Journal of Germanic Linguistics 17.3 (2005):155–181
The Syntax of Old Norse Kviuháttr Meter Kari Ellen Gade Indiana University, Bloomington This paper seeks to explain syntactic and structural features that have puzzled earlier scholars and editors of poems composed in kviuháttr, one of the oldest attested Old Norse skaldic meters. More precisely, I attempt to answer the following questions: first, why do poems in kviuháttr fail to adhere to the stanzaic eight-line division and the fourline helmingr (half-stanza) division, which are so firmly entrenched in the ON skaldic tradition? Second, why does the syntax of kviuháttr poetry depart from other patterns of skaldic (and eddic) verb syntax? It will become clear that the metrical innovation that characterizes kviuháttr, namely, the introduction of regularized catalectic a-lines of Sievers’ Types A, C, and D, imposed significant restrictions on verb syntax. Rather than violating the verb-second constraint in independent clauses, the poets were forced to relax structural rules that are observed strictly in other branches of ON poetry.*
1. Introduction. Unlike other ON alliterative meters, kviuháttr has received scant attention from scholars.1 This scholarly neglect is all the more to be *
I am grateful to my colleague, Robert D. Fulk, for helpful comments on earlier versions of this paper. Thanks also to the two anonymous JGL reviewers for valuable suggestions. 1
For a recent bibliography of the scholarship on ON alliterative poetry see Fulk and Gade 2002:102–23. Kviuháttr is discussed (briefly) in Sievers 1879:291– 294, Sievers 1893:117–118, Heusler 1956:201–204, Kuhn 1983:336, Faulkes 1991:74, 84, and Gade 2002:863–864. Most works devoted to poetry in this meter have been concerned with literary questions and with the dating of the earliest poems (see Marold 1983:114–153, Krag 1991, and Sapp 2000). The most complete treatment of kviuháttr is Åkerlund 1939, but that study is outdated. The meaning and etymology of the word kviuháttr are debated. See Konrá Gíslason 1881:188–189, Möbius 1883:292–293, Sievers 1893:219–220, Wessén 1915:127–141, Noreen 1924:61–65, Indrebø 1928:118, Finnur Jónsson 1929:268, Heusler 1956:215, and de Vries 1977:337–338. © Society for Germanic Linguistics
156
Gade
lamented, because poems in this skaldic meter display unique metrical and syntactic characteristics that set them apart from other ON poetry.2 In the following, I explore and seek to explain certain peculiar syntactic features of kviuháttr poems that have been commented on by earlier scholars, but never explained. Kviuháttr was used in seven longer encomiastic poems totaling 1,926 lines from the ninth century to the thirteenth: namely, jóólfr of Hvini’s Ynglingatal (Yt: 360 lines; end of ninth century?), Egill Skallagrímsson’s Sonatorrek (St: 194 lines; ca. 960) and Arinbjarnarkvia (Abj: 196 lines; ca. 960), Eyvindr Finnson’s Háleygjatal (Hál: 100 lines; ca. 985), órarinn loftunga’s Glælognskvia (Glæ: 76 lines; ca. 1032), the anonymous Nóregskonungatal (Nkt: 664 lines; ca. 1190), and Sturla órarson’s Hákonarkvia (Hkv: 336 lines; 1263–64).3 These poems are transmitted either en bloc in the manuscripts or as narrative units or single stanzas and half-stanzas interspersed with prose in the kings’ sagas and in the Prose Edda. Unlike West Germanic alliterative poetry, ON skaldic poetry is strictly stanzaic. A stanza consists of eight lines that are divided into two four-line half-stanzas, or, to use the conventional ON term, helmingar. The helmingr is syntactically complete and inviolable.4 However, a 2
The earliest forerunner to kviuháttr is the stanza on the Swedish Rök stone, dated to around 850 (see von See 1967:47–48, Gade 1995:234–235, Gade 2002:858–859). It is generally believed that kviuháttr developed as a catalectic variant of fornyrislag (so Kuhn 1933:105, Åkerlund 1939:189, 192–194, Gade 2002:863; but compare Heusler 1956:203). 3
See Skj AI:7–15 (Yt), 40–43 (St), 43–48 (Abj), 68–71 (Hál), 324–327 (Glæ), 579–589 (Nkt), Skj AII:108–119 (Hkv). The dates given for the earliest poetry are tentative (but see Sapp 2000). Kviuháttr is also used in two stanzas by Gísli Súrsson (sts. 10–11; Skj AI:103), in Grettir Ásmundarson’s Ævikvia (7 sts; Skj AI:309–311) and another 8 1/2 stanzas in Grettis saga (Skj AII:439–442), in Merlinusspá (sts. 62–68; Skj AII:20–21), and in Stjrnu-Odda draumr I (5 sts; Skj AII:208). See also Faulkes 1991:84. Note that the line counts refer to halflines, that is, a- and b-lines, and not to long lines. In Old Norse, the conventional terms for half-lines are a- and b-lines (or odd and even lines), not a- and b-verses as in Old English. 4
Syntactic completeness refers to the fact that the four-line stanzas are syntactically independent units, that is, no clause or phrase in one helmingr depends on or is appositive to a syntactic element in another helmingr or stanza (see
Syntax of Old Norse Meter
157
glance at the standard editions reveals that many of the kviuháttr stanzas are syntactically incomplete; that is, they start in medias res, as it were, as is illustrated by the following stanza from Ynglingatal:5 (1) Yt 36
Unz fótverk vi Foldar rm vígmilung of via skyldi. Nú liggr gunndjarfr á Geirstum herkonungr haugi ausinn.
Before foot-ache at (the) Fold’s edge (the) battle-distributor (expl. of) to kill was6 Now lies (the) battle-brave at Gjekstad war-king by-(a)-mound covered.
Prose order: Unz fótverk skyldi of via vígmilung vi Foldar rm. Nú liggr gunndjarfr herkonungr á Geirstum, ausinn haugi. Translation: Before foot-ache was to kill the battle-distributor [WARRIOR] at the Fold’s edge [on the shores of Oslofjorden]. Now the battle-brave war-king lies at Gjekstad, covered by a mound.
Hollander 1947:300, Kuhn 1969, Kuhn 1983:187–188). In ON historical and poetological prose works, stanzas and half-stanzas are routinely interspersed with the prose as historical verification of events discussed or as illustrations of specific metrical and poetic peculiarities. 5
The translations and editions of the poems in kviuháttr are my own. The lines in fornyrislag (and the abbreviated titles of the poems) are cited from Neckel 1962, and the translations are my own. Note that, unlike West Germanic alliterative poetry and eddic fornyrislag, poetry in ON skaldic meters is always printed in columns of half-lines rather than as long-lines divided by metrical caesurae. The referents of the kennings, the poetic circumlocutions that characterize ON skaldic diction, are included in the translation in small capitals within square brackets, for example, ‘the battle-distributor [ WARRIOR]’. 6
The notations (expl. of) and (expl. um) are used to signal the presence of the expletive particle of (later um), which evolved in preliterary ON from earlier prefixes (see Kuhn 1929). Though required by the meter, such particles are meaningless.
158
Gade
Furthermore, sentence boundaries are often obfuscated by helmingr and stanza divisions, which never happens in other skaldic meters. Consider the following four lines: (2) Hál 4
Sævar beins ok sunu marga ndurdís vi Óni gat.
Ocean’s bone’s and sons many (the) ski-goddess with Óinn begot.
Prose order: Sævar beins; ok ndurdís gat marga sunu vi Óni. Translation: Ocean’s bone’s [STONE’S ]; and the ski-goddess [Skai] begot many sons with Óinn. This half-stanza is cited as a single, separate unit by Snorri Sturluson in his Heimskringla (ca. 1235) to verify the mythological marriage of Óinn and Skai. However, the genitive qualifier ocean’s bone’s that introduces the half-stanza does not belong syntactically to the following three lines. Rather, this phrase must have qualified an NP in the preceding, no longer extant half-stanza, thus violating the helmingr division. Clearly, the four-line helmingr was such a structural staple in skaldic poetics that Snorri automatically cited the four lines as a unit, even though the first line is nonsensical in the present context. The question is why poems in kviuháttr meter fail to adhere to the stanzaic eight-line division and the four-line helmingr division that are so firmly entrenched in the ON skaldic tradition. In the following, I first examine the structure of kviuháttr and attempt to answer that question. Second, I show how the poets resorted to different solutions to cope with metrical constraints imposed on syntax. It will become clear that the syntax of kviuháttr departs from other patterns of skaldic (and eddic) verse syntax, but that the departures are principled, and the governing principles are clearly discoverable.
Syntax of Old Norse Meter
159
2. The Metrical Structure of Kviuháttr. Kviuháttr is a variant of eddic fornyrislag, but unlike the extant fornyrislag, it is syllable counting.7 Whereas the b-lines are tetrasyllabic and follow the metrical patterns of fornyrislag, the a-lines are hypometrical, that is, each a-line consists of three syllables instead of four. Because this is the distinguishing feature of the meter, I focus here on the a-lines in an attempt to explain the rather aberrant syntax of kviuháttr poems. Structurally kviuháttr a-lines can best be characterized as catalectic, that is, they follow the same metrical patterns as fornyrislag lines of Sievers’ Types A, C, and D, minus the last, unstressed syllable.8 Types B, D4, and E do not end in an unstressed syllable; therefore there are no catalectic lines conforming to those patterns, and, consequently, no such kviuháttr a-lines. Consider the following metrical schemes (according to Sievers’ Five Types):9
7
Fornyrislag meter is the ON equivalent of the West Germanic alliterative meter, and it served as the basis for Sievers’ formulation of his Five-Type system (see Sievers 1893). 8
That is, whereas fornyrislag a-lines have four metrical positions, a-lines in kviuháttr have three metrical positions. ON poetry in fornyrislag can also contain trisyllabic catalectic a-lines, but such lines are employed fairly infrequently and not according to a regular pattern as in kviuháttr. See Sievers 1893:68, Heusler 1956:178–180. 9
See Sievers 1893:117–118. The alliterating staves have been emphasized in the ON text. Although criticism has been raised concerning the validity of this metrical framework for ON skaldic poetry (see Gade 1995), it provides a convenient descriptive layout for the purposes of the present investigation. The metrical notations used in 3 are the following: – = alliterating lift; – = nonalliterating lift; = alliterating short lift; = nonalliterating short lift; – = secondary stress, long syllable; = secondary stress, short syllable; x = unstressed syllable (dip); | = metrical foot demarcation.
160
Gade
(3) Kviuháttr a-lines according to Sievers’ Five Types A1: – x | – Yt 4/11 glóa garmr embers’ dog A2: – x | – Yt 1/3 feigaror
death-word
A3: – x | – Yt 2/5 ás í stein
when into (the) stone
C1: x – | – Yt 1/5 ok sikling
and (the) lordling
C3: x – | Yt 16/1 Ok sveiurs
or:
x – | and (the) steer’s
D1: – | – – Yt 1/7 vágr vindlauss
or:
– | – – sea windless
D2: – | – Yt 2/3 salvrur
or:
– | – hall-warden
D3: – | – Yt 3/11 mengltur
necklace-destroyer
10
Others: x – Yt 14/1 var Jrundr
or:
x – was Jrundr
x – Yt 2/11 jtunbyggr
or:
x – giant-populated
The main constraint that the catalectic meter places on the syntax is that it makes it difficult to accommodate finite verbs in the trisyllabic alines. As the following overview of the possible placement of such verbs in bound and independent clauses shows, that circumstance in turn has 10
In all ON meters, two short syllables occurring in the first lift are always resolved, that is, two short stressed syllables fill the position of one long stressed syllable (see Sievers 1878:455–456, Kuhn 1983:5–56, Gade 1995:60–66). In Yt 14/1 and 2/11, two short syllables occupy two metrical positions and resolution is suspended in positions 2–3 and 1–2, respectively. Sievers (1893:118) was at a loss to explain these two types.
Syntax of Old Norse Meter
161
consequences for the nature of sentence introduction that can occur in alines. 3. The Placement of Finite Verbs in Independent and Bound Clauses. The terms bound clauses and independent clauses are used here to distinguish between clauses introduced by connectives (such as subordinating conjunctions and ok ‘and’, en ‘but, and’, ea ‘or’, né ‘nor’; bound clauses) and those not so introduced (independent clauses).11 The distinction between bound clauses and independent clauses in poetry is mainly one of word order, in particular with regard to the placement of finite verbs. In independent clauses, a monosyllabic or short disyllabic, metrically resolved verb is usually unstressed and proclitic or enclitic to the first lift in the line.12 The finite verb may also occur in the first position and carry a lift, or, more rarely, it may occur in metrical position 3 or 4 (Types E, D4 in fornyrislag and dróttkvætt), but it cannot be preceded by more than one sentence element without violation of the syntax. In that respect, poetic syntax in independent clauses corresponds to that of prose (the finite verb must occur in syntactic positions 1 or 2, that is, obeying the verb-second constraint).13 Examples 4–7 show the 11
For this distinction, see Kuhn 1933:30, 50–51 and Kuhn 1983:122–123.
12
This is known as Kuhn’s First Law, according to which sentence particles must stand in the first dip of a sentence, proclitically to either the first or the second stressed word of a clause (see Kuhn 1933:8). For a convenient overview and discussion of Kuhn’s First Law in OE poetry, see Stockwell and Minkova 1994.
13
For discussions of the placement of verbs in independent clauses, see Kuhn 1933:30, 51–52, 58, Kuhn 1983:43, 117, 195, and Gade 1995:173–192. In eddic fornyrislag, there are instances of finite verbs in independent clauses occurring further back than position 2, but according to Kuhn (1933:61) the percentage is negligible (1.9%). Kuhn’s percentage is a little skewed, however, because it is based on the ratio of verb placement to lines. If we look at the position of verbs in independent clauses with two constituents or more, the percentages are higher. In mythological fornyrislag poems such as V luspá and Hymiskvia, the placement of finite verbs in main clauses violates the verb-second constraint in 6.7% and 4.2% of the clauses, respectively. The numbers for the heroic poems with native narrative material is lower (Helgakvia Hundingsbana I, 1.7%; Helgakvia Hjrvarssonar, 0%; Helgakvia Hundingsbana II, 2.5%).
162
Gade
possible placements of finite verbs in eddic fornyrislag (the finite verbs are bolded, the alliterating staves italicized). (4) The placement of monosyllabic finite verbs in independent clauses in fornyrislag Type A, metrical position 1: Vsp 42/5 gól um hánom
crowed around him
Type A, metrical position 2: Vsp 58/3 festr mun slitna,
(the) fastening will tear
Type C, metrical position 1: Vsp 28/11 dreccr mi Mímir
drinks mead Mímir
Type D, metrical position 1: Hym 27/1 gecc Hlórrii
went Hlórrii
Type B, metrical position 1: HH I, 27/3 brast rnd vi rnd
crashed rim against rim
Type B, metrical position 2 (rare): HH I, 9/5 hann galt oc gaf he paid and gave Type E, metrical position 1: Br 16/3 svalt alt í sal
chilled everything in (the) hall
Type E, metrical position 2: HH I, 2/1 nótt var í bœ
night fell on (the) farm
Type E, metrical position 3: Grp 25/7 dœgr eitt er ér
day one is for-you
Type E, metrical position 4: Br 19/1 benvnd of lét
wound-wand (expl. of) let (lie)14
The heroic poems in fornyrislag with West Germanic narrative material have the most violations (Brot af Sigurarkvia, 13.5%; Sigurarkvia in skamma, 4.8%; Gurúnarkvia I, 13.5%; Gurúnarkvia II, 15.5%; Gurúnarkvia III, 7.14%; Gurúnarhvt, 16.7; Hamismál, 7.8%). For a discussion of the placement of finite verbs in OE and ON prose, see Swan 1994. 14
In this example, the finite verb is still in syntactic position 2, because the unstressed expletive of does not count as a sentence element (it evolved from a prefix; see note 6 above).
Syntax of Old Norse Meter
163
(5) The placement of disyllabic, short-stemmed metrically nonresolved verbs and trisyllabic verbs in independent clauses in fornyrislag Type A2k, metrical positions 3–4: Vsp 52/5 griótbirg gnata
gravel-mountains tumble
Type A, metrical positions 1–2 (with neutralization in position 2):15 Br 14/1 vacnai Brynhildr awoke Brynhhildr Type C3, metrical positions 2–4 (rare): HH II, 4/5 hon scævai she hastened Type D, metrical positions 2–4: rk 21/5 birg brotnoo
mountains broke
(6) The placement of disyllabic, long-stemmed verbs in independent clauses in fornyrislag Type A, metrical positions 1–2: Vsp 46/7 mælir Óinn
speaks Óinn
Type A, metrical positions 3–4: Grp 1/5 Grípir heitir
Grípir is-called
Type B, metrical positions 2–3: HHv 35/5 hon vissi at
she knew that
Type E, metrical positions 2–3: Hym 13/5 fram gengo eir
forth went they.
In bound clauses, on the other hand, the finite verbs typically occur later than syntactic positions 1 or 2, and they usually carry lifts, even when they are monosyllabic. In the following half-stanza, the bound clause is introduced by the connective enn ‘and’, and the finite verb (knnoo ‘watched’) occurs in syntactic position 5:
15
The term neutralization refers to two unstressed short syllables occupying one metrical position, that is, the unstressed counterpart to resolution (see Gade 1995:61–66).
164
Gade
(7) HH I, 31/5-8
Enn eir siálfir And they themselves
frá Svarinshaugi, from Svarinshaugr
me hermar hug with wrathful mind
her k nnoo. (the) host watched.
Prose order: Enn eir siálfir k nnoo her frá Svarinshaugi me hermar hug. Translation: And they themselves watched the host from Svarinshaugr with wrathful mind. In that respect, then, we may say that independent clauses have V1 or V2 word order, whereas the bound clauses show a strong tendency toward verb-final word order.16 As we see below, this has consequences for sentence introductions in kviuháttr. It must be emphasized, however, that the verb-final word order in bound clauses, which occurs in all skaldic meters as well as in eddic fornyrislag, cannot be observed in the extant ON prose (see Nygaard 1966:270–271, 347, and Swan 1994). 3.1. Finite Monosyllabic Verbs or Short Disyllabic Verbs in Kviuháttr. Because sentence-introductory kviuháttr a-lines are catalectic, monosyllabic and short disyllabic, resolved or neutralized finite verbs can be accommodated only in a small number of positions in independent clauses, as in 8–10 (the finite verbs are boldface).
16
In this respect, the poetic word order in ON bound clauses resembles that of verb-final Modern German subordinate clauses. The main difference is that, unlike in Modern German, connectives such as but and and also trigger this word order (see Kuhn 1933:51–52, 58–68, 108).
Syntax of Old Norse Meter
165
(8) Metrical positions 1 and 2 in catalectic Type A lines of the structure s+s+s17 a. b. c. d.
Yt 8/1 frák, at Dagr Yt 13/1 vasa at bært St 10/1 mik hefr marr St 1/1 mjk erum tregt
heard-I that Dagr was-not that appropriate me has (the) ocean much is-for-me difficult
(9) Metrical position 3 in catalectic Type A lines of the structure s+s+s (extremely rare)18 a. Yt 6/9 núk at veit b. Abj 8/1 vi ví tók c. Glæ 9/ 5 hann of getr
now-I that know of that took-possession he (expl. of) receives
(10) Metrical position 1 in catalectic Type C lines19 a. Yt 11/1 fell Alrekr b. Hkv 20/1 flugu hræleiptr
fell Alrekr flew corpse-lightnings
Hence the fillers of kviuháttr A- and C-lines that introduce independent clauses are very limited from a syntactic point of view. Furthermore, the catalectic nature of the meter prohibits lines ending in an unstressed inflectional syllable, which restricts the inventory of available fillers 17
In the following, s refers to a monosyllable, or a resolved or neutralized disyllable, and ss to a disyllabic word with a long first syllable. The finite verbs are bolded and the alliterating syllables are italicized. 18
As we see below, it appears that, in kviuháttr, adverbs such as nú ‘now’ could be treated as connectives to allow for bound-clause word order, in which case 9a cannot be regarded as an example of an independent clause. Note that the finite verbs in 9b–c are still in syntactic position 2 (as we have seen, the expletive particle of does not count as a sentence element). If the finite verbs had been preceded by two constituents, we would have a violation of the verbsecond constraint in independent clauses (which explains the rarity of such lines in the corpus of kviuháttr poetry).
19
In Egill’s poetry, there are occasional examples of Type D in which the verb in position 1 carries alliteration (St. 14/5, 14/7, 18/5; Abj 3/5), but such lines are not found elsewhere and appear to be an idiosyncrasy of Egill’s.
166
Gade
even more. Consider the following main types: (11) a. b. c. d. e. f. g. h. i. j.
V+Pron+Adv: V+Adv+Adj: V+PP: Pron+V+N: Pron+V+Adv: Adv+V+Pron: Adv+V+N: V+N: Pron/V+Adj: V+Adv:
St 16/1 finnk at opt St 25/5 skalk ó glar Nkt 35/5 stendr í Krists St 10/1 mik hefr marr Nkt 39/5 sá vas norr St 13/1 opt kømr mér Glæ 1/5 ar vas jarl Abj 25/5 hlók lofkst Abj 22/1 hinns fégrimmr Hk 4/1 bar tállaust
find-I that often shall-I yet happy stands in Christ’s me has (the) sea that-one was north often comes to-me there was (the) earl heaped-I praise-pile that-one-is property-grim bore deceitlessly
3.2. Long-Stemmed Finite Disyllabic Verbs in Kviuháttr. In sentence-introductory kviuháttr a-lines, long-stemmed inflected disyllabic verbs in independent clauses can occur only in metrical positions 1–2 in Type A lines of the structure ss + s:20 (12) V+N: Yt 7/1 kvekat dul
speak-I-not folly
Again, the syntactic fillers are extremely restricted because the word in position 3 must be monosyllabic. Trisyllabic finite verbs cannot occur at all in catalectic kviuháttr a-lines because such verbs end in unstressed syllables (compare rk 21/5 birg brotnoo ‘mountains broke’ in 5 above). 3.3. Summary. As we have seen, the catalectic nature of kviuháttr a-lines places severe constraints on the position of finite verbs in independent clauses. However, because kviuháttr b-lines are not catalectic, but correspond to fornyrislag lines with four metrical positions, monosyllabic, disyllabic, and trisyllabic finite verbs can easily be accommodated in b-lines. Yet, the placement of such verbs in b-lines necessarily violates the syntax of independent clauses introduced in a preceding a-line (that is, in syntactic position 1 or 2). Poets composing in kviuháttr meter circumvented that 20
This type of filler is rare: Yt (3 tokens), St (5 tokens), Abj (3 tokens), Hál (1 token), Glæ (1 token), Nkt (18 tokens), Hkv (4 tokens).
Syntax of Old Norse Meter
167
problem in different ways, as discussed in sections 4.1–4.3 below where it is shown how poets depart from an older form, that is, the correlation of stanzaic and helmingr structure with sentence structure, to comply with the requirements of a new form. It is also clear, however, that the poets in some cases were forced by the meter to resort to measures that violated syntax and word order.21 4. Devices to Circumvent Constraints on Syntax Imposed by Meter. 4.1. Run-on Bound Clauses Obliterating Stanzaic Boundaries. As shown above, finite verbs in bound clauses are not subject to the restrictions imposed on finite verbs in independent clauses. They occur later than in syntactic positions 1 or 2 and usually carry a lift.22 Hence poets composing in kviuháttr meter frequently resorted to a concatenation of bound clauses, which allowed for the placement of finite verbs in b-lines. Most frequently, these bound clauses are introduced by en ‘and, but’ or ok ‘and’, but such connectives as ás ‘when’, svát ‘so that’, ars ‘where’, unz ‘until’, ár ‘before’, and vít ‘because’ are also quite common. It may at first seem incongruous that the (now) coordinating conjunctions and and but should trigger the same word order as subordinating conjunctions. However, as Kuhn (1933) has shown, they do so in all branches of Germanic alliterative poetry. In fact, verb-final word order in and-clauses occurs in Old English, Old High German, and Middle High German prose, and it is documented in prose as late as Luther (see Kuhn 1933:63–64, 108). According to Kuhn 1933:51, the anaphoric nature of these connectives caused independent
21
Violations of Kuhn’s First Law are a case in point (see note 12 above). In Egill’s poetry in kviuháttr, for example, finite verbs in bound clauses may occur in unstressed positions in Type C a-lines that are not sentence introductory, as in the following example from St 11/2–4: at í syni mínum | vasa ills egns | efni vaxit ‘that in (the) son of-mine | was-not (a) bad man’s makings grown’ (‘that my son did not have the makings of a bad man’). In this line, the unstressed verb (vasa ‘was-not’), which, owing to its lack of stress, is a sentence particle, does not occur in the licensed position. Kuhn (1933:28–29) notes these instances, but makes no attempt to explain why some poems in kviuháttr behave in this manner. 22
See Kuhn 1933:51–52, Kuhn 1983:43, 118–19, Gade 1995:177–208.
168
Gade
clauses to be reanalyzed as bound clauses, thus causing finite verbs to occur further back and in stressed positions. As stated above, the extant ON prose does not have SOV word order in and- and but-clauses, but as 13 shows, that word order is common in poetry, and more so in kviuháttr than in the other skaldic meters and in fornyrislag (connectives and finite verbs are in boldface). (13) Yt 33–34
Var Gurør inn gfuglæti lómi beittr, sás fyr lngu vas, ok umbr at lum stilli hfu heiptrækt at hilmi dró, ok launsigr inn lómgei su rr af jfri bar, ok bulungr á bei fornum Stíflusunds of stunginn vas.
Became Gurør the grandiose by-deceit overcome, who for long was, and deceit when aledrunk (the) lord (the) head venge-driven against (the) prince plotted, and secret-victory the deceit-minded Ása’s servant over (the) king gained, and (the) ruler on (the) shore old of-Stíflusund (expl. of) pierced was.
Prose order: Inn gfugláti Gurør, sás vas fyr lngu, var beittr lómi, ok at lum stilli dró heiptrækt hfu umbr at hilmi, ok inn lómgei rr su bar launsigr af jfri, ok bulungr vas of stunginn á fornum bei Stíflusunds. Translation: The grandiose Gurør, who lived long ago, was overcome by deceit, and, while the lord was ale-drunk, the vengeance-driven head [Ása] plotted against the prince, and the deceit-minded servant of Ása gained a secret victory over the king, and the ruler was pierced through on the old shore of Stíflusund. Such structures are especially favored by jóólfr in Ynglingatal, but they are also found regularly in other poems in kviuháttr, especially in Hákonarkvia. In longer poems composed in other skaldic meters, such run-on concatenations of bound clauses do not occur. The stanzaic eight-
Syntax of Old Norse Meter
169
line units are also strictly adhered to in encomiastic poems in fornyrislag, as the following stanza from Gísl Illugason’s poem to Magnús berfœttr shows (ca. 1104; Skj AI, 441, st. 5):23 (14)
Hyrr sveimai; hallir urru; gekk hár logi of her eira. Séa knátti ar, es salir fellu, landrá konungs, of lii óris.
Fire surged; halls crumbled; went (the) high flame through (the) counties of-theirs. See (one) could there, as houses crumbled, (the) justice of-(the)-king, around (the) host of-órir.
Prose order: Hyrr sveimai; hallir urru; hár logi gekk of her eira. ar knátti séa konungs landrá, es salir fellu of óris lii. Translation: Fire surged; halls crumbled; the high flame rushed through their counties. There one could see the king’s justice as houses collapsed around órir’s host. There can be no doubt that the abundance of run-on bound clauses that characterize poems in kviuháttr meter and violate the stanzaic boundaries resulted from restrictions imposed by the meter on the syntax. With three exceptions (Yt 3/9, 12/1, 31/5), all the finite verbs in bound clauses in Ynglingatal are found in stressed positions in b-lines, usually in Types B (monosyllabic verb in final position: 43%) or C2 (longstemmed disyllabic verb in positions 3–4: 29%). That circumstance has not gone unnoticed by scholars, and it has been suggested that jóólfr used verb-final placement for psychological reasons, that is, to create a suspense that ultimately reaches its closure with the verb in final position (Åkerlund 1939:249–250; Nordland 1956:124–125). Although we cannot exclude the possibility that the preponderance of Type B and C2 b-lines in Ynglingatal could have been caused by stylistic preferences, it is more likely that we are dealing with an archaic syntactic feature, namely, the necessity to accommodate the finite verbs in bound clauses in stressed, sentence-final positions. In Sonatorrek and Arinbjarnarkvia, there is a similar tendency as far as Type B is concerned (38%, 56%), but in the 23
For the sake of convenience, the stanza is printed as a sequence of half-lines.
170
Gade
later poetry, lines of this type are more negligible (Glæ 18%; Nkt 11%, Hkv 6%). In the later poems, verb-final position occurs much less frequently, allowing for a word order that is closer to that of prose, although V1/V2 in bound clauses is usually avoided. The poets of the eleventh, twelfth, and thirteenth centuries also tend to place the finite verbs in bound clauses in trochaic lines (Type ss+ss: Glæ 30%, Nkt 40%, Hkv 48% as opposed to Yt 15%, St 11%, Abj 12.5%, Hál 21%). Consider the following example from Hákonarkvia (connectives and finite verbs are emphasized):24 (15) Hkv 30
En sú rausn ría knátti Hra grams me himinskautum, hvé vlsungr veita knátti sæmdardag sinnar víxlu.
And that splendor hover could (the) Hrar’s king’s along (the) heaven-corners, how (the) hero entertain could on-(the)-glory-day of-his consecration.
Prose order: En sú rausn grams Hra knátti ría me himinskautum, hvé vlsungr knátti veita sœmdardag sinnar víxlu. Translation: And that splendor of the king of the Hrar could hover along the heavenly corners, how the hero could entertain on the glorious day of his consecration. We see, then, that there seems to be a clear correlation between the placement of finite verbs, the types of metrical lines employed by the poets, and the time of composition. Throughout the period under examination, pronouns and adverbs, which are not connectives, could be treated as such in kviuháttr to allow for bound-clause word order:25 24
In the sequence inf + aux (ría knátti ‘hover could’, veita knátti ‘entertain could’) the inflected aux counts as verb-final, as in Modern German subordinate clauses. 25
Further examples of this irregularity occur in Yt 6/9, 19/5, 29/9; St 12/5; Abj 6/1, 17/1, 19/1, 23/1; Hál 2/1; Glæ 6/1; Nkt 59/1; Hkv 22/1, 25/1, 32/1. Kuhn (1933:61) briefly noted the irregular lines in Yt. See also 9a and note 18 above.
Syntax of Old Norse Meter
(16) Yt 19/5–8
ann hergammr hrægum fœti víts borinn á Vendli sparn.
171
that-one (the) war-buzzard with-(a)-bloody foot from-afar come at Vendel kicked.
Prose order: Hergammr, borinn víts, sparn ann hrægum fœti á Vendli. Translation: The war-buzzard [BIRD OF PREY], come from afar, kicked that one with a bloody foot at Vendel. (17) Hkv 36/1–2
ar sjómjll svífa knátti.
There purse-snow drift could.
Prose order: ar knátti sjómjll svífa. Translation: There purse-snow [SILVER] could drift. In 16, the demonstrative ann (m. acc. sg.) ‘that one’ takes on the function of a connective and causes verb-final word order; and in 17, the adverb ar ‘there’ triggers the same word order (see also the adverb nú ‘now’ in 9a above). Thus, we can say that the syntax of independent clauses in poetry in kviuháttr meter sometimes exhibits verb-final word order, but only if these clauses are introduced by demonstratives or adverbs. There is no natural linguistic basis for treating demonstratives in Old Norse as connectives. Unlike in West Germanic, where demonstratives could occur without relative particles and cause a potential confusion between demonstrative and demonstrative + relative particle (see Kuhn 1933:52; also his note 114), in Old Norse the relative particle er was rarely omitted, and such occurrences are late and part of the learned style (see Nygaard 1966:262). Furthermore, there is no parallel elsewhere in the language to reanalysis of an adverb such as nú ‘now’ as a connective. Clearly, by treating pronouns and adverbs as connectives, the poets increased the inventory of possible bound clauses, which enabled them to avoid VPs in catalectic a-lines.
172
Gade
To be sure, pronouns and adverbs could function as connectives in eddic fornyrislag as well.26 However, the difference is that, in fornyrislag, the verb-second constraint in independent clauses is not as strictly adhered to as in kviuháttr, and verb-final word order also occurs in independent clauses that are not introduced by adverbs and pronouns. Violations are especially prominent in the eddic heroic lays with West Germanic narrative material, and West Germanic influence on the word order of fornyrislag poems cannot be excluded.27 In kviuháttr, however, violations of the verb-second constraint in independent clauses occur only when the clauses are introduced by adverbs and pronouns. It could well be that this is an archaic feature, that is, a remnant of an older verb-final word order in independent clauses.28 In other skaldic meters, there are no instances in which adverbs and demonstratives take on the function of connectives, which suggests that this is another peculiarity in kviuháttr triggered by the necessity to accommodate finite verbs in blines. 4.2. Nominal Elements Obliterating Helmingr or Stanzaic Boundaries. Rather than beginning a new clause in an a-line, poets would postpone nominal elements, thus allowing them to introduce new clauses in blines, even if it meant obfuscating the four- or eight-line metrical divisions that are strictly observed in other ON skaldic meters. These nominal elements are often the subject or appositions to the subject:29
26
See, for example, Vsp 10/5, 20/9, 20/10, 26/3; Hym 1/1; rk 8/5, 10/5, 11/5, 32/5; HH I 4/1, 47/1; HH II, 1/5, 24/5; Rm 23/5.
27
See the percentages given in note 13 above. For West Germanic influence on ON fornyrislag, see Kuhn 1939.
28
That word order is found in such runic inscriptions as the Gallehus inscription from ca. 450 C.E.: ek hlewagasti holtija horna tawido ‘I Hlewagastir descendent-of-Holti (the) horn made’. For this inscription, see Krause 1937:596–598. 29
In 18–20, the verbs and the connectives are bolded, elements belonging to the same NPs are italicized, and the nominal elements obfuscating the metrical boundaries are bolded and italicized.
Syntax of Old Norse Meter
173
(18) Abj 10–11: 10. ar stó mér mrgum betri hddfinnndum á hli ara tryggr vinr minn, sás trúa knáttak, heiróar hverju rái,
There stood for-me than-many better hoard-finders on side one (the) faithful friend of-mine, whom trust could-I, honor-thriving by-each counsel,
11. Arinbj rn, es oss einn of hóf, knía fremstr, frá konungs fjónum, vinr jóans, es vættki ló í herskás hilmis gari.
Arinbj rn, who us alone (expl. of) removed, of-men (the) foremost, from (the) king’s hatred, (the) friend of-(the)-lord, who never lied in (the) warprone monarch’s mansion.
Prose order: ar stó tryggr vinr minn, sás knáttak trúa, á ara hli mér, betri mrgum hoddfinnndum, heiróar hverju rái, Arinbj rn, fremstr knía, es einn of hóf oss frá konungs fjónum, vinr jóans, es vættki ló í herskás hilmis gari. Translation: There stood the faithful friend of mine, whom I could trust, by my one side, better than most hoard-finders, honor-thriving by each counsel, Arinbj rn, foremost of men, who alone removed me (literally ‘us’) from the king’s hatred, the friend of the lord, who never lied in the war-prone monarch’s mansion. The stanzaic division in 18 is that of the standard editions. However, in Abj 10, the NP in line 5 (tryggr vinr minn ‘the faithful friend of mine’), which begins the new helmingr, is the postponed subject from the previous half-stanza, and Arinbjrn (st. 11/1) and vinr jóans ‘the friend of the lord’ (st. 11/5) are NPs belonging to the same complex subject. Stanzas of this type are not found in Ynglingatal, but they are quite common in kviuháttr until the late thirteenth century.
174
Gade
If four lines are quoted separately, the helmingr is often syntactically incomplete unless it is sentence-introductory, as shown in 19. (19)
Hál 16
jólna sumbl; en vér gótum stillis lof sem steinar brú.
(the) gods’ drink; and we fashioned (the) ruler’s praise like (a) stone’s bridge.
Prose order: Jólna sumbl; en vér gótum stillis lof sem brú steinar. Translation: The gods’ drink [POETRY]; and I (literally ‘we’) fashioned the ruler’s praise like a bridge of stone. In this half-stanza from Háleygjatal, which Snorri quotes as a separate unit in the Prose Edda to illustrate a collective name for the ON gods (jóln ‘gods’), the poetic circumlocution in line 1 (jólna sumbl ‘the gods’ drink [POETRY]’) must belong to the preceding, no longer extant helmingr. As was the case in 2 above, Snorri adhered to the four-line half-stanza unit and ignored the fact that, as it stands in the text, line 1 occurs in syntactic isolation. Again it is clear that poets composing in kviuháttr meter departed from a deeply entrenched convention to comply with the requirements imposed by a new form. 4.3. Preposition of Nominal Elements. Quite often nominal elements, usually the subject or appositions to the subject, are preposed to allow for sentence introduction in a following bline. In 20, the adjective in line 5 (lofanlig ‘praiseworthy’) modifies the subject (himna sól ‘sun of the heavens’) of the following bound clause, introduced in line 6:
Syntax of Old Norse Meter
(20) Hkv 40
En elding ifla foldar langa lei lsa knátti, lofanlig, á er litum skipti himna sól vi hafrul.
175
And (the) lightning of-(the)-hawks’ land (a) long way shine could, (the) praise-worthy, when color exchanged heavens’ sun with (the) ocean radiance.
Prose order: En elding ifla foldar knátti lsa langa lei, á er lofanlig sól himna skipti litum vi hafrul. Translation: And the lightning of the hawks’ land [ARM > GOLD] could shine a long way when the praiseworthy sun of the heavens exchanged color with the ocean-radiance [GOLD]. As we have seen, the catalectic a-lines favor nominal fillers (Types A, C, and D), and this led poets to postpone or to prepose concatenations of nominal elements, further obscuring helmingr and stanza boundaries. Such stanzas are not found in Ynglingatal, but they are particularly prominent in Egill’s poetry, in Háleygjatal, in the first and last parts of Nóregs Konungatal, and in Hákonarkvia.30 Earlier scholars have noted that poems in kviuháttr are characterized by syntactic nominal variation in a much higher degree than, for example, ON poems in dróttkvætt meter (Marold 1983:144–153).31 The reason for this is clear: nominal compounds, NPs, and PPs could easily be accommodated in catalectic alines, whereas VPs could not. To be sure, such nominal fillers are 30
See St 2/5, 3/1, 8/5, 17/5, 21/5, 24/5; Abj 1/5, 2/1, 8/5, 10/5, 11/1, 11/5, 13/5, 14/5, 15/5, 18/5, 19/5, 21/5, 22/5, 23/5; Hál 4/1, 6/5, 10/5, 11/5, 12/5, 13/5, 16/1; Nkt 1/5, 71/5, 72/5, 73/5, 74/5, 75/5, 76/5, 77/5, 78/5, 80/5, 81/5, 82/5, 83/5; Hkv 2/5, 6/5, 15/5, 18/5, 24/5, 25/5, 26/5, 27/5, 29/5, 31/5, 33/5, 36/5, 37/5, 38/5, 40/5, 41/5.
31
Nominal variation comprises epic repetition, series of poetic synonyms, and series of appositions and appellatives. Consider the variation in 18: tryggr vinr minn ‘(the) faithful friend of-mine’, betri mrgum hddfinnndum ‘better thanmost hoard-finders’, heiróar hverju rái ‘honor-thriving by-each counsel’, Arinbj rn ‘Arinbjrn’, fremstr knía ‘foremost of-men’, vinr jóans ‘the-friend of the-lord’.
176
Gade
characteristic of a-lines in other Germanic meters as well, but not to the same extent as in kviuháttr. As was the case with the types of fillers discussed in 4.1, there is a marked shift in the types of metrical lines used with these nominal fillers starting with Glælognskvia. Consider the following percentages:32 (21) Metrical Types of kviuháttr a-lines from the ninth century until 1263
Yt St Abj Hál Glæ Nkt Hkv
Type A
Type C
Type D
62% 66% 30% 63% 71% 80% 72%
2% 18% 35% 6% 12% 12% 24%
35% 16% 31% 28% 6% 2% 1%
Poems traditionally (or tentatively) dated to the ninth and tenth centuries show a fairly high percentage of Type D lines, but the frequency of that type decreases dramatically in the eleventh, twelfth, and thirteenth centuries, while at the same time there is an increase in trochaic lines (see also the discussion in 4.1 above). This circumstance would seem to have implications for the debate concerning the date ascribed to a poem like Ynglingatal.33 Both preposition and postponement of nominal (and other) sentence elements occur frequently in dróttkvætt poetry, too, but unlike in kviuháttr, such elements never obfuscate helmingr and stanzaic boundaries in dróttkvætt (see Kuhn 1983:202–205). It is clear that preposition crossing metrical boundaries is another characteristic prompted by metrical restrictions in kviuháttr.
32
The occurrence of Type A (s+s+s) is negligible throughout the centuries, and the type has been omitted in this table. For the dates of the poetic texts, see section 1 above.
33
The distribution of other types of lines over the four centuries examined here also shows significant differences, as was pointed out in 4.1, but a discussion of that material lies outside the scope of the present discussion.
Syntax of Old Norse Meter
177
5. Conclusion. The discussion above has shown that the catalectic kviuháttr a-lines, the main characteristic of this meter, caused a breakdown in the syntactic and metrical structure, and this breakdown sets kviuháttr apart from other ON meters. First of all, the aberrant syntax of kviuháttr, which has been commented on by earlier scholars but never explained, is caused by metrical restrictions imposed on syntax, more specifically, on the placement of finite verbs in catalectic a-lines. Already by the time of the composition of the earliest poetry in kviuháttr meter, the placement of finite verbs in independent clauses was subject to the verb-second constraint. In other skaldic meters and in eddic fornyrislag, monosyllabic, disyllabic, and trisyllabic finite verbs could easily be accommodated in clause-introductory a-lines, but the metrical inventory of alines available to poets composing in kviuháttr meter was restricted to the catalectic versions of Types A, C, and D. In lines of these types, finite verbs could occur only in a small number of positions, that is, positions 1, 2, and 3 in a-lines, and position 1 in c-lines, proclitically to the first or the second stressed word in the sentence.34 Placing the finite verb of independent clauses in the following b-line was not an option, because such a word order violated the verb-second constraint. The finite verbs in bound clauses, on the other hand, could occur further back and in stressed positions, and they do so in all branches of ON poetry. In fact, poetry in kviuháttr meter from as late as 1263 (Hkv) still displays verbfinal syntax in bound clauses, even though that word order is not attested in the earliest ON prose. Furthermore, demonstratives and adverbs could function as connectives to allow for bound-clause word order. In Old Norse, there is no linguistic basis for reanalyzing demonstratives and adverbs as connectives, but this could be an archaic feature that was retained in kviuháttr because it increased the inventory of bound clauses and allowed the poets to place finite verbs in b-lines (sentence-final position). In Hákonarkvia (ca. 1263), this type of bound clause is always introduced by the adverb ar ‘there’ plus a compound (Type C; Hkv 22/1, 25/1, 32/1, 36/1). The construction is clearly no longer productive and has been reduced to a mere syntactic metric stereotype.
34
See, again, Kuhn’s First Law. Recall that finite verbs in metrical position 3 are very rare.
178
Gade
Second, and as a consequence of the need to accommodate finite verbs in b-lines, poems in kviuháttr are characterized by a strong preference for bound clauses. This in turn led to concatenations of such clauses obfuscating the eight-line stanzaic division that is inviolable in all other skaldic meters, as well as in encomiastic fornyrislag poetry (see 13 and 14 above). Third, to allow for clause onset in tetrasyllabic b-lines rather than in catalectic a-lines, the poets often postponed or preposed nominal elements, making extensive nominal variation another feature characteristic of kviuháttr.35 Furthermore, because of nominal variation, the poets often failed to treat the four-line half-stanza as an independent syntactic unit, unlike their practice when composing in other skaldic meters.36 Because the helmingr unit was so deeply entrenched in ON poetic tradition, later authors of prose works, such as Snorri Sturluson, automatically cited such half-stanzas in a prose context, despite the fact that they are syntactically incomplete (see 2 and 19 above). When faced with the restrictions imposed by a new poetic form, then, poets composing in kviuháttr meter were forced to relax structural rules that were strictly adhered to in other skaldic meters. To observe the constraint of word order, and, more specifically, the placement of inflected verbs in independent clauses, they adopted measures that violated such Nordic innovations as the eight-line stanza and the fourline helmingr units. Thus, it is clear that the syntactic and structural abnormalities of kviuháttr poems are both principled and predictable. In that respect, kviuháttr is a valuable heuristic in regard to poetic syntax,
35
Marold (1983:214), in particular, emphasizes the stylistic difference between jóólfr’s dróttkvætt and kviuháttr poems.
36
Egill Skallagrímsson’s poetry in kviuháttr (St. 194 lines; Abj, 196 lines) contains 15 violations of the helmingr unit, but his poetry in the skaldic meters runhent (Hfulausn, 152 lines; Skj AI:35–39) and dróttkvætt (Lausavísur, 352 lines; Skj AI:48–59) has no violations. The corresponding numbers for Eyvindr Finnsson are as follows: kviuháttr has 7 violations (Hál, 100 lines), dróttkvætt has no violations (Lausavísur, 112 lines; Skj AI: 961–70). That also holds true for Sturla órarson: kviuháttr has 14 violations (Hkv, 336 lines), hrynhent has no violations (Hrynhenda, 168 lines; Skj AII:102–108), hagmælt has no violations (Hrafnsmál, 160 lines; Skj AII, 119–124), and dróttkvætt has no violations (Hákonarflokkr, 88 lines; Skj AII:124–127).
Syntax of Old Norse Meter
179
and it lends strong support to our reconstruction of the grammar and prosody of ON poetry. REFERENCES Åkerlund, Walter. 1939. Studier över Ynglingatal. (Skrifter utg. av Vetenskapssocieteten i Lund, 23.) Lund: Gleerup. Faulkes, Anthony (ed.). 1991. Snorri Sturluson Edda: Háttatal. Oxford: Clarendon Press. Finnur Jónsson. 1929. Snorre Sturlusons Háttatal. Arkiv för nordisk filologi 45.229–269. Fulk, Robert D., and Kari Ellen Gade. 2002. A bibliography of Germanic alliterative meters. Internationales Jahrbuch für Germanistik 34.87–186. Gade, Kari Ellen. 1995. The structure of Old Norse dróttkvætt poetry. (Islandica, 49.) Ithaca and London: Cornell University Press. Gade, Kari Ellen. 2002. History of Old Nordic metrics. The Nordic languages, vol 1, ed. by Oskar Bandle, Kurt Braunmüller, Ernst Håkon Jahr, Allan Karker, Hans-Peter Naumann, and Ulf Teleman, 865–870. Berlin: Walter de Gruyter. Heusler, Andreas. 1925. Deutsche Versgeschichte mit Einschluss des altenglischen und altnordischen Stabreimverses, vol. I, parts I–II. Einführendes; Grundbegriffe der Verslehre; der altgermanische Vers. (Grundriss der germanischen Philologie, 8.1.) Berlin: Walter de Gruyter. [2nd, unrevised edition, 1956.] Hollander, Lee M. 1947. Is the skaldic stanza a unit? Germanic Review 22.298– 319. Indrebø, Gustav. 1928. Tri namn or Haakonssoga. Maal og minne, 16–120. Konrá Gíslason. 1881. Nogle bemærkninger angående Ynglingatal. Aarbøger for nordisk oldkyndighed, 185–251. Krause, Wolfgang. 1937. Runeninschriften im älteren Futhark. Halle an der Saale: Max Niemeyer. Krag, Claus. 1991. Ynglingatal og Ynglingesaga: En studie i historiske kilder. (Studia Humaniora, 2.) Oslo: Universitetsforlaget. Kuhn, Hans. 1929. Das Füllwort of-um im Altwestnordischen. Eine Untersuchung zur Geschichte der germanischen Präfixe. Ein Beitrag zur altgermanischen Metrik. Göttingen: Vandenhoeck and Ruprecht. Kuhn, Hans. 1933. Zur Wortstellung und -betonung im Altgermanischen. Beiträge zur Geschichte der deutschen Sprache und Literatur 57.1–109. Kuhn, Hans. 1939. Westgermanisches in der altnordischen Verskunst. Beiträge zur Geschichte der deutschen Sprache und Literatur 63.178–236.
180
Gade
Kuhn, Hans. 1969. Die Dróttkvættstrophe als Kunstwerk. Festschrift für Konstantin Reichardt, ed. by Christian Gellinek, 63–72. Bern and Munich: Francke. Kuhn, Hans. 1983. Das Dróttkvætt. Heidelberg: Carl Winter. Marold, Edith. 1983. Kenningkunst: Ein Beitrag zu einer Poetik der Skaldendichtung. (Quellen und Forschungen zur Sprach- und Kulturgeschichte der germanischen Völker, Neue Folge, 80.) Berlin: Walter de Gruyter. Möbius, Theodor. 1883. Über die Ausdrücke fornyrislag, kviuháttr, ljóaháttr. Arkiv för nordisk filologi 1.288–294. Neckel, Gustav (ed.). 1962. Edda: Die Lieder des Codex Regius nebst verwandten Denkmälern (Germanische Bibliothek, 4th ser.), vol. 1, Texte. 4th rev. ed. by Hans Kuhn. Heidelberg: Carl Winter. Nordland, Odd. 1956. Hfulausn i Egils saga: Ein tradisjonskritisk studie. Oslo: Det norske samlaget. Noreen, Erik. 1924. Kuia: En hypotes. Festschrift Eugen Mogk zum 70. Geburtstag 19. Juli 1924, 61–65. Halle an der Saale: Max Niemeyer. Nygaard, Marius. 1906. Norrøn syntax. Oslo: Aschehoug. [2nd, unrevised edition, 1966.] Sapp, Christopher D. 2000. Dating Ynglingatal—chronological metrical development in kviuháttr. Skandinavistik 30.85–98. See, Klaus von. 1967. Germanische Verskunst. (Sammlung Metzler, 67.) Stuttgart: Metzler. Sievers, Eduard. 1878. Beiträge zur Skaldenmetrik. Beiträge zur Geschichte der deutschen Sprache und Literatur 5.449–518. Sievers, Eduard. 1879. Beiträge zur Skaldenmetrik. Beiträge zur Geschichte der deutschen Sprache und Literatur 6.265–376. Sievers, Eduard. 1893. Altgermanische Metrik. (Sammlung kurzer Grammatiken germanischer Dialekte. Ergänzungsreihe II.) Halle an der Saale: Max Niemeyer. Skj AI–AII: Den norsk-islandske skjaldedigtning, vols AI–AII. Tekst efter haandskrifterne, ed. by Finnur Jónsson. 1912–1915. [Reprint Copenhagen: Rosenkilde og Bagger, 1967.] Stockwell, Robert P., and Donka Minkova. 1994. Kuhn’s laws and the rise of verb-second syntax. Swan et al. 1994, 213–231. Swan, Toril. 1994. A note on Old English and Old Norse initial adverbials and word order with special reference to sentence adverbials. Swan et al. 1994, 233–270. Swan, Toril, Endre Mørck, and Olaf Jansen Westvik (eds.). 1994. Language change and language structure: Older Germanic languages in a comparative perspective. (Trends in Linguistics: Studies and Monographs, 73.) Berlin: Mouton de Gruyter.
Syntax of Old Norse Meter
181
Vries, Jan de. 1962. Altnordisches etymologisches Wörterbuch. Leiden: Brill. [2nd, revised edition, 1977.] Wessén, Elias. 1915. Om kuida i namn på fornnordiska dikter: Ett bidrag till eddadiktningens historia. Edda 4.127–141.
Germanic Studies Indiana University Bloomington, IN 47405 USA [
[email protected]]
Journal of Germanic Linguistics 17.3 (2005):183–217
Aspectual Posture Verb Constructions in Dutch Maarten Lemmens Université Lille 3/ Centre National de la Recherche Scientifique, France This paper discusses the auxiliated posture verb constructions in Dutch (liggen/zitten/staan+te+V ‘sit/lie/stand to V’) that have progressive, durative, or habitual interpretation. My analysis, based on a large corpus of written texts, reveals, first of all, that to a large extent these constructions follow the regular (that is, non-aspectual) use of the three cardinal posture verbs as basic locational verbs. Second, the corpus used for the present study reveals clear experientially based patterns in the type of verbs that occur in the auxiliated posture verb construction. The data suggest that, at least in the written language, the construction has retained a link with the postural (or by extension, the locational) source. This sheds light on some clear semantic differences between the auxiliated posture verb constructions and another common progressive construction in Dutch, aan het V zijn ‘be at the V-inf’. The paper also briefly considers the progressive construction lopen+te+V ‘run to V’, showing that it is less grammaticalized and still predominantly tied to motion events.
1. Introduction: Progressive Constructions in Dutch. As one may read in virtually any Dutch grammar, progressive aspect can be expressed in a variety of ways. Apart from the base form, the two most common constructions are illustrated in the following examples: (1) Ik was aan het lezen / aan het wachten / aan het slapen. I was at the read-INF / at the wait-INF / at the sleep-INF ‘I was reading/waiting/sleeping.’
I would like to thank the anonymous referees of this journal for their comments on a previous version of this paper. Responsibility for any remaining inaccuracies is, of course, mine. © Society for Germanic Linguistics
184 Lemmens (2) Ik zat te lezen / ik stond te wachten / ik lag te slapen. I sat to read-INF / I stood to wait-INF / I lay to sleep-INF ‘I was (sitting and) reading/(standing and) waiting/(lying and) sleeping.’1 Although in both cases, the non-finite complement expresses an ongoing activity, syntactically these constructions are quite different. In the construction in 1 (henceforth PREP-PROGRESSIVE), the semantically light verb zijn ‘be’ is followed by the abstract locational preposition aan ‘at’, which takes a definite noun phrase as a complement with the nominalized infinitival verb as its head.2 In the construction in 2 (henceforth POS-PROGRESSIVE), a cardinal posture verb (CPV) refers to the agent’s posture while carrying out the activity expressed in the infinitival complement. The difference between the pos- and the prep-progressive seems obvious, as the posture verbs are lexically more precise in specifying the agent’s posture. However, this need not always be so, as illustrated by the examples in 3.3
1
The English glosses for the cardinal posture verb (CPV) progressive will be of the type CPV to V. This should be understood as the equivalent of English be+Ving, which in the context of posture can be translated as sitting and Ving, standing and Ving, or lying and Ving. As can be seen in these examples, as well as some other examples quoted in this paper, the posture verbs are fully inflected, just like any other main verb. The posture verbs, as well as lopen ‘run’, are strong verbs having a vowel change in the past forms. 2 3
For a syntactic discussion of the aan het V construction, see Booij 2002.
Unless stated otherwise, all examples have been drawn from the INL corpus on which the present study is based (see p. 187 below for more information on the corpus used). Some minor editing may have been done to avoid irrelevant complications.
Posture Verb Constructions
185
(3) a. Ze zaten met de snelheid van een lift they sat with the speed of a lift tien meters op en neer te suizen. ten meters up and down to whizz ‘They were whizzing up and down ten meters with the speed of a lift.’ b. Onze ploeg stond lamlendig te hockeyen. our team stood sluggishly to hockey ‘Our team was playing hockey sluggishly.’ c. Wat zit ik hier toch rond te lopen? (pers. attestation) what sit I here (toch) around to walk? ‘Why on earth am I walking (around) here?’ In contexts where the agent’s posture is no longer at issue and actually quite incompatible with the activity expressed by the complement verb, the pos- and prep-progressives have become interchangeable. Brisau (1969:75) explicitly states: “the fact that [the verbs] can be easily substituted for one another when they are not in a context, points to their being subordinate to the main verb in meaning: primarily they serve to indicate aspect.” Similarly, the Algemene Nederlandse Spraakkunst (ANS)—the standard comprehensive grammar of Dutch—comments on non-postural uses of the pos-progressives as follows: “De betekenis van deze hulpwerkwoorden […] is echter soms verzwakt tot ‘bezig zijn met’ […]. Zinnen als deze hebben vaak een ondertoon van irritatie en komen vooral in gesproken taal voor” (Haeseryn et al. 1997:973). (The meaning of these auxiliaries has often weakened to ‘be occupied with’ [ML: that is, progressive]. Sentences such as these often indicate irritation and occur mainly in spoken language.)4 The ANS still leaves room for some semantic specificity, but relegated to the connotational domain, or that of register. According to Leys (1985:275–276), constructions in which the CPV no longer refers to posture have the meaning of een zich bevinden ‘being somewhere’, and consequently have become synonymous with the construction with zijn ‘be’. Boogaart (1991) follows Leys in grouping both under the 4
The ANS is also available in electronic form (E-ANS) at (last accessed May 30, 2005):
.
186 Lemmens heading locational.5 In her general overview of progressive constructions in Germanic languages, Ebert (2000:619) cannot detect any clear differences between the two, and even remarks that, at least for agentive verbs, “the choice between prep, pos, and simple forms seems to be partly dependent on personal preferences.” From a cognitive point of view, I would like to defend the idea that these constructions are not mere formal and/or stylistic variations. Consequently, while they may be interchangeable in some contexts, this cannot be done without a change in meaning, and the choice may not be as free as Ebert suggests. At the same time, the semantic differences between the pos- and prep-progressive are not the major concern of the present article.6 The present analysis of the pos-progressive builds on previous elaborate analyses of Dutch posture verbs (Van Oosten 1986, Lemmens 2002). I demonstrate that these constructions, at least as used in the written register, retain a link with the verbs’ locational source semantics. Thus, the analysis developed here is similar to Lødrup’s (2000:122) syntactic analysis of Norwegian CPV och V constructions: “it has often been pointed out that the positional verbs […] are used to express progressive aspect […]. However, they are not really gram-maticalized, they keep their literal meaning, and they may be modified, for example by a locative.” She does not clarify what exactly she means by “literal meaning” (her study focuses on the syntactic status of the different types 5
Boogaart’s grouping is justified given that his goal is to distinguish between these two constructions on the one hand, and the base form that is also often used with the progressive meaning, on the other. 6
In its discussion of the pos-progressives, the ANS mentions two other constructions. The first uses the verb hangen ‘hang’, as in De was hangt te drogen ‘The laundry hangs to dry’. This construction is not dealt with in this article because it has not really grammaticalized in the way that the three other posture verbs have. It is restricted to cases where the agent is indeed suspended. The second construction mentioned in the ANS involves the (general) motion verb lopen ‘run’, as in Hij loopt de hele tijd te zingen, literally ‘He runs all the time to sing’ (= ‘He is singing all the time’). The lopen+te+V construction does merit some analysis, especially in contrast to the (more frequent) pos-progressives. It is discussed briefly at the end of this article, but my main focus remains on the liggen/zitten/staan+te+V constructions, which have a much wider coverage and, logically, a higher degree of grammaticalization.
Posture Verb Constructions
187
of pseudo-coordination); this is a central point of the present article. In particular, examining the type of agent used in the pos-progressive, as well as the type of complement verb that figures in the construction, reveals that, for written language, the construction retains a link with the verb’s source semantics, which explains semantic differences between the pos- and pres-progressives. The data underlying the present study have been drawn from the largest computerized Dutch corpora freely available to researchers, at the Instituut voor Nederlandse Lexicologie, Leiden.7 Unfortunately, this corpus is not fully representative, given the absence of spontaneous spoken data, its restriction to non-fictional prose, and the heavy predominance of Netherlandic Dutch. The subcorpus of contemporary Dutch prose (1970–1995) that has been selected for the present analysis contains nearly 26.4 million words, of which 90% is Netherlandic Dutch. The only set of Belgian Dutch data is a two-month compilation of a respected Flemish newspaper (De Standaard, November–December 1995). Consequently, while the overall analytical value of the present analysis remains, it is impossible to test to the fullest the ANS claim cited above that the semantically bleached construction is mainly restricted to the spoken register. Thus, the present study should be complemented with a study of the influence of register to establish whether CPV constructions are indeed more common in informal language or fiction. Furthermore, there is a need for a more careful analysis of regional differences, which intuitively seem to be pertinent to some of the uses of CPV constructions.8 7
For more information, see
. The use of the INL corpora is gratefully acknowledged.
8
While the Belgian and Netherlandic corpora used here are not fully comparable, there is no significant regional difference in frequency for the three constructions. The normalized frequencies (averaged per 1 million words) in the Netherlandic (N) and Belgian (B) subcorpora being 27.01 (N) versus 23.01 (B) for zitten+te+V; 6.91 (N) versus 6.14 (N) for liggen+te+V; and 25.65 (N) versus 25.31 (B) for staan+te+V. The absence of any significant difference is somewhat unexpected, judging from my own intuitions and comments by Northern Dutch speakers on some of the Belgian Dutch examples that I have cited elsewhere. This result may be due to the regional differences being leveled out in written language.
188 Lemmens All occurrences of zitten, liggen, and staan have been excerpted from this subcorpus. Table 1 provides an overview of the number of attestations retrieved, setting the progressive constructions apart from the nonaspectual uses. The row percentages indicate the relative distribution within the aspectual or the non-aspectual uses. CONSTRUCTION TYPE non-aspectual CPV+te+inf TOTAL % of progressives in grand total
N row % N row % N row %
staan
zitten
liggen
TOTAL
36,883 48.7% 658 48.1% 37,541 48.7%
21,908 28.9% 529 38.6% 22,437 29.1%
16,948 22.4% 182 13.3% 17,130 22.2%
75,739 100% 1,369 100% 77,112 100%
1.8%
2.4%
1.1%
1.8%
Table 1. CPV+te+inf versus other uses of Dutch CPVs. As these figures clearly indicate, the progressive constructions constitute only a small portion of the total: 1.1% for liggen+te+V, 1.8% for staan+te+V, and 2.4% for zitten+te+V. Further comments on the results in table 1 are provided in section 3.1. The remainder of this article is structured as follows. In the next section, I briefly describe the auxiliation process itself. The main part of the paper, section 3, presents a more detailed analysis of the posprogressive as attested in the corpus, discussing the general distribution of the different progressives and the type of figure (section 3.1), the type of complement verb (section 3.2), and some other semantic particulars as reflected in the use of temporal, aspectual, and locational modifiers (section 3.3).9 Section 4 briefly considers the lopen+te+V progressive.
9
The term complement verb is not entirely correct given that it can be regarded as the main verb of the sentence in an auxiliated construction. For ease of formulation, however, I continue to use this term.
Posture Verb Constructions
189
2. From Postural to Aspectual. The grammaticalization of posture verb constructions with aspectual value, often also referred to as auxiliation, has not gone unnoticed in the literature, where it has been shown to be a recurrent pattern in many (unrelated) languages (see, for example, Heine 1991, Bybee et al. 1994, Kuteva 1999). Characterizing the pos- and prep-progressive constructions as locational, as do Leys (1985) and Boogaart (1991), reflects the (generally accepted) idea that the origin of progressive constructions often lies in morphemes and/or constructions having a locational meaning (see, for example, Comrie 1976 or Bertinetto et al. 2000). The pos-progressive that contains a specific locational verb is a salient exponent of such locational constructions.10 Kuteva (1999:192) argues that the prerequisite for auxiliation is that the posture verbs are used as “the unmarked/canonical encodings of position of physical objects in space.” The intuitively appealing logic behind this approach is that the posture verbs must first lose their “postural” semantics, so that they are able to encode the location of any entity, animate or inanimate. Kuteva supports this hypothesis with data from European and non-European sources. Here I do not deal with the question of whether Dutch has followed the same path of auxiliation; this issue requires a careful diachronic study, which, while certainly called for, is beyond the scope of the present paper. However, it can be mentioned in passing that Song’s (2002:378) analysis of Korean progressive posture verb constructions challenges Kuteva’s claim, as the Korean equivalent of the verb lie “cannot encode the spatial position of inanimate entities although it has already been pressed into service to express progressive aspect.” Whatever the evolutionary path of the auxiliation of Dutch posture verbs, it cannot be denied that they are the habitual encoding of an entity’s location. In fact, their use is essentially obligatory when one 10
Obviously, the syntactic patterns of the pos-progressive differ across languages, even within the Germanic family. Most Germanic languages have a coordination pattern (CPV and V); for example, Swedish. Dutch seems to be the only language within the family that has evolved toward a different pattern with te+infinitive (see Van Pottelberghe 2002 for an interesting diachronic account). Leys (1985) mentions that some Dutch and Flemish dialects still have the coordination pattern, but I do not discuss these here.
190 Lemmens wants to express location of an animate or inanimate entity; the general verb zijn ‘be’ is usually unacceptable (unlike in English or in French), as evidenced by the following examples.11 (4) a. Ik {sta / *ben} in de rij. I {stand / *am} in the line ‘I am standing in line.’ b. Mijn sleutels {liggen / *zijn} op de tafel. my keys {lie / *are} on the table ‘My keys are (lying) on the table.’ Many progressive uses of Dutch posture verbs are a logical extension of their basic locational sense. As a result, the inanimacy of the subject in the examples in 5 is not to be attributed to the auxiliation process as such, but to the use of staan, liggen, and zitten as default locational verbs. (5) a. Een vrachtwagen staat voor ons te a truck stands in-front-of us to ‘A truck is unloading in front of us.’
lossen. unload
b. Het schip lag twee jaar later nog weg the ship lay two years later still away te roesten aan een kade in Stockholm. to rust at a quay in Stockholm ‘Two years later the ship was still rusting away at a quay in Stockholm.’ c. Deze wijn zit te popelen in een fles … this wine sits to be-anxious in a bottle ‘This wine is anxious to get out of the bottle …’ Example 5a illustrates the use of staan to describe a situation in which an entity is resting on its base (sta-vlak ‘stand-side’, as Van den Hoek 1971 11
On the inter-Germanic differences in the use of posture verbs, see also Lemmens 1995. While zijn is unacceptable in these sentences, the more formal expression zich bevinden ‘be found’ is generally acceptable, but less common.
Posture Verb Constructions
191
neologistically terms it).12 Example 5b shows the use of liggen to encode the default posture of ships. Example 5c is a typical case of zitten in reference to containment, less common with aspectual uses, but quite productive in its regular use (accounting for some 45% of the locational uses of zitten).13 The next section presents a more detailed discussion of the aspectual constructions in their own right, by looking more closely at, among other things, the types of verbs that occur in this construction. I show that even in cases when there is no further reference to posture, the locational source semantics determines, to a certain extent, the use of CPVconstructions. These cases provide the key to understanding the difference between pos- and prep-progressives. 3. A Corpus-Based Analysis of Pos-Progressives. 3.1. General Distribution of the CPV+te+infinitive Construction. Comparing the frequency of the aspectual uses of the CPV+te+V construction to its regular uses (see table 1 above), one notices that the overall frequency ranking is comparable for both, staan > zitten > liggen, with staan occurring more then twice as often as liggen.14 However, while both categories have the same frequency ranking, their groupinternal ratio is not the same: for staan, both non-aspectual and progressive usage have a comparable frequency (48%), whereas for liggen, the progressive construction is much less frequent compared to the verb’s other uses (a drop of 9%). Zitten, in turn, is the only verb for which the progressive use is more frequent percentage-wise than the regular one. How can the distribution shown in table 1 be explained? Newman and Rice (2004)—a source of inspiration for the present article—present a similar corpus-based analysis for English posture verbs in three different constructions: (i) verb-particle constructions (sit down, 12
Notice further the metonymy in this example, as it is obviously not the truck itself doing the unloading.
13
Since my major concern here is the grammaticalized status of the aspectual constructions, I do not discuss the logic of these locational uses in detail here; see Van Oosten 1986 and Lemmens 2002 for further discussion. 14
Four cases combining different CPVs (such as wie niet voortdurend staat, zit of ligt te konverseren ‘who not continuously stands, sits, or lies to converse’) have been omitted; hence the total of 1,369 instead of 1,373.
192 Lemmens stand up, and lie down), (ii) simultaneous conjunctions (CPVing and Ving, as in sitting and reading), and (iii) consecutive conjunctions (CPVed PART and Ved, as in stood up and walked out). Newman and Rice regard (i) and (ii) as stative uses and (iii) as a dynamic use, as the CPV codes the onset of an action sequence. The frequency ranking for (i) and (ii) in the British National Corpus (BNC) is sit > stand > lie, and for (iii) the frequency ranking is stand > sit > lie. In other words, for the stative uses (i) and (ii), sit is more frequent; for the dynamic CPV and Ved, stand is much more frequent. For the two conjoined constructions, a higher frequency is aligned with the collocation of a greater range of verbs. Newman and Rice see a possible experiential motivation for this distribution that can be related to grammaticalization patterns in other languages. Sitting is the most relaxed, comfortable posture associated with many social events (eating, drinking, etc.) and intellectual activities (deskwork, reading, etc.). A standing posture, by contrast, requires much more muscular effort and is usually maintained for a short period of time only. It is nevertheless still relatively frequent, since it is the starting posture for common activities such as walking and running, and it is also the posture often associated with other common activities such as giving speeches. Finally, a lying posture has a more limited purpose (sleeping, resting, etc.), and may have a negative connotation often associated with weakness, decay, or death. The higher frequency of stand up observed for the dynamic consecutive conjunction can be attributed to the fact that standing up is often the starting point of the event chain. It is thus very close to the inchoative use of Dutch staan, as in Wat staat er te gebeuren? which literally translates as ‘What stands to happen?’ (or ‘What is about to happen?’).15 Clearly, Newman and Rice are well aware of the fact that the English constructions have not grammaticalized as they have in other languages, with the CPVs having retained their lexical load. However, their corpus-based analysis reveals striking similarities in patterns of use of these constructions in English and in languages that do have grammaticalized CPV-constructions. How can we explain the difference in frequency ranking in Dutch data presented in table 1? One may argue that this difference is due to 15
Such inchoative (or ingressive) uses of staan have not been included in the present study, since it focuses on progressive constructions.
Posture Verb Constructions
193
differences in register, since the BNC, the main corpus on which Newman and Rice base their frequency count, also contains spoken data.16 While some caution is warranted, the comparison is still valid for a number of reasons. First, the percentage of spoken material in the BNC is rather low (10%), and is thus unlikely to have a drastic effect on the results. Second, and more importantly, the ranking sit > stand > lie is the same for the spoken BNC data (with, understandably, even a higher ratio of sit down). Third, Newman and Rice point out that these relative frequencies have been found in all the corpora they used, covering both spoken and written English, and both American and British variants. The Dutch data parallel the BNC data in its low frequency of liggen for both aspectual and non-aspectual uses. It is possible that this parallel is again due to the fact that a lying posture allows only a small range of activities and is associated with inactivity, as illustrated below. At the same time, note the high frequency of staan for both categories, which may lend support to the claim made by Van Oosten 1986 and Lemmens 2002 that Dutch staan is associated with a default position for both humans and inanimates much more than English stand is. Recall that to describe an object resting on its base Dutch always uses staan. Interestingly, and in line with the frequency differences observed here, English allows the verb sit in this context. Consider the following contrast in the way Dutch and English may express the location of a computer on a desk: (6) a. Mijn computer staat op mijn bureau. my computer stands on my desk b. My computer is sitting on my desk. In English, the verb sit emphasizes the inactivity (hence Newman’s [2002] term for this usage inactivity sit), whereas no such meaning is present in the Dutch construction with staan; it simply encodes the default posture for computers (on their base). Note that to render such an inactivity reading, Dutch can actually have a stacked staan te staan construction, often with other modal particles added, as shown in 7.
16
I thank one of the anonymous reviewers for drawing my attention to this methodological issue.
194 Lemmens (7) Mijn computer staat daar maar wat te staan. my computer stands there just a-bit to stand ‘My computer is merely sitting there (without being used).’ A similar type of “stacking” is possible with liggen te liggen and zitten te zitten, but all three are relatively uncommon (for example, the corpus used for this study does not contain any such examples). Returning to the higher frequency of Dutch staan, I argue that this is due to the verb being strongly associated with the entity’s canonical and/or functional position when engaged in an activity. The next section discusses in more detail the range of activities that occur with each verb. Before we turn to the complement verbs, let us briefly look at the distribution of the agent in these aspectual constructions. As table 2 shows, the aspectual construction is restricted primarily to a human agent (83.8% in total), as opposed to the non-aspectual uses, which are not restricted in this way (approximately 25%). However, this is not so much due to the CPV as such (inanimate subjects being obviously quite common in the non-aspectual uses), but rather to the semantics of the progressive itself which has an overall preference for animate agents. Indicative in this respect is that the prep-progressive construction aan het V zijn ‘be at the V’, where zijn ‘be’ does not impose any thematic restrictions on its subject, has a similar preference for a human agent (73% of a total of 1,040 occurrences in the INL corpus). At the same time, 65% of the pos-progressives with a human subject are used to refer to posture, as opposed to approximately 10–15% in non-aspectual use.
Posture Verb Constructions
CONSTRUCTION AGENT staan+te+V zitten+te+V liggen+te+V human 564 498 84 col % 85.7% 94.1% 46.2% row % 49.0% 43.3% 7.3% animal 15 19 12 col % 2.3% 3.8% 6.6% row % 31.9% 42.6% 25.5% entity 74 11 84 col % 11.2% 2.1% 46.2% row % 43.2% 6.5% 49.7% abstract 1 10 10 col % 0.2% 1.9% 5.5% concrete 73 1 74 col % 11.1% 0.2% 40.7% plant 6 2 col % 0.9% 1.1% row % 75.0% 25.0% TOTAL 658 529 182 col % 100% 100% 100%
multiple 3 75.0%
1 25.0% 0.6%
1 25.0%
4 100%
195
TOTAL
1,150 83.7% 100% 46 3.4% 100% 170 12.3% 100% 17 1.2% 153 11.1% 8 0.6% 100% 1,373 100%
Table 2. Distribution of subject types for the CPV-progressive. Notice, however, that the overall high percentage of human agent is not observed with respect to liggen: 46.2% of the INL examples have an inanimate subject. The reason for this should be clear from the foregoing discussion of the overall frequency hierarchy (see table 1): for humans, it is quite difficult to be engaged in an activity while lying (and as we shall see, the range of possible activities is more limited too). Thus, human subjects are expected to occur less frequently with liggen. In addition, liggen is the default locational verb that expresses the positioning of ships, symmetrical objects, substances, and geotopographical locations (cities, countries, etc.), all of which are well represented in the corpus. Table 2 also confirms some other typical patterns for the non-aspectual uses of CPVs, such as the common use of zitten for small animals (rabbits, frogs, etc.), birds, and insects (see Lemmens 2002).
196 Lemmens 3.2. Verb Complements of the CPV+te+infinitive Construction. Once again inspired by the corpus study in Newman and Rice 2004, I also investigated the complement verbs that occur in pos-progressives. In addition to the bodily posture meaning, CPVs involve “inherent stative semantics or temporal ‘unboundedness’ of the verb situation” (Kuteva 1999:206).17 In the first stage of the grammaticalization process, the complement verbs are restricted to those expressing activities compatible with the posture expressed by the CPV. In later stages, the complement verb may express activities less compatible with the posture, or activities that do not imply a posture at all. An overview of complement verb types may thus shed more light onto the grammaticalization process itself. Not only may it reveal some of the general patterns (as shown for English by Newman and Rice 2004), but the range of verb types may also give an indication of the degree of grammaticalization, which may be different for each of the three verbs.
COMPLEMENT VERB TYPES
staan 210 48.1%
zitten 157 35.9%
liggen 70 16.0%
TOTAL 357 100%
Table 3. Range of complement verb types. When looking at the Dutch complement verbs occurring in the posprogressive, one notices that the frequency ranking presented in table 1 (staan > zitten > liggen) correlates with a wider choice of complement verb types, as shown in table 3.18 Here Dutch shows a different pattern from the one observed by Newman and Rice (2004) for English. While it has been shown that Dutch pos-progressives also mostly occur in postural contexts, with subjects restricted primarily to human agents, the description above suggests that the Dutch construction may have a higher 17
Kuteva uses the term stative in the sense of temporally unbounded, as opposed to Vendler (1967) or Comrie (1976), who use this term to refer to verbs that express states, and thus, as a rule, are incompatible with the progressive, as in *I am believing or *I am seeing you under the table. 18
The total of complement verb types is not the mere sum of the individual counts, since complement verbs occurring with two or more posture verbs, such as wachten ‘wait’, have been counted only once.
197
Posture Verb Constructions
grammaticalized status than its English counterpart. This is clearly revealed in the range of verbs that occur in these constructions. Table 4 presents a summary of these verbs, listing those for which N10. staan+te+V wachten ‘wait’ kijken ‘watch’ trappelen ‘stamp’ dringen ‘jostle’ opwachten ‘wait (for s.o.)’ springen ‘jump’ juichen ‘cheer’ popelen ‘be anxious’ praten ‘talk’ spelen ‘play’ pronken ‘prance’ slapen ‘sleep’ N<10 TOTAL
N
%
120 18.2 56
8.5
38
5.8
27
4.1
23
3.5
21
3.2
19
2.9
16
2.4
15
2.3
15
2.3
11
1.7
10
1.5
zitten+te+V wachten ‘wait’ kijken ‘watch’ lezen ‘read’ eten ‘eat’ springen ‘jump’ praten ‘talk’ spelen ‘play’ luisteren ‘listen’ mediteren ‘meditate’ schrijven ‘write’ aankomen ‘happen’
287 43.6 N<10 658
N
%
liggen+te+V wachten 147 27.8 ‘wait’ slapen 29 5.5 ‘sleep’ 18
3.4
18
3.4
18
3.4
13
2.5
12
2.3
11
2.1
10
1.9
10
1.9
10
1.9
233 44.0 N<10 529
N
%
45
24.7
44
24.2
93 182
51.1 1,369
Table 4. Complement verbs for CPV-progressives (with N10). One of the results that stands out very clearly is that for all three constructions the verb wachten ‘wait’ is most frequent. This is of course fully in line with the durative semantics of the posture verbs and with our experience that waiting is usually done while sitting, standing, or lying (in that order of frequency). Strikingly, the verb wachten has not been
198 Lemmens attested in the prep-progressive (aan het V zijn ‘be at the V’) in a total of 1,040 occurrences drawn from the same corpus. It is possible in principle to say Ik ben aan het wachten ‘I am at the wait-INF’ (an informal search on the Internet produced a number of examples). Yet there seems to be a strong preference for using the pos-progressive instead, which supports the strength of the experiential association. There is a distinct drop in frequency—by one-half—for both staan and zitten when they occur with the second verb kijken ‘watch’. (The two verbs wachten and kijken account for 401 attestations, or almost 30% of the total.) That the frequency of this verb occurring with liggen is considerably lower (only 3 attestations, 1.6%) comes as no surprise: it is a posture from which one’s visual perception is quite limited. The high frequency of kijken with staan and (less so) zitten may be explained as follows. First, the verb is the prototypical expression of what can be regarded as our most prominent perceptional activity, automatically activated when not sleeping (so mostly when standing or sitting); yet, in itself, a relatively passive activity, and thus in full accord with the posture progressive. Second, the construction staan te kijken frequently occurs in two idiomatic contexts. In one context, it has the metaphorical meaning ‘be surprised’, as in the following examples: (8) a. Verzorgers van een dierenasiel in Engeland stonden wel keepers of an animal-shelter in England stood wel heel vreemd te kijken toen ze de post open maakten. very strange to watch when they the mail open made ‘Keepers of an animal shelter in England were quite surprised when they opened the mail.’
Posture Verb Constructions
199
b. De Nederlanders stonden ervan te kijken. the Dutch stood there-from to watch ‘The Dutch were quite surprised.’ Geklopt op beaten on
het the
jaren meester for-years master
terrein terrain
waar ze zich where they themselves
waanden: de taal. thought: the language
‘Beaten on ground where they thought themselves superior for years: language.’19 The two examples illustrate the two syntactic construction types that typically have this meaning: (i) ADJ staan te kijken with the adjective expressing the strange look on the beholder’s face (as in vreemd or raar both meaning ‘strange, weird’), and (ii) ergens van staan kijken, literally ‘from somewhere stand to watch’, where the locative somewhere refers to the origin of the surprise, and mostly does so anaphorically (as er[van] ‘thereof’ does in 8b above). The other idiomatic usage of staan te kijken that plays a role here is a reference to inactivity: someone standing and watching is often associated with that person not engaging in any action, usually in these contexts evaluated negatively. It can roughly be considered as the equivalent of English stand around doing nothing. The third factor motivating the suitability of kijken, this time as a complement of zitten, is that we often sit and watch all kinds of performances, such as plays, films, concerts, television programs, etc. A more detailed discussion of such social activities is presented below. Table 4 also shows that apart from the high frequencies of the typical combinations discussed above, the frequency of the other verbs, when taken individually, drops sharply. The second verb that occurs with liggen is slapen ‘sleep’, which comes as no surprise. After that, the 19
This sentence refers to the fact that in competitive games between Belgium and the Netherlands dealing with language, it is mostly the Flemish teams who win. The surprise referred to here has to do with the commonly held belief (false but sadly enough still nurtured by some) that the Belgian variant of Dutch is “substandard”. As a native speaker of the latter, I thought this example too nice to be excluded.
200 Lemmens frequency of combinations drops to six and lower. Thus, while the pattern of the pos-progressive is in principle relatively open and productive (within some limits as outlined here), the frequency count clearly shows that there are typical combinations that seem to have acquired strong unit status. From an encoding perspective, then, this means that these units will have a greater chance of being selected than other constructions available to the Dutch speaker, as was illustrated above for the contexts of ongoing waiting: the unit zitten/liggen/staan te wachten is well-entrenched, which makes it the most likely candidate for selection, much more than the aan het wachten zijn alternative. Grouping the individual verbs in larger semantic classes, as does table 5, further reveals interesting experiential patterns. The different classes confirm the general stative (and typically atelic) character of the verbs that occur in the pos-progressive, because they revolve around states, cognition processes, rest, etc. The largest class is that of stative verbs. Here, this term is used to refer to (i) verbs that express the meaning of being located statically, such as is the case for standing, hitchhiking, posing, birds brooding, etc., or (ii) a number of otherwise hard to classify verbs that express a (temporary) state, such as N/ADJ zijn ‘be N/ADJ’. Also, “waiting” has been included in this class, which of course affects the overall frequency of the class.
Posture Verb Constructions
SEMANTIC CLASS
EXAMPLE (English translation) wait, pose
N row % jump, dance N MOTION row % N listen, watch PERCEPTION row % talk, listen N COMMUNICATION row % sleep, rest N REST row % N think, wonder COGNITION row % N rust, ripen, CHANGE rot row % N write, paint CREATE row % N enjoy, rejoice PSYCH row % vomit, cry N BODY PROCESSES row % N eat, drink INGESTION row % N sports, games PLAY row % N shine, flash EMISSION row % N < 1% row % TOTAL FREQ N TOTAL ROW % row % STATIVE
201
staan+ zitten+ liggen+ TOTAL te+V te+V te+V 161 43.6% 152 76.0% 80 56.3% 69 62.2% 12 15.4% 7 9.7% 18 35.3% 5 11.1% 25 55.6% 16 36.4% 8 18.2% 25 61.0% 22 56.4% 58 65.9% 658 48.1%
158 42.8% 33 16.5% 55 38.7% 41 36.9% 11 14.1% 60 83.3% 11 21.6% 40 88.9% 19 42.2% 22 50.0% 36 81.8% 16 39.0% 6 15.4% 21 23.9% 529 38.6%
50 13.6% 15 7.5% 7 4.9% 1 0.9% 55 70.5% 5 6.9% 22 43.1% 0.0% 1 2.2% 6 13.6% 0.0% 0.0% 11 28.2% 9 10.2% 182 13.3%
369 27.0% 200 14.6% 142 10.4% 111 8.1% 78 5.7% 72 5.3% 51 3.7% 45 3.3% 45 3.3% 44 3.2% 44 3.2% 41 3.0% 39 2.8% 88 6.4% 1,369 100.0%
Table 5. Most frequent verb classes for CPV-progressives.
202 Lemmens More interesting are some of the other groupings that confirm patterns observed in earlier analyses of posture verbs.20 It comes as no surprise that staan has the highest number of movement verbs, as this is generally the start position for walking and, by extension, any movement. It should be observed, however, that the movement verbs occurring in the pos-progressive are generally those expressing motion of a body part, such as zwaaien ‘wave’, or motion not affecting the overall bodily posture, such as draaien ‘turn’, wiebelen ‘wobble’, trillen ‘tremble’, etc.21 These types of motion are compatible with the posture, and logically the three constructions differ in the types of motion events that they describe: staan trappelen ‘trample’, dringen ‘jostle’, springen ‘jump’, dansen ‘dance’, aanschuiven ‘line up’, draaien ‘turn’, trillen ‘tremble’, wankelen ‘wobble’, wiebelen ‘wiggle’, schommelen ‘swing’, etc. (46 types, 152 tokens) zitten springen ‘jump’, aankomen ‘arrive’, trappelen ‘trample’, draaien ‘turn’, knikken ‘nod’, etc. (14 types, 43 tokens) liggen rollen ‘roll’, woelen/draaien ‘toss/turn’ (in bed), spartelen/kronkelen ‘squirm/squiggle’, drijven/deinen ‘float/bob’, etc. (10 types; 15 tokens) The verb liggen is compatible with verbs expressing motion that can cooccur with lying. Floating objects also occur in this construction (but not in the two others), since floating objects will naturally assume a horizontal posture, which, as pointed out earlier, also explains why
20
See, among others, Van Oosten 1984, Serra Borneto 1996, Lemmens 2002, Newman 2002, Newman and Rice 2004.
21
This is similar to what Talmy (2000) has termed self-contained motion, although his concern is more with the figure not changing location (as opposed to “translational motion”) rather than its posture. Notice further that basically any activity (selling, scrubbing, etc.) involves motion one way or another, but included in our class of motion events are only those verbs generally considered genuine motion verbs.
Posture Verb Constructions
203
liggen is the default verb used to describe the positioning of ships (see example 5b). The motion verbs that can occur with staan often include in their conceptual structure the notion of maintaining an upright position, as well as using one’s feet as in dancing, trampling, etc. Such verbs may also express balance-related notions, such as wobbling or tottering, which are strongly associated with the standing posture. A related case, included not in the group of motion events but in that of play, concerns sports such as tennis, football, or hockey, where being on one’s feet seems to be the salient posture throughout the game. This explains the existence of constructions such as staan te voetballen, staan te tennissen, staan te hockeyen ‘stand to play football/tennis/hockey’, as in 9. (9) Af en toe stond Krajicek zelfs fantastisch te tennissen. once in a while stood Krajicek even fantastically to play-tennis ‘Now and then, Krajicek was even playing (tennis) fantastically.’ In sum, staan+te+V can be used in reference to TRANSLATIONAL MOTION events where the moving entity changes its actual location, but these contexts generally involve what could be termed a dynamicized standing posture, that is, the agent is moving around but saliently maintains a standing posture.22 The motion verbs one finds with zitten mostly refer to body part motion as well, such as nod or gesture, or those compatible with a sitting posture, as in zitten te fietsen ‘sit to cycle’. Some motion verbs (springen ‘jump’ and aankomen ‘arrive’) seem to be in conflict with zitten. However, this conflict is only apparent, since these verbs are used in a metaphorical sense: zitten te springen means ‘be impatient to act’ and probably derives from the fact that impatient people eager to start doing something may often be described as “jumping” up and down on their chair. Also, many of the self-contained motion verbs—that is, verbs describing situations where the agent does not really change its location such as dringen ‘jostle’, trappelen ‘trample’, springen ‘jump’—are used in this sense when they occur with staan. It is no coincidence that verbs with the meaning ‘be anxious’ (either as a metaphorical extension, or as 22
The term translational motion (in addition to self-contained motion, see note 21) is also borrowed from Talmy 2000.
204 Lemmens their basic meaning, such as popelen ‘be anxious’ whose motion sense has been lost) do not occur with liggen, since it is not the position associated with eagerness and readiness for action. The case of zitten aan te komen ‘sit to arrive’ is more complicated. A typical example is provided in 10. (10) Er zit een nieuwe roman aan there sits a new novel at ‘A new novel is coming up …’
te komen … to arrive
The arrival is, as in all the other nine occurrences of this construction, metaphorical. Notice that the aspectual nature is quite complex: arriving is generally an achievement verb (punctual), yet in this context it has become an ongoing event. Strictly speaking, it is still a progressive, yet the nature of the event leads to an ingressive interpretation as well (‘is about to’). The only cases found in INL in which the actual motion is truly incompatible with sitting is example 3a above, to which we can add the personally attested 3c, both repeated here for convenience, as well as another example drawn from the Internet (using WebCorp).23 (11) a. Ze zaten met de snelheid van een lift they sat with the speed of a lift tien meters op en neer te suizen. ten meters up and down to whizz ‘They were whizzing up and down ten meters with the speed of a lift.’ b. Wat zit ik hier toch rond te lopen? (pers. attestation) what sit I here (toch) around to walk? ‘Why on earth am I walking (around) here?’
23
See (last accessed May 30, 2005).
Posture Verb Constructions
205
(12) Omdat ik achter een trein aan zit te hollen, because I after a train at sit to run, heb ik de trein waar ik eigenlijk in hoor te zitten gemist. have I the train where I actually in have to sit missed ‘Because I was running for a train, I missed the one that I actually had to be on.’ The example in 12 is noteworthy, since it contains aspectual zitten (combined with the motion verb hollen ‘run’), as well as locational zitten (not necessarily exclusively postural, although sitting is pretty much the normal posture when in the train). Given that zitten is the most apostural of the three verbs, it seems natural that precisely this verb acquires various dynamic senses. Liggen and staan retain most of their postural semantics, even if the latter is already somewhat more tolerant to nonpostural uses. This is in line with what one finds in other languages as well. Heine et al. (1991) observe that while grammaticalization toward progressive, durative, or habitual markers is found with all three posture verbs, the ones expressing the concept of sitting seem to be the most common, followed by lie-verbs, and then by stand-verbs.24 Newman (2002), who provides a good overview of aspectual patterns, suggests that the extension through time is perhaps strongest for sitting and lying, that is, for postures that humans or objects are able to maintain for a longer period of time, as opposed to standing, which requires more physical energy. Also, the fact that the sit-verb often covers a more varied range of postures than the lie- and stand-verbs may contribute to its having gone furthest in its semantic bleaching, making it more available for grammaticalization. Clearly, this factor has been influential in Dutch, where zitten is often devoid of any postural semantics, particularly in its non-aspectual use. Some examples of such apostural uses of zitten+te+V are provided below.
24
Paul Roberge (personal communication) indicates that in Afrikaans, lê ‘lie’ is the basic verb for the progressive.
206 Lemmens (13) a. Ik zat een jaar te dubben hoe ik weer aan de slag kon. I sat a year to worry how I again at the beat could ‘For a year, I was worrying over how to get back to work.’ b. Hij zit een groot deel van zijn tijd in harde steen te boren. he sits a large part of his time in hard stone to drill ‘Most of the time, he’s drilling in some hard rock.’ At present, however, uses of zitten combined with translational motion verbs as in the examples above—where virtually nothing remains of the static nature of zitten, and where it merely indicates progressive aspect— are still relatively infrequent and indicative of a more informal register. The Internet is generally a good source for informal language, yet a search via WebCorp did not yield many such cases. An informal Google search for the strings zitten/zit/zat/zaten te V returned a number of examples. While more limited than the WebCorp’s interface, the results may suggest a certain pattern of ongoing grammaticalization worthy of further exploration. The total number of hits is still limited, but in many cases the complement verb means either ‘to run’ (lopen/rennen/hollen) or ‘to walk’ (wandelen). In contrast, no results have been found for other more specific verbs of human (self-propelled) translational motion, such as stappen ‘step, walk’, tuimelen ‘tumble’, hinkelen ‘hop’, or manken ‘limp’. This does not mean that combinations with these verbs would be completely unacceptable, but they are clearly rare, which suggests that the construction is still fairly limited in scope and restricted to the most typical verbs of walking/running. Moreover, often these combinations include the particle rond ‘around’. This particle converts the true translational motion into one occurring (repetitively) within a certain location. The repetitive character of the motion, as well as the motion not really progressing to its endpoint, is what often gives these constructions their negative connotation (see the quote from the ANS above).25 The Google search also yielded some examples of zitten te vliegen ‘sit to fly’. However, these were restricted to insects or birds—for which zitten is the default posture and flying their proto-movement, such as walking/running is for humans—or cases where people are flying 25
As Newman and Rice (2004) show, the particle around is also quite typical for the English V and V construction, such as sitting around and doing nothing.
Posture Verb Constructions
207
virtually, using a flight simulator, and are thus seated. One interesting case, found in a report written by a ten-year-old girl, concerned Harry Potter flying in his friend’s magic car: hij zit in de auto en zit te vliegen ‘he sits in a car and sits to fly’. Here too, the agent is seated while flying. In addition, one may justifiably wonder to what extent the use of zitten te vliegen is triggered by the use of zitten in the preceding clause. The fact that this is a child’s report may be important. A brief count of pos-progressives in other children’s narratives (the frog story data, see Berman and Slobin 1996 and Verhoeven and Strömqvist 2004) by children of 5, 7, and 9 years old, shows that zitten is far more frequent (22 or 67%) than liggen (6 or 18%) or staan (5 or 15%). These figures should be interpreted with caution, given the small sample, as well as the fact that two important characters in the story are a frog and a dog whose default posture is expressed by zitten. Nevertheless, the higher frequency of the verb, as well as its higher number of different verb complements (13, as opposed to four for staan and two for liggen), invite the hypothesis that children may be more sensitive to the apostural use of zitten, and may thus overgeneralize its use. For example, a narrator may use it in contexts in which the figure’s posture is in conflict with that expressed by zitten; for example, a standing boy who zit te schreeuwen ‘sits to yell’. A final point worth mentioning with respect to zitten and motion events is that the verb can also be used in other constructions, without a te+V complement, to express motion. Consider the following Internet examples: 26 (14) a. Als we tikkertje aan het spelen zijn met mijn vrienden, if we tag at the play-INF are with my friends, zit hij altijd achter mij aan! sits he always after me at ‘When we are playing tag, he’s always after me!’ b. De FBI zit de moordenaar achterna. the FBI sits the murderer behind-after ‘The FBI is chasing the murderer.’ 26
Examples 14a–c are from , <www.moviedb.nl>, and <www.deboekenplank.nl/naslag/aut/b/brown_d.htm>, respectively.
208 Lemmens c. De politie en de moordenaar zitten hen op de hielen ... the police and the murderer sit them on the heels ‘The police and the murderer are (following) hot on their heels.’ The common expressions achter iemand aanzitten, literally ‘sit behind at someone’, iemand achterna zitten, literally ‘sit someone behind-after’, and iemand op de hielen zitten, literally ‘sit someone on the heels’—almost untranslatable because of their complex particle constructions—require motion verbs in English, such as chase or follow. The closest English equivalent is the expression be after someone. Dutch has some true motion verbs that more congruently express such followevents; for example, volgen ‘follow’ or achter iemand aanhollen/ aanlopen ‘run after somebody’. The idioms with zitten are often used to profile the closeness of the chase, a logical extension of the notion of contact that is incorporated into the semantics of zitten (see Lemmens 2002:114ff.). The aspectual construction illustrated in 12 above is an example of such idiomatic use. Returning to the groupings in table 5, we observe another pattern that is more clearly reflected in the verb complements for zitten, that is, that most of our everyday activities involve sitting. These activities can be grouped as follows: Social interaction (eating, drinking, talking, meeting, negotiating, etc.) Cognitive activities (reading, thinking, brooding, meditating, etc.) Creative activities (writing, typing, etc.; knitting, sewing, etc.) Visual/auditory perception (watching television, a play; listening, etc.) None of these activities requires strong physical strain, yet some effort (physical and cognitive) is required over a longer period of time. The sitting posture is optimal, as it permits activity while in a comfortable position. I suggest the term ACTIVE REST to denote the notion strongly associated with zitten. Verbs from the visual or auditory domain can also occur with staan, yet these verbs are slightly different from those occurring with zitten, as they all express the emission by the entity itself. Such verbs belong to a separate category of emission verbs and describe events such as shining, glittering, prancing, and showing off. That these verbs occur more
Posture Verb Constructions
209
frequently with staan finds an experiential grounding as well: standing entities are more easily perceived by others (compare with English stand out and outstanding). A final note on the grouping presented in table 5 concerns some of the patterns found with liggen that reflect the verb’s strong association with inactivity (lying being the typical posture for complete rest) and decay (lying being the typical posture when dead or ill). Thus, it is not surprising that complements of this verb include verbs such as rotten ‘rot’, roesten ‘rust’, beschimmelen ‘getting moldy’, and niets doen ‘doing nothing’, whose meanings center around these notions. Mostly, these situations are evaluated negatively, which may explain the use of liggen te V to convey the speaker’s negative attitude, as in Lig niet te zeuren!, literally ‘lie not to whine’, that is ‘Stop whining!’, where the verb typically expresses a speech activity, such as whining or complaining. Intuitively, this usage feels more typical of Northern Dutch—in Belgian Dutch zitten would be more common—but the corpus used in this study does not allow verification of this intuition. To conclude, the types of complement verbs that occur in the posprogressive are largely compatible with the posture verbs’ stative semantics. Moreover, the collocational clusters reveal some typical, experientially grounded associations, which strengthens the idea that the range of verbs occurring in these constructions is not as random as it may appear (and, in fact, quite different from the range of verbs that occurs in the prep-progressive). Staan has the widest range of verb complements, a fact attributable to (i) it being the default posture for humans and many inanimate objects, and (ii) it being the starting posture for activity. At the same time, zitten appears to be most permissive, allowing true translational motion verbs as its complements, in line with the stronger apostural character of the verb. Additional data from the Internet and from children’s narratives suggest some ongoing grammaticalization, a hypothesis to be pursued further. It has been observed that the use of zitten in combination with a translational motion verb is facilitated by use of the particle rond ‘around’, deemphasizing the translational character of the motion. This brings us to another aspect of the posprogressive that has not been discussed, namely, the occurrence of other adverbial or aspectual modifiers. This issue is addressed in the next section.
210 Lemmens 3.3. Durative and Locative Semantics of the Pos-Progressive. This section presents a short discussion of various modifiers that occur in the pos-progressive. As I show below, the distribution of temporal and aspectual modifiers may shed light on the differences between the posprogressive and the prep-progressive aan het V zijn ‘be at the V’. Although a full comparison is beyond the scope of this paper, some observations are worth considering for the sake of completeness. Kuteva (1999:209) observes that auxiliated posture verb constructions often contain temporal adverbial phrases, such as all the time, all day long, etc., that emphasize the durative or progressive aspect, and states that such adverbials are “redundant rather than necessary.” Table 6 provides an overview of the temporal and aspectual modifiers found in the pos-progressive in our corpus (the percentages are computed relative to the total number of pos-progressives). DURATIVE GENERAL 167 98 12.2% 7.1%
REPETITIVE MOMENTARY TOTAL 44 61 370 3.2% 4.4% 26.9%
Table 6. Aspectual and temporal adverbials expressing duration. The results indeed suggest that these adverbials are redundant: only 12.2% (167) of the 1,373 cases have a durative modifier that can take the form of an aspectual marker, such as nog ‘still’, or a temporal marker, such as de hele dag ‘all day’ or jarenlang ‘for years’. In other words, the posture verb seems to do all the work in this domain. In a similar vein, the low percentage of momentary modifiers, such as op dat moment ‘at that moment’, can be attributed to the conflict between punctuality expressed by the modifier and duration expressed by the verb. Note, however, that in the prep-progressive—where no such salient durative focus can be attributed to the general verb zijn ‘be’—the ratio of durative modifiers is not much higher (16%), although momentary modifiers are somewhat more frequent here than in the pos-progressive (11% versus 4.4%). The frequencies of repetitive modifiers, such as vaak ‘often’, elke dag ‘every day’, and of general temporal modifiers that merely specify a certain time frame, such as gisteren ‘yesterday’ or vorig jaar ‘last year’, are also comparable. In short, these aspectual and temporal modifiers are not particularly revealing, at least not in our data selection, although they
Posture Verb Constructions
211
do contribute to the general characterization of the two progressive constructions. Another type of modification worth investigating is the presence of locational modifiers. When posture verbs are used as non-aspectual locational verbs, a locational complement is compulsory, as shown in 15a. In contrast, with the aspectual usage it is optional, as in 15b. (15) a. Hij zit op een stoel. / *Hij zit. he sits on a chair / he sits ‘He sits on a chair.’ b. Hij zit (op een stoel) te lezen. he sits on a chair to read ‘He sits (on a chair) to read.’ (= ‘He is reading.’) Despite its optional character, a locational complement is still expressed in 44% (601) of the aspectual constructions, which is relatively high, especially when compared to the 12% of locational modifiers that occur in the prep-progressive. In the case of other types of modifiers occurring in the posprogressive, they mostly concern a modifier pertaining to the figure or to the manner in which the action is carried out, which is more closely associated with the figure than with the action expressed by the complement verb. In contrast, the modifiers found in prep-progressives typically apply to the event expressed by the complement verb and the speed with which it evolves. Taken together, these observations support the claim that the posprogressive is still very much tied to the verb’s stative and locational character. Therefore, this construction is more typical in contexts where the action is viewed from a wider perspective, situated within the location at hand or as part of the setting. In contrast, the prep-progressive focuses on the action itself. 4. Lopen+te+V: A More Dynamic Alternative. While our main focus is on the pos-progressives, the construction lopen+te+V ‘run to V’ deserves a brief description. ANS treats this construction on a par with the three posture verb constructions, which is not completely unjustified as its usage does overlap to some degree with the others. Below is a typical example from the corpus.
212 Lemmens
(16) Ze liepen de hele dag te sjouwen met kartonnen dozen. they ran the whole day to haul with cardboard boxes ‘They were hauling boxes all day.’ However, the construction with lopen is considerably more restricted than the pos-progressives. First, it is markedly less frequent than the posprogressives: only 98 examples are found in the corpus (little over half the frequency of liggen+te+V). Second, it requires an agentive subject: 100% of the examples contain a human or an animate subject. Third, there is a strong tendency for the complement verb to express a motion event: in 61.2% of the cases, the complement verb expresses either a motion event (as in the two examples above), or an event in which motion is implied. A typical case of the latter is the combination with certain sport activities (16%), such as lopen te voetballen ‘run to play soccer’ or lopen te hockeyen ‘run to play hockey’, which has also been attested for the staan+te+V construction (see examples 3b and 9 above). In other words, the semantic bleaching has gone less far for lopen+te+V, as the latter retains an even stronger link with the source semantics of lopen (animate motion). However, this does not mean it cannot be used in contexts where no (real) motion is at issue, as in 17. (17) a. Er komen bij hem veel there come to him many
mensen people
die al jaren lopen te dokteren. who for years run to doctor ‘Many people come to him who for years have been going to the doctor.’ b. Ik loop al vijftien jaar te roepen I run already fifteen years to yell dat ik piloot wil worden. that I pilot want to-become ‘For fifteen years I’ve been saying that I want to become a pilot.’
Posture Verb Constructions
213
What distinguishes these uses of the lopen+te+V construction from those with liggen, zitten, or staan is that the former strongly imply multiple occurrences of the event over a given time span, mostly marked as having continued for quite some time (which may be a subjective interpretation of the speaker). Example 17 is a good illustration: it expresses how for years many people have been going to, most likely, different doctors on different occasions. The use of multiple agents is quite common for the lopen+te+V construction, and there is also a higher percentage of durational modifiers (as in all the examples cited so far), namely, 24.4% as opposed to 12.2% for the pos-progressives. For the cases where there is no real motion involved, this percentage nearly doubles to 43%. To sum up, as a logical extension of the dynamic motion event expressed by lopen, the general tendency is for lopen+te+V to emphasize combined iterative and durative aspect more than the pos-progressives do. This tendency is reflected in the use of aspectual modifiers, and particularly in the choice of complement verbs, generally restricted to either motion verbs or verbs expressing an activity that is part of a larger motion event. Among complement verbs that do not refer to motion, the most common are speech act verbs and verbs of social interaction with negative connotation (16 out of 28 non-motion events, or 57%). For example, people are said to be running around and screaming or yelling at one another, or getting angry at each other. Such uses of “negative” speech act verbs have also been found in the pos-progressive (see some comments above), but they are markedly less frequent. Two reasons can be suggested for these verbs being more common in the lopen+te+V construction. First, within the smaller scope of the semantics of the construction itself, there is often the idea that these fights, arguments, and the like are reciprocal and/or accumulative, which is consistent with the durative-iterative nature of the construction. Second, in a larger perspective, and as indicated by ANS (see the quotation in section 1), the construction tends to have a negative connotation. This is certainly true for these speech acts, but applies to most other complement verbs as well, as, for example, in 17 above, where dokteren, literally ‘to doctor’, clearly has a negative connotation, as opposed to a more neutral expression such as naar de dokter gaan ‘go to a doctor’.
214 Lemmens A final note pertaining to the lopen+te+V construction concerns a regional difference. Intuitively, the construction is felt to be more common in the Netherlandic variant of Dutch than in the Belgian. As mentioned above, the corpus does not allow for a systematic evaluation of these differences, yet a partial comparison is possible by examining the texts of two newspapers, similar in size, over a period of two months: the Belgian newspaper De Standaard and the Dutch newspaper NRC Handelsblad. While caution is called for, given the small number of tokens, the difference is striking: three cases found in the Belgian newspaper versus 16 in the Netherlandic newspaper. In terms of normalized frequency (per 1 million words), this is a ratio of 1.15 to 4.74, which is a considerable difference. This is not surprising, since in Netherlandic Dutch, lopen has become the default verb for normal human self-propelled motion (walking), whereas in Belgian Dutch, it saliently refers to rapid self-propelled motion (running), gaan ‘go’ being the neutral verb for walking.27 In other words, the higher degree of semantic bleaching of lopen in Netherlandic Dutch may have triggered a higher degree of grammaticalization of the lopen+te+V construction, whereas this is less so for Belgian Dutch where the verb is still strongly associated with rapid bipedal motion. Though telling, these findings need to be corroborated with more extensive data sets. 5. Conclusion. Many uses of the pos-progressive in Dutch should not be attributed to the auxiliation process, since they contain CPVs as default location verbs that are used obligatorily when referring to an entity’s location. The corpus-based analysis of the progressive constructions has confirmed the patterns for non-aspectual uses as discussed in, for instance, Van Oosten 1986 and Lemmens 2002. A further analysis of complement verbs that occur in the pos-progressive revealed that their selection follows clear patterns, many of which can be accounted for in terms of our everyday 27
This difference is bound to give rise to humorous misunderstandings, as I have experienced myself. Some years ago, a Dutch colleague of mine used the phrase lopen naar het station ‘lopen to the station’ to which I replied—at that time insufficiently alerted to the regional difference—that there was still ample time to catch the train and no need to run, a reply that completely puzzled my interlocutor since, to his knowledge, he had never implied that we should run.
Posture Verb Constructions
215
experience. In line with what has been observed for other, unrelated languages, the Dutch data show that zitten is the verb that has gone furthest in its semantic bleaching, and its grammaticalization is probably still continuing. Finally, other elements in the construction (or in the wider context) contribute to the stative, locational character of the posprogressive, setting it apart from the prep-progressive on the one hand, and from the lopen+te+V construction on the other. This indicates that even in contexts where the different constructions seem interchangeable, they are not semantically identical, as they impose their own construal on the event. Due to limited corpus material, the results of the present study essentially reflect what is true for written language, and they may be slightly different for the spoken register, as is suggested by the limited exploration of (informal) data obtained from the Internet and children’s stories. However, even in these samples, the general patterns seem to be confirmed. A more careful contrastive analysis of such register differences as well as regional differences is clearly worth pursuing to arrive at an even better understanding of the use of progressive posture verb constructions. REFERENCES Berman, Ruth A., and Dan I. Slobin (eds.). 1994. Relating events in narrative: A crosslinguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum Associates. Bertinetto, Pier Marco, Karen H. Ebert, and Casper de Groot. 2000. The progressive in Europe. Dahl 2000, 517–558. Boogaart, Ronny. 1991. Progressive aspect in Dutch. Linguistics in the Netherlands, ed. by Frank Drijkoningen and Ans van Kemenade, 1–9. Amsterdam: John Benjamins. Booij, Geert. 2002. Constructional idioms, morphology, and the Dutch lexicon. Journal of Germanic Linguistics 14.301–329. Brisau, André. 1969. English progressive tenses and their Dutch equivalents. Studia Germanica Gandensia 11.73–85. Bybee, Joan L., Revere Perkins, and William Pagliuca. 1994. The evolution of grammar: Tense, aspect and modality. Chicago: University of Chicago Press. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Dahl, Östen (ed.). 2000. Tense and aspect in the languages of Europe. Berlin: Mouton de Gruyter.
216 Lemmens Ebert, Karen H. 2000. Progressive markers in Germanic languages. Dahl 2000, 605–653. Haeseryn, Walter, Kirsten Romijn, Guido Geerts, Jaap de Rooij, and Maarten C. van den Toorn. 1997. Algemene Nederlandse Spraakkunst. Tweede, geheel herziene druk. Groningen and Deurne: Martinus Nijhoffuitgevers/Wolters Plantyn. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer. 1991. Grammaticalization: A conceptual framework. Chicago: University of Chicago Press. Kuteva, Tania. 1999. On ‘sit’/‘stand’/‘lie’ auxiliation. Linguistics 37.191–213. Lemmens, Maarten. 2002. The semantic network of Dutch posture verbs. Newman 2002, 103–139. Lemmens, Maarten. 2005. Motion and location: Toward a cognitive typology. Parcours, détour, contour (Travaux du CIEREC, 117), ed. by Geneviève Girard, 223–242. Saint-Etienne: Publications de l’Université St. Etienne. Leys, Odilon. 1985. De konstruktie staan te + infinitief en verwante konstrukties. Verslagen en Mededelingen van de Koninklijke Akademie voor Nederlandse Taal- en Letterkunde 3.265–277. Lødrup, Helge. 2002. Norwegian pseudocoordinations. Studia Linguistica 56.121–143. Newman, John (ed.). 2002. The linguistics of sitting, standing, and lying. Amsterdam: John Benjamins. Newman, John, and Sally Rice. 2004. English sit, stand, and lie: Patterns of usage and their experiential motivations. Cognitive Linguistics 15.351–396. Serra Borneto, Carlo. 1996. Liegen and stehen in German: A study in horizontality and verticality. Cognitive linguistics in the Redwoods, ed. by Eugene Casad, 458–505. Berlin: Mouton de Gruyter. Song, Jae Jung. 2002. The posture verbs in Korean. Newman 2002, 359–385. Talmy, Leonard. 2000. Toward a cognitive semantics (2 vols.). Cambridge, MA: MIT Press. Van den Hoek, Theo. 1971. Leggen en zetten. Tabu 1.33–37. Van den Toorn, Maarten C. 1972. Over de semantische kenmerken van staan, liggen en zitten. De Nieuwe Taalgids 6.459–464. Van Oosten, Jeanne. 1986. Sitting, standing and lying in Dutch: A cognitive approach to the distribution of the verbs zitten, staan, and liggen. Dutch linguistics at Berkeley, ed. by Jeanne van Oosten and John Snapper, 137–160. Berkeley: University of California Press. Van Pottelberghe, Jeroen. 2002. Nederlandse progressiefconstructies met werkwoorden van lichaamshouding. Nederlandse Taalkunde 7.142–174. Vendler, Zeno. 1967. Linguistics in philosophy. Ithaca, NY: Cornell University Press.
Posture Verb Constructions
217
Verhoeven, Ludo, and Sven Strömqvist (eds). 2004. Relating events in narratives: Typological and contextual perspectives. Mahwah, NJ: Lawrence Erlbaum Associates.
Université Lille3/Centre National de la Recherche Scientifique, France B.P. 60149 59653 Villeneuve d’Ascq Cedex France [[email protected]]
Journal of Germanic Linguistics 17.3 (2005):219–224
REVIEWS
Studies in the History of the English Language II: Unfolding Conversations. Edited by Anne Curzan and Kimberly Emmons. (Topics in English Linguistics, 45.) Berlin and New York: Mouton de Gruyter, 2004. Pp. xii, 500. Hardcover. 94.00. Reviewed by SEIICHI SUZUKI, Kansai Gaidai University This volume is a selection of papers presented at the second Studies in the History of the English Language conference (SHEL 2) held at the University of Washington in 2000. According to the editors, the design and organization of the book were inspired by ongoing productive conversations on wide ranging topics in English historical linguistics. This explains the rather opaque subtitle when taken in isolation, and the persistent appearance of the term conversation in the editors’ introductory remarks. The book consists of the following four sections: Linguistics and Philology, Corpus- and Text-based Studies, Constraintbased Studies, and Dialectology. Beginning with a useful introduction by the editors, each section contains a key article, a detailed commentary on it by an invited respondent, a brief rejoinder by the keynote author, and two or three independent papers. In the key article of section 1, “Philology, linguistics, and the history of [hw]~[w]” (pp. 7–46), Donka Minkova first raises objections on methodological and epistemological grounds to the strict separation of philology and linguistics, calling instead for the need to integrate the two into a single discipline of linguistic philology. As an example of practicing linguistic philology, the author explores the historical development of /hw-/ in English. On the basis of ample evidence consisting largely of alliteration in Old English and Middle English verse and scribal practice in Early Middle English manuscripts, Minkova shows that the merger of /hw-/ with /w-/ was well established in some varieties of Early English in the South. In the sixteenth and seventeenth centuries, however, the contrast was reintroduced among educated southerners through dialect contact. This approach contradicts the prevailing view, according to which the merger was an eighteenth © Society for Germanic Linguistics
220
Reviews
century development. Further, Minkova argues that the subsequent “unmerger” of /hw-/ and /w-/ in Early Modern English would have been motivated by spelling, word frequency, and the prestige of some northern varieties in which the contrast was maintained throughout. Minkova’s proposal is then made more realistic by Lesley Milroy in “An essay in historical sociolinguistics? On Donka Minkova’s ‘Philology, linguistics, and the history of [hw]~[w]’” (pp. 47–53). Milroy adds a sociolinguistic variationist perspective to the picture. She emphasizes the presumable variability and social dynamics underlying the merger by presenting well-documented parallel developments. Arguing on theoretical and cross linguistic grounds that phonemic short diphthongs are not known to exist, David L. White in “Why we should not believe in short diphthongs” (pp. 57–84) reiterates Daunt’s interpretation that the so-called short diphthongs in Old English were devised for representing phonetic velarization by the hand of Irish missionaries on analogy with their native language. In “Extended forms (Streckformen) in English” (pp. 85–110), Anatoly Liberman calls attention to the role of syllable insertion in word formation in the history of English, and offers a detailed case study of ragamuffin (extended with -a-) and hobbledehoy (extended with -de-/-te-). In “Linguistic change in words one owns: How trademarks become ‘generic’” (pp. 111–123), Ronald R. Butters and Jennifer Westerhaus explore some sociolinguistic aspects of the semantic change of trademarks developing into generics (GENERICIDE), notably the public’s understanding of, and the court’s decision on, genericness. Leading section 2 is Susan M. Fitzmaurice’s “The meanings and uses of the progressive construction in an early eighteenth century English network” (pp. 131–173). The author investigates the frequency and distribution of progressive constructions overall, as well as their experiential/subjective use to “express a speaker’s evaluation of or attitude toward what is being talked about” (p. 133), in contrast to their aspectual or truth-conditional function. Her analysis is based on the Network of Eighteenth-Century English Texts, a corpus of writings containing four prose registers (letters, essays, fiction, and drama) produced between 1653 and 1762 by a group of seventeen individuals affiliated in varying degrees with Joseph Addison, a central figure of the social network in question. Largely attentive to a spectrum of parameters involved in the corpus, Fitzmaurice brings to light a multi-dimensional
Journal of Germanic Linguistics 17.3 (2005)
221
picture of varied use and function of the progressive. The construction is discussed from the perspective of register, generation, the varying density of network membership, and a set of lexical and grammatical contexts such as tense and clause type. However, as Erik Smitterberg remarks in “Investigating the expressive progressive: On Susan M. Fitzmaurice’s ‘The meanings and uses of the progressive construction in an early eighteenth-century English network’” (pp. 175–182), the role of gender seems significant and needs to be fully integrated into the analysis. Furthermore, the occurrence of experiential progressives requires a fuller quantitative treatment on the basis of the close readings that Fitzmaurice’s analysis exemplifies. Based on the data from A Representative Corpus of Historical English Registers (ARCHER), Douglas Biber, in “Modal use across registers and time” (pp. 189–216), makes a cross-register comparison of the use of modals and semi-modals through the period of 1650–1990. In “The need for good texts: The case of Henry Machyn’s Day Book, 1550–1563” (pp. 217–228), Richard W. Bailey, by documenting a tradition of (mis)treatments of the work by language historians of the past and present, pleads for the necessity of producing reliable editions for the history of English. In “The perils of firsts: Dating Rawlinson MS Poet. 108 and tracing the development of monolingual English lexicons” (pp. 229–272), Ian Lancashire, focusing on the part of MS Rawlinson Poet. 108 (fols. 45r–81v) regarded by some as the first monolingual English dictionary, examines its dating and lexicographical problems against the background of emerging monolingual English lexicons in the early Modern English period, and presents a transcription of the manuscript fragment at issue. Geoffrey Russom begins the key article of section 3, “The evolution of Middle English alliterative meter” (pp. 279–304), by introducing his four rules of the b-verse in Middle English alliterative meter. Rule 1 (a b-verse must contain a long dip) and Rule 2 (a b-verse must not contain more than one long dip) feature prominently in his analysis of the metrical developments from Old to Middle English. Having identified comparable phenomena inherent in Beowulf, Russom then demonstrates that later works in Old English and in Old Saxon show increasingly stronger tendencies to realize a single long dip in the b-verse. He further claims that Middle English poets would have been led to reanalyze these inherited features as categorical rules. Russom’s account thus entails an
222
Reviews
unbroken continuity in alliterative tradition from Old English through the Middle English period, a hypothesis diametrically opposed to the prevailing view that the Middle English alliterative revival constituted a development out of Ælfric’s rhythmical alliterative prose, independent of the Old English alliterative verse tradition that had long died out. As Russom persuasively shows, however, Ælfric’s work contains many violations of Rules 1 and 2 above, making the derivation of Middle English alliterative verse from Ælfric’s style implausible on structural grounds. Continuing the conversation about the origin of Middle English alliterative verse in “Old English poetry and the alliterative revival: On Geoffrey Russom’s ‘The evolution of Middle English alliterative meter’” (pp. 305–312), Robert D. Fulk points out conceptual and historical difficulties with the dominant view, raising a fundamental question of the distinction between verse and prose in medieval times in general, and in Ælfric’s mind in particular. In “A central metrical prototype for English iambic tetrameter verse: Evidence from Chaucer’s octosyllabic lines” (pp. 315–341), Xingzhong Li proposes a prototype of iambic tetrameter verse that determines gradient metricality according to the four ranked gradient metrical saliency principles at work in the metrical positions, the feet, the hemistichs, and the metrical boundaries. He then adduces statistical evidence from Chaucer’s octosyllabic lines in support of his model. In “Early English clause structure change in a stochastic optimality theory setting” (pp. 343–369), Brady Z. Clark analyzes a change in the preferred order of a finite semi-auxiliary verb and a nonfinite verb from Late Old English to Early Middle English. The analysis is launched in a stochastic Optimality Theory framework, in which constraints are associated with a range of variable values rather than fixed single points. In “The role of perceptual contrast in Verner’s Law” (pp. 371–408), Olga Petrova presents a perception-based Optimality Theoretic account of Verner’s Law in Proto-Germanic (and also in Early Modern English, albeit less convincingly argued) as a contrast preservation phenomenon. Under her approach, the original pitch contour contrast was reanalyzed as the voice contrast in fricatives as a consequence of accent shift. Michael Montgomery and Connie Eble’s key article of the last section, “Historical perspectives on the pen/pin merger in Southern American English” (pp. 415–434), explores the pre-Civil War
Journal of Germanic Linguistics 17.3 (2005)
223
development of the merger of // and /i/ before nasals in the American South on the basis of manuscript evidence. The authors identified two kinds of occasional spellings (phonetic and inverse) in their new three corpora of colloquial documents. Comprising for the most part letters written by Southerners and dating from three roughly consecutive periods between 1761 and 1868, these three corpora contain data from different social and ethnic backgrounds: the educated, landed class of North Carolina; the semi-literate, working class white men (plantation overseers); and African Americans. By sorting out relevant data into three phonological groups (stressed and prenasal, unstressed and prenasal, stressed and non-prenasal) and comparing them across the three corpora in question, Montgomery and Eble conclude that the variation of // and /i/ was already underway in their earliest, late eighteenth-century corpus from North Carolina in all the three phonological contexts. Furthermore, the variation was apparently limited to prenasal environments in the later two corpora of the early nineteenth-century. As Guy Bailey points out in his response, “Digging up the roots of Southern American English: On Michael Montgomery and Connie Eble’s ‘Historical perspectives on the pen/pin merger in Southern American English’” (pp. 435–444), the authors’ data should be better analyzed quantitatively by calculating ratios of occasional spellings to their conventional counterparts. The debate between Montgomery/Eble and Bailey further highlights the fundamental issue of the upsurge of the prenasal (as opposed to the non-prenasal) merger in the late nineteenth and early twentieth centuries. Their findings suggest that the phenomenon of dialect contact requires closer consideration within a specific socio-historical setting (for example, urbanization and migration). In “Vowel merger in west central Indiana: A naughty, knotty project” (pp. 447–457), Betty S. Phillips reports on major findings of her and her students’ class project on the vowel merger in naughty and knotty and similar pairs in Terre Haute, Indiana. In “The spread of negative contraction in early English” (pp. 459–482), Richard M. Hogg carefully argues on the basis of safely localizable Old and Middle English texts that negative contraction—originally a random variation throughout the country—became categorical in the southwest, spreading northwestwards over time, but lost in the east and northeast. Finally, the editors are applauded for bringing out this valuable collection of stimulating, and sometimes even provoking scholarly
224
Reviews
conversations in English historical linguistics, which readers may wish to continue (or discontinue) in their own different ways.
Kansai Gaidai University 16–1 Nakamiya-higashino Hirakata 573-1001 Japan [[email protected]]