Introduction - PDF document

Germanic W eak Preterite Paul Kiparsky University 1 Introduction The weak verbs remains one of the most troublesome chapters of Germanic historical-comparative grammar . The morphological provenience of formative -d- has been debated for nearly two centuries, and there still consensus on whether is a reex of one or more the Indo-European dental sufxes, a grammaticalized form of the light verb d ¯ o or some mix these. The category' s phonological development within early Germanic presents a whole series of other mysteries. Why the effect of syllable weight on umlaut preterite stems differ North W est Germanic, and for that matter why umlaut be sensitive weight at Why the dental undergo two distinct “phases” of umlaut North Germanic, and why this category alone undergo a special of syncope W Germanic? Lahiri 2000 rst the morphological puzzle of the weak preterite' s etymology with some of the main puzzles of divergent evolution within Germanic. She that the form' s periphrastic origin can actually explain peculiarities. The of this paper to pose a different make link, which the same morphological hypothesis, but a much simpler account of the phonological development within Germanic languages. The that the dental from the past tense of the light verb d ¯ `do' goes back at least Bopp 1816. Though not uncontroversial, is the most widely etymology of the dental (Streit- berg 1896, Sverdrup 1929, von Friesen 1925, T ops 1974, Bammesberger 1986). From a morphological point of view , the assumption attractive. The templatic morphology by which verbs formed their past tenses, inherited from the Indo-European perfect, to monosyllabic roots. Longer verbs Germanic would accordingly have formed their perfects with 1 , just as they do Sanskrit (where disyllabic stems like `think' form perfect with or “do”, e.g. cint-ay- ¯ am ¯ as-a (or cint-ay- ¯ am . ca-k ¯ ar-a The Germanic periphrastic forms could have been later grammaticalized into inected forms, exactly the Sanskrit periphrastic forms were grammaticalized Middle Indic. (1) shows this familar grammaticalization trajectory , on the assumption that the reduction from second members compounds sufxes goes through a (Lahiri (The symbols ! and  stand for Prosodic W ord and Prosodic Stem, respectively a. Compounding: [ [ V erb + Sufx ] ! + [ Light V erb + Inection ] ! ] ! Cliticization: [ [ V erb + Sufx ] ! + Clitic + Inection ] ! c. Sufxation: [ [ V erb + Sufx ]  + T + Inection ] ! uncontroversial instance of this trajectory grammaticalization of Latin ¯ are habe ¯ o French chanterai will Lahiri 2000 presents a parallel development Bengali with interesting additional twist. In language, the auxiliary a  c h has been recruited the endings of both the progres- sive and the perfect; former the grammaticalization has gone completion and the erstwhile now just a sufx, while in it only reached the The Germanic dental be assumed have followed a similar from the original light verb d ¯ sufx -d-. (2) Latin [ [ ¯ a + ] ! + [ + ¯ o ] ! ] ! Old Bengali [ [ bO s + i ] + [ a  c + o ] ] Germanic [ [ tal + i ] + [ d ¯ ed + un ] ] F rench: [ chant + er + ai ] ! Modern Bengali [ bO s + +  c + o ] ! Old High German [ zel + i + t + un ] Lahiri 2000 posits that the new dental sufx goes through two stages, one where is a extension, on a par with derivational sufxes, and another where is as an inectional ending. (4) a. [ ( V erb + Sufx + d ) Stem + Inection ] ! [ ( V + Sufx ) Stem + d + Inection ] ! These different morphological structures, she proposes, are associated with ferent phonological effects, along the lines of level-ordered phonology . This a possibility , though one might wonder what drive a reanalysis which ending migrates “downwards” from the stem level, opposite direction from what observed. The between the morphology and phonology of the weak I will explore here more and more The idea simple: umlaut and vowel took place before the light verb morphologized a sufx W Germanic, and after this morphologization North Germanic. Unlike previous accounts, this achieves and maximally simple formulations of umlaut and vowel both of which moreover now be seen invariant throughout Germanic. particular, umlaut applies and heavy syllables alike, and apocope and syncope apply in branches, subject respective syllable and foot structure of each. “phases” or “stages” need be posited for of either process. 2 Let take a closer look at the phonology . a the problem W est and North Germanic seemingly differ conditions under which umlaut and syncope apply — a difference manifested only dental preterites. Whereas W est Germanic umlauts light stems and syncopates heavy stems (see (5a)), North Germanic umlauts heavy stems and syncopates both light and heavy stems (see (5b)). (5) Germanic Old High German Old a. Light stems: **talida da `counted' Heavy stems: *d ¯ omi t ¯ uomta d ¨ o ¯ mda `judged' Old High German, the distribution of medial in preterites governed the weight of the preceding root syllable. The vowel without exception lost after a syllable that heavy of a long root vowel, a consonant cluster, as or a geminate consonant (which then shortens again before the dental sufx), as After roots (whose nal are geminated before the stem-forming sufx ), -i- is normally retained, see (6d). (6) a. tuomta `judged', h ¯ orta `heard' dampfta `steamed' (from dempfen starcta `strengthened' (from sterken dursta `thirsted' (from dursten wanta `turn' (from wenten stalta `put' (from stellen branta `burned' (from brennen `kissed' (from kussen) d. nerita `saved' (from nerren *ner-ja-n `crushed' (from knussen *knus-ja-n) The distribution seems that medial vowels were after heavy syllables Old High German. That what always been as- sumed. Y et this putative syncope process would be extremely problematic, for not apply elsewhere early Old High German, in fact massively contradicted by the data. Outside of weak past tense forms, Old High German regularly retains medial vowels, even heavy syllables. 1 (7) Synkope von urspr ¨ unglichen Mittelvokalen, die ¨ ubrigen Sprachen nach langer Stammsilbe verbreitet ist . . . im Ahd. konsequent nur bei dem i Praet. (Part. Praet.) langsilbigen . V . I z.B. nerita, ginerit¯ er, aber h ¯ orta, gih ¯ ort¯ er
— Sonstige urspr ¨ ungliche Mittelvokale werden Ahd. (abgesehen von {65 3 genannten 2 ) bewahrt; z.B. Part. Praet. -an gibuntan¯ , eigan — eiganemu; offanes . V . offan ¯ ; Adj. -ag manag, heilag Gen. manages, heilages Komparative lengiro lang (Braune/Mitzka 1961:63). The origin of the weak preterites implies that the medial of heavy-stem preterites need not have been lost by medial rather, they could well have been lost by word-nal apocope, at a stage when light verb not yet become a sufx, when stem before independent phonolog- ical word. That assumption immediately resolves the phonology , for is a fact that W est Germanic apocope took place precisely after heavy syllables. The distribu- tion perspicuous Old English and Old Saxon (high vowel 1 See Baesecke 1918:66, 225 ff. for the details. 2 This a words whose original vowels have analogized words original vowels. E.g. meistar  meist(e)ra — originally disyllabic fact a loanword) which follows the pattern originally monosyllabic as  ngra 3 vs. word a much case of considerable analogical reshufing, 3 Old High German reects the same distribution: (8) Old High German stems (original distribution) stems Light wini quiti `saying', turi `door' ( Heavy `guest', anst `favor', `need' ( stems Light `custom', fridu `peace', hu (-u Heavy `hand' the stems) ( attributing the deletion of of the medial vowel weak past tense forms normal word-nal apocope prior grammaticalization eliminate the anomalous W est Germanic syncope process outright — obviously a good result because simply not work outside the weak verb as shown by forms like wirsiro blintemu `blind' (dat.sg.) or those cited Another benet explains the lack of umlaut forms with a medial vowel, as tuomta versus know on independent grounds that word-nal vowels were before they could trigger umlaut — that why stems like `guest' and anst `favor' have no umlaut (contrast meri `sea' from mari Note that the proposed solution reduces the burden the morphology irreducible minimum. All we the phonology right the weak verb a prosodic word W Germanic when took place. not have been a separate morphological word at that time, indeed at any time. The of the morphological categories are also not important for our purposes — the phonology works regardless of whether stem innitive, a verbal or something else, 4 and whether what followed it a word verb an auxiliary , stage (1b)), or already degraded a clitic, a transitional stage between word and sufx (stage (1b)). The even with idea that the dental sufx a conation of the light verb d ¯ with or more the Indo-European dental sufxes (such as the perfect participle -to- long as character from the former . Nor matter whether tense sufx, once reanalyzed, became a class marker an inectional ending (or a stem-level word-level ending). Sorting out all these options engrossing task for future research, but phonological evidence discussed here does not contribute it . same token, the hypothesis that the weak verb had the status of a separate prosodic word at the point at which W Germanic apocope applied stands on own being robustly compatible with a wide of morphologization scenarios. 2 the phonological divergence Now us examine more how the conditioning of apocope and syncope can be understood on these assumptions. With Optimality Theory , we that phonological processes are limited ranked violable constraints dened on 3 , light adopted the heavy heavy joined the -i 4 Sanskrit, is a nominalized verb form derived by afxing - ¯ a the present . E.g. `think' forms the periphrastic cint-ay- ¯ ¯ as-a (or ¯ . ¯ ar-a 4 representations. particular, vowel governed constraints on the prosodic form of words. The most important families such constraints, Germanic and elsewhere, are those on foot structure and on syllable structure. F eet have both a lower and an upper bound on their size. Germanic languages under discussion modern English — the basic metrical unit a moraic that is, a bimoraic unit consisting of two short syllables or a long syllable. uncommon moraic systems, Light–Heavy sequences could be parsed into feet as a last resort avoid metrically homeless syllables (the phenomenon called resolution 5 proviso, formally INALITY word-nal consonants are weightless (“extrametrical”). If must have at least two moras OOT words must at least a foot, then words must minimally bimoraic; this minimum length requirement excludes  and, insofar as nal are weightless, also  Such a minimality requirement will word-nal V - deletion (apocope) after a short syllable, viz.  !  ! . Finnish example just this prosodic constellation. Monosyllabic words may of the form C ¯ but C  are excluded, for they would be monomoraic because nal is weightless. certain registers, Finnish nal -i e.g. olisi ! olis, nousi ! nous, veisi ! veis; this option disallowed precisely C  pesi , kosi *kos The same constraining effect of word minimality apocope Germanic, as illustrated by the Old High German -stems V owel subject on syllables. These typically in- volve upper bound on the complexity of the syllable rhyme, or on the number of moras syllable. A common of this type prohibition of superheavy (3-mora) syllables. (9) *: A rhyme maximally . a with constraint, medial V (syncope) might fail after a long syllable: .CV *CVCC.CV , C ¯ .CV ¯ . An example Cairo Arabic, where such as *kalbna, *baabna are not pos- sible for this reason, and the syncope process seen like c i.di.la ! c id.la `straight' (f.) is blocked yik.ti.bu *yikt.bu `they write'. Germanic, word- nal are weightless, so words of the type baab, kalb are admissible, but, the same token, *bab, *kal are not possible words. Now us apply this idea Germanic. W Germanic, by hypothesis, in tense forms of -jan verbs at a time when stem of weak preterites a word its own hence in forms by apocope. Apocope would have had in -jan verbs heavy stems, but stems, because of OOT Since umlaut-triggering vowel heavy stems, the later umlaut process could take effect only stems. Moreover, need assume an medial process, which be otherwise unattested language. This for all the W est Germanic data discussed. 5 arguments that the languages have trochees as their basic foot type, Kiparsky and for other views, see Lahiri, Riad, 5 Light stems Heavy stems ! ! ¯ ! ta] ! apocope: ! ! ¯ ! ta] ! reanalysis: ! [h ¯ ! stems, the umlaut-triggering is because : *zalita ! heavy stems, apocope takes effect, bleeding umlaut: h ¯ orta `heard' ( ¨ o ¯ rta Polysyllabic stems are correctly predicted with heavy monosyl- labic stems undergoing vowel and no umlaut, for OOT not block apocope them, e.g. [[mahal+i] ! ! ! [[mahal] ! ta] ! `magnied'. The process just discussed, which in W Germanic, did not extend North Germanic, as contrasts such as Old gestir -gastiR) versus demonstrate. 4th-7th century runic forms like aliR, horna, dagar , preserve nal vowels . W e can safely conclude that branch of Germanic word-nal vowels were at the point when main verb with following light verb a single word (at stage (1c)). This fully by runic weak past tense forms from the same period, such as satido, tawi fahido, which medial vowel, conrming that syncope category took place much later, perhaps not until the 7th-8th centuries — well the weak past tense grammaticalized. North Germanic weak past tenses lost their medial vowel when weak preterites were single words, their deletion must have been governed the constraints on medial — not by the constraints on apocope, as W Germanic. deletion governed nal but has of course always been assumed) by medial which constrained by OOT but syllable structure. What the relevant North Germanic syllable structure constraints? The Runic evidence shows that North Germanic, unlike the Old of later written texts, three-mora prohibition (9) *. attested Runic form version R (early c.) iarl ¯ ¯ salas (ca. 400) ¯ Asg ¯ (compound wan ¯ adas (6th c.) V andr ¯ aps (compound This that there a syllable structure of North Germanic between the runic period and later Old dating perhaps 7th century: (12) a. Early North Germanic (Runic): the limited maximally moras (CVC, C ¯ V), excepting monosyllabic words as usual. Later (Old C ¯ VC-syllables became admissible, and arose through syncope, apocope, and processes. Independent evidence for the more restrictive structure of North Germanic comparend W Germanic North Germanic evaded 3-mora syllables both `actively', by deletion of in overlong and by shortening, and `passively', by failing these respects with W Germanic, which 3-mora syllables (Kiparsky 1998): 6 a. North Germanic: d.joo] ! .do] (ON pa stems kept j, e.g. nip. W Germanic: no deletion, e.g. [hirt.joo]. High of the * also blocks Coda gemination North Ger- manic, isogloss between the two groups: (14) a. North Germanic *[tel.jan] (ON telja). W Germanic *[tal.jan] ! *[tell.jan] (OHG ) Another consequence of high-ranked * Apocope, as can be seen in -i stems (here assume that is as an extrasyllabic mora): Apocope stems: di(-R)] ! d(-R)] ( d(-R)]) heavy stems: [gas.ti(-R)] ! [ges.ti(-R)] ( *[gast(-R)]) From infer that North Germanic, medial would have been originally to stems by *. Therefore, syncope would have removed trigger for umlaut stems, leaving scope for to only heavy stems. Hence we have originally had *d ¯ omi ! *d ¨ o ¯ da ¨ o ¯ mda but da ! talda (and not da Light Heavy Structure at sufx stage: [CVC+i+ ! [CVVC+i+ ! Syncope stems: [ta.li. ! ! d ! ¨ o ¯ .mi. W e cannot be sure that this a stable stage, but it is the type that . other hand, readily be modeled Stratal Lexical Phonology). A stem-level straint system with ranking OOT   * provides the input a word-level constraint system, where * . Our that medial applied rst stems and then heavy stems reverses the traditionally assumed chronology . The for the hand- books' assumption of the opposite chronology analogy bles, which did apply rst heavy stems, e.g. umu[n]t, sunu (Helnæs, ca. 800), which be mund, sun ([gupu] ! [mundu] ! , [sunu] ! In view of the considerations presented above, inference invalid. If the mini- mum constraint OOT the maximum constraint * were dominant early Old Norse, and nal weightless INALITY nal apocope would have taken place stems, and medial syncope would have taken place stems. This be a system like one of Cairo Arabic, mentioned above. And ts exactly the syncope data and the distribution of umlaut seen last period of apocope, seems that the effects of become morphologized some extent. stems dominates OOT Light stems: kyn ber `berry', fen `fen' Heavy stems: r ¯ ke `kingdom', ku ¯ æpe `saying', d ¨ o ¯ `judgment' other hand, 3-mora syllables eventually became admissible, so that syncope extended heavy stems: (18) *d ¨ o ¯ da ! d ¨ o ¯ mda, gesti ! gest 7 Analogical developments such as talda � telda and N.Pl. d � d indi- cate morphologization of Umlaut. Apparently became a stem-level and ceased triggered by word level (inectional) sufxes. (Cf. Lahiri 2000, and the parallel later development of -Umlaut, see Kiparsky 1984.) (19) c. Apocope heavy stems light stems, later heavy stems Light stems wini (� vin+r) (Ab)-guti gup Heavy stems gast *gestir � gest+r (Umlaut!) pyrft � purft (Umlaut!) Light stems sunu (� sun) son (� son+r) fridu *frid (� frid) Heavy stems hand ho ¸ nd Light stems beri ber Heavy stems tuom (� stem) d ¯ o ¨ (Umlaut!) Light -wa stems haro, ¸ r `ax' Heavy -wa stems drang (?) pro ¸ ng `throng' I conclude that the proposed perspective less fruitful on the North Ger- manic sweeps away unnatural restriction of umlaut heavy stems posited previous analyses. Umlaut now be assumed identically and generality branches of Germanic. And we bid good riddance any weight conditions on umlaut, which were quite — why a heavy syllable be more likely to assimilate to a following high front vowel? — well suspiciously redundant on account of repeating a condition which for syncope anyway , where make excellent phonetic sense, for the reasons just explained. any case, there of evidence that umlaut did apply also stems, such as the nouns p, lykell, fetell (versus R ¨ ok pR /fatilaVr/, where by syncope). These data then fall straightforwardly . A second gain for North Germanic the difference between the condition- ing of nal and medial derived on principled grounds from the fact that the output of the former (but not of the is minimality 3 Conclusion I summarize my as follows: Germanic has the same grammaticalization trajectory as Romance Mid- Indic, whereby light verbs/auxiliaries become sufxes viz. (1). Lahiri this morphological development as the phonol- ogy of the Germanic weak preterites. The weak preterites grammaticalization, hence by apocope, grammaticalization, hence by syncope. This explains the divergence of forms W Germanic and North Ger- manic. 8 The differences V processes between W est Germanic and North Germanic follow the respective syllable structures of these dialects. d. Umlaut not be to heavy stems North Germanic. The striction that heavy stems do not lose the umlaut-triggering vowel medial because of North Germanic' s resistance super- heavy syllables. e. There need an otherwise unattested medial process for Old High German. 9 Bibliography AMMESBERGER A. Der Aufbau des germanischen V erbsystems Heidel- berg: Winter . AESECKE EORG Einf ¨ Althochdeutsche M ¨ unchen: Beck. RAUNE W., ITZKA Althochdeutsche Grammatik T ¨ Niemeyer . VON RIESEN O. svaga preteritum i germanska spr ak. Skrifter utgivna av kungliga humanistiska vetenskapssamfundet 22(5). IPARSKY , AUL Lexical Phonology of Icelandic, C.-C. Elert, I. and Eva Stangert (eds.), Nordic Prosody . Stockholm: Almquist & Wiksell. IPARSKY , AUL Sievers' Law prosodic optimization. Jasanoff, Craig Lisi Oliver (edd.), M ´ Curad: Studies of Calvert W atkins. Innsbruck: Innsbrucker Beitr ¨ zur Sprachwissenschaft. AHIRI DITI Hierarchical restructuring creation of verbal mor- Bengali and Germanic: Evidence from phonology . Aditi Lahiri (ed.) Analogy , Levelling, Markedness.xdvi AHIRI DITI TOMAS IAD AIKE ACOBS Diachronic prosody . Harry van Hulst (ed.), W ord prosodic systems languages of Eu- rope 335-422. Berlin: Mouton de Gruyter . TREITBERG ILHELM Urgermanische Grammatik. Heidelberg: Win- . VERDRUP , Das germanische Dentalpr ¨ ateritum. OPS GUY A.J. The of the Germanic dental preterit. Leiden: Brill.