Cognition, Vol. 7, No. 1

Editorial board Franccsco Antinucci Institute di Psicologia, CNR, Via dei Monti Tiburtini 509, 00157 Rome, Italy Hiroshi Faculty Tokyo Hongo, Tokyo,

Azuma of Education, University, Bunkyo-ku Japan

Paul Bertelson Laboratoire de Psychologie Experimentale, Universite Libre de Bruxelles 117 Av. Adolphe Buyl, B-l 050 Brussels, Belgium Ned Block Dept. of Philosophy, M.I. T., Cambridge, Mass. 02139, U.S.A.

Anne Cutler Laboratory of Experimental Psychology, Centre for Research on Percep tion and Cognition, University of Sussex, Brighton BNl, Gt. Britain Margaret Donaldson Department of Psychology, University of Edinburgh, 1-7 Roxburgh Street, Edinburgh EH8 9TA, Gt. Britain Peter D. Eimas Walter S. Hunter Laboratory of Psychology, Brown University, Providence, RI. 02912, U.S.A.

Francois Bresson Laboratoire de Psychologie 54 bvd. Raspail, F. 75006 Paris, France

Gunnar Fant Lab. of Speech Transmission, Royal Institute of Technology, S-l 0044 Stockholm 70, Sweden

Roger Brown Dept. of Psychology, Harvard University, Cambridge, Mass. 02138,

Gilles Fauconnier 1 Rue des Guillemites, 75004 Paris, France

U.S.A.

Jerome S. Bruner Dept. of Exp. Psychology, University of Oxford, South Parks Road, Oxford, OX1 3UD, Gt. Britain

David Fay Department of Psychology, University of Illinois, Chicago, Box 4348, Chicago, Ill. 60680, U.S.A.

Noam Chomsky Dept. Modern Languages and Linguistics, M.I.T., Cam bridge, Mass. 02139, U.S.A

Jerry Fodor Dept. of Psychology, M.I. T. El O-034 Cambridge, Mass. 02139,

Eve Clark Department of Linguistics, Stanford University, Stanford, Calif. 94305, U.S.A. Richard Cromer MRC Developmental Psychology Unit, Drayton House, Gordon Street, London, WCIH OAN, Gt. Britain

Kenneth Dept. of Monash Clayton,

U.S.A.

Forster Psychology, University, Vie. 3168, Australia

Merrill Garrett Department of Psychology, M.I. T. El O-034, Cambridge, Mass. 02139, U.S.A.

Lila Gleitman Graduate School of Education, University of Pennsylvania, 3700 Walnut Street, Philadelphia, Pa. 19104, U.S.A. David T. Hakes, Department of Psychology, University of Texas, Austin, Tex. 78712, U.S.A. Henry Hecaen Directeur d’Etudes, Ecole Pratique des Hau tes Etudes, Unite de Recherches Neuropsychologiques, I.N.S.E.R.M., 2, rue d’Alesia, F- 75014 Paris, France Michel Imbert Laboratoire de Neurophysiologie, College de France, 11 place Marcelin Berthelot, F- 75005 Paris, France Barbel Inhelder Faculte de Psychologie et des Sciences de IEducation, Universite de Genkve, CH-1211 Geneva 14, Switzerland Marc Jeannerod Laboratoire de Neuropsychologie Experimentale, 16 kv. Doyen L&pine, F-69500 Bron. France Philip Johnson-Laird Laboratory of Experimental Psychology, Centre for Research on Perception and Cognition, Sussex University, Brighton BNl 9QG, Gt. Britain Peter W. Jusczyk Department of Psychology, Dalhousie University, Halifax, N. S., Canada B3H 4JL

Jerrold J. Katz Dept. of Linguistics, CUNY Graduate Center, 23 W 42nd Street, New York, N. Y. 10036, U.S.A. Mary-Louise Kean Cognitive Science Program, School of Social Sciences, University of California, Irvine, Calif 92717, U.S.A. Edward Klima Dept. of Linguistics, La Jolla, University of California, San Diego, Cali’ 92037, U.S.A. Stephen M. Kosslyn Department of Psychology and Social Relations, Harvard University, William James Hall, 33 Kirkland Street, Cambridge, Mass. 02138, U.S.A. James R. Lackner Department of Psychology, Brandeis University, Waltham, Mass. 02154, U.S.A. Harlan Lane Department of Psychology, Northeastern University, 360 Huntington Avenue, Boston, Mass. 02115, U.S.A. Willem Levelt Psychological Laboratory, Nifmegen University, Erasmuslaan 16, Nijmegen, Netherlands John Lyons Dept. of Linguistics, Adam Ferguson Building, Edinburgh EH8 9LL, Gt. Britain David McNeil1 Department of Behavioral Sciences, Committee on Cognition and Communication, University of Chicago, 5848 South University Avenue, Chicago, Ill. 60637, U.S.A. John Marshall Psychological Laboratory, Niimegen University, Erasmuslaan 16, Nijmegen, Netherlands

JosC Morais Laboratoire de Psvchologie Experimentale, Universitt Libre de Bruxelles, I I 7 Avenue Adolphe Buyl, B-I 050 Brussels, Belgium

Tim Shallicc Psychology Department, The National Hospital for Nervous Diseases, QueenS Square, London WCI, Gt. Britain

John Morton Applied Psychology Unit, 15 Chaucer Road, Cambridge CB2 ZEF, Ct. Britain

Dan 1. Slobin Department of Psychology, University of California, Berkeley, Calif 94720. U.S.A

George Noizet Laboratoire de Psychologie Experimen tale, 29 Av. R. Schuman, F-13 Aix en Provence, France

Sidney Strauss Department of Educational Sciences, Tel Aviv University, Ramat Aviv, Israel

Daniel Osherson ZOC-I 24 (DSRE), M.I. T., Cambridge, Mass. 02139,

Michael Studdert-Kennedy Department of Communication Arts and Sciences, Queens College, City University of New York, Flushing, N. Y. 11367, U.S.A.

U.S.A.

Michael Posner Dept. of Psychology, University of Oregon, Eugene, Ore. 97403, U.S.A. David Premack Psychology Department, University of Pennsylvania, 3813-15 Walnut Street, Philadelphia, Pa. I91 74, U.S.A Zenon Pylyshyn Dept. of Linguistics and Psychology, M.I. T., Cambridge, Mass. 02139,

U.S.A.

Andrk Roth Lecours Hotel-Dieu de Montreal 3840 rue St. Urban, Montreal, Quebec H2W I T8, Canada Steven Rose BioIogy Department, The Open University, Walton Hall, Milton Keynes MK 7 6AA, Gt. Britain Klaus R. Scherer Fachbereich 06, Psychologie der Justus Liebi. Universitat, Otto Behegel Strasse, 6300 Giessen, F.R.G. Scania de SchGnen Laboratoire de Psychologie, 54 Boulevard Raspail, 75270 Paris Cedex 06, France

David Swinney Department of Psychology, Tufts University, Medford, Mass. 02155, U.S.A. Alina Szeminska Olesiska 513. Warsaw. Poland Virginia Valian Ph.D. Program in Psvchologv. C. U.N. Y. Graduate Center,. _ 33 West 42nd Street, New York, N.Y. 10036, U.S.A. Edward Walker Department of Psychology, ML T., Cambridge, Mass. 02139, U.S.A. Peter Wason Psycholinguistics, University College London, Research Unit, 4 Stephenson Way, London NW1 2HE, Gt. Britain Dr. Edgar Zurif Aphasia Research Center, Boston University Medical Center, 150 South Huntington Avenue, Room Cl S-5, Boston, Mass. 02130, U.S.A. Hermina Sinclair de Zwart Centre d’Epistemologie Gknetique. Universite de Geneve, CH-I ZII Geneva, Switzerland

Cognition, 7 (1979) l-2

Editorial

Cognition is entering into its seventh year of existence. We were not ripe enough at seven minus two and we do not want to wait until seven plus two so we have decided to get a number of changes under way right now in order, hopefully, to improve the form and substance of our journal. First, I am very happy to inform you that both John Morton and Michael Posner have agreed to join Thomas Bever, co-founder of Cognition, as Associate Editors of the journal. We expect that these additions will help us to secure a broader coverage of the various aspects of cognition in general and give us better geographical representation. It is the vocation of Cognition to serve as a vehicle for discussion and it is for this purpose that our Discussion section will be preserved in spite of the creation of a new Brief Reports and Letters section. This new section will contain notes, short experiments and Letters to the Editor. We have taken this initiative because a number of short, intelligent experimental contributions have not been accepted due to our bias towards studies with a broad theoretical base. However, both kinds of contributions are desirable and necessary in a field such as ours. Good data presented within an intelligent theoretical framework can be as important as the theoretical developments that they may in turn trigger. This being said, the occasional short papers that we hope to publish are not the sole reason for inaugurating this new section. It is our feeling that quite apart from being a means of conveying research results, the journal should be an instrument which favors the development of easy communication between members of the scientific community. Hence, interesting reflections concerning the field, methods, ways of working, the wider scientific context, can all be of potential interest to others. Short notes on theoretical, historical or scientific matters are all candidates for inclusion in this short section. We therefore encourage you strongly to send us any contributions which you think might fit this description. However, at least at first, we reserve the right to decide on them without appealing to the usual review process. Finally, I would like to draw your attention to a problem which is of some concern to this journal and which might be the object of fruitful discussion. It is often said that the pressures of the academic world make people write papers that serve no real purpose other than that of advancing their author’s careers. If this is so it is certainly a sad turn of events for authors who find themselves, like many other workers, having to produce something which they do not respect and whose purpose they do not fully grasp. However,

this situation is also difficult for those of us who act as reviewers and have to rate a great deal of this production with a vast loss of personal time. As active scientists it is obvious that we can increase both our colleagues’ and students’ awareness of this problem for which I see no simple solution beyond that of true discussion of all the workings of academia to which we contribute and from which we suffer. In closing, may I remind you that we would like readers to let us know in what way Cognition can carry more services to the profession. Some of your ideas may not have occurred to us and others may be germinating. For those that are still in the bud your suggestions will help us to make up our minds and bring projects of mutual interest to fruition. Please let us know what you think. Jacques Mehler

cognirion. 7 (1979) 3-17 @lsevier Sequoia S.A., Lausanne-Printed

1 in the Netherlands

A cross-linguistic study of the processing of causative sentences* MARY SUE AMMON and DAN I. SLOBIN University

of California,

Berkeley

Abstract The comprehension of sentences expressing instigative causation (e.g., The horse makes the camel run) was investigated in children between the ages of 2;0 and 4;4, speaking English, Italian, Serbo-Croatian and Turkish. Crosslinguistic differences in development reveal the roles of morphological (causative particle, case inflection) and syntactic devices (periphrasis, word order) in guiding children’s processing of such constructions. It is suggested that local cues (inflectional suffixes, particles, specialized causative verb forms) contribute to the more rapid development of sentence processing strategies in Serbo-Croatian and Turkish. The word order systems of English and Italian, which require that the listener hold the entire sentence pattern in mind in order to determine underlying semantic relations, contribute to slower development on this task,

Children’s comprehension of causative constructions was studied as one part of a large cross-linguistic investigation conducted in Berkeley, Rome, Dubrovnik and Istanbul (Aksu, 1978; Clancy, Jacobsen and Silva, 1976; Johnston and Slobin, 1977; Radulovic, 1975; Slobin, 1978; Slobin and Bever, 1978). Our overall concern is the effect of grammatical form on the developing ability to express basic concepts in language. The data consists of crosslinguistic differences in the rate and pattern of acquisition of the means for encoding notions of space, time, agency and causation. *This study is part of the Berkeley Cross-Linguistic Acquisition Project (Dan I. Slobin, Principal Investigator), carried out with generous and appreciated support from the William T. Grant Foundation to the Institute of Human Learning and from NIMH to the Language-Behavior Research Laboratory of the University of California. The investigation was designed in collaboration with Ayhan Aksu, Francesco Antinucci, Thomas G. Bever, Susan Ervin-Tripp, Judith R. Johnston and Ljubica Radulovid. We gratefully acknowledge the labors of our testers: Penny Boyes-Braem, Judith R. Johnston and Gail Loewenstein Holland in the United States; Rosanna Bosi and Wanda Gianelli in Italy; Ljubica Radulovid and Emma Zalovic in Yugoslavia; and Ayla Algar and Alev Orhon in Turkey. Reprint requests should be sent to: Dan I. Slobin, Department of Psychology, University of California, Berkeley, Calif. 94720, U.S.A.

4

Mary Sue Ammon and Dan I. Slobin

In the present study we are concerned with forms for the expression of causation-that is, situations in which one agent impels another agent to act. English has a variety of means-lexical, morphological and syntactic-for such expressions. Here we consider the productive expression most accessible to children: the periphrastic construction using make as a causative verb, as in, “You made me do it.” The four languages studied-English, Italian, Serbo-Croatian and Turkish-have different means for the production of such expressions. The first three, all Indo-European SVO (subject-verb-object) languages, have similar periphrastic constructions, differing in regard to the roles of word order and case inflection. By contrast, Turkish, an agglutinative SOV language, encodes causation by the insertion of one or more causative particles in the verb. Table 1 compares the productive causative expressions in the four languages, using one of the sentences from the investigation. instigative

TABLE 1, Comparison of causative constructions in English, Italian, Serbo-Croatian and Turkish

English The horse makes NOUN VERB

causatiue

the camel run NOUN VERB

inftnittue

third person Italian I1 cavallo [the horse] NOUN

fa [makes] VERB causative

correre [to run] VERB

infinitive

il cammello. [the camel] NOUN

third person Serbo-Croatian idrijebe [horse] NOUN nominative

Turkish At [horse] NOUN

tjera

devu

[drives]

[camel]

VERB

NOUN occusatiue*

causative third person

deveyi [camel]

nominattve NOUNaccusative

da

trEi. [runs]

PARTICLE

VERB third perSOn

kostursun.

[should make run] VERB causative optatiue third person

*The noun in the example is feminine, and therefore has a distinct contrast between nominative (deva) and accusative (devu). Neuter nouns, like tdrijebe, however, appear in the identical form in both nominative and accusative. Sentences used in the study exhibited three types of noun combinations: neuter-feminine (object marking), feminineneuter (subject marking), and neuter-neuter (neither).

Processing of causative sentences

5

Note that there are word-order differences among the three SVO languages-roughly: English: The horse makes the camel run; Italian: The horse makes run the camel; Serbo-Croatian: The horse makes the camel that (it) runs. In both English and Serbo-Croatian, the action verb appears at the end of the sentence, with the causative verb in the normal verbal position between the two nouns; in Italian the entire verbal construction remains in normal position. In English and Italian the causative verb (makes, fa) agrees with the subject in person and number, while the action verb is in the infinitive (run, correre). In Serbo-Croatian, a particle separates the causative clause from the action clause, and both verbs are finite. In addition, the case inflectional system marks the instigator as subject and the instigated as object. Also, the causative verb in Serbo-Croatian (tjerati) is a specialized one, used to discuss driving animals and impelling action, whereas the English make and Italian fare are general verbs performing a large number of other functions. In terms of surface marking, therefore, Serbo-Croatian seems to provide the clearest cues of the three Indo-European languages. The Turkish causative retains standard SOV order, embedding a causative particle within the verb. The particle follows the verb root (agreeing with it in vowel harmony) and is followed by particles of person, number, tense, mood and aspect (e.g., kos ‘run’ + -uyor ‘third person’ = ko$uyor ‘he/she runs’; + -tur- ‘causative’ = kogturuyor ‘he/she makes (someone run’). ’ As in Serbo-Croatian, the case inflectional system marks the instigator as subject and the instigated as object. However, unlike Serbo-Croatian, Turkish has no grammatical gender and no exceptions to the regular inflectional paradigm. As a result, the object is always clearly marked in Turkish. In Serbo-Croatian, however, neuter nouns appear in unmarked form for both nominative and accusative, thus making it impossible to mark neuter direct objects inflectionally.2 Turkish, then, provides clear surface cues for the causative, but of a different sort than the Indo-European languages. These contrasts between the four languages allow us to compare the roles of morphological (causative particle, case inflections) and syntactic devices (periphrasis, word order) in facilitating children’s comprehension of causative constructions. Very little work has been done on the development of these forms in child speech (Aksu, 1975; Baron, 1977; Bowerman, 1974, 1977). Primarily what ’ There are actually several different forms of the causative particle in Turkish. In this study we used the most common form,--&(ko@~- ‘make run’, ytizdtir- ‘make swim’, yutzr- ‘make lie down’) and the reduced form -t-, used with polysyallabic verb stems ending in a vowel (at&- ‘make jump’, yiitit‘make walk’, uyuf- ‘make sleep’). There were no differences in results for the two forms of the morpheme, and the data for the two forms are presented together. *The same is true of masculine inanimate nouns in the singular, and neuter and feminine nouns in the plural. As shown in Table 3, only singular neuter and feminine nouns were used in this study.

6

Mary Sue Ammon

and Dan I. Slobin

has been observed is that both the inflectional structure in Turkish and the periphrastic structure in English start appearing in child speech between the ages of two-and-a-half and three. Although the evidence is limited, it is probably the case that children in all four languages hear instigative causatives frequently, and can use them productively sometime in the third year of life. In the present study we examine the ability of children between the ages of 2;0 and 4;4 to comprehend such constructions in an experimental task. Our major concern in this study lies in cross-linguistic comparison of the course of acquisition of these constructions. Control of forms in free speech, supported by communicative intent, generally occurs at an earlier age than is demonstrated in formal comprehension testing, but this fact should not alter the language-specific development patterns which we seek.

Method Subjects The causative comprehension measure reported in this paper was one of 16 linguistic and nonlinguistic tests offered to matched samples of two-, three-, and four-year-olds in the four countries. Our plan was to study 48 children in each field site, in groups of three boys and three girls spaced at four-month intervals between the ages of 2;0 and 4;4. By and large, we completed this plan, though there are gaps in the data at various points, due to the tribulations of carrying out research in foreign countries within a limited time schedule. Table 2 presents a summary of the subject sample participating in this phase of the investigation. Since we were interested in cross-linguistic, and not cross-cultural factors, we tried, as much as possible, to equate our samples on sociocultural grounds. That is, we limited ourselves to children of urban, professional parents, at least one of whom had a college degree. By Table 2.

Number of subjects contributing

to causative comprehension

Group

of children

A B C D Total

Age

2;0-2;4 2;8-3;0 3;4-3;8 4;0-4;4

Number

data

English

Italian

Serbo-Croatian

Turkish

10 11 12 12

10 12 12 11

12 12 12 12

10 8 11 10

45

45

48

39

Processing of causative sentences

7

and large, we feel that we were working with a fairly homogeneous group of children across the four field sites, at least in terms of early material and intellectual experience. Each child was studied extensively, over a period of 1 S-20 hours, within a ten-day span. The causative task was administered midway in the series. The examiner was always female, and always a native speaker. Procedure

Examiner and child sat at the same side of a table. The child was presented with a pair of toy animals or dolls and was asked to demonstrate an action instigated by one and carried out by the other (e.g., The horse makes the camel run). The child was already familiar with these animals from two sessions of a word order comprehension task (Slobin and Bever, 1978), in which s/he was to demonstrate actions involving two animals and a reversible transitive verb (e.g., The horse kicks the camel). A pair of animals was introduced and named, and the relevant action was demonstrated for each animal separately (The camel can run; The horse can run). The child was then instructed to demonstrate the causative interaction. Each sentence was offered up to three times, in a medium flat intonation at normal speaking rate with equal stress on both nouns. If the child’s response, after three presentations, remained unclear, s/he was probed by questions as to what each of the animals had done. A correct response required that the first animal instigate the action (either by direct contact with the second animal, or by verbal instruction uttered by the child for the sake of the first animal) and that the second animal carry out the action. Responses were also counted as correct if, on questioning, the child attributed instigation to the first animal (e.g., The horse

told the camel to run; The camel ran because

the horse wanted him to;

etc.). There were five classes of errors: (1) no response or unclear response, (2) reversal (e.g,, the camel made the horse run), (3) both animals act separately, (4) only the first animal acts, (5) only the second animal acts. Design

There were six causative configurations, as shown in Table 3. The order of the six configurations was always the same. For half of the children, the first participant listed in Table 3 was the instigator; for the other half it was the second. To control for possible biases of physical placement of the animals, each instigator was placed an equal number of times to the right and to the left of the other participant. The syntactic constructions of the stimulus sentences are presented in Table 1. In Serbo-Croatian, feminine nouns have distinctive nominative and accusative forms, whereas neuter nouns appear in uninflected form for both of these cases. In order to test for the effects of

8

Mary Sue Ammon and Dan I. Slobin

Table 3.

Verb-noun

configurations

used in causative comprehension

task

l.:nglish

Italian

Scrbo-Croatian*

Turkish

makes run the horse the camel

FJ corrcrc il cavallo il cammollo

tjera da trti Zdrijcbe (N) dcva (I:)

kosturwn at devc

makes s\vim the dog the cat

fa nuotare il cant il g3 tto

tjcra da pliva ktenc (N) maEe (N)

yiiLdiirsiin kBpck kcdi

makes lie down the boy the girl

fil wdcrc il bambino la bambina

tjera da Ic2e djctc (N) curica (F)

yattlrsin

makes jump the chick the lamb

fa saltarc I3 pallina I’agnellino

tjera da \koEi pile (N) janje (N)

atlat
makes walk the pig

fa camminarc il maialino 13 pecora

tjcra da Scdc prasc (N) “VU (F)

yiiriitsiin domuz koyun

fa dormirc 13 capra la zebra

tjcra da spava ko/le (N) zebra (I:)

uyutsun kqi xbra

the

sheep

makes go to alccp the

goat

the

zebra

*Gender

is indicated

by (N) neuter

Of!lClIl

kiz

and (I:) fcmininc.

inflectional cues in this language, three kinds of participant pairings were used:. (1) neuter-feminine (distinctive object marking), (2) feminine-neuter (distinctive subject marking), (3) neuter-neuter (subject-object relation signalled by word order with no inflectional cue).

Results and Discussion The major analysis focused on the total number of correct acting-out responses on the causative task and the way this score related to language, age and sex. A factorial design analysis of variance was performed on the data. Because the factor of sex of subject did not produce a significant overall effect, and did not interact with any of the other factors, it will not be discussed further. Instead of performing significance tests on the overall main effects and interactions, more specific questions were asked of the data by means of planned contrasts at each level of the analysis.3 Table 4 displays the various 3Sw bottom

of facing page.

9

Processing of causative sentences

Table 4.

Mean percentage

As

Across languages

Language

group

Table 5.

correct by age and language

English

Italian

SerboCroatian

Turkish

A B C D

I 33 12 70

8 60 67 64

68 61 82 95

37 94 96 95

Across ages

38

50

II

81

Significant contrasts on causative comprehension

30 62 19 81

task df

F

1 1

64.141 13.874

Overall Language Effect$ Turkish-Italian Turkish-English SerboCroatian-Italian SerboCroatian-English Turkish and SerboCroatian-English and Italian Turkish-SerboCroatian and English and Italian

1 1 1 1 1 1

26.9541 27.0358 17.1136 22.8410 42.644 14.048

Interaction Effects of Age and LanguageC Turkish-SerboCroatian in quadratic trend SerboCroatian-Italian in quadratic trend

1 1

8.1222 8.0662

Subsidiary Analysis of Differences in Cell Meansd Serbo-Croatian Age A-Italian Age A SerboCroatian Age A-English Age A Turkish Age B-English Age B

1 1 1

19.9303 21.0593 17.3993

=

8.644 7.5076 7.8961 9.4249

Overall Age Effectsa Linear improvement in performance with age (linear trend) Simple departure from linear improvement with age (quadratic

aCritical bCritical ‘Critical dCritical

1~’at F’ at F’ at F’ at

OLof Q of Q of CYof

0.01 0.05 0.09 0.05

for for for for

3 contrasts [CXfor each test = 0.00331 8 contrasts [CXfor each test = 0.006251 18 contrasts [a for each test = O.OOS] 24 contrasts [a for each test = 0.002]

trend)

= = =

3 Instead of alotting the usual alpha level of 0.05 to each of the orthogonal sets of contrasts, the total experimentwise error-rate was divided unequally to allow more power for the important age by language interaction contrasts. Thus (1) an alpha of 0.01 was divided among the three trend contrasts for age, (2) an alpha level of 0.05 was divided among seven language contrasts, and (3) an alpha level of 0.09 was divided among the 18 age trend by language interaction contrasts.

10

Mary Sue Ammon and Dan I. Slobin

means, and Table 5 lists the significant main effect contrasts and significant interaction effect contrasts. It also presents the results of a subsidiary analysis on certain differences of cell means-the significant differences between pairs of languages at the various age levels. It should be noted that the criterion adopted for significance of each of the contrasts was fairly conservative, because of the number of tests performed.4

The significant contrasts within the main effect for age indicate both a significant overall linear improvement in performance with age (linear trend) and a significant departure from this straight line improvement with age (quadratic trend). More simply, this means that performance on this structure is showing improvement within the age range surveyed, but that the rate of growth is not the same from one age period to the next. Consideration of the growth curves for each language in relation to the others further illuminates, as well as complicates, this overall nonlinear growth trend. Language Eight contrasts were made within the main effect for language. The languages marking the direct object of causation with an inflection (Turkish and Serbo-Croatian) were contrasted with the languages marking this case by word order alone (English and Italian). And the languages expressing the causative with a periphrastic construction (English, Italian and Serbo-Croatian) were contrasted with Turkish, which inserts a particle into the verb to carry out this function. In addition, tests were preformed on each of the six possible two-way pairings of the four languages. These tests produced two major findings. First, children learning the two inflectional languages performed better overall on this task than did children learning the two word-order languages. Thus Turkish and Serbo-Croatian, both individuaily and together, have significantly higher means than English and Italian, taken either separately or together. And the Italian mean does not differ significantly from English, nor does Turkish differ overall from Serbo-Croatian. There are a variety of factors which differentiate the two inflectional languages from the two word-order languages, and we are not in a position to pull them apart in this study. But what Serbo-Croatian and Turkish have in common, as opposed to English and Italian, is the presence of morphological particles which may facilitate sentence processing strategies. Both have a direct object inflection. The Serbo-Croatian causative verb, tjera, serves a 4A full summary

table of all of the contrasts

performed

is available

on request.

Processing of causative sentences

11

specialized function in the language, and in this sense is in some way similar to the Turkish causative morpheme, which is also specialized in function. The Serbo-Croatian action verb in the causative construction is finite and is separated from the rest of the sentence by a particle, da, which may also call special sentence processing strategies into play for this type of sentence, or at least serve to block strategies applicable to simple declaratives. Thus, in general terms, Serbo-Croatian and Turkish can be said to provide more local cues within the sentence, in contrast to English and Italian, whose sentence structures require that the listener hold the entire sentence pattern in mind in order to determine the underlying semantic relations. That is to say that in the non-inflectional languages some semantic decisions must be postponed until the entire sentence has been heard, while assignment of semantic roles to participants can be made during the course of sentence processing in the inflectional languages. The second finding relates to another way in which the four languages can be contrasted. English, Italian and Serbo-Croatian express the causative with a periphrastic construction, while Turkish inserts a particle into the verb to carry out this function. And overall performance for Turkish was higher than that for the other three languages combined. Some caution should be exercised in evaluating this result, because (1) this contrast is not independent of the inflectional vs. word-order contrast just discussed, and (2) the Turkish mean was not significantly different from the Serbo-Croatian mean. However, although more evidence is needed to substantiate the proposal that particle constructions may be easier than periphrastic constructions, it is at least clear that such constructions are not more difficult. One might have expected the Turkish form to be more difficult, in that the verb conflates both the causing and the caused phases of the underlying semantic structure (Talmy, 1976). That is, the two phases of action are not as obviously marked on the surface as they are in languages using periphrastic constructions. However, whatever the nature of early sentence processing strategies, this did not pose a problem for the Turkish children. It should also be noted that Turkish has the advantage of a clear and exceptionless inflectional system, which is mastered before the age of two. In addition, the causative sentence follows normal SOV word order. Thus several factors may play a joint role in contributing to the precocious Turkish performance. Language

by Age

The interaction contrasts compared the growth trends across age for each pair of languages. Figure 1 portrays this language by age interaction. The two significant contrasts relate to the nonparallelism of some of the growth curves. The observed differences and similarities both point to particular aspects of

12

Mary Sue Ammon and Dan I. Slobin

Figure

1

Average percentage correct on causative comprehension test by language and age group (E = English, I = Italian, S = Serbo-Croatian, T = Turkish).

100

‘-:,-=-:‘“’

60

A

S

S , -.--

,_.-

,E---_,-__

-----E

/ ,.-.I-.-._.

-.-.-.-,-.-I

./.‘i /

/

/ /

/

/

/

/

/

/

T

/’ /*

/’ /’

/’

20

/’

// / /

/ ,El’

/’ ,’

I/’ A’ E' A

,

C

B

D

Age Group

language development in the four languages investigated. (The differences between growth curves for the four languages cannot be attributed to overall differences between subject samples since the languages patterned differently on other linguistic tasks in our battery [Slobin, 19781.) In this part of the analysis a significant difference can be seen between Turkish and Serbo-Croatian in the shape of the growth curves (the nonparallel quadratic components). While Turkish shows accelerated development between the first and second age periods and reaches an early ceiling, SerboCroatian starts out with a relatively high level of performance but improves more slowly. Though none of the differences in the means between Turkish and Serbo-Croatian within any age is significant, the rapid development in Turkish and the lack of Serbo-Croatian development at the second age level seem mostly responsible for this significant contrast in patterns of growth between the two languages.

13

Processing of causative sentences

Possible reasons for the shape of the Turkish growth curve have already been discussed, but the Serbo-Croatian developmental pattern needs further examination. Several sorts of evidence suggest that at this second age level the Yugoslav children begin to have difficulty attending to both word order and inflectional cues. Although word order and inflection never present contradictory information in this task (that is, the second noun is always the direct object, whether inflectionally marked or not), in the language in general, inflectional marking allows for nonstandard word orders. It seems that at the second age level, the children begin to be aware that case inflections can countermand basic word-order strategies in sentence processing, but they have not yet mastered the necessary inflectional strategies to allow them to identify grammatical relations on the basis of inflectional cues alone. In Table 6 we compare correct performance on the three types of sentences in SerboCroatian-object marked, subject marked and unmarked. It is interesting that performance at the second age level shows a clear drop only for the object marked sentences, that is, those with clear inflectional cue for direct object. We suggest that these children have begun to realize that there is some relation between a marked accusative noun in a sentence and the possibility of reversed interpretation of the order of nouns in that sentence, but they are not yet clear on the interaction of word order and inflection in their language. Evidence from performance on a word order comprehension test (Slobin and Bever, 1978) provides some support for this interpretation. As mentioned above, the same subjects were also asked to act out sentences made up of two nouns and a reversible verb (e.g., The horse kicks the camel). These sentences were presented in all six possible orders of subject, verb and object, and represented the same three pairings of feminine and neuter nouns as were used in the causative task. At the second age level, performance was significantly different from chance only on those sentences which provide 120 inflectional cue to either subject or object-that is, the sentences with two neuter nouns. It is as if the presence of any sort of inflection-subject or Table 6.

Word order task: Percentage type Sentence

type

Neuter-feminine (object marked) Feminine-neuter (subject marked) Neuter-neuter (unmarked)

correct response in Serbo-Croatian

by sentence

Age group A

B

C

D

71 67 67

54 67 63

83 92 73

100 92 92

14

Mary Sue Ammon

and Dan I. Slobin

object, on either noun-leads to conflicting strategies, resulting in an overall picture of random performance. In the absence of marked inflections, a standard subject-object word order strategy is followed. By the third age level, inflectional cues aid performance, even making it possible for children to identify an initial object-marked noun as the patient of action. By this age, then, word order and inflectional strategies can operate separately and in interaction with one another. Note in Table 6 that at the third age level it is the inflectionally marked causative sentences which show the greatest increase in correct performance, with the object-marked sentences reaching 100% by the fourth age. This suggests that after the second age level children have become much more proficient at using inflections to aid sentence processing. Table 7 breaks down error types by age for Serbo-Croatian. It is striking that for the first two age groups, all types of errors are distributed across all three sentence types. This suggests a general confusion in the application of sentence processing strategies to the causative sentences. At the third age level, however, although the numbers are small, a pattern seems to be present. Most of the errors, as pointed out above, are due to the unmarked, neuterneuter sentences; and here the errors are all of the reversal type-that is, taking the second noun to be instigator. These are the only sentences with no inflectional cues, and it may well be that by this age children rely on such cues above word order in sentence processing. Left with two unmarked nouns in memory, the Serbo-Croatian child may be less clear about the grammatical relations between these two nouns than a child speaking a strict word-order language. That is to say, when short-term memory difficulties happen to leave a Serbo-Croatian child with only word-order information, lacking the cusTable 7.

Causative comprehension and age in Serbo-Croatian Sentence

task: Frequency

type

of error types by sentence

type

Age group A

B

C

D

Neuter-feminine (object marked)

reversal only one animal acts both animals act similarly

1 5 1

3 8 0

2 1 0

0 0 0

Feminine-neuter (subject marked)

reversal only one animal acts both animals act similarly

5 3 0

3 3 1

1 0 1

1 1 0

Neuter-neuter (unmarked)

reversal only one animal acts both animals act similarly

1 6 1

4 5 0

I 0 0

1 1 0

Processing ofcausative sentences

15

tomary local cues, s/he may have special difficulty in processing word-order cues alone. However, such reversal errors do include both animals in a causative interaction, reflecting Serbo-Croatian children’s generally high level of control of this construction. By contrast, English and Italian errors tend to be non-causative, simply having one animal (usually the second) act alone, reflecting a lower level of comprehension of this structure. The other significant age-by-language contrast involves the contrast between the Serbo-Croatian and Italian quadratic trends. Part of this is due to the lowered performance of the Serbo-Croatian children at the second age level just discussed, but part is due to the lack of continued development by the Italian children at the last two ages after a steep early rise in performance. In fact, the Italian and Serbo-Croatian curves move in opposite directions. It is not the case that only a few children are committing most of the errors at these later stages in Italian. At the oldest age level, 7/l 1 of the Italian children made one or more errors. The nature of the curve in the Italian data may be due to several factors. Some of these may explain why the English growth curve does not have the same nonlinearity as the Italian curve. If a trend analysis is performed on each language separately, the Italian curve has both a significant linear and quadratic component, and the English curve has only a significant linear component. In both languages, the causative is expressed syntactically, requiring that the whole sentence pattern be attended to, as we noted above. Consequently, performance at the first age level is quite poor, probably as a result of processing limitations. However, the word-order properties of the causative in Italian-Noun, Compound Verb, Noun-may make the nature of the action (i.e., who is the agent of the causative action) more unambiguous than in English, where the Noun, Verb, Noun, Verb order may be more confusing. The continued errors at later ages by both word-order languages may be due to two different factors. (As in Italian, many English-speaking children-9/1 2-continue to make errors at the oldest age level). The form of the compound verb in Italian is homonymous with one form of the imperative. Thus older Italian children may be hearing the sentences as, for example, The horse. Make run the camel. In fact, most of the errors in the Italian children at these later ages involved either (1) having the second-named participant perform the action, or (2) touching the first-named participant and then having the second act. Considering the nature of the task-performing actions according to specifications of the examiner-hearing the sentences as imperatives seems not to have been a bad guess on the part of these children. While errors of having only the second animal act were also made by older Englishspeaking children, they did not include the component of touching the firstnamed participant. But it is evident that the particular word-order configura-

16

Mary Sue Ammon and Dan I. Slobin

tions of the causative constructions pose problems to both Italian and American children throughout the age range tested. In this regard it should be noted that neither of these groups reach the ceiling of near-perfect performance evidenced by the Yugoslav and Turkish children. In summary, the results of our causative comprehension study suggest that sentence processing is especially aided by what we have called local cuesthat is, surface markings which identify the roles of particular words, such as case inflections on nouns and particles affixed to or immediately adjacent to verbs. More difficult for processing are constructions which require that one hold an entire sentence pattern in memory in order to determine semantic relations, such as is required for the application of word-order rules. The acquisition of a system of local cues can, at points, be retarded in languages where both word order and inflectional cues operate to express the same grammatical relations, as in Serbo-Croatian. In this case, two types of strategies must be kept available. However, the existence of local cues still facilitates Serbo-Croatian sentence processing overall.

References A. (1978) The acquisition of causal connectives in Turkish. Papers and Reports on Child Language Developnrent (Stanford University, Department of Linguistics), 15, 1299139. Aksu, A. (1975) Development of the expression of cause-effect relations in Turkish. Unpublished paper, Department of Psychology, University of California, Berkeley. Baron, N. S. (1977) Language Acquisition and Historical Change. Amsterdam, North-Holland. Bowerman, M. (1974) Learning the structure of causative verbs: A study in the relationship of cognitive, semantic, and syntactic development. Papers and Reports on Child Language Development (Stanford University, Department of Linguistics), 8, 142-178. Bowerman, M. (1977) The acquisition of rules governing “possible lexical items”: Evidence from spontaneous speech errors. Papers and Reports on Child Language Development (Stanford University, Department of Linguistics), 13, 148-156. Clancy, P., Jacobsen, T. and Silva, M. (1976) The acquisition of conjunction: A cross-linguistic study. Papers and Reports on Child Language Development (Stanford University, Department of Linguistics), 12, 71-80. Johnston, J. R. and Slobin, D. I. (1977) The development of locative expressions in IQrglish, Italian, ScrboCroatian and Turkish. Papers and Reports on Child Language Development (Stanford University, Department of Linguistics), 13, 134-147. (Revised version in press.J. Child Lang./. Radulovic, L. (1975) Acquisition of language: Studies of Dubrovnic children. Unpublished doctoral dissertation, Department of Lducational Psychology, University of California, Berkeley. Slobin, D. 1. (1978) Universal and particular in language acquisition. Paper presented at Conference on “Language Acquisition: The State of the Art”; University of Pennsylvania, Philadelphia. (To be published in proceedings, edited by Lila Gleitman and Eric Wanner). Slobin, D. I. and Bever, T. G. (1978) Cross-linguistic acquisition of canonical arrangements of word order and inflections. Unpublished paper. Tahny, L. (1976) Semantic causative types. In M. Shibatani (ed.), Syntax and Senlantics. Vol. 6: The Grarnmzr of Causative Constructions. New York, Academic Press. Aksu,

Processing of causative sentences

17

Cctte Etude pork sur la comprchcnsion de phrases exprimant un cffet dfi a unc incitation (par excmple le chcval fait courir le chamcau). Les recherchcs ont iit6 mcndcs aupres d’enfants agds dc 2 ans a 4 arts parlant anglais, italien, serbo-create ou turc. Des differences ont cte obtcnues scion la langue parlee qui montrcnt que Its performances des cnfants pour cc type dc construction dcpendcnt dc factcurs morphologiques (particule causative, flexions indiquant le cas) et de mekanismes syntaxiques (periphrase, ordre dcs mots) Its indicts locaux tcls Its suffixes flexionncls ct lcs particules causativcs spccifiques favorisent le developpement plus rapide des strategies de calcul chez les enfants Serbo-croates et Turcs. Lcs systcmes de I’anglais et de I’italicn fond& sur I’ordre des mots requicrt de I’auditeur qu’il garde i l’esprit lc pattern de la phrase entiere pour en determiner les relations syntaxiques sous-jacentes; cela contribue $ ralentir le dcveloppement des enfants dans cette tache.

Cognilion, 7 (1979) @Elsevier

Sequoia

19-33 S.A., Lausanne

2 - Printed

in the Netherlands

The cognitive and neurological basis of developmental dyslexia: A theoretical framework and review ANTHONY

F. JORM

Deakin University, Australia *

Abstract It is argued that developmental dyslexics have difficulty in accessing the meaning of written words via phonological recoding, although they can successfully access meaning by a direct visual route. This difficulty with phonological recording is explained in terms of a short-term memory deficit. It is suggested that developmental dyslexia is a genetically-based dysfunction of the inferior parietal lobule and evidence is reviewed that this region is important in both reading and short-term memory. Implications for remedial instruction are discussed.

At present there is considerable disagreement concerning the most appropriate nomenclature and classification for children with reading problems. In this article, I will use the term dyslexia to describe children “who, despite fail to attain the language skills of conventional classroom experience, reading, writing, and spelling commensurate with their intellectual abilities (Critchley, 1970, p. 11)“. In certain sections of the article I use the term developmental dyslexia to distinguish this disability from acquired dyslexia in brain-damaged adults. Although the authors of many of the studies cited in this article use alternative terms such as reading disability, learning disability, and reading retardation, in each case they appear to be researching a similar group of children. Typically, these researchers have used the strategy of comparing the performance of a group of children low in reading achievement with a group of normal readers of equivalent IQ and schooling. Although the concept of dyslexia adopted here is a broad one, 1 feel that any more detailed characterisation of the disorder would beg the question which this article attempts to answer. Undoubtedly, a more refined definition of the disorder will come with greater theoretical understanding. Over the years, numerous theories of dyslexia have been proposed. Usually these theories have attempted to account for only one particular facet of *Requests of Education,

for reprints should Deakin University,

be sent to: A. F. Jorm, Cognitive Psychology Research P.O. Box 267, Belmont, Victoria, 3216, Australia.

Group,

School

20

Anthony

I;. Jorm

dyslexia such as the nature of the reading difficulty (Firth, 1972), the cognitive deficit underlying the reading difficulty (Birch & Belmont, 1964; Corkin, 1974; Richardson, 1974; Stanley, 1975; Vellutino, 1977), or the neurological basis of the disorder (Delacato, 1966; Satz & Sparrow, 1970). Ideally, however, a theory of developmental dyslexia should account for the disorder at all these levels of investigation. The purpose of this article is to put forward a theory which can integrate much of the research evidence at each of these levels.

The Nature of the Reading Difficulties

in Dyslexia

Wernicke (1874; 1966) long ago suggested that readers can access the meaning of a written word by either of two routes: (a) a visual route in which the written form of a word is used to directly retrieve the meaning from memory, and (b) a phonological route in which the written word is converted to a phonological representation and this is then used to retrieve the meaning. Bradshaw (1975) and Coltheart (in press) have reviewed modern research which suggests that readers can, in fact, access word meanings by either of these routes. I will use this dual access theory as a framework for considering the nature of the reading deficit in dyslexia. Firth (1972) carried out what is probably the most comprehensive investigation into the components of the reading process which are affected in dyslexia. He devised a large battery of tests which he felt tapped various component skills in the reading process and administered these tests to groups of dyslexics and normals. Firth found that the following tests best discriminated the two groups: (1) Nonsense word test. In this test the child had to generate pronunciations for nonsense words like nate and iston. (2) Auditory blending test. In this test the child had to blend elements like a - t and c - all into words. (3) Auditory analysis test. In this test the child had to listen to a word and tell what would be left if a certain sound (e.g., the t in stand) were removed. Firth concluded from his research that the children were deficient in phonics skills, that is the ability to carry out a phonological recoding on a written word. Despite their difficulties with phonological recoding of written words, dyslexics seem to be quite capable of accessing meaning directly from the visual form of a word. Supporting evidence comes from Rozin, Poritsky and Sotsky’s (197 1) demonstration that children who have difficulty generating

Cognitive and neurological basis of developmental dyslexia

21

pronunciations for written nonsense words can easily learn the meanings of Chinese characters (which do not represent phonological information). Similarly, Firth (1972) and Jorm (1977a) have reported that dyslexics perform as well as normal readers when they are required to directly associate meanings with word-like visual forms. Contrary to the view taken here, Boder (197 1) has suggested that there are distinct subgroups of dyslexics having difficulties in either the visual reading route (dyseidetic dyslexics) or the phonological reading route (dysphonetic dyslexics). However, Boder’s subgroups cannot be taken seriously since Firth’s (1972) study found no children with an inability to learn symbolmeaning associations (i.e., dyseidetic dyslexics) and Holmes and Peper (1977) were unable to identify these subgroups on the basis of an analysis of spelling errors. It therefore seems that developmental dyslexics have a specific impairment of the ability to access meanings via the phonological route. So far I have only considered reading difficulties involving the processing of single words. However, research by Guthrie (1973) shows that the reading deficit in dyslexia involves comprehension as well. In his study, Guthrie matched older dyslexics with younger good readers for IQ and ability to read individual words. By using a task which required the reader to select the missing words from a passage, he was able to show that the dyslexics had difficulty in comprehending passages even though they were as capable of reading the individual words in the passage as controls. So it seems that dyslexics have difficulty in integrating the information they receive from individual words as well as in the phonological recoding of individual words.

The Cognitive Deficit in Dyslexia The dyslexic’s difficulties in the phonological recoding of written words and in the comprehension of text must be due to some deficiency in the basic cognitive abilities which are crucial to these processes. Contrary to what has often been argued in the past (e.g., Young and Lindsley, 1970) dyslexia does not involve any defect of visual perception. For example, Morrison, Giordani and Nagy (1977) found that dyslexic children could report visual information presented less than 300 msec previously, as well as could controls, although they were poor at reporting the information after longer delays. Morrison et al. (1977) took the view that reports of information presented less than 300 msec previously reflect perceptual registration and iconic storage, while reports of information after 300 msec reflect memory processes. They therefore concluded that dyslexics have normal

22

Arlthony

I? Jot-m

perceptual registration and iconic storage, but some sort of memory deficit. Further evidence against the perceptual deficit hypothesis comes from a study by Vellutino, Steger and Kandel (1972) which used a task requiring children to reproduce single designs, numbers, or letters which had been presented immediately before. Such a task involves a very small memory load and is therefore sensitive to perceptual deficits. However, dyslexics were found to perform normally on this task. Similary, Katz and Wicklund (1972) were able to show that dyslexics performed as well as normals on a task requiring them to visually scan a row of letters for the presence or absence of a predetermined target letter. However, when the research on memory performance in dyslexics is examined, they seem to show impaired performance on some memory tasks but normal performance on others. I will attempt to account for these findings in terms of a currently acceptable view of memory. The view taken here is that memory performance can be best accounted for by positing various separate but interrelated memory stores. Memorystores approaches have been criticized by Craik and Lockhart (1972), so some justification is required for adopting one here. The main virtue of approaches which posit separate memory stores is that they seem to be most compatible with the current neuropsychological evidence. Warrington (1971) and Warrington and Weiskrantz (1973) have reviewed the evidence collected using brain-damaged patients with memory deficits. This evidence indicates that lesions to particular regions of the brain can result in impairment of certain types of memory performance while leaving other aspects of memory performance intact. The reviews of Warrington (197 1) and Warrington and Weiskrantz ( 1973) suggest that it is reasonable to posit short-term stores which are respectively specialized for auditory-verbal and visual information as well as a long-term store which is specialized for storing information in a semantic form. Contrary to some memory-stores models (e.g., Atkinson and Shiffrin, 197 1), incoming information is able to be directly processed by the long-term store without necessarily passing through a short-term store. There is also considerable psychological evidence which is consistent with the notion of such stores (see Baddeley, 1976, for a review). If the research on memory performance in dyslexic children is interpreted in terms of this view of memory, it seems clear that they have a deficit of both auditory-verbal and visual short-term storage, as indicated by their difficulty in retaining both types of material with fast presentation rates. Several studies provide evidence consistent with the view that dyslexics are deficient in auditory-verbal short-term storage. For instance, dyslexics tend to perform poorly on the Digit Span subtest of the Weschler Intelligence

Cognitive and neurological basis of developmental

dyslexia

23

Scale for Children (Rugel, 1974) and the Auditory Sequential Memory subtest of the Illinois Test of Psycholinguistic Abilities (Stanley, 1975). Similarly, other researchers have shown dyslexics to be worse than normal readers at retaining auditorily-presented digits (Corkin, 1974; Spring 1976) and letters (Bakker, 1972; Jorm, 1977b). Contrary evidence comes from Perfetti and colleagues (Perfetti and Goldman, 1976; Perfetti and Lesgold, 1977) who report the results of an experiment showing that a group of retarded readers did not differ from a group of IQ-matched good readers on an auditory digit memory task. However, the two groups did differ on a task requiring memory for words in a passage. This discrepant result may be due to the comparatively small difference in reading ability between the two groups. Nevertheless, the findings are interesting in that they indicate that memory for words in a sentence is a better correlate of reading ability than digit memory. This may be because memory for the wording of discourse is solely dependent on the auditory-verbal short-term store, whereas digit span is only partly dependent on this store (Baddeley and Hitch, 1974). A related finding comes from Waller (1976) who showed that dyslexics could remember the semantic content of sentences as well as normal readers, but were worse at remembering aspects of the specific wording of the sentences. Further evidence for a deficit in auditory-verbal short-term storage comes from the finding that dyslexics make fewer phonological confusions than normal readers in the short-term retention of words (Mark, Shankweiler, Liberman and Fowler, 1977) and letters (Liberman, Shankweiler, Liberman, Fowler and Fischer, 1977). Such a lack of phonological confusions suggests that dyslexics are not relying on the auditory-verbal short-term store to the same degree as normal readers. A deficit in the auditory-verbal short-term store may also be responsible for the frequent temporal order errors noted in the short-term memory of dyslexics (Bakker, 1972; Corkin, 1974; Senf, 1969; Senf and Feshback, 1970). A similar short-term memory deficit is apparent with visually-presented items. This short-term memory deficit is evident not only with items such as visually-presented digits (Spring and Caps, 1974), letters (Morrison et al., 1977; Stanley and Hall, 1973) and simple pictures (Cummings and Faw, 1976; Morrison et al., 1977) which can be encoded by the subjects in either a visual or auditory form, but also with abstract forms which are not so readily translated into an auditory-verbal code (Noelker and Schumsky, 1973; Morrison et al., 1977; Stanley, 1970). The poor short-term memory performance with abstract forms supports the notion that dyslexics also have a deficiency of the visual short-term store. Just as the deficiency of the auditory-

24

Anthony

F. Jorm

verbal short-term store may be responsible for temporal-order errors in memory, a deficiency of the visual short-term store could be associated with spatial-order errors in the short-term memory of dyslexics (Noelker and Schumsky, 1973; Morrison et al., 1977). Recently, Torgeson (1977) and Torgeson and Goldman (1977) have argued against the notion that the short-term memory deficit of dyslexic children is due to a deficiency in memory storage capacity and have explained it in terms of a lack of inclination or ability to use efficient task strategies. As this view is contrary to the one expressed in this article, their work merits some discussion. Torgeson (1977) and Torgeson and Goldman (1977) offered two findings to support their view. Firstly, they showed that dyslexics were less likely to spontaneously use verbal rehearsal strategies while studying pictorial stimuli than normal readers and, secondly, they were able to demonstrate that the difference between dyslexics and normal readers was reduced if both groups were instructed to use mnemonic strategies such as giving verbal labels to the pictures and encoding them into semantically-meaningful categories. These findings can be readily interpreted within the theoretical framework proposed here. The first finding, that dyslexics are less likely to use a cumulative rehearsal strategy is probably simply a reflection of their auditory-verbal shortterm storage deficit, since it is this storage system that seems to be responsible for rehearsal of verbal material (see Baddeley, 1976). The second finding, that dyslexic short-term memory performance can be improved, may be due to the fact that the strategies in which the children were trained are such as to be likely to promote direct long-term storage, and, as the evidence reviewed below shows, dyslexics show normal long-term storage. Studies comparing paired-associate learning in dyslexics and normals indicate that long-term storage is not deficient in dyslexia. Dyslexics have been shown to perform as well as normal readers when required to associate words with words (Budoff and Quinlan, 1964; Weinstein and Rabinovitch, 1971), visual symbols with visual symbols (Vellutino, Harding, Phillips and Steger, 1975; Vellutino, Steger and Pruzek, 1973) words with visual symbols (Firth, 1972; Jorm, 1977a), and non-verbal sounds (e.g., hisses and clicks) with visual symbols (Vellutino, Harding, Phillips and Steger, 1975; Vellutino et al., 1973). Waller’s (1976) finding that dyslexics are as good as normal readers at retaining the semantic content of sentences also supports the view that their long-term storage is not deficient. However, dyslexics perform poorly at associating nonsense words with visual symbols (Gascon and Goodglass, 1970; Otto, 196 1; Vellutino, Harding, Phillips and Steger, 1975; Vellutino, Steger, Harding and Phillips, 1975). It thus seems that the semantic richness of the material is a crucial determinant

Cognitive and neurological basis of developmental

dyslexia

25

of the ease with which dyslexics can encode information in long-term storage. Perhaps the reason that dyslexics have such difficulty in the paired-associate learning of semantically-meaningless material such as nonsense words is that it must be encoded temporarily in short-term storage to give the long-term store time to encode it semantically. Because of their deficiencies in shortterm storage, dyslexics would have greater difficulty with such material. However, semantically-meaningful material can be directly encoded into long-term storage. Support for this view comes from Vellutino, Steger, Harding and Phillips (1975) finding that, in learning to associate nonsense words with visual symbols, dyslexics tend to give real words as errors whereas normal readers tend to give phonemically-similar nonsense words as errors. Thus, it appears that dyslexics were attempting to carry out a direct semantic encoding of the nonsense words (a specialty of long-term storage), whereas the normal readers are attempting a phonemic encoding (a specialty of the auditory-verbal shortterm store). Further supporting evidence comes from Otto’s (1961) study which showed that although dyslexics are slower at initial learning of symbolnonsense syllable paired associates, they are as fast as normal readers at relearning the pairs after a 24 hour delay. Although the nonsense syllables would have been meaningless to the children on initial learning, this would not have been the case on relearning as they would already be quite familiar. In summary, the evidence seems to be consistent with the notion that dyslexics have an impairment of auditory-verbal and visual short-term storage, but their long-term store seems to operate normally. How then does such a short-term storage deficit produce problems in the phonological recoding of written words and in reading comprehension? It seems likely that individual differences in short-term memory storage could produce differences in both these aspects of reading performance. For instance, when a child recodes a written word into a phonological representation he must be able to remember the phonemes which result from applying each grapheme-phoneme correspondence rule and the order in which the phonemes are to be output. He must also remember which letters of the word he has already analysed so that he will avoid going over letters which have already been accounted for, or omitting strings of letters which have not. Similarly, a deficiency of auditory-verbal short-term storage could affect reading comprehension. Shallice and Warrington (1970) and Shallice (1975) have argued that this store plays an important role in the comprehension of speech, while Kleiman (1975) and Levy (1975, 1977) have presented evidence to show that speech recoding is important to reading comprehension. It is also possible that the visual short-term store has a role in holding information about the reader’s current position in the text.

26

Anthony

F. Jot-m

The Neurological

Basis of Dyslexia

A fully adequate theory of the neurological basis of dyslexia should be able to show that: (a) the neurological system hypothesised to be involved in the disorder is crucial to those components of the reading process which are affected in dyslexia, (b) this neurological system is also crucial to the fundamental cognitive process which is deficient in dyslexia, and (c) this neurological system does not function normally in dyslexics. I will show that the theory advanced in this article fulfills all of these criteria. The position taken here is that the inferior parietal lobule is not fully developed in dyslexics. This view is similar to the very first theory of developmental dyslexia which was proposed by Morgan in 1896. Claiborne (1906), Geschwind (1965) and Hermann (1959) have also suggested the parietal region is involved in dyslexia. There are a number of points of evidence suggestive of a left inferior parietal lobule involvement in developmental dyslexia. Firstly, Dejerine (189 1, 1892) described an acquired reading disorder, alexia with agraphia, which resulted from a lesion to the angular gyrus of the left inferior parietal lobule. Modern neurological opinion confirms that alexia with agraphia results from lesions to this region (Benson and Geschwind, 1969). On the basis of a study of the oral reading ability of patients with this disorder, Gardner and Zurif (1975) concluded that “the alexic with agraphia has lost the ability to proceed from the visual grapheme to its corresponding individual sounds but retains the capacity to proceed from a familiar visual configuration to its meaning (p. 189)“. Marshall and Newcombe (1973) have applied the term deep dyslexia to describe acquired reading disorders which involve an impairment of the phonological reading route, while Shallice and Warrington (1975) use the term phonemic dyslexia. Therefore, all three terms can be used synonymously to describe an acquired deficit of the ability to carry out a phonological recoding of written words, but with the visual reading route still intact. As I have pointed out elsewhere (Jorm, 1977a), this form of acquired dyslexia has certain functional similarities to developmental dyslexia. Like developmental dyslexics, acquired dyslexics have great difficulty reading simple nonsense words (Firth, 1972; Patterson and Marcel, 1977; Saffran and Marin, 1977). Also, in both disorders high-imagery words are easy to read and low-imagery words are difficult (Jorm, 1977a; Richardson, 1975a, 1975b; Shallice and Warrington, 1975), and there is a common tendency to make visual errors in reading (Jorm, 1977a; Shallice and Warrington, 1975). Although these acquired dyslexics often seem to have a deficiency in speech comprehension (as measured by De Renzi and Vignolo’s, 1962, Token Test) (Patterson and Marcel, 1977; Saffran and Marin, 1977; Warrington, Logue and Pratt, 197 1), there has as yet been no systematic investigation

Cognitive and neurological basis of developmental dyslexia

27

of their reading comprehension which is comparable to Guthrie’s (1973) study with developmental dyslexics. A further argument in favour of the idea of a left inferior parietal lobule dysfunction comes from research indicating that this region has an important involvement in short-term memory function. On the basis of a study of 3 patients with an auditory-verbal short-term storage deficit, Warrington et al. (197 1) have suggested that lesions to the left inferior parietal lobule are responsible. A similar patient described by Saffran and Marin (1975) also had a lesion in this region. Similarly, Warrington and Rabin (1971) found that disorders of visual short-term storage were associated with left posterior cerebral lesions. Although acquired dyslexia and short-term memory deficiencies are thought to arise from lesions to the same region, there has been little attempt in the literature to explore the functional relationship between the two disorders. Warrington, Logue and Pratt (197 1) reported on the reading performance of their three patients with auditory-verbal short-term storage deficits. One of these patients had a severe reading deficit and has also featured in studies on acquired dyslexia (Shallice and Warrington, 1975), but another of these patients had a reading ability within the normal range. However, we do not know to what extent this patient was relying on her ability to directly access the meaning of written words without phonological recoding. As Wernicke (1874, 1966) pointed out, a lesion which produces a severe reading deficit in a poorly educated reader may have little apparent effect on a highly practiced reader because of the latter’s ability to read without phonological recoding. Similarly, although there has been no systematic investigation into the relationship between visual short-term storage deficits and reading ability, Warrington and Rabin (197 1) say that reading difficulties are a typical feature of these patients. Many studies on acquired dyslexia make no report of the short-term memory performance of their patients. Exceptions are Shallice and Warrington’s (1975) patient, who has also featured in studies of auditory-verbal short-term memory deficit and one of Marshall and Newcombe’s (1973) two patients, who is reported as having a severe deficit of digit- and word-span. Other studies of acquired dyslexic patients make no mention of their memory performance (Gardner and Zurif, 1975; Patterson and Marcel, 1977; Saffran and Mar-in, 1977; Saffran, Schwartz and Marin, 1976). Given that the left inferior parietal lobule may be crucially involved in both reading and short-term memory, what evidence is there to suggest that dyslexics have a dysfunction of this region? There are four studies of visual evoked potentials in dyslexics which are consistent with this notion (Connors,

28

Anthony

F. Jorrn

1970; Preston, Guthrie and Childs, 1974; Preston, Guthrie, Kirsch, Gertman and Childs, 1977; Symann-Lovett, Fascon, Matsumiya and Lombroso, 1977). In each of these studies, electrodes were placed above the occipital and parietal regions of both hemispheres and it was found that only the response recorded from above the left parietal region differentiated the dyslexics from the normal readers. Furthermore, studies of the development of the human brain have shown that the inferior parietal lobules develop slowly in cortical size (Blinkov and Glezer, 1968), dendrite formation (Bonin and Bailey, 196 1) and myelination (Bailey and Bonin, 195 1; Kaes, 1907; Yakovlev, 1962; Yakovlev and Lecours, 1967). In fact, the evidence reviewed by Lecours (1975) on the development of myelination suggests that the inferior parietal lobules continue to mature well into adolescence if not later. Since the inferior parietal lobules are still developing throughout the school years, it seems possible that there could be individual differences in the rate of development of the regions which could give rise to differences in reading achievement. It is possible that genetic factors could give rise to individual differences in neurological development which then affect reading acquisition. There is evidence from twin studies showing a higher concordance rate for dyslexia in monozygotic twins than in dizygotic twins (Bakwin, 1973; Hallgren, 1950; Hermann, 1959) (although the criteria for defining dyslexia in Bakwin’s (1973) study were quite inadequate). The results of family pedigree studies are also consistent with the notion of a genetic influence (Finucci, Guthrie, Childs, Abbey and Childs, 1976), but unfortunately are also amenable to an explanation in terms of cultural transmission. Lesions to the left inferior parietal lobule produce other cognitive deficits besides those in short-term memory and reading. Therefore, the theory that dyslexia results from a genetically-based dysfunction of this region predicts that similar cognitive deficits should be observable in dyslexic children. For instance, both patients with an auditory-verbal short-term memory deficit and acquired dyslexics seem to have a subtle defect in the ability to comprehend syntactically-complex sentences (Patterson and Marcel, 1977; Saffran and Marin, 1975, 1977; Warrington et al., 1971). The possibility of such a deficit in dyslexic children has yet to be investigated but, according to the theory I have proposed, should occur. Similar predictions will be possible as more becomes known about the function of the left inferior parietal lobule.

Some Remedial

Implications

The theory proposed here has some implication for remediation. It suggests that dyslexics are as capable as normal readers of learning to access the mean-

Cognitive and neurological basis of developmental dyslexia

29

ing of a printed word visually, but they will have great difficulty in accessing meaning via phonological recoding. Therefore, methods of instruction which capitalise on this intact ability (e.g., look-say method) may be more appropriate than methods which try to teach grapheme-phoneme correspondence rules (phonics methods), even though the latter methods seem to be more effective for the general population of children (Chall, 1967). The reason phonics methods are more effective for the majority of children may be that they allow the child to access the meanings of novel words without the aid of a teacher. In a sense, grapheme-phoneme correspondence rules act as a built-in teacher because, after they have been used to read a novel word on several occasions, the child is able to access the meaning of the word directly from print. If this view is correct, the dyslexic child is poor at reading because he cannot manage to read novel words without the help of a teacher and therefore gets less experience at associating the visual forms of words directly with their meanings. So dyslexics may require a more individualised mode of instruction than normal children in order to have the frequent exposures to the associations between print and meaning which they require to use their intact skills effectively.

References Atkinson, R. C. and Shiffrin, R. M. (1971) The control of short-term memory. Sci. Amer., 225, 82-89. Baddeley, A. D. (1976) The Psychology ofMemory. New York, Basic Books. Baddeley, A. D. and Hitch, G. (1974) Working memory. In G. A. Bower (ed.), The Psychology of Learning and Motivation, (Vol. 8) New York, Academic Press. Bailey, P. and Bonin, G. V. (1951) The Isocortex of Man, Urbana, University of Illinois Press. Bakker, D. J. (1972) Temporal Order in Disturbed Reading, Rotterdam, Rotterdam University Press. Bakwin, H. (1973) Reading disability in twins. Dev. Med. child Neural. 15, 184-187. Benson, D. F. and Geschwind, N. (1969) The alexias. In P. J. Vinken and G. W. Bruyn (eds.), Handbook of Clinical Neurology (Vol. 4), Amsterdam, North Holland Publishing Company. Birch, H. G. and Belmont, L. (1964) Auditory-visual integration in normal and retarded readers. Am. J. Orthopsychiat., 34, 852-862. Blinkov, S. M. and Glezer, I. I. (1968) The Human Brain in Figures and Tables, New York, Plenum Press. Boder, E. (1971), Developmental dyslexia: Prevailing diagnostic concepts and a new diagnostic approach. In H. R. Myklebust (ed.), Progress in Learning Disabilities, (Vol. 2), New York, Grune and Stratton. Bonin, G. V. and Bailey, P. (1961) Pattern of the cerebral isocortex. Primatologziz, 2, (10). Bradshaw, J. L. (1975) Three interrelated problems in reading: A review, Mem. Cog., 3, 123-134. Budoff, M. and Quinlan, D. (1964) Reading progress as related to efficiency of visual and aural learning in the primary grades. J. educ. Psychol., 55, 247-252. Chall, J. (1967) Learning to Read: The Great Debate, New York, McGraw-Hill.

30

Anthony

F. Jorm

Claiborne, J. H. (1906) Types of congential symbol amblyopia. J. amer. Med. Ass., 47, 1813-1816. Coltheart, M. (in press) Lexical access in simple reading tasks. In G. Underwood (ed.), Strategies of Information Processing, New York, Academic Press. Connors, C. K. (1970) Cortical visual evoked response in children with learning disorders, Psychophysiol.. 7, 418-428. Corkin, S. (1974) Serial-ordering deficits in inferior readers. Neuropsychol. 12, 347-354. Craik, F. I. M. and Lockhart, R. S. (1972) Levels of processing: A framework for memory research. J. verb. Learn. verb. Behav., 11, 671-684. Critchley, M. (1970) The Dyslexic Child. London, Heinemann. Cummings, E. M. and Faw, T. T. (1976) Short-term memory and equivalence judgements in normal and retarded readers. Child Duel., 47, 286-289. Dcjerine, J. J. (1891) Sur un cas de c&cite verbale avec agraphie, suivi d’autopsie. Sot. Biol. C. R. Hebd. Games et Mimoires, 3, 197-201. Dejerine, J. J. (1892) Contribution $ I’dtude Anatomo-pathologique ct clinique des diff&entes variktcs de c&citt! verbale. Sot. Biol. C. R. Hebd. Skances et MPmoires, 4, 6 l-90. Delcato, C. H. (1966) Neurological Organization and Reading. Springfield, Charles C. Thomas. De Renzi, E. and Vignolo, L. A. (1962) The token test: A sensitive test to detect receptive disturbances in aphasics. Brain, 85, 665-678. Finucci, J. M., Cuthrie, J. T., Childs, A. L., Abbey, H. and Childs, B. (1976) The genetics of specific reading disability. Ann. hum. Genet., Lond., 40, l-23. Firth, I. (1972) Components of Reading Disability. Unpublished doctoral dissertation, University of New South Wales. Gardner, H. and Zurif, E. (1975) Bee but not be: Oral reading of single words in aphasia and alexia. Neuropsychol., 13, 181-190. Gascon, G. and Goodglass, H. (1970) Reading retardation and the information content of stimuli in paired associate learning. Cortex, 6, 417-429. Geschwind, N. (1965) Disconnexion syndromes in animals and man. Brain, 88, 237-294. Guthrie, J. T. (1973) Reading comprehension and syntactic responses in good and poor readers. J. educ. Psycho/., 65, 294-299. Hallgren, B. (1950) Specific dyslexia. Acta Psych. Neur., 65, l-287. Hermann, K. (1959) Reading Disability. Copenhagen, Munkgaard. Holmes, D. L. and Peper, R. J. (1977) An evaluation of the use of spelling error analysis in the diagnosis of reading disabilities. Child Devel., 48, 1708-l 711. Jorm, A. F. (1977) Effect of word imagery on reading performance as a function of reader ability. J. educ. Psychol., 69, 46-54. (a). Jorm, A. F. (1977) Parietal lobe function in developmentaldyslexia.Neuropsychol., 15, 841-844. (b). Kaes, T. (1907) Die Grosshirnrinde des Menschen in ihrem Massen und in ihrem Fasengehalt, Jena, Fischer. Katz, L. and Wicklund, D. T. (1972) Letter scanning rate for good and poor readers in grades two and six. J. educ. Psychol., 63, 363-367. Kleiman, G. M. (1975) Speech recoding in reading. J. verb. Learn. verb. Behav., 14, 323-339. Lecours, A. R. (1975) Myelogenetic correlates of the development of speech and language. In E. H. Lenneberg and C. Lenneberg (eds.), Foundations of Language Development: A Multidisciplinary Approach. (Vol. l), New York, Academic Press. Levy, B. A. (1975) Vocalization and suppression effects in sentence memory. J. verb. Learn. verb. Behav., 14, 304-316. Levy, B. A. (1977) Reading: Speech and meaning processes. J. verb. Learn. verb. Behav., 16, 623-638. Liberman, I. Y., Shankweiler, D., Liberman, A. M., Fowler, C. and Fischer, F. W. (1977) Phonetic segmentation and recoding in the beginning reader. In A. S. Reber and D. Scarborough (eds.),

Cognitive and neurological basis of developmental

dyslexia

31

Toward a Psychology of Reading: The Proceedings of the CUNY Conferences. Hillsdale, Lawrence Erlbaum. Mark, L. S., Shankweiler, D., Liberman, I. Y. and Fowler, C. A. (1977) Phonetic recoding and reading difficulty in beginning readers. Mem. Cog., 5, 623-629. Marshall, J. C. and Newcombe, F. (1973) Patterns of paralexia: A psycholinguistic approach. J. ps.Ycholinguist. Res.. 2, 715-199. Morgan, W. P. (1896) A. case of congenital word blindness. Br. Med. J., 2, 1378. Morrison, F. J., Giordani, B. and Nagy, 1. (1977) Reading disability: An information-processing analysis., Science, 196, 77-79. Noelker, R. W. and Schumsky, D. A. (1973) Memory for sequence, form, and position as related to the identification of reading retardates. J. educ. Psycho/., 64, 22-25. Otto, W. (1961) The acquisition and retention of paired associates by good, average and poor readers. J. educ. Psychol., 52, 241-248. Patterson, K. E. and Marcel, A. J. (1977) Aphasia, dyslexia and the phonological coding of written words. Quart. J. exp. Psychol., 29, 307-318. Perfetti, C. A. and Goldman, S. R. (1976) Discourse memory and reading comprehension skill. J. verb. Learn. verb. Behav., 14, 33-42. Perfetti, C. A. and Lesgold, A. M. (1977) Discourse comprehension and sources of individual differences. In M. A. Just and P. A. Carpenter (eds.), Cognitive Processes in Comprehension, New York, Wiley. Preston, M. S., Guthrie, J. T. and Childs, B. (1974). Visual evoked responses (VERs) in normal and disabled readers. Psychophysiol., 1 I, 452-457. Preston, M. S., Guthrie, J. T., Kirsch, I., Certman, D. and Childs, B. (1977) VERs in normal and disabled adult readers. Psychophysiol., 14, 8-14. Richardson, G. (1974) The Cartesian frame of reference: A structure unifying the description of dyslexia, J. psycholinguist. Res., 3, 15-63. Richardson, J. T. E. (1975) The effect of word imageability in acquired dyslexia. Neuropsychol., 13, 281-288 (a). Richardson, J. T. E. (1975) Further evidence on the effect of word imageability in dyslexia. Quart. J. exp. Psychol., 27, 445-449 (b). Rozin, P., Poritsky, S. and Sotsky, R. (1971) American children with reading problems can easily learn to read English represented by Chinese characters. Science, 171, 1264-1267. Rugel, R. P. (1974) WISC subtest scores of disabled readers: A review with respect to Bannatyne’s recategorisation. J. learn. Dis., I, 48-64. Saffran, E. M. and Marin, 0. S. M. (1975) Immediate memory for word lists and sentences in a patient with deficient auditory short-term memory. Brain Lang., 2, 420-433. Saffran, E. M. and Marin, 0. S. M. (1977) Reading without phonology: Evidence from aphasia. Quart. J. exp. Psychol., 29, 515-525. Saffran, E. M., Schwartz, M. F. and Marin, 0. S. M. (1976) Semantic mechanisms in paralexia. Brain Lang., 3, 255-265. Satz, P. and Sparrow, S. (1970) Specific developmental dyslexia: A theoretical reformulation. In D. J. Bakker and P. Satz (eds.), Specific Reading Disability: Advances in Theory and Method. Rotterdam, Rotterdam University Press. Scnf, G. M. (1969) Development of immediate memory for bisensory stimuli in normal children and children with learning disorders. Devel. Psychol. Mong., 1969, 1 (6), Part 2. Senf, G. M. and Feshback, S. (1970) Development of bisensory memory in culturally deprived, dyslexic, and normal readers. J. educ. Psychol., 61, 46 l-470. Shalllce, T. (1975) On the contents of primary memory. In P. M. A. Rabbitt and S. Dornic (eds.), Attention and Performance V. London, Academic Press.

32

Shallice,

Anthony

F. Jam

T. and Warrington, E. K. (1970) Independent functioning of verbal memory stores: A neuropsychological study. Quart. J. exp. Psychol., 22, 261-273. Shallice, T. and Warrington, E. K. (1975) Word recognition in a phonemic dyslexic patient. Quart. J. exp. Psychol., 27, 187-199. Spring, C. (1976) Encoding speed and memory span in dyslexic children. J. Sp. Educ., 10, 35540. Spring, C. and Capps, C. (1974) Encoding speed, rehearsal and probed recall of dyslexic boys. J. educ. Psychol., 66, 780-786. Stanley, G. (1975) Visual memory processes in dyslexia. In D. Deutsch and J. A. Deutsch (eds.), Short-term Memory. New York, Academic Press. Stanley, G. and Hall, R. (1973) A comparison of dyslexics and normals in recalling letter arrays after brief presentation. Brit. J. educ. Psychol., 43, 301-304. Symann-Lovett, N., Gascon, G. G., Matsumiya, Y. and Lombroso, C. T. (1977) Wave form difference in visual evoked responses between normal and reading disabled children. Neurol., 27, 156-159. Torgeson, J. K. (1977) Memorization processes in reading-disabled children. J. educ. PsychoZ.. 69, 571-578. Torgeson, J. and Goldman, T. (1977) Verbal rehearsal and short-term memory in reading-disabled children. Child Devel., 48, 56660. Vellutino, F. R. (1977) Alternative conceptualizations of dyslexia: Evidence in support of a verbaldeficit hypothesis. Harvard Educ. Rev., 47, 334-354. VeButino, 1:. R., Harding, C. .I., Phillips, F. and Steger, J. A. (1975) Differential transfer in poor and normal readers. J. genet. Psychol., 126, 3-18. Vellutino, F. R., Steger, J. A. and Kandel, G. (1972) Reading disability: An investigation of the perceptual deficit hypothesis. Cortex, 8, 1066118. Vellutino, 1‘. R., Steger, J. A. and Pruzck, R. M. (1973) Inter- vs intra-sensory deficit in paired associate learning in poor and normal readers. Canad. J. behau. Sci., 5, 11l- 123. Vellutino, F. R., Steger, J. A., Harding, C. J. and Phillips, F. (1975) Verbal vs non-verbal pairedassociates learning in poor and normal readers. Neuropsychol., 23, 75-82. Waller, T. G. (1976) Children’s recognition memory for written sentences: A comparison of good and poor readers. Child Devel., 47, 90-95. Warrington, E. K. (1971) Neurological disorders of memory. Br. Med. Bull., 27, 243-247. Warrington, E. K., Logue, V. and Pratt, R. T. C. (1971) The anatomical localisation of selective impairment of auditory verbal short-term memory. Neuropsychol., 9, 377-387. Warrington, E. K. and Rabin, P. (1971) Visual span of apprehension in patients with unilateral cerebral lesions. Quart. J. exp. Psychol., 23, 423-431. Warrington, E. K. and Weiskrantz, L. (1973) An analysis of short-term and long-term memory defects in man. In J. A. Dcutsch (ed.), The Physiological Basis of Memory, New York Academic Press. Weinstein, R. and Robinovitch, M. S. (1971) Sentence structure and retention in good and poor readers. J. educ. Psychol., 62, 25-30. Wernicke, C. (1874) Deraphasische Symptornenkomplex. Breslau, Cohn and Weigert. Wernicke, C. (1966) The symptom complex of aphasia. In R. S. Cohen and M. W. Wartofsky (eds.), Boston Studies in the Philosophy of Science (Vol. 4), Dordrecht, D. Reidel. Yakovlev, P. 1. (1962) Morphological criteria of growth and maturation of the nervous system in man. Res. Pub. Assoc. Nerv. Ment. Dis., 39, 3-46. Yakovlev, P. I. and Lecours, A. R. (1967) The myelogenetic cycles of regional maturation of the brain. In A. Minowski (ed.), Regional Development of the Brain in Early Life, Oxford, Blackwell. Young, I:. A. and Lindsley, D. B. (eds.) (1970) Early Experience and Visual Information Processing in Perceptualand Reading Disorders, Washington, D. C., National Academy of Sciences.

Cognitive and neurological basis of developmental dyslexia

33

Dam cet article, on souticnt la these que des enfants dyslcxiqucs ont des difficultds a acceder au scns dcs mots Ccrits a travcrs le recodage phonologique bien que ccs enfants ac&dent corrcctemcnt au scns par la voie visuelle dircctc. On rend compte de cettc difficult6 du recodagc phonologique par un deficit de la memoire a court terme. I1 est suggCrC que la dyslexie se fonde sur une dysfonction genetique du lobe pariental inferieur et on donne la preuve que cette region est importante a la fois pour la lecture et pour la memoire a court terme. Des consequences therapeutiques sont disc&es.

Cognition, I(19791 35-47 @Elsevier Sequoia S.A., Lausanne

Brief Report - Printed

in the Netherlands

Remembering plurals: Unit of coding and form of coding during serial recall* HUGO VAN DER MOLEN** JOHN MORTON MRC Applied Cambridge,

Psychology England * * *

and

Unit,

Abstract Subjects were required to recall lists of six words which had been presented visually in sequence. One or two of the words might be plural nouns. A substantial number of errors occurred in which the plural form became detached from its original root. This is taken as evidence for a morpheme-based code as opposed to a unitary word code. A significantly high proportion of these errors maintained the form of the plural C/S/, /Z/ or (a Z/J which is considered evidence for a primarily phonological coding of the plural morpheme. There is however a suggestion that on some occasions the plural is coded morphemically-i.e., in a way which does not distinguish between the various plural endings.

It is currently common to talk in terms of the code in which material is registered mentally in particular tasks. By means of a simple illustrative experiment we wish to make a distinction between two factors; the units on which the code is “based” and the form of the code. The distinction between the two can best be described by means of an example. The word CATS can be represented in a number of ways. There could be a single unit, which would be at the word size. Alternatively the code could be morpheme-based in which case there would be two units CAT + plural morpheme. A third possibility is that the characters are the only units. If the characters were letters then at this size of coding there would be no difference between the forms of representation of CATS, ACTS and SCAT. Given that we have units of a particular size we then have options as to the form of the code. The plural morpheme could thus be represented mor*The research was carried out in 1974 while the first author was supported by a Graduate Fellowship from the Rotary Foundation. **Now at the University of Groningen, The Netherlands. ***Requests for reprints should be sent to John Morton, M.R.C. Applied Psychology Unit, 15 Chaucer Road, Cambridge, CB2 2EF, England.

36

Hugo van der Moletz and Johtz Morton

phemically, in which case the plural morphemes for CATS, DOGS and MEN will be identical. Alternatively the plural morpheme could be coded phonologically, in which case that for CATS would be represented as /s/ and that for DOGS as /z/. With a visual code the two have an identical form, namely ‘S’. In discussing the internal representation of words we will prefer to talk about the dominant unit of coding rather than trying to establish a particular unit as the only one. Equally we will see that in practice the form of the code may not be restricted to one type in the course of a single task. The task we will examine is that of serial recall where words were presented visually to our subjects. The units and form of the coding will be reflected in the type of errors found in recall. Our interest was focussed specifically on the way in which plural nouns are represented internally. There are a number of possible ways in which the plural might be coded and linked to the root morpheme.

Word-based

and character-based

coding

In the extreme case of word-based coding one would expect relatively few errors occurring in which the plural inflection becomes detached from its original root, or in which the separated plural inflection becomes attached to another root. In the extreme case of character based coding the plural inflection is related to the rest of the word in just the same way as any other component. Thus, in a word like SPOTS, the final S would be linked to the preceding T in just the same way as the initial S was linked to the P. This could apply equally whether the code has a visual form or a phonological form. Thus errors of the SPOTS-SPOT type should be no more common than errors of the SPOTS-POTS type (given suitable controls for frequency etc). And in general one would expect omission errors to involve both parts of the word equally. Morpheme-based

coding

In this case the root and plural parts of the word are linked together in a way which is different from the intercharacter links in the root. Specifically, the former would be weaker. Thus we would expect recall errors of the form SPOTS-SPOT considerably more often than errors like SPOTS-POTS. We have a further decision to make concerning whether the plural part of the stimulus word can survive in memory separated from its root. If it can then we would expect there to be errors which look like the transfer of the plural from one stimulus to another. Thus: SPOTS CAT-SPOT CATS.

Remembering plurals

37

One study which examines the possibility of the morpheme as unit of coding is by Murrell and Morton (1974) who showed that the prevailing units involved in facilitation of visual recognition of words were the morphemes, rather than the words or characters. Taft and Forster (1975) came to similar conclusions with a lexical decision task. Within the framework of the logogen ‘model (Morton, 1968, 1969, 1970) this would lead to the prediction of morpheme-based coding in the short term serial recall paradigm. Concerning the form of the coding there would be two main alternatives, given that the code was morpheme-based. (a) Phonological code. In such a case the plural morpheme would be represented as one of the three alternative regular forms of the plural Is/, /z/ and /az/ or the feature list which correspond to these. Three equivalent classes of noun are formed as a function of the final phoneme in the root noun. Thus /az/ appears after sibilants and affricates as in ‘horses, judges’, /z/ appears after all other voiced phonemes as in ‘dogs, windows’, and Is/ after all other voiceless phonemes as in ‘cats, mops’. The same symbol would be used to represent the endings of 3rd person singular present tense verb forms such as “teaches, opens, sits”. (b) Morphemic code. Here all plurals would be coded in exactly the same way as, say, [ PL] . So CATS would be coded as [CAT] + [ PL] , DOGS would be [DOG] + [ PL] , and MEN would be [MAN] + [ PL] . The verb forms, however, would be coded differently, SITS being represented, perhaps, as [SIT] + [SING] + [3rd] + [PRES] . Note that the [CAT] component of CATS would be identical to that of CAT. In principle we can decide between the alternatives (a) and (b) in a fairly simple way. We present the subjects with a list of words which contains a plural, say, BEDS, and three singular nouns, say CAT, DOG and HORSE. We are particularly interested in the occasions where BED is made as a response. In this case the plural has become detached from the root noun and we would suppose that it would be floating, as it were, in memory and could be available for recall with some other noun. The important question is whether or not the three classes of noun are equally likely to appear in plural form. If the plural is coded morphemically then CAT, DOG and HORSE are equally likely to appear in the plural form. If the code were phonological, however, the particular form of the plural, in this case /z/, would require that it went only to a noun of the correct class-in this case DOG. Thus by examining the fate of the lost plurals we can decide between phonological and morphemic forms of coding. Unfortunately, the predictions are not quite clear as it seems. If one allows the possibility of decay or interference during the recall inter-

38

Hugo van der Molerzand John Morton

val then a /z/ form could be changed into a /s/ form etc., in store, giving rise to an error which would appear to support the notion of morphemic coding. Two further predictions can be made. If the plural is coded morphemically then the loss of a plural from a stimulus noun should never lead to an error on a verb such as producing KNOWS instead of KNOW. The suffix on a verb does not indicate a plural. If the coding were phonological, however, there would be no obstruction to such a transfer (on the assumption that the derivational history and semantic or syntactic force of a suffix is not maintained during those processes involving a phonological code). With a morphemic code there would be no obstruction to the formation of plurals from irregular forms (such as MEN for MAN). With a phonological code this would be impossible. There is a little prior work which would lead to particular predictions other than the quite generally reported finding that phonological confusions are found in material presented visually. As it currently stands the logogen model is indifferent as to morphemic or phonologically coded plurals in this paradigm.

Method

Stimulus Materials The stimulus lists each consisted of six words. They can best be seen as being made up of a set of three test words and a set of three ‘context’ words. The words in the test sets were all monosyllabic singular nouns. Each set contained one noun from each of the main plural-forming classes. The three words in each set were chosen to have roughly the same frequency of occurrence as in Kucera and Francis (1967). An additional constraint was that the plurals of the three nouns in a test set should also have the same frequency of occurrence. The sets of context words contained either 2 (in 6 cases), 1 (in 3 cases) or no plurals (in 3 cases). Such plurals were always regular and where there were two plurals they were of the same phoneme type. The three types of plural were equally represented, each being found in three sets. The rest of the context sets were made up of singular nouns which take irregular plurals, 3rd person plural present tense verbs, adjectives and prepositions. The sets are given in the Appendix. In the stimulus lists one test set and one context set were combined so that the three test words followed each other in the beginning, the middle or the end of the list. The place of the plurals in the lists varied but was balanced across the plural types. Different combinations of test and context sets were used for the six subject groups tested. Subjects, all adult female volunteers, were tested in groups of about 15 each. A total of 86 were tested.

Remembering plurals

39

Stimulus Presentation The words were projected

from slides-white letters on a black field, and were presented at a rate of 1.7 set (1 set exposure and 0.7 set between slides). Following the last word in a list a dim, blank slide was presented as a sign for the subjects to write down their responses. About 30 set were allowed for the responses. Each list was written on a separate page in a booklet. Two practice lists were given, after each of which the correct responses were written on a blackboard to familiarise the subjects with the procedure. Neither of these lists contained plurals and no mention of plurals was made during the instruction period. There followed the twelve test lists which were presented once through with, again, 30 set for written responses. After the experiment, subjects were asked to write down their comments on the experiment, notes on any strategies they had used or difficulties they had encountered. Only one of the subjects made any mention of the plurals, to the effect that they were more difficult. Other comments which were made include that the similarity of succeeding words made recall more difficult, that alliteration helped; that function words were more difficult than nouns. We were satisfied that the purpose of the experiment was not apparent to the subjects and that we could be confident that no special strategies were being used.

Results and Discussion Recall errors were only scored with respect to the plurals. For scoring purposes we refer to the root upon which the plural was formed as the plural word or W, and to the plural morpheme itself as P. Either or both of these components could be recalled. When the word, W, is omitted P can still be recalled attached to another word which may be either one which was in the list (either one of the test set or the context set) or an intrusion. In those lists containing two plurals it would be possible for either or both to be recalled with or without the plural. In the cases where P is attached to some word other than W, that word may belong to the same or a different phonological class as the stimulus W. Word-, Morpheme-

or Character-based

We can first summarise There are two kinds of morpheme have become 1. Loss of the plural: 406 were unambiguously

coding

the evidence that the coding is strongly morphemic. event which indicated that the root and the plural separated. Of the 1290 plural nouns presented to the subjects, forgotten (i.e., not counting possible replacements

40

Hugo van der Molen and John Morton

such as faces for phrases). Of the 884 nouns recalled 1.50 were recalled in the singular form. We take this 17% of errors as an indication of the degree of independence of the plural morpheme. The alternative explanation of the omission is that it was the final character (letter of phoneme) which was omitted rather than the plural morpheme. This would imply that any other final character would have the same status as the plural and be equally likely to be omitted. By simple extension the initial character might be supposed to be equally vulnerable. (a) The probability of an /s/ being omitted from the beginning of a stimulus word. There were only two possible cases-spaces which was reproduced as paces by only one of the 86 subjects, and sit which was never reproduced as it. (b) The probability of other final phonemes being omitted. A number of possibilities occurred in the stimulus lists. Any of guest, court, lawn, warm, tooth, wall, write, team and lamp, could lose the final phoneme and still produce an English word. In the data only two examples were found with TEA substituted for TEAM and LAMB for LAMP. (c) The probability of other initial phonemes being deleted. From the set of possibilities train, price, bridge, queen and fruit only FRUIT + ROOT was found from one subject. None of these baselines were designed for. However, the number of occasions on which evidence leads us to suppose the coding to be character based is very low indeed, less than 1% of the possible occasions. 2. On a total of 150 of the 1032 lists plurals were added to one or two of the test set nouns. The probability of such an addition was heavily dependent upon whether or not a plural had been in the stimulus list and whether or not an error had been made in the recall of the plural nouns. The figures are given in Table 1 (a, b). Firstly it is clear that the likelihood of a plural being added is very small in the case that no plural had been in the stimulus list. That is we are not seeing a generalized tendency to add plurals to the list regardless of the stimulus. Secondly, the probability of there being a spare plural is much greater if an error had occurred on a plural noun. It didn’t make more difference whether the error had been the omission of a plural from a remembered stem or the omission of the stem itself. In both cases, then, we can imagine that unattached plural morphemes were available in the appropriate storage system. A further check that these additions were plural rather than characters is given by the probability of an /s/ being added at the beginning of a word. It would, of course, be necessary for the result also to be an English word. Such a control was not designed for but the stimulus words included lake, park, team, tops and inch which could be turned into slake, spark, steam, stops

Remembering plurals

41

Table la. No.of lists

All stimulus lists with plurals No plurals in stimulus list

714 258 Total

Table 1b.

No. of lists with plurals added in response

%

141 9

18.2 3.5

-

1032

150

No.of

No. of lists with plurals added in response

%

113 28

22.3 10.5

(for stimulus lists with plurals)

lists

Response error in a plural noun No response error in a plural noun

507 261

and cinch (the latter with a phonological code only, the others with either phonological or visual). None of these errors occurred. Using this data as a control does involve some minimal extension to the underlying model but not excessively so. Therefore we confidently conclude that the unit of coding of the material during the recall interval is indeed strongly morpheme-based, in that the linkage between the root and the plural is weaker than the links within the root. A recent paper by Glanzer and Raze1 (1974) showed that free recall of a list of 15 bimorphemic nouns such as chestnut, teaspoon, earthquake was as good as recall of disyllabic single morpheme nouns like donkey, satin, fibre and slightly better than recall of monosyllabic words such as elm, yawn, scrap. Their interpretation is that the unit of storage is a word rather than a syllable or a morpheme. The same conclusion was drawn by Osgood and Hussain (1972) who showed that no transfer occurred from compounds like STOCK MARKET to the subsequent recognition of MARKET. The difference between this pair of results and ours lie in the nature of the complex words. We have used inflections, as did Murrell and Morton, whereas

42

Hugo van der Molen and John Morton

the experiments of Glazer and Raze1 and Osgood and Hussain used nouns compounded from two free morphemes. It seems likely that the psychology being adequate for the needs two terms for the linguists’ one, “morpheme” latter but not for our purposes.

Phonological

vs Morphemic

code of the Plural

Having decided that the plural is coded separately from the root morpheme we needed to decide the nature of the code. We pointed out in the introduction that if the plural is coded phonologically then having become detached from its root noun it could only reattach to a noun of the same type. If the plural is coded morphemically then it could be reattached to any other noun. The most extreme ‘morphemic’ code would be the subject just remembering something like “there was a plural noun somewhere, so let’s just make one of the responses plural”. In the analysis we will deal separately with displaced plurals and duplicated plurals since they could in principle give different answers. In analysing the fate of displaced plurals we excluded from the analysis those plural responses which were clearly the result of a simple error in the root morpheme. These included changes in a single feature such as the devoicing of the /z/ in PHASES to give FACES, insertion of an additional phoneme such as PHRASES instead of PHASES (an error made by 3 subjects) and the homophone error of PEARS for PAIRS (4 subjects). To include such errors would be to bias the results in favour of phonological coding. We were left with 154 errors to examine from a total of 774 lists. There were 3 1 changes involving the context set; these are special and will be discussed further below. There remain a total of 123 errors whose distribution is given in Table 2. In 84 cases a plural was added to one of the test set nouns, there were 36 plurals on intrusion nouns, and 3 cases of intrusions involving verbs with 3rd person singular endings. Table 2.

Errors involving a switch of the plural to another word Stimulus

Error type

Total

Iad Is/ IZI

list type

/ad

IS/

Id

Total

29 2 11

10 23 8

4 16 20

43 41 39

42

41

40

123

Remembering plurals

43

It should be recalled that the test lists contained a plural or plurals of one particular class (/s/, /z/, or /a~/). For each class we tested the null hypothesis that the plural intrusions were spread over the same phonological class or one of the different phonological classes with probabilities of l/3 and 2/3 respectively, using 2-tailed binomial tests with a correction for continuity. This test requires independence of the occurrences that are counted in the columns of the matrix in Table 2. Although this requirement is not completely met it is closely approximated since we have about 40 errors in each column spread over 86 subjects, and any one subject rarely contributes more than one error to the same column. In the exceptional cases there was no evidence of errors due to a single subject being related. The null hypothesis was rejected at the 1% level for the /z/ and /s/ lists and at the 5% level for the /z/ lists. We can be satisfied, then, that this class of errors is biased towards the same phonological class as the stimulus lists. Such biases could not exist unless the plurals concerned were coded phonemically. Duplication

errors

A further class of errors involved those 267 lists where the plural nouns were correctly recalled. On 29 occasions one of the test set nouns was given a plural, there were 5 intrusive plurals and 18 errors involving the context set. Ignoring the latter for the moment we have 34 errors (13%) whose distribution is shown in Table 3. The columns of this matrix reject the null hypothesis only in the case of the /s/ lists (p < 0.01). For the case of the /s/ lists, then, we have weak evidence of a response bias induced by the stimulus list. In practice this could be brought about by the reduplication of a phonologically coded plural morpheme in memory. Precise accounts will depend upon the theoretical superstructure one wishes to adopt. The extent to which this is a matter of general experimental set can be estimated by the fact that in answer to the 258 control lists which contained no plurals, only 9 plurals (3.5%) were given in a reply vs the level of 13% for the errors of duplication. Table 3.

Errors involving the addition of a plural Stimulus list type

Error type

Total

Ia4 IS/ Id

iad

Id

I4

4 2 1

2 11 3

4 6 1

10 19 5

I

16

11

34

Total

44

Hugo van der Molen and John Morton

We have treated plural errors involving nouns and verbs as equivalent in Table 2. This is legitimate if one is considering a phonological code. The problem is more complex if one is thinking in terms of morphemes since the inflected form, e.g., SINGS, is restricted to the 3rd person singular present. If the inflection is to be associated with number it would thus be considered singular. If errors involving verbs were thought to be morphemic then it would be the [SINGULAR] marker not the [PLURAL] one which could lead to the error (the addition of an s). If this were the case then we would expect a higher proportion of inflected verbs in the control lists. That is if the stimulus list included: NOUN + PLURAL dogs i.e. know VERB the possible error would lead to: NOUN dog i.e. know. VERB + PLURAL With the control lists we would have: NOUN + SINGULAR dog i.e. know VERB changing to the error form: NOUN dog i.e. knows VERB + SINGULAR (with assumptions about the nature of unmarked number and person). In fact there were no inflected verbs given as responses to the control lists. Errors involving verbs can provide further evidence in favour of low level coding. The context sets with a single plural noun also included a verb. These verbs were chosen to be of the same phonological class as the plural noun, sit, know and teach being associated with tops, beds, and spaces respectively. If the plural is coded phonemically then if an error were made it could as well end up with the verb (giving sits, knows or teaches) as with the test set noun of the same class. In those 258 lists there were 27 errors involving same class nouns and 29 involving verbs. This could not be the case if morphemic or semantic information were carried with the inflection. Evidence for morphemic code We have presented the evidence supporting a phonological code. The existence of such evidence does not preclude the possibility of a morphemic code also being involved. The main evidence for a morphemic code is that the

Remembering

plurals

45

biases in Table 2 are incomplete. In particular a /z/ plural, lost from the stimulus plural reappears equally often as /z/ and /s/. There seem to be two possibilities in respect of such errors; either they arise from morphemic coding (from some of the subjects or all of the subjects some of the time-there are insufficient errors from individual subjects to evaluate these alternatives) or the phonologically coded plural is subject to some kind of transformation as a result, perhaps, of interference or decay. In such a case we must observe that there are 5 1 such errors of the 123, giving a probability of 0.41 that a detached plural will be altered in form. While we have no base-rate against which to compare this figure it does seem rather high. Eight of the context sets contained singular nouns which form irregular plurals. These nouns were included in an exploratory way only, as it proved difficult to match them all in frequency of singular and plural forms with other nouns. They were recalled in plural form on 20 occasions, involving the responses of WOMEN (11 times), MEN (3), FEET (2), TEETH (2), and MICE (2). LICE, GEESE and PENCE were never given as responses. Again we have trouble with a base-line against which to evaluate such figures. The responses all involve a change in the vowel alone (in both vowels in the case of WOMEN). Looking at responses to the other words in the stimulus lists we found only four which involved just a vowel change; FAR was given twice for FIRE and TIPS and SET were given once each for TOPS and SIT. A large number of other similar possibilities never occurred such as ROSE-RISE, LIKE-LOOK, HEART-HURT, FOOT-FIT, SHIP-SHOP, WALL-WOOL. We take the 20 irregular plurals, then, as weak evidence in favour of some morpheme coding of the plurals. More detailed discussion of the implications of this kind of error would be contingent on the way chosen to describe such plurals (see Palmer, 197 1, p. 112 ff. for an account of some of the problems). It seems, then, that we have evidence of at least two forms of coding of the plurals. The position with respect to the root nouns is equally complex. The existence of homophonic substitutions might be taken as evidence in favour of the essentially phonological code of the material being remembered. Thus WRITE was variously reproduced as RIGHT (4 times) and WRIGHT (twice). However, we may view this data from the other side and point out that in about 50 cases WRITE was reproduced correctly. Such accuracy might be claimed to constitute evidence in favour of a visual or a semantic code at least supplementing the phonological code in the majority of cases. The other examples of phonological errors were COURT which was once produced as CAUGHT and THERE which was once produced as THEIR. In conclusion we claim to have shown in a situation where subjects are visually presented with a list of words for immediate serial recall, that plurals are coded separately from the root nouns they modify. There is evidence

46

Hugo van der Molen and John Morton

from the distribution of errors showing that the plurals are primarily coded in a phonological form but some other evidence suggests the presence of a morphemic code some of the time. It is possible that the two-kinds of error occur at different stages of the task (e.g., storage vs retrieval): equally it is possible that information is liable to be coded in many forms at the same time.

References Glanzer, M. and Raze], M. (1974) The size of the unit in short-term

storage J. verb. Learn. verb.

Behav., 13, 114-131.

Kucera, H. and Francis, W. N. (1967) Computational Analysis of Present-day American English. Rhode Island, Brown University Press. Morton, J. (1968) Grammar and computation in language behaviour. In J. C. Catford (ed.), Studies in Language and Language Behauiour C.R. L.L.B. Progress Report, 6, University of Michigan. Morton, J. (1969) Interaction of information in word recognition. Psychol. Rev., 76, (2), 165-178. Morton, J. (1970) A functional model for memory. In: Norman, D. A. (ed.), Models of HUmanMemory. New York, Academic Press. MurreIl, G. A. and Morton, J. (1974) Word recognition and morphemic structure. J. exper. Psychol., 102, (6) 963-968.

Osgood, C. E. and Hoosain, R. (1974) Salience of the word as a unit in the perception

of language,

Percep. Psychophys., 15, 168-192. Palmer, F. R. (1971) Grammar, Harmondsworth,

Penguin Books. Taft, M. and Forster, K. I. (1975) Lexical storage and retrieval of prefixed words. J. verb. Learn. verb. Behav., 14, 638-647.

APPENDIX Testword Sets chance heart club

edge ship bar

bridge park team

price test game

case point boy

voice court table

axe lamp lawn

inch guest yard

there

tops

beds

climb

queen write

dark sit

man know

spaces feet teach

trains pens woman

phases ranches louse

cups books goose

pairs plans penny

fences pieces tooth

noise

rose

fruit queen

lake king

horse unit wall

house fact home

Context Sets from warm

gay far

sing rocks aunts mouse

Remembering

plurals

47

On a demand& i dcs sujets de rdpkter des listcs de 6 mots present& visucllement. Parmi ccs mats SC trouvaicnt un ou dew noms au pluriel. On trbuve un nombrc important d’crreurs qui SC traduisent par Ie ditachemcnt de la forme pluriellc dc la racine originale. On considke cela comme une preuve en faveur d’un codagc fond& sur le morpheme plutot qu’un codage fond6 sur lc mot comme unit& Dans une proportion significativement 61evde ces crreurs mainticnnent la forme du pluriel (/s/, /z/ ou /aZ/) dont on pensc qu’clle prouve la priorit& du codage phonologiquc du morpheme pluricl. Cependant il est possible que le pluriel soit parfois cod@ morphondmiquement c’cst i dire de faqon i ce quc les diverses terminaisons plurielles ne soient pas distinguables.

Cognition, @Elsevier

l(1979) 49-59 Sequoia S.A., Lausanne

Brief Report ~ Printed

in the Netherlands

Semantic focus and sentence comprehension* ANNE CUTLER Massachusetts

and JERRY A. FODOR

institute

of Technology

Abstract Reaction time to detect a phoneme target in a sentence was found to be faster when the word in which the target occurred formed part of the semantic focus of the sentence. Focus was determined by asking a question before the sentence; that part of the sentence which comprised the answer to the sentence was assumed to be focussed. This procedure made it possible to vary position offocus within the sentence while holding all acoustic aspects of the sentence itself constant. It is argued that sentence understanding is facilitated by rapid identification of focussed information. Since focussed words are usually accented, it is further argued that the active search for accented words demonstrated in previous research should be interpreted as a search for seman tic focus. The meanings of individual words and the combinatorial information provided by syntax do not always constitute the entire meaning of a sentence. The relationship of a sentence to its context is indicated, for example, by the focus structure of the sentence, where the focus is that information which is new and unrecoverable from preceding discourse. Speakers commonly indicate focus by assigning primary sentence accent to focussed words. Thus the new information in (1) is that it was the man on the corner, not some other man, who wore the blue hat, whereas in (2) the new information is that the hat was blue and not some other color: (1) (2)

The man on the CORNER was wearing the blue hat. The man on the corner was wearing the BLUE hat.

Although accenting a word is the most usual way of expressing focus, various syntactic devices can produce the same effect, for instance the topicalisation of “blue” in (3): *This research was supported by N. I. H. grant number 5-ROI-HD05168-05 to Jerry A. Fodor and Merrill F. Garrett. The authors are grateful for comments from Dianne Bradley, David Fay, Merrill Garrett, David Swinney, Virginia Valian and Edward Walker. Our thanks also to Dumont Billings, who ran the experiment, and to Edward Walker for providing the ALICE statistical analysis program. Reprint requests should be addressed to the first author at the Laboratory of Experimental Psychology, University of Sussex, Brighton, BNl 90G, England.

50

(3)

Anne Cutler and Jerry A. Fodor

Blue, the hat was, that the man on the corner was wearing.

Linguists have devoted considerable attention to the manifestations of focus (see e.g., Halliday (1967) Jackendoff (1972)). Psycholinguistic studies of comprehension have not, so far, explicitly addressed the question of how hearers identify the focus of a sentence. Nevertheless, some indirect evidence can be adduced; focus is very often expressed as accent, and the processing of a sentence’s suprasegmental structure (accent and intonation contour) has been extensively investigated. The results of a series of phoneme-monitoring studies of accent appear to offer a starting-point for investigating the comprehension of focus. In the phoneme-monitoring task, subjects are asked to comprehend sentences and at the same time to listen for the occurrence within them of a word beginning with a specified target sound. Reaction times to the target sound in this task are faster if the target begins a word which bears accent (sentence stress) than if it begins a non-accented word (Shields, McHugh and Martin, 1974; Cutler and Foss, 1977). Since accented words are acoustically clearer in several ways (Lehiste, 1970) it might be argued that this result reflects merely the greater ease of identifying acoustically clearer words. Cutler (1976a), however, concluded that at least part of the reaction time advantage of accented targets was due to an active search for accented words on the part of the listener. In Cutler’s experiment each sentence was produced in two suprasegmental versions, one in which the target-bearing word was accented and one in which it was not. The target-bearing word was then spliced out of each version and replaced by identical copies of the same word taken from a third, fairly monotone rendition of the same sentence. Thus the target-bearing word was acoustically identical in the two versions of each sentence to be compared, but the suprasegmental context in which it occurred differed greatly from one version to another: in one version the intonation of the early part of the sentence was consistent with accent falling on the target-bearing word, in the other version it was consistent with accent falling elsewhere. Reaction times were significantly faster to targets which occurred in positions which should have borne accent than to targets which occurred where no accent was predicted. Since the acoustic clarity of the target-bearing word did not differ for the two versions, the reaction time difference in this experiment must reflect the suprasegmental differences in the part of the sentence which preceded the target. In other words, cues in the intonation contour must have enabled listeners to direct their attention to a part of the sentence where accent was about to occur. The function of sentence accent is, above all, to indicate new information. It is clear that locating accented words as quickly as possible could, therefore,

Semantic focus and sentence comprehension

51

be a useful strategy in sentence comprehension. Once the accented words are identified, the hearer knows which part of the sentence is new information, i.e., which part is most essential for grasping the speaker’s message. Thus the reaction time advantage of accented words in phoneme-monitoring seemed to us to be due at least in part to the fact that accent expresses focus. To test this explanation, we designed the present experiment. If the reason that accent facilitates phoneme-monitoring reaction times is because it indicates focus, then, we predicted, a similar reaction time advantage for focussed words should be obtainable when focus is indicated by non-suprasegmental cues. The main design problem with such an experiment lies in the very premise on which it is based, namely, that sentence focus and accent coincide; since accent affects phoneme-monitoring reaction times, to demonstrate an analogeous effect for focus we must remove the confounding with accent, i.e., keep the acoustic representation of the sentence constant across changes in focus. However, as we have mentioned, accent is not the only method of focussing a word in a sentence.’ In this experiment focus was manipulated by preceding the sentence with a question, the answer to which was the word to be focussed. Thus when (4) is preceded by (5) it can be said to have a focus structure identical to that of (1 ), whereas preceding it by (6) will produce the focus structure of (2): (4) (5) (6)

The man on the corner was wearing the blue hat. Which man was wearing the hat? What hat was the man wearing?

In the present experiment, a set of sentences similar to (4) was presented to subjects. Half the subjects heard the sentence preceded by (5) half heard it preceded by (6). The experimental tapes were spliced together from separate recordings of sentences and questions, so that the monitored sentence (4) was actually acoustically identical in each presentation. Because acoustic cues in the sentence could favor one monitoring location over another, each sentence contained two possible phoneme targets; within each of the two question groups, half the subjects listened for one target, half for the other. Thus ’ These methods include topicalisation, clefting (“It is the man who is wearing the hat”) and pseudoclefting (“The one who is wearing the hat is the man”), all of which are less suitable than the method actually adopted since they involve variations of the surface structure of the sentences to be compared. Allen and O’Shaughnessy (forthcoming) report that various devices used to indicate sentence focus-clefting, pseudo-clefting, topicalisation, preposed question-produce reliable and essentially similar effects on the suprasegmental contour; fundamental frequency accent occurs in each case on the element which is focussed. (In the present experiment these explicit cues were of course not present since the questions were produced separately from the sentences).

52

Anne Cutler and Jerry A. Fodor

in (4) half the subjects listened for a word beginning with /k/, half listened for a word beginning with /b/. If phoneme-monitoring reaction time were indeed to prove sensitive to focus variation in the manner suggested by the results from the accent experiments, then it was expected that subjects monitoring in (4) for the /k/ sound on the word “corner” would respond faster if (4) was preceded by (5) than if it was preceded by (6), while subjects monitoring in the same sentence for the /b/ sound on “blue” would produce faster reaction times if (4) was preceded by (6) than if it was preceded by (5). Thus we predicted an interaction between the two variables of Focus Position and Target Position, such that reaction times to earlier targets would be faster when the preceding question focussed upon the earlier part of the sentence, whereas reaction times to later targets would be faster when focus was upon the later part of the sentence. On the basis of other findings, a main effect of Target Position was also predicted. It is a reliable finding in phoneme-monitoring experiments that targets near the end of a sentence produce faster reaction times than targets near the beginning of a sentence (Foss, 1969; Shields, McHugh and Martin, 1974). It has been argued (Cutler, 1976b) that this effect reflects the construction and testing of semantic hypotheses during sentence comprehension, with a greater probability of correct hypotheses being constructed as the sentence progresses towards its end.

Methods Materials

Thirty-two experimental sentences were constructed and recorded along with thirty-two distractor sentences. Also recorded were two questions for each experimental sentence, one question directed to information in the earlier part, one to information in the later part of each sentence. For each experimental sentence two possible target specifications were determined, with one target-bearing word occuring in each of the two parts of the sentence to which the questions were directed (as in (4) above). Thus each experimental sentence occurred in four versions, with all possible combinations of target and question. Each distractor sentence was also preceded by a target specification and a question. The experimental sentences with their questions are listed in Appendix I. Care was taken in recording the experimental sentences to keep the intonation contours as neutral as possible; in particular, no potential target phrase was assigned emphatic accent.

Semantic focus and sentence comprehension

53

Since the targets in the experimental sentences always occurred in modifiers (attributive adjectives or prepositional phrases), the distractor sentences were designed to prevent this regularity becoming apparent. Targets in distractor sentences began verbs, head nouns of noun phrases, adjectives in predicate position and prepositions. Twelve distractor sentences did not contain an instance of the specified target. Four materials sets were constructed, each containing all distractor sentences and a version of each experimental sentence. No experimental sentence occurred in the same version on more than one tape, and the two variables of Target Position (early versus late) and Focus Position (early versus late) were counterbalanced across sets. By means of tape-splicing the need for multiple recording of the sentences was avoided; the experimental sentences were acoustically identical in each set although preceded by different targetquestion combinations. Subjects

Seventy-nine members of the MIT community, recruited by advertisement on campus, took part in the experiment. Each subject was paid $2 for participating. Twenty subjects heard each of three materials sets, nineteen subjects heard the fourth materials set. Procedure

Subjects were tested in groups of 2 or 3 at a time. They were instructed first to pay careful attention and comprehend the sentences as they would be tested on them later, and second to listen within the sentences for the occurrence of the target sound for that sentence, and to press a button as soon as they heard a word beginning with that sound. The target sound, which could be /b/, /d/ or /k/, was specified immediately prior to the question which preceded the sentence. The subjects were informed that the targetbearing word would occur only in the sentence, never in the question. No question contained a word beginning with the target sound for the sentence following it. The instructions and materials were presented binaurally over Telephonics TDH-49 headphones. A signal on the tape, synchronised with the onset of the target sound but inaudible to the subjects, started a separate timer for each subject which was stopped by the subject’s action in pressing the button. After hearing the sentences the subjects were given a post-test which covered all 32 experimental sentences. The subjects were required to make the correct choice among four alternatives replacing one word in the sentence. In half of these judgments the subjects were providing words which had constituted the answer to the question they had heard, in half they were pro-

54

Anne Cutler and Jerry A. Fodor

viding words from the other part of the sentence; in half of the cases, the words provided had been target-bearing words, in half they had not.

Results 1. Reaction

Times

The data from three subjects who had missed a large number of targets were omitted from this analysis. Seven per cent of the remaining data was lost for varying reasons: of a total of 2432 (76 X 32) target presentations, 26 were not recorded because of equipment failure or experimenter error; in 57 cases, the subject failed to respond; 16 very long reaction times were excluded as possible reprocessing (“double take”); and 73 very short reaction times were excluded as anticipations. No more than 5 responses were lost for any one subject. A mean reaction time for each condition was computed for each of the remaining 76 subjects; the overall means for each condition are shown in Table 1. A mean reaction time for each condition was also computed for each sentence. Separate analyses of variance were performed on these two sets of means; the combined result of the two analyses allowed computation of the min F’ statistic (Clark, 1973). Both main effects and the interaction between them reached significance. Reaction times to targets in later positions were significantly faster than reaction times to targets in earlier positions (min F’ (1,50) = 20.86, p < 0.001). Reaction times were faster when the preceding question focussed upon the earlier part of the sentence than when it focussed upon the later part of the sentence (min F’ (1,95) = 8.45, p < 0.01). The predicted interaction was also found: reaction times to the earlier target were faster in the early than in the late focus condition, whereas reaction times to the late target were faster in the late than in the early focus condition (min F’ (1,84) = 19.36, p < 0.001). Table 1.

Mean RTs (msec) to presence of target photreme Target

position

early

late

x

376

360

368

445 410

344 352

394

I’ocus position

Semantic focus and sentence comprehension

55

The’ difference between focussed and non-focussed targets in the late target condition, which was small, was examined separately and found to be significant (t(3 1) = 1.7, p < 0.05 (one-tailed)). The additional variable of Materials Sets in the analysis by Subjects was not significant, and did not interact with the Target Position variable or with the Target Position by Focus Position interaction. It did, however, interact weakly with the Focus Position variable (F (3,72) = 2.9 1, p < 0.05); although all four Materials Sets showed a Focus Position effect in the same direction, one set showed a stronger effect than the others. It is doubtful whether much importance should be attributed to this finding. 2. Post-test Overall performance on this test maintained a high level: mean percent correct across all 79 subjects was 70% (chance performance would be 25%). When the word to be chosen had been the target-bearing word the mean number of errors across subjects was 4.66, when it had not borne the target the mean number of errors was 4.99, a non-significant difference (t (78) = 0.288, p > 0.1). However, as had been expected, words which had been focussed were retained better, so that significantly fewer errors were made when the word to be chosen had been the focussed word than when it had not been focussed (4.23 versus 5.41; t (77) = 2.368, p < 0.025 (one-tailed)). Discussion As predicted, reaction times to the earlier target in a sentence were faster when the preceding question had focussed on the earlier part of the sentence, reaction times to the later target were faster when the focus was upon the later part of the sentence. Furthermore, as predicted, targets near the end of the sentence were responded to faster than earlier targets. Reaction times were also faster when the preceding question had focussed upon the earlier part of the sentence than when it focussed upon the later part. This result presumably reflects completion of the focus-identifying (question-answering) task, freeing greater processing capacity for other tasks. Subjects in phoneme-monitoring experiments are in any case performing a divided attention task, listening for the target and comprehending the sentence simultaneously. In this experiment the presence of the preceding question provided yet another task, namely determining the answer to the question, despite the fact that the subjects were not required to give the answer after hearing the sentence. Extensive work in recent years on the division of attention has shown that simultaneous performance of two competing tasks can lead to a decrement in performance on one or another or both in com-

56

Anne Cutler and Jerry A. Fodor

parison with the level of performance on the same tasks when performed in isolation. In the present case it is likely that simultaneous performance of the question-answering task would interfere with performance of the monitoring task, and the reaction time data lend support to this assumption. (Note that it is unlikely that the question-answering task, or for that matter the monitoring task, would interfere with comprehension. Highly overlearned tasks appear to be resistant to interference from competing tasks (Moray, 1969) and sentence comprehension is probably the most highly overlearned task that subjects in a laboratory experiment are ever asked to perform). Our chief prediction, however, concerned the interaction between target position and position of focus. We have shown that reaction time to a phoneme target on a particular word is faster when the word is focussed than when it is not. We would like to argue that this result indicates that the previous finding of a reaction time advantage of accented words reflects at least partly the role of accent as an indicator of focus, or new information. Similarly, the finding by Cutler (1976a) that listeners actively search for words which will be accented can be interpreted in the light of the present experiment as a search for sentence focus. The most impressive aspect of the operation of the sentence comprehension device is its speed. For a sentence to be understood, the part which above all needs to be apprehended must be the focus-the part which is totally new. Any strategy which allowed the comprehension device to find the sentence focus rapidly would presumably facilitate understanding. Thus we suggest that our results indicate that particular attention is directed to the processing of focussed words when a sentence is being understood. Once the focussed segment has been located (usually by tracking the intonation contour to locate the main accent), processing of it begins immediately, even though in some cases processing of preceding segments is not complete. Thus phoneme targets on focussed words can be identified more rapidly than targets on non-focussed words. We should note that a possible alternative explanation would stress not the prior processing of focussed words, but rather their relative availability for controlling responses. On this account, the order of perceptual analysis of material in the sentence would be independent of focus, but once the focussed segment has been identified it becomes the item most accessible for report. The present data are compatible with this interpretation, but it seems to us relatively implausible. Phoneme-monitoring reaction times appear to be predominantly affected by on-line processes, above all lexical access (Cutler and Norris, 1979). This latter account would in any case still bolster our principal conclusion, that the identification of semantic focus is an essential component of sentence comprehension.

Semantic focus and sentence comprehension

57

References Allen, J. and O’Shaughnessy, D. (forthcoming) The effect of focus shifting transformations on fundamental frequency contours. Clark, H. H. (1973) The language-as-fixed effect fallacy: A critique of language statistics in psychological research. J. verb. Learn. verb. Eehav., 12, 335-359. Cutler, A, (1976a) Phoneme-monitoring reaction times as a function of preceding intonation contour. Percep. Psychophys., 20, 55-60. Cutler, A. (1976b) Beyond parsing and lexical look-up. In R. J. Wales and E. C. T. Walker (eds.), New Approaches to Language Mechanisms. Amsterdam, North-Holland. Cutler, A. and Foss, D. J. (1977) On the role of sentence stress in sentence processing. Lang. Speech, 20, l-10. Cutler, A. and Norris, D. (1979) Monitoring sentence comprehension. In W. E. Cooper and E. C. T. Walker (eds.), Sentence Processing: Psycholinguistic Essays Presented to Merrill Garrett. Hillsdale, N.J., Erlbaum. Foss, D. J. (1969) Decision processes during sentence comprehension: effects of lexical item difficulty and position upon decision times. J. verb. Learn. verb. Behav., 8, 457-462. HaUiday, M. A. K. (1967) Notes on transitivity and theme in English. Part II. J. Ling., 3, 199-244. Jackendoff, R. (1972) Semantic Interpretation in Generative Grammar. Cambridge, Mass., MIT Press. Lehiste, I. (1970) Suprasegmentals. Cambridge, Mass., MIT Press. Moray, N. (1969) Attention: Selective Processes in Vision and Hearing. London, Hutchinson. Shields, J. L., McHugh, A. and Martin, J. G. (1974) Reaction time to phoneme targets as a function of rhythmic cues in continuous speech. J. exper. Psychol., 102, 250-255.

APPENDIX I Sentences used in the experiment 1. The man on the corner was wearing the blue hat. Which man was wearing the hat? Which hat was the man wearing? 2. The reporter with the daily newspaper was responsible for the candid story. Which reporter was responsible for the story? Which story was the reporter responsible for? 3. The opening of the concert was spoiled by the director’s outburst. Which opening was spoilt by the outburst? Whose outburst spoilt the opening? 4. The value of the bonds was altered with the devalued currency. Which value was altered with the currency? Which currency altered the value? 5. The house with the carport must belong to the doctor’s widow. Which house must belong to the widow? Which widow must the house belong to? 6. The woman with the bag went into the dentist’s office. Which woman went into the office? Which office was it that the woman went into? 7. The checking of the ballots was interrupted when the computer failed. What checking was interrupted when the failure happened? Which failure interrupted the checking? 8. The author of the bestseller refused to go to the Congressman’s party. Which author refused to go to the party? Which party did the author refuse to go to? 9. The motion to dismiss was proposed by the Californian senator. Which motion was proposed by the senator? Which senator proposed the motion?

58

Anne Cutler and Jerry A. Fodor

10. The manager of the dairy will check on his bank account. Which manager will check on his account? Which account will the manager check on? 11. The mother of two daughters wrote to the boarding school. Which mother was it that wrote to the school? Which school was it that the mother wrote to? 12. The interviewer from the digest questioned the book’s publisher. Which interviewer was it that questioned the publisher? Which publisher was it that the interviewer questioned? 13. The association of consumers objected to the new delivery procedure. Which association was it that objected to the new procedure? Which new procedure was it that the association objected to? 14. The price of bread was raised at the Council session. Which price was raised at the session? Which session raised the price? 15. The group of dancers negotiated with the Broadway agent. Which group negotiated with the agent? Which agent was it that the group negotiated with? 16. The owner of the department store fired the bicycle salesman. Which owner fired the salesman? Which salesman was it that the owner fired? 17. The voice of the caller was hard to hear on the defective telephone. Which voice was hard to hear on the telephone? Which telephone was the voice hard to hear on? 18. The road to the Cape was washed out in the December rains. Which road was washed out in the rains? Which rains washed out the road? 19. The suspension of the boy was protested at the college meeting. Whose suspension was protested at the meeting? Which meeting protested the suspension? 20. The flavor of the coffee was ruined by the dirty water. Which flavor was ruined by the water? Which water ruined the flavor? 21. The remains of the camp were found by the deer hunter. Which remains were found by the hunter? Which hunter found the remains? 22. The actions of the committee focussed on the dangerous situation. Which actions focussed on the situation? Which situation was it that the actions focussed on? 23. The attitudes of the businessman aroused his colleagues’ anger. Whose attitudes aroused anger? Whose anger did the attitudes arouse? 24. The watcher on the balcony saw the driver’s escape. Which watcher was it that saw the escape? Whose escape was it that the watcher saw? 25. The regulations of the commission set new billboard standards. Which regulations set new standards? Which new standards did the regulations set? 26. The residents of the district were annoyed at the building plans. Which residents were annoyed at the plans? Which plans were the residents annoyed at? 27. The program about Britain interested the common viewers. Which program interested the viewers? Which viewers did the program interest? 28. The fear of death inspired the courageous fighters. Which fear inspired the fighters? Which fighters were inspired by the fear? 29. The tourists from Denmark photographed the Bicentennial parade. Which tourists photographed the parade? Which parade was it that the tourists photographed? 30. The personnel officer of the company interviewed the baseball player. Which personnel officer interviewed the player? Which player did the personnel officer interview?

Semun tic focus and sentence comprehension

59

3 1. The rising price of boxes worried the candy manufacturer. Which rising price worried the manufacturer? Which manufacturer was worried at the rising price? 32. The janitor at the ballpark joined the custodians’ union. Which janitor joined the union? Which union did the janitor join?

R&M& On a trouve que le temps de reaction pour detecter un phoneme cible est plus rapide quand cette cible se trouve dans le mot qui fait partie du focus semantique de la phrase. Ce focus est determine par la pose d’une question avant I’enonce de la phrase: on suppose que la part de la phrase qui contient la reponse a cette question est le focus. Cette procedure permet de varier les positions du focus dans une phrase tout en n’en tenant constants les aspects acoustiques. On soutient que la comprehension d’une phrase est facihtee par I’identification rapide de l’information focalisee. Les mots focalises etant le plus souvent accent&s, on peut aussi soutenir que la recherche active des mots accent&s que des experiences precedentes avaient mis en evidence peut Ctre interpret&e comme une recherche semantique.

Co@tion, @Elsevier

7 (1979) 61-68 Sequoia S.A., Lausanne-Printed

Discussion in the Netherlands

Two unargued linguistic assumptions in Kean’s “phonological” interpretation of agrammatism JOHN KLOSEK Graduate

Center,

CUNY

*

It has long been recognized that damage to Broca’s area is associated with highly restricted speech output and with the loss of inflectional endings and functional words.’ These symptoms have for the most part been interpreted as resulting from a syntactic or morphological disorder. Kean (1977) in a complicated linguistic analysis purports to show that such an interpretation is untenable and that these deficits are, in fact, due to a phonological disorder. Essential to Kean’s hypothesis are two claims which will be shown to have no linguistic motivation; the first is the assertion that the plural morpheme is derivational-not inflectional-and the second is the postulation of the notion of “phonological word”. Many of the early quantitative studies of agrammatism (Goodglass and Hunt, 1958; Goodglass and Mayer, 1958; Goodglass and Berko, 1960) revealed that of the three homophonous -S morphemes in English (plural, possessive, and third person singular indicative) the most often preserved was the plural. The differential omission of the three morphemes was interpreted as resulting from a disruption of morphological and/or syntactic capacities.’ Kean, however, claims to have a deeper, more insightful theory which attributes this deficit to a purely phonological disorder. Given that aphasics omit homophonous morphemes at different rates, Kean was faced with the task of showing that there is some deeper phonological difference among the various morphemes, which is not obvious from the phonetic structure of the language. Her reasoning is as follows: There are two types of affixes in English-word-boundary and non-word-boundary (formative boundary); the former are set off by # and the latter by + in *Reprint requests should be addressed to John Klossek, Ph.D. Program Linguistics, CUNY, 33 West 42nd Street, New York, N.Y. 10036. ’ I would like to thank Jason Brown for bringing Kean’s paper to my attention and taking an interest in my reply, and D. Terence Langendoen, Anita Janda, Robert Vago, Robert May, Constantine Kaniklidis, Miroslav Rensky and Donald Byrd for commenting on earlier drafts of this paper. The present work is a much-shortened version of an earlier paper which covered many of the same points as Kolk (1978). ‘If we are to accept the phonological hypothesis, phonologically-like representations should be either retained or lost as a class without regard to syntactic, semantic, or lexical differences among them. This obviously is not what happens with the morphemes in question.

62

John Klosek

generative-phonology notation. It is the word-boundary affixes which tend agrammatic speech. However, this analysis is inadequate since the morphemes agrammatic speech. However, this analysis is inadequate singe the morphemes in question are all word-boundary affixes. In order to get around this, Kean had to show that there was some other phonological difference between the plural and the other two morphemes, thus the claim that the plural is derivational. What is the evidence on which Kean based this decision? From Bloomfield to Chomsky, linguists have considered the English plural an inflectional morpheme. Kean provides no linguistic arguments for the derivational status of the plural; she merely refers to Aronoff (1976) for a further discussion of word formation rules. Aronoff does indeed discuss the distinction between derivation and inflection, but nowhere does he state that the plural is derivais generally viewed as tional. On the contrary, Aronoff writes, “Inflection encompassing the ‘purely grammatical’ markers, those for tense, aspect, person, number, gender, case, etc.” (A ronoff, 1976, p. 2, italics added). Derivation, on the other hand, usually changes the grammatical category of a word; for exampie, the verb bake becomes the noun baker when the derivational affix -er is added; the adjective nice becomes the adverb nicely when the derivational affix -Iv is added, etc. Aside from the traditional-grammar classification, a number of arguments can be made to support the view that the plural morpheme in English is not derivational, but inflectional. Kean maintains that syntactically derived morphemes are inflectional (p. 23). The following sentences illustrate that the presence of the plural morpheme is “a function of the syntactic structure of the sentence” (p. 23); we can have, I like books but not *I like book, and I like a book that is long but not *I like a books that is long. Therefore, according to Kean’s own criterion, the plural morpheme must be inflectional. Furthermore, Chomsky (1965) has at least three good reasons supporting the view that the syntactic feature [+ Plural] , which marks the place in the tree where the plural morpheme is to be inserted, is syntactically determined. His position is the following: (1) (a) [+ Plural] is “inherent to the Phrase-marker rather than the lexical item, . . . and become[sl part of the formative only after it is inserted into the Phrase-marker” (p. 177, italics added). (b) The syntactic feature [+ Plural1 may be introduced by a contextfree rule of the base or a transformational rule (pp. 172, 181). rules, (c) [+ Plural] plays a role in the application of transformational e.g., relativization (p. 177). Since [+ Plural] is a syntactic feature, its presence in the syntactic structure of a sentence determines the occurence of the plural morpheme. In other

Kean S interpretation of agrammatism

63

words, the plural is syntactically determined, and by Kean’s own definition all such morphemes are inflectional. Bloomfield (1933) and Aronoff (1976) point out that inflectional morphemes are always the last in a series of morphemes. Aronoff writes, “In the English word compart+ment+al#ize#d, for example, the last morpheme, #d is inflectional, and those internal to it are derivational. The two sets may not be interspersed. Thus the word compart+ment+al#iz+ation#s is possible, though the word *compart+ment+al#ize#d+ation#s is not” (p. 2). If the plural allomorphs were derivational, then it would be possible for other morphemes to follow them, but if they are inflectional, no other morphemes should follow. Consider the following words: Sing.

Sing. Poss.

Plural

Plural Poss.

cats’ [k&l cat’s [k&s] cats [k&s] cat [k&l game’s [ geymz I games [geymz 1 games’ [geymzl game [germ 1 buses [ bhsaz] buses’ [ bAsaz] bus [bAs] bus’s [ bhsazl Notice that there is no phonetic difference among the singular possessive, the plural, and the plural possessive. If the plural possessives are formed from the plurals, one would expect the following phonetic representations: [kaetsazl , But the ‘possessive morpheme is not realized [geymzaz], and [basazazl. phonetically. Why is this so? The answer is simple if the plural allomorphs are inflectional: inflectional morphemes (the plural) may not be followed by other morphemes (the possessive).3 Even if the plural turned out to be a derivational morpheme, contrary to the evidence thus presented, it would not lend any support to Kean’s argument because traditionally the difference between derivation and inflection has generally been treated in terms of morphology or syntax, not phonology. In such an unlikely eventuality, Kean would have to argue that the difference between derivation and inflection is necessarily phonological. The second linguistically unjustified assumption central to Kean’s hypothesis is the concept of “phonological word”. Kean maintains that only those strings which are “phonological words”, or may be construed as such through a speaker’s normal linguistic capacities, will be retained in the speech of Broca’s aphasics. This claim is embodied in Kean’s principle (19): “A Broca’s aphasic tends to reduce the structure of a sentence to the minimal string of elements which can be lexically construed as phonological words in his 31rregular plurals may have overt possessive morphemes, e.g., geese’s [giysaz] . Thus it seems that Aronoff (1976) is justified in treating morphemes as sequences of segments, not as abstract features. This fact, however, has no bearing on the criticisms made here. It should also be said that there is no phonetic constraint on the number of [z] sounds that can occur in a sequence. The following are possible English words: Katz’s [katsaz] , James’s [jeymzaz] , and Isis’s [aysazaz] . AU these words end with sounds identical to the plural allomorphs but take the possessive inflection, nonetheless.

language” (p. 25). It will be shown below that “phonological words” are nothing more than secondary correlates of the major lexical categories. The concept of “phonological word” explains nothing and only obscures the true nature of Broca’s aphasia. The reasoning which led to the notion of “phonological word” may be outlined as follows: (2) (a) If a word is a N, V, Adj., or Adv., assign # boundaries to both sides of it (pp. 21, 24, 29). (b) Assign stress only to words flanked by # boundaries (p. 24). word [is] the domain over which the assignment (c) “A phonological of stress [takes] place” (p. 24). Apparently at least four related concepts are contained in (2), which may be diagrammed as follows:

phonological

stressed

word

word

word flanked by fi

N, V Adj ., Adv.

It is clear that all four concepts define one and the same class of strings because no “phonological word” exists, according to Kean’s analysis, which is not stressed, flanked by #, or a member of the major lexical categories, N, V, Adj ., and Adv. They may be represented in the following composite diagram: Figure 2

phonological word

= stressed = word = N,V word flanked by # Adj., Adv.

The domains defined by each of these concepts are coextensive. The concept of “phonological word” is not so much invalid as it is trivial since it is based on the series of reductions in (2), ultimately founded on lexico-syntactic criteria. The only words that can ever be called “phonological words” are those which are stressed, and the only words which are stressed are those flanked by + boundaries, and the only words which are flanked by # boundaries are the major lexical categories, N, V, Adj., and Adv. In the final analysis “phonological word” is defined as any string which is a member of a major lexical category. Therefore, “phonological word” should not be construed as anything more than a convenient mnemonic device for the major lexical categories. The use of “phonological word” in any other sense will inevitably lead

Kean’s interpretation

of agrammatism

65

equivocation. Calling certain members of the major lexical categories “phonological words” does not prove that Broca’s aphasia is a phonological disorder, just as calling apples “red oranges” does not prove that New York State is covered with orange groves. Moreover, the concept of “phonological word” as used by Kean has no precedent in linguistic theory. Attempts have been made over the years to define words in terms of phonological criteria. In examining several possible definitions, Lyons (1968) rejects the phonological one outright: “As for the phonological criterion: although, as we shall see later, words are phonologically determined in various ways in many languages, such phonological features are never more than secondary correlations. We shall therefore concentrate upon defining the word in purely grammatical terms” (p. 200, italics added). Chomsky and Halle (1968) take essentially the same view when they assign labeled bracketing to lexical strings. They write: It is important to bear in mind, however, that a word, in the phonologically rele-

to

vant sense, is not simply determined as a string bounded by occurences of #. The situation is somewhat more complex... Expressions such as differing, ringing, metalanguage, establishment are single words from the point of view of phonological rules, and the definition of “word” must certainly be designed to express this fact. But convention (11.5) will assign an internal #boundary in these expressions because they contain the elements differ, ring, language, establish, each of which belongs to a lexical category (p. 367).

It is clear from the above quotation that Chomsky and Halle’s concept of what a word is phonologically differs considerably from Kean’s; they maintain that it “is not simply determined as a string bounded by occurrences of #.” This statement clearly contradicts Kean’s definition of “phonological word” which was defined in precisely the way Chomsky and Halle maintain it cannot be. Kean’s definition is not in accord with the principles of phonological theory in general and English phonology in particular. Furthermore, Chomsky and Halle maintain that “word” must be defined as the string over which phonological rules apply. Phonological rules must be taken into account if any purported phonological definition of “word” is to be truly phonological and not merely a phonological correlate of non-phonological phenomena.4 The mention of phonological rules is conspicuously absent from Kean’s definition of “phonological word”, and since no other relevant phonological arguments were provided for the linguistic viability of this concept, any claims about its explanatory role in aphasiology cannot be taken seriously. 4This condition does not require that a generative grammar be taken as a performance model. However, since Kean’s argument is based on theoretical linguistic studies, it should not deviate from the theory without justification. With regard to behavioral data, what is ultimately significant is not the simple listing of structural representations but the underlying rules and strategies which speakers employ.

66

John Klosek

Kean’s view of phonology is also at odds with the lexicalist hypothesis, which envisions a separate morphological (i.e., word-formation) component operating in the lexicon (Aronoff, 1976; Halle, 1973). Word-formation rules of the lexicon may SuperIicially have the appearance of phonological rules, but they are not rules of the phonological component. They may not be ordered among the rules of the phonological component; they apply prior to them (Aronoff, 1976, p. 72; Halle, 1973, p. 15). According to Kean, it is in the lexicon that “phonological words” are defined (construed?). “The lexical rules which account for the structure of definitive relate it to another word in English, definite; the rules which characterize the morphological structure of object do not, however, relate either ob or ject to some word (ob or ject, respectively) which does occur in English” (p. 25). If Kean wants to claim that “phonological word” is indeed a phonological concept, she has to argue that the rules which Aronoff and Halle maintain are morphological are really no different from the garden variety phonological rules. This may have been possible in pre-lexicalist grammar, but not in the theory she professes to espouse. One could argue, with equal vigor, that what Kean calls a “phonological word” is really a “syntactic word” or “semantic word” since the lexicon also contains syntactic elements and semantic information. We could define a “syntactic word” as a minimal string which is a member of a major lexical category and which contains no other major lexical categories. We could also provide a semantically-based definition by claiming that major lexical categories carry greater semantic weight. The latter approach is supported by the preponderance of major lexical categories, especially referential nouns, in the speech of Broca’s aphasics. We could therefore claim that Broca’s aphasics tend to retain “semantic words” and that Broca’s aphasia is essentially a semantic disorder. These two approaches would be just as far-fetched as Kean’s if they were alleged to explain all the deficits of Broca’s aphasia. What Kean seems to have overlooked are the more recent studies (Gleason et al., 1975; Goodglass et al., 1972, among others) which have found that Broca’s aphasics’ speech differs structurally from normal speech, in addition to the fact that function words and morphemes tend to be lost. One of the reasons why the notion of “phonological word” was introduced in the first place is the claim that function words and affixes form a natural “phonological” class.’ Function words and affixes are never flanked by ’ If instead of looking at what tends to be lost, we look at what tends to be retained, Kean’s phonologically “natural class consisting of function words and some bound morphemes” (p. 9) becomes superfluous. That is, what tends to be retained is the class of major lexical items containing no internal labelled brackets.

Kean 6 interpretation of agrammatism

67

word-boundary symbols; therefore, they are not stressed, leaving them vulnerable to omission. All this would be fine were it not for the fact that there are many instances when function words are not omitted as predicted. Goodglass et al. (1967) found that agrammatic aphasics tended to retain unstressed function words (articles, prepositions, auxiliary verbs) if they occurred between stressed words in a sentence (“Dogs can bark”) and tended to lose them in initial position (“Do birds fly?“) (p. 157). Both can and do are not major lexical items, yet the former is retained while the latter is omitted in these examples. Kean’s theory predicts that both should be lost with equal frequency since neither is a “phonological word”. There are also many instances when function words are stressed, but no phonological theory (to my knowledge) exists today which can correctly assign stress to them. Can’t in Gin’t he dance?, according to Kean’s theory, should not belong to the class of stressed words (i.e., “phonological words”), but it is stressed and tends to be retained by aphasics, even in initial position (Goodglass et al., 1967). There are also many instances when members of major lexical categories are not stressed, contrary to the predictions of Kean’s theory, e.g., “This is the miin I was tklling you about” from Schmerling, 1976, p. 78). In an exhaustive examination of the relationship between stress and a number of other factors, Schmerling (1976) concludes that any theory which posits a relationship between stress and lexical category membership (cf. (2) above) is “forced to make either false or empirically vacuous claims” (p. 80). Kean’s theory has no way of assigning stress to negated auxiliaries (and preventing stress assignment to major lexical items in certain contexts); therefore, it has no way of accounting for the fact that negated auxiliaries tend of be retained in the speech of Broca’s aphasics. Kean’s only way out would be to claim that negated auxiliaries are “construed” to be “phonological words”. However, the notion of construal is itself illicit since it is empirically unfalsifiable. Furthermore, it is based on purported speech errors which do not even remotely resemble the types of errors made by Broca’s aphasics (Jason Brown, personal communication). It seems to have been included in the theory for the express purpose of handling any and all exceptions. Kean has no valid arguments of any kind as to why her solution should be chosen over previous explanations other than the assertions that her theory is “the only possible interpretation of the syndrome” and it is “the only systematic account which is consistent with what is known of the structure of the language processing system” (p. 10). As it has been shown above, there is good reason to believe that Kean’s is not even a possible interpretation. As far as the language processing system is concerned, Kean’s is a languagespecific account, based on nothing more than a handful of speech errors. As an explanation of speech errors, it is inferior to the more general analyses

proposed bly apply study of language

by Fromkin (1971, 1973) and Garrett (1975), which could probato all languages, not just English. Moreover, it is doubtful that the speech errors alone is sufficient to reveal the true nature of the processmg system.

References Aronoff, M. (I 976) Word Ebrrnat~on an Generame Grommur, Cambridge, Mass., MIT Press. Rlooml’ield, L. (1933) Lmguge, New York, Holt, Rinehart and Winston. <‘hornsky, N. (1Yb5) Asprc~s of the Theory ofSyntax, Cambridge Mass., MIT Press. Chomsky, N. and Halle, M. (1968) lhe Sound Putrern ofEngfish, New York, Harper and Row. Fromkin, V. (1971) The nonanomaious nature of anomalous utterances. Lang., 47, 27-52. lromkm, V. (1973) Slips of the tongue. Sci. Amer., 229, 110-l 16. Garrett. M. F. (1975’) The analysis of sentence production. In Bower, G. (ed.), The Psychology of Learning and Motrvarion: Advances in Research and Theory, 9, New York, Academic Press. Gleason, J. B., Goodglass, Ii., Green, E., Ackerman, N. and Hyde, M. R. (1975) The retrieval of syntax in Broca’s apnasia, Brain Lang., 2, 45 I-471 Goodglass, H. and Bcrko, J. (1960) Agrammation and mtlectional morphology in English, J. speech hear. Kes., 3, 257-267. Goodglass, h., Fodor, I. and Schulhoff, C. (1967) Prosodic factors in grammar-evidence from aphasia, J. speech hear. Kes., IV, S-20. Goodgtass, H., Gleason, J. B., bernholtz, N. A. and Hyde, M. R. (1972) Some linguistic structures in the speech of a Broca’s aphasic. Lbrfex, 8, 191-212. Goodglass, H. and Hunt, J. (1958) Grammatical complexity and aphasic speech, Word, 14, 197-207. Good&ass, H. and Mayer, J. (195X) Agrammatism in aphasia, J. speech hear. Dis., 23, 99-l 11. Halle, M. (1973) Prolegomena to a theory of word-formation, Ling. Inq., 4, 3-16. Kean. M-L. (1977) The linguistic interpretation of aphasic syndromes: Agrammation in Broca’s aphasia, an example, Cog., 5, 9-46. Kolk, II. H. J. (1977) The linguistic interpretation of Broca’s aphasia. A reply to M-L. Kean, Cog., 6, 249-262. Lyons, J. (196X)Introducfion to Theoretical Linguistics, Cambridge, Cambridge Univ. Press. Schmerling, S. F. (1976) Aspects of EngZish Sentence Stress. Austin, Univ. of Texas Press.

Cognition, 7 (1979) 69-83 @Elsevicr Sequoia S.A., Lausanne

Discussion - Printed

in the Netherlands

Agrammatism:

A phonological

MARY-LOUISE Univefsity

deficit?

KEAN

of California,

Irvine

*

The comprehensive linguistic analysis of any systematic language deficit will have two parts: (a) a characterization of that component of the language faculty which has been impaired, and (b) a characterization of the observed behavior in terms of an interaction of the impaired component with those components of the language faculty which remain intact. That this is so is a matter of logical necessity; for any complex function characterized as a partially ordered set of components, impairment ro any component of that system may well have the effect of distorting the normal realizations of those components which remain intact.’ It is this general view which lies behind the analysis of the language deficit known as agrammatism which is put forward in Kean (1977a).’ Agrammatism is characteristically described as the tendency to fail to exploit so-called functors in all aspects of language use, while at the sarne time the so-called major class elements of language tend to be retained. Based on an explicit model of the language faculty, in the papers cited it is proposed that this deficit is (a) to be grammatically characterized with respect to phonological representations, and (b) that to a fairly large extent aspects of agrammatism should not be considered in terms of the functioning of some impaired component of the system, but rather in terms of contributions of intact components of the language faculty. Recently both parts of this analysis have been challenged (Kolk, 1978; Klosek, 1979).3 While both Kolk and Klosek do not *This work was prepared while the author was at MIT, and its preparation was supported by NIMH fellowship l-F32-MH0718941. Requests for reprints should be addressed to Mary-Louise Kean, School of Social Sciences, University of California, Irvine, Irvine, Calif. 92717, U.S.A. ‘See Kean (forthcoming b) for an extended discussion. ‘The analysis of agrammatism is also discussed in Kean (1978, forthcoming a, b). Unless otherwise noted, cited page references refer to Kean (1977a). 31n this paper I will only be concerned with the analysis of agrammatism put forward in the references cited, and not deal with the other deficits which are discussed in Kean )1977a, 1978). For a variety of reasons, including evidence such as that KoIk cites, I would not now argue that in Broca’s aphasia there is an “impairment [which] is distributed across the entire domain of phonology” @. 41). In contradiction to this claim, in Kean (1977a, 1978) it is argued that dysprosody does not reflect an impairment to the language faculty, but rather arises as a consequence of such factors as the rate at (Continued overleaf)

70

Mary-Louise Kean

accept the particular analysis, neither puts forward any arguments against the general framework nor against the particular model of the language faculty which is assumed; neither provides arguments which are directed toward the particular analysis nor any explicit counterproposal for the interpretation of agrammatism. This being so, it can be assumed that they accept both the general framework and the model of the language faculty, and that what is at issue is whether or not the specific proposal is justified in terms of both the general framework and the particular model. As the justification for the analysis in question relies on a particular hypothesis as to the structure of the language faculty, before turning to the details of the analysis of agrammatism it is necessary to at least outline those assumptions about the structure of the human language faculty. It is assumed that the language faculty is to be conceptualized in terms of a grammar, characterizing an individual’s knowledge of his language, and a processor, characterizing the computational systems by which such knowledge is exploited in use. The grammar consists of a partially ordered set of components, rule systems, which are universally characterizable (and distinguishable) in terms of the formal properties of their rules and in terms of the classes of substantive elements those rules apply on. For any well-formed string in a language, the output of any component of the grammar is a representation of that string with respect to that component. It follows from the fact that the components of the grammar are universally determined that the characteristics of their associated representations are also universally determined, The set of representations of a string generated by the components of a grammar is the structural description of that string; that is, what a grammar does is generate the structural descriptions of strings. It is assumed that the structural descriptions of strings are realized in language use. From this it follows that the grammar is a component of what has been called a performance system. The representations generated by one component of the grammar serve as the inputs to the ‘next’ component. As each component of the system has a distinct function, the input to each must be structures which are appropriate

which agrammatic aphasics typically speak. This is an empirical hypothesis, and at this time I know of no data which refutes it. In Kean (forthcoming b, c), it is argued that the data on segmental (phonemic) paraphasias (notably that of Blumstein, 1973, and Lecours and Lhermitte, 1969) pattern with the segmental substitutions found in normal spontaneous speech errors (Shattuck-Hufnagel and Klatt, forthcoming), and that these in turn pattern with the pairs of segments which may alternate grammatically under a single phonological rule (Kean, 1975, 1977b). In sum, these data suggest that segmental paraphasias are not apparently analyzable as true impairments to an individual’s phonological capacity.

Agrammatism: A phonological deficit?

11

for the application of its rules. If the rule system of some component is sensitive to some substantive distinction(s) among the items in a string, then those distinct classes of elements must be uniformly and systematically distinguished in the input representation. If the output representation of some component does not contain all the substantive partitions of the items in a string which are required by the rule system of the ‘next’ component, then there must exist a universal convention which applies to that output to instantiate the needed substantive partition. While the substantive partitions of the items in a string at any level of representation are grammatically motivated, that does not entail that those classifications of items do not play a crucial and systematic role in non grammatical areas of language use (e.g., lexical access); any realized substantive classification of the items in a string may be exploited in non grammatical components of the processing system. Given such a conception of the language faculty, in considering any deficit which appears to involve the systematic differentiation of classes of items in a string, it is appropriate to consider what level of grammatical representation provides the appropriate substantive partition of elements into these classes, keeping in mind, of course, that what we take to be a single deficit may prove under analysis to be a family of deficits and therefore require characterization in terms of several levels of representation. Agrammatism is just such a deficit; any account of agrammatism will have to include a characterization of the elements of a sentence into those items which tend to be unexploited and those items which tend to be retained, and that partition of the data will have to be principled and systematic in terms of some explicit linguistic hypothesis as to the substantive partitions of items in a string with respect to some component representation(s) of the structural description of a string, the structural description including the set of possible partitions of elements. In Kean (1977a) it is argued that those representations which provide the input to the phonological component of the grammar provide the required partition of substantive elements which is necessary for any analysis of agrammatism, and that characterizations with respect to other levels of systematic representation are ad hoc. The argument for this claim rests on consideration of what substantive partitions of elements are available at all levels of representation. That no level save the phonological offers the necessary partition of items can be illustrated by consideration of the representations which the grammar assigns to strings; let us consider a case in point. The structure (la) is derived through the operation of two distinct sets of rules, the phase structure rules and the lexical rules of the base; the phrase structure rules provide a characterization of the sentence in terms of a hierarchical constituent structure (i.e., the labelled bracketing of (la)) and specify

72

Maty-I,ouise Kean

(1)

The boys played on the bank4 a.[[ s NP [ Specfhel $boy,

WI 1 [VPIA,lx+wfl [VPlvl &,$,onl [NP [Spec the] +&f

+&f

INbank 1111 b.[[

[

the1 [Nboy,

S NP Spec +M

WI 1 Lvp[VPW, w;~sfl[pp[p~nl [

[

the]

NP Spec

+3rd

+dcf

INbank 1111 the]

c. t#t# [

S NP Spec

[#J&bo~#l s#I [ffp [A+[hpv#l d#l [$prfon#l

+dcf

‘EP’Spec +def

d. I#+[# [

the] [#Nbank#l #I #] #I #I

S NP Spec

the] [4&[fib&-t] 4 Itp I#V[@W$l d#l I#p[ponl

+dcf the] [#Nbank#l #I #I #I #I %P[Spec +def more simply stated:

LN#bank4 #I #I 11 certain features on that bracketing (in this case the features [+ plural] and [+pastl ). The lexical rules apply to insert the lexical items into this constituent structure (i.e., bo_b?,play, on, bank). The structure (lb), the surface structure, is derived from (la) by the application of syntactic rules, in this case notably the rules of Affix Hopping and Subject-Verb Agreement, which specifies the verb for person and number. If we consider the items in this “The reader will note that in (1 a, b) the articles appear. It has been argued (see Fiengo, 1974) that articles, like the tense markers, etc., are grammatical formatives which are syntactically represented in terms of features, and which receive their phonological specifications through late spelling rules. Such issues in the analysis of articles have no bearing on the matter at hand.

,&rammntism:

4 phonollpictrl deficit?

73

string which are typically retained in agrammatism we note that all those items are the syntactic heads of phrases and are inserted by lexical rules. However, while this is so, it does not provide any argument for an analytic distinction between what tends to be retained and what tends to be unexploited in agrammatism. Prepositions. particularly short ones, tend to be unexploited in the spontaneous language use of aprammatics; prepositions, like the oy1 in this example, are both the syntactic heads of phrases and items inserted by lexical rule: that is, they are syntactically in the same class as nouns and verbs. Therefore, there is apparently no systematic account of what is typically retained in terms of the syntactic structure of the sentence. By the same token, there is no available systematic syntactic account of what is lost either; the preposition on and the grammatical formatives characterized in terms of the features plural. past, etc. are tokens of distinct types. Considering (la) we see on the one hand we are dealing with a lexical formative derived by lexical rules, and on the other hand with formatives which are characterized by the phrase structure rules. The surface structure (1 b) offers no more promise of a principled partition than does (1 a). There is nothing in the representation which provides a classification of on, the definite articles, and the plural and tense/person/number grammatical formatives as a single coherent set in contrast to the other items in the string, boy, play, and bank. To attribute the distinction between these two classes of items to the syntax is, therefore, simply to resort to an ad hoc restatement of the data.s In the example (l), those elements which belong to the class of items which is typically unexploited in agrammatism are of two types: on the one hand, there are elements which are lexical items inserted by rule, and on the other hand, grammatical formatives specified by the phrase structure rules (abstracting away for the moment from the person and number specifications on the verb in (1 b)). There is a third type of element which also is typically unexploited in agrammatism, what we can call pure inflectional morphemes. This class of elements consists entirely of those elements which are assigned to a string in virtue of its hierarchical structure and which must be ordered after (at least some) syntactic transformations. A ready example of such morphemes comes from consideration of a class of nominalizations.

‘Klosek asserts that “one could argue with equal vigor [to define the set of items retained in agrammatism as] ‘syntactic word[s] ’ ... We could define a ‘syntactic word’ as a minimal string which is a member of a major lexical category and which contains no other major lexical categories”, a view which overlooks the fact that Preposition is a “major lexical category” (a non syntactic notion, it might be noted; see also footnote 8), and that prepositions like, e.g., nouns, are the syntactic heads of phrases.

74

Mary-Louise Kean

The object of a derived nominal, when not preposed, is invariably preceded by of in English, as in the destruction of the city. When such a nominal is passivized and the object is preposed it invariably is marked by the genitive (“possessive”) affix, as in the city’s destruction. In this case we are dealing with syntactic markers whose distribution and occurrence is totally predictable from the structure of the sentence. Note that these grammatical formatives are, in this regard, quite distinct from the plural and tense formatives of example (1) which are generated in the base - one cannot predict from the structure of the sentence whether or not the subject will be singular or plural (i.e., boy or boys) or whether the verb will be past tense or present. The of in the destruction of the city is a grammatical formative, distinct in kind from the lexical prepositions (including the lexical of of he took aduantage of Martha), and is, like the genitive in the above example, assigned automatically by rule. Thus, in terms of the syntax of sentences we find that the formatives which tend to go unexploited in agrammatism are from three distinct classes of formatives: the set of lexical formatives, the set of phrase structure generated grammatical formatives, and the structurally predictable class of grammatical formatives called pure inflections here.6 In contrast to this set of items we have those items which tend to be retained, a class the members of which are all lexical formatives. That is, to reemphasize the point, those items which are retained in agrammatism as well as one class of the items which are unexploited belong to a single syntactic class, the set 6 Klosek observes, correctly, that my use of the terms derivational and inflectional in Kean (1977a) is not in conformity with traditional grammatical practice; the same is true of the use of these terms in Kean (1978). Traditionally, only those morphemes which are purely lexical are called derivational, while both phrase structure generated formatives (e.g., plural) and pure inflections are called inflectional morphemes. In Kean (1977a, 1978) the term derivational is applied not only to the lexical derivational formatives, but also to the phrase structure generated formatives, in particular the plural. I apologize for any confusion this infelicitous terminology may have led to, and have attempted to avoid that problem here by referring to pure inflections, on the one hand, and phrase strucfure generated formatives, on the other; derivational is now restricted to its traditional use with reference to lexical morphology. It is one thing to note poor use of terminology, and quite another to show that the distinctions drawn are incoherent. First, Klosek provides no argument that one should not distinguish pure inf[ections from other formatives; this distinction, while not terminologically pervasive, is certainly analytically pervasive (cf., Chomsky, 1965, 1970). Secondly, while objecting to my claim that plural must be a feature represented in the lexicon, he fails to provide any account of irregular plurals in which that feature is not mentioned in the lexicon; the accounts in the literature with which I am familiar all provide lexical analyses of the irregular forms. Thirdly, while Klosek claims that the distinction I draw demonstrates that I have adopted a “pre-lexicalist grammar” and “not the lexicalist theory she he overlooks the fact that such distinctions are found in lexicalist syntax professes to espouse”, (Chomsky, 1970). Finally, he does not make reference to data such as that reported in Forster (1978) which indicates that plurality does have some influence lexically in processing. In sum, Klosek has only noted a terminological confusion, and has failed to address the analytic substance of the distinction which was explicitly drawn.

Agrammatism:

A phonological deficit?

15

of lexical formatives. Not only is it impossible to give a uniform syntactic account of what is retained, it is also impossible to give a systematic syntactic account of the set of items which is typically unexploited as that set consists of two types of grammatical formatives as well as a subset of the set of lexical formatives. Before turning to the question of whether or not phonological representations do provide an appropriate partition of the elements of a string into those which tend to be retained in agrammatism and those which tend to go unexploited, let us briefly consider one other type of representation which might provide the appropriate partition of elements: semantic representations.8 If one considers the level of logical form, it is evident that that level

7Kolk suggests at the end of his paper that the “evidence” suggests that “one should look for syntactic instead of phonological models [for the analysis of agrammatism] “. He fails, however, to enlighten us on what particular model of syntax he has in mind, and does not address the problems which a syntactic analysis would face under the particular theory of syntax which has been consistently assumed in my work. As to the evidence he cites, it is far from compelling. First, there is the data from Goodglass et al. (1972) in which the patient says the girl tall and the boy little rather than the girl was taller, the most common response by normal subjects. As John Marshall (personal communication) has pointed out, it is an option available to every normal speaker of English to say the girl wus tall and the boy was little; thus, the “typical response” cited by Goodglass et al. reflects a syntactic option available to any speaker, modulo the agrammatic deletion of the copula. Secondly, Kolk cites some data from his own research. The experiment in question involved training agrammatic aphasics on so-called eager type sentences (using story completion); having been trained on such constructions the subject was then given a context in which the completion required a so-called easy type sentence. While they are superficially very similar, eager and easy constructions such as John is eager fo please and John is easy to please are syntactically quite distinct, a fact long noted in the literature. Having been trained on eager constructions, the subjects in Kolk’s study did not demonstrate a transference of training to the easy constructions. Given the superficial similarity of these constructions, it can plausibly be assumed that the failure of transference was a function of the fact that the two types of constructions are syntactically distinct. For such a distinction to interfere with performance on the task (i.e., block transference), it would, it would seem, be necessary that a subject’s syntactic capacity be sufficiently intact to be sensitive to the distinction. Thus, these results do not in any way support the suggestion that we “look for syntactic instead of phonological models”, at least in so far as accounting for the impairment in agrammatism is concerned. Data such as the errors made by Kolk’s subjects are, of course, quite interesting, and demanding of careful analysis. Any such analysis will, however, have to be based on some explicit hypothesis as to the nature of the impairment which leads to agrammatism; the analysis would then account for the errors in terms of an interaction of the impaired component(s) of the language faculty with those which remain intact. These errors may prove to be particularly interesting and enlightening because they all are deviances from a structurally restricted set, and, consequently, may be more readily subject to close analysis than errors from a structurally open set. aTaking another approach, one might try to argue that agrammatism cannot be accounted for in terms of a single representational level, but rather must be analyzed in terms of a family of representational levels. Under such an approach one might rely on the syntax to provide the distinction between both phrase structure generated and pure inflectional grammatical formatives and all other items, and look to other components of the system to distinguish lexical prepositions from the other categories of lexical items. One obvious candidate under such an approach would be the structure of lexical (Continued overleaf)

76

Mary-Louise Kean

will not provide the relevant partition. While determiners such as the carry logically consequential information (definiteness and cardinality), such is not the case with the pure inflections. If we look to lexical semantics, under either a view such as that espoused by Katz (1972) where meanings are to be represented in terms of semantic features, or that proposed in Fodor, Fodor, and Garrett (1975) where it is claimed that meanings are represented in terms of meaning postulates, there is no distinction in form between the meanings assigned to lexical items of any one category as opposed to those assigned to all other categories. That is, there is no theory of lexical semantics which distinguishes lexical prepositions from elements in other major lexical categories.9 Having said something of the levels of representation which do not provide any systematic partition of the items in a string into the two relevant classes for the characterization of agrammatism, let us turn now to those representations which serve as the input to the phonology, and reconsider whether or not the evidence supports the claim that such representations do in fact provide the appropriate partition of elements. The syntactic surface structure of a string provides the foundation for such representations. Observations such as noting that contraction is possible in the second conjunct of Bill 1: 1 in the kitchen and Fred f 2 1in th e 1sundry morn, but not possible in

items themselves; if prepositions must be distinguished structurally (other than by category stipulation) from the other categories, then the derivational morphology of the lexicon would provide a foundation for making the appropriate representational partition. However, prepositions have in all relevant respects the same properties as the other categories. Thus, for example, looking at an earlier stage of English, we find prepositions and members of other categories all derived off the same stem beyond and yonder > be + yond and yond + er, behind and hindmost > be + hind and hind + most. Similarly one finds elements derived off prepositions just as one finds elements derived off members of other categories, thus, the noun behind from the preposition behind just as one finds, for example, the noun permit from the verb permit. While English is rather impoverished with regard to such derivations involving prepositions, they do exist in English and therefore are sufficient for making the case that in terms of the derivational structure of words the category Preposition exhibits essentially the same properties as do the other lexical categories. ‘Klosek claims that one could come up with a semantic account of agrammatism, one in which the items retained were “semantic words”, the set of semantic words consisting of items from the categories N, A, V, Adv. He does not, however, provide any reference to any semantic theory in which lexical items of these categories are systematically distinguished from lexical propositions. While he says that N, A, V, and Adv “carry greater semantic weight” than, presumably, prepositions, for example, the notion of “semantic weight” is undefined, and one is left wondering what the measure to be invoked is. It should be noted that it is incumbant on anyone advocating such a position to develop a definition whereby a noun such as thing (a pseudo-designator) or an adverb such as very (an intensifier) will carry more “semantic weight” than prepositions such as up and from which are, by Klosek’s stipulation, not “semantic words” and, therefore, of less “semantic weight”.

Agrammatism: A phonological deficit?

II

the second conjunct of Bill f ti 1 in the kitchen and Fred 1 i: 1 too serve to show that the possible phonetic realizations of any sentence as assigned by the phonology are crucially contingent on the constituent structure of sentences and not simply on the linear string of segments which make up a string. The rules of the phonological component are also sensitive to the category membership of items - thus, for example, in English the stress retraction in Latinate deverbal nouns with monosyllabic stems, e.g., pe’rmit. However, categories and constituent structures no not provide all the information required for distinguishing substantive classes of elements with respect to the operations of the phonology. The rules of the phonological component of a grammar must distinguish two classes of items, clitics and non clitics (the latter also being called phonological words). lo Before turning to some examples of why this distinction is required for the proper application of phonological rules, what the distinction actually is will first be presented. The clitic/non clitic distinction is instantiated on syntactic surface structures subsequent to the spelling out of grammatical formatives; the distinction is instantiated by a universal convention (i.e., a convention which applies automatically in all languages to such representations). The convention assigns what are known as wordboundaries (#‘s) to a string: Assign a word-boundary, #, to the left and right of every lexical category and every category dominating a lexical category. I1 (lc) is a representation of the type yielded through this convention. In English, by a readjustment rule, the boundaries to the right and left of monosyllabic prepositions are erased.‘* Thus, (Id) is derived; (Id) is the

“Klosek states that my “concept of ‘phonological word’ has no precedent in linguistic theory”. If this were so, then it would also, of course, have to be true that my concept of clitic would also have to be with “no precedent in linguistic theory” since everything which is not a component of a phonological word is a clitic, and vice versa. The term clitic is taken from traditional grammar. As to the linguistic precedent for distinguishing clitics from non clitics (or phonological words) references abound ranging from Morris-Jones (1913) to Selkirk (1972), to cite a traditional and a more recent work. As I noted, the notion of a phonologicalword “does not correspond to the ordinary notion of the term [word] ” (p. 22). Against the notion of a phonological word Klosek cites a reference from Lyons (1968) and a reference from Chomsky and Halle (1968). The former quotation is taken from a discussion of developing a definition of word to cover the ordinary usage of that term, and the latter quote is taken from a discussion of “words” at the grammatical “word-level” where strings such as the boys are single words. There is a systematic grammatical distinction between word-level words and phonological words: it is therefore not surprising that those two systematically distinguished classes “differ considerably”. The whole question of grammatical definitions of words with respect to phonological structure is taken up in detail in Kean (forthcoming a). See also footnote 15. ” Kolk states that this convention “is called ‘lexical construal’ “. To the best of my knowledge, there are no references in the literature where the boundary assignment convention is called “lexical construal”; the references to “lexical construal” with which I am familar all involve discussion of non grhmmatical processes of sentence production (Kean, 1977a, 1978; Kean and Garrett, forthcoming). See Chomsky and Hahe (1968) and Kean (forthcoming a) for discussions of readjustment rules.

78

Mary-Louise Kean

phonological representation of the sentence (l), that is, the representation on which the rules of the phonological component apply. A phonological word, i.e., a non clitic, is a properly bracketed string of the form # _ -#, where contains no occurrences of #; all the other formatives in the sentence are clitics. Thus, in (Id), boy, play, and bank are phonological words, while the, s, d, 0~1, and the are clitics. In English, the clitic/non clitic distinction is crucial to the proper formulation of the rules which assign stress to words: clitics are not assigned word stress, and make no contribution to the stress of words. In positive active declarative, sentences, clitics do not participate or contribute to the stress pattern of the sentence. Thus, whether or not an item participates in stress phenomena in the neutral unmarked case in English is a diagnostic as to whether or not that item is a clitic.13 All phrase structure generated affixes and pure inflections are clitics in English. It should, perhaps, be emphasized that the clitic/non clitic distinction is not a stress/stressless distinction universally (or even through the full range of English constructions), rather it is a universal distinction which just happens to interact with stress assignment in English.14 To bring home the point that

13Klosek states that “the mention of phonological rules is conspicuously absent from Kean’s definition of ‘phonological word’ “. Given that the notion of a phonological word is developed in Kean (1977a) in the context of a discussion of some of the rules involved in stress assignment in English, this comment is without foundation. The sentences discussed in Kean (1977a, 1978) were active affirmative declaratives, and the rules of stress assignment in question were those that apply to such constructions. That items which are clitics receive stress in questions such as do hirds fly is a datum which has bearing only on the rule for stress assignment in questions, a topic not discussed at all in Kean (1977a). Noting such constructions in Kean (1978), in the context of a discussion of Goodglass et al. (1964) and Goodglass et al. (1967). it is noted that “while it is not obvious that any metrical study of the speech of Broca’s aphasics would yield interesting generalizations about the pattern of omission, to ignore the possibility that such generalizations might emerge seems unwarranted” (p. 115). Thus, to the extent Klosek’s reference to such questions is relevant, that relevance is to point out that they are suggestive of an area for further research, a position with which quite obviously 1 agree. As to Klosek’s mention of the fact that can ‘t is stressed in can ‘t he dance, since can ‘r is not a clitic (Selkirk, 1972) the example is simply beside the point. 14Kolk states, “More specifically, it [a phonological word] is ‘the domain over which stress assignment takes place’ (p. 24)“: later, in his discussion of what I have to say about predictions having to do with Russian, he notes that “stress relevance is no longer a criterion” for determining whether or not an item is a phonological word. Stress relevance was never claimed to be universally applicable in determining whether or not an item was a phonological word in Kean (1977a) (or any of my other papers). Kolk’s quote from me cited above is taken from a sentence which begins, “Recall that we said a phonological word was defined in English as the domain...,” [emphasis added]. In his further comments on Russian, Kolk notes that “in Russian, unlike English, the inflectional affix of the third person singular present tense form of the verb does play a role in the assignment of stress.” In Russian many affixes and stems have inherent (i.e., lexical or morphological) stress, and stress assignment applies to select out the rightmost of such stresses in a form with one or more inherent stresses while in forms with no inherent stresses, the leftmost syllable is stressed (IIalle, 1971). Given that Russian is a very different language from English, it is not surprising that its stress rule is quite different, nor that the domain of stress assignment includes clitic affixes.

Agrammatism: A phonological deficit?

79

the clitic/non clitic distinction is not a stress distinction, consider the following two cases of rules which have nothing to do with stress but which are sensitive to the distinction between clitic formatives and other formatives. In Klamath, an American Indian language, proclitics (clitic prefixes) trigger vowel deletion when the succeeding item is of the form CVCV..., and vowel reduction when the succeeding item is of the form CVCC..., whereas regular prefixes do not trigger either reduction or deletion baga, ‘smokes’, snapgu (causative)papgu (distributive), but, note pagpuga (intensive); bo~wa ‘drinks’, hosbanwa (causative), bobanwa (distributive), but note, bonbonwa (intensive) (Kean, 1974)). In Russian, word-final consonants are voiceless when the following word begins with a sonorant segment, e.g., v,os ### atcu, ‘he drove his father’. In Russian, as in English, there is a readjustment rule which erases the #‘s flanking monosyllabic prepositions. One does not find consonant final devoicing in clitic prepositions, but one does in non clitic prepositions, e.g., b,iz # atca, ‘without a father’, but p,er,it ## atcom, ‘before a father’.” The distinction between phonological words and clitics seems to closely fit the distinction between what is retained and what is unexploited in the language use of agrammatics: those items which are typically retained are phonological words, and those items which are typically unexploited are clitics. The claim of the phonological analysis of agrammatism is then simply this: phonological representations are the only systematic level of representation within a grammar which properly partitions the items in a string into those items which are typically retained and those which are typically not retained. Needless to say, in taking such a position one is making a claim about a range of data which extend far beyond the currently available corpus; this is in no way either surprising or objectionable. Rather, quite to the contrary. To be content with such notions as “major lexical category” and “functor” without assigning to such terms systematic grammatical definitions is to be content with ad hoc analyses of the data available, where one can, for any given datum, assign it stipulatively or operationally to either one class or the other.16 “This example is taken from Chomsky and Halle (1968), appearing in the section entitled “The Boundary ‘#’ and the Notion ‘Word’ ” (pp. 366-370). While Klosek quotes a passage from this section (noted in footnote 9) which has no bearing on the coherence of the notion phonological word, he ignores an example which is directly addressed to the distinction between phonological words and clitics that appears three pages later. For some additional comments on Russian prepositions in the context of a discussion of phonological words and clitics, see Kean (forthcoming a). 16Klosek states, “In the final analysis ‘phonological word’ is defined as any string which is a member of a major lexical category [these stipulated to be N, A, V, Adv]. Therefore, ‘phonological word’ should not be construed as anything more than a convenient mnemonic device for the major lexical categories.” This is to put the cart before the horse; phonological word, unlike the notion major lexical category, is a term with a principled grammatical definition; therefore, if either is a mnemonic for the other, it is major lexical category which is a mnemonic for phonological word. It should, however, be emphasized, that like all mnemonics this one does not really properly characterize the domain of elements in question.

80

Mary-Louise Kean

Such an approach fails to make any attempt at reaching a level of descriptive adequacy, much less explanatory adequacy; it is without empirical force in that it makes no predictions which extend beyond the corpus at hand. In the case at hand the analysis goes considerably further than the available data in making claims about what is and is not likely to be retained in agrammatic language use. To take one example, not all the lexical items in a language are monomorphemic; in each language there are lexical rules which concatenate morphemes to make lexical items. Some of the morphemes on which such rules apply are invariably bound, that is never occur freely as “words” in the language. In English, for example, the suffixes -YZCSS and -ive are such morphemes; +zess is, in English, a clitic suffix definite # ness, while -ive is not, definite + ive. I7 It follows from the hypothesis that agrammatism involves the selective retention of phonological words that -yless should, like other clitics, go unexploited, ceteris paribm. While there may well be no data available at this time which bears on whether or not such detailed predictions are in fact supported, that in itself is hardly reason for rejecting the analysis. That the analysis does make such predictions is a virtue in that it means the analysis has empirical content and suggests new avenues of reaserch for furthering our understanding of the nature of agrammatism. It is, of course, a long noted observation that the likelihood of retention of clitics varies. In Kean (1977a, 1978, forthcoming a) it is suggested that this variation may simply be a reflection of one of those aspects of the language faculty which remains essentially intact. The argument for this was based on the analysis of normal speech errors, in particular, those errors in which two lexical items exchange and in which affixes which are bound to those items may be stranded (e.g., ‘you ordered up ending).18 The argument presented was this: in normal speech errors some suffixes are more likely to be stranded in exchange errors than others; the likelihood of failure to “In Kean (1977a, 1978) it is suggested that definifive could by a non grammatical and non phonological process of lexical construal be realized as the morphologically simpler lexically related form d&ire; the same process of lexical construal is appealed to in the suggested analysis of Russian. Kolk is simply in error in claiming that my analysis of agrammatism in Russian requires postulating construal capacities distinct from those required for English. As is noted in Kean and Garrett (forthcoming) one must postulate mechanisms of lexical construal to account for normal spontaneous speech errors. “What Klosek refers to as the “purported speech errors” cited as examples in presenting this argument all appear in the references cited in Kean (1977a) or come from the “MIT Corpus” (the data on which Garrett’s work that Klosek favorably refers to is based). Furthermore, as it was never suggested that aphasic errors and normal speech errors were superficially alike, Klosck’s comment that the normal errors cited “do not even remotely resemble the types of errors made by Broca’s aphasics” is totally beside the point. It was, however, suggested that there was a similarity in variation of clitic errors in the two types of data.

Agrammatism:

A phonological

deficit?

81

exploit a clitic in agrammatism parallels the likelihood of a clitic’s being stranded; therefore, it would seem that the degree of adhesion of clitics to words is the same in agrammatism as in normal language use, and, consequently, the variation in clitic realization in agrammatism must reflect some intact component of the language faculty. l9 Secondly, it was suggested that general hierarchy of clitic adhesion, starting with those clitics which are least adherent, went from pure inflections, through the phrase structure generated clitics affixes, to purely lexical clitics. As with the hypothesis that the phonological distinction between clitics and non clitics provides the appropriate general partition of the data on agrammatism, these hypotheses went beyond the corpus available in the predictions made. For example, the third person present singular verb affix -S in English is both a phrase structure generated affix (in terms of tense) and a pure inflection (in terms of person and number); therefore, it should fall between the pure inflection genitive -s, and the phrase structure generated plural -s. There is, to my knowledge, no data in the literature which bears on this. ‘O Whether or not such hypotheses will “Kolk notes that Garrett (1975) attributes the exchange errors involving stranding to a syntactic level of processing. Since what is at issue in the analysis of the normal error data with respect to agrammatic language use is a purportedly normal aspect of the agrammatic’s language use, whether or not this level is a syntactic level in processing is of no consequence to the point at issue. Garrett (1975) does not consider the possibility of these errors involving the phonological level of representation; however, Kean and Garrett (forthcoming) do consider this possibility, and conclude that phonological representations do provide an appropriate partition of elements, the clitic/phonological word partition. Klosek faults the analysis of the normal errors in Kean (1977a) on two grounds: it is based on only a “handful” of errors, and it is a “language specific account”. and he concludes that it is “inferior to the more general analyses proposed in Fromkin (1971, 1973) and Garrett (1975)“. The references he cites are the source of the “handful” of errors the analysis was based on. Klosek cites no errors from those corpera (or any other), and, consequently, provides no evidence that the examples cited are not properly representative of the type in question. As both Fromkin and Garrett direct their attention to the analysis of English speech errors, it is unclear in what, if any, way their analyses are not equally “language specific”. Finally, as both Fromkin and Garrett are concerned with avariety of different types of errors, t&e is an obvious sense in which their analyses are “more general”, though this is hardly germane. Curiously, Kolk suggests that the data in Goodglass and Hunt (1958) “might be evidence that goes directly against it [the argument to hierarchy of affllal adhesion] “; Goodglass and Hunt is, however, one of the sources cited in Kean (1977a) in the comparison of normal speech errors with agrammatic data. The Goodglass and Hunt data do support the distinction between the plural and the other two -s suffixes; they do not, however, support the distinction between the verbal -s and the pure inflection genitive -s. That this is so does not argue against the hierarchy since the experiment in question did not control for the fact that in addition to the pure inflection genitive, the genitive also may be generated in the base by phrase structure rules (as in the “possessive”). While Kolk is correct in noting that I claim that there is “only one phonological structure involved” with these -s affixes, that is irrelevant to the analysis in question, an analysis explicitly based on distinguishing the different sources of these affixes, all of which are noted to be non phonological. In relation to the omission of the -s affixes and to the past tense affix -ed I suggest, as Kolk observes, that the sonorance hierarchy might provide the basis for a partial account of observed omissions. What I said was that “the linkelihood of their [the -s and -(e/d affixes] omission appears” to correlate with the sonorance hierarchy (p. 17, italics added), and cited some data from Goodglass et al. (1972). I

(Continued overleaf)

82

Mary-Louise Kean

prove to be viable bases for an extended and more detailed analysis of agrammatism remains to be seen; in the absence of any counterarguments and any research results relevant to the predictions made, insofar as these suggestions do provide an account for the data available they must be accepted as viable hypotheses, suggestive of further lines of research. The analysis of agrammatism in terms of phonological representations makes the claim that the basic bifurcation of the agrammatic data into the set of those items which tend to go unexploited in language use and those items which tend to be retained is the same as the distinction embodied in the clitic/phonological word distinction found in phonological representations. This is an empirical hypothesis, and as such is subject to falsification. It has also been argued here that other levels of grammatical representation do not provide a proper partition of the agrammatic data. This analytic claim is also subject to falsification. However, it should be clear that to show the latter analytic claim to be in error would not in itself vitiate the former hypothesis, rather it would leave us with competing analyses. Similarly, the suggestion that the variation in the likelihood of realization of clitics is essentially a reflection of the functioning of an intact component of the language faculty is a suggestion which is subject to empirical examination; it is also logically independent of the other two claims. The analysis is made within the context of a specific theory of the structure of the human language faculty. If that view were shown to be wrong, then of course the analysis would lose its foundation. By the same token, the analysis cannot be shown to be wrong by simply proposing a competing analysis which is based on a distinct theory.

References Blumstein, S. (1973) A Phonological Investigation of Aphasic Speech. The Hague, Mouton. Chomsky, N. (1965) Aspects of the Theory ofSyntax. Cambridge, MIT Press. Chomsky, N. (1970) Remarks on nominaliiations. In R. Jacobs’and P. Rosenbaum teds.), in English Transformational Grammar. Waltham, Mass., Ginn & Co. Chomsky, N. and M. Halle (1968) The Sound Pattern ofEnglish. New York, Harper & Row. Fiengo, R. (1974) Semantic conditions on surface structure. Unpublished Ph.D. thesis, MIT.

Readings

noted that this area “has not been very extensively investigated beyond the basic research of Goodglass and Berko (1960)” (p. 17), but did not claim, as Kolk implies I did, that that research dealt solely with Broca’s aphasics. Furthermore, for the data in question, the terms “post-vocalic” and “post-consonantal” are extensionally equivalent with “syllabic” and “nonsyllabic”, though Kolk implies they are not. That brief discussion in Kean (1977a) concludes with the observation that “more detailed and systematic study in this area” should be pursued (p. 17).

Agrammatism:

A phonological

deficit?

83

J., J. Fodor and M. F. Garrett (1975) The psycholoEica1 unreality of semantic representations. _. ‘Li&. Inq.,6, 515-532. Forster. K. I. (1978) Accessing the mental lexicon. In E. Walker (ed.). Explorations in the Biology of .&g&e. Montgomery, Vt., Bradford Books. Fromkin, V. (1971) The nonanomalous nature of anomalous utterances. Lung., 47, 27-52. Fromkin, V., ed. (1973) Speech Errors as Linguistic Evidence. The Hague, Mouton. Garrett, M. F. (1975) The analysis of sentence production. In G. Bower (ed.), The psychology of learning and motivation: advances in research and theory, Vol. 9. New York, Academic Press. Goodglass, H. and J. Hunt (1958) Grammatical complexity and aphasic speech. Word, 14, 197-207. Goodglass, H. and J. Berko (1960) Aphasia and inflectional morphology. J. sp. hear. Res., 3, 257-267. Goodglass, H., C. Quadfasel and W. H. Timberlake (1964) Phrase length and the type and severity of aphasic disturbances. Cortex, I, 133-153. Goodglass, H., I. Fodor and C. Schulhoff (1967) Prosodic features in grammar - evidence from aphasia. J. sp. hear. Res., 10, S-20. Goodglass, H., J. Gleason, N. Bernholz and M. R. Hyde (1972) Some linguistic structures in the speech of a Broca’s aphasic. Cortex, 8, 191-212. Katz, J. J. (1972) Semantic Theory. New York, Harper & Row. Kean, M.-L. (1974) The strict cycle in phonology. Ling. Znq., 5, 179-203. Kean, M.-L. (1975) The theory of markedness in generative grammar. [Jnpublished Ph.D. thesis, MIT. Kean, M.-L. (1977a) The linguistic interpretation of aphasic syndromes: agrammatism in Broca’s aphasia, an example. Cog., 5, 9-46. Kean, M.-L. (1977b) ‘Natural processes’ and ‘learned rules’in markedness theory. NELS, VIZ, 135-146. Kean, M.-L. (1978) The linguistic interpretation of aphasic syndromes. In E. Walker (ed.), Explorations in the Biology of Language. Montgomery, Vt., Bradford Books. pp. 67-138. [The first two sections of this paper appear as Kean (1977a) in a slightly revised form.] Kean, M.-L. (forthcoming a) Grammatical representations and the description of processing. To appear in D. Caplan (ed.), Biological Studies ofMental Processes. Cambridge, Mass., MIT Press. Kean, M.-L. (forthcoming b) Explanation in neurolinguistics. To appear in N. Hornstein and D. Lightfoot (eds.), Explanation in Linguistics. Kean, M.-L. (forthcoming c) The ‘psychological reality’ of markedness theories. Kean, M.-L. and M. F. Garrett (forthcoming) Levels of representation and the analysis of speech errors. To appear in M.-L. Kean and M. Aronoff (eds.), Juncture. Klosek, J. (1979) Two unargued linguistic assumptions in Kean’s ‘phonological’ interpretation of agrammatism. Cog., 7, 61-68. Kolk, H. (1978) The linguistic interpretation of Broca’s aphasia: a reply to M.-L. Kean. Cog., 6, 349362. Lecours, A. R. and F. Lhermitte (1969) Phonemic paraphasias: linguistic structures and tentative hypotheses. Cortex, 5, 193-228. Lyons, J. (1968) Introduction to Theoretical Linguistics. Cambridge, Cambridge University Press. Morris-Jones, J. (1913) A Welsh Grammar: Historical and Comparative. London, Oxford University Press. Selkirk, E. (1972) The phrase phonology of English and French. Unpublished Ph.D. thesis, MIT. Shattuck-Hufnagel, S. and D. Klatt (forthcoming) The limited use of distinctive features and markedness in speech production. J. verb. Learn. verb. Behav. Fodor.

Cognition, 7 (1979) 85-91 @Elsevier Sequoia S.A., Lausanne

Discussion - Printed

in the Netherlands

A reevaluation of the basic operations hypothesis” HELEN GOODLUCK University

of Wisconsin,

LAWRENCE University

Madison

SOLAN

of Massachusetts,

Amherst

Some recent work in language acquisition has emphasized the importance of linguistic theory to our understanding of the data from child language. In their paper, “Transformations, Basic Operations and Language Acquisition”, Mayer, Erreich and Valian propose that certain data from child language reflects the analysis of syntactic movement rules as a combination of two “basic operations”: copying and deletion. Errors of the type in (1 ), (1)

Whose is that is?

are explained by Mayer et al.‘s claim that the auxiliary verb is is copied into presubject position, but is not deleted from its underlying post-subject position, as adult grammar requires. Mayer et al., observe that failure to delete is widespread in children’s language, but this type of error is not found when the moved constituent is a w&word, although their analysis leaves this and other gaps unaccounted for.’ We will suggest that the occurrence of gaps in the errors predicted under the copying without deletion hypothesis is not accidental, but may follow from properties of grammar discussed in recent linguistic analyses of the constructions involved.

Two types of transformation Transformational rules can be divided into two types: those whose structural descriptions involve an essential variable, and those whose structural descriptions can be stated in terms of a finite string of specified terms with no variables (Postal (197 1); Bresnan (1977a)). We will refer to the former as “un*We wish to thank Emmon Bach, Charles Clifton, David Fay, Steve Lapointe and Tom Roeper for their valuable discussion and suggestions. Reprint Requests should be addressed to Helen Goodluck, Department of English, University of Wisconsin - Madison, 600 North Park Street, Madison, Wise. 53706, U.S.A. ‘The other gaps, negative placement and ing hopping, will be discussed in some detail below.

86

Helen Goodluck and Lawrence Solan

bounded rules,” and to the latter as “local rules”. W/z-movement is an example of an unbounded rule. As the sentences in (2) illustrate, there is no limit to the distance between the position of a wh-word in the surface structure, and its underlying position.

(2) a. Who did John see __? b.

Who did John see Harry persuading

Tough movement, below in (3).2 (3)

a. b.

another

example

Bill to talk to __?

of an unbounded

rule,

is illustrated

It is tough to get Harry to please John. John is tough to get Harry to please __.

(It is actually possible that these unbounded rules do not involve movement at all; both wh movement and tough movement can be formulated as rules that delete constituents under identity (see Lasnik and Fiengo (1974). Crucial for the point we will make is that these rules are unbounded, whatever their precise formulation.) In contrast to wh-movement and tough movement, subject auxiliary inversion is a local rule. The auxiliary is moved to the left of an immediately preceding noun phrase; no other material intervenes. The rule allows us to derive sentences such as (4b) from underlying structures such as (4a). (4)

a. [sQ[NpJohn] [Auxis] [vpplaying b. Is John playing tennis?

tennis1

Dative movement, as illustrated in (5), and particle in (6), are also examples of local rules. (5) (6)

a. b. a. b.

Mary Mary John John

gave gave took took

movement,

as illustrated

a book to Susan. Susan a book. off his shoes. his shoes off.

In both cases, only adjacent

elements

are involved in the movement.

21t has been disputed that tough movement is an unbounded rule (Postal (1971)). Examples such as (3b) argue that it is unbounded, at least for our dialects. Nanni (1978) presents additional evidence that rough movement is an unbounded rule.

Reevaluation of the basic operations hypothesis

81

Copying without deletion3 as a property of local rules The cases of copying without deletion that Mayer et al., cite - tense hopping, subject auxiliary inversion, dative movement and particle movement - are all examples of local rules. (7) (8) (9) (10)

I did broke it (tense hopping) What did you did (subject auxiliary inversion) Could you get me a banana for me (dative movement)” The barber cut off his hair off (particle movement)

If we assume that copying without deletion is confined to local rules, then the failure to copy without deleting in the case of w/z-movement, which Mayer et al. assume to be an unexplained gap in their data, need not be viewed as an exception at all. Rather, the fact that copying without deletion does not occur with unbounded rules can be taken as evidence that children are sensitive to the distinction between rule types outlined above.5 As our claim predicts, cases such as the following have not been reported in the literature: (11) (12)

*Who did Bill shout at who (w/z-movement) *John is tough to please John (tough movement)

Properties of some local rules Is the distinction between unbounded and local rules appropriate to account for the range of observed and non-observed copying without deletion errors in the literature? We have pointed out that no errors have been observed involving unbounded rules. However, Mayer et al. also note the absence of errors with two other rules, which are local. 3We use copying withour deletion as a descriptive term. We will suggest below that certain of these errors should be explained differently. 4This example was cited by the authors in a paper presented at the Language Acquisition Workshop, University of Massachusetts at Amherst, April, 1978. No examples of dative movement are cited in their article. ‘Foss and Fay (1975) suggest that the range of children’s errors may correlate with the distinction between major and minor transformations drawn by Bach (1971). Bach proposed that there is a finite list of transformations made available to every language by linguistic theory; these are the major rules, of which Wh-movement would be an example. A rule such as subject auxiliary inversion, which does not belong to this list, is an example of a minor rule in Bach’s typology. Thus Foss and Fay propose that copying without deletion errors be restricted to minor rules. It may be that the unbounded vs. bounded distinction that we focus on provides an appropriate formal basis for distinguishing between major and minor transformations.

88

(13) (14)

Helen Goodluck and Lawrence Man

*He ising going to the store (ing hopping)6 What not can’t I do (negative placement)

In this section we will suggest that the absence of the first of these errors follows from an analysis of the English auxiliary in which movement rules are not involved. The presence of tense hopping errors and the absence of copying errors with negatives can also be handled within this analysis. We will look first at tense hopping, and then examine ing hopping and negative placement. Brame (1978) and Lapointe (1977) have proposed that the so-called tense hopping rule used in traditional analyses of the English auxiliary can be replaced by subcategorial restrictions on auxiliary verbs. Under Lapointe’s analysis, which makes crucial use of a precise formulation of the syntactic features defining the verbal system, (7) can be accounted for as the result of incomplete mastery of the adult subcategorization frames for auxiliary verbs. In Lapointe’s analysis a positive feature specification for tensing ([+tin(ite)] ) is confined to the first verb. This is ensured by the subcategorization frames for tensed verbs. It is possible that children at a certain stage do not observe this restriction, and permit tensed verbs to be inserted in second position in the auxiliary system also. Lapointe also accounts for the distribution of progressive and perfective verbal affixes (ing and en) by subcategorization. The syntactic feature [+part(iciple)] that is realized as the progressive ing suffix on the verb may not occur in the specification of a tensed verb. The fact that the features [ +fin] and [ +part] may not occur on the same verb will follow from universal principles of well-formedness on the specification of the verb, since all verbs with the specification [+part] will (redundantly) be specified [-fin], and no verb can be specified both [+fin] and [-fin]. Thus, once the child knows that the first verb in the matrix sentence is always tensed, then the failure to find forms such as (13) follows automatically from the constraint on cooccurrence of the features [+part] and [+fin] . In Lapointe’s analysis of the auxiliary system, the negative morpheme not is inserted directly in its surface position by the lexical insertion rules of the base, rather than being moved to its surface position from sentence initial position by a movement rule. The absence of copying errors in the case of negative sentences such as (14) is thus not an unexplained gap, given

6The illustrative sentence Mayer et al., give for copying without deletion of the -ing affix is tenseless (I being going fo the store). We assume that children will not produce tenseless matrix sentences.

Reevaluation of the basic operations hypothesis

89

an analysis of the English auxiliary system in which no movement is involved in determining the surface position of the negative morpheme.7 In sum, we see that an analysis of the English auxiliary system in which the occurrence of tensing, progressive and perfective marking and the position of the negative morpheme does not involve movement in any case does not predict any copying without deletion errors for (13) and (14). Errors of this kind are found only in the case of tensing. We have proposed that this may follow from the child not having mastered a restriction on the occurrence of tensing to the first verb in the matrix sentence.

Extending

non-transformational

analyses

The cases of copying without deletion errors that do not involve the internal structure of the auxiliary (subject auxiliary inversion, dative movement and particle movement) can all be accounted for by their localness. It is worth noting that recent analyses of some local rules in addition to those involving the auxiliary system claim that the co-occurrence restrictions and synonomy relations holding between structures related by such rules should not be accounted for by transformations. This type of reanalysis may with further research provide the basis for a non-transformational account of the copying without deletion errors that do occur. Thus Brame (1978) and Bresnan (1977b) make the claim that certain lexically governed rules such as dative movement are not transformations. According to Bresnan’s analysis, the two sentences in (5) are both base generated, and are related by a redundancy rule (lecture, MIT, 1977). A similar analysis can be proposed for the particle movement sentence in (6). This kind of analysis suggests that the copying without deletion errors for dative movement and particle movement may be the result of a “blend” of subcategorization frames.8 Extending a non-transformational analysis to subject auxiliary inversion would not be plausible, however, at least in adult grammar, since the rule operates independent of the particular lexical items involved. If a subcate‘Actually, Lapointe’s analysis of negative placement does not preclude the occurrence of such errors. If they were to occur, an analysis similar to that used to explain the double tensing errors could be used to account for them. *Some produced

utterances from a 4 year old child we have studied appear to support this view. The child copying without deletion errors with the particle verb put on ([He’s] putting on the blue beads on), and also produced the following sentence: [He’s] putting the hat on on his head. This error can be accounted for as failure to consider the subcategorization frames for put as mutally exclusive (He put the hat on; He put the hat on his head). It is difficult to see how a copying without deletion analysis can generalize to account for this data. A subcategorial analysis treats errors of the form put on his hut on and put his hat on on his head similarly.

90

Helen Goodluck and Lawrence Man

gorial analysis of rules such as dative movement is adopted, this leaves subject auxiliary inversion as the only strong case for the copying without deletion analysis of children’s errors, regardless of whether or not it is claimed that these errors are restricted to local rules. The strength of this case also deserves further scrutiny. Limber (1973; p. 178) observes that most auxiliary verbs enter the child’s language in questions slightly before they enter in declaratives. This indicates that an analysis under which questions are generated by a separate phrase structure rule from that for declaratives and under which verbs are lexically selected for these structures might be possible for children’s grammar. If so, all observed copying without deletion errors could be described in terms of base rules and subcategorization frames. Whether such an analysis can be maintained will depend crucially on the results of additional research on the development of questions in children. Conclusion The basic operations hypothesis may be correct. However, if interpreted as a general principle governing the acquisition of all movement rules, this hypothesis may obscure the fact that children distinguish between unbounded and local rules. That the pattern of errors supports the presence of this distinction in child language lends credence to linguistic theories in which the two rule types have separate status. In their response to Fay (1978), Maratsos and Kuczaj (1978), having argued against a particular transformational account of double tensing errors, conclude that “these are errors of production, rather than representing underlying rule structure,” (p. 343). We share Maratsos and Kuczaj’s concern with arriving at an analysis that will eliminate or reduce the number of unsupported predictions that the theory of language acquisition makes concerning copying without deletion errors. However, it seems to us that, in spite of possible shortcomings in particular transformational analyses, underlying rule structure is not necessarily irrelevant in accounting for errors of this type, and that this question is independent of whether the copying without deletion phenomena reflect production errors or non-adult rules of grammar. Thus we hold that the fact that copying without deletion errors are restricted to rules of a particular type provides a principled basis for accounting for the errors observed and so supports the presence of the relevant distinctions in the child’s grammar. Whether these errors represent incorrectly formulated rules in this area of the child’s grammar or production errors is a separate issue.

Reevaluation

of the basic operations

hypothesis

91

If, as we have suggested, the correct generalization is that copying without deletion errors are confined to rules whose statement does not involve an essential variable, then future research should be directed at discovering properties of the language acquisition device which correlate with this formal property of linguistic rules. The distinction between unbounded and local rules corresponds in many cases to that between structure preserving and structure destroying (“root”) transformations proposed by Emonds (1976). T. Roeper has suggested to us that the restriction of copying without deletion errors may follow from the child’s desire to mark underlying sentential relations in surface structure: such relations are destroyed only by local rules. We expect that work which explores further the relationship between linguistic theories such as those of Emonds and Bresnan and the data from language acquisition will lead to insight into the cognitive and linguistic development of the child.

References E. (1971) Syntax since Aspects, in Report of the 22nd Annual Round Table Meeting on Linguistics and Language Study, No. 24. Brame, M. (1978) The base hypothesis and the spelling prohibition, Ling. Anal, 4, l-30. Bresnan, J. (1977a) Variables in the theory of transformations. ln P. Culicover, T. Wasow and A. Akmajian (eds.), Formal Synfax. New York, Academic Press. Bresnan, J. (1977b) A realistic transformational grammar. To appear in M. Halle, J. Bresnan and G. Miller (eds.), Linguistic Theory and Psychological Reality. Cambridge, Mass., MIT Press. Emonds, J. (1976) A Transformational Approach fo English Syntax. New York, Academic Press. Fay, D. (1978) Transformations as mental operations: a reply to Kuczaj.J. Child Lang., 5,143-149. Foss, D. and D. Fay (1975) Linguistic theory and performance models. In J. Wirth and D. Cohen (eds.), Testing Linguistic Hypotheses. New York, Hemisphere Press. ,Lapointe, 8.. (1977) A lexical reanalysis of the English auxiliary system, ms. University of Massachusetts, Amherst. Lasnik, H. and R. Fiengo (1974) Complement object deletion,Ling. Inq., 5,535-572. Limber, J. (1973) The genesis of complex sentences. In T. E. Moore (ed.), Cognitive Development and the Acquisition of Language. New York, Academic Press. Maratsos, M. and S. Kuczaj (1978) Against the transformational account: a simpler analysis of auxiliary overmarkings,J. Child Lang., 5,337-345. Mayer, J., A. Erreich and V. Valian (1978) Transformations, basic operations and language acquisition. Cog., 6, 1-13. Nanni, D. (1978) The Easy Class of Adjectives in English, doctoral dissertation, University of Massachusetts, Amherst. Distributed by the Graduate Linguistic Student Association, University of Massachusetts, Amherst. Postal, P. (197 1) Cross-Over Phenomena. New York, Holt, Rinehart & Winston. Bach,

Discussion

Cognition, 1(1979) 93-95 @Elsevier Sequoia %A., Lausanne - Printed in the Netherlands

In reply to Philip Johnson-Laird JERRY

A. FODOR

Massachusetts

institute

of Technology

*

Professor Johnson-Laird in “What’s Wrong with Grandma’s Guide to procedural Semantics: A Reply to Jerry Fodor” (1978) has three points. (A) he agrees that the position I attacked in “Tom Swift” is indefensible; (B) he denies that anybody has held that position, except for (C) me, for Goodness sake.’ About A: Good. About B: I’m more than content to leave the issue of textual exegesis to those who have read the texts. I site a number of these in “Tom Swift”, and there are plenty more around.’ ‘He thinks he has a fourth; viz. that semantic interpretations ought to be specified over formulae of higher-level languages directly, and not over their machine language translations. This point “largely wrecks” my paper. Johnson-Laird doesn’t say just what in “Tom Swift” turns upon this point, and that is well-advised of him. For: (a) There is no argument in “Tom Swift” that implicates this issue; there is no argument in “Tom Swift” that even requires that there be a distinction between higher level programming languages and ML. (b) The question which level of representation to specify interpretations over is entirely tactical; it raises no matter of principle (as Johnson-Laird himself almost notices). (c) It’s patent that to fix an interpretation for the formula F is also to fix an interpretation for any formula viewed as synonymous with F. But, as 1 went to pains to point out in “Tom Swift”, compiling is understood as translation according to PS principles. So, if we interpret the formulae of a high-level language L, we thereby fix interpretations for the relevant machine language formulae of any device on which L is implemented; and vice versa, translation being a symmetrical relation. Given the total irrelevance of these matters to the issues examined in ‘Tom Swift”, I can’t imagine what could have prompted Johnson-Laird to raise them here; except, perhaps, for a desire to put out smoke. *One kind of evidence that’s supposed to show the inaccuracy of my reading of PS is the fact that much AI work on vision (scene analysis programs, for example) is neither reductionistic nor sensationalistic. This is entirely correct and utterly irrelevant. The question is not whether some practitioners in AI (# PS, by the way) have understood that Locke won’t do. It’s whether such non-Empiricist accounts will cohere with the PS doctrine about the way that perception and language are interfaced. To repeat the doctrine of “Tom Swift”: such coherence could be purchased only by abandoning the PS account of language and the world; either PS interprets internal representations via transducer language, or it has no account whatever of how they are to be interpreted. This argument has the form of a dilemma and cannot be met by embracing both horns simultaneously, as Prof. Johnson-Laird appears to be inclined to do. *Reprint requests should be addressed to Jerry A. Fodor, Department 034, Cambridge, Mass. 02139.

of Psychology, M.I.T. ElO-

94

Jerry A. F&or

About

C: This really

is egregious.

The third passage that Johnson-Laird of a discussion of the question whether, given that a translation algorithm can’t provide a theory of language and the world, there may nevertheless be reason to translate English into some canonical interpreted language in the course of providing a semantic theory of English. The doctrine in LOT (and, again explicitly, in “Tom Swift”) was that, yes indeed, there probably is good reason to adopt that tactic. The kicker is, of course, in the observation that translation into canonical fixes an interpretation only if canonical is itself interpreted. So, the discussion in LOT starts with the remark: “. . . . . mere translation of the formulae of L into those of a canonical language does not provide an account of the way that the formulae of L relate to the world. . . It follows that translational semantics, unlike real semantics, does not say how symbols relate to whatever it is they symbolize.” (page 120). Johson-Laird would like to convince the reader that “Tom Swift” is really a quibble over how to use the term “semantics”. But it is not. Lots of texts, including, by all means, Katz and Fodor (1963), LOT and, for that matter, most of contemporary linguistics, use the term “semantics” equivocally to refer either to a function from natural language onto canonical or to a function from canonical onto the world. It’s not a particularly desirable practice, but it’s harmless unless you get confused. The point of “Tom Swift” is that the procedural semanticists have gotten confused; that they’ve thought of semantics as a theory which somehow accomplishes both the ends of translation and interpretation. Or, rather more accurately, they’ve failed to see that these are different ends and composed theories which merely fall between them. The verificationism ubiquitous in PS is the inevitable consequence, as “Tom Swift” seeks to show. I’m convinced that this is the right diagnosis of what PS is suffering from though, given point A, diagnosis is perhaps academic. The patient is dead. Similar remarks apply to the way that Johnson-Laird uses the other passages from LOT. For example, the fourth quote is part of a discussion of a number of possib2e ways of avoiding nativistic consequences of standard theories of word and concept learning. The succeeding chapter of LOT contains some thirty pages of argument against the definition/decomposition way out of these consequences which, in fact, the book contends ought to be endorsed. The third quote, on the other hand, expresses a doctrine that LOT does accept, but it’s quite compatible with what I say in “Tom Swift” and is, indeed, repeated there. Once again, for the record: you might well think of an account of the operations involved in getting from English into canonical as a model of language comprehension. This is a reasonable view precisely

cites from Language of Thought occurs in the context

In

reply to PhilipJohnson-Laird

95

because such an account isn’t a semantic theory, hence doesn’t have to provide a theory of language and the world (an interpretation of canonical). There’s no inconsistency in saying: (1) a semantic theory specifies the interpretation of canonical; (2) a function that takes you from English onto formulae of an interpreted canonical determines the interpretation of English; and (3) a computational model which realizes that function would be a candidate theory of the speaker/hearer. l-3 can’t be inconsistant, because they’re all true. There’s more to say about Johnson-Laird’s handling of LOT. For example, the fact that he thinks that the second quote is relevant to the present discussion suggests that Johnson-Laird hasn’t grasped the distinction between claiming that sensory data confirm perceptual hypotheses (which nobody in his right mind has ever doubted) and claiming that the content of perceptual hypotheses is exhaused by the sensory data relevant to their confirmation. Missing this distinction is precisely what made epistemologists verificationists in their day and what makes procedural semantics verificationists in ours. Enough, however. Johnson-Laird asserts that he never meant what he seems to say, and I assert that I never meant what I didn’t euen seem to say. A vast silence might now profitably descend.

References Fodor, J. A. (1975) The Language of Thought, New York, Thomas Y. Cromwell Co. Johnson-Laird, P. N. (1978) What’s Wrong with Grandma’s Guide to Procedural Semantics: A reply to Jerry Fodor, Cog., 6, 249-261. Katz, J. J. and Fodor, J. A. (1963) The structure of a semantic theory, Lang., 39, 170-210.

97

Cognilion,7 (1979) 97-98

Books received

H. Putnam, Meaning and the Moral Sciences, Routledge, Kegan Paul, Henley on Thames, 1978. R. M. Jones, The New Psychology of Dreaming, Penguin Books, Harmondsworth, sex, 1978

Middle-

W. K. Estes, Handbook of Learning and Cognitive Processes, Lawrence Erlbaum. ciates, New Jersey, 1978.

Asso-

R. A. Monty and J. W. Senders (eds.), Eye Movements and Psychological Processes, Lawrence Erlbaum Associates, New Jersey, 1978. A. R. Luria, The Neuropsychologv of Memory, K. Winston and Sons, Washington, 1978.

D.C.,

L. P. Lipsitt (ed.), Developmental Psychology, Lawrence Erlbaum Associates, New Jersey, 1978. R. G. Crowder, Principles of Learning and Memory, Lawrence Erlbaum Associates, New Jersey, 1978. N. Kogan, Cognitive Styles in Infancy and Early Childhood, Lawrence ciates, New Jersey, 1978. J. R. Andersoon, Jersey, 1978.

Erlbaum

Asso-

Language, Memory and Thought, Lawrence Erlbaum Associates, New

J. Schenken (ed .), Studies in the Organization of Conversational Interaction, Academic Press, New York, 1978. H. I. Pick, E. Satzman (eds.), Modes of Perceiving and Processing Information, Lawrence Erlbaum Associates, New Jersey, 1978. E. L. J. Leeuwenberg and H. F. J. Buffart (eds.), Formal Theories of VisualPerception, John Wiley & Sons, New York, 1978. B. R. Clifford and Ray Bull (eds.), The Psychology of Person Identification, Routledge and Kegan Paul, 1978. A. Sloman, The Computer Revolution in Philosophy, Harvester Press, Sussex, 1978.

98

Books Received

A. F. Parker-Rhodes,

Inferential Semantics,

Harvester Press, Sussex, 1978.

W. K. Estes (ed.), The Handbook of Learning and Cognitive Processes, Lawrence Erlbaum, New Jersey, 1978. R. Revlin and R. Mayer, Human Reasoning, Winston-Wiley, Castellan & Restle (eds.), Cognitive Theory, 1978. Leon Colburn Lack, Selective Attention Publishers, The Hague, 1978. K. T. Strongman,

Lawrence Erlbaum

Associates, New Jersey,

and the Control of Binocular Rivalry, Mouton

The Psychology of Emotion,

D. W. Churchill, Language ofAutistic

Sussex, 1978

John Wiley & Sons, New York, 1978.

Children, John Wiley & Sons, New York, 1978.

P. Siple (ed.), Understanding Language Through Sign Language Research, Academic Press, New York, 1978. P. B. Warr, Psychology at Work, Penguin Books, Middlesex, 1978. Yando, Seitz and Zigler, Imitation: Associates, New Jersey, 1978. Szalay and Deese, Subjective Jersey, 1978.

A Development

Meaning and Culture,

of Perspective,

Lawrence Erlbaum

Lawrence Erlbaum

Associates, New

W. Levelt & G. Flores d’Arcais, Studies in the Perception of Language, John Wiley & Sons, New York, 1978.