restitutive reading only in languages that permit resultatives. (G$) A goal PP construction in combination with again has a restitutive reading only in languages that have the positive setting of the Complex Predicate Parameter.

(G#)

15 I should note that some of the material presented here is a further development of ideas that were first introduced in Beck & Snyder (2001b). A proper subset of the data in section 4.1. has been reported in Beck & Snyder (2001b). Also, the analysis of goal PP constructions in (+R) languages and the restitutive reading, in sections 5.1. and 5.2. respectively, is in the spirit of that paper, although the details of the analysis differ.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a. Suresh walked to the village for 10 minutes. b. Suresh ist fu¨r 10 Minuten zum Dorf gelaufen. Suresh is for 10 minutes to-the village walked (55#) Suresh walked to the village, and he stayed there for 10 minutes. (56) a. ?? Suresh anduvo hasta la aldea durante diez minutos. Suresh walked to the village for 10 minutes b. ?? Dan halax el ha-kfar lemeSex eser dakot. Dan walked to the-village for 10 minutes

26 There and Back Again: A Semantic Analysis

4.1 Data and method The availability of a restitutive reading was tested for (minor variations of) the two sentences in (57) in 18 languages: American Sign Language (ASL), Bahasa Indonesia, English, French, German, Hebrew, Hindi/ Urdu, Hungarian, Inuttut (an Inuktitut language spoken in Labrador), Japanese, Kannada, Khmer, Korean, Lingala (a Bantu language spoken in the Democratic Republic of Congo), Mandarin Chinese, Serbian/ Croatian, Spanish and Tagalog.16 (57)

a. Sally opened the door again. b. Suresh walked to the village again.

(58)

a. open the door again Sally built a wardrobe. The last thing she made was the door. She set it on its hinges and it looked fine. But when she closed the door, it didn’t quite fit. So she opened it again and took it off to sand the edges.

16 In a previous report of this research (Beck & Snyder (2001b)), Russian was included in the survey. The Russian goal PP data tested rendered unclear results—an issue we had been unable to resolve. A more careful look at the language has since revealed that the crucial data are systematically not testable (similar to the case of Basque). Previously, we had regarded (ia) as a goal PP construction. However, as (ib) reveals, the verb we used is not an activity predicate. The verb in (iia) is an activity predicate, but combination with a goal PP is unacceptable. Other verbs behave in a parallel way (e.g. ‘swim’ plavat’, ‘fly’ letat’). Russian thus does not seem to permit us to construct what we call a goal PP construction, and I have taken it out of the survey.

(i) a.

Suresh prishel v derevnju. Suresh Perf-walked to village b. * Suresh prishel 10 minutes Suresh Perf-walked for 10 minutes (ii) a. Suresh shel 10 minutes Suresh walked for 10 minutes b. * Suresh shel v derevnju. Suresh walked to village

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Judgments were elicited by presenting the crucial sentence in the context of (minor variations of) the two stories in (58). Note that the story makes it clear that no event of the kind described in the sentence happened before. Thus the sentence would be a presupposition violation on the repetitive reading and should be judged inappropriate. Only the restitutive reading fits the story. If an informant accepted the crucial sentence in the story, it was concluded that a restitutive reading was available. Was the story rejected on the basis of the sentence, it was concluded that a restitutive reading was unavailable.

Sigrid Beck 27

b. walk to the village again Suresh was born in a tiny village on a mountain top in Nepal. It is accessible only by a footpath through the mountains. He left the village for the first time when he was ten, and went to a school in the city for twelve years without going home. He walked to the village again when he was 22.

17

It is fairly easy to find free range examples of the relevant type for both languages. (ia–d) provide some illustration of restitutive readings with English goal PP constructions (the construction type in (ie) is perhaps less clear). (ii) are examples with lexical accomplishments. In all cases, the context which the sentence occurs in makes it clear that the intended reading is the restitutive one (i.e. Dumbledore had never climbed out of the trunk before, Julia had never run to the bedroom before, and so on). Examples (ic,d,e) and (iia–e) are all from the e-texts on the Project Gutenberg official homesite http://promo.net/pg/ (i)

a.

b. c. d. e.

Dumbledore covered Moody in the cloak, tucked it around him, and clambered out of the trunk again. (Harry Potter and the Goblet of Fire, by J.K. Rowling, 2000, Scholastic Press, 1st American Edition, hardcover, p. 681) Julia ran to the bedroom again. (Time and Again, by Jack Finney (1970), Recorded Books Unabridged, 1995, tape 10) [. . .], with which I plunge downward to the surgery again, [. . .] (Hospital Sketches, by Louisa May Alcott) They all looked exceedingly foolish; and Edward seemed to have as great an inclination to walk out of the room again, as to advance further into it. (Sense and Sensibility, by Jane Austen) The forms filed into the cabinet again, [. . .]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The actual data used for the survey are provided in Appendix B. The same comments on the choice of translation for the predicate apply as before (and in those languages which occur in both surveys, the predicates are identical to those in Appendix A). Moreover, the strategy was, uniformly, to try to create the best possible circumstances for acceptability of a restitutive reading. This implied in particular choosing that translation of the word again that was most likely acceptable with a restitutive reading (remember from section 2 that not all adverbs with the basic meaning of again can be restitutive), and choosing a sentential context that supports the restitutive reading (remember, also from section 2, that word order can have an effect on availability of restitutive readings). Focus has also been observed to play a role (see, for example, Klein 2001 for discussion). All of these factors are interfering factors that, for our purposes, have to be excluded. We are interested in the question whether our predicates support restitutive readings at all. The results of the cross-linguistic study are summarized in the table below. The first column specifies the language, the second the number of consultants for that language (the data for English and German come from the literature and agree with our intuitions; I did not formally elicit judgments for either).17 Next, I give the setting of the (R)

28 There and Back Again: A Semantic Analysis parameter for the language. The last two columns report the judgements obtained for our ‘open the door again’ example and our ‘walk to the village again’ example, respectively. The entry ‘ok’ means that the restitutive reading was accepted, ‘*’ means it was rejected; one solitary dissenting opinion is recorded by bracketing the judgment and noting that. If genuinely mixed judgments were obtained, this is reported as ‘%’, with numbers provided in parentheses. (59) Language

Number of consultants

Restitutive reading w/lexical accomplishment

Restitutive reading w/goal PP construction

ok ok ok (*) (1 acc.) % (4 acc., 3 rej.) ok (ok) (1 ?) % (1 acc., 2 ??)

ok ok ok % (3acc,1?, 1rej) % (4 acc., 3 rej.) ok ok ok

ok (ok) (1 rej.) ok % (3 acc., 2 rej.) * ok * % (2 acc., 3 rej.) ok ok

* * * * * * * * * *

ASL English German Hungarian Japanese Khmer Korean Mandarin

(3)

(5) (7) (1) (5) (3)

+ + + + (+) + + +

Bahasa Indonesia French Hebrew Hindi/Urdu Inuttut Kannada Lingala Serbian/Croatian Spanish Tagalog

(1) (3) (5) (5) (3) (2) (1) (5) (7) (3)













(ii) a. b. c. d. e.

(How Spirits Materialize, author anonymous, in: David Phelps Abbot, Fraudulent Spiritualism Unveiled) The splendour of the argument took Jill’s breath away, and before she got it again, in came Frank and Ralph with two clothes-baskets of treasures to be hung upon the tree. (Jack and Jill, by Louisa May Alcott) Jack opened his lips to speak, but shut them again, [. . .] (Jack and Jill, by Louisa May Alcott) As she said this, she sunk into a reverie for a few moments;–but rousing herself again, [. . .] (Sense and Sensibility, by Jane Austen) I might have sold it again, the next day, for more than I gave: [. . .] (Sense and Sensibility, by Jane Austen) And away she went; but returning again in a moment, [. . .] (Sense and Sensibility, by Jane Austen)

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(R)

Sigrid Beck 29

4.2 Discussion

(60)

A kira‘lyne’ u$jra fel-ment a torony-ba. the queen again up-went the tower-(in)to The queen climbed the tower again.

In short, we see that restitutive readings with goal-PPCs are only acceptable in a language if that language has the positive setting of the (R) Parameter.18 The first conclusion that we draw from these data is that we have found further confirmation for the (R) Parameter. We have found another test that distinguishes the same two groups of languages as availability of resultatives and Snyder’s other criteria. It follows that linguistic theory has to make a connection between availability of resultatives and availability of restitutive again with goal-PPCs. 18 The probability of obtaining this result by accident was caluclated by Fisher Exact Test (leaving out Japanese, and using Hungarian2). The outcome is p < 0.001. This result is significant, indicating that the correlation between the setting of the (R) parameter and the availability of a restitutive reading with goal PP constructions is not accidental.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Let’s first consider the judgments we obtained for the goal-PPCs. There is a very good correlation between availability of a restitutive reading for goal-PPCs and the setting of the (R) Parameter. Exceptions are Japanese and Hungarian, where the judgements in the last column are mixed. For Japanese, we already noted that there is genuine variation between speakers as to the acceptability of resultatives. It is not surprising that this resurfaces with the restitutive reading and gives us mixed judgments. Hence Japanese actually fits the correlation quite well. Hungarian, on the other hand, clearly has the positive setting of the (R) Parameter. Regarding the goal-PPCs, it seems that there is a dialect (let’s call it Hungarian1) that accepts restitutive readings, but there is also a dialect, Hungarian2, in which restitutive readings with goal-PPCs are unacceptable. Judgments with a verb-particle construction (example given in (60)) were similarly mixed (3acc., 1?, 1*) ((60) contains a PP ‘to the tower’, similar to our goal PP example; note, however, that the verb has already combined with a particle—thus (60) is not a goal-PPC but a different kind of complex predicate construction). The fact that judgements for (60) are quite parallel to the judgements for Hungarian ‘walk to the village’ supports our claim that goal-PPCs are a complex predicate construction in Hungarian. Our view of Hungarian is, then, that speakers do not uniformly allow restitutive readings with complex predicate constructions, including goal-PPCs—speakers of Hungarian1 do, but speakers of Hungarian2 do not.

30 There and Back Again: A Semantic Analysis

(A)

a (
R) language never permits restitutive readings with goal PP constructions. (B) a (
R) language may or may not permit a restitutive reading with lexical accomplisments. (C) a (+R) language very often (but not necessarily) does permit restitutive readings with goal PP constructions. (D) In a (+R) language, a restitutive reading with a lexical accomplishment is as acceptable or less acceptable as with a goal PP construction (but not more acceptable). (E) there is room for variation in the judgments for lexical accomplishments. The behaviour of restitutive again confirms the claim from section 3 that the interpretation of goal-PPCs in (+R) languages differs from their interpretation in (
R) languages. That semantic difference has to be the key to the pattern of restitutive again. The next section develops an analysis. 5 ANALYSIS The task that our cross-linguistic data present to the semanticist is to assign different interpretations to goal-PPCs in (+R) v. (
R) languages,

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Next, let’s consider our findings for lexical accomplishment predicates. We observe first that there is no correlation of availability of a restitutive reading with lexical accomplishments, and the setting of the (R) parameter. In (+R) languages, note that when there is a difference in acceptability between lexical accomplishments and goal PPs, restitutive again is always less acceptable with lexical accomplishments (Mandarin, Hungarian). In (
R) languages on the other hand, restitutive readings with lexical accomplishment predicates tend to be better than with goal-PPCs, in that they are frequently acceptable. They seem completely acceptable in Bahasa Indonesia, Hebrew, Kannada, Spanish and Tagalog. French, Hindi/Urdu and Serbian/Croatian show mixed results, and they were rejected only in Lingala and in Inuttut. It is striking that there can be considerable variation between speakers within a given language (e.g. Japanese, Hindi/Urdu, Serbian/ Croatian). What’s more, a particular speaker may accept a restitutive reading with one lexical accomplishment predicate, but reject it with another (several speakers accepted a restitutive reading with the lexical accomplishment return, while rejecting the ‘open the door again’ example—in Japanese (where the verb return is kaetta), Mandarin (xui), Lingala (jonga) and Serbian/Croatian (vrati)). To summarize, linguistic theory has to predict that:

Sigrid Beck 31

5.1 Semantics of goal PP constructions 5.1.1 The Semantics of Goal PP constructions in (+R) Languages. I suggest, with Beck & Snyder (2001b), that in (+R) languages, the compositional analysis of resultatives by Stechow (1995) can apply to goal-PPCs. Recall that we introduced Principle (R), repeated below, to combine a verb with a result state. (8)

Principle (R) (Stechow (1995)): If a ¼ [Vc SCb] and b# is of type Æi, tæ and c# is of type Æe, . . . Æe,Æi, tæææ (an n-place predicate), then a# ¼ kx1 . . . kxnke.c#e(x1) . . . (xn) & $e#[BECOMEe#(b#) & CAUSE(e#)(e)]

I propose that goal PPs in English and other (+R) languages can be interpreted as result phrases, using principle (R). This suggestion per se is not new: the connection between resultatives and what we call goal PPs has been made before and has been much debated in the literature; compare for example Goldberg (1995), Jackendoff (1990), Levin & Rappaport Hovav (1995) and references therein, also once more Aske (1989). We will here work out a proposal on the basis of (8). Thus we assign to (61a) the structure in (61b), where the PP is a small clause with a PRO subject.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

in a way that is suitable to capture the different behaviour of again in the two groups and that sheds some light on the other differences between them observed in section 3. In this section, I make a proposal to that effect. Section 5.1.1. suggests a semantics for goal-PPCs in (+R) languages. In section 5.1.2. we take some steps towards assigning a semantics to goal-PPCs in (
R) languages. Section 5.2. uses those suggestions to derive the cross-linguistic data with restitutive again, based on the structural theory of again. Finally, in section 5.3. we compare the predictions of our analysis to the predictions that a lexical ambiguity analysis of again could make. I should note that my suggestions in section 5.1. are geared specifically towards finding an explanation for the behaviour of restitutive again. I do not presume to do justice to all the semantic considerations that should enter into developing an analysis of goal-PPCs, much less to verbs of motion and path phrases in general. See, for example, Cresswell (1974, 1978), Jackendoff (1996), Krifka (1998) and Tenny (1995). However, our data from section 4 will ultimately have to be explained by any successful analysis of goal-PPCs, and crucial properties of my analysis can be expected to carry over.

32 There and Back Again: A Semantic Analysis (61)

a. Sally walked to the summit. b. [ Sally [1[ t1 [ walked [PP PRO1 to the summit]]]]]

(62) (63)

[PRO1 to the summit] / ke.ate(the_summit)(x1) a. [walked [PP PRO1 to the summit]] / kxke.walke(x) & $e#[BECOMEe#(ke* .ate*(the_summit)(x1)) & CAUSE(e#)(e)] b. Sally walked to the summit / ke.walke(S) & $e#[BECOMEe#(ke*.ate*(the_summit)(S)) & CAUSE(e#)(e)] Sally’s walking caused her to come to be at the summit.

Thus the sentence in (61a) will be true of an event iff it is a walking by Sally, and causes another event which is Sally’s coming to be at the summit. This is an appropriate description of the meaning of (61). (64) goes through the same steps for a second example (the reading described is the directional interpretation of (64a)). (64)

a. John swam under the bridge. b. [ John [1[ t1 [ swam [PP PRO1 under the bridge]]]]] c. ke.swime(J) & $e#[BECOMEe#(ke*.undere*(the_brigde)(J)) & CAUSE(e#)(e)] John’s swimming caused him to come to be under the bridge.

Note that (just like in the case of resultatives) application of principle (R) turns the predicate ‘walk to the summit’ or ‘swim under the bridge’ into an accomplishment, by virtue of introducing CAUSE BECOME. So if goal-PPCs are interpreted via principle (R) in languages that have resultatives (and therefore have to have principle (R)), we expect them to behave like accomplishments even if the verb lexically is not an accomplishment (as in the case of walk). On this proposal, goal-PPCs are a type of complex predicate construction in

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Let’s assume for the moment that to means at (this will be reconsidered in the next subsection). Then the PP denotes the predicate of events in (62), and hence would not be combinable with the verb walk by standard principles of interpretation, just like the result phrase of a resultative. Principle (R) applies as shown in (63). Thus ‘walk to the summit’ is a predicate that is true of an individual and an event iff the event is a walking by the individual, and causes a second event which is a coming to be at the summit of the subject of the small clause PP. Once more, interpretation of the whole structure results in binding of the small clause subject and filling of the individual argument slot.

Sigrid Beck 33

5.1.2 The semantics of goal PP constructions in (
R) languages. We now turn to the interpretation of goal-PPCs in (
R) languages. Obviously, the interpretation strategy developed for (+R) languages above cannot apply, since (
R) languages by assumption lack the relevant interpretation principle; nor do we want it to apply, since goalPPCs do not behave like accomplishments in (
R) languages. Hence we need an alternative way of combining the goal PP with the verb. I must acknowledge that the non-accomplishment-like behaviour of goal-PPCs in (
R) languages needs to be investigated in more detail than I can in this paper. I regard the analysis below as tentative. I will make a suggestion for an interpretation mechanism alternative to principle (R) to be involved that represents an independently motivated way of combining a verb with a PP modifier. The analysis meets two desiderata: Firstly, it doesn’t predict the combination of verb and PP to be an accomplishment, or telic, predicate. This is a requirement given the data discussed in section 3.2. Secondly, we want to locate the variation in the semantics of goal-PPCs between (+R) and (
R) languages in the way in which the PP is integrated compositionally, not with the semantics of the PP itself. It is not plausible that availability of priniciple (R) correlates with a systematic difference in the meaning of prepositions across languages. Therefore, we want to choose an analysis of goal PPs in which the semantics of the preposition can be the same for (+R) and (
R) languages. In particular, the semantics must be usable for the interpretation strategy with prinicple (R), and suitable for the description of the restitutive reading. These desiderata put limitations on the options we can consider. Note in particular that the goal of any analysis developed for, say, English goal PP predicates, will

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(+R) languages, as anticipated in section 3.2. In contrast to some of the other complex predicate constructions like resultatives and verbparticle constructions, they are grammatical strings in (
R) languages. The difference is not in grammaticality but in interpretation. Before proceeding to discuss the interpretation of goal-PPCs in (
R) languages, I should note that goal-PPCs in (+R) languages come out essentially as subject oriented resultatives. The existence of such resultatives seems to be a matter still under debate, cf. the discussion in Rothstein (to appear). It is beyond the scope of this paper to address the issue. We merely note that there have been some recent suggestions to regard data like (61) as subject oriented resultatives (e.g. Wechsler (1997); Rothstein (to appear) cites Rappaport Hovav & Levin (1999)); on the other hand, this is incompatible with Rothstein’s analysis of resultatives.

34 There and Back Again: A Semantic Analysis be to account for their accomplishment-like behaviour, which automatically makes it unsuitable for my purposes. An interpretation strategy that I found adaptible to present purposes emerges from Cresswell’s (1974, 1978) discussion. This is an analysis as a Davidsonian event modifier. Such an analysis amounts to the suggestion that the analysis of PP modifiers like ‘in the park’ in (65a) can be extended to (65b). (65)

a. Sally slept in the park. b. Sally walked in the park (from the station, along the beach, . . .).

(66)

a. ke.sleepe(S) & ine(the_park)(S) b. ke.walke(S) & ine(the_park)(S) c. [in the park]/ kxke.ine(the_park)(x)

(67)

Generalized PM: If a ¼ [b c] and b# is of type Æe,Æi, tææ and c# is of type Æe,Æi, tææ, then a#¼kxke.b#(x)(e) & c#(x)(e)

I suggest that this independently attested mechanism is available for the interpretation of goal-PPCs in (-R) languages; example (68a) is from Spanish. The resulting semantic representation is given in (68b). The

19 The informed reader will have noticed that our analysis departs from a strictly Davidsonian analysis in giving the preposition a subject argument (an individual); so in is not just a relation between an event and a location (and similarly for to etc.). This step is made necessary by our restitutive again data: the restitutive reading of (ia) is (ib), not (ic), hence the result state must include information on the holder of the result state; cf. our desideratum above. Thus if we want a uniform semantics for prepositions, it has to be one in which they take an individual argument. The same desideratum also precludes an analysis strictly following Krifka (1998). Note that, beyond the requirements of restitutive again, our semantics for prepositions allows interpretations for (ii) that are more straightforward.

(i) a. b. c.

Sally walked to the village again. Sally walked to the village, and she had been at the village before. Sally walked to the village, and there had been a previous event at the village.

(ii) a. b.

Sally is in the park. Sally pushed/kicked the box into the room.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In our framework (65a,b) would be interpreted as in (66a,b). (66b) is true of an event iff that event is a walking by Sally and Sally is in the park in (during) that event. The PP itself, then, has to denote a relation between individuals and events, as indicated in (66c).19 When the PP combines with the verb, we are thus trying to combine two expressions of type Æe, Æi, tææ. The mode of combination is basically conjunctive. I propose to use a generalized principle of predicate modification for this, given in (67).

Sigrid Beck 35

only difference between (68b) and (66b) is that we now have the preposition to. (68)

a. Pablo camino hasta la cima. Pablo walked up-to the summit b. ke.walke(Pablo) & toe(the_summit)(Pablo)

(69)

a. [[to]] (e)(x)(y) ¼ 1 iff at end(e), y is at the final point of a journey p(e) that ends at x b. [[end]] is that f: DÆiæ / DÆiæ such that for any e: f(e) ¼ the temporal end point of e c. p(e) (a path or a journey) is a function that maps times that are part of the running time of e to a spatial region (the space occupied by the path at that time).

(70)

Pablo walked, and by the end of the walk he was at the final point of a journey that ends at the summit.

A lot more can be said about paths. Compare for example, Cresswell (1978), Krifka (1998). For our purposes, we should note, with Cresswell, that (i) paths, or journeys, can be hypothetical, i.e. do not actually have to be made; and (ii) that contextual salience plays a role in choosing the relevant path associated with the event we are looking at. Point (i) is illustrated by data like (71). The ambiguity of (72) is plausibly a result of the context dependent underdeterminacy of paths. (71)

a. The band was playing across the river. b. The band was playing at the end of a hypothetical journey across the river. (72) Alice walked across the meadow. (72#) a. Alice walked along a path that crossed the meadow. b. Alice walked in a place that is across the meadow from here. Cresswell makes these issues much more explicit in his paper than I do here. I will simply assume that the paths in our interpretations can be

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

To make this a meaningful proposal, it is crucial to consider the semantic contribution of the preposition. We follow Cresswell (1978) in assuming that the semantics of prepositions like to makes use of the concept of a path, and in deriving the notion of a path from progress through time. Such a semantics for the preposition is given in (69a). (69a) says that to holds between an event e and individuals x and y iff at the end of e, y is at the final point of some path that ends at x. The function end is defined in (69b). The central notion of a path is defined in (69c). (68b) is paraphrased in (70).

36 There and Back Again: A Semantic Analysis

(73)

a. Juan nado debajo del puente (*en una hora). Juan swam under the bridge (in an hour) b. ke.swime(Juan) & undere(the_bridge)(Juan) Juan swam, and he was under the bridge during the swimming.

The assumption in (69) about the meaning of to is compatible with our analysis of (+R) goal PPs above. (63) for example becomes (74). Thus this analysis does not rely on a semantic difference in the prepositions in goal PPs — they may be exactly the same in a given pair of a (
R) and a (+R) language. The difference I rely on is in how the PP is integrated compositionally. It should be admitted, though, that nothing forces the preposition to be to rather than at in English; i.e. I predict that ‘Sally walked at the summit’ should have the same directional interpretation as ‘John swam under the bridge’. This problem exists for English to, into and onto, which offer directional alternatives to at, in and on. I do not predict that there is this restriction. The way it shows up is specific to English (Korean, for example, requires a directional marker with all prepositions). My hunch is that our understanding of the role of the BECOME component may not be quite right (see also Beck & Johnson (2004) on this issue). I will leave it as an open problem here. (74)

a. Sally walked to the summit/ ke.walke(S) & $e#[BECOMEe#(ke*.toe*(the_summit)(S)) & CAUSE(e#)(e)] b. Sally walked, and her walking caused her to come to be at the end of a path to the summit.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

hypothetical, and that in the case of motion verbs like walk, the verb plausibly brings a most salient path with it. This accounts for the fact that (72#a) is the more salient interpretation of (72), and also for the fact that it is not the only interpretation. Under this analysis, (68) is true of an event iff that event is a walking by Pablo and Pablo is at the summit by the end of it—more precisely: at the end of the event, Pablo is at the end point of some path to the summit. Assuming that we normally associate the path in the semantics of to with the motion expressed by the verb, this seems an adequate description of the meaning of (68) in (
R) languages. The denotation in (68b) thus seems suitably similar to goal PPs in (+R) languages. (73) provides another example from Spanish with a locative preposition; in contrast to (70#) the example only has a locative interpretation. The effect of presence v. absence of Principle (R) is in a way clearer with such ‘non-directional’ prepositions.

Sigrid Beck 37

Note, finally, that our analysis gives us no reason to believe that the predicate modifier way of combining the goal PP with the verb is unavailable in English and other (+R) languages. What’s crucial is that (+R) languages do have the combination via prinicple (R). We can follow Grolla (2003) in speculating that the ambiguity of (75a,b) below, observed by Higginbotham (2000), has its source in the two choices that we have in English for how to compositionally integrate the PP. (75)

a. I walked around the fountain (for 10 minutes/in 10 minutes). b. I ran upstairs (for 10 minutes/in 10 minutes).

(76)

*Sally walked along the beach in an hour.

To summarize the analysis: in (+R) languages, goal-PPCs are interpreted by the same principle that interprets resultatives. They are grammatical strings in languages that don’t have complex predicate constructions. However, they are not semantically identical. The goal PP expresses the result state of a complex predicate in (+R) languages, but a predicate modifier of some kind in (
R) languages. Goal-PPCs are therefore accomplishments, and telic predicates, in (+R) languages, but not in (
R) languages. This is the source of the cross-linguistic variation in the data in tables (51) and (59).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Before we turn to restitutive again, let us briefly consider modifiability with expressions like ‘in an hour’ on the basis of the proposed analyses of goal-PPCs. Dowty (1979) suggests that the semantics of the predicate must guarantee that there is a unique shortest time interval at which the predicate is true, for ‘in an hour’ to be acceptable. This distinguishes accomplishments from activities. Krifka’s (1998) analysis is similar in spirit: if a telic predicate is true of an event e, it is not true of any subevent that is temporally shorter. Krifka then defines the semantics of ‘in an hour’ in such a way that its implicatures can only be satisfied by predicates of events with a unique shortest running time. I hypothesize that the semantics of goal-PPCs in (
R) languages does not guarantee uniqueness. That is, if ‘walk to the summit’ is true of an event e, it is not guaranteed by the semantics that the predicate isn’t also true of a shorter subevent. The semantics suggested above is weak enough to incorporate this idea. Goal-PPCs in (
R) languages come out as non-telic, in Krifka’s sense. We then expect that modification with ‘in an hour’ is no better than it is with predicates like ‘walk along the beach’, cf. (76). On the other hand, the presence of BECOME in (+R) languages will suffice to satisfy the uniqueness requirement.

38 There and Back Again: A Semantic Analysis

5.2 The structural analysis of again

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

We are now in a position to provide an analysis of the data reported in table (59). We assume the structural theory of restitutive again from section 2.1, in which, recall, a restitutive reading requires an independent result state predicate that again can modify. And we will work with the semantics for goal-PPCs developed in the previous subsection. Let’s first look at what happens in (+R) languages. Given the semantics developed in section 5.1, the ambiguity of the example in (77) can be derived in the same way as the ambiguity of the resultative example in section 2.1. Once more, there are two possibilities as to what again modifies: either the result PP or the entire VP. The two LFs receive the translations in (79) and (80) respectively and yield the intuitive meanings of the repetitive and the restitutive reading of (77). (77) Sally walked to the summit again. (77#) a. Sally walked to the summit, and she had done that before. b. Sally walked to the summit, and she had been there before. (78) a. [ Sally 1 [VP [VP t1 walked [PP PRO1 to the summit]] again]] b. [ Sally 1 [VP t1 walked [PP [PP PRO1 to the summit] again]]] (79) ke$.againe$(ke.walke(S) & $e#[BECOMEe#(ke*.toe*(t_s)(S)) & CAUSE(e#)(e)]) (79#) Once more, Sally#s walking caused her to come to be at the end of a journey to the summit. (80) ke.walke(S) & $e#[BECOMEe#(ke$.againe$(ke*.toe*(t_s)(S)) & CAUSE(e#)(e)]) (80#) Sally’s walking caused her to come to be once more at the end of a journey to the summit. Let us take a step back and consider the general cross-linguistic predictions made by this theory. Remember that we expect a restitutive reading to be possible if there is a result state denoting constituent that the adverb corresponding to again can adjoin to, and if that particular adverb can find that constituent (in the sense of the visibility parameter for adverbs from section 2.1.). In (+R) languages, the PP in a goal-PPC is always an appropriate adjunction site for the adverb. In terms of visibility, what the adverb can do here should parallel other complex predicate constructions. Thus we expect a (+R) language to permit a restitutive reading with goal-PPCs as long as that language has an adverb to which complex predicate result states are visible (i.e. which has at least setting (ii) of the visibility parameter). This, as the table in (59) shows us, is overwhelmingly the

Sigrid Beck 39

case. Moving on to the predictions that the structural theory makes for restitutive readings with lexical accomplishment predicates, we expect that they are less freely available than with goal-PPCs. The reason is that visibility of the result state denoting constituent decreases; thus it would contradict our predictions to find a (+R) language in which restitutive readings with lexical accomplishment predicates are more acceptable than with complex predicate constructions. We find in table (59) the general tendency that we expect, and no counterexample of the sort just mentioned. We conclude that the structural theory of restitutive again makes predictions (C) and (D) formulated in section 4.2: (C)

Before we consider (
R) languages, a comment on Hungarian. I suggest that Hungarian does not have a genuine decomposition adverb again. The result states within a decomposition structure are almost universally invisible, in the sense of section 2.1. The result states of complex predicates seem to be visible to u#jra for most speakers, but not all. This is the difference between the two dialects of Hungarian. In a broad sense, then, Hungarian meets our expectations. Under the structural analysis of restitutive again combined with the visibility parameter, we expect that there are languages (like English) that allow restitutive readings with both lexical and complex predicate accomplishments. There may be languages (like Hungarian1, and Mandarin for the majority of speakers consulted) that allow restitutive readings with complex predicates, but not with lexical accomplishments. And there may be languages (like Hungarian2) that don’t allow a restitutive reading with either.20 Turning to (
R) languages, what can the structural theory of again say about why a restitutive reading for goal-PPCs in those languages is unavailable? I suggest that the behaviour of goal PPs in (
R) languages can be understood in view of the behaviour of PP modifiers in combination with again. Note that the data in (81)–(82) do not permit the (c)-reading. 20 Despite the fact that restitutive readings with lexical accomplisments are not very good in Japanese either, Japanese does not lend itself to such an analysis: there are speakers who accept a restitutive reading with lexcial accomplishements, but not with goal PP constructions. Besides it is clear that there is variation in the setting of the (R) parameter, so this is definitely the most plausible cause of the judgments I got for goal PPs.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

a (+R) language very often (but not necessarily) does permit restitutive readings with goal PP constructions. (D) In a (+R) language, a restitutive reading with a lexical accomplishment is as acceptable or less acceptable as with a goal PP construction (but not more acceptable).

40 There and Back Again: A Semantic Analysis (81)

a. b.

Sally slept/walked in the park again. Sally slept/walked in the park, and she had done that before. c. # Sally slept/walked in the park, and she had been there before. (82) a. Sally walked along the beach again. b. Sally walked along the beach, and she had done that before. c. # Sally walked along the beach, and she had moved along the beach before.

(A) (B)

a (
R) language never permits restitutive readings with goal PP constructions. a (
R) language may or may not permit a restitutive reading with lexical accomplisments.

Finally, there is prediction (E). As for between-speaker variation with lexical accomplishments, I propose that not everybody actually arrives at setting (iii) of the visibility parameter of the adverb. Within-speaker 21

This fact is interesting of itself and not necessarily expected under all analyses of PP modifiers —e.g. a strictly Davidsonian analysis, in which the PPs would be of the right semantic type to be modified by again (cf. fn 19). 22 Rapp & Stechow (1999) propose that the preposition to is decomposed into BECOME and at. If this were possible, then a restitutive reading could arise by applying again to the at-predicate. Their suggestion is thus incompatible with our cross-linguistic data.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

These data show us that again cannot apply to just the PP modifier.21 A goal PP would have to be interpreted in the same way as the PPs in (81)–(82) under our assumptions (an interpretation as a result phrase is impossible due to lack of principle (R)). Hence whatever prevents again from modifying ‘in the park’ in (81) and ‘along the beach’ in (82) will prevent it from modifying ‘to the village’ in (
R) languages. On a more technical level, the modifiers are of the wrong semantic type to be combined with again (they are Æe, Æi, tææ, not Æi, tæ), so they are not a possible adjunction site for again. Thus our framework formally captures the lack of a restitutive reading.22 I conclude that the structural theory can make prediction (A), given plausible assumptions about the modifier role of the goal PP in (
R) languages and given the general pattern of again with modifiers. We still have to address prediction (B). Prima facie, the structural theory does not make a prediction about whether a given (
R) language has a decomposition adverb corresponding to again or not. Our result that there are languages that do and others that do not have such an adverb is quite compatible with the structural theory.

Sigrid Beck 41

variation across different lexical accomplishments might suggest that a speaker has to learn that a decomposition analysis is available for a given verb. (E)

there is room for variation in the judgments for lexical accomplishments.

In conclusion, the structural theory of again, in combination with an appropriate analysis of how the semantics of goal-PPCs varies along with the setting of the (R) parameter, can explain the cross-linguistic pattern in the acceptability of restitutive again and generalization (G$). (G$)

A comment on further empirical predictions: The cross-linguistic variation in the availability of a result state modifying reading of a forPP is amenable to the same analysis—recall the contrast from section 3 between English and Spanish in (83). Only in (+R) languages can the goal PP express a result state. The for-PP can modify a result state (English), but not a simple modifier (Spanish). Compare Stechow (1995) for a suitable analysis of for-PPs. We make the further prediction that the behaviour of other decomposition adverbs (specifically almost) should be similar to that of again and ‘for an hour’—in particular, it should not be able to modify a (
R) goal PP. Testing the prediction is left for another occasion. (83)

a. b. ?? c.

Suresh walked to the village for 10 minutes. Suresh anduvo hasta la aldea durante diez minutos. Suresh walked to he village for 10 minutes Suresh walked to the village, and he stayed there for 10 minutes.

5.3 The lexical analysis of again In this subsection, I examine the predictions that the lexical ambiguity theory makes regarding our cross-linguistic data. We begin by considering how a restitutive reading for our goal PP example in (84a) would be obtained under this analysis. (84b) is true of an event e iff e is a walking to the village by Suresh, and there is a previous event e# of which the counterdirectional predicate Pc is true, and whose result state was the starting point of Suresh’s walk to the village. Let’s make the assumptions in (85) about the concepts involved; the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

A goal PP construction in combination with again has a restitutive reading only in languages that have the positive setting of the complex predicate parameter.

42 There and Back Again: A Semantic Analysis counterdirectional predicate to a walking to the village by Suresh is Suresh leaving the village (or possibly Suresh walking away from the village). (84b) then means that Suresh walked to the village, and that there is a preceding event in which Suresh leaves the village, and the result of the leaving is the prestate of the walk to the village. This seems to be an equally appropriate description of the restitutive reading as the one we derived before, using the structural analysis. (84)

&

Pc(e#)

&

predicate P: ke.walk_to_the_villagee(S) counterdirectional predicate Pc: ke.leavee(the_village)(S) ks. ; ats(the_village)(S) possible prestates of P, preP: possible result states of P, resP: ks.ats(the_village)(S)

Observe that in the representations above, I leave ‘walk to the village’ as an internally unanalyzed predicate of events. Its internal composition is irrelevant for the purposes of the counterdirectional again analysis; hence its proponents need not be committed to any particular analysis of the goal-PPC, as long as the resulting semantics is that of a reversible predicate of events. Remember that what is required for a restitutive reading on this analysis is an adverb corresponding to counterdirectional again on the one hand, and a predicate of events that makes available a counterdirectional predicate (plus suitable prestates and result states) on the other. With that in mind, let’s now consider the cross-linguistic predictions of the lexical ambiguity theory. Note that the counterdirectional predicates for ‘walk to the village’ and ‘open the door’ are in fact given explicitly in the contexts in which our sentences are presented (leaving the village and closing the door, respectively). The result states of those counterdirectional predicates are indeed the prestates of our (attempted) counterdirectional readings in both stories. In all our languages, the goal-PPC is a change of state predicate (moving from a state of not being at the village to being at the village). Therefore, I conclude that the predicate of events expressed by the goal-PPC should be reversible in all the languages I considered. There are no conceptual reasons that would prevent a restitutive/counterdirectional reading from being available. What could still go wrong is that a language lacks counterdirectional again. Given that, how much of a correlation between the (R) parameter and availability of restitutive readings with goal-PPCs can we expect the lexical analysis to capture?

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(85)

a. Suresh walked to the village again. b. ke.walk_to_the_villagee(S) & $e#[e#<e resPc(e#) ¼ preP(e)]

Sigrid Beck 43

(A) (B)

A (
R) language never permits restitutive readings with goal PP constructions. A (
R) language may or may not permit a restitutive reading with lexical accomplishments.

Thus we see that it is to the advantage of the structural theory that it makes reference to the syntactic make-up and internal composition of the predicate of events that again combines with. This allows us to capture the connection to the setting of the (R) parameter. Conversely, the fact that the internal structure and composition of a predicate are irrelevant under the lexical theory prevent it from making that connection. Goal-PPCs lead us to the conclusion that restitutive again requires a structure-based analysis.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The crucial problem are (
R) languages. There is no reason to expect any connection between the setting of the (R) parameter and existence of counterdirectional again in a given language. That is, there is no theoretical connection between absence of resultatives and unavailability of a restitutive/counterdirectional reading with goalPPCs. This is because the internal structure of the goal PP predicate is irrelevant under this analysis. The clear correlation that we observe is thus on general grounds quite unexpected. In the (
R) languages, the data for goal-PPCs should pattern with the results for lexical accomplishments in table (59) (since everything hinges on whether or not the language has counterdirectional again), but they do not. Restitutive readings with lexical accomplishments are much more widely accepted among (
R) languages. Concretely, the judgments we collected for lexical accomplishments in Indonesian, Hebrew, Kannada, Spanish and Tagalog clearly show that these languages would have to have counterdirectional again . Hence, any theory that derives those judgments from the existence of a counterdirectional/restitutive meaning of again wrongly predicts these (
R) languages to permit a restitutive reading with goal-PPCs. I conclude that the lexical theory cannot systematically make prediction (A), and concretely, it cannot make prediction (A) while making prediction (B). It is worth pointing out that this conclusion does not depend on the exact nature of the lexical ambiguity. I chose Fabricius-Hansen’s analysis as a representative, but no matter what particular meaning for ‘restitutive’ again a lexical analysis assumes, the systematic connection of restitutive readings with goalPPCs with the (R) parameter, and the absence of such a connection with lexical accomplishments, is not expected under this type of analysis.

44 There and Back Again: A Semantic Analysis 6 CONCLUSIONS The main results of this paper are summarized below. Needless to say, there are other arguments to consider in the comparison of the structural and the lexical analysis of restitutive again. And there are more considerations that figure into a proper semantics for goal-PPCs than the ones discussed here. A new type of evidence has been brought forth that has to be taken into account by any analysis of goal-PPCs, and by any theory of restitutive again. (i)

Besides the obvious consequences which I have argued for throughout the paper, the analysis has some further repercussions. It supports Snyder’s Complex Predicate Parameter. It also supports the idea that resultatives are to be interpreted by some non-standard mode of composition, by showing that (non-) availability of that mode of composition has widespread and systematic consequences. Hence our position that principle (R) is a crucial part of the Complex Predicate Parameter. This makes it a semantic parameter: (non-)availability of a principle of semantic composition. Strengthening the structural theory of again is itself not without consequences. A strictly structural theory implies that restitutive again shows us which properties of events syntax makes available. This has repercussions for how much of a Parsonian event semantics can be mapped into the syntax (see Stechow (1996) for some discussion) and for the semantics of PP modifiers (cf. the discussion in section 5). These issues will have to be left for another occasion. Acknowledgments The work reported in this paper started out as a joint project with William Snyder (Beck & Snyder (2001a,b)). His influence and the influence of his work are visible throughout the paper, and are here gratefully acknowledged. The following people have provided invaluable help with collecting and understanding the data: Lydia Grebenyova, Elaine Grolla, Shin-Sook Kim, Diane

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The internal composition of goal PP constructions, and the resulting semantics, are subject to systematic cross-linguistic variation. They are complex predicates and denote accomplishments in (+R) languages, but not in (
R) languages. (ii) The cross-linguistic pattern of availability of restitutive readings shows that the analysis of restitutive again must make reference to structural factors. An analysis that depends on conceptual and lexical factors only cannot derive the correlation with the setting of the (R) parameter.

Sigrid Beck 45

Received: 20.11.03 Final version received: 08.07.04

SIGRID BECK University of Connecticut 337 Mansfield Road, U-1145 Storrs, CT 06269 USA e-mail: [email protected]

APPENDIX A: GOAL PPS AND TEMPORAL IN-PHRASES ACROSS LANGUAGES French: German: Hebrew: Hindi/ Urdu:

Jean a marche´ au sommet en une heure. Jean has walked to-the summit in one hour (2) Ottilie ist in einer Stunde zum Gipfel gelaufen. Ottilie is in one hour to-the summit walked (3)* Dan halax el ha-kfar tox Sa’a. Dan walked to the-village in an hour (1)*

(4)* Veneeta do ghante Veneeta two hours

mein in

summit ki taraf chal-ii summit towards walk-Perf. Gen-F.Sg F.Sg

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Lillo-Martin, Luisa Martı´, Deborah Chen Pichler, Katie Plavetic, Yael Sharvit, Arthur Stepanov, Koji Sugisaki and Douglas Wharram. Our (other) informants have been extremely helpful, thorough and patient—thanks to Alimah (Bahasa Indonesia); Xabier Artiagoitia and Ricardo Etxepare (Basque); Ce´dric Boeckx, Caroline Fe´ry and Miche`le Bacholle (French); Ron Artstein, Nurit Assayag, Daphna Heller and Orr Ravitz (Hebrew); Miriam Butt, Veneeta Dayal, Tipu Fateh, Utpal Lahiri and Anees Shaik (Hindi/Urdu); Eva Bar-Shalom, Gabor Bobok, Gabor Drasny, Janina Rado and Anna Szabolcsi (Hungarian); Maria Mikaratsuk Dicker, Nat Igliorte and Ellen Ford (Inuttut); Takako Iseda, Keiko Ishii, Nobu Miyoshi, Fumi Niinuma, Masashi Nomura and Toshiko Oda (Japanese); Jyothirmai Narayan and Chandu Visweswariah (Kannada); Raymond Lamm (Khmer); Inkie Chung, Bosook Kang, Sei-Rang Oh and Bum-Sik Park (Korean); Haldor Heimer (Lingala); Liang Chen, Pan Ning Chen and Kirsten Y Lin (Mandarin); Eva Bar-Shalom, Evgeni Gousev, Alex Lavrentiev, Arkady Polyak and Natasha Rakhlin (Russian); Zeljko Boskovic, Nada Jevtic, Marin Pichler, Sandra Stepanovic and Mihajlo (Serbian/Croatian); Adolfo Ausin, Rene Builes, Katrin, Regine and Paco Garcia, Patricia Molina, Lara Reglero and Emma Ticio (Spanish); Teresita Graham, Regina So and Maricar Tarun (Tagalog). The ASL stories had to be filmed, which would never have been possible without the help of Deborah Chen Pichler; thanks also to signer Doreen Simmons-Marques, and to consultants Nancy Fuller, Robin Shaw and Mary Tomasian. I would like to thank Cathrine Fabricius-Hansen and Arnim von Stechow for their help. Many thanks also to several annonymous reviewers, and to the audiences at the University of Connecticut, SULA (UMass Amherst), Universita¨t Konstanz, Universita¨t Tu¨bingen and Universita¨t Stuttgart, for comments and feedback. And finally, I am very grateful to Thilo Go¨tz, Emma Hoy, Laura Kosbar and Reinier van den Born.

46 There and Back Again: A Semantic Analysis Hungarian: (5) Japanese: Kannada: Khmer:23 Korean:24

Suresh 2 o’ra alatt else’ta’lt a faluig. Suresh 2 hour in Perf.walked the village-to. (6) Suresh-ga 2 jikan-de mura-made arui-ta. Suresh-Nom 2 hours-in village-to walk-Past (7) Suresh eradu ganteyalli hallige nadedanu. Suresh two hours-in village-to walked (8) Joe dae(r) tiu kompul knong mu:ey maong. Joe walk go/to summit within one hour (9) Suresh-nun 10 pun-mane maul-lo tallie ka-(a)ss-ta. Suresh-Top 10 minute-in village-Dir run go-Past-Decl

APPENDIX B: RESTITUTIVE AGAIN ACROSS LANGUAGES ASL:25

(12)

TURN-TABLE-UPSIDE-DOWN AGAIN ‘I turned the table over again.’

23 Khmer does not seem to distinguish between verb and preposition, hence the ambiguous gloss for tiu. 24 The Korean example contains two verbs, tallie ‘run’ and ka ‘go’. This is obligatory in Korean—the simple verb ‘run’ cannot combine with a goal PP. One might be concerned that the combination ‘run-go’ is no longer an activity predicate, but rather by itself already an accomplishment. In that case Korean would not have a goal PP construction in our sense, and the Korean data would be irrelevant to the point I am making. However, the second verb in these constructions (‘come’ or ‘go’), while obligatory in Korean, is optional in Japanese. Interestingly, presence or absence of this verb made no difference to the judgments I got for Japanese. In particular, the speakers who rejected a restitutive reading with a goal PP construction did so both with and without the second verb. At the same time, some of those speakers accepted restitutive readings with lexical accomplishments. This suggests that the second verb (‘go’ in our examples) does not turn the predicate into an accomplishment. I therefore leave the examples in our paradigm as goal PP constructions. One could consider an analysis as a perspective marker for the verbs ‘come’ and ‘go’ in these constructions. This seems to be Slobin & Hoiting’s (1994) idea, who call these verbs deictic. 25 Signs are represented by English glosses morpheme for morpheme in these examples. The ASL data reported here differ from the other examples in the survey because the signed versions of our standard examples suffered from interfering factors specific to ASL. TURN-TABLE-UPSIDEDOWN is one verb signed with classifiers, which we counted as a lexical accomplishment predicate. We should note that Slobin and Hoiting (1994) would classify our example of a goal PP construction SWIM ARRIVE ISLAND as a serial verb construction, presumably implying a structure [ [SWIM ARRIVE] ISLAND] rather than the, for our purposes relevant, [ SWIM [ARRIVE ISLAND]]. They do not, however, specify a reason for this that would be compelling under our theoretical assumptions. We leave the ASL data in the survey with the proviso that the issue needs to be clarified.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Mandarin: (10) Wo shi fen zhong nei (qu) zou dao le nei-ge cunzi. I ten minutes in (go) walk to Past that village Spanish: (11)* Juan anduvo hasta la cima de la montana en una hora Juan walked to the summit of the mountain in one hour

Sigrid Beck 47 (13) Bahasa (14) Indonesia: (15) French:

(16) (17)

German:

(18)

Hebrew:

(20)

Dan Dan (21) Dan Dan

Hindi/Urdu: (22)

(23)

Hungarian:26

(24) (25)

Inuttut:

Japanese:

patax opened halax walked Sally Sally

et Object-marker el ha-kfar to the-village

ne Erg

phir se again Inst

ha-xalon me-xadaS. the-window again me-xadaS. again daarwaazaa door.M.Sg.Nom

khol-aa openperf.M.Sg Veneeta phir gaun k-i taraf chal-ii. Veneeta again village.M.Sg Gen-F.Sg towards walkPerf.F.Sg

Sally u#jra kinyitotta az ajto’t. Sally again opened the door Suresh u#jra else#talt a faluig. Suresh again Perf-walked the village-to

(26)

Suresh KakKaup kaangaanut pisu-gialla-juk. Suresh mountain’s top walk-again-3sAgr (27) Holda ukuamik ukuisi-gialla-juk. Holda door open-again-3sAgr (28)

Sally-ga futatabi doa-o ake-ta. Sally-Nom again door-Acc open-Past (29) Suresh-ga futatabi mura-made arui-ta. Suresh-Nom again village-to walk-Past

26 I chose a prefective form of the verb ‘walk’ in the Hungarian example (25). The reader might be concerned that the perfective aspect suffices to turn the verb into an accomplishment predicate, even without the goal PP. Note, however, that this is not the case in Serbian/Croatian, where we also chose a perfective form of ‘walk’. A restitutive reading is impossible in Serbian/ Croatian despite that.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(19)

SWIM ARRIVE ISLAND AGAIN ‘He swam to the island again.’ Sally membuka pintu lagi. Sally opened door again Suresh jalan lagi ke desa. Suresh walked again to village Sally a ouvert de nouveau la porte. Sally has opened again the door Jean a marche´ de nouveau au sommet. Jean has walked again to-the summit Sally hat die Tu¨r wieder geo¨ffnet. Sally has the door again opened Suresh ist wieder zum Dorf gelaufen. Suresh is again to-the village walked

48 There and Back Again: A Semantic Analysis Kannada:

(36)

Sally akangoli ekuke lisusu. Sally open.3rd.Sg.F.Past door again (37) Suresh atamboli na mboka Suresh walk.3rd.Sg.M.Past (to) home

Mandarin:

(38)

Sally you kai le men. Sally again open Past door (39) a. Suresh you zou dao nei-ge cunzi qu le. Suresh again walk to that village go Past b. Xiao niao you fei jin le niao chao. little bird again fly into Past bird nest

Serbian/Croatian:

Spanish:

Tagalog:

lisusu again

Sally je ponovo otvrila vrata. Sally Aux again opened door (41) Ponovo je odpesacio u selo. again Aux Perf.walked.3rd.Sg to village (42) Sally la abrio` otra vez. Sally it opened again (43) Suresh anduvo hasta la aldea otra vez. Suresh walked to the village again (44)

(40)

Binuksan Opened (45) Lumakat walk

ulit ni again Det ulit si again Det

Sally Sally Suresh Suresh

ang pinto. Det door sa bayan. Det village

27 For Korean and Mandarin I report two goal PP constructions because the first data I tested contained a confounding factor that confused the judgments somewhat. This turned out to be irrelevant to our concerns; the judgments I got for the second sentence were clear and unanimous.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Suresh tirga/punaha hallige nadedanu. Suresh again village-to walked (31) Sally thirga/punaha bagalu thegedalu. Sally again door opened Khmer: (32) Suresh dae(r) tiu phu:m(i) ven. Suresh walk go/to village again (33) Sally baek tvi:e(r) ven. Sally open door again Korean:27 (34) Sally-nun mun-ul tasi yel-ess-ta. Sally-Top door-Acc again open-Past-Decl (35) a. Suresh-ka tasi maul-lo tallie ka-(a)ss-ta. Suresh-Nom again village-Dir run go-Past-Decl b. Sae-tul-i tasi tungchi-lo nal-a ka-(a)ss-ta. bird-Pl-Nom again nest-Dir fly go-Past-Decl Lingala:

(30)

Sigrid Beck 49

REFERENCES Fabricius-Hansen, Cathrine (1995) ‘Wiedersehen mit wieder’. Unpublished lecture notes. Fabricius-Hansen, Cathrine (2001) ‘Wi(e)der and Again(st)’. In C. Fery & W. Sternefeld (eds). Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow. 101–130. Fery, C. & Sternefeld, W. (eds) (2001) Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow. Akademie Verlag. Berlin. Goldberg, A. (1995) Constructions. A Construction Grammar Approach to Argument Structure. The University of Chicago Press. Chicago and London. Green, G. (1973) ‘A syntactic syncretism in English and French’. In B. Kachru et al. (eds), Issues in Linguistics. University of Illinois Press. Urbanachampaign. 257–278. Grolla, E. (2003) ‘Prepositions, scales and telicity: A case study’. To appear in the Proceedings of WCCFL 23. Higginbotham, J. (2000) ‘On events in linguistic semantics’. In J. Higginbotham, F. Pianesi & A. C. Varzi (eds), Speaking of Events. Oxford University Press. Oxford. Jackendoff, R. (1990) Semantic Structures. MIT Press. Cambridge, MA. Jackendoff, R. (1996) ‘The proper treatment of measuring out, telicity and perhaps even quantification in English’. Natural Language and Linguistic Theory 14:305–354. Ja¨ger, G. & Blutner, R. (2000) ‘Against lexical decomposition in syntax’. In A. Z. Wyner (ed), Israeli Association for Linguistics 15. Haifa University. 113–137. Kamp, H. & Rossdeutscher, A. (1994) ‘DRS-construction and lexicallydriven inference’. Theoretical Linguistics 20:165–235.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Aske, J. (1989) ‘Path predicates in English and Spanish: A closer look’. In Proceedings of Berkely Linguistic Society 15. Beck, S. & Snyder, W. (2001a) ‘Complex predicates and goal PPs: Evidence for a semantic parameter.’ In: A. H.-J. Do, L. Dominguez and A. Johansen (eds), Proceedings of the 25th Boston University Conference on Language Development. Cascadilla Press. Somerville, MA. 114–122. Beck, S. & Snyder, W. (2001b) ‘The resultative parameter and restitutive again’. In C. Fery & W. Sternefeld (eds). 48–69. Beck, S. & Johnson, K. (2004) ‘Double objects Again.’ Linguistic Inquiry 35:97–124. Bittner, M. (1999) ‘Concealed causatives.’ Natural Language Semantics 7: 1–78. Cresswell, M. (1974) ‘Adverbs and events.’ Synthese 28:455–481. Cresswell, M. (1978) Prepositions and points of view. Linguistics and Philosophy 2:1–41. Davidson, D. (1967) ‘The logical form of action sentences’. In N. Rescher, (ed), The Logic of Decision and Action. Pittsburgh University Press. Pittsburgh 81–95. Den Dikken, M. (1995) Particles: On the Syntax of Verb-Particle, Triadic and Causative Constructions. Oxford University Press. New York. Dowty, D. (1979) Word Meaning and Montague Grammar. Kluwer. Dordrecht. Fabricius-Hansen, C. (1983) ‘Wieder ein wieder? Zur Semantik von wieder.’ In R. Baeuerle, C. Schwarze, & A. von Stechow (eds), Meaning, Use and Interpretation of Language. De Gruyter. Berlin, 97–120.

50 There and Back Again: A Semantic Analysis for functional adverbs’. Journal of Semantics 16:149–204. Rothstein, S. (to appear) Secondary predication and aspectual structure. In C. Fabricius-Hansen, E. Lang & C. Maienborn (eds). ZAS Papers in Linguistics. Slobin, D. & Hoiting, N. (1994) ‘Reference to movement in spoken and signed languages: Typological considerations’. In Proceedings of Berkeley Linguistic Society 20:487–505. Snyder, W. (1995) Language Acquisition and Language Variation: The Role of Morphology. PhD dissertation, MIT. Distributed by MIT Working Papers in Linguistics, Cambridge, MA. Snyder, W. (2001) ’On the nature of syntactic variation: Evidence from complex predicates and complex wordformation. Language 77:324–342. Stechow, A. von (1995) ‘Lexical decomposition in syntax’. In U. Egli, P. E. Pause, Ch. Schwarze, A. von Stechow & G. Wienhold (eds), The Lexicon in the Organization of Language. John Benjamins. Amsterdam and Philadelphia, 81–118. Stechow, A. von (1996) ‘The Different Readings of Wieder ‘‘Again’’: A structural account’. Journal of Semantics 13:87–38. Stowell, T. (1983) ‘Subjects across categories’. The Linguistic Review 2– 3:285–312. Talmy, L. (1985) Lexicalization Patterns: Semantic Structures in Lexical Forms. In T. Shopen (ed.), Language Typology and Syntactic Description. Volume 3: Grammatical Categories and the Lexicon. Cambridge University Press. Cambridge, 57–149. Talmy, L. (1991). Path to Realization— via Aspect and Result. In Proceedings of Berkeley Linguistic Society 17:480–519.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Klein, W. (2001) ‘Time and again’. In C. Fery & W. Sternefeld (eds), Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow. 267–286. Krifka, M. (1998) ‘The origins of telicity’. In S. Rothstein (ed.), Events and Grammar. Kluwer. Dordrecht, 197–235. Larson, R. (1988) ‘Light predicate raising’. Lexicon Project Working Papers #27. MIT Center for Cognitive Science. Cambridge, MA. Levin, B. & Rapoport, T. R. (1988) ‘Lexical subordination’. In L. MacLeod, G. Larson & D. Brentari (eds). Proceedings of CLS 24. Chicago. Chicago Linguistics Society. Levin, B. & Rappaport Hovav, M. R. (1995) Unaccusativity. At the SyntaxLexical Semantics Interface. MIT Press. Cambridge, MA. Lewis, D. (1973) ‘Causation’. The Journal of Philosophy 70:556–567. MacWhinney, B. & Snow, C. (1985) ‘The child language data exchange system’. Journal of Child Language 12:271–296. MacWhinney, B. & Snow, C. (1990) ‘The child language data exchange system: An update’. Journal of Child Language 17:457–472. MacWhinney, B. (2000) The CHILDES Database: Tools for Analyzing Talk. 3rd edition, volume 2: The Database. Lawrence Erlbaum Associates. Mahwah, New Jersey. McCawley, J. D. (1968) ‘The role of semantics in a grammar’. In E. Back & R. Harms (eds), Universals in Linguistic Theory. Holt, Rinhart & Winston. New York, 124–169. Rappaport Hovav, M. & Levin, B. (1999). ‘A reevaluation of the direct object restriction on English Resulatives’. MS, Bar-Ilan University and Stanford. Rapp, I. & Stechow, A. von (1999). ‘Fast ‘‘almost’’ and the visibility parameter

Sigrid Beck 51 Tenny, C. (1995) ‘How motion verbs are special’. Pragmatics and Cognition 3:31–74. Vendler, Z. (1967) Linguistics in Philosophy. Cornell University Press. Ithaca, New York.

Washio, R. (1997). ‘Resultatives, Compositionality and Language Variation’. Journal of East Asian Linguistics 6:1–49. Wechsler, S. (1997) ‘Resultative predicates and control’. Proceedings of the 1997 Texas Linguistics Society Conference. 307–321.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Journal of Semantics 22: 53–96 doi:10.1093/jos/ffh017

What’s in two names? THOMAS EDE ZIMMERMANN Johann Wolfgang Goethe Universita¨t

Abstract As first pointed out by Saul (1997a), the co-referential names in sentences like (1) and (2) defy substitution salva veritate: Clark Kent went into the phone booth, and Superman came out. I never made it to Karl-Marx-Stadt, but I visited Chemnitz last year.

This paper elaborates and compares two solutions to Saul’s substitution problem, both of which turn on an asymmetry between names that share their bearers. According to the first solution there is a semantic distinction between neutral names (like ‘Superman’ and ‘Karl-Marx-Stadt’) and restricted names (like ‘Clark Kent’ and ‘Chemnitz’). According to the second solution, only neutral names are properly used, whereas the use of (what would be) restricted names involves code-switching and pragmatic re-interpretation. As it turns out, the semantic approach deals more easily with changing names as in (2), whereas the pragmatic account is more adequate in explaining hidden identity cases like (1).

1 TWO OBSERVATIONS Common prejudice has it that nothing and nobody bears more than one name. When being introduced to someone, you do not ask for one of their names; neither do you expect what you find on a place name sign to be just one way of calling the town. To be sure, the town may have different names in different tongues, and a person may be called different names by different people or on different occasions, but within one language and context, one tends to think of naming as a one-one mapping. Expectations seem deep-rooted in this respect. This is why semantics texts usually have to remind students that distinct proper names may denote the same object nevertheless.1 The observation I am interested in is not that the naive expectation is not borne out. Rather it concerns the more basic fact that there is such a naive expectation in the first place. For clearly, from the 1 Cf. Dowty et al. (1981: 17): ‘It is [. . .] allowed for one and the same individual to have two or more names (just as ‘‘Samuel Clemens’’ and ‘‘Mark Twain’’ are names of the same person).’


The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(1) (2)

54 What’s in Two Names pre-theoretic point of view, the case of ‘Mark Twain’ and ‘Samuel Clemens’ is exceptional.2

Observation 1

There are simple sentences [footnote suppressed] which evoke anti-substitution intuitions quite similar to those evoked by attitude-reporting sentences. Saul (1997a: 102) The point is:

Observation 2 Non-substitutivity of coreferential names may occur in extensional contexts. The example Saul uses to drive the point home is: (1) Clark Kent went into the phone booth, and Superman came out. 2 Observation 1 is usually regarded as irrelevant to semantic theory, though it has been connected to the pragmatics of proper names; cf. Lerner & Zimmermann (1991: 364), where the following principle is defended along largely Gricean lines:

(M) Do not use different names for the same thing in the same context. 3

It is questionable, of course, whether Frege’s (1892) original examples, the German noun phrases ‘der Morgenstern’ and ‘der Abendstern’, are proper names at all (as opposed to definite descriptions). Frege himself did not seem to take the matter important.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Names are generally expected to be unique. It seems as if the observed expectation is not always fulfilled. Exceptional though Mark Twain may be, the fact that he has two names is not. Quite generally, pen names and other pseudonyms, historical place names, under cover identities, erroneous re-namings etc. all go to show that the contention that names are unique is a naive overgeneralization. Once this has been recognized, double names are made out all over the place. In particular, they are found in the semanticists’ and philosophers’ of language favourite environment of attitude reports, where they appear to fulfil the specific function of presenting their bearers from different angles, and where they defy substitution salva veritate. Starting with Frege’s Venus,3 the neverending debate has focussed on the question of how presentational function and non-substitutivity of double names pertain to their meanings and how they relate to the interpretation of attitude reports at large. However, more recently dug up examples indicate that neither double naming-cum-presenting nor substitution failure are a matter of attitudes or even intensional contexts in general:

Thomas Ede Zimmermann 55

Here a substitution of ‘Superman’ by ‘Clark Kent’ appears unwarranted, even though the two names are supposed to be known to belong to the same person. Moreover, as Saul (1997a: 103) already mentioned, Observation 2 can also be substantiated by examples involving changing place names: (2)

I never made it to Karl-Marx-Stadt, but I visited Chemnitz last year.

2 CHANGING NAMES: A SEMANTIC ACCOUNT In this section, I will introduce a rather obvious semantic way of dealing with examples like (2). At the end of the section I will discuss the kind of asymmetry announced above and adapt the semantics of changing names accordingly (which will not present a great problem). Only then, in Section 3, will I turn to the question of whether the semantic approach of the present section can be extended to double names in general.

2.1 Restricted names As far as Chemnitz is concerned, we may distinguish three phases: the time before, while, and after it was called ‘Karl-Marx-Stadt’. The middle phase will be referred to as the S-Phase; the other eras are C-Phases. The distinction concerns the history of the official names of the city of Chemnitz. An official name of a place is one that is used in legal

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(I have allowed myself to adapt Saul’s original example to my German habitat.) In the debate following Saul’s discovery, various proposals for accounting for Observation 2 have been made, some of which will be discussed in sections 2 and 3. In the course of the discussion it will be revealed that one aspect of the data appears to have gone unnoticed. The aspect concerns certain substitutional asymmetries that go against a hidden assumption made throughout the debate, viz. that the various names of a given object are, as it were, semantically on a par. In section 2, substitution puzzles like (2) will be accounted for in semantic terms, reducing the asymmetry of changing place names to (systematic) ambiguity. In Section 3 it will be shown that such a semantic approach runs into problems when it comes to cases like (1) involving hidden identities of persons. In section 4 I will therefore explore a pragmatic explanation that builds on the prejudice that no object is supposed to have more than one name. It will turn out that the pragmatic approach fares better with cases like (1) than its semantic does. On the other hand, unlike the latter it cannot naturally deal with (2). If this is right, the two examples appear to call for different accounts and are not as parallel as it seems.

56 What’s in Two Names

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

documents, government publications and the like, and at a given time it is unique (as far as I know). However, as the case of Chemnitz shows, official names may change over time. Such changes take place instantaneously, usually at midnight or immediately after a declaration. Concomitant to the instantaneous change of an official name there is frequently a gradual change of speakers’ habits. For instance, once Chemnitz had been officially renamed in 1953, more and more people in that city started using the new name when referring to their hometown, thereby eventually replacing the old official name. And the habit spread through East Germany and other countries. Still, for whatever reasons, a few traditionalists may have continued to use the old name; as far as their referential habits were concerned, the official change was non-existent. After the German reunification history repeated itself, with the place names swapping their roles. And again the instantaneous change of official names was followed by a gradual change of people’s referring habits, leaving out a few hardliners who may continue to use the name ‘Karl-Marx-Stadt’ (deploring linguistic imperialism) and those traditionalists who did not have to change (having seen it coming all along). However, the majority appears to have behaved conformistically, gradually switching from the ideology-loaded name back to the original one. In the present context, we may restrict attention to the conformists’ usage of the names in our time, i.e. long after the fall of the iron curtain in 1990. Neither linguistic change, nor sociological variation will be relevant to what I am saying, but it is important to realize that these phenomena exist, if only to not confuse them with the problems at stake. Even if linguistic conformists like me have long stopped calling Chemnitz ‘Karl-Marx-Stadt’, we still understand and use sentences like (2). Hence although we do not normally call the city by its communist name, we have not stopped using that name altogether. We do use it, but only in particular connections. For instance, (2) means that Fritz never visited Chemnitz between 1953 and 1990, but only before 1953 or after 1990. Thus, intuitively speaking, the name ‘Karl-Marx-Stadt’ is used to refer to Chemnitz during a certain time. What I will be concerned with is precisely this use of the name. There are various ways of implementing the intuition that ‘KarlMarx-Stadt’refers to Chemnitz at a certain time. I will only look into one of these accounts, not because they are all equivalent—they are not—nor because the account is more adequate than its contenders—it isn’t; but it is simpler, and the differences between the various possible analyses do not bear upon the present discussion: a more adequate analysis along the same lines would treat the restriction (3b) as a presupposition (rather than part of the assertion) and the name as a designator (rather than a quantifier).

Thomas Ede Zimmermann 57

However, the differences only come out in certain environments that I have chosen to ignore here. The idea is to interpret the name as a secondorder property (or quantifier) that applies to properties that Chemnitz had during the S-Phase: (3)

‘Karl-Marx-Stadt’ as a restricted name A predication of the form ‘Karl-Marx-Stadt P’ uttered in a context (w0, t0) is true of a situation (w, t) if[ and only i]f a. in (w, t), Chemnitz has the property the predicate P expresses in (w0, t0) and b. t is part of some S-Phase in w.

(4) iff

In a context (w0, t0), ‘Fritz lived in Karl-Marx-Stadt’ is true of a situation (w, t) at some time t# that is both before the time of utterance t0 and part of an S-Phase, the relation of dwelling holds between Fritz and Chemnitz in w.

One assumption behind (4) is that the main verb must be split into the lexical form ‘live-’, which expresses a relation between individuals and 4

In his (1979) and (1980) David Lewis gave independent reasons why this identification should not be made. In the present context, it is harmless. I will, however, switch to more fine-grained contexts in the formal account in Section 4.5.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

A predication is not itself a sentence, but rather corresponds to an infinitival like ‘Karl-Marx-Stadt be polluted’. The analysis rests on the assumption that predications, like sentences, describe—or are true of—situations. Situations may be thought of as points in Logical Space, i.e. maximally specific possibilities. Following standard terminology, the situations described by an expression will be referred to as indices. (More precisely, the term ‘index’ relates to the roˆle of being described by an expression.) For the present purposes one may take indices to be possible worlds w at times t and thus identify them with pairs (w, t), as in (3). Note that, according to condition (3b), what the S-Phases are depends on the index world w. This would have to be justified independently by looking at intensional constructions in which the index parameter is bound. I will ignore this complication here. (3) makes reference to the context of utterance. This is because in general the content of an expression may depend on the circumstances under which it is uttered, which is reflected by the general format of semantic analyses. For ease of exposition I will assume that the context of utterance is itself a situation (w0, t0).4 To see how (3) works, one may apply it to a simple sentence:

58 What’s in Two Names

5

ð$tÞ½½-ed

w0 ;t0 ðw; tÞðkt ½½Fritz live-in Karl-Marx-Stadt

w0 ;t0 ðw; tÞÞ

5 5

ð$tÞ ½t < t0 ^ ½½Fritz live-in Karl-Marx-Stadt

w0 ;t0 ðw; tÞ
...

According to (4), the sentence ‘Fritz lived in Karl-Marx-Stadt’ is true of a situation if that situation is set before the time of utterance, when Fritz used to live in Chemnitz while it was called ‘Karl-Marx-Stadt’. (4) says under which circumstances a certain sentence is true of a given situation. Under normal circumstances the situation described by an unembedded declarative sentence will be the context of utterance. Hence (4) also says under which (normal) circumstances an utterance of that sentence will be true tout court, viz. if it is true of the context of utterance (w0, t0), which in turn is the case if, at some time t# that is both before the time of utterance t0 and part of an S-Phase, the relation of dwelling holds between Fritz and Chemnitz in w0. This, I take it, is an adequate truth-conditional account of the sentence, confirming the analysis (3) of the name ‘Karl-MarxStadt’. Further evidence comes from considering the present tense version of the same sentence: (5) In a context (w0, t0), ‘Fritz lived in Karl-Marx-Stadt’ is true of a situation (w, t) iff at the time of utterance t0, which is part of an S-Phase, the relation of dwelling holds between Fritz and Chemnitz in w (5) is quite analogous to (4). The only further assumption is that the present tense identifies context time and index time: t ¼ t0. Again, the result of the analysis is summed up in (5), according to which the sentence can only be used truthfully during the S-Phase, i.e. at a time

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

places, and the past morpheme ‘-ed’, which is interpreted as relating index time (the time t talked about) and context time (the time t0 at which the utterance is made) in the obvious way: t is before t0; this treatment of tense as a relation between index and context times is somewhat inspired by Katz (2001) but largely simplified. An additional assumption is that the past tense morpheme must be moved to a sentence-initial position in order to be interpreted; this may be motivated by the fact that, in a sentence like ‘The king was away’, the subject may (but need not) refer to whoever happened to be king at the time talked about, and not (or not necessarily) to whoever happens to be king at the time of utterance. Hence (4) can be derived from a fuller analysis along the lines of (4*), the details of which are left to the reader: ð4*Þ ½½Fritz lived in Karl-Marx-Stadt

w0 ;t0 ðw; tÞ

Thomas Ede Zimmermann 59

while Chemnitz is called ‘Karl-Marx-Stadt’. In particular, since the place no longer bears that name, it would not be true if uttered now. Again, this prediction is no doubt correct.

2.2 Asymmetry Given the above analysis (3) of ‘Karl-Marx-Stadt’, one may expect ‘Chemnitz’ to denote the same place during the C-Phases: (6)

Example (7) confirms this analysis: (7)

Fritz lived in Karl-Marx-Stadt, not in Chemnitz.

At least one plausible construal of (7) comes out on the analyses (3) and (6), once the ellipsis is expanded in the obvious way. According to it, (7) merely says that Fritz lived in Chemnitz during (part of) the S-Phase. This reading is given in (8), which can be derived as indicated in (8*): (8)

Analysis of (7), based on (3) and (6) At some time before the time of utterance and during an S-Phase, which was not a time during a C-Phase, does the relation of dwelling hold between Fritz and Chemnitz in w.

ð8*Þ ð$tÞ½½½-ed

w0 ;t0 ðw; tÞðkt½½Fritz live-in Karl-Marx-Stadt

w0 ;t0 ðw; tÞÞ ½½not

5

w0 ;t0

ðw; tÞð½½ Fritz live-in Chemnitzr



w0 ;t0

^

ðw; tÞÞ


ð$tÞ½t < t0 ^ ð$ts Þ½S-Phaseðw; ts Þ ^ t 4 ts ^ Lw;t ðf ;cÞ ^ :½C-Phaseðw; ts Þ ^ t 4 ts ^ Lw;t ðf ;cÞ


5

ð$tÞð$ts Þ½t < t0 ^ S-Phaseðw; ts Þ ^ t 4 ts ^ Lw;t ðf ;cÞ ^ :C-Phaseðw; ts Þ


(7) may also have a stronger reading, according to which Fritz did not live in Chemnitz at any time before 1953 or after 1990, and I suspect that it, too, can be derived using the above interpretations of the rules but allowing for a variable scope of the existential closure operation associated with the tense morpheme. (Cf. Katz (2001) for more on this.) However, I

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

‘Chemnitz’ as a restricted name A predication of the form ‘Chemnitz P’ uttered in a context (w0, t0) is true of a situation (w, t) iff a. in (w, t), Chemnitz has the property the predicate P expresses in (w0, t0) and b. t is part of some C-Phase.

60 What’s in Two Names do not want to go too deeply into subtleties of tense and ellipsis and thus leave the matter open. In any case, (6) cannot be the whole story about the name ‘Chemnitz’, as the paraphrase (9) of (7) shows: (9) Fritz lived in Chemnitz while it was called ‘Karl-Marx-Stadt’.

(10) Analysis of (9), based on (3) and (6) There is a time before the time of utterance which is part of an S-Phase and part of a C-Phase and at which the relation of dwelling holds between Fritz and Chemnitz in w. ð10*Þ ½½Fritz lived in Chemnitzr while it was called ‘KMS’

w0 ;t0 ðw; tÞ
5 ð$tÞ½½while

w0 ;t0 ðw; tÞ kt½½-ed

w0 ;t0 ðw; tÞ ðkt½½it be-called ‘KMS’

w0 ;t0 ðw; tÞÞÞ
 kt½½-ed

w0 ;t0 ðw; tÞðkt½½Fritz live-in Chemnitzr

w0 ;t0 ðw; tÞ 5

ð$tÞð$t#Þ½t 4 t# ^ t# < t0 ^ ½½it be-called ‘KMS’

w0 ;t0 ðw; t#Þ ^ ½kt½t < t0 ^ ½½Fritz live-in Chemnitzr

w0 ;t0 ðw; tÞ

ðtÞ


5

ð$tk Þð$tc Þð$tÞð$t#Þ½S-Phaseðw; tk Þ ^ C-Phaseðw; tc Þ ^ t# < t0 ^ t 4 t# ^ t# 4 tk ^ t 4 tc ^ Lw;t ðf ;cÞ


5

ð$tk Þð$tc Þð$tÞ½S-Phaseðw; tk Þ ^ C-Phaseðw; tc Þ ^ t < t0 ^ t 4 tk 4 t 4 tc ^ Lw;t ðf ;cÞ


Since the S-Phase does not overlap the C-Phases (in our world), no time satisfying the underlined condition exists, so that (9) would be bound to be false, no matter where Fritz used to live. Surely, this is an unwelcome result; but it is obvious what’s going on here. In (9), ‘Chemnitz’ is used as a name of the town as such, independently of what it is called. In fact, in the meta-language, I have been using the name in that sense all along (as the reader may have noticed). Hence

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In order to see what the above account of the name ‘Chemnitz’ does to (9), let us assume that the underlined temporal modifier restricts the time talked about, the index time, to parts of S-Phases—after all, this is how the term S-Phase was introduced above—and that, moreover, the temporal conjunction ‘while’ expresses that the main clause is true of a situation that is a temporal part of a situation of which the subordinate clause is true. Then, according to (6), (9) is analysed as in (10)—or, more explicitly, (10*):

Thomas Ede Zimmermann 61

apart from the reading given in (6), the name should also receive a more standard interpretation: (11)

‘Chemnitz’ as a neutral name A predication of the form ‘Chemnitz P’ uttered in a context (w0, t0) is true of a situation (w, t) iff (!) in (w, t) Chemnitz has the property the predicate P expresses in (w0, t0).

(12)

Fritz lived in Karl-Marx-Stadt while it was [still] called ‘Chemnitz’.

In fact, the name ‘Karl-Marx-Stadt’ invariably appears odd when relating to a time that is not entirely part of the S-Phase. In other words, the following pattern emerges, where the upper and lower judgement marks ‘#’ [‘odd’] and ‘U’ [‘ok’] relate to the ‘Chemnitz’ and ‘Karl-Marx-Stadt’ variants, respectively: ð14Þ

8U 9 1840 until 1860 > > # > > >U > > > > > 1940 until 1960 > > # >   < U 1940 until 1999 > = Chemnitz Fritz lived in from # : U Karl-Marx-Stadt > 1960 until 1980 > > > U > > U > > > > > # 1960 until 1999 > > : U 1994 until 1999 > ; #

It thus turns out that ‘Chemnitz’, having a neutral sense, is always fine (in positive contexts), whereas ‘Karl-Marx-Stadt’ is inherently

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(11) is essentially Montague’s (1970) typed-shifted account of proper names as rigid designators. I am making use of it because it minimally deviates from the analyses (3) and (6). According to (11), the name ‘Chemnitz’ may be combined with any predicate whatsoever, resulting in a predication attributing the property expressed by that predicate to the city of Chemnitz. Under this analysis, the name ‘Chemnitz’ does little more than designating the object Chemnitz. One may suspect that (11) is all there is to the name ‘Chemnitz’ in that it covers the more specific uses when the name is combined with predicates that are true of Chemnitz during the C-Phases. However, the interpretation given in (6) is still needed as a separate reading. Otherwise (7) would come out contradictory, as the reader may verify. Since (7) may well be true even, the name ‘Chemnitz’ must be ambiguous between a restricted reading (6) and a neutral reading (11). But the name ‘Karl-Marx-Stadt’does not show this kind of ambiguity, witness the oddness of (12) in the (presentday) conformists’ dialect (which is the one I am looking at):

62 What’s in Two Names

3 DOUBLE LIVES I: SEMANTIC ACCOUNTS Example (2) illustrates that truth-preserving substitution of codesignative names in extensional contexts may be blocked. Since the distinction between neutral and restricted names offers a semantic account of this phenomenon, it is tempting to try and adapt it to apparently parallel cases like (1) that do not involve changing place names.

3.1 Restricted names Let us first generalize the crucial ingredients of the above analysis. A neutral name is just a name. It rigidly designates one particular object; it can be used to refer to that object under all kinds of circumstances. And even though one may represent it by a quantifier, this type shift is only 5 The following variation on (14), due to an anonymous referee, appear to cast doubt on the generality of the observation:

(14#) Fritz continued to live in Karl-Marx-Stadt even after its name was changed to ‘Chemnitz’. However, a comparison with (14$) shows that the alleged neutral use of the name ‘Karl-Marx-Stadt’ is licensed by the presupposition triggered by the verb ‘continue’: (14$) Fritz continued to love his wife even after they had been divorced. For (14$) to be true of a time t, the person loved by Fritz obviously does not have to satisfy the description ‘his wife’ at t, but only at a time before t at which the presupposition holds that Fritz used to love her. Similarly, for (14#) to be true of a time t, the town lived in by Fritz does not have to satisfy the description (3) at t, but only at a time before t at which the presupposition holds that Fritz used to live there.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

restricted. This is the earlier-mentioned asymmetry between the two names. Given the above analyses, the easiest way to account for it is to treat the name ‘Chemnitz’ as ambiguous between a neutral and a restricted reading. Of course, this ambiguity cannot be the kind of accidental homonymy exemplified by words like ‘bank’ or ‘can’. For the case at hand is not an isolated one. Very often when a place changed its name in history, the older name(s) remain in currency—though not as neutral names for the place as such but only as restricted names of the place in olden times; and in these cases the ordinary name may also be used in a contrastive sense, restricted to the bearer in modern times. ‘Byzantium’, ‘Constantinople’, and ‘Istanbul’ all refer to the same city, but only the third one is its proper neutral name in present-day English. So the ambiguity must be systematic, to be explained in general semantic or pragmatic terms. When a place changes its official name, only one of its names remains or becomes the one that has a neutral reading. Changing place names, then, confirm rather than refute naive prejudice.5

Thomas Ede Zimmermann 63

made for convenience and does not change the designating function of the name. The general format of a neutral name is given in: (NN)

Semantic behaviour of neutral names (Montagovian version) If N is a neutral name, there is some (uniquely identifiable) individual x (the bearer of N) such that any predication N P uttered in a context (w0, t0), is true of a situation (w, t) iff (!) in (w, t), x has the property expressed by the predicate P in (w0, t0)

(RN)

N is a restricted name iff N is not a neutral name and there is some (uniquely identifiable) individual x (the bearer of N) and a condition u such that any predication of the form N P uttered in a context (w0, t0) is true of a situation (w, t) iff a. in (w, t), x has the property the predicate P expresses in (w0, t0) and b. u

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(NN) is a straightforward generalisation of the interpretation (11) of ‘Chemnitz’ as a neutral name; I only replaced the word ‘Chemnitz’ by an arbitrary neutral name N and its bearer, the city of Chemnitz, by the bearer of N. According to (NN), combinations of neutral names and predicates express that the bearer of the name has the property expressed by the predicate. As mentioned above, this is the standard semantics of proper names. In order to arrive at a general notion of a restricted name, we need to generalize (3) and (6). Both analyses are conjunctions of two conditions (a) and (b), where (a) is precisely as the condition (!) in the interpretation of a neutral name and (b) is some extra condition relating to official names. The most obvious generalization is to replace (b) by something about official names in general. Since the aim is to adapt the analysis of place names to double identities, this won’t do: for one thing, it is not always clear what the official name of a person is—just think of Superman; for another thing, even if one of the names of a person is clearly the official one, it need not be the current one, the one that is used as a neutral name to refer to that person—Mark Twain being a case in point: present-day speakers normally refer to that author by his nom de plume, whether or not they know that his official name was ‘Samuel Langhorne Clemens’. So if the distinction between restricted vs. neutral names is to carry over to cases of double identity, condition (b) must be allowed to bring in other aspects than official usages of names. Before discussing which aspects these may, let me give a very general version of the interpretation of restricted names, one that simply adds an arbitrary condition to the predication (a):

64 What’s in Two Names

3.2 The apparel oft proclaims the man If (1) and (2) are as parallel as Saul and intuition suggest, the nonsubstitutivity ought to have a common source. Hence, following the analysis of the preceding section, either ‘Superman’ or ‘Clark Kent’ or both would have to be restricted names. The tricky question is not so much which of them is, but rather what the relevant condition might be. A simple answer is suggested by the paraphrase (15) of (1), due to Forbes (1997: 111): (15)

Clark, so-attired, entered the phone booth and Superman, soattired, emerged.

Forbes not only came up with a paraphrase, he also gave a logical analysis, which I am repeating in (16) with slight notational changes: (16)

entered(Clarki,b) & (the a: L(a, soi)) [attired(heClark, a)] & then (emerged(Supermanj, b) & (the a: L(a, soj)) [attired(heSuperman, a)])

The co-indexing in (16) is meant to relate (occurrences of) expressions and not their referents. Forbes uses arrows leading from occurrences of ‘so’ to the relevant occurrence of the corresponding name. Following this analysis, the condition ‘so-attired’ in (15) relates an individual—Superman in both cases—and the object Labelled by the referent of ‘so’. According to (16), the first ‘so’ in (15) is construed as referring to the subject of the first conjunct, i.e. the name ‘Clark’; and this name can be understood as Labelling a certain way of dressing, an appearance. Since, on the same analysis, the second ‘so’ refers to a different name, viz. ‘Superman’, the second occurrence of ‘so-attired’ expresses a different condition, provided that that name Labels a different kind of attire. The

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

According to (RN), a restricted name is like a neutral name in that it refers to its bearer to which a predicate may attribute a property; but unlike a neutral name, it makes this predication subject to an additional condition u. The nature of that condition is left maximally open in that satisfaction of u may depend on (i) the context of utterance, (ii) the time and world talked about, (iii) the bearer of the name, and even (iv) the property expressed by the predicate. For the time being one may think of u as expressing a property the bearer must have at the index, thus only showing dependencies (ii) and (iii); for instance, in the case of ‘Karl-Marx-Stadt’, the relevant property could be being officially called ‘Karl-Marx-Stadt’. However, in Section 3.3 we will see reasons to consider a more liberal notion of condition u, one that does not correspond to a property of the bearer.

Thomas Ede Zimmermann 65

(17)

‘Clark [Kent]’ as a restricted name (attire version) A predication of the form ‘Clark P’ uttered in a context (w0, t0) is true of a situation (w, t) iff a. in (w, t) Superman has the property the predicate P expresses in (w0, t0) and b. Superman is wearing Kentian attire in (w, t).

3.3 Personae As Saul (1997b: 116) pointed out in her reply to Forbes, one problem with this particular proposal is that attires are not enough: (18) (19)

Lois slept with Superman before she slept with Clark Kent. Lois slept with Superman, so-attired, before she slept with Clark Kent, so-attired.

(18) cannot truthfully describe a situation in which Lois Lane, having saved herself for the hero for so long, finally starts an affair with Clark Kent, and . . . ‘[. . .] due to her long-standing fascination with the Man of Steel, asks Clark to put on a little cape, take off those spectacles, etc.’ Saul (1997b: 116) It thus seems that the conditions u to be associated with the restricted names ‘Superman’ and ‘Clark Kent’ must be somewhat less

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

idea is, of course, that ‘Clark’ does Label a different kind of attire, viz. a business suit etc., whereas ‘Superman’ Labels a certain kind of overall. Forbes’s analysis can be cast in the present framework of restricted names. At first glance, this may appear impossible, because (16) makes reference to an apparatus of demonstratives, which does not seem to play any role in the purely semantic account (RN). However, I think the difference is cosmetic. After all, the original sentence to be analysed, viz. Saul’s (1), does not contain any of the demonstratives on which the paraphrase (15) and its analysis (16) rely. So, given that we are after a systematic account of how the original sentence is understood, we would have to provide a source for it, either from within the sentence, or from the context of utterance, or both. Now, the only source for the crucial addition of ‘so-attired’ that I can think of are the two names themselves; for if they were replaced by descriptions or ordinary names, no such paraphrase seems to be called for, or even suitable. But if the two names are the sources of the critical material, we may as well enrich their interpretation with it, thus arriving at the semantic reformulation (17) of Forbes’s indexical account:

66 What’s in Two Names

(20)

‘Clark [Kent]’ as a restricted name (adverbial version) A predication of the form ‘Clark P’ uttered in a context (w0, t0) is true of a situation (w, t) iff a. in (w, t) Superman has the property Q that the predicate P expresses in (w0, t0) and b. Q is a C-property in (w, t).

It is important for (20) to work properly that being a C-property is itself a situation-dependent (second-order) property. In other words, what is a C-property in one situation may not be a C-property in another situation: though entering a call box is a C-property in a situation as described in (1), it is not necessarily one in a situation described in (20). Following this lead, it is easy to give a similar analysis of ‘Superman’. Moreover, like ‘Chemnitz’, ‘Superman’ may also be used as a neutral name. Not only have I been using it like that here, but this is how the name is generally used by those who are in the know about the identity of its bearer (and for the time being I will ignore the others). Examples (21) and (22) suggest this: (21) (22)

Superman frequently disguises as Clark Kent. Clark Kent frequently disguises as Superman.

Whereas (21) is a perfectly normal, if quite obvious, thing to say, (22) is somewhat odd. And this is so because in (21) ‘Superman’ may be understood as a name of the person as such, whereas the subject in (22) does not allow for such a reading. Unlike ‘Clark Kent’ but like

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

superficial than (17b), marking a difference in character rather than in outward appearance. However, as Graeme Forbes (1999) pointed out, it is not enough to trade clothes for characters and just have the names label personae instead of ways of dressing; for Superman may express his double personality simultaneously, by simultaneously sitting in a bar in his fancy costume and talking shop to Lois Lane over the phone. What is missing, then, is a connection between the individual predications in (1) and the personae (to use Forbes’s term) indicated by the names. In other words, the truth condition of the first conjunct must make sure that the property of entering the phone booth is attributed to Superman as Clark Kent; similarly, the second conjunct would have to express that he has the property of leaving the phone booth as Superman. So, although in a (typical) situation as described by (1), Superman has both properties, he has one of them as Clark Kent and the other one as Superman. Referring to properties that Superman has as Clark Kent as C-Properties, we may replace (17) by:

Thomas Ede Zimmermann 67

(23) (24)

Clark Kent is Superman. Clark Kent admires Superman.

(23) is unequivocally true. In terms of the distinction between Superman’s two kinds of properties this means that being Superman must be a C-property; but the predication made in (23) does not relate the bearer of the subject name to an eventuality. One may take this to be an indication that the subject in (23) should not be construed along the lines of (20). But then, other things being equal, a unified analysis of the name ‘Clark Kent’ is certainly to be preferred to one based on a case distinction (pace Forbes (1999) who still seems to prefer an analysis of (1) in terms of attires to a more unified account in terms of personae). Moreover, while (23) might be analysed as a metalinguistic predication, no such interpretation offers itself for (24), which may be used to describe the kind of character displayed by Superman performing his Kent act.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

‘Chemnitz’, then, ‘Superman’ appears to call for both a neutral interpretation as a rigid designator and a restricted interpretation along the lines of (20), trading C-properties for S-properties. What makes a property a C-property, or an S-property, in a given situation? To attack this question, let us look at some specific examples first. In (1), the answer is pretty straightforward: Superman has the property of entering the phone booth due to his performing a certain kind of action (entering the phone booth), and while doing so, he is wearing his Kentian attire. However, the phone call from the bar shows that matters are not always that simple: though he is wearing his Superman costume, he is talking to Lois Lane as Clark Kent. Hence, in that scene, the property of making a phone call to Lois Lane is a Cproperty, whereas sitting in the bar is an S-property. We may even imagine him making two phone calls at the same time, by uttering ‘I can’t talk now, but I will get back to you tomorrow’ into two portables one of which is connected to Lois, who is waiting for Clark’s call, while on the other phone line there is Lex Luthor expecting a call from his arch-enemy. In that case, the property of talking to Lois Lane is a Cproperty, while the property of talking to Lex is an S-property. Since the two properties are realized by two different speech acts performed by the same person with one articulatory act, the difference between them may be captured by a Davidsonian account of adverbial modification as conditions on events (in a broad sense), as indeed Graeme Forbes (1999) proposed. However, (20) also allows for distinctions that may not manifest themselves in eventualities, and I suspect that this freedom is sometimes needed. (23) are cases in point:

68 What’s in Two Names

3.4 More asymmetry In order to motivate and formulate the alternative pragmatic analysis of non-substitutivity, let me discuss one more example, one that brings out more clearly what I take to be the point.6 A few years ago, I was looking for a paperback edition of No Night is Too Long, a piece of crime fiction written by the British author Ruth Rendell. One thing you have to know about Ruth Rendell in order to understand the example is that she writes under two names, viz. ‘Ruth Rendell’ and ‘Barbara Vine’. It so happens that neither of these names is the author’s full official name (cf. Section 4.2 below), but the fact that they refer to one and the same person was never hidden from the reading public. However, as far as the latter are concerned, there is an asymmetry between the author’s two names. For a long time, starting in 1964, she only published books under the name of ‘Ruth Rendell’, and only from 1986 onward did she also use the name ‘Barbara Vine’ for some of her novels, the difference in name supposedly coinciding with a difference in the styles of writing. The book I had been looking 6 The example originated in Zimmermann (1996), where I defended a semantic solution along the lines of the previous subsection. I had not been aware of Saul’s work at the time.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

What, then, are the criteria by which a given property should be judged a C-property in a given situation? A simple-minded answer to this question employs a descriptive construal of names as abbreviating descriptions, or as bundling properties, of their bearers: C-properties are those that are part of the sense, or the descriptive content, of the name. I am not calling this answer ‘simple-minded’ because I think that names should not be associated with properties but because the properties associated with the name ‘Clark Kent’ simply cannot be the C-properties referred to in (20). As a case in point, take the property of entering the phone booth, which, as we have seen, is among the C-properties in certain situations. Yet, clearly, entering a phone booth is not among the properties implied by the sense of the name ‘Clark Kent’: otherwise speakers using the name in that sense would have to believe that its bearer is someone who is constantly entering phone booths. Neither is the property of entering the phone booth in a particular situation part of the sense of that name; for it is not part of any speaker’s understanding of that name that Superman is entering a phone booth on a particular occasion, however he may be dressed for it.

Thomas Ede Zimmermann 69

for back in 1995 was the latest novel written under the name ‘Barbara Vine’ at the time. Someone had recommended a bookstore which specialized in crime fiction to me, and entering the store I enquired about the book (which they did not have, but that is another story) and immediately found myself engaged in a conversation with the owner of the shop, a devoted connoisseur of British crime fiction. One of the things he told me was (25), adding that he liked all of these novels—except the most recent one, which was the one I had been looking for. (25)

I’ve read all of Barbara Vine’s books.

(26)

I’ve read all of Ruth Rendell’s books.

On a straightforward construal of (26), ‘Ruth Rendell’ is a neutral name, whereas ‘Barbara Vine’ was not used neutrally in (25). Moreover, although the bookseller could have used both names in a restricted sense as in (27), no neutral reading of ‘Barbara Vine’ seems to have been available to him: (27)

I’ve read all of Ruth Rendell’s books, though not all of Barbara Vine’s.

The difference between the two names in (27) is thus reminiscent of the asymmetries observed above. And while could be cast in a semantic account based on a distinction between Ruth Rendell’s R-Properties and her V-Properties,7 I would like to explore an alternative, pragmatic approach.

7 In that case the examples (25)–(27) must not be construed as predications. Otherwise a treatment of the two names along the above lines would assign to (25) an unattested reading according to which being an author all whose books the speaker has read, is one of Ruth Rendell’s V-Properties—and hence one of her properties; but then the speaker would have to read all of her books (written under whatever name) to make (25) true.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

It was obvious that he was not talking about all of Ruth Rendell’s books, even though we were both clearly aware of the fact that Barbara Vine and Ruth Rendell are the same person. Rather, the utterance was unmistakably about those books that the British queen of crime wrote under the name of ‘Barbara Vine’. Why is that so? I think that again the asymmetry between the two names comes into play here. For if, instead of (25), the bookseller had uttered (26), I may have concluded that the bookseller meant all of the books written by Ruth Rendell, no matter under which name they had been published:

70 What’s in Two Names 4 DOUBLE LIVES II: A PRAGMATIC APPROACH

4.1 Neutrality Unlike other pragmatic explanations of substitution puzzles,8 the approach to be presented does not depend on contrastive uses as in (27) but also works if what looks like a restricted name comes on its own, as in (25). However, strictly speaking there are no restricted names: (NP)

Neutrality principle All names are neutral.

4.2 A failure of uniqueness Let me give the naive prejudice about names a more official form: (U) Principle of Uniqueness No two names of the same language have the same bearer. 8 Cf. Barber (2000), where a principle similar to (M) in fn. 2 is invoked to arrive at roughly the interpretation given in section 4.3. The problem with (M) is that it only applies to utterances of two coreferential names in the same context. 9 The concept of a name must be taken as pre-semantic lest the Neutrality Principle be circular. Semantically, a name is a (lexical) expression the meaning of which is neutral in the sense to be made more precise in the next subsection. Pre-semantically, a name is a linguistic expression that is conventionally associated with an object (its bearer), the relevant conventions being partly nonlinguistic. I do not have anything to say about the nature of these conventions, but I trust there is some way of distinguishing names from other linguistic expressions.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Neutrality is a semantic notion concerning the literal meaning of proper names: a neutral name is one that merely presents its bearer, not adding any descriptive information to the content of expressions in which it occurs. And since the semantic function of a neutral name does not go beyond specifying its bearer, any two names with the same bearer will also have the same literal meaning. Consequently, they may appear to be always substitutable salvo sensu; but they are not, as we will see. Like all linguistic expressions, names are language-dependent. What may be a name of your language, need not be one in mine.9 Consequently, since languages are (or at least determine) form-meaning mappings, there will hardly be any two speakers of the same language. On the other hand, names may be shared between languages and if they are, they also share their referents. These truisms, which will be given a (partial) formal reconstruction in Section 4.5, are crucial to the pragmatic account of non-substitutivity of coreferential names. For the time being it is important to keep in mind that a name is always a name in some language and that languages may differ in the names they contain.

Thomas Ede Zimmermann 71

(UP)

Uniqueness Presumption Speakers presume that (U) holds and that speakers presume that it holds.

In order to avoid the impression that (U) can be justified by recourse to rationality principles underlying cooperative communication, I have refrained from formulating it as an imperative a` la Grice (1975). I would like to stay uncommitted on this question. Bringing in the notion of a name, both (U) and (UP) are language-dependent. According to (UP), a language can only contain two distinct, coreferential names as long as its speakers are not aware of this fact. Once they are, they will drop one of the names, thus becoming speakers of a different language, one that contains fewer names. From then on, any utterances of coreferential names must either be instances of code-switching or else deviations from ordinary usage to be explained by pragmatic strategies of reinterpretation. The latter will be addressed in Section 4.3. To see how (U) and (UP) square with the examples discussed above, let us start with Chemnitz. Some fifty years ago, the semantic situation was simple. If you wished to name the place, you had no choice: ‘Chemnitz’ was the only name in town. Clearly, in the language of O[lden] spoken at the time, ‘Chemnitz’ was no counter-example to (U). Then came the time when the Chemnitzers were split into two linguistic communities. The first of these continued to speak O whereas in the language N[ewspeak] spoken by the majority, Chemnitz was now called ‘Karl-Marx-Stadt’. Presumably, the two communities were not entirely stable. Apart from natural turnover, there may have been some fluctuation due to changes of political camps, convenience, etc. Moreover, most (if not all) Chemnitzers were bilingual and quite capable of following and conducting conversations in either language.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(U) is not intended as an empirical generalization. As such it would not only be false but also miss the point. It would be false because there are languages containing two (neutral) names of the same bearer—even though, as I will argue, it takes more to reject (U) than two coreferential names within one utterance. As an empirical generalization, (U) misses the point in that the naive prejudice only concerns speakers’ expectations about language whereas the empirical correctness of (U) partly depends on extra-linguistic facts: speakers may be convinced that their language satisfies (U) and they may still be wrong about these facts. In other words, there cannot be two distinct names that a speaker knows to corefer. So (U) should be read as an ideal that speakers avow to for whatever reason, although they may fail to live up to it. The following formulation captures this subjective construal of (U):

72 What’s in Two Names

(2) I never made it to Karl-Marx-Stadt, but I visited Chemnitz last year. It seems that something other than, or maybe something on top of, code-switching is going on here. Precisely what this is, will be elaborated in Section 4.4. In Superman’s case, time is less important than knowledge. As has been emphasized before,12 we must distinguish between those ignorant speakers of a language I who believe that there are two distinct persons by the respective names of ‘Superman’ and ‘Clark Kent’, and us enlightened 10 Speaking of two languages is an oversimplification: other coreferential names will cause other dialect splits, so that the total number of languages may well be equal to, or even exceed, the number of speakers. 11 Graeme Forbes (p.c.) pointed out to me that there are non-temporal ways of interpreting (2), which I suspect might be explained along these lines; but then as far as I can see, such construals are clearly distinct from the temporal readings discussed in Section 2 above. 12 Moore (1999); the linguistic turn is mine though. Note that, like anybody in the debate initiated by Saul (1997a), I am pretending that the Superman myth is fact rather than fiction. As to ‘Superman’ and ‘Clark Kent’ as names of fictional characters, I can only refer to Parsons (1980).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

All in all, it may be assumed that the two communities led a peaceful co-existence in Chemnitz—and elsewhere, where the same split could be observed. In fact, the split still exists, though after the reunification of Germany the number of N-speakers has dramatically shrunk. More importantly, however, both languages10 conform to (U), by containing either ‘Chemnitz’ or ‘Karl-Marx-Stadt’, but not both, as a (neutral) name for Chemnitz. Being bilingual, sometimes speakers of one language would make utterances in the other language. In particular, as long as N was the official dialect of the political authorities, an O-speaker may have preferred to adapt herself to it, given the right kind of circumstances; at the same time an N-speaker may have mocked his neighbour by making an utterance in O. So sometimes O-speakers speak N without becoming (permanent) N-speakers, and vice versa. The code-switching may occur instantaneous and it may trigger pragmatic side-effects. For instance, speakers may out themselves, or misidentify themselves, as members of a particular community by making an utterance in that community’s language; or they may indicate their sympathy, their dislike, or their profound knowledge of the other community by using their language in the right kind of situation. None of this is too exciting; nor is it characteristic of the Chemnitz situation. But I think it can help explain away some apparent counter-evidence to (U). Still, it is hard to see how (2)—repeated here for the readers’ convenience—could be construed as an instance of code switching from O to N or back:11

Thomas Ede Zimmermann 73

I tend to divide friends and relatives into the ‘Ruth people’ and the ‘Barbara people’. Both names are equally familiar to me, equally ‘my’ names. [. . .] And I don’t mind which I am called so long as people don’t try to change in, so to speak, midstream. There is for me something grotesque in a Barbara person trying to become a Ruth person, or vice-versa. [signed ‘RR’] These remarks about unmotivated code-switching are certainly in line with (U) (and may well have served as a source of inspiration). However, for the present purposes I will concentrate on the community of crime fiction readers, among whom I will also ignore those who do not know Ruth Rendell under either, or only one, of her names. The rest of her 13

As regards E and I, the same caveats apply as in the ‘Chemnitz’/‘Karl-Marx-Stadt’ case (cf. fn. 10). Moreover, it should be noted that I is the language spoken in the fictitious world(s) of the comic strip, whereas E is common in the actual world. 14 The quotation is taken from an announcement of the first Barbara Vine novel, A Dark Adapted Eye, (Viking 1986); the announcement, which has been published in various places, can also be found on the internet under www.gusworld.com.au/books/vine/why.htm.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

speakers of E who know that the two names refer to the same person. It is cases like I for which a subjective construal of (U) becomes important: after all, the language does contain two distinct names for the same person. But not only does this escape its users; in line with (UP), they even assume that ‘Superman’ and ‘Clark Kent’ are the names of two different persons. The situation with E is different. I take it those who know the ‘facts’ about the comic paper hero’s double life, call him ‘Superman’, which is his neutral name in E; and according to (U), Superman ought to have no other name in our language. Hence uses of ‘Clark Kent’ by speakers of E must be either ungrammatical or else instances of codeswitching. The latter uses are extremely rare, simply because normally people get acquainted with Superman by reading the comics, watching the movies, or listening to an Enlightened account, thereby becoming Enlightened themselves; so there are no Ignorami and, consequently, it seems that there is hardly any need for speaking their language. Still, as will become clear from the discussion in the next two subsections, codeswitching from E to I does occur precisely in the cases discussed by Saul—though, again, this is not the full story.13 The case of ‘Ruth Rendell’ and ‘Barbara Vine’ is slightly more involved. Neither of these names is a genuine nom de plume. The author’s maiden name is ‘Vine’, ‘Rendell’ is her married name, and she has two Christian names, ‘Barbara’ and ‘Ruth’, which are apparently enjoying equal currency among her family and personal acquaintances—but:14

74 What’s in Two Names

(25) (27)

I’ve read all of Barbara Vine’s books. I’ve read all of Ruth Rendell’s books, though not all of Barbara Vine’s.

15 Cf. Stalnaker (1976: 89): ‘Let us consider what happens when a person comes to know that Hesperus is identical to Phosphorus after first being in doubt about it. If the possible-world analysis of knowledge is right, then one ought to be able to understand this change in the person’s state of knowledge as the elimination of certain epistemically possible worlds.’ [Emphasis added]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

readers is divided into the Experts who know that both names refer to the same person, and the Laymen who do not. As in the Superman case, there is a clear asymmetry: Experts invariably use ‘Ruth Rendell’ to neutrally refer to the author, preserving ‘Barbara Vine’ for special occasions, the above interchange between me and the bookseller being a case in point; further examples can be gleaned from the secondary literature. Now for the Laymen. In line with a certain epistemological tradition, it would be natural to characterize them, or at least some of them, as being undecided as to whether the two names corefer: in some worlds compatible with a typical Layman’s perspective they would, in others they would not.15 I will not follow this tradition here, but instead claim that even someone who has read, say, A Fatal Inversion (by Barbara Vine) and The Lake of Darkness (by Ruth Rendell) and discovered certain similarities in style, plot, or characters without being aware that they have been written by the same person, will take it that they have not—simply because the covers give different author names. This claim is somewhat hard to corroborate: after all, doubts about identity may always be elicited, given strict enough standards of knowledge and belief; cf. Lewis (1996). But even in the presence of doubt, the distinctness assumption may prevail until the Layman becomes an Expert. According to (UP), this is what should happen. And in the interest of simplicity and transparency, I will work on the assumption that it does. For although it will turn out that the pragmatic reinterpretation strategy only depends on a Uniqueness Presumption made by Experts about Laymen, it is hard to see why they should make this presumption if it were not for its apparent universality. I thus take it that what distinguishes the Layman from the Expert is his being positive that ‘Ruth Rendell’ and ‘Barbara Vine’ refer to different persons. In particular, then, both the Layman and the Expert conform to (UP). According to (UP), coreferential names are accidents that only happen as long as speakers do not notice them. As soon as the Layman finds out about the true identity of the author he used to call ‘Barbara Vine’, he becomes an Expert and erases that name from his language. Hence in his new language E there is no room for sentences like (25) or (27) above and repeated below:

Thomas Ede Zimmermann 75

In other words, (25) and (27) are not interpretable as utterances in Expert talk. So they must be instances of code-switching, or else they must be construed somewhat differently. In the next subsection, I will argue that the latter is the case.

4.3 A counterfactual reading

16

This is not the only escape from this Fregean dilemma, of course. In an earlier version of this paper I opted for giving up (NN) instead, replacing it by a descriptive account of names along the lines of Lerner & Zimmermann (1991) and Haas-Spohn (1995: ch. 4). Though I am not aware of any principled obstacles to this strategy, I now think the more natural option is to follow Kaplan (1989a: 562) and Stalnaker (2001) and treat referential uncertainties about proper names as lack of linguistic knowledge.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Though what the bookseller told me did not make any literal sense, I was still able to figure out what he meant. I think this was mainly because I recognized (25) as an utterance in a Layman’s language; Experts do not use the name ‘Barbara Vine’ to neutrally refer to the author of A Fatal Inversion. So my first inclination might have been to interpret the bookseller’s utterance simply as an utterance in a foreign language, L. However, even in L the speaker would have said that he had read all of Ruth Rendell’s books—whichever name appeared on their covers: as a sentence of L, (25) is true in that situation iff the bookseller had read all of the books written by the bearer of the name ‘Barbara Vine’, i.e. Ruth Rendell. A mere shift of language would not have led me to the right interpretation. What did, I will now argue, was a shift in perspective: instead of taking the sentence uttered as a piece of Laytalk, it rather seems that I interpreted the utterance as if it had been made by a Layman, while at the same time realizing that it was not. Let me explain. It helps to first take a look at a Layman’s perspective on his own language. I am assuming that he conforms to (UP), but does not realize that L fails to satisfy (U). How can this be? After all, according to (38), any two neutral names of the same individual are synonymous. Consequently, knowing L, the Layman ought to be able to see that ‘Ruth Rendell’ and ‘Barbara Vine’ have the same meaning. More generally, if the speakers know their language and thus the meaning of each word, they can tell whether there are any pairs of distinct coreferential names. In other words, if a language fails to meet (U), it would seem that its speakers, knowing their language, ought to be aware of this. However, the Layman is not; and according to (UP), speakers in general are never aware of any failure of (U). One may conclude that speakers do not always know their language.16 The kind of linguistic ignorance that is at stake here is not at all uncommon. In fact, whether or not (UP) holds, ordinary speakers

76 What’s in Two Names

17 Hence in each of his subjective languages, ‘Ruth Rendell’ and ‘Barbara Vine’ refer to different persons. But who are these persons? Whoever they are, they cannot be the same; a fortiori, they cannot both be identical to Ruth Rendell. Any further specification of them depends on the metaphysics and epistemology of cross-world identification. Cf. Lewis (1981).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

arguably never fully grasp the meanings of (neutral) names as interpreted in (NN)—on an internalist analysis of grasping, that is; cf. Frege (1892: 27), or Lewis (1981). A fortiori practically never fully know the language they actually speak. And whether or not (NN) holds, speakers may be uncertain about the exact meaning of an expression like ‘identical twins’ although they do use it as part of their language. This tension between what an expression actually means and what a speaker takes it to mean points to an important distinction between two senses of what a speaker’s language is. His objective language is the one the speaker actually speaks. Its expressions receive their meanings by social convention, and the speaker may or may not know which meaning is associated with which expression. A speaker’s subjective language is a language the speaker takes himself to speak. Since, as we have just seen, speakers may be uncertain as to what precisely their language is, they will in general have more than one subjective language in the sense that there are several possibilities of matching form and meaning none of which the speaker can exclude to be in full accord with her (objective) language. For instance, if I am uncertain as to the exact meaning of ‘identical twins’, then in one of my subjective languages it may be paraphrasable by ‘twins emerging from the same egg’, in another one it would mean the same as ‘twins emerging from the same sperm’ (though I may exclude that it is synonymous with ‘twins of the same sex’). The concept of a subjective language as well as its formal reconstruction (to be given in Section 4.5) are closely related to Haas-Spohn’s (1995) theory of subjective meaning, over which in Kaplan’s (1989b: 574) terms, a metasemantic stance is taken. (As to the concept of an objective language, I am glancing over a lot of interesting, but largely irrelevant subtleties. For example, there is a limit to linguistic ignorance in that speakers need to know enough about the meanings of enough expressions to count as speakers in the first place; and there is also a good deal of vagueness as to what the socially determined meanings are.) As a matter of historic fact, in L ‘Ruth Rendell’ and ‘Barbara Vine’ both refer to Ruth Rendell. Hence the Layman’s objective language L obviously violates (U). On the other hand, his subjective languages do not; for the Layman is devoted to (UP) (as I continue to assume).17 An immediate consequence of this observation is that L cannot be among the Layman’s subjective languages. This loss of objectivity among the

Thomas Ede Zimmermann 77

(28)

I would have said S, had I been a Layman.

That (28) delivers the intended interpretation in the case at hand crucially depends on the precise construal of the counterfactual condition, the nature of which will be investigated in the next subsection. The idea is that, had the bookshop owner been a Layman, his beliefs about the coreference of the names ‘Barbara Vine’ and ‘Ruth Rendell’ would have been different, but everything else would have been as it actually is. In particular, he would have had read the very same books that he actually read. Of course, he would have been under

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

subjective alternatives indicates that the Layman has an erroneous belief about his (objective) language, viz. that it satisfy (U). Concomitant to this linguistic error are false beliefs about extra-linguistic matters, like the conviction that A Fatal Inversion and The Lake of Darkness (both of which may be identified by plot rather than title) have been written by two different women. This is why a Layman who wants to express that he has read all of the books published under the author name ‘Barbara Vine’, would utter (25). Given that he is a speaker of L, where ‘Barbara Vine’ refers to Ruth Rendell, objectively he would therefore express that he has read all of Ruth Rendell’s books—including The Lake of Darkness. However, subjectively he would be under the impression to have only expressed that he has read (at least) all of the novels written under the author name ‘Barbara Vine’. Moreover, and somewhat paradoxically, the discrepancy between his subjective understanding and the meaning objectively expressed need not thwart his communicative intents: if the hearer is another Layman, he would be subject to the very same (mis-) understanding; and if she is an Expert, she will recognize his utterance as a Layman’s talk and gather that he does not realize that the person called ‘Barbara Vine’ also wrote books under a different name. In both cases, then, the hearer will understand that the speaker did not mean to say anything about the latter. By the time the bookseller uttered (25), it had become common knowledge between us that he was an Expert, which even his Laymanlike linguistic behaviour could not change. This is why I did not take him to simply make an utterance in L, to be interpreted literally. Rather, I took him to communicate precisely what he would have communicated had he been a Layman, viz. that he has read all of the books Ruth Rendell wrote under the author name ‘Barbara Vine’. In other words, I took his utterance to be a counterfactual speech act. More generally and schematically, a sentence S of L would thus be amenable to the following counterfactual reinterpretation when uttered by an Expert:

78 What’s in Two Names

(29)

I would have said S had I believed that the bearer of the name Barbara Vine (in L) is distinct from the bearer of the name Ruth Rendell (in L).

Saul’s case of ‘Superman’ and ‘Clark Kent’ is quite parallel. As pointed out above, the pertinent construal of (1) arises if the sentence is used in communication between Enlightened speakers: by using the name ‘Clark Kent’, which is absent from his language E, the Enlightened speaker behaves like an Ignoramus, who would have communicated that a person whom he refers to as ‘Clark Kent’ entered a phone booth, which was (subsequently) left by a different person, viz. the one he refers to as ‘Superman’. Given the speaker’s counterfactual speech act, the Enlightened hearer may conclude that, were it not for the fact that the speaker was Enlightened too, the latter had reason to believe that two distinct persons were involved in the scenario reported by him. If the speaker’s report is based on an eye-witnessing, it is obvious that what would have led him (as an Ignoramus) to the conclusion that the enterer was called ‘Clark Kent’ and the leaver was called ‘Superman’ must have been that they looked like these two persons, which is most likely due to the different attire the true common bearer of the two

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

the impression that the ones written under the name ‘Barbara Vine’ were the only ones written by the bearer of that name. Hence, by uttering (25), he would have been under the impression of only saying that he read all of those books. Assuming that, whether Layman or not, he would have spoken truthfully, it follows that the bookseller, had he been a Layman, had read all of the books written under the name ‘Barbara Vine’. Since the books that he had read had he been a Layman, are the books he actually read, one may conclude that he has actually read all of those books. On the other hand, had he been a Layman and uttered (25), he would have been under the impression of not committing himself to having read all of Ruth Rendell’s books. Hence his being a Layman is compatible with him truthfully uttering (25) without having read all of Ruth Rendell’s books. Hence, since the books he would have read as a Layman are the books he has actually read, the counterfactual reinterpretation (28) of his utterance does not imply that he has read all of Ruth Rendell’s books. This, I take it, is an adequate interpretation of the bookseller’s utterance. Since, by hypothesis, L only differs from E by the presence of an additional name for Ruth Rendell and the consequent violation of the Uniqueness Principle, the counterfactual reinterpretation of an Expert’s Layman’s talk can be given the following rough paraphrase that will come in handy in a minute:

Thomas Ede Zimmermann 79

(30) (31)

I would have said S, had I been a speaker of O. I would have said S had I believed that we are currently in an SPhase.

In section 4.5, I will briefly address the question of what it takes to have the counterfactual speech act of a Newspeaker’s utterance come out in the sense of (31). Whether such formalization is adequate in the first place, I will largely leave open here, devoting the remainder of this paper to the non-temporal examples.

4.4 Mechanisms of reinterpretation In order to see how the non-literal construal of the bookseller’s utterance (25) can be arrived at in a systematic fashion, combining literal meaning and contextual background, I will first isolate three features that distinguish the counterfactual speech act from an ordinary assertion. As the formalization to be presented in section 4.5 will make clear, the reinterpretation process roughly described in (28) may then be seen as a combination of these three features. By containing a name that is not part of the language spoken among Experts, the utterance was taken, and intended to be taken, as an instance of code switching from Expertese, which had been the language of our conversation so far, to Laytalk. One of the prerequisites of any successful code switching is that the language switched to is intelligible to all participants in the conversation. In the case at hand this means

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

names wore on the two occasions. Since Superman’s different appearances would have formed the ground of the speaker’s erroneous judgment had he been an Ignoramus, Superman must actually have looked differently in the two parts of the scenario described; for the counterfactual situation only differs from the reported reality in the speaker’s information state, which can hardly affect the way Superman dresses. Hence, according to the speaker’s report, Superman must have changed his outward appearance between entering and leaving the phone booth. Again, I take it that this is an adequate interpretation of the utterance. Epistemic perspectives seem to have little to do with the difference between ‘Chemnitz’ and ‘Karl-Marx-Stadt’. Hence a treatment of (2) along the lines of (1) does not look all too promising. Even if one somehow managed to render (30) equivalent to (31) and both equivalent to the result of interpreting ‘Karl-Marx-Stadt’ as a restricted name, this construal does not seem to reflect a general and natural strategy of interpretation.

80 What’s in Two Names

(26) is a sentence in Expertese, and I would probably have taken it as such had the bookseller uttered it instead of (25). However, (26) is also a sentence of L and as such has the same literal meaning as it has in E: ‘Ruth Rendell’ is a (neutral) name for Ruth Rendell in both languages, which do not disagree in the interpretation of any of the other parts of (26) either. By the same token it is readily seen that (25) and (26) must have the same literal meaning in L. Still, I insist, I would have taken an utterance of (26) as a sentence of E, not of L, because—and as long as—there was no indication of code switching. Fair enough, one may object, but what would have been the difference, given the synonymy just observed? The difference would have been (I reply) that, confronted with a code switch from E to L, I would not have taken the utterance in its literal meaning. A brief look at (27) reveals this: the speaker would have left E with the second clause; consequently the first clause is also (preferably) understood as an utterance in L, and as such it is not interpreted literally but as part of a counterfactual speech act. Hence it seems to be the very use of the name ‘Barbara Vine’ that indicates a switch from E to L and triggers the non-literal interpretation. But how come that it is L, rather than some other language that the speaker successfully alludes to? After all, there are other languages containing ‘Barbara Vine’ as a name for Ruth Rendell (or maybe someone else). For instance, why couldn’t the speaker have uttered (25) as a sentence of a language that contains ‘Barbara Vine’ instead of ‘Ruth Rendell’ and otherwise coincides with E? I think this

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

that the bookseller and I were in a position to speak and understand utterances made by Laymen. L only differs from E by one name (and the differences that result from this lexical difference), and so for us Experts to understand Laytalk means to realize that ‘Barbara Vine’ (as used in L) is but another name for Ruth Rendell—and there can be no doubt among us that we do realize this. More importantly though, apart from intelligibility there is another general prerequisite for code switching, viz. that the language in which the utterance was made is recognized. (Strictly speaking, in view of the kind of linguistic ignorance discussed in the previous subsection, demanding full identification of the intended language by the hearer or the speaker would obviously be too much.) To see what is at stake here, let us briefly recapitulate the bookseller’s alternative utterances discussed earlier: (25) I’ve read all of Barbara Vine’s books. (26) I’ve read all of Ruth Rendell’s books. (27) I’ve read all of Ruth Rendell’s books, though not all of Barbara Vine’s.

Thomas Ede Zimmermann 81

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

is a matter of contextual background and expectation: we all know, or can easily imagine, people who have come across the two author names without ever realizing that they are coreferential. Unlike these Laymen, people who are using the name ‘Barbara Vine’ without using ‘Ruth Rendell’ are somehow harder to come across or even imagine—for whatever reason: the name ‘Barbara Vine’ simply has a Laymannish ring. If code switching were all that is going on in the bookseller’s utterance of (25), I would have interpreted it literally, thus ending up with the same result as if he had uttered (26). What made the interpretive difference was that I not only recognized the sentence as one of a foreign language, but also took the speaker to act as if he had been a speaker of that language. In other words, code switching indicated a context shift, albeit a hypothetical one. And just like the target language of the linguistic switch, so would the target context of the contextual shift have to be identifiable by his partner in conversation, i.e. me. Hence the hypothetical context hinted at by the speaker was not just any old context in which the language differed from ours. It was the context that was exactly like our actual situation (then, in New Brunswick, New Jersey, by the way)—except that L and not E was the language of conversation. For this to be the case, I take it, the speaker would have had to be a Layman, although he would still have been a bookseller, and he would have read and liked the same books as he actually did, although being a Layman, he would have been ill-informed somehow about the authorship of some of them. Maybe, in that hypothetical and counterfactual context, he had never come across, or chosen to ignore, or forgotten about, the author’s announcement quoted earlier in which she reveals her identity. This is hard to imagine, given the fact that the same ‘revelation’ can be found on the cover of practically any book written under the author name ‘Barbara Vine’, but then the context we are dealing with is a hypothetical and counterfactual one. And it is this context to which the bookseller draws my attention by code switching. Of course, the hypothetical context shift is by no means a regular side-effect of code switching, but lies in the very nature of the target code, which is the Layman’s language. Violating (U), the latter is defective, and the typical Expert knows this. Of course, not everybody who speaks E and is thus an Expert in our semi-technical sense does; however, as already announced in Section 4.2 above, I would like to concentrate on speakers who are aware of both of Ruth Rendell’s names, and that goes for the Laymen we are considering as well as the Experts. ‘Experts’ who have never heard of ‘Barbara Vine’ would hardly have reason to use a sentence like (25). But the bookseller must have had a reason for lapsing into Laymanese, and one that I would

82 What’s in Two Names

(32)

I’ve read all of the Barbara Vine books.

In (32) the italicised noun phrase is a name of the name ‘Barbara Vine’, a meta-name (if you wish), which I take to be part of both the Expert’s and the Layman’s language.18 Of course, this does not mean that the Layman takes (25) and (32) to be synonymous; one is about an author, the other one about her name. In fact, this kind of difference has played an important roˆle in ruling out a simplistic descriptive approach to names; cf. chapter 3.4 of Haas-Spohn (1995: 105ff.) for some discussion and history. But for all intents and purposes, (32) may go proxy for (25)—or so the Layman erroneously believes. According to him, uttering (32) would have had the same communicative effect as uttering (25): in either case he would have informed me that he has read all of the books that were written under the name ‘Barbara Vine’. And it is this subjective understanding of (25) that the bookseller is hinting at when he puts on his Layman act. Given the close resemblance of the hypothetical and the actual context, I could then easily figure out that what he would have tried to get across as a Layman was actually the case, viz. that (32) is true of him, the Expert. The rationale behind the counterfactual speech act is that code switching creates a hypothetical context shift—an actual context shift being out, because the bookseller does not become a Layman by simply speaking like one—and that the context shift invites a subjective interpretation, 18 I am indebted to Robert Stalnaker for suggesting to focus on the contrast between (25) and (32)—there is of course some freedom, or context-dependence, in the interpretation of the compound ‘‘‘Barbara Vine’’ book’. However, to facilitate the present discussion, I am assuming that it expresses the property of being a book published under the author name ‘Barbara Vine’. An analogous simplification has been made for the interpretation of the Saxon genitives in (25)–(26), where the possessor relation is supposed to express authorship. If any of this is felt as a problem, the reader is invited to replace the sentences by unequivocal paraphrases.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

have had to be able to reconstruct. Merely using a different name for Ruth Rendell does not appear to be a good reason, especially since this would not have changed the content of what the speaker would have expressed by uttering (26). However, instead of intending the utterance to be interpreted as is, he wanted it to be understood as if it had been made from a Layman’s perspective; in other words, instead of relating it to the actual utterance context, he wanted it to be an expression of his—hypothetical and counterfactual—subjective background. For even though he would have had expressed the same proposition by uttering (25) as he would have done by uttering (26), as a Layman he would not have been aware of this fact. Instead, he would have been under the impression of saying something close to:

Thomas Ede Zimmermann 83

 code switching;  hypothetically shifting the context; and  interpreting the utterance subjectively. About the first one, I do not have anything to add to what I have already said, except that determining the target of the switch, though certainly a matter of contextual background knowledge, cannot be part of the interpretive process. After all, ‘semantics cannot tell us what expression was uttered or what language it was uttered in’—as David Kaplan (1989a: 559; emphasis added) observed. (The quotation is taken from a passage that is mainly directed against treating ambiguity as context dependence.) As to the context shift, I hope to have already made it clear that it may be thought of as an implicit counterfactual

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

because the literal meaning of the sentence uttered can somehow be excluded. As to the latter proviso, I am not sure about the reasons. After all, code switching as such may occur for a variety of reasons, and so why didn’t the speaker merely use the Layman’s language in a literal way? I suppose that the fact that he may as well have used (26) then is reason enough for not supposing that merely an otherwise unmotivated code switch took place and thus look for a non-literal interpretation. The fact that the language he uses is defective—as the Experts know, L violates (U)—may then be of help in arriving at a subjective interpretation: the latter is in line with (UP). I admit that this reasoning is still a far cry from a systematic derivation, but I think it is on the right track. Understanding how the intended non-literal construal of the utterance comes about is one thing. Explaining why the speaker has chosen such a roundabout way of expressing himself is another one, which is the more pressing if he had simpler alternatives at hand. And he did. For he could have reached the communicative effect of his counterfactual speech act by simply uttering (32), which is easily recognizable as a piece of Expert talk, ready to be interpreted literally. So why did he not do that? Again I can only speculate. Maybe, by relying on the fact that speaker and hearer were both Experts, the counterfactual speech act established a certain sense of belonging, as a switch into a familiar dialect may do. Certainly more cases will have to be explored to settle this question. Generalizing from the example, a speaker performs a counterfactual speech act if he uses a sentence that is recognizably part of a language L not spoken in the utterance context and wants it to be understood as having the communicative effect he would have intended it to have had he spoken L. Once again, the three ingredients that lead from the sentence uttered to its intended interpretation are:

84 What’s in Two Names

4.5 Formal modelling The following formal development is rather sketchy and gappy. To a large extent, this is so because filling in the details would be a distracting and tedious task, but it is obvious that it can be done, and how it can be done. To an equally large extent though, gaps occur when it it is not so obvious that they can be filled, or how precisely they should be filled. I still hope that the network of formal definitions, assumptions, and observations presented on the following pages, offers a glimpse of what a fully fledged pragmatic account of the extensional substitution puzzles would amount to. Somewhat simplifyingly, languages will be identified with sets of (disambiguated declarative) sentences denoting Kaplanian characters. More precisely, given the sets W, T, and D of possible worlds, times, and speakers respectively, let us pick a set C 4 W 3 T 3 D and a family (Lc)c 2 C of functions Lc : S Lc /ˆðW 3 TÞC ; where S Lc 6¼ B is the set of sentences of Lc. The elements c ¼ (wc, tc, dc) of C are the (possible utterance) contexts; Lc is the speaker’s objective language in the context c. If L is a language (i.e. the speaker’s objective language in some context20), u 2 S L and c 2 C, then ½½u

cL :¼ LðuÞðcÞ is the (horizontal) proposition expressed by u in c (according to L). A sentence u 2 S L of some language L is L-true in a context c 2 C if, and only if, 19 The usefulness of two-dimensionalism in the analysis of subjectivity has already been speculated on in Kaplan (1989a: 529ff.). The central roˆle the diagonal has been first emphasized in Stalnaker (1978). 20 Note that, according to this definition, not every mapping from the sentences SL of a given language L can be a language. Otherwise one could construct a diagonal language L* by putting L*(u)(c) :¼ (W 3 T)\Lc (u)(c) (for any c 2 C and u2 SL)—and end up in a contradiction: by definition of a language, L ¼ Lc*, for some c*, and so Lc* (u)(c) ¼ L*(u)(c*) ¼ (W 3 T)\Lc* (u)(c*)!

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

conditional along the lines of the paraphrase (28) above. One advantage of this understanding, and of isolating the context shift in the first place, is that it thus becomes amenable to a formal treatment along familiar lines. And as to subjective interpretation, it is well known that twodimensionalism, i.e. the distinction between context and index, may do some service here. More precisely, if doxastic uncertainty is modelled by sets of utterance contexts, separating out the contexts being in which the speaker does not exclude (on the basis of his beliefs), then the speaker’s subjective construal of a sentence may be identified with all those contexts at which it would have been true. This account of subjectivity, diagonalization, is a well-respected tool in formal pragmatics, and it will be put to work in the following subsection, where some of what I have said so far is made more precise.19

Thomas Ede Zimmermann 85

ðwc ; tc Þ 2 ½½u

cL ; the (semantic) diagonal of u (according to L) is the set DL(u) :¼ fc 2 C j u is L-true in cg. Although a full formal treatment of Neutrality would call for a more fine-grained individuation of languages, the present framework does allow for a general and abstract formulation. For each language L, let EL 4 S L be the set of equations between any two distinct names.21 Then the following formalization is a close as we get to (NP): (NPF) For any language L and u 2 EL it holds that:  either:  or else:

½½u

cL ¼ W 3 T for all c 2 C, ½½u

cL ¼ B for all c 2 C.

21

Unlike names (cf. fn. 9) equations may not be identifiable without interpretation. If so, both (NPF) and the formal version (UPF) of the Uniqueness Presumption below would have to be consequences of underlying semantic principles and cannot replace them.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

We will also make use of a relation of syntactic (and lexical) isomorphism ; between languages L and L#; the following crude characterization will be sufficient: L ; L# whenever S L ¼ S L# and EL ¼ EL# : Subjectivity is captured by a function P with domain C and values that are sets of contexts: Pc 4 C, where P(w,t,d) 6¼ B is (speaker) d’s doxastic perspective in w at t, i.e. the set of those contexts she may be in, for all (d thinks) d knows in w at t; I will refer to the elements of Pc as the (speaker’s) doxastic alternatives at c. The general approach to subjectivity in terms of alternatives is due to Hintikka (1962), of course; since any c 2 C provides a speaker, the perspectives form the basis of self-locating belief, i.e. belief de se in the sense of Lewis (1979). We will say that the speaker (in a context c 2 C) is reliable as long as c 2 Pc; although this notion inherits the notorious plague of coarseness from the possible worlds background, it will do for our purposes. The speaker’s subjective languages (in a context c 2 C) are those languages L for which there are contexts c’ such that c# 2 Pc and Lc# ¼ L; the set of subjective languages in c 2 C is Rc. For simplicity, I will assume full syntactic (and lexical) competence: L ; Lc whenever L 2 Rc. Moreover I assume that there is no objective language without subjective language: Rc ¼ B whenever c 2 C. In case Lc ; Rc, the speaker has some misconception about her objective language (and is not reliable); and if Rc contains more than one element, then—whether or not she has a misconception—the speaker is uncertain about its exact (semantic) nature. She may also be uncertain as to which proposition she would express by using a given sentence u of her objective language Lc; for there may be two contexts neither of which she excludes to be in and in which

86 What’s in Two Names

(33)

Pc 4 D(u).

A full formal treatment of the Uniqueness Presumption would again call for a more fine-grained individuation of languages; but, as in the case of (NPF), the present framework allows for a general and abstract formulation. According to (UP), all members of EL ought to be false in all subjective languages L, which is why we adopt: ðUPFÞ ½½u

cL ; ¼ B; for any c 2 C, L 2 Rc and u 2 EL. Whatever counterfactuality is needed will be supplied by a function + associating with each context c 2 C and language L a context c + L that is like c except for the fact that L is the speaker’s objective language in c + L. Hence two minimal requirements on + are S[electivity] and S[trong] C[entering], which we take to hold for all languages L and contexts c; we also assume that code-switching does not change denotations, i.e. that the proposition a sentence of a given language expresses and the truth value it has are preserved in a context in which a different language would be spoken:

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

u expresses distinct propositions—whether or not the (objective) languages at those contexts are the same: c#2 Pc, c$2Pc, but c$ ½½u

c# Lc# 6¼ ½½u

Lc$ (whether or not Lc# ¼ Lc$). In such cases of uncertainty—which, as regards to sentences u of the speaker’s objective languages may well be the rule rather than the exception—she may still be able to use such u to express whatever u may express for all she knows, or takes herself to know. By uttering a sentence u 2 S Lc in a context c the speaker intends (or would intend to express) a character v: C / ˆ(W 3 T) just in case vðc#Þ ¼ ½½u

c# Lc# ; for any c# 2 Pc . A sentence u 2 S Lc of the speaker’s objective language Lc in a context c 2 C is subjectively true in c if, and only if, u is Lc#-true in all contexts c# 2 Pc; following Stalnaker (2001) the metasemantic diagonal of u is defined as D(u) :¼ fc 2 C j u is subjectively true in cg. And the speaker is, or would be, sincere in (L-) asserting u in c if, and only if, by uttering u in c she intends a character v such that (wc#, tc#) 2 v(c#), for all contexts c# 2 Pc. Two remarks on the last two definitions are in order. First, given that the condition on the intended character v is normally satisfied by a lot of functions that only differ in the propositions they assign to contexts outside the speaker’s doxastic perspective, this is a rather coarse approximation to subjective meaning; but I think it will do for my present purposes. Secondly, it may be noted that the speaker’s sincerity in uttering yu in c could have been defined equivalently in terms of the metasemantic diagonal of u:

Thomas Ede Zimmermann 87

Assumptions about + (for any languages L and L#, u 2 SL and c 2 C): (Sel) (SC) ðIntÞ ðExtÞ (TI)

Lc+L ¼ L. c + Lc ¼ c. ½½u

cL ¼ ½½u

Lc+L# : ðwc ; tc Þ 2 ½½u

cL iff ðwc+L#; tc Þ 2 ½½u

cL : tc ¼ tc+L.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

One may think of + as determined by a more general selection function f: ˆ(C) 3 C / C along the lines of Stalnaker (1968): c + L ¼ f(fc# 2 C j L ¼ Lc#g,c); (Sel) and (SC) then follow from corresponding assumptions about f. (Int) generalizes Kripke’s (1972: 290) remarks about ‘describing a possible world or counterfactual situation in which people, including ourselves, did speak in a certain way different from the way we speak’: ‘we use English with our meanings and our references’ (ibid.). (Ext) obviously imposes some harmless restrictions on expressive power by excluding reference to foreign (objective) languages, as in the informal paraphrase (29) of counterfactual assertion. (TI) reflects the fact that counterfactual context shifts only affect the world and will be needed later. A similar intertia condition on the speaker would be equally natural but does not play a roˆle in the formal reconstruction. Further, more specific properties of + will emerge in the formal treatment of the above examples. We are now in a position to capture the counterfactual speech act performed by uttering a sentence that is not part of the speaker’s (objective or subjective) language. If c 2 C and L is a language such that u 2 S L \ S Lc then the speaker in c is (or would be) open in counterfactually (L-) asserting u if, and only if, the (same) speaker would have been sincere in uttering u in c + L. As we will see, openness is to counterfactual assertion what sincerity is to ordinary assertion; the two concepts are primarily distinguished for reasons of terminological transparency. The previous definition still depends on the language L to which a hypothetically uttered sentence is meant to belong. Given the above strict identity criteria for languages, a sentence that is not part of the language of the utterance context will be part of many other languages. Hence, in order to interpret a counterfactual speech act, one of these languages would have to be picked out first. I suspect that, where no natural choice can be made, the counterfactual interpretation is simply out. But even in the cases envisaged above, no single language in the sense defined here will suggest itself. It rather seems that a counterfactual speech act can be made in cases where a sufficiently well-defined class of alternative languages containing the sentences uttered is equally obvious

88 What’s in Two Names

(34) [¼

C(w,t#,d) 4 C(w,t,d) \ fc# 2 C j in c#, the speaker is sincere in Lasserting ug —by (33)] C(w,t,d) \ fc# 2 C j Pc#+L 4 D(u)g

If, on the other hand, L 6¼ Lc, then the utterance may be taken as a counterfactual assertion, in which case the communicative effect is that the speaker had been open rather than sincere: (35) [¼

C(w,t#,d) 4 C(w,t,d) \ fc# 2 C j in c#, the speaker is open in counterfactually L-asserting ug C(w,t,d) \ fc# 2 C j Pc#+L 4 D(u)g —by (33)]

22 An earlier version of this paper contained an alternative treatment that does justice to the indeterminacy of the language hinted at; it has been omitted for reasons of space. 23 (34) is in the spirit of Gazdar’s (1979: 46) reconstruction of Grice’s (1975: 45) Maxim of Quality. Given trust, (34) implies, but is not equivalent to:

(34#) C(w,t#,d) 4 C(w,t,d) \ D(u), which is a metasemantic variant of Stalnakerian Updating that in the present framework would read: (34$) C(w,t#,d) 4 C(w,t,d) \ DL(u). According to Stalnaker (1978), updates can be reached by accepting the horizontal proposition— provided it is identifiable given the common ground—or by accepting the semantic diagonal. In both cases (which are distinguished for independent reasons) the resulting context will satisfy (34$). Note that (34$) would only be satisfied in the unlikely event of full identifiablity of the objective language, i.e. if ¼ R(w,t,d) ¼ fLg.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

to the speaker and the hearer, in which case the resulting interpretation will be somewhat underdetermined. Though such uncertainty about the language hinted at by uttering a non-sentence can clearly be captured within the present framework, I am afraid this additional complication would create more confusion than insight. I therefore abstain from implementing it here, presuming that a single obvious alternative to the language of the utterance context is always available.22 We must finally say something about the communicative effect of utterances. To this end, let us take the common ground underlying communication in a context c to form a set Cc 4 C of contexts, for which we will assume syntactic competence—Lc# ; Lc whenever c# 2 Cc—as well as trust: the speaker is reliable in all c# 2 Cc. It is this common ground that is primarily affected by the speaker’s utterance of a sentence u 2 SL of some language L in the context c, in one of the following two ways. If L ¼ Lc, then the utterance may be taken literally, and at the time t# immediately following it we have the following communicative effect of an (ordinary) assertion:23

Thomas Ede Zimmermann 89

In order to see what this amounts to in the cases under scrutiny, we must recast them within the present formal framework. I will only do so for the case ‘Ruth Rendell’ v. ‘Barbara Vine’, leaving the other examples to the reader. Let us therefore see what the present formalization can do for the following sentences. I’ve read all of Barbara Vine’s books. I’ve read all of Ruth Rendell’s books. I’ve read all of the ‘Barbara Vine’ books. I’ve read all of the ‘Ruth Rendell’ books. Barbara Vine is Ruth Rendell. [The name] ‘Barbara Vine’ is [the name] ‘Ruth Rendell’.

The two objective languages involved are L and E. The following assumptions about them are fairly straightforward: Assumptions about L and E (and +): (a) SE 4 SL; EE 4 EL; (25) 2 SL\ SE; f(26),(32),(36)g 4 SE; (37) 2 EL\ SE; (38)2 EE. (b) L(25) ¼ L(26) ¼ E(26); L(32) ¼ E(32); L(36) ¼ E(36). ðcÞ ½½ð37Þ

cL ¼ W 3 T 6¼ ½½ð38Þ

cL ¼ ½½ð38Þ

cE ¼ B—for any c 2 C. ðdÞ ½½ð37Þ

cL ¼ ½½ð38Þ

cL ¼ B—whenever Lc ¼ L and L 2 Rc. ðeÞ ½½ð25Þ

cL 6¼ ½½ð26Þ

cL ; ½½ð32Þ

cL 6¼ ½½ð36Þ

cL —for any c 2 C such thatLc ¼ L and any L 2 Rc . (f) D(25) \ Pc ¼ D(32) \ Pc 6¼ D(26) \ Pc ¼ D(36) \ Pc—for any c 2 C such that Lc ¼ L. The syntactic assumptions in (a) are all straightforward: only the Layman’s language contains the name ‘Barbara Vine’; but both the Layman’s and the Expert’s language contain the (italicized) names of the alleged noms de plume ‘Barbara Vine’ and ‘Ruth Rendell’. (b) reflects the fact that L and E agree on the meanings of all expressions that do not contain the name ‘Barbara Vine’, which—although speakers are not aware of it—objectively refers to Ruth Rendell (in L) and is thus synonymous with ‘Ruth Rendell’ (in L and E). (c) expresses the semantic effect of this hidden synonymy on the proposition expressed by the equation (37) in L, as well as the contrary effect of the referential distinctness of the two meta-names. On a fuller semantic account, (b) and (c) could be derived compositionally along the lines indicated in the previous sections. Together with (a) and syntactic competence, (UPF) implies (d)—but does not imply (e), which would

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(25) (26) (32) (36) (37) (38)

90 What’s in Two Names not even follow had we bothered to include a level of subsentential compositionality. In fact, (d) does not even imply the considerably weaker inequalities: (e#)

L(25) 6¼ L(26); L(25) 6¼ L(26).

Assumptions about c* (and +): (g) Lc* ¼ E. (h) Pc* 4 fc j for some L 2 Rc*+L, c + L 2 Pc*+Lg; Pc*+L 4 fc + L j c 2 Pc* & L 2 Rc*+Lg. (i) Lc(32) ¼ Lc#(32); Lc(36) ¼ Lc#(36); Lc(38) ¼ Lc#(38) —whenever c 2 Pc* and c# 2 Pc*+L such that c# ¼ c + Lc#. About ( g) nothing needs to be said. (h), which has been formulated as a specific property of c* although it may well be generalizable to all c 2 C, describes the effect of the hypothetical language shift we are interested in on the subject’s perspective. One may have expected a more direct connection, viz: (h#)

Pc*+L ¼ fc + L j c 2 Pcg.

However due to (UPF), L cannot be a subjective language, and so this door is closed. More precisely, if we had c + L 2 Pc*+L (which we 24 Cf. Rooth (1985: 85f.). Due to general limitations of the possible worlds framework, this cannot be expected of all sentences though.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Still, (e) is certainly a highly natural assumption given (c): if a name is replaced by a referentially distinct one, the resulting sentence ought to express a different proposition.24 In fact, the second inequality in (e) should also hold for the Layman’s and the Expert’s respective objective languages L and E. The idea behind (f ) is that, the Layman takes ‘Barbara Vine’ to be coreferential with ‘the person called ‘‘Barbara Vine’’’ and is thus under the impression that (25) and (32), though not expressing the same proposition, have the same truth value; similarly for (26) and (36), which have been included for illustration only. In order to arrive at the desired construal of the bookseller’s utterance of (25), we need to implement some more specific observations about the particular context involved. Let us identify the scene described in Section 3.4 with an utterance context c*¼ (w*, t*, d*) 2 C, where w* is (or represents, for that matter) the actual world, t* is some particular time in May 1995, and d* is the bookseller I had been talking to. The following assumptions about c* turn out to be sufficient to make the points announced and sketched in the previous subsections:

Thomas Ede Zimmermann 91

(h$) Pc*+L ¼ fc + L j c 2 Pc* & L 2 Rc*+Lg. Finally, (i) says that, as far as d*’s subjective interpretation is concerned, certain sentences of E are not affected by the counterfactual language shift: a minimal deviation from a given context in the speaker’s perspective should not require more changes than those due to the shift from one (subjective) language to the other. Given that the languages overlap considerably, most sentences would come out as (subjectively) meaning the same in the two (subjective) contexts, although their meanings do not have to be stable across the subject’s doxastic alternatives.25 The reader is now invited to verify that the above assumptions (a)–(i) imply the following predictions: ( j)

d* is open in counterfactually L-asserting (25) in c* iff d* would be sincere in E-asserting (32). (k) If d* is reliable and would be open in counterfactually L-asserting (25) in c*, then (32) is E-true in c*. (k) is immediate, given (j): d*’s openness and reliability imply that c* c* 2 Pc* 4½½ð32Þ

c* Lc* ; i.e.: c* 2 ½½ð32Þ

E , by (g).

25

Do all sentences of E (and hence in SL \ SE) emerge unscathed from the (subjective) language shift? This seems questionable in view of cases like (26). I am afraid I will have to leave the matter open here.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

have been assuming to be non-empty), then Lc+L ¼ L (by (Sel)) and thus: L 2 Rc*+L (¼ fLc# j c# 2 Pc*+Lg); hence, given that (37) 2 SL (by assumption (a)), ½½ð37Þ

c# L ¼ B (for any c# 2 C), according to (UPF)—but contradicting (c). Instead, the idea behind (h) is that the contexts compatible with d*’s counterfactual belief are as close to the contexts compatible with his actual belief (at the time) as is consistent with the fact that he would be a speaker of what he would take to be the Layman’s language. Since different doxastic alternatives at c* may resolve d*’s semantic uncertainty in different ways, there need not be a unique such language he takes to be E, i.e. Rc* need not be a singleton. In that case each doxastic alternative c 2 Pc* will give rise to a way the world may be like according to him as a Layman, viz. like c except that the language he speaks (and takes to be L) would be one that minimally deviates from Lc by resolving all semantic uncertainties in the same way as Lc does. But then not every combination c + L of a doxastic alternative c 2 Pc* and a subjective language L 2 Rc*+L will qualify as a doxastic alternative at c + L, which is why we should not endorse the following strengthening of (h):

92 What’s in Two Names In order to also capture the communicative effect of the bookseller’s utterance we have to take an intersubjective turn, applying the reasoning that led to (j) and (k) from within the common ground, provided the latter covers whatever we have been assuming about the context c*. Let us therefore make the following Assumptions about Cc* (and +):

. . . and immediately obtain the following predictions (for any c0 2 Cc*): (jC) The speaker in c0 would be open in counterfactually L-asserting (25) in c0 iff the speaker in c0 would be sincere in E-asserting (32). (kC) Suppose the speaker in c0 would be open in counterfactually Lasserting (25) in c0; then (32) is E-true in c0. . . . as respective corollaries to (j) and (k). Note that, unlike (k), (kC) can do without any appeal to reliability because we have been assuming trust as a property of the common ground. Alternatively, we could have replaced the doxastic by epistemic perspectives all along, subsuming reliability under the properties of the accessibility relation P: c 2 Pc, for all c 2 C. I am not sure which option is to be preferred in the long run. Of course, our assumptions about the common ground Cc* are crucial here. For instance, in a context in which the speaker is not known to be an Expert and for which (gC) thus fails, (jC) and (kC) should not hold. Neither should the following conclusion, which I take to be the main point of the present subsection: (39)

Cc* \ fc 2 C j Pc+L 4Dð25Þg ¼ Cc* \ DE ð32Þ:

In other words, asserting (32) would have had the same communicative effect in c* as counterfactually asserting (25) actually had. I take it that Saul’s Superman example can be treated along the same lines, trading Experts for Enlightened speakers and Laymen for Ignorami. But, as already discussed in the Section 4.3, historical place names call for a more radical revision (if their use is a matter of counterfactual speech acts at all). For although it may seem plausible to reconstruct the purely temporal condition (3b#) given in Section 4.3 in terms of a purely temporal context shift, the assumption that wc+O ¼ wc

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

For any c0 2 Cc* the following holds: (gC) Lc0 ¼ E. (hC) Pc04fc j for some L 2 Rc0 + L, c + L 2 Pc0+Lg; Pc0+L4 fc + L j c 2 Pc0 & L 2 Rc0 + Lg. (iC) Lc(32) ¼ Lc#(32); Lc(36) ¼ Lc#(36); Lc(38) ¼ Lc#(38) —whenever c 2 Pc0 and c# 2 Pc0 +L such that c# ¼ c + Lc# :

Thomas Ede Zimmermann 93

26 In particular, if the common ground is identified with mutual belief in the sense of Schiffer (1972), then Pc*4Cc*, which implies that Rc*4 fLc j c 2 Cc*g. And even if, as Stalnaker (2002: 704f.) pointed out, this identification in general is rash, it is still arguable that it holds for a context like c*.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

whenever Lc ¼ N would play havoc with subjectivity: the N-speaker’s actual doxastic perspective will have changed since the days of Olden (if she is old enough in the first place), but even when uttering a sentence of O she will perform her counterfactual speech act on the basis of her present perspective. To account for this constancy of the perspective, one would have to assume that Rc+O ¼ Rc (at least for the kind of contexts c considered above) and see to it that the temporal parameter coincides with the S-Phase (of wc+O and wc, which I suppose would be the same). I am not at all sure how all this tampering with the revision function + could be motivated independently and whether it would ultimately lead to the same—apparently correct—predictions of the semantic account given in Section 2. There are some general issues that the formalization leaves open. In particular, I have been rather vague on questions of identifying languages and expressions. As a case in point, it is not obvious, which facts determine what the objective language spoken in a given context is. Is it, e.g. possible to have two contexts c and c# that only differ in their objective language? As far as I can see, it is not easy to give this question a precise formulation, let alone answer it in a principled way; but the whole formal framework is incomplete as long as this and related questions are left open. Similarly, it is not clear under which circumstances two (disambiguated) expressions of different languages should count as being identical. In principle, one should be prepared to identify such diverse forms as French ‘Louis-quatorze’ and German ‘Ludwig der Vierzehnte’. Let me finally point out an inadequacy in the above account of the common ground Cc*. Condition (gC), which had been adapted from the corresponding assumption (g) about c*, smuggles the objective language E into the (inter-) subjective beliefs. More specifically, (gC) implies that speaker and hearer agree on what the objective language is, which—given minimal assumptions about the connection between perspectives and common grounds26—implies full identifiability of the objective language: Rc* ¼ fEg. This consequence of (iC) certainly goes against the spirit of the enterprise, according to which speakers in general only have incomplete knowledge of their language. Hence a more adequate account would have to replace (iC) by a number of conditions on the language spoken that suffice to derive (39) without necessarily fully identifying it as E. Of course, the pertinent conditions

94 What’s in Two Names are essentially those laid out in (a)–( f), which would have to be reformulated in turn, having only been given for fully specified languages. Though such a reformulation can be given, it turns out to be rather tedious, which is why I have opted for the more transparent formulation above.

5 CONCLUDING REMARKS

(23) (37) (40)

Clark Kent is Superman. Barbara Vine is Ruth Rendell. Hesperus is Phosphorus.

In Section 3.2 I argued that the semantic account cannot explain why whoever may utter (23) on whatever occasion would have said something true. But it seems that neither can the pragmatic account.27 To be sure, if the sentence is uttered by an Ignoramus and taken literally, it always expresses a (necessarily) true proposition. However, if it is uttered by an Enlightened speaker, it would have to be taken as a counterfactual speech act and as expressing the subjective perspective of Ignoramus, who takes it to be necessarily false; the communicative effect would thus be disastrous. And, of course, the same goes for the other two of the above examples. I think there is a natural way out of this embarrassment. Any truly informative use of an identity statement like (23) would Enlighten an Ignoramus. In such a conversation 27

I am grateful to Robert van Rooy for bringing up Frege’s (1892) puzzle in this connection.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

The upshot of this comparison of two radically different approaches to extensional substitution puzzles is that neither of them is perfect. However, whereas the semantic account fares well with the temporal examples and leads to conceptual difficulties when it comes to double identities, the pragmatic account offers a (to my mind) plausible strategy of interpreting the latter but would have to be unduly stretched if it were to cover the former in an adequate way. One may speculate whether this shows that the initially obvious parallel between changing places and double personalities is not as close as it would seem. However, I still believe it is because I do not see a principled reason why changed names should invariably be encoded in semantics, whereas the use of hidden names are a matter of pragmatics. Let me end this somewhat inconclusive comparison with a classical puzzle which surprisingly does not find a simple and natural solution in either of the two accounts:

Thomas Ede Zimmermann 95

between unequal partners, the common ground would have to be undecided as to the coreferentiality of the two names (in I). And although the Enlightened speaker uses a sentence of I, the Ignoramus would not think this odd and would therefore not have any reason to reinterpret it. Rather, he would take it at face-value, i.e. in its literal meaning. Given the above treatment of ordinary assertions, he would thus learn that, according to the reliable speaker, the two names are coreferential, thus calling for a revision of his own beliefs—but not of the pragmatic approach to non-substitutivity.

The above text is a considerably revised version of the Sixth Weatherhead Lecture in Philosophy of Language, which I had the honour to give at Tulane University in October 2002. Part of the material has also been presented at a workshop on Robert Stalnaker’s philosophy, at the Sinn und Bedeutung 7 (both in Konstanz), to the Berlin Semantikzirkel, and in talks at the Tu¨bingen linguistics department and at Reid Hall, Paris. I owe a great debt to comments and criticism that I received from various members of my audiences. In particular, I would like to thank Manfred Bierwisch, Graeme Forbes, Hans-Martin Ga¨rtner, Ulrike Haas-Spohn, Aaron Konopasky, Ce´cile Meier, Friederike Moltmann, Robert van Rooy, Robert Stalnaker, Peter Staudacher, Ju¨rgen Konradi, as well as two anonymous reviewers for the Journal of Semantics. THOMAS EDE ZIMMERMANN Institut fu¨r Kognitive Linguisitik Johann Wolfgang Goethe Universita¨t Gru¨neburgplatz 1 D-60629 Frankfurt Germany email: [email protected]

Received: 05.08.03 Final version received: 01.09.04

REFERENCES Barber, A. (2000) ‘A pragmatic treatment of simple sentences’. Analysis 60:300–308. Dowty, D., Wall, R. & Peters, S. (1981) Introduction to Montague Semantics. Reidel. Dordrecht. Forbes, G. (1997) ‘How much substitutivity?’. Analysis 57:109–113. Forbes, G. (1999) ‘Enlightened semantics for simple sentences’. Analysis 59: 86–90.

¨ ber Sinn und Frege, G. (1892) ‘U Bedeutung’. Zeitschrift fu¨r Philosophie und philosophische Kritik (NF) 100: 25–50. [English translation: (1960) ‘On sense and reference’. In P. T. Geach & M. Black (eds). Transations from the Philosophical Writings of Gottlob Frege. Oxford, 56–78.] Gazdar, G. (1979) Pragmatics. Academic Press. New York.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Acknowledgements

96 What’s in Two Names Montague, R. (1970) ‘Universal grammar’. Theoria 36:373–398. Moore, J. G. (1999) ‘Saving substitutivity in simple sentences’. Analysis 57: 91–105. Parsons, T. (1980) Nonexistent Objects. Yale University Press. New Haven. Rooth, M. (1985) Association with Focus. Ph.D. Dissertation University of Massachusetts at Amherst. Schiffer, S. (1972) Meaning. Clarendon Press. Oxford. Saul, J. M. (1997a) ‘Substitution and simple sentences’. Analysis 57:102– 108. Saul, J. M. (1997b) ‘Reply to Forbes’. Analysis 57: 114–118. Stalnaker, R. C. (1968) ‘A Theory of Conditionals’. In N. Rescher (ed.), Studies in Logical Theory. Blackwell. Oxford, 41–55. Stalnaker, R. C. (1976) ‘Propositions’. In A. Mackay & D. Merrill (eds), Issues in the Philosophy of Language. Yale University Press. New Haven, 79–91. Stalnaker, R. C. (1978) ‘Assertion’. In P. Cole (ed.), Syntax and Semantics 9: Pragmatics. Academic Press. New York, 315–332. Stalnaker, R. C. (2001) ‘On considering a possible world as actual’. Proceedings of the Aristotelian Society, Supplementary Volume 65:141–156. Stalnaker, R. C. (2002) ‘Common Ground’. Linguistics and Philosophy 25: 701–721. Yablo, S. (2002) ‘Coulda, woulda, shoulda’. In T. S. Gendler & J. Hawthorne (eds), Conceivabiltity and Possibility. Clarendon Press. Oxford, 441–492. Zimmermann, T. E. (1996) ‘Hat jedes Ding seinen Namen?’ Talk presented at a workshop on proper names. Max-Planck-Arbeitsgruppe fu¨r strukturelle Grammatik, Berlin, October.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Grice, H. P. (1975) ‘Logic and conversation’. In P. Cole & J. Morgan (eds.). Syntax and Semantics, Vol. 3. Academic Press. New York, 41–58. Haas-Spohn, U. (1995) Versteckte Indexikalita¨t und subjektive Bedeutung. Akademie Verlag. Berlin. [English translation available under http:// www2.sfs.nphil.uni-tuebingen.de/ Alumni/Dissertationen/ullidiss/] Katz, G. (2001) ‘(A)temporal complements’. In C. Fe´ry & W. Sternefeld (eds). Audiatur Vox Sapientiae. Akademie Verlag. Berlin, 240–258. Kaplan, D. (1989a) ‘Demonstratives’. In: J. Almog et al. (eds.), Themes from Kaplan. Oxford University Press. Oxford, 481–563. Kaplan, D. (1989b) ‘Afterthoughts’. In J. Almog et al. (eds), Themes from Kaplan, Oxford University Press. Oxford, 565–614. Kripke, S. (1972) ‘Naming and necessity’. In D. Davidson & G. Harman (eds), Semantics of Natural Language. Reidel. Dordrecht, 253–355. Lerner, J.-Y. & Zimmermann, T. E. (1984) Bedeutung und Inhalt von Eigennamen. SFB 99, Universita¨t Konstanz Papier Nr. 94. Lerner, J.-Y. & Zimmermann, T. E. (1991) ‘Eigennamen’. In A. v. Stechow & D. Wunderlich (eds). Semantik/Semantics. De Gruyter. Berlin/ New York, 349–370. Lewis, D. K. (1979) ‘Attitudes de dicto and de se’. Philosophical Review 88: 513–543. Lewis, D. K. (1980) ‘Index, context, and ¨ hman content’. In S. Kanger & S. O (eds). Philosophy and Grammar, Dordrecht, 79–100. Lewis, D. K. (1981) ‘What puzzling pierre does not believe’. Australasian Journal of Philosophy 59:283–289. Lewis, D. K. (1996) ‘Elusive knowledge’. Australasian Journal of Philosophy 74:549–567.

Journal of Semantics 22: 97–117 doi:10.1093/jos/ffh018

Monotonicity and Processing Load BART GEURTS AND FRANS VAN DER SLIK University of Nijmegen

Abstract




If the monotonicity profiles of two quantifying expressions are the same, they should be equally easy or hard to process, ceteris paribus.
Sentences containing both upward and downward entailing quantifiers are more difficult than sentences with upward entailing quantifiers only.
Downward-entailing quantifiers built from cardinals, like ‘at most three’, are more difficult than others.
Inferences from subsets to supersets are easier than inferences in the opposite direction. We present experimental evidence confirming these predictions.

1 INTRODUCTION This paper is about quantification, a topic that has been central to natural language semantics since the very inception of the field. Many semantics textbooks (and all the good ones) will have at least one chapter on quantification, and will recount in more or less detail that there is a rich and widely accepted framework for treating quantified expressions, i.e. the theory of generalized quantifiers. Somewhat surprisingly, this framework has almost completely failed to affect the branches of psychology that might have benefited from it, such as the psychologies of language, acquisition, and reasoning. Our aim is to demonstrate that this neglect is not justified. More concretely, it will be shown how the theory of generalized quantifiers can contribute in a novel and non-trivial way to our understanding of cognitive complexity. Some expressions are more difficult to process than others, and in certain cases the difference is due to semantic rather than structural factors. We are concerned with such semantic sources of complexity. Ó The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Starting out from the assumption that monotonicity plays a central role in interpretation and inference, we derive a number of predictions about the complexity of processing quantified sentences. A quantifier may be upward entailing (i.e. license inferences from subsets to supersets) or downward entailing (i.e. license inferences from supersets to subsets). Our main predictions are the following:

98 Monotonicity and Processing Load The paper is organized as follows. We start with a quick overview of generalized-quantifier theory, with emphasis on the concept of monotonicity (section 2), and discuss in some detail the linguistic and psychological evidence for the importance of this concept (section 3). After a brief excursus on the interpretation of number words (section 4), we present our predictions (section 5), and report on an experimental study we carried out to test them (section 6). 2 QUANTIFICATION AND MONOTONICITY

(1) a. b. c. d.

‘No A B’ is true iff A \ B ¼ Ø ‘Some A B’ is true iff A \ B 6¼ Ø ‘At least five A B’ is true iff jA \ Bj > 5 ‘Most A B’ is true iff jA \ Bj > jA  Bj

A quantifier is upward entailing (or monotone increasing) if it licenses inferences from subsets to supersets, and it is is downward entailing (or monotone decreasing) if it licenses inferences in the opposite direction. Less informally: ‘Det A’ is upward entailing iff, for any B and C such that B # C, ‘Det A B’ entails ‘Det A C’; and ‘Det A’ is downward entailing iff, for any B and C such that C # B, ‘Det A B’ entails ‘Det A C’. A quantifier is monotonic iff it is either upward or downward entailing; otherwise it is non-monotonic. We will always confine our attention to the second argument of a determiner, and ignore the first one, so no harm will be done if we say that a determiner Det is downward (upward) entailing iff ‘Det A’ is downward (upward) entailing, for any A; similarly for monotonicity and non-monotonicity. Table 1 illustrates the monotonicity properties of some determiners. The concept of monotonicity is generally applicable to any expression whose denotation is partially ordered. For example, the prepositional object in (2a) is upward entailing, as witness the fact that (2a) entails (2b): (2) a. Betty lives in Berlin. b. Betty lives in Germany.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In accordance with standard practice in semantics we assume that determiners like ‘all’, ‘some’, ‘most’, and so on, combine with setdenoting expressions to form compound expressions that denote properties of sets. The denotation of ‘all A’, for instance, is the family of sets of which A is a subset, and consequently ‘All A B’ is true iff B is a member of the denotation of ‘all A’; i.e. ‘All A B’ is true iff A # B. Similarly:

Bart Geurts and Frans van der Slik 99 Table 1 Monotonicity properties of some determiners

Upward entailing all most some at least n Downward entailing no fewer than n

No dots are red 0 No dots are scarlet Fewer than five dots are red 0 Fewer than five dots are scarlet Exactly five dots are scarlet L Exactly five dots are red Exactly five dots are red L Exactly five dots are scarlet

If we negate (2a), the direction of monotonicity is reversed: (3a) entails (3b), rather than the other way round: (3) a. Betty doesn’t live in Germany. b. Betty doesn’t live in Berlin. The direction of monotonicity can be reversed again by embedding in yet another negative context: (4) a. It’s not true that Betty doesn’t live in Berlin. b. It’s not true that Betty doesn’t live in Germany. These examples illustrate a point that is of some importance for the objective of this paper, namely, that the monotonicity properties of a position in a sentence may be affected by the interplay between several expressions. In (3) and (4) this was shown by embedding a sentence under negation. In the following examples, different combinations of quantifiers are used with similar effects (we use up and down arrows to indicate upward and downward entailment, respectively): (5) a. b. c. d.

At At At At

least three hunters shot [[more than five [[colleagues]]. least three hunters shot [[fewer than five [[colleagues]]. most three hunters shot Y[more than five [[colleagues]]. most three hunters shot Y[fewer than five [[colleagues]].

The syntactic structure of each of these sentences is same: ‘Det1 hunters shot Det2 colleagues’. Semantically, however, they are rather different. In the first two sentences the object position is upward entailing, because the subject position is occupied by an upward entailing quantifier. In (5a) the

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Non-monotonic exactly n

All dots are scarlet 0 All dots are red Most dots are scarlet 0 Most dots are red Some dots are scarlet 0 Some dots are red At least five dots are scarlet 0 At least five dots are red

100 Monotonicity and Processing Load

3 MONOTONICITY MATTERS It is one thing to show that linguistic expressions have mononoticity properties. It is quite another thing to show that such properties matter to languages and their speakers. (After all, every linguistic expression, and indeed every thing, has an indefinite number of properties that aren’t of interest to anyone.) So let us consider some of the linguistic and psychological evidence for the relevance of monotonicity. To begin with, monotonicity is crucially involved in the following general constraint on lexicalization: (6) Lexical NPs and NPs built from simple determiners are always monotonic, and usually upward entailing (Keenan and Westersta˚hl 1997). Lexical NPs are pronouns, demonstratives, proper names, and lexical quantifiers such as ‘everybody’, ‘someone’, and ‘nobody’. NPs built from simple determiners are ‘every aardvark’, ‘no bonsai’, ‘most cassowaries’, and so on. That all these expressions should be monotonic is quite remarkable, for it means that the overwhelming majority of quantifiers expressible in English cannot be expressed either lexically or by NPs built from simple determiners. The constraint entails, for example, that there couldn’t be a lexical NP meaning ‘everyone but myself ’ or ‘about five dolphins’. A possible exception to the constraint in (6) are NPs with bare cardinals, like ‘three elephants’. In many cases, a quantifier of the form

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

object NP is upward entailing, as well, and therefore the position occupied by the noun ‘colleagues’ is upward entailing, too. In (5b), by contrast, the object NP is downward entailing, and the noun position follows suit. In (5c,d) the subject quantifier makes the object position downward entailing. In combination with an upward entailing object, this makes the position of the noun downward entailing, (5c); when combined with a downward entailing quantifier, the same position is upward entailing, (5d). Intuitively speaking, the monotonicity profile of (5a) is more harmonic than that of the other sentences in (5), in the sense that (5b–d) contain conflicting clues as to what monotonicity property of the last noun position is. Our main empirical tenet is that harmony, in this sense of the word, is among the factors that determine semantic complexity. A sentence whose parts are upward entailing throughout, for example, is less complex than a sentence whose monotonicity profile is mixed. However, before we can work out this idea in more detail, there are some preliminary issues that need to be addressed.

Bart Geurts and Frans van der Slik 101

‘n A’ will be construed as ‘no more and no fewer than n A’. For example, the following sentence would normally be taken to mean that the exact number of elephants photographed by Fred was three: (7) Fred photographed three elephants.

(8) a. Wilma f*has/doesn’t haveg any luck. b. f*Someone/No oneg has any luck. On closer inspection, it turns out that negative polarity items do not necessarily require a negative environment, though there certainly are constraints on where they may occur, as witness: (9) a. b. (10) a. b.

If Wilma has any luck, she will pass the exam. *If Wilma passes the exam, she must have any luck. Everyone who has any luck will pass the exam. *Everyone who passes the exam must have any luck.

The generalization is that negative polarity items may only occur in downward entailing positions. In effect, a negative polarity item serves to signal that the environment in which it occurs is downward entailing.1 This is a linguistic fact, but it is a fact about language users, too, for it can hardly be an accident that speakers will produce sentences like (9a) and (10a), and refrain from producing sentences like (9b) or (10b). The most plausible explanation for this behaviour is that speakers know that ‘any’ may only occur in downward entailing environments, and that they routinely compute the monotonicity 1 Actually the facts about polarity are rather more complex than this brief discussion might suggest. See (van der Wouden 1997) for an overview of the main issues. One particular problem we finess here is that in fact conditional antecedents may not be downward entailing in a straightforward semantic sense (see Heim 1984).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Taken at face value, this observation implies that ‘three elephants’ is non-monotonic, and therefore an exception to the constraint in (6). However, there are rather good reasons for assuming that the lexical meaning of ‘three’ is ‘at least three’, and that the ‘exactly’ reading of (7) is due to a conversational implicature (Horn 1989). If we accept the standard arguments, quantifiers of the form ‘n A’ are upward entailing, and conform to the constraint in (6). We will return to the semantics and pragmatics of cardinals in the next section. All languages have negative polarity items, which are so-called because they can occur within the scope of a negative expression, and are banned from positive environments. English ‘any’ is a case in point:

102 Monotonicity and Processing Load

(11) a. Puppet: None of the Ninja Turtles got anything from Santa. Child: No, this one got something from Santa. b. Puppet: Only one of the reindeer found anything to eat. Child: No, every reindeer found something to eat. O’Leary and Crain found that negative polarity items were practically never used in corrections, which indicates rather strongly that young children are aware of the distributional constraints on ‘anything’. See Musolino et al. (2000) for further discussion of monotonicity in language acquisition. Our last illustration of the importance of monotonicity relates to the interpretation of, and inferences licensed by, so-called ‘donkey sentences’ (Kanazawa 1994, Geurts 2002): (12) Every farmer who owns a donkey beats it. There is a much-discussed problem with this kind of sentences. It is that native speakers’ intuitions about their truth conditions are insecure in certain respects. To explain how, consider the following situation. We have three farmers each of whom owns a single donkey and beats it, and we have a fourth farmer who owns two donkeys, and beats only one of them. Question: Is (12) true in this situation? It turns out that speakers’ intuitions diverge: according to some the sentence is false in the given situation, while others judge the sentence to be true or cannot make up their minds. In brief: speakers disagree about the exact truth conditions of sentences like (12). This being so it is somewhat paradoxical that speakers appear to agree about certain inferences involving (12). For example, there is a consensus that (12) entails (13a) and is entailed by (13b). (13) a. Every farmer who owns a male donkey beats it. b. Every farmer who owns a donkey beats it with a stick.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

properties of incoming and outgoing utterances. In short, polarity items provide evidence that monotonicity is relevant to the psychology of production and comprehension. There are several studies on language acquisition which suggest that already at a very tender age, speakers are attuned to monotonicity properties. For example, in an elicited production study, O’Leary and Crain (1994) showed that even before they reach the age of four children consistently avoided using ‘any’ in environments like (10b). In this study children were presented with incorrect descriptions of stories previously acted out by the experimenters. Target sentences, which were produced by a puppet, contained the negative polarity item ‘anything’ and in their corrections children were expected to avoid using this expression:

Bart Geurts and Frans van der Slik 103

4 CARDINALS Some quantifiers, we have seen, are upward entailing, while others are downward entailing; ‘at least n A’ is an example of the former, ‘at most n A’ of the latter. Some quantifiers are non-monotonic, i.e. neither upward nor downward entailing; a case in point is ‘exactly n’ (see Table 1 for examples). What is the monotonicity profile of bare ‘n’? The answer that comes to mind first, perhaps, is that ‘n’ is synonymous with ‘exactly n’, and therefore non-monotonic. But things are not as simple as that: (14) Fred had (*exactly) three oranges and possibly four. if not four. and for all I know four. If ‘three’ was synonymous with ‘exactly three’, the two should always be interchangeable, but it is plain that in (14) they are not. It would seem that bare ‘n’ is ambiguous between ‘at least n’ as well as ‘exactly n’. If so, the ambiguity becomes manifest in cases like the following: (15) If you score five points, you get a teddy bear. The teddy bear may be yours if you score five points or more, or perhaps you’ll win a Barbie doll (instead of a teddy bear) if you score more than five points. However, it would be unsatisfactory simply to assume that cardinals are ambiguous between an ‘at least’ and an ‘exactly’ reading, and leave

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Note that both of these inferences are based on the monotonicity properties of ‘every’: given a proposition of the form ‘Every A B’, inferences to subsets are permitted in the A-part, while inferences to supersets are permitted in the B-part. Apparently, monotonicity inferences may go through even if the interpretation of a sentence remains underdetermined or unclear in certain respects. Not only do these observations confirm the importance of monotonicity, they also suggest a reason why this should be the case. Monotonicity inferences as such are very simple; they are just a matter of replacing one term by another that is either less or more specific. Furthermore, these inferences are shallow in the sense that they can be ‘read off ’ a sentence simply by checking for the occurrence of logical words such as quantifiers, negative particles, and so on; they do not require a deep understanding of what a sentence means. This explains why speakers boldly draw monotonicity inferences from sentences whose truth conditions they are uncertain about.

104 Monotonicity and Processing Load

(16) a. She can have 2000 calories without putting on weight. b. The council houses are big enough for families with three kids. c. You may attend six courses. While it is true that the cardinal expressions in these examples can be paraphrased with ‘at most’, this does not show that there is an ‘at most’ reading just as there are ‘at least’ and ‘exactly’ readings. First, once it is

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

the matter at that. This worry has led neo-Griceans like Horn (1972, 1989) to argue that the basic meaning of ‘n’ is ‘at least n’, and that, depending on the context, pragmatic reasoning may strengthen this to an ‘exactly’ interpretation. On the neo-Gricean view, the ‘exactly’ interpretation of the number word ‘seventeen’, say, actually consists of two components: the literal meaning of the word (‘at least seventeen’) and a scalar implicature (‘not more than seventeen’). There are problems with this treatment of cardinals, which, incidentally, led Horn (1992) to throw in the towel after 20 years, leaving Levinson (2000) as the sole defender of what seems to be a lost cause. If it were true that cardinals are scalar expressions, the ‘at least’ interpretation of a cardinal should be more basic than the ‘exactly’ interpretation, and there are various kinds of evidence which suggest that this expectation is not borne out by the facts (Horn 1992; Carston 1998; Geurts 1998). Most importantly, for our present purposes, there are experimental data indicating that young children, who in general seem to find it easier to obtain ‘at least’ interpretations than adults, prefer ‘exactly’ readings for cardinals (Noveck 2001; Papafragou and Musolino 2003; Musolino 2004). In brief, the semantics (and pragmatics) of number words, simple though they may seem at first, turns out to be remarkably thorny, but in this paper we will not attempt an analysis of our own. We will assume that both the ‘exactly’ and the ‘at least’ readings of cardinals are easily obtained—and this much is uncontroversial. We will not make any assumptions about how the two readings are related. In particular, we will not assume that one is derived from the other. The only thing that matters to our purposes is that both are readily available. In saying that the ‘at least’ and ‘exactly’ readings of bare ‘n’ are readily avaible we intend to implicate that they are the only readings that are readily available—and this is a controversial point. It has been argued by Carston (1998), among others, that in addition to the readings discussed so far there is an ‘at most’ reading, which is on a par with the others. Some of Carston’s examples are:

Bart Geurts and Frans van der Slik 105

granted that cardinals may have ‘exactly’ readings, it can be argued that the intended readings of (16a–c) result from pragmatic inferences. For example, if (16c) is read as, ‘There are deontically possible worlds in which the addressee attends exactly six courses’, it is natural to infer that there are deontically possible worlds in which the addressee attends fewer than six courses, and no such worlds in which the addressee attends more than six courses. Secondly, even if ‘at most’ readings are not derived in this way, they only come to the fore in very special circumstances. It would be most unusual for a simple sentence like (17), for instance, to give rise to an ‘at most’ construal of ‘fifty’: (17) Betty has fifty guilders.

5 MONOTONICITY AND INFERENCE The notion that monotonicity might play an important role in reasoning (especially reasoning with quantifiers) is not new; it can be traced back in part to medieval times and in part even to Aristotle. More recent developments in semantic theory have helped to further develop the idea and incorporate it in the framework of generalizedquantifier theory (Sa´nchez Valencia 1991). The first sustained attempt at showing that monotonicity is relevant to the psychology of reasoning, as well, is due to Geurts (2003a), who proposes a theory of syllogistic reasoning that hinges on the concept of monotonicity.2 Geurts presents a processing model in the tradition of ‘mental logic’; that it to say, the core of his system is a collection of weighted inference rules that assigns a complexity index to every argument form within its scope. Methodologically, the strategy pursued in this paper is different: we will for the most part ignore the question how people process inferences, and argue simply that some arguments are bound to be more difficult than others because they involve expressions that, for semantic reasons, are more complex. Hence, the level at which we derive our predictions about human reasoning is different from that of Geurts (2003a). But the key idea is the same; it is that certain patterns of reasoning crucially involve monotonicity inferences. We will now discuss a number of predictions that are derivable from this premiss. 2

See Newstead (2003) for critical discussion of Geurts’s theory, and Geurts (2003b) for a reply.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

On the strength of these observations, we will assume that the ‘at most’ reading of a cardinal, if available at all, is exceptional.

106 Monotonicity and Processing Load

5.1 Feasible inferences Our first prediction is seemingly trivial: it is that certain reasoning tasks should be feasible. (18) More than half of the tenors have the flu. More than half of the tenors are sick.

(19) All the tenors have the flu. All the tenors are sick. Logically speaking, the difference between (18) and (19) is profound. For, whereas can be (18) expressed in first-order predicate logic, (18) requires more powerful means (as noted above), and in this sense is more complex. However, ‘more than half ’ and ‘all’ are both upward 3

As Larry Horn pointed out to us, the inference (18) in may become problematic if it is given that all tenors are sick. More generally, people will tend to avoid endorsing conclusions that are pragmatically infelicitous. See Geurts (2003b) for further discussion of this point.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Intuitively speaking, this inference is valid and obviously so. From a strictly logical point of view this is quite remarkable. On the one hand, it is a familiar fact that first-order logic cannot capture the meaning of the determiner ‘more than half ’ (Barwise and Cooper 1981); a more powerful logic is needed for that. On the other hand, to the extent that it is decidable at all, first-order logic is known to computationally intractable, and therefore a more powerful logic can only be worse. So how do we humans manage to see that arguments like (18) are valid? The answer is fairly obvious: by trading completeness for efficiency. The inference strategies we employ cannot fail to be radically incomplete, but at least they are reasonably efficient. One of the reasons why monotonicity inferences are important to us is that they are highly efficient. As we noted at the end of section 3, a system for producing monotonicity inferences can be very simple, because it requires only a shallow understanding of the representations it operates on. In the case of (18), for example, such a system would merely need to know that having the flu entails being sick and that ‘more than half ’ is upward entailing; the exact meaning of ‘more than half ’ is immaterial. That is why (18) is easy.3 Monotonocity inferences are simple because they don’t require a full-blown interpretation; a superficial understanding usually will do. By the same token, we predict that, in so far as human reasoning is based on monotonicity inferences, it will be sensitive only to the logical bare bones of an argument. To illustrate this point, compare (18) with the following argument:

Bart Geurts and Frans van der Slik 107

entailing, so with respect to monotonicity inferences (18) and (19) should be equally easy. As we will see, this prediction turns out to be correct.

5.2 Harmony While the concept of monotonicity may simplify our models of reasoning in certain ways, as we have just seen, it also helps to reveal sources of complexity that would remain hidden without it. To explain how, compare the following sentences:

While the surface forms of these sentences are nearly identical, their monotonicity profiles are different, and we believe that the difference matters. Whereas the two quantifiers in (20a) are both upward entailing, their counterparts in (20b) are upward and downward entailing, respectively. Let us say that the monotonicity profile of the former sentence is more harmonic than that of the latter. We predict that harmony, in this sense of the word, is a measure of semantic complexity: all other things being equal, (20a) will be easier to process than (20b), because its monotonicity profile is more harmonic. The rationale behind this prediction is as follows. Both (20a) and (20b) are of the form ‘Some of the sopranos sang with . . .’, where the position marked by the dots is upward entailing. In (20a) this position is occupied by a quantifier that is itself upward entailing, but in (20b) the second quantifier is downward entailing. As a consequence, the latter sentence gives conflicting clues as to what its monotonicity profile is—which should make it harder to process than (20a). (20b) is less harmonic than (20a), we have said, because its monotonicity profile is mixed: while the two quantifiers in (20a) are both upward entailing, the quantifiers in (20b) point in opposite directions, so to speak. But what if the quantifiers are both downward entailing, as in the following sentence? (21) None of the sopranos sang with fewer than three of the tenors. Even though both quantifiers occurring in (21) are downward entailing, we should expect this sentence to be less harmonic than (20a), because it involves a monotonicity reversal: the position occupied by the noun in ‘fewer than three of the tenors’ is downward entailing, but as the quantifier fills an argument slot that is itself

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(20) a. Some of the sopranos sang with more than three of the tenors. b. Some of the sopranos sang with fewer than three of the tenors.

108 Monotonicity and Processing Load downward entailing, the ‘tenors’ position becomes upward entailing, as the validity of following argument confirms: (22) None of the sopranos sang with fewer than three of the tenors. None of the sopranos sang with fewer than three of the male singers.

Hence, (22) should be more difficult than (23), which is upward entailing throughout: (23) Some of the sopranos sang with more than three of the tenors. Some of the sopranos sang with more than three of the male singers.

(24) a. The first number on the list is not odd. b. The first number on the list is even. As shown by Clark (1974), if someone has to evaluate (24a), he will often decide to mentally rephrase the sentence as (24b), and evaluate that instead.4 This ‘conversion method’, as Clark calls it, applies not only to double negatives, but also to sentences like (21), in which one downward entailing quantifier sits in the scope of another. Now if, in the process of assessing (22), the conversion method is applied, the task is made more manageable, obviously. This is not to say that (22) will be as easy as (23) is; the procedure for deciding (22) will be more complex than for (23), whether or not the conversion method is used. But it is to say that the conversion method will make (22) more manageable than it would be otherwise. Consequently, it may well turn out that (22) is easier than analogous arguments with a mixed monotonicity profile. We are not in a position to predict that it will be easier (or harder, for that matter), because that would require a estimate of the absolute complexity index of the procedures and representations involved in solving problems like (22) and (23). But at least we are able to say that certain patterns in the data can be made sense of within the general approach adopted here. To sum up the foregoing discussion: given a corpus of sentences of the form ‘DetA A . . . DetB B’, where the latter quantifier is in the scope of the former, and both may be either upward ([) or downward 4

Occasionally, this procedure is overextended and applied in cases in which it is not valid; see Geurts (2002) for an example. The psychological literature on negation is discussed at length and in depth by Horn (1989).

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

This prediction is intuitively correct (i.e. (23) does seem to be easier than (22)), and is corroborated by experimental data, as we will see, but it is possible that the complexity of (22) is mitigated by an additional factor. It is a well-established fact that in processing negation, many people replace double negatives with the appropriate positive expression.

Bart Geurts and Frans van der Slik 109

(Y) entailing, we expect to find one of the following patterns (where ‘X < Y’ means that X is easier than Y): [DetA YDetB [DetA [DetB < YDetA [DetB YDetA YDetB [DetA [DetB < YDetA YDetB <

[DetA YDetB YDetA [DetB

(25) a. b. c. d.

some kangaroos at least five kangaroos at most five kangaroos no kangaroos

Semantically, (25a) and (25b) are on a par in that they are upward entailing, while (25c) and (25d) are both downward entailing. But in another respect (25c) is the odd one out: it is a downward entailing phrase that contains an expression that is not downward entailing. None of the other phrases in (25) have this property. In particular, (25b) is no less harmonic than (25a) or (25d), assuming as we do that one of the readily available construals of ‘five’ is upward entailing. We predict, therefore, that (25c) will be harder to process than the other phrases in (25).5

5.3 Up is easier than down To the extent that it relies on monotonicity, reasoning is a matter of interchanging set-denoting representations, going from a set to one of its subsets, or vice versa. 5

Geurts (2003a) derived a similar prediction by assuming that ‘at least’ is inherently negative. There are several problems with this; most importantly, it is ad hoc and fails to predict a difference between ‘at most’ and ‘no’.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

That is, the [DetA[DetB sentences should be easier than all others; the mixed cases should be equally hard; and the YDetAYDetB sentences may be of intermediate complexity. We argued in section 4 that bare number words are either upward entailing or non-monotonic. If this is so, considerations of harmony yield an interesting prediction: expressions like ‘at most five’ or ‘fewer than five’ should be harder to process than others. Hence, (25c) should be more difficult than the other phrases in (25):

110 Monotonicity and Processing Load (26) a. Some A are B All B are C Some A are C

b. Some A are B All C are B All C are B

6

The locus classicus is Clark (1973). See Horn (1989) for a more recent overview.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In both of these arguments, the conclusion is obtained by replacing the B-term in the major premiss with the C-term. But (26a) moves upward in the sense that B # C, whereas in (26b) the inference goes in the opposite direction, because C # B. (Note that we are using the up/ down terminology in two related but different senses. If we say that expression X is upward or downward entailing, we are describing the inferences licensed by X. If, on the other hand, we say that an inference moves upward or downward, we are describing an actual inference, be it valid or not.) There are reasons to expect that downward inferences may be more demanding than upward ones. It is well known that pairs of concepts like big/small, long/short, up/down, above/below, and so forth, which would appear to be symmetric, in fact show a fundamental cognitive asymmetry: by several criteria, the first member of each pair is privileged over the second one.6 To begin with, the words ‘big’, ‘long’, etc. are unmarked in the linguistic sense of the word. We normally ask how long, not how short, something is, and the name for the measure is ‘length’ rather than ‘shortness’. In many cases markedness is reflected morphologically, as in the pair ‘happy/unhappy’, unmarked words are more frequent than their marked counterparts, and tend to be learned earlier. Furthermore, these regularities hold across languages; so if any language has words that translate into English as ‘big’ and ‘small’, the first one will be the unmarked form—which already indicates that what linguists call markedness does not depend on linguistic convention alone. Even more compelling evidence for the claim that the asymmetries in question are language-independent is provided by McGonigle and Chalmers’s (1986, 1996) discovery that they are present in non-human primates, too. In one experiment, McGonigle and Chalmers presented squirrel monkeys with series of objects of varying sizes. If all the objects were black, subjects had to select the largest item, if all were white, the smallest. Once their performance was perfect, the monkeys’ decision times were recorded, and it was found that they were consistently faster when they had to find the largest item in a set. This and other results confirm that the asymmetry between large and small is not just a linguistic matter.

Bart Geurts and Frans van der Slik 111

6 EXPERIMENTAL EVIDENCE In the foregoing we have discussed several predictions that follow naturally from the idea that monotonicity inferences play an important part in human reasoning:
Even if the quantifiers involved are non-trivial, logically speaking, monotonicity inferences may be very easy.
If the monotonicity profiles of two quantifiers are the same or similar, we expect this to be reflected in experimental data. For example, since they are both upward entailing, monotonicity inferences involving ‘all’ will be about as difficult or easy as the corresponding inferences with ‘most’.
Complex expressions all of whose parts are upward entailing are easier to process than compounds whose parts have mixed monotonicity profiles. In particular, downward entailing quantifiers built from cardinals (‘at most n’, ‘fewer than’) are more complex than others.
Upward inferences are easier than downward inferences.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

In many though not all cases, the dimension to which an opposition pair applies has a natural end point, and if it has the direction away from it is the favoured one. Take size, for example. The extent of an object must be greater than zero, or it wouldn’t exist, and accordingly the natural direction is from smaller to greater size. And sure enough, as we noted already, ‘big’ is the unmarked form vis a` vis ‘small’, and in other respects too the concepts related to larger size are privileged compared with concepts related to smallness. Another example is provided by the opposition between up and down. In this case the relevant dimension has no intrinsic end point: there is no contradiction in the notion of something going up or down forever. But, as observed by Clark (1973), for humans and other agents there is a canonical end point, which is the ground on which they stand. If this is correct, it should be the case that the favoured direction is upwards, and this prediction, too, has been confirmed (Clark 1974). Thus far we have concentrated our attention on pairs of spatial concepts, but what we have said carries over to the temporal domain, for example, and to more abstract opposites like good/bad, clever/dumb, strong/weak, and so on. This being so, we conjecture that there is an asymmetry between upward and downward monotonicity inferences, and if such an asymmetry exists, upward inferences should be easiest, because no set is smaller than the empty set.

112 Monotonicity and Processing Load In order to test these predictions, we conducted the following experiment.

6.1 Materials and procedure Forty-five first-year students at the University of Nijmegen received a 24-page booklet with instructions printed on the cover. Each page presented an argument of the following form:

In each argument DetA was instantiated with one of the following: ‘every’, ‘most’, ‘some’, ‘at least 3’, ‘at most 3’, ‘no’; DetB was either ‘more than 2’ or ‘fewer than 2’. A, B, and C were arbitrary substantives like ‘forester’, ‘nurse’, ‘socialist’, etc. Each subject was taken through the individual tasks in a different, randomly determined order. The instructions explained that the arguments referred to an imaginary tennis tournament, in which stewardesses played against foresters, communists against nurses, and so on. Subjects were to decide whether arguments were valid or not, and the notion of validity was summarily elucidated, with due emphasis on the key elements (‘IF the premisses are true, the conclusion MUST be true as well’).

6.2 Results The results of the experiment are presented in Table 2. To analyse these data, a repeated measures analysis of variance was conducted with three within-subject factors: the first determiner of the major premiss (DetA), the second determiner of the major premiss (DetB), and the orientation of the minor premiss (Minor), which was either upward or downward. As the sphericity assumption was not satisfied, we used the HuynhFeldt epsilon. This yielded main effects for all factors (DetA: F(4,185) ¼ 12.631, p < 0.001; DetB: F(1,44) ¼ 17.956, p < 0.001; Minor: F(1,44) ¼ 25.507, p < 0.001). There were interactions between DetA and DetB (F(2,88) ¼ 3.267, p < 0.05) as well as between DetB and Minor (F(1,44) ¼ 7.977, p < 0.01). Other potential interactions failed to reach significance. Pairwise comparisons between the DetA determiners revealed significant differences between ‘at most’ and all of the others (p < 0.001, throughout). Otherwise, differences between DetA determiners were not significant. With a Bonferroni adjustment for the number of

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

MAJOR PREMISS: DetA A played against DetB B. MINOR PREMISS: EITHER: All B were C OR: All C were B. CONCLUSION: DetA A played against DetB C.

Bart Geurts and Frans van der Slik 113 Table 2 Percentages of correct answers (n ¼ 45).

DetA

DetB

every

more than fewer than

at least

more than fewer than more than fewer than

% (s)

DetA

DetB more than

all all all all

BC CB BC CB

91 69 71 58

(29) (47) (46) (50)

most

all all all all

BC CB BC CB

96 69 53 51

(21) (47) (51) (51)

some

all all all all

BC CB BC CB

51 38 36 49

(51) (49) (48) (51)

no

fewer than more than fewer than more than fewer than

Minor

% (s)

all all all all

BC CB BC CB

91 67 62 60

(29) (48) (49) (50)

all all all all

BC CB BC CB

87 67 60 62

(34) (48) (50) (49)

all all all all

BC CB BC CB

69 53 73 64

(47) (50) (45) (48)

equations, pairwise comparisons between DetA/‘more than’ and DetA/ ‘fewer than’ showed no significant differences for DetA ¼ ‘no’ and ‘at most’, significant differences for DetA ¼ ‘at least’ and ‘most’ (t ¼ 4.401, p < 0.0005, and t ¼ 2.874, p ¼ 0.003, respectively), and differences of borderline significance for DetA ¼ ‘every’ and ‘some’ (t ¼ 2.319, p ¼ 0.012, and t ¼ 2.049, p ¼ 0.023, respectively).

6.3 Discussion The results generally accord with our predictions. To begin with, it is evident that this type of task need not be particularly difficult, as we predicted. Some of the arguments are correctly assessed in over 90% of the cases, even if they involve compound quantifiers like ‘at least 2 nurses’ or quantifiers that are not first-order definable, like ‘most communists’. Secondly, the response patterns associated with individual determiners are in line with our predictions, too: while the upward entailing determiners ‘all’, ‘most’, ‘some’, and ‘at least’ evoke very similar response patterns, the downward entailing determiners are clearly different. Thirdly, as predicted, arguments with ‘at most’ are exceptionally difficult: their scores are at or below chance level in all conditions. Note, in particular, that ‘at most’ deviates from ‘no’, even though both determiners are downward entailing, and that the contrast between ‘at most’ and ‘no’ is not mirrored by a contrast in the counterpart pair ‘at least’/‘some’, which suggests very strongly that quantifiers with ‘at most’ are difficult not for structural but for semantic reasons. Hence, the data confirm what is perhaps the most striking

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

at most

Minor

114 Monotonicity and Processing Load

prediction of our theory, viz. that downward-entailing quantifiers built from cardinals deviate from all others. The interaction between DetA and DetB confirms our expectations, too (see Figure 1). The easiest arguments are those in which DetA and DetB are both upward entailing. With DetA upward and DetB downward entailing, an argument is harder to process. With DetA ¼ ‘no’ the pattern is reversed, and arguments with ‘at most’ are so complex, apparently, that the distinction between upward and downward entailing DetB doesn’t register anymore. One notable finding is that overall arguments with ‘no’ do not appear to be more complex than arguments with upward entailing determiners, which goes against the general trend, in the experimental literature, for positive expressions to be easier than their negated counterparts (Clark 1974, Horn 1989). A closer look at our data reveals that the explanation for this apparent discrepancy may be in line with our speculations in section 5.2. To explain this, we set aside ‘at most’ and compare the arguments with DetA ¼ ‘no’ to the arguments in which DetA is upward entailing. In this group, the ‘no’ arguments score worse than the others with DetB ¼ ‘more than’ and better than all the others with DetB ¼ ‘fewer than’. More accurately, whereas arguments with upward entailing DetA are, on the whole, significantly easier with DetB ¼ ‘more than’ than with DetB ¼ ‘fewer than’, the difference is obliterated wih DetA ¼ ‘no’. Hence, ‘no’ arguments confirm our hypothesis that sentences with mixed monotonicity profiles are more complex than others. Furthermore, it appears that the initial impression

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

Figure 1 Interaction between DetA and DetB (percentages of correct answers).

Bart Geurts and Frans van der Slik 115

7 CONCLUDING REMARKS Natural language semantics is not a branch of psychology. Rather, we view linguistic meaning as a topic lying at the crossroads of several

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

that subjects perform surprisingly well with ‘no’ is misleading: they do well on one particular class of ‘no’ arguments, i.e. the ones in which DetB ¼ ‘fewer than’. That is to say, the positive effect is due entirely to those arguments whose major premiss contains two downwardentailing determiners. This makes sense if we assume, for reasons discussed in section 5.2, that subjects tend to use the conversion method, and before making monotonicity inferences translate ‘No A played against fewer than 2 B’ into ‘All A played against at least 2 B’. However, even if the conversion method is used to simplify an argument, the procedure for solving experimental tasks with ‘no’ sentences is more complex than for tasks with ‘most’, for example. Why isn’t this reflected in our data? First, it isn’t quite correct to say that the difference isn’t reflected in the data, for in absolute terms subjects perform worse on tasks with ‘no’ than on tasks with upward entailing determiners: with ‘no’ 64.75% of the answers are correct, while the upward entailing determiners range between 67.75% and 72.25% (for ‘at least’ and ‘every’, respectively) with a mean of 69.6%. These differences are not statistically significant, to be sure, but it is at least possible that a weak effect would be found for a larger n. Secondly, it should be noted that the experimental task in our study is different from the experimental tasks that have been used to argue that negative expressions are more complex than positive ones. The cognitive processes studied here are slow and long-winded in comparison to, for instance, Clark’s (1974) reaction time experiments, which makes it less surprising that effects demonstrated in one paradigm fail to appear in another. The observed interaction between DetB and Minor seems a bit puzzling at first. As it turns out, if DetB is upward entailing the orientation of the minor premiss is significant: if it goes upwards, the average rate of correct responses is 0.81 and this figure drops to 0.61 if the direction of the argument is reversed. For downward entailing DetB, by contrast, the two rates nearly coincide: 0.59 and 0.57, respectively. Why is it that the orientation of the minor premiss makes a difference only in the case of an upward entailing DetB? The answer to this question, we suspect, is simply that this is a floor effect: subjects’ performance approximates chance level already with downward entailing DetB and upward Minor, so the effect of choosing a downward Minor instead is bound to be limited.

116 Monotonicity and Processing Load

Acknowledgements We would like to thank Larry Horn, Keith Stenning, and the anonymous reviewers for the Journal of Semantics for their helpful comments on earlier versions of this paper.

Received: 28.05.04 BART GEURTS AND FRANS VAN DER SLIK Final version received: 12.10.04 Philosophy Department/Department of Linguistics University of Nijmegen P. O. Box 9103 NL-6500 HD Nijmegen The Netherlands e-mail: [email protected]

REFERENCES Barwise, J. & R. Cooper (1981) ‘Generalized quantifiers and natural language’. Linguistics and Philosophy 4:159–219. Carston, R. (1998) ‘Informativeness, relevance and scalar implicature’. In

R. Carston and S. Uchida (eds). RelevanceTheory: Applications and Implications. 179–236. Clark, H. H. (1973) ‘Space, time, semantics, and the child’. In T. Moore

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

disciplines, of which psychology is one. This being so, it is to be expected that semantics is directly relevant to human psychology. The results presented in the foregoing illustrate how semantic theory can be brought to bear on behavioural data. We have proposed a (partial) semantic measure of linguistic complexity, and derived from it predictions about the behaviour of experimental subjects, without using a processing model. The predictions made by our account are based on a competence theory of quantification, which is motivated mainly, though not exclusively, by semantic evidence. Therefore, in some quarters of psychology, our account would not qualify as psychological. Most theories in the psychology of reasoning, for instance, involve a detailed architecture of the reasoning faculty, and it is often supposed, overtly or tacitly, that this is a sine qua non for psychological theorizing. In our opinion, such views are seriously mistaken. Like all natural phenonema, behaviour can be described and explained on many levels, and we see no reason to believe that any one of these levels is more pertinent to psychology than the others. Part of the game of science is to find the level at which a given phenomenon is best explained, and for certain aspects of human reasoning, we maintain, that level is semantical.

Bart Geurts and Frans van der Slik 117 Levinson, S. C. (2000) Presumptive Meanings. MIT Press. Cambridge, MA. McGonigle, B. & M. Chalmers (1986) ‘Representations and strategies during inference’. In T. Myers, K. Brown, & B. McGonigle (eds). Reasoning and Discourse Processes. Academic Press. London, 141–164. McGonigle, B. & M. Chalmers (1996) ‘The ontology of order’. In L. Smith (ed.). Critical Readings on Piaget. Routledge. London, 279–311. Musolino, J. (2004) ‘The semantics and acquisition of number words: integrating linguistic and developmental perspectives’. Cognition 93: 1–41. Musolino, J., S. Crain, & R. Thornton (2000) ‘Navigating negative quantificational space’. Linguistics 38:1–32. Newstead, S. E. (2003) ‘Can natural language semantics explain syllogistic reasoning?’ Cognition 90:193–199. Noveck, I. (2001) ‘When children are more logical than adults: experimental investigations of scalar implicature’. Cognition 78:165–188. O’Leary, C. & S. Crain (1994) ‘Negative polarity items (a positive result), positive polarity items (a negative result)’. Paper presented at the 19th Boston University Conference on Language Development. Papafragou, A. & J. Musolino (2003) ‘Scalar implicatures: experiments at the syntax-semantics interface’. Cognition 86:253–282. Sa´nchez Valencia, V. (1991) Studies on natural logic and categorial grammar. Doctoral dissertation, University of Amsterdam. van der Wouden, T. (1997) Negative Contexts: Collocation, Polarity and Multiple Negation. Routledge. London.

Downloaded from jos.oxfordjournals.org by guest on January 1, 2011

(ed.). Cognitive Development and the Acquisition of Language. Academic Press. New York, 27–63. Clark, H. H. (1974) ‘Semantics and comprehension’. In T. Sebeok (ed.). Current Trends in Linguistics, volume 12. Mouton. The Hague, 1291–1428. Also published separately under the same title: 1976, Mouton, The Hague. Geurts, B. (1998) ‘Scalars’. In P. Ludewig & B. Geurts (eds). Lexikalische Semantik aus kognitiver Sicht. Gunter Narr. Tu¨bingen, 85–117. Geurts, B. (2002) ‘Donkey business’. Linguistics and Philosophy 25:129–156. Geurts, B. (2003a) ‘Reasoning with quantifiers’. Cognition 86:223–251. Geurts, B. (2003b) ‘Monotonicity and syllogistic inference: a reply to Newstead’. Cognition 90:201–204. Heim, I. (1984) ‘A note on negative polarity and downward entailingness’. In Proceedings of NELS 14, pp. 98–107. Horn, L. R. (1972) On the semantic properties of the logical operators in English. Ph. D. thesis, University of California at Los Angeles. Horn, L. R. (1989) A Natural History of Negation. Chicago University Press. Chicago. Horn, L. R. (1992) ‘The said and the unsaid’. In C. Barker & D. Dowty (eds). Proceedings of SALT 2, pp. 163–192. Kanazawa, M. (1994) ‘Weak vs. strong readings of donkey sentences and monotonicity inference in a dynamic setting’. Linguistics and Philosophy 17:109–158. Keenan, E. L. & D. Westersta˚hl (1997) ‘Generalized quantifiers in linguistics and logic’. In J. van Benthem & A. ter Meulen (eds). Handbook of Logic and Language. Elsevier/MIT Press. Amsterdam/Cambridge, MA, pp. 837–893.