¹ Quellen und Forschungen zur Sprach- und Kulturgeschichte der germanischen Völker 32.

² Kluge posited it as ∗éyēt, which would have become PGmc. ∗ijē.

³ The best established example is the suffix of the weak preterit singular, PGmc. ∗-dōⁿ, ∗-dēs, ∗-dē, pointing to a root aorist ∗dhōm (with unexplained o-vocalism), ∗dhēs, ∗dhēt.

⁴ If, as seems likely, Toch. ∗ā cannot directly represent a Proto-Indo-European long vowel, and if, as Werner Winter suggested at the 1959 Texas Conference in Indo-European Linguistics, resonant plus laryngeal remained uncontracted in Tocharian, B iyā- and A yā might possibly be from PIE ∗iA, the otherwise unattested zero grade of ∗yā (= ∗yeA). Franklin Edgerton's view, Lg. 10.244–5, that Proto-Indo-European initial ∗y- and ∗iy-alternated in accordance with Sievers' Law is not supported by any data available to me.

⁵ Spelled pa-a-an-zi.

⁶ Cf. Ernout-Meillet³ 354, s.v. eō.

⁷ Hom. eísato belongs to híemai; cf. Frisk, GEW s.v. eísomai.

⁸ The precise origin of this stem, recurring in Umbr. iust ‘ierit’, is problematic, but I think it most probably a perfect ∗īy- with the same reduplication ∗i- and zero-grade root ∗-iy- as Skt. īyúr; cf. Leumann-Hofmann, Lateinische Grammatik 336–7, and R. G. Kent, The forms of Latin 120, and note the parallel formation of Fal. fifiked ‘finxit’ and OLat. scicidit (so quoted by Priscian, while Gellius seems to have found the form as sescidit or the like; cf. Neue-Wagener 3.344–5). Elsewhere I may discuss iī and its problems in more detail.

⁹ The prehistory of Lith. jo ‘went’ (beside Latv. gãja) is obscure to me.

¹⁰ For this term see Howard Garey, Lg. 33.106.

¹¹ This view of the basic aspects of ∗ey, ∗g^wem, and ∗g^wā I owe to Paul Tedesco.

¹² Aside from rare preterit presents like Go. wait ‘I know’, Lat. meminī ‘I remember’, OIr. ‘dúthraccar ‘I desire (d)'.

¹³ On the root allomorph -iy- see §16 below.

¹⁴ Or ∗-tha; Germanic provides no evidence for or against a ‘voiceless aspirate’ here. Cf. Ingerid Dal, NTS 16.330–1, and note Anglian earÐ, arÐ ‘thou art’, whose Ð can hardly be anything but the Proto-Germanic ∗þ that we expect from pre-Germanic ∗t after a resonant preceded by an accented vowel.

¹⁵ Whether such a-roots ablauted to o before the falling together of ∗a and ∗o in pre-Germanic is a question of no importance for the present investigation. I shall arbitrarily reconstruct perfects in ∗a, without Abtönung.

¹⁶ For the (ir-)uor which R. Kögel, PBB 16.502, incorrectly inferred from spellings of the synonymous ir-fuor in some late manuscripts see J. Schatz, IF Anz. 41.30 and Althochdeutsche Grammatik 291.

¹⁷ Unconvincing Flasdieck, op.cit. 334.

¹⁸ So Flasdieck, loc.cit.; otherwise Feist, PBB 32.508.

¹⁹ Not attested, but probable on the evidence of Skt. ániti ‘breathes’ and OIr. anaid ‘stays’, impv. an.

²⁰ We need not discuss the origin of this formation here.

²¹ To judge by Skt. adanti, Hitt. adanzi, Gk. édousi, Lat. edunt, ∗d had been replaced by ∗ed as zero grade in this verb already in the proto-language. Only the semantically isolated ∗dónt-, dṇt- ‘tooth’ kept the regular zero grade.

²² I would not postulate an original ∗Ee-Ed-, as Benveniste has done for Lat. ēdī, Archivum linguisticum 1.19. I doubt strongly that laryngeals still existed in pre-Germanic and pre-Italic when these dialects created perfects to ∗ed, and furthermore I can find no good evidence for an initial laryngeal in this root (cf. the discussion of ∗ey in §18). Indeed, the lack of an initial vowel in ‘tooth’ outside Greek and Armenian, contrasted, say, with Lat. apiō from ∗Ep-, seems evidence against one.

²³ The few occurrences of MHG ăz and the like show only that medieval Teutons found ēt as anomalous as we do.

²⁴ That loss of reduplication in pre-Germanic was later than loss of intervocalic ∗y is indicated by its failure to occur in ∗eóye (cf. §14); this implies that it was also later than the Germanic obstruent shift and (presumably) the falling together of ∗a and ∗o (cf. ftn. 28). For the moment I see no evidence showing whether it occurred before or after the loss of final ∗-a and ∗-e, but to avoid citing numerous alternative forms I will arbitrarily assume that it was later.

²⁵ The Proto-Indo-European lengthened-grade perfect or aorist which figures so prominently in many treatments of the Germanic preterit since Brugmann and Ljungstedt is a myth, invented by scholars who had grown weary trying to explain attested forms only from prehistoric forms for which there is good evidence. An original ∗e-ed- with e-grade in the perfect singular (Osthoff, Perf. 123–4) lacks support since Kurylowicz's brilliant and convincing explanation of the short a in Vedic 1st singular perfects of the type jagama, Symbolae … Rozwadowski 1.103 (cf. also F. B. J. Kuiper, India antiqua 200). A newly created ∗e-ed- (Flasdieck, Anglia 60.335) is against all analogy and not to be seriously considered unless all other explanations fail. An augmented imperfect ∗ēd- cognate with Skt. ādam ‘I ate’ would be open to the same morphologic objections as Kluge and Möller's equation of iddja with áyāt.

²⁶ This change can be described (not explained) as part of a general tendency in Germanic to replace reduplication by ablaut as the formal mark of the preterit wherever possible. Perfect singulars of long-vowel roots like ∗wādh ‘vadere’ with Germanic a-presents and not complicated by Verner's Law may have set it off, since they were presumably already of the type ∗wewōd and needed only to jettison their reduplication to fall in with their plurals ∗wōdun etc. Incidentally, the peculiarity of ∗ear ‘plowed’ may be that it generalized the singular ∗ear- instead of the plural ∗ōr-, but I still can suggest no reason for this special treatment.

²⁷ The type gōl, ōl must have existed before ∗gegab was dereduplicated; otherwise the failure of ∗eat to become ∗at would be incomprehensible.

²⁸ On this chronology cf. Lg. 35.9 fn. 17. It is also possible that a perfect to ∗éyti was not created until after ∗a and ∗o had fallen together, so that its oldest form would be ∗eáye (or even ∗éaye).

²⁹ I am not able to find any decisive material for the treatment of initial ∗yi-. Verb forms with ji- like OHG gihit ‘declares’ can easily be analogic to other forms of the paradigm like inf. jehan and pret. jah, and the more isolated OE gif ‘if’ and gicel ‘icicle’ may be contaminations of ∗ib-, ∗ikila- and the ∗jeb-, ∗jekula- of OFr. jef and OIc. jkull (cf. the slightly different gecilae, gecile of the early Old English glossaries). On the other hand, forms with i- like OHG ibu ‘if’, Go. iba(i) ‘lest’ and (if reliable) OHG ihilla ‘icicle’ (AHDGll. 2.726.36) may reflect zero-grade ablaut variants with i- from the outset. Analogy of ∗wu-, which clearly stays (cf. ‘wool’, ‘wolf’, ‘wound‘), favors ∗ji- as the regular development.

³⁰ If anyone should be willing to accept loss of ∗y before ∗e but not before ∗a, he can consider the PGmc. ∗eō with which I operate in the following pages to be 3d singular only.

³¹ In Lg. 35.9 I carelessly forgot this and derived the 2d singular imperative ending ∗-õ of the second weak conjugation from PIE ∗-āye through intermediate steps ∗-ōji and ∗-ōï instead of through ∗-ōje and ∗-ōe.

³² I cannot forbear saying, though, that I do not understand why Hamp seems not to have taken into consideration the possibility that the difference between nom. OE giefu, OIc. spk and acc. OE giefe, OIc. spaka reflects a change of nonnasal two-mora ∗-ō in absolute final to ∗-ū in North and West Germanic before the loss of nasality in ∗-ōⁿ.

³³ I here arbitrarily assume that pre-Germanic had already replaced the original r-ending of the 3d plural perfect reflected in Skt. vid-úr, Lat. ded-ēre, Toch. B tak-āre, etc. with the ∗-ṇt of the s-aorist.

³⁴ Tedesco suggests to me that the model was rather dissyllabic bases like bhī ‘fear’ (PIE ∗bheyH-), whose Indic perfect plural he thinks was originally ∗bhi-bhiy-úr, with the root syllabic because its resonant was at one time flanked by consonants (∗-bhiH-). But I personally am quite uncertain about the behavior in Aryan of root-final laryngeals in reduplicated formations. Germanic seems to have preserved no perfects to roots ending in ∗yH, so that they hardly come into consideration as likely models for ∗eiyt.

³⁵ Not /ii/, because /ji/ and /ij/ contrast in words like ∗īsarna- ‘iron’ and ∗jihid/þ ‘declares’ (whether the ∗ji- of the latter is regular or not; cf. fn. 29).

³⁶ Cf. J. Marchand, Lg. 32.285–7. Note also OIc. svilar ‘men whose wives are sisters of each other’, which, as W. Wissmann has pointed out, Münch. Stud. 6.129 fn. 28, cannot continue the PIE ∗swelo- given e.g. by Frisk, GEW 24, and Pokorny, IEW 1046. But svilar can reflect a PGmc. ∗swilijōz or ∗swilijaniz (the word is an n-stem in Icelandic; for the phonetics cf. hirÐar from ∗hirdijōz), developed from PIE ∗sweliyo- or ∗sweliyon- in defiance of ‘Sievers’ Law'; and ∗sweliyo-, ∗sweliyon- are precisely the forms needed to explain the completely synonymous Greek aélioi, aílioi, eilíones (on which cf. Frisk, loc.cit.).

³⁷ If we use laryngeals freely enough, e.g. by writing ∗sweliHo- instead of ∗sweliyo-.

³⁸ The present study can be considered as a start toward such a weeding out. Several cases where there is evidence for laryngeal and yet no Verschärfung are pointed out by Ladislav Zgusta in his short but trenchant critique of Lehmann's chapter on Verschärfung, Archiv orientální 23.198–201 (1955).

³⁹ In Eng. Stud. 3.158 he adds that the unattested 2d singular indicative can hardly have been anything but iddjes.

⁴⁰ I see nothing in favor of Fourquet's view, Studia neophilologica 14.423 (1941–2), that the first part of the paradigm to be so remade was the optative.

⁴¹ Joseph Wright, English dialect grammar 284, gives numerous preterits from modern English dialects formed by adding the weak suffix to the old strong preterit, e.g. droved, rosed, broked, gaved, felled, flewed, knewed.

⁴² It is interesting to speculate that the form displaced by OIr. luid ‘went’, do'coïd ‘has gone’ and W aeth, MBr. aez, Co. eth ‘went’ was probably a perfect of ∗ey similar in origin to iddja, ēode, and iī. Although I would not venture to try to say exactly what a Celtic perfect to ∗ey would have looked like, it seems safe to say that regular sound change would have reduced it to practically nothing in the attested Insular Celtic languages.