1. Introduction
2. Position of the stress
3. The regular paradigms:

3.1 Present indicative
3.2 Present subjunctive
3.3 Imperfect
3.4 Preterite
3.5 Imperfect subjunctive
3.6 Imperative
3.7 Future and conditional

4. Root vocalism (radical changing verbs):

4.1 The -ar and -er conjugations
4.2 The -ir conjugation

5. Root alternations involving consonants

5.1 Decir and hacer
5.2 Verbs like nacer and merecer
5.3 Insertion of /g/ in verbs like poner, salir and caer
5.4 Root-final /ʝ/
5.5 Caber and saber

6. The pretérito grave
7. Gerund
8. Past participle
9. Suppletion and merger

While Spanish has three conjugations, Latin had four, the distinction between the second and third ones depending primarily on what was orginally a contrast in vowel length:

-āre verbs: parāre ‘to prepare’, collocāre ‘to place’, pēnsāre ‘to ponder’ etc.
-ēre verbs: habēre ‘to have’, tĕnēre ‘to hold’, tĭmēre ‘to fear’ etc.
-ĕre verbs: sūmĕre ‘to take’, facĕre ‘to do’, cŭrrĕre ‘to run’ etc.
-īre verbs: audīre ‘to hear’, vĕstīre ‘to dress’, fĕrīre ‘to strike’ etc.

With the loss of vowel length as a distinctive feature in the sound system, the -ēre and -ĕre conjugations became poorly distinguished in speech and they merged in the spoken Latin of central and western Iberia. With the exception of a few individual forms, such as the now obsolete femos ‘we do’ (< facĭmŭs), there is no residue in Spanish (Old or Modern) of those aspects of the -ĕre morphology that distinguished it from the -ēre morphology.

All Latin -āre verbs that survive in modern Spanish have remained in what is now the -ar class. The greater part of the surviving -ēre and -ĕre verbs appear in the modern -er class, although some have migrated to the -ir class. And the bulk of the surviving -īre verbs belong to the -ir class, although a small number migrated to the -er conjugation. Migration between the -ir and -er conjugations is relatively easy, given that the relevant endings differ only in the infinitive, the plural imperative and the first and second persons plural of the present indicative.

Latin also had a number of deponent verbs, which exhibited only the endings associated with the passive voice but which had an active meaning. For example, mŏritur ‘he/she dies’ has the appearance of a passive (compare audītur ‘he/she is heard’), but is not passive in meaning; nor does it have a corresponding ‘active’ form, as audītur, for example, does, viz. audit ‘he/she hears’. The Latin deponent verbs were either lost or remodelled with active endings, surfacing in Spanish mainly in the -ir class. For example, the passive-looking ending -ī in the infinitival form mŏrī ‘to die’ was replaced by the active ending -īre, producing the analogical form *mŏrīre, from which modern Spanish morir descends, and the passive-looking ending -itur in the third person singular present tense form mŏritur ‘he/she dies’ was replaced by the equivalent active ending -it, giving the analogical form *mŏrit, from which modern muere descends (the diphthong in the root reflecting diphthongization of Latin ŏ in stressed syllabus).

2. Position of the stress
Spanish maintains the stress pattern of the Latin verb paradigms, except in the imperfect (indicative) and what are now the two subparadigms of the imperfect subjunctive. In each of these, the stress in the first and second persons plural has been retracted one syllable, by analogy with all the other forms in the subparadigm. This change is illustrated in Table 1 below, in which the stressed syllable in each verb form is shown in bold.

Some stress reassignment also occured in those verbs that belonged to the -ĕre conjugation, as part of its merger with the -ēre conjugation. While the latter verb class, like the verb conjugations of modern Spanish, exhibited stressed endings in the first and second persons plural of the present indicative, as well as in the infinitive and the plural imperative, the -ĕre conjugation stressed these forms on the verbal root. This was because the relevant syllable in the -ĕre conjugation was short, requiring the stress to be retracted from its usual position on the penultimate syllable to the antepenult. The absorption of the -ĕre conjugation into the -ēre one thus implies stress shift from the root to the ending in the following forms (bold font marks the stressed syllable):

1. The first person plural present indicative, e.g. capĭmus > cabemos
2. The second person plural present indicative e.g. capĭtis > cabéis
3. The infinitive, e.g. capĕre > caber ‘to fit’
4. The plural imperative, e.g. capĭte > cabed

In general, however, with the loss of the intertonic vowel, verb forms that contained a short penult suffered a reduction in syllable count, unless the penult also happened to be the initial syllable or its vowel was /a/. Predictable reduction of this type can be seen in forms such as the following (which, in terms of the modern language, belong to the verbs colgar ‘to hang’, vengar ‘to avenge’ and recobrar ‘to recover’ respectively):

cŏllŏcō, cŏllŏcās etc. > cuelgo, cuelgas
vĭndĭcō, vĭndĭcās etc. > vengo, vengas
rĕcŭpĕrō, rĕcŭpĕrās etc. > recobro, recobras

On the other hand, syncope was sometimes blocked if the verbal root was morphologically complex, the antepenult actually being a prefix and the short penult the lexical root. In this type of case, speakers often moved the stress rightwards from the prefix to the penult. For example, forms like renŏvō ‘I renew’ and renŏvās ‘you renew’, with stress on the prefix re-, have reflexes in modern Spanish that are stressed on the penult, viz. renuevo, renuevas etc., implying that in spoken Latin the latter syllable acquired the stress hitherto applied to the prefix, its vowel thereby escaping syncope (and dipthongizing like any other stressed ŏ).

Finally, it is worth noting that some verbs had variable stress across their rhizotonic forms, i.e. those in which the stress fell on the verbal root rather than on the ending. This variation was due to the fact that, in the relevant verbs, certain forms of the present tense exhibited a root increment, i or e, which in Classical Latin created an extra syllable, enabling a short syllable that would otherwise also be a penult to bear the stress: a-pĕ-ri-ō ‘I open’, re-cĭ-pi-ō ‘I take back’ etc. In the rhizotonic forms that lacked the root increment, the short syllable was the penult and hence the stress fell on the antepenult: a-pĕ-rīs, re-cĭ-pĭs etc. With the loss of the hiatus, this difference in syllable count was erased and, as a by-product, the stress was equalized across the various rhizotonic forms. In some cases, such equalization was achieved by retracting the stress by one syllable in the forms that contained the root increment, as happened with the verb abrir ‘to open’:

apĕriō > abro
apĕrīs > abres
apĕrĭt > abre etc.

In other cases, the converse occurred. That is to say, the stress was advanced one syllable in those forms that lacked the root increment. This pattern of development can be seen in the verb recibir ‘to receive’:

recĭpiō > recibo
recĭpĭs > recibes
recĭpĭt > recibe etc.

In both types of case, however, the effect of the stress shift was the same, viz. elimination of the variable stress that had previously existed among the rhizotonic forms of the relevant verbs.

3.1 Present indicative
The present indicative endings derive from their Latin counterparts largely through regular sound change (merger of /e/ and /i/ in final syllables, loss of final /t/ etc.), as is evident from Table 2 below. Note that this table does not show the root increment, i or e, that occurred in certain forms of the -īre and -ēre conjugations (as well as in some individual -ĕre verbs). This latter item was lost before the Old Spanish period and has no effect on the verb endings themselves. Its effect on the verbal root has been profound, however, and this is discussed in Section 4.2 and Section 5.

The one cell in the above table in which the Spanish form is not strictly etymological is that of the third person plural of the -ir conjugation, viz. -en. Obviously enough, this item cannot be a reflex of the equivalent ending in the Classical Latin -īre conjugation, which was -ŭnt. Instead, it must be assumed to descend from -ĕnt, which must have been borrowed into the -īre verb class from the -ēre paradigm in late spoken Latin (the -ŭnt ending in the -ĕre paradigm must also have been displaced by -ĕnt).

It can also be seen from Table 2 that the endings of the second person plural have been reduced from two syllables to one. This occurred after intervocalic /t/ was voiced during the lenition phase, resulting in the Old Spanish reflexes of -ātis, -ētis/-ĭtĭs and -ītis being -ades, -edes and -ides. The /-d-/ in these endings was subsequently lost (roughly in the fifteenth century) and the resulting hiatus was eliminated, entailing a reduction in syllable count.

The second person plural forms in the -ar and -er conjugations reflect dissimilation, in the sense that the post-hiatus vowel /e/ was modified to [j], which was phonetically less similar than /e/ to the pre-hiatus vowel, i.e. /a/ in the -ar conjugation and /e/ in the -er one. In the -ir conjugation, however, the second person plural form of the present indicative reflects assimilation, in the sense that the post-hiatus vowel /e/ coalesced with pre-hiatus vowel /i/.

Assimilated second person plural forms were in fact also available for -ar and -er verbs and, throughout the later Middle Ages and the early modern period, they co-existed with the corresponding dissimilated forms shown in Table 2. The various competing forms are illustrated in Table 3 below.

Except in the -ir conjugation, where such endings were the only ones available, the assimilated endings were abandoned in the Peninsula by the mid-sixteenth century, but they survive in many parts of Latin America. There, about 40% of speakers exhibit voseo, which is the use of vos as a familiar second person singular pronoun. Vos (a cognate of French vous) was originally a plural item and it continues to select verb forms that descend from old plural forms. These are often a continuation of the old assimilated second person plural forms – as in vos hablás and vos comés – but in certain regions (Zulia state in Venezuela being a case in point) vos goes with the corresponding dissimilated forms, in the -ar conjugation at least.

3.2 Present subjunctive
Exactly like modern Spanish, Latin generated its present subjunctive forms from the corresponding indicative ones primarily through a process of ‘conjugation vowel reversal’, the stem vowel /a/ of the -āre indicative becoming /e/ and the stem vowel /e/ or /i/ of the indicative in the three other verb classes becoming /a/. This is illustrated below:

pēnsat ‘he/she ponders’ [indicative] → pēnset ‘he/she ponders’ [subjunctive]
dīcit ‘he/she says’ [indicative] → dīcat ‘he/she says’ [subjunctive]

pēnsō ‘I ponder’ [indicative] → pēnsem ‘I ponder’ [subjunctive]
dīcō ‘I say’ [indicative] → dīcam ‘I say’ [subjunctive]

3.3 The Imperfect
The imperfect endings were characterized in Latin by the presence of the formative /-ba-/, which was inserted between the stem vowel and the person marker. This is illustrated below, using the second person singular forms of the verbs clāmāre ‘to shout’, tĭmēre ‘to fear’, and vĕnīre ‘to come’:

clām-ā-bā-s ‘you shouted’
tĭm-ē-bā-s ‘you feared’
vĕn-iē-bā-s ‘you were coming’

As a marker of imperfectivity, /-ba-/ survives intact in modern Spanish in the -ar conjugation, but it has been reduced to just /-a-/ in the -er and -ir conjugations, where, in addition, the stem vowel has been equalized to /i/:

clām-ā-bā-s > llam-a-ba-s
tĭm-ē-bā-s > tem-í-a-s
vĕn-iē-bā-s > ven-í-a-s

In the -er and -ir conjugations of Old Spanish, the sequence /-ia-/, comprising the merged stem vowel and the residue of the imperfective formative /-ba-/, was for a long time pronounced as /-ie-/, due to assimilation between the /a/ and the /i/, although rarely in the first person singular. This pronunciation is reflected in pre-fifteenth century spellings such as tenies, tenien, vivies, vivien etc.

This /-ie-/ must often have been realized as [-je-], given that Old Spanish imperfect forms in -ie- commonly show raising of the root vowel, suggesting metaphony induced by [j]. This can be seen in examples such as the following:

durmies (later: dormías ‘you were sleeping’)
sirvien (later: servían ‘they served’)
murien (later: morían ‘they were dying’)

By the fifteenth century, imperfect forms in -ie- had largely disappeared, as had the associated tendency for the root vowel to become raised.

3.4 The Preterite
In modern Spanish, the regular preterite has two sets of endings, one for -ar verbs and one for -er and -ir verbs. The -ar endings are a continuation of the corresponding endings in the Latin -āre conjugation, while the -er/-ir endings can be traced back to the endings associated with the -īre conjugation. As regards the preterite or ‘perfect’ of the Latin -ēre and -ĕre conjugations, the only traces of this in modern Spanish are confined to the small class of irregular verb forms known as the pretérito grave; for which, see Section 6 below. Thus the Latin endings that need to be considered at this point are just those of the -āre and -īre paradigms which, in their Classical form, are illustrated in Table 4 below:

The endings shown above were contracted in spoken Latin, usually through elision of the perfective marker -v-, which occurred between the stem vowel and the remainder of the ending. In the first person singular in both conjugations, loss of -v- yields -āī and -īī respectively. The latter was probably pronounced [i] as soon as the element -v- was lost and continues into modern Spanish as -í, e.g. in dormí ‘I slept’. The former, pronounced [aj], reduced naturally to /e/ (compare bāsium > [bajso] > beso ‘kiss’), whence the modern ending -é, as in hablé ‘I spoke’.

In contrast to what happened in the first person singular, the consonantal element -v- appears to have been retained in the third person singular, and the contraction consisted in the elision of the vowel ĭ that came immediately after it. The endings -āvĭt and -īvĭt were thus reduced to [-awt] and [-iwt] respectively.

In the -āre ending, [aw] underwent its normal reduction to /o/, producing the ending -ot which is found in early Old Spanish. In the -īre ending, the semivowel [w] became syllabic, i.e. [ˈiwt] > [ˈiut]. As happened generally, /u/ in final syllables was replaced by /o/, resulting in the early Old Spanish ending -iot, initially pronounced [ˈiot] but over time evolving to [ˈjot]. The final /t/ of both -ot and -iot went the way of Latin final /t/ generally, not surviving much beyond the eleventh century. Following this latter development, the third person singular endings were as they are in modern Spanish, namely -ó and -ió (although the practice of indicating the stress with a written accent dates from the early modern period).

As regards the other forms (second person singular and all of the plural forms), the transition to Spanish can be explained as follows. In the -āre class, once the consonantal element -v- was lost, assimilation appears to have occurred, the stem vowel /a/ absorbing the now adjacent vowel /e/ (= ĭ or ē). On the other hand, in the -īre class, loss of -v- appears to have produced competing assimilated and dissimilated forms, the latter resulting from semivocalization of the stem vowel /i/. Both types of outcome are shown in Table 5 below.

In the third person plural, the dissimilated ending [-jeɾon(t)] (> modern ­ieron) appears to have ousted [-iɾon(t)] before Spanish came to be written down, but competition between assimilated and dissimilated endings in the second person (singular and plural) as well as in the first person plural continued into the High Middle Ages. Thus, as late as the thirteenth century, forms with dissimilated endings, such as dixiemos ‘we said’ and recibiestes ‘you [plural] received’, co-existed with forms that had assimilated endings, such as sobimos ‘we climbed’ and salistes ‘you [plural] left’. Eventually, however, the dissimilated endings fell into obsolescence.

The diphthong -ei- (phonetically [ej]) of the modern second person plural endings -asteis and -isteis is a seventeenth-century innovation, clearly inspired by identical or similar diphthongs in the second person plural in many of the other tenses (coméis, cantabais, durmierais etc.).

3.5 Imperfect subjunctive
While the term ‘imperfect subjunctive’ appears to be borrowed from Latin grammar writing, the Latin imperfect subjunctive does not itself survive in Spanish. Rather, the two subparadigms of the modern imperfect subjunctive, in -ra- and -se- respectively, descend from the indicative and the subjunctive of the Latin pluperfect. The latter were members of the perfectum group of subparadigms, meaning that they exhibited the perfective marker -v- that characterized the perfect tense, as well as inheriting any root irregularities that existed in the perfect of individual verbs.

As in the preterite, the -er and -ir conjugations have merged in the imperfect subjunctive, the shared modern endings being most straightforwardly derived from those of the Latin -īre verb class. The corresponding endings in the -ar conjugation are transparently descended from those of the -āre class.

Looking first at the -ar conjugation, both the original Latin forms and their reflexes in modern Spanish are illustrated in the table below (the stressed syllable is indicated in bold font).

The above Latin forms appear to have been regularly contracted in speech, through elision of the perfective marker -v-, together with its following vowel /e/ (= ĕ or ĭ). This produces in each case a form that evolves regularly into the modern form through one or more of the following processes: simplification of intervocalic /ss/ to /s/, loss of final /t/ and lenition of intervocalic /t/ followed by elimination of the resultant hiatus. (For the retraction of the stress in the first and second persons plural, see Section 2.)

Analogous remarks apply to the corresponding forms in the -īre verb class, shown in Table 7 below, although in this case retention must be posited of the /e/ that immediately followed the -v- element (but see Penny 2002: 201–204 for a different approach). In the rhizotonic forms, the stress must then have been transferred from the stem vowel ī to the aforementioned /e/, creating the diphthong [je] that characterizes the imperfect subjunctive of the modern -ir conjugation. It is the yod in this diphthong that accounts for the high root vowel found in the imperfect subjunctive in verbs such as dormir, morir, medir, sentir etc. The same [j] would arise naturally in the first and second persons plural, given that in these forms the stem vowel ī was unstressed – for the retraction of the stress in these forms, see Section 2.

3.6 Imperative
The Spanish imperative endings derive straightforwardly from the Latin present imperative, which in the singular had forms in -ā, -ē/ĕ, and -ī and in the plural had forms in -āte, -ēte/-ĭte and -īte. Of the singular endings, only -ā and -ē/ĕ (= /a/ and /e/) survive, given the vowel mergers that took place generally in final syllables. We thus have e.g. entra ‘¡enter!’ (< intrā) and bebe ‘¡drink!’ (< bĭbĕ), but sube ‘¡go up!’ (< subī) rather than *subi.

Final /-e/ was often lost in Old Spanish after intervocalic dental and alveolar consonants. This development affected the plural imperative generally (e.g. cantāte > cantade > cantad ‘sing’) and also the singular imperative forms of salir ‘to go out’, poner ‘to put’, venir ‘to come’, tener ‘to have’, and hacer ‘to do’, which in modern Spanish are ‘irregular’ for that very reason. The forms in question are sal, pon, ven, ten and haz. The first three of these are etymological, descending from salī, pōnĕ and vĕnī respectively, final /i/ in the latter form preventing diphthongization of the root vowel ĕ. In contrast, the root vowel in ten (< tĕnē) is analogical, given that, in the absence of any trigger for metaphony, stressed ĕ would be expected to yield [je], while haz comes not from fac, the irregular singular imperative of Latin facĕre (> Spanish hacer), but from the Old Spanish form faze, which reflects remodelling of the imperative through analogy with present indicative forms such as fazes ‘you do’, faze ‘he does’ and fazen ‘they do’.

On the other hand, Spanish di ‘¡say!’ (sing.) is derived straightforwardly from the irregular Latin form dīc, given that /k/ was not among the small number of Latin consonants that survive in word-final position.

3.7 Future and conditional
In Spanish, as throughout most of western Romance, the future and conditional evolved through agglutination of contracted present or imperfect forms of habēre ‘to have’ and a preceding infinitive. This structure appears to have been available already in late Latin, although there habēre retains its status as a full auxiliary verb. An oft-cited example is the one below from St. Augustine:

tempestas illa tollere habet totam paleam (In evang. Joh. 4.1.2)
‘this storm will carry off all the chaff’

The development of the future and conditional subparadigms in Spanish is schematized in the table below. In the conditional, the inherited ē-based of habēre were replaced endings with /i/-based ones, as described in 3.3 above.

In the new agglutinated forms, the conjugation vowel of the erstwhile infinitive (i.e. /-a-/, /-e-/ or /-i-/) occupied an internal unstressed position and so, with the exception of /a/, was subject to elimination. However, of the resulting syncopated forms, e.g. vivran (now vivirán), movrien (now moverían) and luzra (now lucirá), only a limited number have survived. Of these, some have undergone no additional adjustment, e.g. habré, cabrá, querré, while others show epenthesis, e.g. pondré, tendrá, valdría, saldré. The modern form dirá ‘will say’ is based on the Old Sp. infinitive dizir, rather than modern decir, and reflects elimination of the root-final -z- in the infinitival base.

Like other Old Romance varieties, Old Spanish allowed clitic pronous to be inserted into future and conditional verb forms between the erstwhile infinitive and the contracted form of habēre, the latter reviving in this construction its original role as an auxiliary verb. The two examples below are from Poema de Mio Cid (lines 161 and 1808).

dar le ien .vi. çientos marcos
= le darían seiscientos marcos

ir vos hedes sin falla
= os iréis sin falla

Split futures and conditionals were at their most productive at the beginning of the Old Spanish period. They gradually declined in frequency before dying out completely in the 17th century. Rivero (1993) analyses this construction as an instance of so-called Long Head Movement, whereby the main verb (in this case the infinitive) moves leftwards across the ‘clitic + auxiliary’ complex to a higher position in clause structure. According to Rivero, the cause of this movement is the need to prevent the clitic occupying clause-initial position, where it would violate the Tobler–Mussafia Law.

4. Root vocalism (radical-changing verbs)

4.1 The -ar and -er conjugations
Among the -ar and -er verbs, the modern e ~ ie and o ~ ue alternations that are exhibited in the present tense directly reflect (in principle at least) the diphthongization of Vulgar Latin /ε/ and /ɔ/. The latter sounds, because they occurred only in stressed syllables, appeared as root vowels only in the rhizotonic forms of the paradigm. This situation is reflected in modern Spanish in the fact that the diphthongs ie and ue occur in the rhizotonic forms of the present tense, whereas the corresponding monophthongs e and o occur in the forms in which the stress falls on the ending. This is illustrated below:

siego ‘I reap’ (< sĕcō [ˈsεko]), cuelgas ‘you hang’ (< cŏll(ŏ)cās [ˈkɔll(o)kas])
versus:
segamos ‘we reap’ (< sĕcāmus [seˈkamos]), colgamos ‘we hang’ (< cŏll(ŏ)cāmus [koll(o)ˈkamos]

However, in quite a few radical-changing -ar and -er verbs, the root alternations are actually due to analogical pressure. For example, sēm(i)nō and cōlat, with /e/ and /o/ as the stem vowel, would normally have evolved to monophthongal forms, rather than siembro ‘I sow’ and cuela ‘he/she strains’, which is what we actually find in modern Spanish.

On the other hand, analogical pressure has in some cases worked the other way, causing monophthongization of previously diphthongized forms. Thus some verbs with unalterable root vowels in modern Spanish, such as entregar ‘to hand over’ and prestar ‘to lend’, were root-alternating in Old Spanish: entriego, priesto etc.

4.2 The -ir conjugation
In modern Spanish, the root vowel alternations in the -ir verb class can be grouped into two main patterns or paradigms, exemplifiable by medir ‘to measure’ and mentir ‘to lie’ respectively. In the first case, the root vowel /e/ raises to /i/ in (i) all forms of the present tense (indicative and subjunctive) except the first and second persons plural in the indicative, (ii) the third person (singular and plural) of the preterite, (iii) all forms of the imperfect subjunctive, (iv) the singular imperative and (v) the gerund. This is shown in Table 9 below.

The root increment [j] (inflectional yod)
All verbs in the Latin -īre and -ēre conjugations (as well as some in the -ĕre class) displayed a root increment, which took the form of an unstressed front vowel (either -ĕ- or -ĭ-) inserted between the verbal root and the ending proper, in the first person singular of the present indicative and in all the forms of the present subjunctive, as illustrated below:

dēb-ĕ-ō ‘I must’ (indicative)
aud-ĭ-ō ‘I hear’ (indicative)
dēb-ĕ-am ‘I must’ (subjunctive)
aud-ĭ-am ‘I hear’ (subjunctive)
dēb-ĕ-ās ‘you must’ (subjunctive)
aud-ĭ-ās ‘you hear’ (subjunctive)
etc.

ceñir ‘to adjust/encircle/hug’ (< cĭngĕre: cĭngō, cĭngam etc.)
teñir ‘to dye’ (< tĭngĕre: tĭngō, tĭngam etc.)
pedir ‘to ask for’ (< pĕtĕre: pĕtō, pĕtam etc.)

VL mĕntiāmus [menˈtjamos] > mintamos
VL mĕntiātis [menˈtjates] > mintáis

The absence of /i/ from the root of the other present tense forms is a consequence of the fact that the original root vowel was short ĕ rather than the long ē found in verbs like mētīre. This meant that verbs like mentir originally exhibited a root alternation between /ɛ/ and /e/ depending on whether the root was stressed or not:

VL mĕntiō [ˈmɛntjo] ‘I lie’
VL mĕntīs [ˈmɛntis] ‘you lie’ etc.
versus:
VL mĕntīmus [menˈtimos] ‘we lie’
VL mĕntiāmus [menˈtjamos] ‘we lie (subjunctive)’ etc.

In principle, this alternation survives into modern Spanish in the form of a contrast between the diphthong [je] and a pure vowel, which is either /e/ or /i/ depending on whether the relevant verb form was subject to metaphony. In the modern language, then, the etymologically regular forms contrast in the following way:

mientes, miente, mienten (diphthing)
versus:
mentimos, mentís, mintamos, mintáis (pure vowel)

On the other hand, the diphthongized root in †miento (first person singular present indicative) and in all the rhizotonic forms of the present subjunctive (†mienta, †mientas etc.) is analogical. In these forms, the inflectional yod would have raised the original root vowel /ɛ/ to /e/, thereby blocking diphthongization. What has happened, therefore, is that the root diphthong of the forms mientes, miente and mienten has spread to all the other rhizotonic forms, thus equalizing their root vocalism. In this respect, the mentir type of verb ends up converging with the medir type, in the sense that both types equalize their root in the rhizotonic forms, the difference being that medir-type verbs achieve this with the pure vowel /i/ whereas mentir-type ones achieve it with the diphthong [je].

A number of verbs developed along identical lines to mentir, including herir ‘to wound’ (< fĕrīre), sentir ‘to feel’ (< sĕntīre) and hervir ‘to boil’ (< fĕrvĕre), the latter also undergoing migration from the -ĕre conjugation to the -īre one. However, many verbs that could have followed the same diachronic trajectory as mentir have come to exhibit the medir type of root vocalism. This could be due to analogical attraction or it could reflect spontaneous reduction of an earlier diphthong [je] to /i/, as occurred in a number of nouns, e.g. castillo ‘castle’ < Old Spanish castiello < Latin castĕllum. Verbs in this latter category include the following:

vestir ‘to dress/wear’ (< vĕstīre)
seguir ‘to follow’ (< deponent sĕquī)
servir ‘to serve’ (< sĕrvīre)

Verbs like subir
For verbs in the -ir class that have a back root vowel, the equivalent of the historical /e/ > /i/ sound change was /o/ > /u/. However, in almost all cases the resulting /o/ ~ /u/ alternations have been levelled to /u/, implying that the high vowel which arose through regular sound change in forms such as the first person singular of the present indicative has spread analogically throughout the entire verbal paradigm. The verb subir ‘to go up’ (< sŭbīre) is an exemplar of this pattern. Thus, alongside historically regular forms such as subo ‘I go up’ (< sŭbeō [ˈsoβjo]) and suba (< sŭbeam [ˈsoβja]), we find not just †subes, †sube and †suben (which are direct analogues of †mides, †mide and †miden) but also †subimos, †subís, †subir, †subí, †subiste, †subid and so on.

The unimpeded nature of the analogical spread of the high root vowel among verbs like subir appears to reflect the fact that /u/ is unlikely to dissimilate with respect to /i/; that is to say, the factor which militated against forms such as midir and midimos (for medir and medimos) was absent among verbs in which the root vowel was a back one. It thus transpires that there is no modern equivalent to the medir pattern among verbs that have a back root vowel. Instead, verbs that could in principle have an /o/ ~ /u/ alternation analogous to the /e/ ~ /i/ alternation in medir now exhibit uniform root vocalism in /u/, which disguises the fact that in (later) Latin their root vowel was /o/.

Furthermore, the pattern involving uniform root vocalism in /u/ has exercised a powerful effect of inter-paradigmatic analogy, in that it has attracted all but two of the -ir verbs whose root vowel was originally ŏ (= /ɔ/ alternating with /o/). These latter verbs, a selection of which is given immediately below, should in principle have ended up with the same root vocalism as dormir and morir, which are discussed in the next subsection. Instead, however, they follow the subir pattern:

excŏnspuĕre > †escupir ‘to spit’ (†escupe, †escupimos etc.)
abhŏrrēre > †aburrir ‘to bore’ (†aburre, †aburrimos etc.)
cŏŏp(e)rīre > †cubrir ‘to cover’ (†cubre, †cubrimos etc.)
cŏmplēre > †cumplir ‘to fulfil’ (†cumple, †cumplimos etc.)

Dormir and morir
The only two verbs with roots in -ŏ- which escaped being attracted to the pattern illustrated by subir are dormir ‘to sleep’ and morir ‘to die’, whose root vocalism is shown in Table 11 above. These verbs exactly parallel mentir, both in their modern pattern of root vocalism and in their historical development. Thus in later Latin, when ŏ was only distinguished from ō in stressed syllables, these two verbs must have exhibited a root alternation between /ɔ/ and /o/, according as the stress fell on the root or not: dŏrmiō [ˈdɔɾmjo] ‘I sleep’, dŏrmīs [ˈdɔɾmis] ‘you sleep’, but dŏrmīmus [doɾˈmimos] ‘we sleep’, dŏrmiāmus [doɾˈmjamos] ‘we sleep (subj.)’ etc. The vowel /ɔ/ subsequently diphthongized to [we] – e.g. dŏrmīs [ˈdɔɾmis] > duermes, dŏrmit [ˈdɔɾmet] > duerme – and the root vowel /o/ was raised to /u/ in those forms that contained the root increment [j], e.g. dŏrmiāmus [doɾˈmjamus] > durmamos. However, modelled on duermes, duerme and duermen, we also find analogical †duermo, †duerma, †duermas, †duerman, where the root increment [j] would normally have raised the root vowel to /o/, forestalling any diphthongization.

The verbs abhŏrrēre ‘to abhor’, cŏŏp(e)rīre ‘to cover’ and cŏmplēre ‘to fill up/satisfy’ could in principle have followed the same historical trajectory as dŏrmīre and VL mŏrīre (for Classical Latin mŏrī). They were, however, attracted to the pattern represented by subir, with uniform root vocalism in /u/, yielding the modern Spanish verbs aburrir, cubrir and cumplir.

The verb venir
Among the -ir verbs, the root vocalism of venir ‘to come’ (< vĕnīre) is sui generis. Given the phonetic form of its Latin etymon, it could in principle have developed in a similar way to mentir, in the present tense at least. However, in that tense it is actually like tener ‘to have’ (< tĕnēre), in the sense that the root vowel /ε/ (corresponding to Latin ĕ) in the rhizotonic forms raised predictably under the influence of [j] (e.g. vĕniō > vengo, vĕniam > venga rather than *viengo, *vienga) but the /e/ corresponding to ĕ in the ending-stressed forms did not (e.g. vĕniāmus > vengamos rather than *vingamos), which is exactly the reverse of what happened with mentir and verbs like it. As in the mentir pattern, however, /ε/ diphthongized (as expected) in those rhizotonic forms that lacked the root increment [j], namely the second and third persons singular and the third person plural of the present indicative: vĕnīs > vienes, vĕnit > viene, vĕnent (for CL vĕniunt) > vienen.

It should also be noted that the high root vowel /i/ found in the preterite forms – vine ‘I came’, viniste ‘you came’, vino ‘he came’ etc. – is not due to the action of [j] at all.

Freír, reír and decir
Finally, a number of additional -ir verbs with roots originally in /i/ were attracted to the medir pattern. For example, †freír ‘to fry’, †reír ‘to laugh’ exhibit the medir-type root vowel alternation between /e/ and /i/ but descend from verbs which had root vowel /i/ throughout their paradigms, viz. frīgĕre and rīdēre. The verb dīcĕre ‘to say’ was similarly adjusted, but preserves its etymological /i/ throughout its preterite.

5.1. Decir and hacer
The alternation between /-θ-/ and /-g-/ exhibited by the roots of decir ‘to say’ (< dīcĕre) and hacer ‘to do’ (< facĕre) results directly from the bifurcating evolution of Latin /k/, which, in general, evolved over time to /θ/ when followed by a front vowel but survived as a velar when followed by a non-front vowel, albeit voiced to /g/ if intervocalic. Thus dices, dice, decimos etc. result from forms in which the velar preceded a front vowel, viz. dīcis, dīcit, dīcimus, whereas digo, diga etc. result from forms in which the velar preceded a non-front vowel, viz. dīcō, dīcam etc.

The explanation for hago, haga, hagas etc. versus haces, hace, hacemes etc. is essentially the same, except that we have to disregard the [j] which came to appear immediately after the root-final /k/ in the Latin etyma of the forms that now exhibit /g/, viz. facĭō, facĭam, facĭās etc. The originally full vowel ĭ which appears immediately before the ending in these latter forms came, with the generalized loss of the hiatus, to be pronounced as the palatal semivowel [j]. In principle, this latter sound has the effect of palatalizing a preceding /k/, producing /ts/ or /dz/ in Old Spanish and ultimately /θ/ in the modern language. However, the [j] in verbs forms like facĭō, facĭam, facĭās is an instance of the so-called inflectional [j] (la yod flexional), which seems to have been treated by speakers in a partially different way from the [j] that occurred in other parts of speech, such as nouns and adjectives. As is noted by Menéndez Pidal (1987:292–3), the inflectional [j] had no effect on the final consonant of the verbal root, except when that consonant was /b/, /d/ or /g/, although in all cases it disappeared. In the case of forms like hago, haga and hagas, the original root-final consonant was /k/, which, in conformity with Menéndez Pidal’s generalization, was unaffected by the following [j]. Portuguese provides an interesting contrast in this regard, as the forms corresponding to Spanish hago, hagas, haga etc., i.e. faço, faça, faças, do reflect palatalization of root-final /k/ by [j].

A further relevant contrast is with verbs forms like capiō (> quepo ‘I fit’) and sapiat (> sepa ‘he/she knows [pres. subj.]’), where the root-final consonant and the inflectional [j] underwent metathesis (i.e. transposition), causing the root vowel /a/ to raise to /e/. Self-evidently, this did not occur in forms like faciō, faciam and faciās, the [j] in question leaving no trace in either Old Spanish or Modern Spanish. The fact that metathesis did not occur in the facĕre paradigm is not necessarily surprising, however, given that metathesis is by its nature subject to a degree of randomness in terms of its occurrence.

5.2. Verbs like nacer and merecer
Quite a few Spanish verbs alternate in the present tense between roots in /-θ-/ and roots in /-θk-/, the latter occurring in the indicative first person singular and throughout the subjunctive. A useful exemplar is the verb nacer ‘to be born’:

Present indicative: nazco, naces, nace, nacemos, nacéis, nacen
Present subjunctive: nazca, nazcas, nazca, nazcamos, nazcáis, nazcan

This alternation has a similar origin to that of the /-θ-/ ~ /-g-/ alternation discussed in 5.1, in the sense that it arose as a consequence of the way Latin /k/ had a different evolution depending on whether it preceded a front vowel or a non-front vowel. In those verbs in which the /-θ-/ ~ /-θk-/ alternation is etymological, or at least not completely analogical, it goes back to the Latin inchoative infix -sc- which appeared throughout the finite paradigms of a number of third conjugation verbs, i.e. those with active infinitives in -ĕre. The Classical Latin ancestor of Spanish nacer had this infix and was a member of the third conjugation, although it was actually a deponent verb, the (present) infinitive being nāscī rather than *nāscĕre. However, with the loss of the deponent pattern of inflection, it can be assumed that nāscī was remodelled in Vulgar Latin as nāscĕre, a verb with active endings, as illustrated below:

VL nāscĕre for deponent nāscī
VL nāscō for deponent nāscor
VL nāscis for deponent nāscĕris
VL nāscam for deponent nāscar
VL nāscās for deponent nāscāris

From the phonological point of view, the -sc- infix exhibited by nāscī/nāscĕre and similar verbs corresponds to the cluster /sk/. As can be seen from the way in which nominal forms such as Latin pĭscēs ‘fish’ developed, this cluster has /ts/ and then /θ/ as its reflex when it precedes a front vowel. Conversely, it usually survives unmodified before a non-front vowel, as can be seen from words like tosco ‘coarse/vulgar’ (< VL tŭscum ‘Etruscan’) and mosca ‘fly’ (< mŭscam). Given this dual evolution, verbs that had the -sc- infix in Latin show a root alternation in Old Spanish between /-ts-/ and /-sk-/, the former appearing before front vowels and the latter before /o/ and /a/:

nāscĕre > Old Sp. naçer [naˈtseɾ]
nāscis > Old Sp. naçes [ˈnatses]
nāscō > Old Sp. nasco [ˈnasko]
nāscam > Old Sp. nasca [ˈnaska]

Old Spanish /ts/ evolved regularly to /θ/ and, over the course of the thirteenth and fourteenth centuries, the /s/ in the unchanged /sk/ cluster was analogically adjusted under the influence of the forms in which /sk/ had evolved to /ts/ or, later, /θ/. These two changes result in the creation of the familiar modern alternation between /-θ-/ and /-θk-/, as in e.g. nacen versus nazcan (‘they are born’ indicative and subjunctive).

The /-θ-/ ~ /-θk-/ alternation in the verbs merecer ‘to merit/deserve’ (< VL merēscĕre for CL merēre), florecer ‘to flower’ (< flōrēscĕre) and conocer ‘to know’ (< cŏgnōscĕre) developed exactly as it did in nacer.

In contrast, in Spanish verbs ending in -ucir (e.g. lucir ‘to shine’, conducir ‘to lead/drive’, aducir ‘to adduce’ etc.) the alternation is entirely analogical. The Classical Latin ancestors of these verbs – lūcēre, condūcĕre, addūcĕre etc. – lacked the -sc- infix altogether and they do not appear to have been reassigned to the -scĕre class in Vulgar Latin, as were verbs like merēre, mentioned just above. Instead, they have reflexes in Old Spanish that exhibit an alternation between g and z (= /dz/), which is exactly what would be expected given their phonetic form in Classical Latin. The verb aduzir (aducir in modern Spanish), for example, is well attested in thirteenth-century manuscripts with its etymological present tense: adugo ‘I bring [ind.]’, aduga ‘I bring [subj.]’ versus aduzes ‘you bring [ind.]’, aduze ‘he/she brings [ind.]’ etc. All of these verbs were later remodelled on the basis of the pattern established by nacer, florecer, merecer etc.

5.3. Insertion of /g/ in verbs like poner, salir and caer
The insertion of /g/ before /o/ or /a/ in the paradigms of poner ‘to put’, tener ‘to have’, salir ‘to go out’, venir ‘to come’, valer ‘to be worth’, caer ‘to fall’, traer ‘to bring’ and oír ‘to hear’ is a salient feature of the modern Spanish verbal morphology. However, this instance of root-final /g/ is actually analogical, because the corresponding Latin forms have no velar at all. This can be seen from the following comparisons of the Spanish and Latin first person singular present indicative forms for these verbs (the equivalent subjunctive forms can be constructed by replacing -o and -ō with -a and -am respectively):

pongo – pōnō
tengo – teneō
salgo – saliō
vengo – veniō
valgo – valeō
caigo – cadō
traigo – trahō
oigo – audiō

The present tense (indicative and subjunctive) of the above verbs appears to have been analogically remodelled during the Middle Ages in imitation of a pattern established by verbs like plañir ‘to lament’ (< Lat. plangĕre). The sequence /-Ng-/ in verbal roots palatalized quite generally to /ɲ/ before a front vowel (Menéndez Pidal 1987: 138) but was left unchanged before central and back vowels. This dual development delivered a common Old Spanish pattern in which the first person singular present indicative of such verbs, together with all of the present subjunctive forms, had a root that ended in /g/, which contrasted with the usual root-final consonant /-ɲ/:

plango, planga, plangas etc.
versus:
plañir, plañes, plañimos etc.

Where this pattern was etymological, i.e. in verbs like plañir, it was subsequently destroyed by analogical levelling, with forms like plango, planga etc. being replaced by plaño, plaña etc. It lives on, however, in the modern paradigms of poner, tener, salir etc., where, as was discussed above, it is actually analogical.

5.4. Root-final /ʝ/
A further irregularity that is in part at least the reflex of a root-final velar is the voiced palatal fricative /ʝ/ (corresponding to the letter y) which appears in the following verb forms:

(i)    the rhizotonic present indicative forms of oír ‘to hear’ (other than the first person singular)
(ii)   the rhizotonic present indicative forms and in all of the present subjunctive forms of verbs like huir ‘to flee’, destruir ‘to destroy’ and disminuir ‘to diminish’
(iii)  the present subjunctive forms of the auxiliary verb haber and of ir ‘to go’

In the paradigm of huir (< fŭgĕre) root-final /ʝ/ is etymological in the 1st person singular of the present indicative (huyo < fŭgiō) and throughout the present subjunctive (e.g. huya < fŭgiam). However, it is analogical in the remainder of the rhizotonic present indicative forms, viz. huyes (< Lat. fŭgis), huye (< fŭgit) and huyen (< VL *fŭgent for CL fŭgiunt). In the absence of analogy, we would expect *hoes, *hoe and *hoen, given that /g/ before a front vowel usually has zero as its reflex, as can be seen, for example, in maestro ‘teacher’ < magistrum (the root vowel in huyes, huye, huyen is also analogical). Verbs like destruir and disminuir are cultismos (learnèd borrowings) and the insertion of /ʝ/ in these verbs (destruyo, destruyes, destruya etc.) appears to reflect imitation of the pattern established by huir.

The /ʝ/ that appears in three of the four rhizotonic present indicative forms of oír (< audīre), viz. oyes, oye, oyen, is also analogical, because the corresponding Latin forms are audīs, audit and VL audent (for audiunt), which would be expected to deliver *o(d)es, *o(d)e and *o(d)en through regular sound change. The forms oyes, oye, oyen were no doubt modelled on Old Spanish oyo (1st person sing. present indicative) together with the old subjunctive forms oya, oyas, oyamos etc. Both oyo and the old subjunctive forms were etymologically regular, given Latin audiō ‘I hear’, audiam ‘I hear [subj.]’ etc., in which the sequence [dj] would be expected to have /ʝ/ as its reflex. In later Old Spanish, /g/ was inserted into these forms (see 5.3 above) at the end of the verbal root, causing the existing root-final consonant /ʝ/ to be modified to [j] (given that /ʝ/ can only be intervocalic in Spanish): oigo, oiga, oigas, oigamos etc. The latter are the forms that persist into modern Spanish.

The /ʝ/ found in the present subjunctive forms of the verb haber, viz. haya, hayas, hayamos, hayáis and hayan, evolved through regular sound change, given that the ancestral Latin forms were habeam, habeās etc. Once the hiatus had been lost, the sequence -be- inside these latter forms was pronounced [βj], a phonetic group which in the majority of cases survived unmodified (as in labium > labio ‘lip’) but in some instances reduced to /ʝ/, as can be seen in the topographical term hoya ‘bowl/depression’, which descends from Latin foveam ‘pit’ (for the equivalence of Latin -v- and -b-, click here). This latter outcome is the one that applies in the case of the verb forms haya, hayas etc.

The Asturian forms eba (< habeam), ebas (< habeās) etc. offer an interesting point of contrast with Castilian haya, hayas etc., as they imply that the [j] and the [β] in the [βj] sequence were transposed, initially giving [ajβa], [ajβas] etc., with [aj] subsequently evolving to /e/, as happened in Castilian saber and caber. The transposition of [j] and the [β] in the Asturian forms also prevents the [βj] > /ʝ/ development exhibited by Castilian haya, hayas etc.

In contrast to the subparadigm haya, hayas, hayamos etc., the present subjunctive forms of ir ‘to go’, viz. vaya, vayas, vayamos, vayáis and vayan, are analogical. The ancestral Latin forms belong in fact to the verb vādĕre, also meaning ‘to go’, rather than to īre, the verb which supplies the infinitive form ir. The present subjunctive forms of vādĕre were vādam, vādās, vādat etc., whose intervocalic /d/ could not possibly deliver /ʝ/ through regular sound change. Presumably the model for vaya, vayas etc. is supplied by forms like haya, hayas etc. from haber, as well as now defunct Old Spanish forms such as oya, oyas, seya, seyas etc. For the present indicative form of ir, see Section 9 below.

5.5. Caber and saber
A further type of root alternation is the /-ep-/ ~ /-ab-/ pattern found in the two verbs caber ‘to fit’ and saber ‘to know’. In caber, for example, the present indicative is quepo, cabes, cabe, cabemos etc., while the present subjunctive has /-ep-/ throughout the subparadigm: quepa, quepas, quepa, quepamos, quepáis, quepan.

Caber and saber descend from Latin capĕre and sapĕre respectively, both of which contained the root increment [j], or ‘inflectional yod’, in the relevant forms of the present tense, i.e. the first person singular of the indicative and in all the forms of the subjunctive. This particular instance of the inflectional yod, like the yod of words such as bāsium ‘kiss’ (> [ˈbajso] > beso ), was transposed with its preceding consonant, triggering metaphony with respect to the root vowel /a/ and thereby raising it to /e/, before being absorbed into the latter sound, as illustrated below:

sapiat [ˈsapjat > [ˈsajpat] > [ˈsepa] sepa ‘he/she knows [subj.]’
capiō [ˈkapjo] > [ˈkajpo] > [ˈkepo] quepo ‘I fit [ind.]’

The voiceless bilabial consonant /p/ which accompanies the raised root vowel in these forms of saber and caber is also due to the presence of the semivowel [j]. In preconsonantal position, this had a similar effect to a consonant, in the sense that it forestalled the voicing of what would otherwise have been an intervocalic consonant (the semivowel [w] had a parallel effect in the preterite of saber). In contrast, the bilabial consonant in the forms that lacked the inflectional yod, such as cabe (< capĭt) and sabe (< sapĭt), was always between two full (i.e. syllabic) vowels and hence did undergo voicing.

Portuguese forms like saiba and saibas, the 1st/3rd and 2nd person singular present subjunctive of saber ‘to know’, reflect transposition of [j] and the original Latin root consonant /p/, exactly as in Spanish, but they also evidence an absence of metaphony and a failure of the semivowel [j] to prevent voicing of the earlier /p/.

The 1st person singular present indicative of Spanish saber, viz. sé, is analogical, modelled perhaps on he the 1st person singular present indicative of haber, although there do not appear to be any historical attestations of the expected form *sepo.

6. The pretérito grave (‘strong’ perfect/preterite)
Most verbs in the -āre and -īre classes had perfects that were uniformly stressed on the verb ending. In contrast to this, all verbs in the Latin -ĕre conjugation as well as most in -ēre had perfects that included rhizotonic forms, such as in fēcĭt ‘he/she did’ or stĕtĭmus ‘we stood’. Despite the fact that, even among these latter verbs, the forms of the second person (singular and plural) were always stressed on the endings (e.g. dīxistī, dīxistis > dijiste, dijisteis) and the third person plural was often stressed on the ending (e.g. habuērunt > hubieron), it is usual to refer to the whole subparadigm in which these rhizotonic forms figured as the strong perfect, where ‘strong’ effectively means ‘rhizotonic’.

In late Latin and Old Spanish, the strong perfect of most verbs was replaced by the ending-stressed pattern that evolved in the -īre conjugation. Where this did not happen, the rhizotonic pattern of stress survives only in the first and third persons singular (e.g. hice, hizo, supe, supo, vine, vino), but even here the third person singular ending in -o is modelled on regular habló, comió etc., as the Latin -ĭt of, say, vēnĭt ‘he/she came’ would give -e. Verbs that exhibit these vestiges of the old rhizotonic stress pattern are said to belong to the pretérito grave or ‘strong’ perfect/preterite.

Besides an irregular stress pattern, the modern pretérito grave also typically exhibits root irregularities, particularly in terms of vowels. Some of these are the reflex of regular sound change, but in the majority of cases analogy obscures the relationship between the modern form and its etymon.

The root vowel /u/ (pude, supe, estuve etc.)
A number of verbs in the pretérito grave change their usual root vowel to /u/, often acquiring a root-final bilabial consonant as a corollary. Some examples are given below:

poder ‘to be able to’ → pude, pudo etc.
saber ‘to know’ → supe, supo etc.
tener ‘to have’ → tuve, tuvo etc.
estar ‘to be’ → estuve, estuvo etc.

The /u/ in this pattern is an indirect reflex of the vowel -ŭ- which appeared immediately after the root in the perfect of the Latin -ēre conjugation as well as that of a small number of -ĕre verbs, notably sapĕre ‘to be wise’ and, through analogy, possibly also capĕre ‘to take’ (but note this latter verb’s Classical Latin perfect: cēpī, cēpĕstī etc.).

In the rhizotonic forms, this -ŭ- was unstressed and so must have come to be pronounced [w] following the loss of the hiatus. It is plausible to assume that this [w] was attracted into the root, producing the diphthong [ɔw] or [aw], depending on whether the original root vowel was ŏ (e.g. pŏtŭī ‘I was able to’) or a (e.g. habŭī ‘I had’). In the latter case, the reflex in old Spanish is /o/ through regular sound change:

habŭī [ˈaβwi] > [ˈawβi] > Old Sp. ove ‘I had’
sapŭī [ˈsapwi] > [ˈsawpi] > Old Sp. sope ‘I knew’

In contrast, the Old Spanish reflex of the [-ɔw-] type of root, at least in the first and third persons singular, is the high back vowel /u/:

pŏtŭī [ˈpɔtwi] > [ˈpɔwti] > pude ‘I was able to’
pŏsŭī [ˈpɔswi] > [ˈpɔwsi] > puse ‘I put’

This /u/ was ultimately generalized to all the other preterite forms in the relevant verbs (e.g. podiste > †pudiste, posieron > †pusieron), as well as to verbs like saber and haber (sope > †supe, ovo > †hubo).

However, before acquiring the higher root vowel, the pattern estblished by verbs like saber and haber, i.e. preterite roots in /o/ followed by a bilabial consonant, had been analogically extended to quite a large number of common verbs. Notable members of this latter category are tener ‘to have’, estar ‘to be/stand’ and andar ‘to walk’:

Old Sp. †tove; Classical Latin: tĕnŭī
Old Sp. †estove (previously estide, †estude); Classical Latin: stĕtī
Old Sp. †andove (previously †andide, †andude); Reconstructed Latin: ambitāvī

With the generalization of the high root vowel /u/, forms such as the above acquired their modern phonetic shape: †tuve, †estuve, †anduve etc. Other medieval preterite forms that had arisen through attraction to the /o/ + bilabial pattern, such as †crovo ‘he/she believed’ (Latin: crēdĭtĭt), simply fell into obsolescence.

Root-final /x/ (dije, traje, conduje)
Some Latin verbs with strong perfects amended their root-final consonant to -x- (= /-ks-/). For example, trahĕre ‘to bring’ has trāxī, trāxĭstī, trāxĭt etc. in its perfect. This cluster /ks/ evolved regularly to /ʃ/ in Old Spanish and ultimately to /x/ in the modern language, the latter consonant, written as j, now appearing root-finally in the preterite of verbs like decir ‘to say’, aducir ‘to adduce’ (also conducir ‘to lead’) and traer ‘to bring’:

dīxī > Old Sp. dixe > dije
addūxī > Old Sp. aduxe ‘I brought’ > aduje
trāxī > Old Sp. troxe > †traje

Though it looks etymological, the root vowel of modern traje, trajes, trajo etc. is actually modelled on the /a/ that occurs elsewhere in the paradigm of traer, e.g. the present indicative: trae, traes etc. The equivalent forms in Old Spanish show /o/, troxe, troxes, troxo etc. (later, /u/: truxe etc.), in imitation of the ove/sope type.

Verbs in this group are notable for having a third person plural ending that lacks the [j] (= -i-) which normally appears in this ending: dijeron, trajeron, adujeron etc. The equivalent forms in Old Spanish were in fact spelled with an -i- – dixieron, troxieron, aduxieron – implying that, originally at least, the usual palatal semivowel [j] was present. Over time, however, this [j] must have been absorbed into the palato-alveolar consonant /ʃ/ of the medieval root: [-ˈʃjeɾon] > [-ˈʃeɾon].

Root vowel /i/ (hacer, venir)
The root vowel /i/ that appears in the modern preterite of hacer ‘to do’ and venir ‘to come’ is the result of metaphony on the root vowel /e/, induced by the final /i/ of the ending of the first person singular. In the remaining forms of their preterite, then, this /i/ is analogical:

fēcī [ˈheci] > Old Sp. fiz(e) > mod. Sp. hice ‘I did’
versus:
fēcĭstī [heˈcesti] > Old Sp. feziste > mod. Sp. †hiciste ‘you did’

vēnī [ˈβeni] > vine ‘I came’
versus:
vēnĭstī [βeˈnesti] > Old Sp. veniste > mod. Sp. †viniste ‘you came’

Querer
The preterite of querer ‘to want’, which amends its root to quis-, is an isolated relic of a pattern that was once common in Spanish, one which included such now obsolete verb forms as miso (now: †metió ‘he/she put’), riso (now: †rió ‘he/she laughed’), †conquiso (now: conquistó ‘he/she conquered’) and †priso (now: †prendió ‘he/she captured’).

This pattern goes back to the so-called sigmatic Latin perfect, which was characterized by endings in -s-, e.g. mīsī, mīsĭstī, mīsĭt etc., from mĭttĕre ‘to throw/send’. The ancestor of querer, viz. quaerĕre had this type of perfect, implying that the -s- in modern forms like quise and quiso is etymological. However, the root vowel /i/ in these latter forms is analogical, as the original root vowel ae equated in stressed syllables to /ε/, which even in the first person singular, where metaphony would no doubt have occurred, would raise to /e/ rather than /i/:

quaesĭī > †quise
quaesĭĭt > †quiso etc.

Ver and dar
The verbs ver ‘to see’ and dar ‘to give’ show normal loss of intervocalic /d/, although their plural forms were restructured in the Middle Ages by analogy with the regular preterite endings shared by the -ir and -er conjugations:

vī(d)ī, vī(d)ĭstī, vī(d)ĭmus > Old Sp. vi, viste, viemos > vi, viste, †vimos
dĕ(d)ī, dĕ(d)ĭstī, dĕ(dĭ)mus > Old Sp. di, diste, diemos > di, diste, †dimos

As regards the forms di, diste and viste, the final /i/ in the Latin etyma must be assumed to have induced metaphony on the stressed vowel, giving /e/, /i/ and /i/ respectively: [ˈde(d)i], [deˈ(d)isti], [βiˈ(d)isti]. Within these latter forms, the vowel sequences produced by the loss of the intervocalic /d/ must each have reduced to /i/.

dĕ(d)ĭt > †[ˈdjeo] > †dio

fŭī > Old Sp. fue > †fui
fŭĭstī > Old Sp. †fueste (expected form: fuiste) > †fuiste
fŭĭt > fue
fŭĭmus > Old Sp. fuemos > †fuimos
fŭĭstĭs > Old Sp. fuestes > †fuisteis
fŭērunt > fueron

In each verb form, the root vowel ŭ should be assumed to have semivocalized to [w], implying rightwards stress shift in the first and third persons singular, where this ŭ (= /o/) must originally have been stressed.

7. Gerund
The modern forms descend from the ablative of the Latin gerund: cantando < cantandō, temiendo < tĭmĕndō (after [ε] > [je]), durmiendo < dormiendō. The ablative of the gerund could already in Latin be used as an adverbial modifier – e.g. lĕgĕndō discitur ‘with reading one learns’ – a function that is retained by its Spanish descendant (compare Se aprende leyendo). In addition, however, the modern gerund can occur adnominally (i.e. modifying a noun), as in El rector inaugurando la facultad…, and in the periphrastic progressive construction está hablando etc.

8. Past Participle
The descendants of -āre and -īre verbs have past participles in historically regular -ado (-ātum) and -ido (-ītum). Most -ēre and -ĕre verbs had rhizotonic participles but a few had ending-stressed forms in -ūtum and one or two in -ētum. Both of these latter endings (or their reflexes) have come to be replaced by -ido, although -udo (< -ūtum) was frequent until the 13th century: conosçudo, vençudo, sabudo, defendudo etc.

As regards the rhizotonic participles, these were typically replaced by ending-stressed forms in -ido, but a number survive into modern Spanish, including the following:

puesto ‘put’ (< pŏs(i)tum)
abierto ‘open(ed)’ (< apĕrtum)
cubierto ‘covered’ (< cŏŏpĕrtum)
escrito ‘written’ (< scrīptum)
dicho ‘said’ (< dictum)
hecho ‘done’ (< factum)

The participle visto ‘seen’ is analogical (Lat. vīsum) while vuelto ‘returned’ and (re-, ab-, di-) suelto ‘set loose’ stem from originally ending-stressed forms (volūtum and solūtum) that became rhizotonic, with the erstwhile stressed vowel being eventually syncopated.

When combined with an appropriate tense of haber, the modern past participle is used in the formation of the compound perfective tenses: he hablado, había hablado, habré hablado, hubiera hablado. In Old Spanish, however, s(e)er was frequently used as the perfective auxiliary with unaccusative and reflexive verbs, as in the examples below:

Venidos son a Castiella aquestos ospedados
El Çid y sos hyernos en Valençia son rastados
‘Those guests have come to Castile
The Cid and his sons-in-law have remained in Valencia’ (Poema de mio Cid 2269)

éranse ya los otros mucho alongados
‘the others had already got a long way away’ (Primera crónica general)

This practice presumably represents a continuation of the use of Latin esse ‘to be’ with the past participle to form the perfect, pluperfect and future perfect of deponent verbs, as in e.g. mentītus sum ‘I have lied’. However, a major reanalysis of the deponent perfect must have occurred in the pre-Spanish period, given that many of the Old Spanish verbs that selected s(e)er in the perfect were not deponent in Latin. For example, while Old Spanish has ido es ‘he has gone’, there is no equivalent form *itus est in Latin (although itum est, with an impersonal neuter subject, was available in the sense of ‘people have gone’). Conversely, many of the Spanish verbs that descend from Latin deponents did not select se(e)r in the perfect in Old Spanish, implying that at some unspecified period they ceased to be eligible to select esse. For example, although mentir ‘to lie’ and hablar ‘to speak’ come from Latin deponent verbs, viz. mentīrī and fābulārī, it has never been possible to use, say, es mentido or es fablado to mean ha mentido ‘he has lied’ or ha hablado ‘he has spoken’.

The past participle also combines, as in Latin, with the auxiliary ser/esse to form a syntactic passive. However, whereas in Latin this was an intrinsically perfective construction, in modern Spanish the perfective aspect must be supplied by the morphology of the auxiliary. For instance, in order to capture the perfective meaning of Latin captīvī factī sunt ‘They have been taken prisoner’, with a present tense passive auxiliary, Spanish requires its own passive auxiliary to be in either the perfect (Han sido hechos prisioneros) or the preterite (Fueron hechos prisioneros). Similarly, the Latin sentence captīvī factī erant, with esse in the imperfect, corresponds not to Eran hechos prisioneros but to Habían sido hechos prisioneros ‘They had been taken prisoner’. Identical remarks apply mutatis mutandis to captīvī factī erunt, with esse in the future (i.e. the tense of the sentence is future perfect – habrán sido ‘they will have been’ – rather than future).

A further difference between the Spanish ser-passive and the Latin esse-passive is that the latter was not available, in Classical Latin at least, for the present, imperfect and future, because inflected forms were used in these cases: audior ‘I am (being) heard’, audiēbar ‘I am (being) heard’, audiar ‘I will be heard’.

9. Suppletion and merger

The verb ir ‘to go’
The present tense of ir (< īre) is suppletive, in the sense that it is borrowed from the verb vādĕre, which must have been used commonly in speech as an alternative to īre. The Latin phrase quo vadis ‘where are you going?’ survives in Christian discourse and is the title of a 1951 Hollywood epic.

The modern Spanish forms vas ‘you go’, va ‘he/she goes’ and van ‘they go’ come from vādis, vādit and *vādent (for CL vādunt) respectively, with early loss of the intervocalic /d/ and reduction of the resultant hiatus /ae/ to just the stressed vowel /a/. Note that the corresponding subjunctive forms of vādĕre, viz. vādās, vādat and vādant, should give exactly the same result in Spanish. However, merger of the relevant forms was prevented by the creation of a subjunctive subparadigm that was entirely analogical, manifesting a root-final consonant /ʝ/ modelled on the /ʝ/ of forms like haya (from haber ‘to have’) and Old Spanish oya (from oír ‘to listen to’).

In the case of vo(y) ‘I go’, the ancestral form is vādō. This latter item, following the loss of /-d-/, must initially have produced *vao, in which the sequence ao is likely to have been pronounced as the diphthong [aw]. Modern Portuguese and western dialects of Leonese have vou ‘I go’, suggesting that the syllabic vowel /a/ in *vao was generally retracted to /o/ in Ibero-Romance, with Castilian additionally eliminating the final glide [w], resulting in the attested Old Spanish form vo (usually written as uo). This additional development may well have been encouraged by the existence in Old Spanish of the forms do ‘I give’, so ‘I am’ and esto ‘I stand’. In any case, all four of these first person forms acquired a final [j], now written as y, at the end of the Middle Ages, resulting in modern voy, doy, soy and estoy.

In the first and second persons plural, the indicative forms imos and ides are attested in Old Spanish. These come respectively from īmus and ītes, which belonged to the īre paradigm rather than to that of vādĕre. At the same time, it seems that the modern indicative forms vamos and vais had a subjunctive value in Old Spanish. In the case of vamos, this original subjunctive meaning persists in first person plural imperative structures such as reflexive vámonos ‘let’s go’ (compare subjunctive sentémonos ‘let’s sit down’, bailemos ‘let’s dance’ etc.). As regards vais as a subjunctive, Menéndez Pidal (1987: 306) gives the illustrative example below:

Hacedme merced que os vais.
‘Do me the favour of leaving.’

It is thus plausible to suppose that the now indicative forms vamos and vais are reflexes of the first and second persons plural of vādĕre’s present subjunctive. Indeed, these latter forms are the ones which offer the most direct diachronic pathway to vamos and vais, loss of the intervocalic /d/ and subsequent reduction of /aa/ to /a/ sufficing to account for the fate of the root, with the inflectional endings evolving in the usual way (for the -t- in the second person plural ending see 3.1):

vā(d)āmus > vamos
vā(d)ātis > vades > vais

In the modern language vamos and vais, qua subjunctives, are replaced by the analogical forms vayamos and vayáis.

In the second person imperative of ir, we see a split between the plural form id, which comes from īte, a member of the īre paradigm, and the singular form ve, which comes from vāde. The derivation of ve implies loss of /-d-/ followed by a reduction of /ae/ to [aj] through dissimilation, the sequence [aj] subsequently evolving to /e/ in Castilian but surviving in Portugues vai.

The preterite of īre also failed to survive into Spanish, where it is replaced by the preterite of ser.

The verb ser ‘to be’
The paradigm of ser reflects merger of the verbs ĕssĕ ‘to be’ and sĕdēre ‘to sit’. The majority of the forms come from ĕssĕ, but the present subjunctive forms as well as those of the infinitive, the imperative and the gerund descend from forms that orginally belonged to the sĕdēre conjugation:

sea, seáis etc. < Old Sp. seya, seyades < sĕdeam, sĕdeātis
ser < Old Sp. seer < sĕdēre
sé, sed < sĕdē, sĕdēte
siendo < Old Sp. seyendo < sĕdĕndō

In the present indicative, soy, es, somos and son can be traced to the Latin present indicative forms sŭm, est, sŭmus and sŭnt, respectively. Eres, however, comes from the future form eris, while sois – originally sodes – must be assumed to come from an (unattested) analogical form *sŭtis, modelled on sŭmus and sŭnt.

References
Menéndez Pidal, Ramón. 1987. Manual de gramática histórica española. Madrid: Espasa Calpe.

Penny, Ralph. 2002. A history of the Spanish language. Cambridge: Cambridge University Press.

Rivero, María Luisa. 1993. ‘Long Head Movement versus V2, and null subjects in Old Romance.’ Lingua 89: 217–245.