* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download as a PDF
French grammar wikipedia , lookup
Ojibwe grammar wikipedia , lookup
Modern Hebrew grammar wikipedia , lookup
Lexical semantics wikipedia , lookup
Polish grammar wikipedia , lookup
Scottish Gaelic grammar wikipedia , lookup
Ancient Greek grammar wikipedia , lookup
Modern Greek grammar wikipedia , lookup
Proto-Indo-European verbs wikipedia , lookup
Sanskrit grammar wikipedia , lookup
Georgian grammar wikipedia , lookup
Latin syntax wikipedia , lookup
Japanese grammar wikipedia , lookup
Portuguese grammar wikipedia , lookup
Ukrainian grammar wikipedia , lookup
Yiddish grammar wikipedia , lookup
Germanic weak verb wikipedia , lookup
Sotho verbs wikipedia , lookup
Serbo-Croatian grammar wikipedia , lookup
Swedish grammar wikipedia , lookup
Old Norse morphology wikipedia , lookup
Icelandic grammar wikipedia , lookup
Germanic strong verb wikipedia , lookup
Pipil grammar wikipedia , lookup
Russian grammar wikipedia , lookup
Latin conjugation wikipedia , lookup
Arabic verbs wikipedia , lookup
Finnish verb conjugation wikipedia , lookup
Kagoshima verb conjugations wikipedia , lookup
Old Irish grammar wikipedia , lookup
1 VOWEL ALTERNATION IN THE SPANISH -IR VERBS Hugh E. Wilkinson 1. Introduction. All students of Spanish have no doubt been bothered at some time or other by the problem of learning where to change an e into an ie or an i in verbs of the -ir conjugation. This vowel alternation is one of the striking features of this conjugation, and takes two forms: the alternation between e and ie or i, and similarly between o and ue or u, in the stem vowel (taking the root without the thematic vowel to be the stem) in the rhizotonic forms of the present indicative and subjunctive and the imperative, and the substitution of i for e and u for o in the arrhizotonic forms of the subjunctive and before a secondary yod. Another thing that stands out, too, is that in the majority of cases e alternates only with i, not with ie in the present, whereas this kind of alternation is never found in the other conjugations; and in the older language there was also mainly an alternation between o and u, which was later eliminated by the extension of u-forms throughout. Moreover, amongst the e/i verbs are many from the Latin -ĒRE and -ĔRE conjugations, which, for one reason or another, have found their way into this conjugation, so that it seems to have become a repository for all verbs of this type. (A certain number are learned borrowings which have been adapted to this type; equally, certain other learned words have been adapted to the e/ie/i type, while a few do not alternate at all.) The 1 2 result is that, in Modern Spanish, as far as the popularly derived verbs are concerned, the -er conjugation contains only verbs with stem vowel e/e < Latin Ē, Ĭ, such as vender, e/ie < Latin Ĕ, such as perder, o/o < Latin Ō, Ŏ, Ŭ such as romper, or o/ue < Latin such as morder, while the -ir conjugation has e/ie/i, as sentir, e/i, as pedir, i/i, as vivir, o/ue/u, as dormir, u/u for older o/u, as subir, and u/u, old and modern, as conducir, with e/e, o/o hardly found, and then only in learned borrowings. It is also noteworthy that, with the exception of hervir and the anomalous erguir, earlier erzer, -er verbs with the alternation e/ie did not migrate to the -ir conjugation, while even these two verbs have produced i-forms, irgo and popular hirvo. (In Old Spanish rendir/-er also fluctuated between rindo and riendo. The learned compounds of cerner, verter, such as concernir, convertir, follow the e/ie pattern of the simple verbs.) In the case of the back vowels, there are several verbs from the -ERE conjugations which had stem vowel Ŏ in Latin, such as nucir, cumplir, tullir, tundir, cuntir, muñir, and dialectal espurrir, escurrir and cullir, and the reasons behind these changes are as various as in the case of the front vowels. 2. What verbs are involved. In seeking the origin of these alternations, we should obviously look only at the verbs popularly or semi-popularly derived from Proto-Romance (PR), and not at learned borrowings. And in so doing, we need to look 2 3 at the same time at the developments in the neighbouring languages, Portuguese, Catalan and Provençal, as well as the other Spanish dialects. When we do so, we find that these languages have also developed special characteristics in the -ir conjugation, sometimes similar to the Spanish ones, though often the process has been a different one. (Source materials referred to by abbreviations in the text will be found listed in the Bibliography at the end.) First let us begin by listing the verbs concerned, by classes, taking the OSp. forms: (1) e/ie/i; Latin Ĕ, AE: ferir, mentir, sentir, (ar)repentirse, fervir, premir, tremir, -cernir, convertir, erguir, re-querir (pes-, con-), venir, aterirse, and compounds. (2) e/i, OSp. also e/ie/i; Latin Ĕ: vestir, embestir, servir, seguir, rendir, pedir, ex-pedir (des-), gemir, regir and compounds. (3) e/i or i/i; Latin Ē, Ĭ: medir (co-), re-çebir (con-, (a)per-, de- (F.J.)), desçir, fenchir, redemir, ceñir, costreñir, estreñir, teñir, feñir, reñir. (4) e/i or i/i; Latin Ī: bevir, escrevir, reir, freir, dezir. (5) e/i; etyma doubtful: arrecirse, derretir, engreirse, desleir. 3 4 (6) e/e/i; Latin Ĕ: (7) o/ue/u; Latin Ŏ: exir. dormir, morir and compounds. (8) u/u, or OSp. o/u, also o/ue/u; Latin Ŏ: cobrir, ofrir, sofrir (with the open o of ofrir) aborrir, nozir, complir, contir, escorrir; add MSp. tullir, tundir (1), surtir (1), muñir, curtir ?, mullir, cundir, pulir, dial. espurrir, cullir. (9) u/u, or OSp. o/u; Latin Ō, Ŭ: sa-codir (a-, re-, per-), sobir, foir, roir, destroir, escopir, dial. tusir, and cusir, bollir, sobollir, engollir, groñir, tundir (2), surtir (2), ordir, englotir, nodrir, podrir, surgir, unzir, dial. mocir, MSp. confundir. (10) u/u, OSp. also o/u; Latin Ū: aduzir, sumir, luzir, mugir, dial. pruir. (11) o/o, OSp. also o/o/u; Latin AU: oir. 3. Narrowing down the field. It will be seen that in the old language the front and back vowels had the same range of alternation. However, several notes need to be added to modify 4 5 the picture: (1) Fervir, premir, tremir, and erguir are flanked by old and dial. infinitives ferver, premer, tremer and erzer (coming from *ĔRGERE for classical ĒRIGERE; -er also in Portuguese, which has old erger, later erguer); premir, tremir, found as old variants of premer, tremer, are perhaps learned formations. Concernir, convertir etc., beside the simple verbs cerner, verter, also have a learned flavour, but -querir beside querer reflects a widespread popular development. Aterirse is of doubtful origin, but is also flanked by aterecerse (also OPtg.), so forms like ateresco probably antedate rhizotonic forms. In the case of repentir, there are, in addition, -er forms (dial. and Ptg.; cf. also p.ps. repeso/repiso/arrepentudo in L.B.A.); further, unlike Malkiel (Mal.; see Bibiliography), I would say that the forms with -t- go back to early PR syncopation, with subsequent attraction to the sentir class (cf. It. pentire), while those with -d- come from unsyncopated ones as preserved in Cat. penedir, Prov. peneder. Venir, being irregular, only partly belongs here, but exerted strong influence. (I ignore perir, as it was an interfix verb, with only arrhizotonic -ir forms; also enxerir, though it seems to have had some popular development in Spanish and Portuguese, because we do not know enough about its forms.) (2) Seguir comes from a widespread PR -IRE variant of *SEQUERE. Rendir also had an infinitive render (cf. Portuguese). Expedir and despedir are semi-learned words attracted to pedir. Gemir is also a learned form alongside gemer/emer (Ptg. gemer). 5 6 Regir too is learned, but has wide popular use in many Romance languages, and correger is found in Old Leonese (cf. Ptg. reger); the g may come from a late Latin geminate, cf. Cat., Prov. regir paralleling llegir, legir. (I ignore obedecer, empedernir/-ecer, denegrir/-ecer, aguerrir being basically interfix verbs.) (3) Desçir, re-çebir etc. and fenchir correspond to Ptg. -er infinitives descer, receber etc., encher, also OLe. receber (F.J., with recebe, quoted by Hanssen, and p.ps. recibudo, percebudo; cf. similar OAr. recebuda, G. H. 99, and apercebudo in L.B.A.), encher, Ar. impler/-ir. However, recebir may be older, with -ir from a PR *-IRE variant, based on the yod forms; note modern recibir, (a)percibir but concebir. (Pedir also goes back to PR *PETIRE.) The origin and development of desçir are problematical. Asturian/Leonese and Old Portuguese also have -ir or -er for the verbs of the -ñir class (which exhibit a specially Spanish change of conjugation, see §11 below). Redemir/remedir, modern redimir, has undergone learned influence. Medir and pedir have influenced each other. (Fender was not attracted to this class because it had ę from PR *FĔNDERE, cf. It. fęndere.) (4) For the change of conjugation in these verbs see §11 below. Bevir and escrevir now have i throughout. Dezir only partly belongs here, being irregular, but is very influential. -Er infinitives are common in the dialects for escrevir, bevir, reir and dezir, and it is not clear which is the earlier; however, Ptg. escrever seems to point to an earlier *escrevir from escrivir 6 7 (attested in Old Galician) with dissimilation (but cf. popular vever, dezer etc., Nunes, p. 275), and there is no sign of a Ptg. *rier beside riir/rir, though Portuguese is very prone to have -er for Sp. -ir; also Galician has vivir and dizir beside viver, dizer, so the -ir forms may be earlier, seeing that the trend is from -ir to -er. (Fenir has no place here, because it had become an interfix verb.) (5) Arrecirse virtually only exists in the form arrecido, and is, perhaps, originally an interfix verb like the cognate Rum. răci: cf. Gal. arrecerse (Cor. Dic.). Derretir corresponds to Ptg. derreter, so may be an old -er verb. Engreir seems to have started from a p.p. engreido, and an infinitive engreyecerse is also found (Cor. Dic.), suggesting that rhizotonic forms are late. These verbs, and desleir, because of their uncertain etyma, cannot be used as the basis of any argument, but are, perhaps, of secondary formation anyway. (6) Exir does not produce ye- forms in Castilian, because of the palatal, but ye- forms are found in the dialects. (7) We must exclude adormir as being an interfix verb. (8) As in the cognate languages, cobrir, ofrir, sofrir (and Ar. obrir) tended to share the same development; the change of conjugation seen in ofrir, sofrir is PR. But ofrir early changed to ofrecer, and sofrir partly to sofrer; from complir (semilearned) an OAr. interfix form complezca is also found. For nozir an OAr. infinitive nozer is found (Alv., Zam.), and Galician and Old Portuguese, less conclusively, have nocer/-ir, nozer/-ir, 7 8 suggesting that the verb may have been attracted to the -ir class by aduzir, luzir. Aborrir/aborrecer and contir/contecer, while exhibiting some old rhizotonic forms (aburra, cunta/cuenta in the L.B.A.), belong rather to the interfix conjugation. Tullir is a new formation replacing old toller, starting from tollido > tullido, and Ar. cullir corresponds to standard coger. For espurrir (< EX-PORRIGERE) for escorrir (< Asturian also has esporrer (pres. espuerre), and EX-CORRIGERE) Galician also has escorrer (Gar. D.). We have insufficient evidence to hand about the old forms of moñir/muñir (< MONĒRE) and polir/pulir. Sortir/surtir, like its cognates in the other languages, seems to come from two sources, SORTIRI and *SURRECTIRE and so also tondir/tundir (TONDĒRE/TUNDĔRE), condir/cundir (CONDĔRE/CONDIRE), the first two with original stem vowels both o and u. Mullir, being formed from an adjective, would seem to be a survival from an earlier mollir/mollecer (cf. enmollecer) with interfix in the present (so also enfurtir, aturdir, OSp. estordir/estordecer). The origin of curtir is uncertain. (9) For foir, roir, destroir (see §15 below), PR seems also to have had forms with (whether CL RUGIRE Ū, had Ŭ to judge by the cognate languages or Ū is uncertain); roir/ruir gave way to learned rugir. Confundir has been substituted for old cofonder. Coser/cusir seems to have been an -er verb attracted to this class (cf. It. cucire/SIt. cusere, cusire, Cat. cosir/dial. cúser; -ĔRE forms elsewhere), with tusir/toser showing the contrary development; the change of conjugation in escopir goes back to 8 9 PR. Surgir seems learned, but has popular cognates in the other Romance languages, and perhaps in OSp. surzir, modern zurcir. Engollir/engullir is an Iberian and Provençal formation (cf. Fr. engouler, It. ingollare), and follows bullir (older bollir/bollecer). OSp. sobollir comes from *SEPULLIRE for SEPELIRE (×SUB-), and seemingly survives in modern za(m)bullir. Nodrir/-ecer and podrir/-ecer belong to the interfix conjugation, and pudro is modern. Unzir (< IUNGERE), mocir (< MUNGERE) show a multiplicity of forms, including -er infinitives in Asturian and Galician. Sacodir (< SUCCUTERE), like foir, recebir, goes back to a PR *-IRE variant based on the yod forms. (10) Aduzir (see §15 below) only partly belongs here, being irregular. Pruir (< *PRŪDIRE for PRŪRIRE) is dialectal (Gal. also proer). Mugir seems learned, as rugir, unless it is dialectal. Sumir is presumed to have come from SŪMERE. (11) Oir, with o < AU, was not subject to diphthongization, but arrhizotonic u-forms are also found, as udir, udamos, huyé, huyendo, udié, udieron, and Berceo also uses imper. udi, subj. udas. From the above it is seen that the nucleus of forms to be considered decisive in setting this pattern is as follows: (1) ferir, mentir, sentir, (ar)repentirse, venir; (2) vestir, embestir, servir, seguir, pedir; (3) medir, re-çebir etc., fenchir, ceñir, costreñir, estreñir, teñir, feñir, reñir; 9 10 (4) bevir, escrevir, reir, freir, dezir; (5) none; (6) exir; (7) dormir, morir; (8) cobrir, ofrir, nozir, complir, escorrir, espurrir; (9) sa-codir etc., sobir, sofrir, foir, destroir, ordir, escopir, bollir, groñir, englotir, unzir etc., mocir etc.; (10) aduzir, sumir, luzir; (11) oir. 4. Alternation, where found. To get down to the origin of the vowel alternation found today, we need to study the effect of the interaction of all these types on each other, and not only in Spanish but in the neighbouring languages. It is my belief that there is no one single force at work, such as the “metaphony” often spoken about (called on, it would seem, as a useful formula to explain irregular vowel changes, without any examination of why metaphony should operate in one case and not in another), but a combination of many. It is such a tangled web that I cannot hope to untangle it completely, but I hope at least to isolate some of the strands. The vowel changes involved take place, as we have seen, under three sets of circumstances: (l) in rhizotonic forms, such as siento, sientes, sienta or sirvo, sirves, sirva; (2) in an unaccented syllable preceding a yod, as sintió, sirvió; (3) in an unaccented syllable when there is no yod following, as sintamos, sirvamos, 10 11 OSp. sintía, sirvía, sintrá, sirvirá. Similarly for the back vowels: (1) duermo, duermes, duerma, subo, subes, suba; (2) durmió, subió; (3) durmamos, subamos, OSp. durmía, subía, durmirá, subirá (with infinitive sobir/subir). Portuguese and Catalan also show alternation, and in all three languages there is a significant contrast between the -er and -ir verbs in this respect. In modern Portuguese the forms for (l) are as firo, feres, fira or agrido, agrides, agrida, and for (3) firamos, agridamos, and in the case of the back vowels (l) durmo, dormes, durma or surto, surtes, surta, and (3) durmamos, surtamos. (Note that whereas in the Spanish present indicative forms persons 1, 2, 3 and 6 have the same vowel, in Portuguese 1 is contrasted with the others.) The Portuguese distribution is a highly artificial one, cutting across etymological distinctions; thus frigir goes with ferir, and subir with dormir, in the rhizotonic forms and all the subjunctive. In Catalan the alternation shows the pattern OCat. sent, sents, senta, sintam; other similar old forms found are sintau, consintrá, vistam, iscam, firam, firén, sofirén, vinent, tinent, fugén, durmía, muría, sirvía, vindrá, tindrá, rechirás, provehiscam, conferiscats, vulía, vulríen, excluding others stemming from diphthongization (cf. also Prov. sirvén, covinén, iscam, firatz). 5. PR yod forms. First we will start with the alternation in the stressed syllable. Here, in comparing the conjugations to find a point of difference to account for this, scholars have traditionally sought for an explanation in the presence of the PR yod (though 11 12 this was also found in the PR -ĒRE verbs), which was supposed to produce the effect of metaphony. Thus Menéndez Pidal, for example, says (Men. M., 113.1) that in the -ir conjugation the yod affected the preceding vowel but that the analogy of the other persons did not allow it to affect the consonant; so mido comes from METIO(R) and recudo from RECUTIO (114), whereas in the nouns you get the contrary development, as vezo, pozo. Now it can be seen at once that this is a purely arbitrary theory; and I prefer to follow established principles and say that if VITIU gave vezo, and *ADVITIO gave avezo, then METIO(R) should be expected to give *mezo. Moreover, we have Ptg. meço to prove that this kind of thing did happen in the -ir conjugation as much as anywhere else; and we also have an OPtg. mido for which an alternative explanation must be found, as it must for Sp. mido. However, all this is not necessarily to say that the yod did not exercise any action, and we must first examine the extant yod forms in the three languages and see what light they throw. In Spanish we have, first, from the PR -ĒRE verbs: he, haya, seyo, seya, veyo, veya, OLe., OAr. valla, OUp.Ar. plaça, OLe. açamos (Alex.), muño, muña, cayo, caya, sé, sepa, quepo, quepa (hincho, hincha probably do not contain yod); to these can be added the forms with a velar interfix, which generally betoken a foregoing yod: valgo, valga, duelgo, duelga, suelgo, suelga, tengo, tenga, remanga, yazgo/-co, yazga/-ca, plazgo/-co, plazga/-ca, caigo, caiga, OSp. and popular haiga, veigo, veiga, while poder has OLe. possa/posca, OAr. posca/puesca/puysquen, 12 13 and OUp.Ar. voler has vuyllguan. The siegat etc. alternating with segamus in the Glosses shows the normal Aragonese diphthongization of the tonic vowel before a yod, though it is not clear whether the g represents a y, or is a velar interfix. In the latter case the word would parallel Ar. tienga for an earlier *tieña in contrast to the Castilian forms, where diphthongization is inhibited by the yod. Be it noted here that in many cases the earliest forms have no yod: yago, yaga, OLe. plaga, valo, vala, duelo, duela, suelo, suela, huelo, huela, OAr. nuega (contrasting with Berceo’s nuzo), Cast. puedo, pueda, devo, deva, muevo, mueva (plega, yega are formed from plogo, yogo on the proportion sopo:sepa; plogo:plega). We may also mention here the subjunctives formed with -ia, such as OLe., MAs. podia, molia, OAr. plaçia, sapian, and other MAs. forms like cabio, cabia/quepia (also cabo, caba) and similar forms from saber. (There is also a Moz. kumplya (G. H. 57) which seems to be a subjunctive.) It will be seen that muñir stands alone among the n-stems in its development, perhaps affected by MUNIRE (cf. the mediaeval confusion of Latin praemonere and praemunire). Henchir was attracted to the -ñir verbs. PR -ĔRE verbs affected are : trayo, traya, destruyo, destruya, with y from the other persons (PR *-GO, *-GES; OLe. has the original trago, traga), and riyo, riya, rayo, raya, royo, roya, vaya, OSp. creya, with analogical extension of the y, alongside rio, rao, roo, crea, OSp. vaa; leo, lea perhaps also represent older *leyo, *leya, as we would expect *liego, *liega < LEGO, 13 14 LEGAM; g has crept in in the popular forms creiga, leiga (Ar. liga), reiga, and in traigo, traiga, OSp. destruigo, destruiga, raigo, raiga, roigo, roiga, OSp. and popular vaiga, and in pongo, ponga, OSp. tuelgo, tuelga, cuezgo, conduzgo for earlier cuego, adugo, now cuezo, conduzco, and conozgo, parezgo for earlier -sco, now -zco, OAr. prengo, prenga, OLe. (F. de Av.) perga, MLe., MAs. muelgo, muelga/molguia. (Ciño etc. are not yod forms, but have the ñ from the other persons.) In the PR -IRE verbs the yod presents are, strangely enough, very few: muero < moiro (quoted by Hanssen, 198, without *, confirmed by modern muöiru in Sisterna), muera/OCas. mojra (G. H. 17), fuyo, fuya, where the y is supported by the other persons, as fuyes, if it did not originate there; old oyo, oya seem to follow fuyo, fuya, as the Leonese forms ozo, oza represent the expected development from AUDIO, AUDIAM (cf. gozo; PR -AUDfalls in with -ALD-, creating a consonant group); in Aragonese the ll of sallo, salla has been extended throughout; forms with velar interfix are salgo, salga for earlier salo, sala, similarly asgo, asga for earlier aso, asa, oigo, oiga, Le. ozga/ozca/osca, luzco, luzca (ealier luga), old exco, exca, huiga, OLe., OAr. fierga, with firgades in the Cid. Vengo, venga are based on earlier *veño, *veña where diphthongization was inhibited in Castilian by the yod; elsewhere ie appears, so As., Le., Ar. viengo, vienga, also Ar. vingamos, vingades. Leaving Portuguese to the last, because it presents the greatest contrast, and taking Catalan next, we find the forms 14 15 OCat. jas, plas, placia, vuyl, vulla, duyl, dulla, vail, valla, calla, puix, puixa, he, haja, tinya, capia, se, sapia, veig, veja, deig, deja, caja, vaig, vaja, faç, faça, llig, llija, muyr, muyra, oig, oja, fuig, fuja, viny, vinya, fira, soffira, requir, requira. These verbs also developed velar interfix forms, as did, in particular, the majority of the -ERE verbs; note siga, sapiga, for earlier sia, sapia, and isc/ixc, isca (similarly Valencian tixc for tix and old tixca). The overall position in Catalan is seen to be basically the same as in Spanish; in modern Catalan, however, there has been wide extension of the -ESC- interfix. (Provençal has some forms which can be compared to the Catalan ones, such as feira, firatz, sofieira, qu(i)eira, iesc, iesca, iscam.) When we come to Portuguese, however, we find a great many more yod forms: from PR -ĒRE verbs, OPtg. jaço, jaça (also jasco; the ç replaced by the z of the other persons in praza — but Gal. prasa — , nuza — but old nusso), hei, haja, sejo, seja, vejo, veja, valho, valha, doio, doia, soio, soia, tenho, tenha, remanho, remanha, caio, caia, caibo/cabio, caiba/cabia, sei, saiba/sabia, feço, feça; from PR -ĔRE verbs arço, arça, perço, perça, now perco, perca (and Gal. perga/perja), faço, faça, recebio, recebia, choiva/chovia, coimo/comio, coima/comia, ponho, ponha, tolho, tolha, queiro, queira, Gal. moio, moia; and, from PR -IRE verbs, fujo, fuja, mujo, muja, rujo, ruja, moiro, moira, feiro, feira, ofeiro, ofeira, pairo, paira, popular aibro, aibra, saio/Gal. sallo, saia/Gal. salla, grunho, grunha, venho, venha, peço, peça, meço, meça, senço, sença, menço, mença, 15 16 ouço, ouça, (z for ç in luzo, luza), servio/servho, servia, dormio/dormho, dormia. In comparison with Spanish it will be seen that Portuguese preserves more of the yod presents of the -IRE verbs and some others, makes a few more extensions than Spanish, e.g. arço, perço, but does not extend the yod to those other -DERE verbs, such as raer, roer, to which it has been extended in Spanish, nor make use of the velar interfix (perco is a special case). The significant points that emerge from an examination of the state of affairs in Portuguese are firstly that Spanish and Catalan have only a limited number of yod forms in the -ir verbs (agreeing with the other Romance languages in this), whereas Portuguese has more, and secondly that forms like pido, mido are clearly seen to be of a different origin when compared with Ptg. peço, meço, so that Spanish must either have had no yod forms here from the start, or lost them in pre-literary times. A further proof that Sp. pido, mido are new formations is provided by the existence of pido, mido also in Old Portuguese, and the replacement of OPtg. menço, senço, servio, dormio by mento/minto, sento/sinto, servo/sirvo, dormo/durmo. It should be noted here that Portuguese is far from showing yod forms for all the -ir verbs; there are none, for example, for vestir, bulir, tossir, sacudir. But in the case of some of the other stems ending in a labial, the modern forms may conceal an earlier yod; for example, subo may come from *suibo (for *subbio, thus giving b, not v), and cuspo from *cuspio/*cuispo, 16 17 as chuva comes from chuiva. And as recebio gave way to recebo, *devio might have given way to devo (for recebio we would expect *recibio/*recibo, as siba, vendima, but evidently the -er conjugation did not admit of such metaphony). Now, supposing we recreate yod forms in the Spanish -ir verbs that show metaphony, to correspond to the Portuguese forms, we get: *recibio, *requero, with -er- as in madera (cf. Ptg. requeiro as madeira) and dial. requiero, agreeing with the form without yod, similarly *ofero (dial. *ofiero), *fero (dial. fiero, corresponding to the form without yod and to fier-ga), *veño, which corresponds to ven-go (and dial. *vieño, corresponding to vien-go), *mezo, *piezo (or *pezo, if attracted to *mezo), *sienço, *mienço (as lienço/-zo), *servio (as nervio), *dormio, to which we may also add *subio, *escupio (or *sobio, *escopio, cf. rubio/ruyo/royo, gubia and gobio). The u of gruño could be due to the yod (cf. cuña), perhaps so also the u of non-Castilian bullo (cf. Ptg. gorgulho; the Castilian form would be *bojo, as gorgojo), but fuyo, fuyes seem rather to be traceable to PR *FŪGIO, *FŪGIS, as do the forms of the other Romance languages (in which case Ptg. foges is secondary, as somes). The Catalan forms also agree with some of these reconstructions: requir, fira, soffira, viny/vinc, escup, gruny. From the above it can be seen that the only forms in Spanish which might go back to yod forms, apart from fuyo, gruño, bullo (?), are those from stems ending in an r or a labial, as these consonants were not changed by a following yod. On 17 18 this basis, muero, with its diphthong coinciding with that of mueres < *MORIS, can be traced directly back to *MORIO, but in the case of the other verbs, even supposing the yod was later absorbed (which we cannot be sure of), we are still far from arriving at the pattern we are seeking; we are left with recibo, recebes (recebe is found in the F.J., but then so also is the infinitive receber), subo, *sobes, *dormo, duermes, *servo, sierves (sierve is found dialectally), *fero, fieres, in all of which the first person is in contrast with the others, as in Portuguese, rather than with the characteristic Spanish pattern of unaccented e alternating with accented ie or i, and o with ue or u. So we must look further afield for possible influences. In passing, I might say that it is not impossible to conceive of the existence of a secondary yod, as seen in Le., As. podia, extracted from the primary yod of forms like veya, riya (cf. Fr. -ions, It. -iamo etc.), which had the metaphonic effect seen in limpio, tibio/tebio. But one is still left with the problem of why and how it was levelled out afterwards. 6. Arrhizotonic subjunctive forms. At this point, while we are still considering primary yod, it may be good to turn aside and look at the arrhizotonic forms of the subjunctive and see what effect the yod might have had on pretonic e and o, confining ourselves once again to the forms where the yod did not affect the preceding consonant. To take the position before an r, in nominal forms, which should show an independent development, 18 19 you have maderaje but ciruela (< *CERJÓLA < CERÉOLA), agorero, corambre but culantro, and also salmorada in Alex. P as against salmoyrada (whence salmuerada) in O; on balance it looks as if the yod here disappeared without affecting the vowel. Before a labial you have liviano, and vendimiar, uviar, if these forms can be relied on as typical (cf. Ptg. vindimar, old oviar/uviar/uivar, Cat. veremar, old ujar), and before ñ, ll, cuñado, dial. muller, señor but riñon, dial. mellor/millor, and before a single s, tusón. On this basis we could explain durmamos, subamos, escupamos, sirvamos, gruñamos, dial. bullamos, Ar. vin-gamos, dial. cusamos, but not firamos or muramos (in contrast to OPtg. feiramos, moiramos, as madeirar, coirama). But this is still only a drop in the bucket. And on the other hand we have some old forms without metaphony, such as servamos in the F.J. (quoted by Hanssen, 200), mentades (Yúçuf), descobrades (L.B.A., 879b), OAr. differades (G. H. 119), backed up by ven-gamos, which suggest that the metaphony may have come later; a special case is presented by oyamos, oyades with o < AU, which are, however, flanked by Berceo’s udamos, udades. Equally, the -er verbs show no metaphony due to yod, though in this case levelling was at work also. On the other hand there are cases of pretonic Spanish e and o changing without a following yod, e.g. dinero, cigüeña, piojo, timón, hinojo (both words), hinchar, tijeras, culebra, dudar, pulgar, lugar, jugar, vulpeja, pulmón, so that there might have been a tendency to such a change in the subjunctive too, though it would need reinforcing from other 19 20 sources to make it permanent. (A similar tendency exists in Catalan, Provençal and Portuguese too, and could have had the same effect on the verbs.) This tendency would also account for the metaphony found in Old Spanish in the future (sintrá), and in the imperfect (sintía), though this last form would also be affected by the secondary yod in sintié. The alternation exca/yscamos seems to favour this latter interpretation, and we can also explain muramos, firamos, firgades this way, as well as the forms of every other verb. 7. Secondary yod. When it comes to this secondary yod, causing metaphony as in the nominal forms siniestra, hiniesta, luci(e)llo etc., which affected the Old Spanish imperfect and the perfect paradigm, the metaphonic effect is too vacillating and uncertain to be decisive by itself, without some other influences at work. This change was completely resisted by the -er verbs (Malkiel, however, gives a few examples of change, Mal., p. 468, quiriendo, trimió, (con)tiniendo, virtió, all, to my mind, to be explained by the analogy of compounds in -ir or conjugation change; for other exceptions, see below), because there was no pattern of vowel alternation for it to fit into, while in the -ir verbs in Old Spanish one could find equally siguió/seguió, pidió/pedió, durmió/dormió, subió/sobió, diziendo/deziendo, comidiendo/comediendo and so on. One influence in favour of i and u came from the strong perfects where the i and u of the first person were extended throughout the paradigm (aided by the 20 21 presence of secondary yod). So first fezo, veno, preso, queso, poso were changed to fizo, vino, priso, quiso, puso, and then fezieron, venieron, posieron etc. were changed to fizieron, vinieron, pusieron (in the old language there is fluctuation, with even forms like dexieron as dexiste < dixiste by dissimilation). In the case of poder, the similarity between podiendo and podieron produced a new participle pudiendo to match pudieron, and even a conditional OAr. puriedes (G. H. 101), creating an instance of o/ue/u alternation in an -er verb; similarly in ver the influence of vieron etc. produced a participle viendo for earlier veyendo, and also an imperfect vía, vié, now disused (cf. ría, rié from reir). In the case of ser, siendo and sido have replaced seyendo and seido, perhaps as unaccented forms, since the very rare dialectal perfect sivieron can hardly have had any influence (note also an isolated lyendo in the L.B.A.). Other apparent alternations in -er verbs are due to a change of conjugation; Ar. vinció, atrivió, cullió, for example, belong to infinitives vincir, atrivir, cullir. Anyway, to return to the -ir verbs, it seems that another influence besides metaphony was necessary to tip the balance in favour of i and u rather than e and o. 8. The imperative. Leaving this subject now for the moment, and returning once again to the rhizotonic forms, we find one more source of change limited to the -IRE verbs, and that is the imperative. Here Old Portuguese, once more, is very instructive. It has forms like subi/sube, riimi/rime, fuge, dispe, where the 21 22 final -Ī changed the Ē and Ō of PR to i and u. This metaphony evidently became a sign of the imperative, as we find it extended to all verbs, thus segui/sigui/sigue, pidi/pide, cubri/cubre, fire, sirve, minte, sinte, viste, durme, and to the -er verbs as well (see Nunes, pp. 286-7). So on the basis of the Portuguese forms we could say that Sp. mide, recibe (if from *RECIPĪ), sacude, sube, fuye, bulle, gruñe, escupe (and Cat. acut, fuig, bull, gruny, escup, tus, cus) would be inherited forms. In the case of ven and Cat. vin-a the final -Ī, when coming before a vowel in the next word (“veni ad me”), has acted as a yod, impeding diphthongization in Spanish, and, on the other hand, causing it in Cat. vin < *viein (cf. similar Cat. (a)hir, OPtg. eire < (AD)HERĪ). 9. Monophthongization of ie. The next possible influence we need to examine is one discussed by Malkiel (Mal.), in which ie tends to be changed to i in the neighbourhood of an s in association with r or v, or, occasionally, in isolation (cf. prisa, avispa, ristra, siglo for earlier priessa, aviespa, riestra, sieglo). This could have affected three verbs, servir, vestir and seguir. We have seen that the etymological reflex of SERVIO would have been *servio, as nervio, which is not found, nor is any yod form found for vestir, while it is unlikely that one ever existed for seguir, as this verb was a late changer to the -IRE class (cf. Fr. suivre/sivir, Prov. segre/seguir (rare), Cat. seguir but also cossegre, It. seguire with presumably an -ere form in the south, where this is usual; Gal. seguer is probably a later formation). 22 23 The forms *siervo, *viesto, *siego are not found either, but we have in the subjunctive, which should have preserved the yod if it had existed, dial. sierva, siervan, bestanlo(s) (this in a text from Sobrarbe which is hesitant about diphthongization, thus tiega/tenganlo; G. H. 92) pointing to PR forms without the yod, in line with OPtg. vesto, vestas (once), sega (Nunes, p. 340). In forms 3 and 6 of the indicative we also have the similar dial. sieguen, sierve, sierven/sierben, dessierven, vieste, viesten, notably in Alex. O, where Castilian and P have i-forms, this situation agreeing chronologically with dialectal retention of ie as against Cas. i in nominal forms (for a wide-ranging documentation see Mal., 455-6; I have also found other examples of Moz. syerbe and Nav. siervan in G.H. 57 and 77, while the L.B.A. has undiphthongized servi (imper.) and serven, and the F.J. segue beside siegue). The ie in these forms then changed to i, or had already done so in Castilian, producing the specifically Spanish pattern of agreement between persons 1, 2, 3 and 6 as against 4 and 5 in the indicative, and a uniform subjunctive, if we suppose metaphony in the plural. At the same time there is a possibility that pedir and medir were attracted into this class. Malkiel quotes pieden from Alex. O, where the imperfect pidien is required, as found in P, and mieden (Alex. O; miden in P), miedan (F.J.); I find pieden also occurs in the F.J. (Llera). The question is whether these ie-forms are Proto-Spanish, with medir < *MĒTIRE, with Ē, affected by pedir, or whether the ie is a typical Leonese substitution for e or i (cf. 23 24 F.J. tiemen, vieno, similarly F. de Ov. uieno/F. de Av. ueno, this last probably on the analogy of dieo/deo (F. de Av.) beside dio). Malkiel observes that ie-forms are not found in Aragonese, which has the old form pitent in the Glosas Silenses, and suggests that Proto-Spanish may never have had ie. On the other hand, piede is the form we would expect, unless pedir was attracted to medir, and its absence from Aragonese may be an indication that this verb (and medir) came into Aragonese from Castilian. Pedir and medir are neither of them found in Catalan, and Aragonese often follows Catalan in its vocabulary; and the Glosas Silenses do, after all, come from Castilian territory. It is tempting to see here an influence for changing riendo, yergo etc. to rindo, irgo, but it is doubtful if the changes are contemporaneous. And a strong group of PR -IRE verbs with ie, ferir, requerir (con-), sentir, mentir, repentir, were left unaffected (but see repi(e)nte, Mal., 460, and requiro (Apol.), Mal., 465), so it looks as if this change from ie to i only affected the verbs mentioned above. (Ferir may have resisted change because of older forms *fero, *fera < *feiro, *feira with yod; *fero, *fera were then adapted to fieres, fiere.) A special case is presented by exir; in Castilian the rhizotonic forms are exco, exes, exe, exen, exca, ex (contrasted with yscamos, Cid, 685), but the other dialects have both yex-/yess- (yexen, Alex. O) and yx-/yss- (yxen, Alex. P), besides yesca (F. de Ar.), esca (F. de Av., which does not usually have diphthongs). In Castilian it forms a class of its own. In parts of Navarre and Aragon we could regard 24 25 yx- as coming from ye(i)x- by reduction of the triphthong as in vin-ga, tin-go and Cat. ix, tix, llig, fira etc. (cf. also muyr, cu(y)ll, vu(y)ll etc.); in other areas it is best regarded, as Malkiel says, as paralleling the change from sierve to sirve etc. 10. The -INGERE verbs. We have now exhausted the possible sources of influence to be found in the original PR -IRE verbs, as far as the front vowels are concerned, and we have seen how we could get i for ie, but, apart from the influence of the imperative, we have not found how we could get rhizotonic i for e, e.g. mides, recibes for *medes, *recebes (taking recebir as a PR -IRE verb). The only e-form found is the recebe of the F.J. mentioned in § 3 (3) above, which can however be referred to an infinitive receber (Malkiel’s receba (Mal., 467) appears to be a misreading of Hanssen). We therefore need now to cast further afield and explore the other verbs which appear in the -ir class in Ibero-Romance. The first of these groups is the -NGERE verbs. Here the position in Spanish is very confused, as the different dialects show different treatment. In Old Castilian, in the -ANGERE verbs, we have tañer, atañer, plañer, and frañer/franzer, with a change to -ir in modern plañir, franjir (given by Velázquez as ‘obsolete’), that is, there was no change to -ir in the old language (-ñ- was originally post-tonic and -nz- pretonic, and then the two forms became interchanged). Modern Leonese and Asturian have francer/francir/frañir. tañir, Mozarabic atañir/atanguer, has franne-, changir, Aragonese 25 26 planyer. Galician has changer, tanger, Old Portuguese changer, tanger and franger, now frangir/franzir (crossed with the cognate of Sp. fruncir). Catalan has tànyer, atànyer, plànyer, frànyer. As far as the a-stems go, then, the tendency is to keep the -er infinitive. When we come to the -INGERE verbs, we find a different picture. Old Castilian has only infinitives in -ir, ceñir, feñir, estreñir, restreñir, co(n)streñir, teñir, reñir, with present tense cingo, ciñes, ceñimos, constringo, constriñes, constreñimos, and subj. cinga, constringa etc. There also appear to be no -nz- forms. (We may also add fenchir, which was attracted to this class.) For Castilian, fingir and restringir must be classed as learned, though they need not be so in the other dialects. For a form with exceptional e for i, note çentura in the Razón de Amor. Variations come as we go west. Asturian (Gar. D.) has (earlier) rancer/(later) riñir, restringir (cf. changir), with present as ringo, riñes/(later) rances (Zamora Vicente also gives rañer). Malkiel (Mal., 465) quotes Hanssen for western variants costrener (with n = nn)/constrennir/constringir, reñer/reñir, fiñir, encher/enchir, and to this add çinir in Alex. O. In the F.J. we find enfinne/enfiñe, and for constreñir, (r)estreñir an almost limitless number of infinitive variants formed by the interchange of e/i with ñ/nn/ng and -er/-ir, and almost equally various rhizotonic forms constrenne/-inne/constrenge/-inge, co(n)streng(u)a/co(n)stringa. Galician (Gar. D. ) similarly has the pairs renger/ringir (I am arbitrarily using traditional orthography for the sound which now 26 27 has the value /∫/), cenger/cingir/cinguir, fenger/fingir, tengir, costrenger, estrenger, with present as tingo, tinges, while Portuguese (Nunes, p. 277, and Michaelis, Taylor) likewise has old -er, modern -ir, as cinger/-ir, tinger/-ir, finger/-ir, renger/ranger/rengir/ringir, co(n)stranger/-ir/constrenger/-ingir, (r)estringir, estingir/extinguir, impingir, atingir, with similar present OPtg. cingo, now cinjo, cinges, OPtg. fingo, OPtg. cõstrengo. (The two forms ranger/renger and OPtg. cõstrengo make it appear that constranger is a phonetic variant of constrenger, rather than a hypercorrection as Malkiel (Mal., 454) would have it; I find no reference to an *atanger beside atingir which would have acted as a model.) Doubling back to the east, in Aragonese and Catalan we find the verbs divided among two infinitive types; thus Catalan has cenyir, tenyir, renyir, with present cinc (?), ciny, cinyen, tiny, riny, but also cínyer, tínyer/tènyer, and estrènyer, costrènyer, empènyer, espènyer, atènyer, acontènyer (Cor. Dic.), fènyer (Moll, 163b), with present estrenc (now estrenyo/-nyc), estrenys, empenc (empenyo), empenys etc., while Aragonese (Gar. D., G. H., et al.) has costrennyer/-eyner/-eigner, empeñer, conteigneria, likewise p.p. costrenido with n equivalent to ñ (cf. Gl. Sil. constrinitu), present constrengo, constrenga(n) (but pres. p. estrinyendo), as against Gl. Sil. tingen. There is also an OAr. tennieu, with the -ieu that belongs to the -er verbs, though this was occasionally extended to the -ir verbs. (Documentation for Catalan appears to be scanty, and many -ir verbs now have the interfix; I have not 27 28 actually found a present form *cinc (now cenyeixo), but there is a strange perfect form sincs (Rus., 41:75) which appears to equal *cinc(h) < *CINGUIT, as planch, strench, empench, subj. atanguessen, so with the support of cinyen (Rus., 68:72) I deduce rhizotonic i throughout.) Provençal also provides some related material; Anglade (p. 61) gives alternative p.p. forms sincha/sencha, fincha/fencha, and Bec (pp. 277–8) says cinta is Gascon and West Languedocian, while cencha is East Occitanian. So much for the extant forms, but the problem is to sort out what the Proto-Iberian forms might have been. As far as Cas. cingo, ciñes go, we would expect to have *cengo, *ceñes, as lengua, leña, ceño < CINGULU (but cf. on the other hand domingo, jeringa, tiña, lonninco, prominco (G. H. 56), and As. llingua, Ptg. lingua, domingo, seringa, provinco). Here the explanation of the i of cingo, cinga probably goes back to Latin CĪNXI, CĪNCTUS, with I regularly lengthened before NX, NCT (Buck, 99-2); for the hesitation between e and i from Latin Ĭ before a palatal cf. the various forms in the Romance languages derived from CILIUM, CONSILIUM, MIRABILIA, LIGNUM, SIGNUM (see Buck, ib.), TINEA and -ICULUM, and especially the alternation found in Prov. celh(a)/cilh(a), conselh/consilh, meravelha/meravilha, lenh/linh, senh/sinh, tenha/tinha, abelha/abilha. In all the Iberian languages, then, it looks as if the hesitation between e and i goes back to Latin times; the other Romance languages levelled one way or the other, Tuscan to i (other Italian dialects retaining the 28 29 e of the present paradigm), French and Provençal to e. It is noteworthy that Aragonese and Catalan have divided the conjugations according to the vowel of the rhizotonic forms — cenyir:cinyen, estrenyer:estrenyen; could it be that cinyen and tinyen established their i-forms because of the p.ps. cint, tint, whereas estrenyer, empenyer, atenyer had p.ps. estret, empès (and substantival empènta), atès (Rus., 73:88) with e ? In the west, likewise, the existence of co(n)streñer, costrenger, co(n)stranger and reñer, rancer, renger, ranger, but no *cenger, *tenger, may be an indication that there was once a similar division, which was later obscured by the confusion of the -ir and -er types, giving a double conjugation for many verbs. Then in Castilian, typically, the -ir type prevailed. In Galician, on the other hand, this confusion could have set the pattern for a vast migration of -ir verbs into the -er class, resulting in doublets like pidir/peder, vistir/vester, sintir/senter, mintir/menter and siguir/seguer, with presents such as visto, vestes, sinto, sentes (Gar. D.), as also in Portuguese. (Whether cingo, cenges could be an early alternation setting the pattern for modern Portuguese is a question, but it seems doubtful to me.) Why the rhizotonic i came to be associated with the -ir infinitive is another question, made all the more interesting because it affects all three Iberian languages; the answer may lie with the next group of verbs I wish to discuss. (The -UNGERE verbs show the same line of development as the -INGERE verbs, but I will leave them till later, when I discuss the back vowels.) The reflexes of IMPLERE, 29 30 evidently being attracted to all the different types, show great vacillation: Cat. omplir, omple/old umple, Cas. fenchir, finche, Ar. empler/impler/implir/enplirnosamus (Gl. Em.), ORioj. imple (Mal., 461), Le., As. encher (in Sisterna, incher)/enchir, Gal., Ptg. encher. One more old -ERE verb, redemir/-imir/remedir, has redimo, redimes through learned influence. 11. Dezir, etc. The next and final group to be considered in the front vowel series are those old -ERE verbs (all -ĔRE in PR) with stem vowel I, which all changed over to -ir in Castilian. These are, as we have seen above, VIVERE, SCRIBERE, RIDERE, FRIGERE, DICERE and compounds, which give OCas. bevir, escrevir, reir, freir, dezir. In the dialects you have the variants Ar. (ar)ri(y)er/arriguer/reir, dir (OAr.)/decir/escontraecer, Le., As. ri(y)er, escrever/escriver, viver, dizer (dicer)/contradizer/-ir and a present dece (F.J.). Galician has viver/-ir, escriver/-ir, r(i)ir, firgir, decer/-ir (dez-, dic-, diz-), and Portuguese viver, escrever (old 3 escrive, pointing to an infinitive escriver but giving no clear indication of the existence here of escrivir, as found in Gal.), r(i)ir, frigir, dizer, with popular forms vever, fregir, dezer (Nunes, p. 275). Catalan has viure, escriure, riure, fregir, dir. The old Latin accentuation also survives still in the place names OSp. Benbiber, Benvivre, now Belbimbre (Oríg.); note also the old futures biuré/viuré (Alex. P and O). We now need to look at each of these groups of verbs, and see what effect their special development had on the -ir verbs as 30 31 a whole. In Castilian the alternation e/i (with e due to dissimilation before a following i, cf. vezino and OPtg. vezinho, Cat. veí) is especially interesting from this point of view. The first question is, why did these verbs move over into the -ir conjugation in Castilian? I have hazarded an answer to this in my earlier papers (Ronshu, 8 & 10), and it is worth repeating here, with a side glance at the other dialects and languages. The etymological derivatives of the Latin infinitives would be *vivre (cf. Benvivre), *escrivre, *rire, *frire, *dire (from *DIGERE). If you then suppose analogical *vire, *escrire (cf. Fr. écrire), and take away the final e’s, you are left with *vir, *escrir, *rir, *frir, dir (found in Old Aragonese, and forming the basis for the future diré), which are easily converted into the Castilian forms by reforming on the stem as found in the arrhizotonic forms. Thus what was the stem vowel has become the vowel of the infinitive ending. (Similarly fer < *faire becomes fazer; for the u-stems see below.) On this basis, the notably western forms viver, escriver/escrever and ri(y)er would be later formations. As far as Galician and Portuguese are concerned, viver (beside Gal. vivir) and escrever (beside Gal. escrivir) may exhibit the typical trend away from -ir to -er; in fact, escrever, escrevo are only understandable on the basis of an earlier *escrevir with dissimilation (Gar. D. gives Gal. escrebir), unless we imagine an earlier paradigm escrivo (attested), escreves, escreve following sirvo, serves, serve, though this is unlikely, given the earlier escrives, escrive; similarly freges, frege, unless likewise 31 32 analogical, may point to the antiquity of the popular form fregir (here Galician has firges, firge). Old Portuguese riir would have come from an earlier *rir, and then have been changed back again to this form later. In the case of dizer, the existence of Gal. dezir/dizir suggests that dizer, like viver, may be a later form. The same arguments hold good for Leonese and Asturian. In Aragonese, reir/riyer are parallelled by leir/liyer, so there may have been interaction between the two here and in Asturo-Leonese; note that Aragonese at first kept the D of RIDERE, so this would not have given *rir. The changes in Catalan that are interesting from our point of view are as follows. Viure, escriure, dir form the plural of the present vivim, escrivim, deim/diem etc. (and dur likewise makes duim/duem). They can thus be compared with the Castilian forms, and perhaps came about in the same way, that is, with the sense of an infinitive ending in -ir; the imperfect forms vivia, escrivia, deia would also have helped. In the case of riure, the imperfect reia has given rise to the form reim in the Balearics (Moll, 310), and the same thing has happened in the case of creim, queim, feim, seim, treim, veim. Fregir is from *frigir by dissimilation; forms without interfix are still found in Cataluña and the Balearics — frig, friges, subj. frigi (Moll, 349), as llig, lliges, lligi from llegir. Of these verbs, clearly the most important from the point of view of frequency of occurrence is DICERE. We have seen that an alternation exists in Castilian, Galician, and Catalan, thus: 32 33 OCas. digo, dizes, dezimos, dezía, diga, dizendo, dixiesse; OGal. digo, dizes, dezimos, dezia, diga, dizendo, dissesse/dixesse; OCat. dic, dius, de(z)im, de(z)ia, diga, di(z)én, dixés. But in Galician and Catalan both de- and di- forms are found pretonically (later in all positions in Gal.), as the dissimilation of the two i’s was not observed so strictly (cf. MPtg. divino, vizinho for OPtg. devino, vezinho). But the results in Castilian are very significant. Whereas in the other languages this alternation was blurred by levelling in both directions, in Castilian it was maintained. Now it will be seen that the alternation e/i which is now the mark of the verbs of the servir type is precisely that found in dezir, reir, freir, bevir, escrevir, and furthermore that it is only in these verbs that it is a phonetically inherited type, fenir having become an interfix verb long since. In other words, it looks very strongly as if, while developments like *siervo > sirvo, sierve > sirve may also have played their part, along with dialectal forms like yexe > yxe, if it had not been for this strong type, which in its turn probably helped to fuse the ceñir type, this alternation would not have become firmly established in Castilian. (Note that the ceñir type also shows -ir forms right across the Peninsula; the alternation between cingo, ciñes, with i as in cinxe, cinto < CĪNXI, CĪNCTU, and ceñía < CĬNGEBAM/*CĪNGEBAM would have identified it with the type digo, dizes, dezía, so drawing it into the -ir conjugation.) Once established, this type then probably first attracted the now anomalous-seeming verbs with stem vowel e: recebir, redemir, medir and perhaps pedir, if it 33 34 had been attracted to medir. (We have seen that recebir might already have had a yod form recibo, reciba, and that the imperatives had become recibe, mide and perhaps pide; redemir was probably always subject to influence from Latin REDIMO, REDIMIS.) How the other verbs, that is, the verbs with e/ie alternation from the -er conjugation, such as rendir, hervir, erguir, were attracted to the -ir conjugation, is a question of pure speculation. In other cases, such as gemir, convertir, one can invoke the mystic words “learned borrowing” to justify adoption into this conjugation (though GEMIRE is found in late Latin). All that we can be sure of is the final result of a contrast between the -er and -ir conjugations which amounts to a kind of vowel harmony, -er being associated with e or ie, and -ir with i. Though we have not found a fully satisfying explanation to cover all cases, the phenomena being too complicated and many factors having been at work, yet it can fairly be said that the above analysis helps to show at least what the prime movers in the process may have been. And most importantly, it puts the effect of the yod in its proper perspective as being only limited, thus establishing that the agreement between persons 1, 2, 3 and 6 is natural in Spanish (always excepting the case of Cas. vengo, vienes etc.), in contrast to the state of affairs in Portuguese. 12. The back vowels; o/ue/u. We must now try and trace the pattern for the back vowels, which we left earlier, having shown 34 35 that the yod had effect in the case of foir and morir, and might also have affected sobir, escopir, groñir and bollir. We also saw that the final -Ī might have affected the imperative, that pretonic o could change to u by metaphony, and that foir probably had an inherited u in the rhizotonic forms. So let us now take up the subject again, beginning with the verbs with stem vowel PR Ŏ, morir, dormir, cobrir, ofrir, nozir, complir, escorrir and esporrir, and also taking in sofrir, which was attracted to this type. In the case of morir, the ue < oi of the yod forms agreed with the ue < ǫ of the other persons, producing a unified type; in the case of dormir we can suppose ue < ǫ in all persons, in the absence of any yod. Durmamos can also come from *DORMAMUS. Muramos, however, and equally the OSp. mueramos (Mal., 466) cannot come from *MORIAMUS, if corambre shows the normal development, so both must be analogical. For the other verbs we have Le., Ar. cuebres, cuebre, descuebre, encuebre (Men. M., Gar. G., Cor. Dic.), and descobre in MSS. G & C of the Libro de Buen Amor (1. 569c), descobrades (ib., 879b), otherwise rhizotonic and subjunctive cubr-; uffre in Berceo (Cor. Dic.), but generally an early change to the interfix class; nuzo in Berceo, nozir/nozer, nueze, nuezen in Alex. (this last perhaps the “OCas.” nuezen quoted by Hanssen), and other instances of Le., Ar. nozer, nueze, nueçe, nuecen, also Ar. nuega (Cor., Alv., Zam., Gar. G.), otherwise 35 36 rhizotonic and subjunctive nuz-; “OCas.” cuemplan (Hanssen), F. de Av. compla (Mal., 470), otherwise rhizotonic and subjunctive cumpl- (Malkiel, Mal., 449, also speaks of a hesitation “between, say, 2 cobres and cubres, or conples and cunples” in some MSS.); escurra, escurren in the Cid; espuorre/espuerre in Asturian (Gar. D.); suefres, suefre in Alex. O (Cor.), also OLe. suefro/sofro, suefra/sofra (Hanssen, quoted by Malkiel, Mal., 465), otherwise rhizotonic and subjunctive sufr-. There are also interesting variants cuenta, cuentan, in MSS. C & T of L. B. A. (1. 1400b), where S has cunta, from contir; all three are all the more interesting because I would expect contesca (another variant) to be the only form found, but perhaps this is a verb, like ofrir, fallir, which changed to the interfix class within the history of Spanish, and not in PWR times. The same document has aburra (114b, confirmed by the rhyme), where I would certainly have expected aborresca to be the earlier form, in line with the interfix conjugation of the other Romance languages, including Rumanian (though Italian has both aborro and aborrisco). The ue of suefro etc. accords with OFr., OProv. suefre and the open o of It. soffre, and shows that in the case of each language the verb was attracted to cobrir, ofrir and their respective cognates. The picture that emerges is similar to that of the front vowels, namely that the diphthong ue has been replaced by the single high vowel u, with a few examples of the older state of affairs preserved in the dialects. However, in this case there is no 36 37 natural sound change to account for it (the only alternation between ue and u is in vidueño/viduño (Men. M., 14.2d), but here the etymon has Ō, so it is unlikely that muñir was similarly affected), and we are thus forced to speculate as to what analogical forces could be at work. One point of difference from the front vowels is that the verbs with a close stem vowel in PR are more numerous than those with an open one. So it is possible that the forms cobre, conple, sofre indicate that these verbs were attracted to the sobir class (or sofrir/-er to romper), unless again these are just forms in which ǫ has not diphthongized (cf. F.J. tuello/tollo); conversely, as we have seen with the front vowels, in Leonese and Asturian diphthongs appear for the close vowels, thus F.J. puenen, ruempe, As. cues(e) (Zam.), ascuendo, respuendo (Men. M., 13.4). The other possibility, more applicable to the next class of verbs, is that the sense of a conjugational pattern of metaphony was created, with the back vowels conforming to the alternation e/i by setting up an alternation o/u (later, of course, to be eliminated by levelling in favour of u). We should also consider the possibility that the change from ue to u might have been assisted by that from ie to i; in both cases some verbs were left unaffected, thus duermo, muero as hiervo, yergo, hiero, -quiero. 13. The back vowels; o/u. In the verbs with stem vowels PR and Ū Ō we find a division of types in Portuguese and Catalan, 37 38 which is worth studying to see if it throws any light on the position in Spanish. In Portuguese all the back vowels are divided into two types irrespective of origin (as were the front vowels, see § 4 above, §16 below). Those with u/o in the present are: dormir, cobrir, sa-cudir etc., subir, fugir, de-struir (con-; this pair also just with u), cuspir, tossir, bulir, engolir, surgir (Bourciez, 360), (con)sumir; Gal. rugir, mu(n)gir/mogir, resurdir, pruir/proer, lucir (so also the popular forms of Ptg. luzir). Cumprir formerly belonged to this type; ungir, mungir formerly had forms *onger (p.p. onjudo)/ongir, monger (no alternative given by Nunes for jungir). All the rest of the verbs with u in the infinitive keep the u throughout, though some also have o-forms, preferred in popular use (Nunes, pp. 288-9); those with o hesitate, as exturco/extorco, or avoid using the forms in question. Morrer, sofrer and coser (Gal. also sofrir, cusir), of course, have o (Gal. still moiro). The position here is evidently so confused as to be uninstructive. In Catalan PR Ŏ gives rhizotonic o: 3 dorm, mor, obre, comple (as omple; OCat.), cobre (OCat.); PR Ō, Ū both give u, but with a varying distribution of vowels in the infinitive: (l) cosir, escopir, sortir, tossir, (2) a-cudir etc., bullir, (es)munyir, fugir, grunyir, engullir, lluir, pruir, pudir, OCat. also destruir, junyir (now only with interfix). Collir makes cu(y)ll by resolution of the triphthong *cueyll (cf. OAr. cuillgades, as Pyrenean vuyllgua from voler, where Catalan. has vulla, later vulga). The only possible light thrown here comes from the verbs in class (2) coming from PR Ō. 38 39 The first is fugir, which agrees, as noted above, with all the Romance languages except Portuguese in having rhizotonic u throughout, probably going back to PR Ū (so Ptg. foges would be a later formation, like somes, lozes). The same can probably be said of destruir and rugir (interfix class in Catalan). Bullir, engullir may have been influenced by collir; but the change may also have originated earlier, as Provençal has both bolh and bulh (Angl., p. 81), and in Roussillon they use forms like coscull, ginull for standard coscoll, genoll. Similarly, grunyir may have followed the -UNGERE verbs, or it may have been subject to the fluctuations seen in puny/poin (Rus., 42:2), vergonya/Rous. vergunya, Prov. conh/cunh, ponh/punh (Angl., ib.). In these three cases the change would have originated with the yod forms and been extended to the other persons; and then this whole group, backed up by the metaphony in the imperative, would have affected cosir, escopir, sortir, tossir and acudir. (This last word, by the way, and sacudir, percudir, are rare examples of Catalan agreeing with Spanish as against Provençal, which has escodre, socodre; the Spanish-looking a of sacudir makes me wonder if they are not Spanish importations.) If we suppose the same state of affairs in Spanish, we get fuyo, fuyes < *FŪGIO, *FŪGIS, and gruño as puño, cuña; but for bullir, engullir and za(m)bullir < sobollir/sebellir (Cor. Dic.) we would expect forms like *bojo, as coscojo, hinojo, in Castilian, but with -llo in the dialects. However, in the case of bullir, we may be dealing with a PWR BŪLLIRE variant. Even so, we still 39 40 have not much evidence of rhizotonic u in the PR -IRE verbs, the only other occurrences being in the imperative and perhaps the yod forms *subio, *escupio, backed up by a tendency to adopt arrhizotonic u-forms. Our last hope, as with the front vowels, is in the -UNGERE verbs, and the old -ERE verbs with stem vowel PR Ū. 14. The -UNGERE verbs. Of the first group, only MUNGERE IUNGERE, (×MULGERE), and to a slight extent PUNGERE (e.g. barua punnientes in the Razón de Amor) are represented in Spanish, IUNGERE having ousted UNG(U)ERE for phonetic reasons. But Old Portuguese has *onger (extrapolated from onjudo)/ongir, now ungir, jungir, pungir, and monger/-ir/mugir, now mungir, with Gal. onger/-ir, moger/-ir/mu(n)gir, and Catalan has junyir (plus a form júnyer given by Moll, 59), punyir/púnyer and munyir/múnyer. In Spanish the dialects have innumerable forms, outmatching those of constreñir or reñir: for Cas. unzir/uncir (also yungir, juncir, uñir in Velázquez, with juñir added by Men. M.) the F.J. has yungir/yuncir/yunzir/junner/junnir/yunnir, pres. yunge/yunze/yunne, perf. yungio/yunço/yunno/iunneo/iunnio; Corominas (Dic.) gives Le. unguir, ungan and junir (Alex. O; = juñir, as in P) and a whole spate of modern dialectal forms; Asturian has (earlier) joncer/juñir/juncir (Santander), with present jun go, (earlier) juñe/jonce; Mozarabic has yunnir. MUNGERE and MULGERE are, in part, geographically separated: in the east, Ar. muir/mullé (Val d’Arán), in the west Le., As. 40 41 mocir/(es)mucir/mocer/mecer, also muñir, with muñe/munce. As in the case of the -INGERE verbs, we can find the influence of the perfect, as IŪNXI, IŪNCTUS, corroborated once more by Prov. jonher/junher, onher/unher (Angl., p. 81), with jonch, ponch in the east, junt, punt in the west and Gascony (Bec, pp. 277-8, 534). (Cf. also the unusual Ptg. ponto, ponta, perhaps pointing to an earlier *ponger.) The result is the extension of the u in most areas of the Peninsula, but in this case virtually without any arrhizotonic o-forms — a point of contrast with the front vowels. (However, in Portuguese and Catalan a sense of alternation would be provided because the pretonic pronunciation of o as u made pretonic u here equivalent to an o. The chances of Latin IŬNGO giving a u are uncertain — cf. Sp. hongo, tronco but junco; Ptg. fungo, tronco, junco — but IŬNGES might have done so; see gruño above.) 15. Aduzir, etc. In the second group we have AD-DUCERE, DESTRUERE, SUMERE, which join the PR -IRE verbs *LUCIRE, *PUTIRE, *PRUIRE. Of these, the first two are the most interesting, because of their change of conjugation; sumir was probably attracted into the -ir class later, when the u-i as opposed to the o-e, correspondence was established. For PR we must suppose *AD-DŪGERE, *DESTRŪGERE/*DESTRŬGERE, to account for the forms in all the Romance languages; these would give Proto-Iberian forms *adúire, *destrúire, thence, by way of a change of accent to *aduír(e), *destruír(e), and remodelling on 41 42 the arrhizotonic forms, we arrive at Cat. (adur)/conduir, destruir, OCas. aduzir, destruir, and OPtg. aduzir/-er, destruir. (A shortened form *ADDŪ(GE)RE is the basis of the old future forms (cf. faré, diré etc.) and perhaps Cat. adur, where we would expect *aduire, as coyre; OLe., OPtg. aduzer will have to be secondary, if this theory is correct.) Like dir, Cat. (a)dur has (a)duim/-em in the present. Once again, then, the back vowels form a parallel series to the front vowels, but without alternation. However, this was sometimes provided analogically, as in the case of OSp. adozir beside aduzir, or somir/sumir (somidas, somiese in L.B.A.; cf. destroir/destruir, foir/fuir, roir/ruir, Cat. destro(u)ir (old)/destruir, fogir (old)/fugir, and possibly Ptg. 2 destróis beside destruis, foges as against old fuges, where PR seems to have had forms with both Ŭ and Ū), Gal. lucir/locir/-er. The situation in the back vowels is not as clear as that in the front vowels, but we now have perhaps enough examples both of rhizotonic and of arrhizotonic u to account for the re-formation of the verbs with o. (That certain verbs must originally have had rhizotonic o is shown by the changes of conjugation in Cas. toser, on the one hand, and dial. cusir on the other, and by variants like As. tos/tus(e) (Zam.), sómete/súmete (Cor.); in fact forms like tos, some and recebe may have given rise to -er infinitives (cf. sumería, Mal., 448). Thus we may say that for the back vowels, as for the front, we have been able to indicate certain factors giving the impetus to change, without being able to present a precise picture of the process. And once more we are left with 42 43 the feeling of a kind of conjugational vowel harmony, -ir being associated with u, where -er is with o or ue. However, there is one great difference in modern Spanish between the front and back vowel series. Whereas in the case of the front vowels Old Spanish hesitated between arrhizotonic i and e, the modern language has mostly settled for e before accented i, because the tendency to dissimilation was uppermost; but in the case of the back vowels this principle did not operate, with the result that the u was generalized, thus eliminating vowel alternation, the only verbs to preserve the o (in some parts) being those where it continued to alternate with both ue and u, namely dormir and morir (apart from oir, where the o coming from AU did not diphthongize, even in Aragonese). We may say that the preponderance of rhizotonic u over ue (cf. the situation in the front vowels) gave the feeling that ue was anomalous, and so most of the verbs with ue changed over to u. 16. Portuguese. Before summing up, it may by valuable to tabulate the varieties of metaphony found in Portuguese and Catalan, as we have had occasion to refer to these languages very often. In Portuguese, some -ir verbs came to change their conjugation, giving forms such as morrer (with -rr- from the future morrei, perhaps), sofrer, oferer/oferecer, repender, receber (con-, de-, (a)per-), while the old language or Old 43 44 Galician had -ir forms rather than -er, e.g. sofrirã, Gal. recebir, repintir (but no *morir, only moiro). We have also established the likelihood that escrever, viver, dizer and aduzer were preceded by -ir forms, as found in Gal. Galician has gone further in the matter of conjugation change, producing vester, menter etc. beside vistir, mintir. Of the remaining -ir verbs, those with stem vowel e (apart from some learned words) alternate as servir, sirvo, serves, sirva, sirvamos. These are seguir, sentir, mentir, vestir, ferir, despir, with learned additions like discernir, advertir, aderir, repetir, preferir, conferir, emergir, imergir (OPtg. immerger/amerger). We have already seen that pedir and medir (and vir) keep the old yod forms, which mentir, sentir, servir, ferir and oferir once had (cf. also receber/Gal. recebir, querer, requerer/Gal. requirir, with recebio, (re)queiro). Besides these forms, which give the alternation peço, pedes, meço, medes, there were also formerly forms as pido, pides, pida, pide (imper.). Similarly there were also formerly mento and mintes, sento and sintes, servo and sirves, sega and sigues etc., with imperatives sigue, sirve, minte, sinte, viste, fire, (d)ispe, rime etc. (Nunes, pp. 287, 337-41). Those with stem vowel i throughout are basically the old -INGERE verbs cingir, fingir, re-stringir (con-), tingir, estingir, impingir, atingir, ringir, which we have seen also had -er forms like cinger, renger. Others are r(i)ir, frigir, which now alternates frijo, freges (but Gal. firgir has firgo, firges, as irgo, irges from erguer — *erguir is unrecorded, though Michaelis lists ergir), and learned afligir, 44 45 resistir, dirigir, corrigir, erigir, inquirir, imprimir, transmitir etc. (old forms correger/-ir, ereger/-ir, enquerer/inquerir, empremer/-ir, tra-/tremeter; remir now takes its rhizotonic forms from redimir). Other verbs, all learned, have arrhizotonic e and rhizotonic i throughout (including all forms of the subjunctive), such as agredir, progredir, transgredir, prevenir, denegrir, also serzir/sirgir. They thus follow the pido, pides seen above, and seem like Hispanicisms, except that in Spanish agredir, transgredir show no alternation, prevenir is like venir, and denegrir has only arrhizotonic forms. The old -IRE verbs, it will have been seen, all had stem vowel PR Ĕ, except medir, nor were there any with stem vowel Ī, OBOEDIRE and FINIRE both having passed to the interfix class in PWR. Among the verbs with stem vowel o alternating with u, as dormir, durmo, dormes, durma, durmamos, there are both verbs with PR Ŏ, such as dormir, cobrir, and those with PR Ō, such as subir, bulir, tossir, engolir, cuspir, sa-cudir (a-, re-), surgir, and also some from the Ū (or Ū/Ŭ) class, such as sumir, fugir (formerly also with u throughout), destruir (also with u throughout), and popularly luzir, with others in Galician, like rugir, resurdir, mungir. Cumprir also formerly belonged to this type, with compre, also subj. compra. Other verbs have u in rhizotonic forms and throughout the subjunctive, such as puir/poir and polir, sortir, OPtg. nozir/nuzir, jungir, pungir, ungir, luzir, aduzir. The tendency, as in Spanish, is to have u only, with no alternation. 45 46 It will be seen that both groupings are highly artificial; we have verbs with and without alternation in the singular of the present, whatever their original vowel and the present stem vowel of the infinitive, thus servir, frigir, dormir, subir with alternation, agredir, corrigir, sortir, jungir with i or u throughout. The only verbs to preserve e (now open ę, except before a nasal) throughout the present are those where the yod has affected the consonant, as peço, meço, venho; the rest have changed to the -er conjugation (as receber, repender), as have some o-stems such as morrer, sofrer. We have, then, a much more confused position than is found in Spanish, and one which seems virtually to defy clarification. In spite of the fact that many old yod presents, senço, menço, servio, feiro, ofeiro, requeiro, recebio, dormio, moiro, have been lost, it seems still as if the pattern of 1 i, u against 2, 3, 6 e, o (in which Portuguese differs notably from Spanish), must go back to the yod (which is noticeably better preserved in Portuguese than in Spanish), as we have seen that other forms like subo, cuspa could go back to yod forms. As to the origin of i and u in the modern first persons, we have seen that sinto, minto, sirvo, durmo are late formations, being preceded at an intermediate stage by sento, mento, servo and dormo. This change seems to be similar to that from esso, esto, todo to isso, isto, tudo, generally explained as due to o-metaphony (cf. the equally late u-metaphony in Asturian, e.g. puirtu, timpu from puerto, tiempo). This metaphony did not, however, take place in 46 47 all words; the above three words, for example, are the “neuter” forms, while the corresponding masculine, todo, shows no metaphony (isso, isto correspond to masc. êsse, êste). But in the case of the verbs one must assume a double metaphony: first sẹrvo < sęrvio, as nẹrvo < dọrmo < dǫ̣rmio as fọrça < NERVIU, FORTIA, sobẹrba < SUPERBIA and and then sirvo < sẹrvo, durmo < dọrmo. Sẹnto, mẹnto (with close ẹ before a nasal) would have replaced senço, menço by analogy; they then changed to sinto, minto in the 15th century, presumably on the same analogy. (But note that this did not happen in the -er verbs.) However, nothing is certain, except the final result. In this respect, where Spanish has a vowel harmony contrast between the -er verbs and the -ir verbs, Portuguese has three contrasting types in the three conjugations: the -ar verbs have open ę and ǫ in all positions, whatever their origin, thus lęvo, lęva, espęro, espęra, rǫgo, rǫga, cǫrto, cǫrta, no doubt because of the opening influence of the post-tonic a; the -er verbs occupy a middle position, in which open and close e and o once more coincide to produce a new type of alternation, as vẹrto, vęrtes, vẹrta, mẹto, mętes, mẹta, vọlvo, vǫlves, vọlva, cọrro, cǫrres, cọrra; here vẹrto and vọlvo have close vowels by metaphony, and the subjunctive follows the first person, as in the yod verbs, while mẹta and cọrra have resisted opening for the same reason, whereas mętes and cǫrres follow vęrtes and vǫlves; finally the -ir 47 48 verbs have extended the alternation one degree further, in harmony, as it were, with the infinitive, giving i and u instead of ẹ and ọ while the other persons have open ę and ǫ̣, like the -er verbs (see Will., 176). Thus Spanish and Portuguese have both, by different routes, arrived at a kind of vowel harmony. The verbs with i and u throughout would appear to have followed the lead of verbs like cingir, dizir (Gal.), jungir, aduzir, with the change from old forms like cenges to cinges perhaps helping the process. Or, considering that many of them are learned words, we may see Spanish influence at work, and perhaps also in the alternative forms pido, pides, mido, mides. The alternation between pretonic o and u has no significance, as both are pronounced alike, and many verbs have been written either way, e.g. tossir/tussir, gronhir/grunhir, cospir/cuspir, ordir/urdir, cobrir/cubrir, bolir/bulir, fogir/fugir, sobir/subir, somir/sumir, acodir/acudir, until recent orthographical reforms settled on one or the other. 17. Catalan. In Catalan the position is simpler. The front vowels can be divided into three types: (l) OCat. sent, sents, sentim, senta, sintam, consintrá (Bad. H.), where pretonic i is found except before a stressed i (but seguent; cf. Prov. servén/sirvén); thus we have vistam (Rus.), and also provehiscam, conferiscats (Bad. H.), where the i may, however, represent the extension of the i of -ir etc. to the interfix (cf. partescam, desgarnescam); (2) 48 49 the verbs with yod, ferir, oferir, soferir, requerir (con-), tenir, venir, which have rhizotonic forms such as requir, requers, requira, and arrhizotonic forms based on requer- or requir- (Gar. D.), thus requerim/requirim, recherás/rechirás (Rus., 15:100), conquirats, and so also fira (Moll),7 firam, 3 fer, firén, ferrá (Rus.), ofir, ofira, subj. soffira and soffiram, soferán, sofiren(t) (Rus., Bad. H.), tinc/tenc, tens, tinga/tenga, tingam/tengam, tinent/tenent, tindré/tendré, and similarly vinc/venc, vens, vinga/venga, vingam/vengam, vinent/venent, vindré/vendré, imper. vin(a); these have i for *iei before the original yod, with a tendency to extend it to the pretonic position, except before a following i (requirim is based on an alternative root requir-); in contrast, tenir and venir also have a tendency to replace the i by the e of the other forms; (3) the verbs with a following palatal which diphthongizes the e with the same result, e(i)xir, teixir, llegir, which have ix/ixc, ixes, exim, ixca, ixcam, ixént (Rus., 41: 63) and also infinitive ixir, similarly OCat. tix/tixc, tixes, tixca, and llig/llixch, lliges, subj. lligi, found in Cataluña and the Balearics (cf. OCat. subj. ligen, Rus.), and lisca, OCat. ligiren (Rus.) and ligenda < LEGENDA (Bad. H.); in the same class we have fregir, frig, friges, frigi (with inherited i, also preserved in the same areas), and formerly also regir (reg/rig), afegir (afig/afisch), elegir (eleg/elig), and the whole class was joined by cenyir, tenyir, renyir (*cinc, cinyen), and in part by dir, riure (deim, deia, reim (Bal.), reia); now, however, exir is the only one of these verbs (apart from the last two) unaffected by the 49 50 interfix. The distribution here is seen to be roughly the same as in the Spanish servir class. In the case of the back vowels there seem to be no verbs corresponding to type (l) above, as dormir has the old subjunctive dormam; durmia, however, occurs in the imperfect. In class (2) we have morir, with muyr, mors, morim, muyra, and pretonic mor-/mur-, as morie/muria, and in class (3) collir, with cullo/cullc (for *colc), culls, collim, culli/OCat. colgan, and pretonic coll- as the general rule, but also old cuylí, rechulirá, aculids, acullentz. As we have seen, the other verbs, with the exception of obrir, omplir, and cobrir, complir, these two now with interfix, follow the pattern of collir, some with arrhizotonic o, others with u; but as the two vowels have fallen together in pronunciation, as in Portuguese, there is hesitation in the texts, thus ubertes, umplert, umplit, (des)cubert, descubrirei, fogí, destro(u)ir (though these two may reflect a hesitation between PR Ū and Ŭ), and even rhizotonic umple. Among the -er verbs also, there is alternation in poder, doler, soler, voler, due to the yod. 18. Conclusion. In conclusion we can say that we have established the following points. Firstly, that in contrast with Portuguese there is not much trace of yod in the Spanish -ir verbs. Consequently we find agreement in the stem vowels of persons 1, 2, 3 and 6 in Spanish, whereas in Portuguese 1 is 50 51 generally opposed to 2, 3 and 6. And whatever slight metaphony may have been caused by a yod, in certain verbs only, was not sufficient of itself to establish a pattern of change. Secondly, metaphony of close vowels could, on the other hand, have been caused by the final -Ī of the imperative, as is demonstrated in Portuguese. Thirdly, the tendency to close pretonic e and o to i and u, whether or not before a yod (primary or secondary), is a hesitant one and would need reinforcing by other influences in order to become dominant. The fact that this tendency was resisted by the -er verbs shows that this is so. Fourthly, certain verbs like servir changed rhizotonic ie to i, thus forming a class of verbs with arrhizotonic e before i alternating with i in other positions (also arrhizotonic i before i). This class of verbs was joined by two groups, firstly the -INGERE verbs like ceñir, and then the group consisting of dezir, reir, freir, escrevir, bevir, both of which had e alternating with i in the same positions as the verbs of the servir class, this time because the stem vowel i was dissimilated to e before an accented i. Thus the alternation e/i in the front vowels became the dominant pattern for the -ir verbs. In the case of the back vowels, the same tendencies are observable, that is, a change from ue to u as in cubre for cuebre, and the parallel introduction of -ERE verbs with stem vowel u, such as unzir, mucir, aduzir, but the actual processes of change are not as clear as in the case of the front vowels, and the final 51 52 result is the elimination of alternation, as there are no influences, such as dissimilation, to impede the introduction of u in all positions. All the processes have been finally congealed by a certain sense of conjugational vowel harmony, which requires that the rhizotonic vowel in the -ir conjugation shall be i or u, whereas in the -e conjugation it is ie or e, ue or o. In this sense of vowel harmony Spanish agrees with Portuguese. BIBLIOGRAPHY Spanish, standard language and dialects: (Men. M.) R. Menéndez Pidal, Manual de Gramática Histórica Española, 13th ed., Madrid, 1968. (Oríg.) R. Menéndez Pidal, Orígenes del Español, 3rd ed., Madrid, 1950. (Cid) R. Menéndez Pidal, Cantar del mio Cid, 3rd ed., Madrid, 1954-6. (Ent.) W. J. Entwistle, The Spanish Language, London, 1936. (G. H.) D. J. Gifford & F. W. Hodcroft, Textos Lingüísticos del Medioevo Español, Oxford, 1959. (Hanssen) F. Hanssen, Gramática Histórica de la 52 53 Lengua Castellana, Halle, 1913. (Gar. E.) V. García de Diego, Elementos de Gramática Histórica Castellana, Burgos, 1914. (Gar. G.) V. García de Diego, Gramática Histórica Española, Madrid, 1951. (Gar. D.) V. García de Diego, Manual de Dialectología Española, Madrid, 1967. (Alv.) M. Alvar, El Dialecto Aragonés, Madrid, 1953. (Zam.) A. Zamora Vicente, Dialectología Española, 2nd ed., Madrid, 1967. (Llera, F.J.) V. Fernández Llera, Gramática y Vocabulario del Fuero Juzgo, Madrid, 1929. (L.B.A.) J. Corominas ed., Juan Ruiz, Libro de Buen Amor, Madrid, 1967. (Cor. Dic.) J.Corominas, Diccionario CríticoEtimológico de la Lengua Castellana, Madrid, 1954-7. (Mal.) Y.Malkiel, “Diphthongization, Monophthongization, Metaphony”, in Language, Vol. 42, pp. 430-472, 1966. (Velázquez) M. Velázquez de la Cadena, Spanish and English Dictionary, Chicago, 1967. Portuguese: (Will.) E. B. Williams, From Latin to Portuguese, 53 54 2nd ed., Philadelphia, 1962. (Nunes) J. J. Nunes, Compêndio de Gramática Histórica Portuguêsa, 7th ed., Lisbon, 1969. (Michaelis) H.Michaelis, Portugês e Inglês Dicionário, New York, 1945. (Taylor) J. L. Taylor, A Portuguese-English Dictionary, Stanford, 1958. Catalan: (Bad. G.) A. M. Badía Margarit, Gramática Catalana, Madrid, 1962. (Bad. H.) A. M. Badía Margarit, Gramática Histórica Catalana, Barcelona, 1951. (Moll) F. de B. Moll, Gramática Histórica Catalana, Madrid, 1952. (Rus.) P. Russell-Gebbett, Mediaeval Catalan Linguistic Texts, Oxford, 1965. Provençal: (Angl.) J. Anglade, Grammaire de l’Ancien Provençal, Paris, 1921. (Bec) P. Bec, Manuel Pratique de Philologie Romane, TomeⅠ, Paris, 1970. 54 55 Latin: (Buck) C. D. Buck, Comparative Grammar of Greek and Latin, Chicago, 1933. General: (Bourciez) E.Bourciez, Eĺ éments de Linguistique Romane, 5th ed., Paris, 1967. Where no reference has been given for forms in the Spanish dialects, and also sometimes in Galician and Catalan, they are generally to be found in the relevant sections of García de Diego’s Dialectología. I have also found certain forms in the Oxford Dictionary of English Etymology. ****** This is an edited version of a paper originally published in Ronshu, Vol.12 ( Tokyo, Aoyama Gakuin University, 1971). 55 56 56