6.2.1 sit and siet

The classical present subjunctive paradigm of esse has no variation: sim, sīs, sit, sīmus, sītis, sint. In very ‘early Latin’, the singular forms are siem, siēs and siēt, and the third person plural is sient. The classical paradigm is the result of morphological levelling based on the first and second person plural. The inherited singular forms continue the full grade of the Indo-European optative suffix, *-ieH₁-, while the plural forms continue the zero-grade, *iH₁-. The reason why the inherited third person plural is sient rather than sint, despite the zero-grade modal marker, is that the ending was *-ent, with a full grade, rather than *-nti.

If we contrast the older forms with the newer ones, we must leave out sīmus and sītis, since these do not allow for variation. Table 6.1 presents the figures for the old and new forms in Plautus and Terence.Footnote ¹

The first thing to note is that in both authors, the modern forms outnumber the old ones by far. The older forms make up less than 18% in Plautus and less than 26% in Terence. They are already on their way out, but they are still frequent enough for us not to regard every single token as stylistically significant. They are part of the Kunstsprache of comedy, being significant in their totality rather than in individual instances (see Maltby, Chapter 12, p. 264 for the occurrence of these forms in tragedy). However, what is surprising about our data is that in Terence the overall percentage of archaic forms is higher than in Plautus. We will come back to this.

Since esse is often clitic, monosyllabic forms are not infrequent at line end. Among the modern forms, Plautus has 77 out of 785 at line end, amounting to 9.81%, and Terence has 24 out of 215 at line end, amounting to 11.16%. However, it is the archaic, disyllabic forms which prefer the position at the end of the line. In Plautus, 149 of the 170 forms are at line end, giving us a total of 87.65%, and in Terence, 65 of the 74 forms are at line end, giving us a total of 87.84%. Given that large metrical incisions, such as the median caesura in an iambic septenarius, follow restrictions similar to line end, we should also take into account that, in Plautus and Terence, nine and two forms, respectively, occur before such big breaks. If we treat these big breaks like line end, 92.94% of the older forms occur at line end in Plautus, and 90.54% of the forms of Terence are found there. What this shows very clearly is that in both Plautus and Terence, siem is a line-end archaism, an archaism used predominantly for metrical reasons. If they had written prose, such archaic forms would be considerably rarer.

The question remains why Terence has a higher percentage of older forms than Plautus, despite generally being more modern linguistically and adopting a more neutral tone. The answer is perhaps that this is a purely metrical phenomenon. Plautus has a greater variety of metres and is in this respect more artistic and sophisticated, but Terenceʼs iambo-trochaic lines normally read more smoothly and naturally, with much less of the Plautine tendency to let line ends and syntactic breaks coincide (Deufert Reference Deufert2002). What we have here is one of the few cases where Terence is more dependent on an artificial metrical ‘crutch’ than Plautus.

6.2.2 amari and amarier

Neither Indo-European nor Proto-Italic had infinitives. After the break-up of Proto-Italic, the different branches grammaticalised various cases of verbal nouns to infinitives. Latin esse ‘to be’ continues the locative of an s-stem to the root *H₁es- ‘to be’, *H₁es-s-i. The Oscan equivalent is ezum (Lu 1.10), without the s-stem suffix and with an accusative ending.

The classical medio-passive ending -ī goes back to -ei, a dative ending. In ‘early Latin’, we also commonly get an ending -iēr, of unclear origin; Meiser (Reference Meiser1998: 225) suggests that we are dealing with an instrumental ending *-eH₁- ‘expanded’ by a preceding glide or vowel and followed by *-r as a medio-passive marker. However that may be, doublets exist for every single conjugation (Table 6.2).

In classical prose, the ending -ier is almost non-existent. On a few occasions, Cicero quotes the archaic legal phrase inter bonos bene agier ʻthat among good people one should act wellʼ (Top. 66, Off. 3. 61 and other places). Similar forms occur in classical poetry, as a deliberate archaism. This means that while both infinitive endings may be equally old, the type in -iēr became obsolete faster and disappeared first, so that I shall from now on refer to it as the ʻarchaicʼ form, because that is what it is from a classical perspective.

But what is the situation like in ‘early Latin’? In Plautus, the two forms occur side by side, and sometimes even in the same line:

However, already in Plautus and Terence it is the classical forms that predominate. Table 6.3 gives the overall figures, without differentiating between conjugation classes.

As we can see, the archaic forms are already rare in Plautus and even rarer in Terence. In Plautus they make up less than 18% of all medio-passive present infinitives and in Terence they only reach a meagre 13%. This means that they are clearly on their way out, but not yet rare enough for us to regard every single occurrence of such forms as stylistically highly significant.

Table 6.3 does not differentiate between different conjugation classes. However, conjugation class turns out to be an important factor as far as frequency is concerned. While in Plautus, the forms in -iēr are generally slightly more frequent than in Terence, the conjugation classes behave alike in these two authors. For instance, forms in -iēr are more common in the first conjugation than in the others, and this is as true in Plautus as it is in Terence. Therefore I shall simply add up the figures for these two authors in what follows.

In the first conjugation, Plautus and Terence have 311 classical forms and 123 archaic ones, giving us a total of 434 tokens; the archaic forms make up 28.34% of the total. In the fourth conjugation, there are 64 classical forms and 21 archaic ones, yielding a total of 85 tokens; the archaic forms make up 24.71% of the total, a slightly lower figure than in the first conjugation. In the second conjugation, we get 58 classical forms and 18 archaic ones, providing us with a total of 76 forms; the archaic forms make up 23.68% of the total, which is marginally lower than in the fourth conjugation.

Altogether, the differences between these three conjugations are minor. The real outlier is the third conjugation. Next to 653 classical forms, we find only 54 archaic ones. In this total of 707 forms, the archaic ones make up no more than 7.64%. This is a substantial, significant difference from the other conjugations.

How can we explain this discrepancy? The first thing to notice is that the classical forms are not restricted in any way, while the archaic forms are almost entirely confined to line end. This means that they are little more than metrically convenient alternatives for Plautus and Terence, and, had they written prose instead of verse, they would undoubtedly have used far fewer archaic medio-passive infinitives.

But why the discrepancy between the third conjugation and the others? Iambo-trochaic lines end in ¯ ̆ ˟, and the endings of the archaic medio-passive infinitives of the first, second and fourth conjugations fit this metrical pattern perfectly. In the third conjugation, on the other hand, the antepenultimate syllable can be either heavy (type fāriēr ‘to speak’Footnote ²) or light (type ăgiēr ‘to be done’). The third-conjugation type with a heavy antepenultimate does fit without problems, but is considerably rarer than the third-conjugation type with a light antepenultimate. This light type is excluded; while the verse ending ¯ ̆ ˟ can be replaced by ̆̆ ̆ ˟, the element ̆̆ may not be divided between two words (Ritschl’s law). We can thus conclude that the differences in behaviour between the conjugation classes do not mirror differences in normal, spoken language; rather, they are purely metrically conditioned. Had Plautus and Terence written prose, the figures for all archaic infinitives would be much closer to the 7.64% that we find for the third conjugation than the 28.34% that we find for the first; in fact, they would probably be even lower than the 7.64% of the third conjugation because some of these forms, the types fāriēr and ăbı̆gı̆ēr, do fit at line end and may have been chosen purely for metrical convenience.

What makes medio-passive infinitives in -ier more exciting than forms like siem is that among the infinitives, those of the third conjugation are metrical outliers that can function as a control group. This control group shows that line-end archaisms in metrical texts are by and large a metrical phenomenon. The same forms would not simply be placed elsewhere in a prose sentence; rather, they would be avoided altogether.

6.2.3 sciebam and scibam, sciam and sciboFootnote ³

In the fourth conjugation, we find variation in the imperfect and the simple future. In the imperfect, we find forms like seruiēbāt (Ter. Ph. 83) next to seruībās (Ter. An. 38), and in the simple future, there is sciēs (Pl. Mil. 520) next to scībis (Pl. Mil. 1365). The question now arises whether such imperfect and future forms are simply metrical variants or whether there are certain differences of register.

In the classical period, there are such differences: classical writers almost exclusively use scies rather than scibis and seruiebas rather than seruibas. The types scibis and seruibas survive mainly as archaisms a poet may use for metrical reasons. Propertius has lēnībunt in a pentameter (3.21.32) and Ovid has audībat in a hexameter (fast. 3.507). In both instances, the classical forms would not fit into the metre.

The pre-classical forms eventually became so rare that copyists often ‘corrected’ or misunderstood them. Thus all manuscripts of Terence present us with the ‘corrected’, but unmetrical form seruiebas in An. 38, a form which we also find in Donatus’ and Eugraphius’ comments on this passage.

In Plautus and Terence, the situation is less clear than in classical Latin and we may wonder whether the types scibis and seruibas were already archaic or still unmarked. But before we can examine the situation in Plautus and Terence in more detail, it will be helpful to look at some facts of synchronic and diachronic morphology.

Table 6.4 lists the forms of the imperfect and the simple future in Plautus according to their conjugation.

As we can see, only the fourth conjugation has alternative imperfect and future forms; in the other conjugation classes there is no variation.

From the classical perspective scibam and scibo are archaic forms, but that does not mean that they have always been more archaic than sciebam and sciam.

The imperfect scibam is older than sciebam. The suffix -bā- was originally attached directly to the verb stem, just as in the first and second conjugation: amā-bā-m, monē-bā-m. Thus scī-bā-m is formed in the same way. Originally there must also have been forms like *agĕ-bā-m in the third conjugation. Because of reanalysis in the second conjugation (Meiser Reference Meiser1998, 197–8), speakers began to understand -ēbā- instead of -bā- as imperfect suffix, which was then extended to the third and fourth conjugations. In the case of the third conjugation, the substitution process (*agĕbām → agēbam) was already completed before our earliest texts, whereas in the fourth conjugation the forms scībam and sciēbam continued to coexist.

In the future, scībō is actually more recent than sciam. The futures -ā- and -ē- (sciam, sciēs) are original and continue old subjunctive formations. The future in -b- is an innovation of Latin and Faliscan based on analogy with the imperfect: eram: erō:: amābam: amābō (Leumann Reference Leumann1977: 579). The new formation replaced the inherited forms in the first and second conjugations before our earliest texts; by contrast, in the third conjugation the new formation never became successful. In the fourth conjugation these new analogical formations began to spread, but they never replaced the inherited forms and finally disappeared. It is also clear that forms like scībō, since they are based on corresponding imperfects, must have been created at a time when the normal imperfect was still scībam and not yet sciēbam.

Let us now look at the imperfect (Table 6.5).

The older forms in -ibam are still the norm in Plautus and Terence, where they make up around 85% of the total. The new forms in -iebam are normal in the classical language, but in Plautus they are still exceptional.

And let us contrast this with the future (Table 6.6).

The future in -ibo is rare in Plautus and Terence (20% and 13% respectively). We are equally used to the future in -ibo and the imperfect in -ibam, but only because in comedy the future is more frequent than the imperfect in absolute terms.

In the classical language the forms in -ibam and -ibo are archaic, but in Plautus and Terence those in -ibam are still considered normal whereas their equivalents in -iebam are still rare. By contrast, the forms in -ibo are already rare in Plautus and Terence, whereas the classical form of the future is the norm.

We can go further if we look at individual verbs. It is remarkable that of the 46 futures in -ibo in Plautus, 21 belong to the verb scire. If almost half of the futures in -ibo belong to one single verb, we have to check the futures in -iam. Here we find that among the 184 futures in Plautus, 43 belong to scire, slightly less than a quarter. We can conclude that scire is frequent in the simple future regardless of how this future is formed; but we can also conclude that the future in -ibo is undergoing a process of fossilisation and is formed easily from scire, but less easily from other verbs. In fact, the only verbs in Plautus which have more than one token of a future in -ibo are seruire (three tokens plus one of inseruire), mentiri and opperiri (two tokens each). Note that in scire the -i- is part of the stem, which may be the reason for this curious frequency pattern.

The situation is essentially the same in Terence. Seven out of the eleven futures in -ibo belong to the verb scire – if anything, the fossilisation process of this future type has progressed. By contrast, only 13 out of the 72 futures in -iam belong to scire. The remaining forms in -ibo belong to opperiri (two tokens) and seruire and experiri (one token each).

What all of this goes to show is that the future in -ibo is not only a later creation than the imperfect in -ibam, but also that its decline precedes that of the corresponding imperfect. The future in -ibo was an innovation which was unsuccessful and died out quickly.

6.3.1 me and med

In our earliest texts, the ablative singular of all noun declensions except for the consonant stems of the third declension ended in -d (lupōd ‘from the wolf’, turrīd ‘from the tower’, but cōnsule ‘from the consul). This final -d was inherited in the ablative of the second declension and spread from there. Among personal pronouns, the original ablative forms were mēd, tēd and sēd, and for reasons that are largely unclear (see Leumann Reference Leumann1977: 461–2) these same forms came to be used for the accusative. Final -d was lost after long vowel, first in polysyllables, then in the monosyllabic pronouns. Already in Plautus, the process is almost complete: not a single noun has an ablative in -d, and among the pronouns only mēd and tēd are still common.

Given that only the monosyllables can still have final -d in Plautus and Terence, the question arises whether the tokens we have are stylistically unmarked leftovers, or whether they are already artificial. I believe the latter to be the case. But before I can argue in this direction, I need to say a few words about the manuscript situation. Quite a few tokens of med and ted were ‘normalised’ to classical me and te in the transmission process, so we have to rely heavily on metre. Given that hiatus is allowed in certain positions in early Latin verse, and given that there is no completely accepted consensus on where it should or should not be allowed, different editors follow different practices in restoring the archaic forms. For the sake of consistency, therefore, I have looked exclusively at my Loeb edition of Plautus, leaving Terence aside. However, I have reason to believe that my findings would not be substantially different there.

Mē scans as a heavy syllable before a consonant. Before a vowel, different scansions are possible: we find complete elision, shortening in hiatus (mĕ + V) and the full form mē if it stands before a main caesura. Mēd, naturally, always scans as a heavy syllable. There are 77 tokens of mēd and 2284 of mē; mēd makes up 3.26% of the total. That alone is an indication that the form is no longer normal. This first hunch becomes a certainty once we look at the distribution patterns. Mē has no restrictions. By contrast, in the manuscripts there are no traces of mēd before words beginning with a consonant. Since mē with a long vowel can only stand before a word with a vocalic onset if mē is before a major caesura, it is obvious that mēd is an archaism that is used purely for metrical reasons, in order to have a heavy syllable in front of a word beginning with a vowel.

The same can be said for tē and tēd. The latter makes up 3.02% of the total (55 tokens of tēd versus 1767 of tē) and the distribution pattern is identical.

If there is still any lingering doubt about mēd and tēd being purely metrical archaisms, it is worth looking at the third person. The reflexive pronoun is sē, and there are 312 tokens in Plautus. An alternative sēd is attested in inscriptions, but not in Plautus and Terence. But why not? The answer is that there is no need for such a form. Mēd and tēd are used only in order to provide heavy syllables before vocalic onsets. Sē has a moderately emphatic by-form sēsē, and if we need a heavy monosyllabic alternative for sē before a vowel, we can use sēsē, with elision of the final vowel. Sēsē is attested 119 times in Plautus; 48 times before a consonant and 71 times before a vowel. There are disproportionately many tokens before vocalic onset; some of these have to be explained purely metrically.

Mēd and tēd are rare, but probably not rare enough to argue that each token is stylistically significant. In fact, since these are archaisms used for purely metrical reasons, they are probably simply a feature of Plautine Kunstsprache.

6.3.2 familiae and familiai

In the first declension, the inherited genitive singular ends in -ās, and Livius Andronicus still uses such forms in his epic, but not in his tragedies, which is an indication that already before Plautus such forms were considered very old-fashioned. In Plautus and Terence, the inherited genitive survives only in the fixed and stylistically neutral phrase pater / māter familiās. Elsewhere, the second-declension genitive -ī influenced the genitive of the first declension, and -ās was replaced by disyllabic -āī, which then turned into monosyllabic -ai /-ae by regular sound change. In metrical inscriptions, the spelling -ae always indicates a monosyllabic genitive, while -ai is ambiguous and can scan as one or two syllables.

Modern editions spell the disyllabic form as -ai and the monosyllabic one as -ae. This is a purely modern convention that is not only not followed in inscriptions, but not even in later manuscripts. In the manuscripts of Plautus and Terence, the spelling -ai is all but eliminated in favour of the modern -ae, and it is up to editors to decide whether a monosyllabic or a disyllabic scansion is appropriate, and to make this decision clear via the orthography.

Since there are many cases where both scansions would be possible, depending on one’s views on legitimate and illegitimate hiatus, I have again decided, for the sake of consistency, to look only at my Loeb Plautus. In there, there are 42 tokens of the disyllabic genitive, as opposed to hundreds and hundreds of monosyllabic forms; I have not counted the latter because quite a few of these are ambiguous between genitive and dative interpretations. At any rate, the 42 tokens of disyllabic genitives make up a very small percentage of the total.

In this case, there is less reason to argue for metrically conditioned archaisms. The forms occur before consonants (31 tokens) and, mostly with elision of -ī, before vowels (11 tokens, one of them without elision). If there is no metrical rationale, we are justified in looking for specific phrases and stylistic differences.

The phrase filiai nuptiis occurs four times in the Aulularia, as a recurring theme; aquai is also found four times, in various plays, followed by comoediai (three times in prologues) and animai, Dianai and familiai (twice each). However, there does not seem to be any real restriction to specific phrases or semantic fields; we find superbiai, nostrai and other genitives as well.

On the other hand, most of the tokens are used in contexts where elevated language is appropriate: in prologues, in formal speeches and so on. An example like Mil. 103 is typical, where a man is sent to Naupactus magnai rei publicai gratia, ‘for the sake of the great state’. Disyllabic genitives are thus used very deliberately, as markers of a higher register. In Ennius’ Annals and in Lucretius, such disyllabic genitives are extremely common and metrically secure, but our data from Plautus shows that this is a very artificial situation that is part of epic language.

6.3.3 deorum and deum

As the Greek ending -ων shows, the inherited ending of the genitive plural of the second declension was *-ōm, which survives in Latin as -om or -um. In classical Latin, there are few traces of this old ending. It is used for coins and in a few fixed phrases, such aedis deum consentium ‘temple of the united gods’; Varro (L. 8.71) says explicitly that in this collocation one cannot say deorum. Elsewhere, the ending is -ōrum, taken from various demonstrative pronouns.

Cic. orat. 155 provides many examples of the earlier ending from various poets and says that in certain expressions, it was normal. As an example of a ‘rather harsh’ phrase he gives us the following line from Pacuvius:

Here we have three genitives in -um where Cicero would have expected -orum. Although he is not explicit about it, Cicero seems to believe that the over-use of such forms is meant for a specific high-style effect.

What is the situation like in Plautus? If we count second-declension nouns, numerals and possessive pronouns, we have 119 genitives in -orum and 65 in -um. At first sight, then, the older forms could almost be argued to be in free variation with the newer ones, but a closer look reveals that this is not the case at all.

The genitive in -orum is subject to hardly any restrictions, but the genitive in -um is mostly found in collocations. Thus, among possessive pronouns, we have three tokens of meorum and six of tuorum. They contrast with four tokens of meum (and none of tuum), but all four tokens are in the phrase maiorum meum, a phrase in which meorum is found only once.

Among the plural pronouns, nostrum is found eleven times, uostrum seven times, nostrorum three times and uostrorum five times. Here, classical usage demands forms in -orum if we are dealing with possessive pronouns modifying nouns, but -um if we are dealing with a partitive genitive (‘of us’). In Plautus, each of the forms is found in both functions.

Among the nouns, deum occurs fourteen times, as opposed to nine instances of deorum. But deum is quite restricted: seven tokens are in the phrase deum uirtute ‘thanks to the gods’ and three are in the phrase deum fidem ‘the good faith of the gods’. In these collocations, deorum is not attested. In classical Latin, diuus is an adjective (‘divine’), but in ‘early Latin’ the word can, in accordance with its etymology, still function as a doublet of deus.Footnote ⁴ There are five attestations of diuom, but none of diuorum. Again, diuom is almost entirely restricted to the collocation diuom atque hominum, with just one exception in a deliberately pompous phrase (quem te diuom nominem, As. 716).

There are six tokens of uerborum and four of uerbum. Again, the older form is used for fixed collocations. Three of the four tokens are in the phrase uerbum sat est, ‘it is enough (of) words’. The modern form is found only once in uerborum satis est.

Given that the genitive in -um is the norm for currency terms in classical Latin, we would expect that situation to reflect earlier usage. This is not entirely true. Plautus has one instance of nummorum and twelve of nummum. The instance of nummorum is in Trin. 152, nummorum Philippeum ad tria milia, ‘around three thousand Philippic coins’. All twelve instances of nummum are combined with a numeral that is not modified by ad or anything else. We can assume that Plautus still had a choice in some instances, but that the combination of plain numeral with currency term had a regular genitive in -um. In the above example we can also see a genitive plural Philippum; this, or Philippeum, is attested seven times, always with a plain numeral, while we have no tokens of a genitive in -orum here.

It is hard to find other second-declension genitives in -um. In Men. 134, we have nostrum salute socium, ‘with the welfare of our allies intact’, in a mock-military passage. Sociorum is also attested once, in Vid. 56, in a more mundane context. In conclusion, then, it appears that Cicero got it right: the genitive in -um was common in certain collocations in ‘early Latin’, but outside these collocations it could be used for bombastic effects (for the occurrence of these forms in tragedy see Maltby, Chapter 12, pp. 265–6).

6.4.1 Absence of Subject AccusativesFootnote ⁵

The accusative-and-infinitive construction is a non-finite subordinate clause in which the verbal action is expressed by an infinitive, of any tense or voice, and in which the subject stands in the accusative. In finite clauses, ellipsis of subjects (in the nominative) is of course very common; in finite verbs, the endings tell us about person and number, so if an entity is recoverable from the endings and the wider context, it is often left out. On the other hand, infinitives like amare ‘to love’ or amari ‘to be loved’ do not provide us with information about subjects, so the presence of a subject accusative is normally considered obligatory or near-obligatory. However, there are exceptions:

Here, one might expect the infinitive fecisse to be combined with an accusative te, but it is left out because even though there is no verb ending to help us, the subject of the infinitive can be recovered from context. Common sense demands that, all things being equal, the ellipsis of subject accusatives should be the more common, the more easily such a subject can be retrieved from context. If there is not enough help from context, the subject accusative should be expressed in a full noun phrase. If there is some help, a pronominal subject accusative should suffice. And if the context is very clear, ellipsis should be allowed. Since the helpfulness of context is a matter of degree and of judgment rather than a black-and-white distinction, we expect that an author might hesitate between a full noun phrase and a pronoun, or between a pronoun and ellipsis; but rarely between a full noun phrase and complete ellipsis.

Of course other factors may be at play, and the one most commonly cited is register. Hofmann and Szantyr (Reference Hofmann and Szantyr1965: 362) and Landgraf (Reference Landgraf1914: 129) believe that ellipsis is favoured in colloquial registers. However, the only argument advanced in favour of this is that in modern European languages, ellipsis of various elements is more common in spoken varieties. In early Latin tragedy, where colloquialisms are much rarer than in comedy, we also find ellipsis of subject accusatives:

Here, a prescriptive grammar would expect fecisse to be combined with me. This example is by no means the only one from tragedy. It is impossible to compile reliable statistics for early tragedy because we do not have enough material, but the distribution indicates that we are not dealing with a colloquialism.

In my two examples, the subject of the superordinate verb is identical with the subject of the infinitive. One might think that this is a factor, but a complete analysis of comedy shows that it is not.

Sjögren (Reference Sjögren1906) and Lindsay (Reference Lindsay1907) both noticed that subject ellipsis is especially common with the future infinitive, and this is an interesting observation that makes little sense at first sight. In de Melo (Reference de Melo2006: 14), I presented the data for all combinations of tense and voice, contrasting accusative-and-infinitive constructions with subject pronouns and those with ellipsis. I am reproducing this material here, but without the present infinitives that have future meaning, as there are further complications there (Table 6.7).

There is not enough material for the future medio-passive infinitive, but perfect active, present active and present passive all pattern together. Here, ellipsis occurs in around one-fifth of all tokens, and we can assume that in these instances, context helps us sufficiently to recover subjects. By contrast, ellipsis is much more common in the future active and the perfect passive, where subjects are left out in one-third of the tokens. This means that in these two combinations of tense and voice, subjects are easier to recover. The reason for that is not that the context is clearer in such cases. It is that future active and perfect passive infinitives consist of esse and a participle, and participles are marked for gender and number. This makes it easier to retrieve subjects, and hence ellipsis is more common here.

What we can see here is that the label ‘colloquialism’ was applied rashly in previous scholarship. Register seems to be irrelevant for the alternation between pronominal subject accusatives and ellipsis; ellipsis is allowed if we can still recover a subject accusative, and it is more common in the future active and the perfect passive because here morphology gives us a helping hand. The moral of the story is obvious: only when an alternation cannot be explained through phonology, morphology or syntax, should we consider it to be driven by register.

6.4.2 Absence of ut

In Latin subject and object clauses introduced by ut, this subordinator is often absent, particularly in Plautus and Terence, and it is generally assumed that this absence is colloquial, particularly because it remains frequent in Cicero's letters, in the type fac + subjunctive. There can be little doubt that the subjunctive without ut is original, and that ut was first used as a voluntary modification that became more and more established, while the erstwhile construction remained common in specific construction types such as fac + subjunctive.

As with the ellipsis of subject accusatives, it is worth looking for syntactic and semantic factors influencing the choice between presence or absence of ut, before claiming a stylistic rationale. A multitude of superordinate verbs could be studied, but for the present it should suffice to look at two very frequent ones, facere and uelle. As there is little difference between Plautus and Terence here, I have conflated the tables found in Mazzanti (Reference Mazzanti2018: 76) into one (Table 6.8).

As we can see, ut is left out in one-third of all instances if the superordinate verb is facere, and in two-thirds of all instances if it is uelle. This difference is remarkable.

What is perhaps even more remarkable is that there are further restrictions for facere, but not for uelle. With uelle, ut can be absent regardless of the person, number, tense or mood of the superordinate verb. With facere, absence is only really common if the superordinate verb is in the imperative.

A verb like uelle takes a complement clause that indicates that something should be done. This ‘deontic’ quality of the subordinate clause exists regardless of what tense or mood uelle is in; it is a natural consequence of the meaning of the verb. With facere, on the other hand, the subordinate clause does not necessarily mark something that should be done. Such a deontic meaning only really arises if the superordinate verb is in the imperative or an equivalent construction. It is this semantic difference that explains why uelle and facere behave differently syntactically. More research is needed on this topic and the reasons for the diachronic restriction process, but impressionistically the findings presented in Mazzanti (Reference Mazzanti2018) hold for other verbs as well, which means that we may not have to look for stylistic differences between the presence and absence of ut.