This is an analysis of the morphophonology of Latin verbs. Based on research by Morris Halle, David Embick, and Andrea Calabrese, including “Aspects of the Morphophonology of the Verb in Latin and in German and English” and other unpublished results.
Latin has the usual five vowels /a e i o u/, with phonemic length contrasts for each, typically indicated with macrons /ā ē ī ō ū/. Latin also has three diphthongs /ae oe ui/, but those are not relevant here. Following Halle’s analysis, we posit an underlying sixth vowel /ɨ/ that is deleted in some contexts (namely, before vowels) and otherwise pronounced as /i/.
What follows is a breakdown of each vowel sound into its features. Each vowel can, of course, be long [+long] or short [-long]; this is not represented in the table below. The two back vowels /o u/ are round, so the feature [±round] is left out in this analysis. The two vowels /a ɨ/ are considered central: that is, neither front nor back, and they group together this way for the purposes of prevocalic deletion.
| a | e | i | ɨ | o | u | |
| Front | - | + | + | - | - | - |
| Back | - | - | - | - | + | + |
| High | - | - | + | + | - | + |
| Low | + | - | - | - | - | - |
Latin verbs follow strict patterns of inflection. Each pattern is called a conjugation, typically analyzed in one of four conjugations (I, II, III, IV), with the fifth (III(i)) being a variant of the existing ones (namely III and IV). Each conjugation is usually associated with a theme vowel, as detailed below, that features prominently in the present tense, and also influences other parts of the conjugations.
In addition to present stem and their conjugation, in order to conjugate a Latin verb properly one needs to know its perfect and supine stem. Often these are derived from the present stem, using patterns based on the conjugation, but sometimes they need to be memorized. Irregularities in conjugation based on the perfect stem are rare, and the supine stem is almost completely regular, if it exists.
The vast majority of verbs belong to one of these regular conjugations, only a handful of verbs are considered irregular, and even then, most forms are usually regular.
| Name | Theme vowel | Example | ||
|---|---|---|---|---|
| Conjugation I | ‑ā‑ | laud‑∅‑ō | laud‑ā‑mus | laud‑ā‑re |
| Conjugation II | ‑ē‑ | mon‑ē‑ō | mon‑ē‑mus | mond‑ē‑re |
| Conjugation III | ‑ɨ‑ | dūc‑∅‑ō | dūc‑ɨ‑mus | dūc‑e‑re |
| Conjugation III(i) | ‑i‑ | cap‑i‑ō | cap‑i‑mus | cap‑e‑re |
| Conjugation IV | ‑ī‑ | aud‑i‑ō | aud‑ī‑mus | aud‑īre |
| Athematic | ‑∅‑ | sum | sumus | es‑∅‑se |
| fer‑ō | fer‑i‑mus | fer‑re | ||
We start by looking at the morphological structure within the verb. It acquires this complex inflectional structure by head-to-head movement in the syntactic tree. After this movement, the morphemes are spelled out and phonological rules apply to get the final spoken form.
The parts of the verbal structure (v, ASP, T, M, and AGR) are explained in the chart below. Each part (with the exception of AGR, which always is a final suffix at the end), has a theme vowel associated with it. The theme vowel associated with v is the main theme vowel associated with the conjugation.
| √Root | v+TH | ASP+TH | T+TH | M+TH | AGR |
|---|---|---|---|---|---|
| The root of the verb form | The (unused) verbalizer together with the main theme vowel of the verb’s conjugation. | The aspect of the verb (perf or null), and the aspect’s theme vowel (if not null). | The tense of the verb (pres, past, fut, or null for the infinitive), and the theme vowel of the tense (if not null). | The mood of the verb (subj or null), and the theme vowel of the mood (if not null). | The (non-semantic) agreement of the verb with its subject, for finite tense. |
The morpheme ASP[perf] has various realizations, depending on the root it is attached to. The most common realization (almost universally for the first conjugation, occasionally in the second, and frequently for the fourth conjugation) is as /v/ after a theme vowel (c.f. laudāvī, dēlēvī, audīvī). Sometimes third conjugation verbs even pick up a long-i theme and have a perfect like the fourth conjugation (c.f. cupīvī). Usually for the third conjugation, however, as well as in other cases, the verb loses its theme vowel in the perfect, and a number of cases occur. Standard phonetic behavior dictates that /v/ becomes /u/ in that context (c.f. monuī). (In Roman times, the distinction was not even notated, v and u were written the same, although the phonetic distinction was known.) However, with verbs that become athematic in the perfect, other allomorphs of ASP[perf] occur, including /s/ (c.f. saepsī) and /∅/, where alteration of the root sometimes occurs with the null allomorph, usually vowel lengthening/ablaut in the stem (c.f. cēpī, ēgī) or reduplication (c.f. spo(s)pondī, tetigī).
When a verb is athematic, following Calabrese, this analysis states that the v+TH node is completely deleted (not just filled by nulls), which allows the ASP[perf] to influence the root, since they become adjacent. Similarly, this analysis also says that T[pres] is deleted, whenever it occurs; it is used to mark finite tense but disappears phonetically.
There are several patterns in the other morphemes. Most of the morphemes start out as /s/ plus a theme vowel, but in most cases it is rhotacized to become /r/, with the notable exception of the pluperfect subjunctive, where T[past]‑M[subj] is realized as /s+∅‑s+ē/. More specifically, both T[past] and T[fut] are realized with /s/ (plus a vowel) after AST[perf]. Otherwise, they are realized with a /b/ (plus a vowel). In both cases (/s/ and /b/), T[past] has theme vowel /ā/ and T[fut] has theme vowel /ɨ/. However, there is one exception: after the high theme vowels of conjugations III, III(i), and IV, T[past] gains /ē/ and T[fut] is realized as just /ē/. The final exception, which is not accounted for in this analysis, is that the first-person singular future indicative after high theme vowels (i.e. v+TH[+high]T[fut]AGR[1S]) is realized as /‑am/ (instead of the expected */‑eō/).
Properly the allomorph should be selected by the verb root via vocabulary insertion. However, this is not modeled in the widget above.
NOT rhotacised, since it is immediately followed /s+ē/ realizing M[subj].
Note: not sure if this is null, or if M[subj] is null, or maybe T[past] is /s+∅/ and M[subj] is /∅+ē/ …
Rhotacised.
Rhotacised.
Historically this was related to the subjunctive mood. (Citation?)
Historically this is related to the participle of esse, “futurus”. (Citation?)
Rhotacised in imperfect tense, not pluperfect.
Rhotacised.
Applies only for the first conjugation. Triggers deletion of the preceding theme vowel /ā/.
For all the other conjugations.
This deletes /a/ and /ɨ/ for conjugations I and III.
This is a common sound change in the development of Latin phonology, and occurs in many contexts (c.f. mōs, mōris). Notably for our analysis it allows us to posit that the infinitive /es‑se/ is actually a regular form, being athematic.
This takes care of vowel shortening for Conjugations I, II, and IV in various contexts, such as before /‑ō/ in the first-person forms.
This applies to vowels in a syllable with a nonstrident consonant in the coda. This takes care of vowel shortening for Conjugations I, II, and IV in the remaining contexts, such as in the third-person forms before /‑t/ and /‑nt/ and /‑n‑tur/ (but importantly not before the third-person passive /‑tur/, which starts a new syllable and still has a long vowel preceding it).
Note: this was implemented in a hacky way, in lieu of implementing consonant features and proper syllabification. In particular, the relevant agreement endings were equipped with a “shortening phoneme” "˘" that shortened the preceding vowel and was deleted later.
Finally, as the last sound change, any remaining /ɨ/ are pronounced as /i/.
Note: this is currently applied indescriminately every time before /r/.
The ordering of rules is, of course, important to ensure the correct results. In this case, there are a few rules that are important to be ordered relative to each other. Obviously central-vowel deletion must be before high-vowel fronting, otherwise it would not apply to the underlying /ɨ/, which was the whole point of introducing that vowel. Lowering must apply after high-vowel fronting (or it could be generalized …).