Dormivit et resurgit. A language-ecology approach to the diachrony of the Latin ingressive perfect

4.1 Language-external influences on the occurrence rate of ingressivity

In Table 3, the results of our close-reading analysis of a representative sample of 227 data points selected evenly from five separate time periods and five text types (depending on availability) are presented in a distribution over these time periods and text types.^[14] By necessity, some cohorts are less well represented: from the earliest period, texts other than Plautus’ and Terence’s comedies are rare or fragmentary; Christian texts in Latin occurred exclusively after around 200 CE; and apparently, texts that reflect the everyday language of their time more closely than other texts (e.g. comedies, personal letters not meant for publication, epigraphic texts such as curse tablets or wooden tablets not meant to last beyond the immediate communicative context) hardly contain any instances of the target observations (i.e. perfectum tense forms of the 19 selected stative verbs), with the exception of Plautus’ and Terence’s comedies from the Early Latin period.^[15] From this table, and in line with our hypotheses regarding the dynamics of the diachrony of the ingressive function of the perfectum stem tenses, first as a residual archaism and later with renewed productivity under the influence of Greek, it would appear that both poetic texts and Christian texts display more attestations of ingressivity with the target observations than literary prose texts (both about 40 %, vs. 16 %). Note, on the other hand, the low occurrence rate of ingressivity in the early, though colloquial, comedies of Plautus and Terence. From the 3rd century onwards, also technical texts display more ingressives with the target observations than before (50 %); simultaneously, the ingressive meaning reaches also a stable occurrence rate across text types of about 38 %, i.e. as (part of) the function of over one-third of our target observations.

As a possible counterargument against relating the correlation with poetic texts to archaic language use, one could claim that the perfect tense was chosen only metri causa, i.e. to obtain the desired number of syllables for the verse metre. However, as illustrated by the example in (18), from Manilius’ 1st-century astronomical treatise in verse, the competition is with the present tense rather than with other past tenses: as discussed above, a ‘(present) anterior ingressive event’ like iacuerunt ‘they have lain down’ is semantically quite similar (if not identical) to the (present) situation ‘they lie dead’, expressed in a less marked fashion by iacent. Note that the status of the cum-clause with iacuerunt as a background situation (more specifically, a circumstantial situation; cf. Aerts 2021b: 179), is made explicit by the (underlined) narration of the events that happen against that background in the subsequent clauses.

As such, while the pf. ind. can certainly be considered a marked tense choice to express a present situation like iacuerunt ‘they lie dead’ in (18), and while many such choices for ingressive perfects over unmarked present tense forms could indeed be considered to be related to metrical considerations, the competition at stake is not between two past tense forms such as the pf. ind. and impf. ind., which share a much larger segment of their semantic potential and for which the choice depends on much more subtle textual and interpersonal motivations.^[16] The fact of the matter remains that, for perfectum stem forms such as iacuerunt in (18) to be grammatical as a form to convey a present situation, ingressivity must have been considered part of their semantic potential.

In a similar way, one could argue that many of the attestations in texts influenced by Greek and its aorist tense could be accounted for by translators’ more or less consistent choices to translate the ingressive function of the Greek aorist with the Latin tense that is usually also selected to represent the other, more frequent uses of the aorist, i.e. the pf. ind. Again, however, it seems implausible that a proficient native speaker like Augustinus or Hieronymus would use the pf. ind. for ingressivity if it struck them as ungrammatical within the Latin tense system.^[17]

In (19), a passage in Greek from Eusebius’ 4th-century Historia ecclesiastica is accompanied by two translations into Latin, by Anastasius the Librarian (9th century) and by Rufinus (5th century), respectively. The Greek original features three instances of the perfect form κεκοίμηται: originally, the default meaning of the perfect would have led to the interpretation ‘(s)he sleeps’, i.e. the state that results from a previously completed action of a person lying down to sleep. However, by the 4th century, the semantic potential of the Greek perfect is much more diverse, complex and possibly ambiguous: as concluded by Crellin (2020: 475), even though his data reveal different trends depending on register, anteriority and therefore some degree of convergence with the aorist seem to have become part of its semantics (i.e. ‘[s]he has died’). Interestingly, Anastasius’ consistent choice for the pr. ind. dormiunt and dormit reflects his interpretation of the Greek perfect κεκοίμηται as a present stative situation, i.e. ‘they sleep’. Note also, as suggested by his choice for quiescit, that he understands the Greek present ἀναπαύεται as ‘she rests’. As such, Anastasius’ versio is more a description of where Philip, his daughters and John are buried. Rufinus’ versio, on the other hand, reflects more closely the later meaning – especially on a textual level – of the Greek perfect as an anterior event (i.e. ‘he has lain down’): not only does he opt for the pf. ind. twice (dormierunt, dormivit), he also elects to narrate or report as events the ageing (consenuere ‘they grew old’) of Philip’s virgin daughters and the fact that another daughter of his remained (remansit) in Ephesus until her death, thus diverging from the versions of Eusebius and especially Anastasius by conveying the information of where their graves can be found by profiling the current relevance of anterior events (i.e. ‘they died/grew old/stayed there’) rather than present situations (i.e. ‘they sleep there’) themselves. Importantly, for that to work with dormio ‘to sleep’, the pf. ind. has to combine anteriority with ingressivity, i.e. ‘they (have) started to sleep’ and hence, in context, ‘they (have) died’. Still, however, for the location of John’s grave he does select the pr. ind. dormit: given Eusebius’ consistent use of κεκοίμηται, and given the contents of the passage, it seems more likely that Rufinus considered dormivit and dormit to ultimately convey the same basic meaning (i.e. present situations) rather than that he wanted to distinguish the third instance of dormio from the first two.^[18]

Likewise, the ingressive function of the Greek aorist is met with various ‘forms’ in Latin, suggesting again that the language system provides for its expression in ways that allow for more subtle nuances depending on the translator’s interpretation of the original. In the 7th-century Glosses on the Psalms, the anonymous author narrates the earthquake following the death of Christ on the cross by means of the pf. ind. of tremo ‘to tremble’.^[19] A non-ingressive reading of the perfect tense in a narrative passage such as this one would entail that the earthquake or ‘trembling of the earth’ was over when the stars were concealed and the rocks were split (complexive perfect, cf. note 12), but such a reading appears rather counterintuitive. Still, a more common way to refer to earthquakes in Latin is the translation opted for in the corresponding passage in the Vulgate (Mt 27:51), i.e. terra mota est ‘the earth was moved’, which features an unmarked (narrative) use of the pf. ind. of a telic event (moveo). The original aorist form ἐσείσθη in the New Testament, however, which should be interpreted as ‘started to tremble’, corresponds more closely to tremuit. Given these two versions in Latin translations, a look at the Vetus Latina database might reveal more about their history: there are many examples of terra mota est (also terra commota est, terre motus factus est, terra movebatur), but also attestations of terra tremuit as well as terra contremuit (e.g. Ambr. inst. virg. 7, 46, 5; Ambr. fid. 5, 14, 34; cf. Arnob. nat. 1, 53), terra intremuit (e.g. Alc. Avit. Carm. 6, 236), and terram tremere facit (Ambr. fid. 5, 54 apud Ioh. Maxent. Conc.^S IV 2 48), terram tremuisse (Ps. Avg. quaest. test. 94, 3). As such, the option of a (grammatical) ingressive tremuit to refer to the earthquake is made explicit by its coexistence with lexical alternatives (with affixed corradicals) for the same content, which also illustrates that the reading of the Greek original as an event (i.e. the initiation of the trembling) rather than a globally viewed situation (cf. note 12) was a common one.^[20] In particular, in an excerpt from Pseudo-Augustine (Ps. Avg. serm. ed. Caillau 1, 26, 7), the version where the other portents occur ‘during the earthquake’, i.e. after the earth had started trembling, is rendered even more clearly by the use of in terraemotu.^[21]

From our discussions of (19) and (20), we can deduce not only that the ingressive function of the pf. ind. was already part of the native Latin tense system, but also that it interacted with similar form-function pairings in the language system to the point that the choice for a certain form resulted in more subtle, pragmatic differences (related to e.g. discourse mode, sequentiality and textual status among foregrounded events).^[22] As such, it would appear that native speakers of Latin were able to make use of the subtleties present in their own language system when dealing with Greek tenses, and that these form-function pairings were not simply ‘calqued’ into Latin. However, given the immensity of Christian literature in Latin from the 3rd century onwards, where much of the language is based on a Greek original where ingressivity was likely still more productive, an increase of attestations of the ingressive meaning of perfectum stem tense forms should not be surprising.^[23] More interesting still is the subsequently increasing occurrence rate, suggested by the data in Table 3, of ingressives with perfectum stem tenses in other text types as well, particularly in technical texts. In that sense, whereas the ingressive function might not have been inspired by translations of the Greek aorist encountered in Christian literature, it might have been rekindled simply because it was used (or found ‘useful’) more often than during the preceding centuries.

4.2 Language-internal influences on the occurrence rate of ingressivity

To account for the ‘dormant’ phase of Latin ingressivity, we now turn to language-internal influences on its diachrony. One of these influences stems from the lexicon and its internal dynamics. Table 4 presents the attestation rate of ingressivity for the various lexemes that were included in our corpus queries. Naturally, only the 227 data points selected for the representative sample in our qualitative phase are included. As can be observed in the table, there are vast differences between our lexemes: for sedeo ‘to be seated’ and haereo ‘to be stuck’, the rate of ingressivity is around 85 %; others, like dormio occurred with ingressivity half of the time (but notice the low number of data points in our sample) or 40 % of the time (iaceo ‘to be in a lying position’), but most verbs are used with ingressive meaning less than 30 % of the time. Recall, at this point, that contrary to Haverling (2003, 2013], 2018], we did not focus our corpus queries on stative verbs that coexist with an affixed corradical verb for the expression of ingressivity in a lexical way. Moreover, from our verbs that do have such corradicals, pateo, ardeo and sileo were disregarded in our further analyses as their perfectum stem forms (e.g. patuit, arduit, siluit) are identical to those of their corradicals (patesco, ardesco, silesco).

However, as it turns out, for the most common lexemes in the corpora (cf. Table 2), there is a striking difference in occurrence rate of ingressivity between those that coexist with such a corradical (especially taceo [5.6 %], for which conticesco is the lexical alternative) and those that do not (especially sedeo [86.4 %] and haereo [83.3 %]).^[24] Moreover, for sedeo, as evidenced by its descendants in Romance, the ingressive meaning at some point became part of the core meaning of the unaffixed verb itself: in Italian, sedere is used for both ‘to be seated’ and ‘to sit down’, although for the latter sedersi is often preferred; in French, the immediate descendant seoir meant ‘to sit down’ as well, although it is now defective and replaced, through a derivational strategy similar to the one for its Italian sibling, by s’asseoir.^[25]

Verbs with less impressive occurrence rates for ingressivity seem to confirm that same connection with a later inclusion of the ingressive variant into the base meaning of the lexeme.^[26] First, the Romance descendants of iaceo ‘to lie, to be in a lying position’, which occurred with ingressivity 40 % of the time in our sample, did not incorporate the ingressive variant ‘to recline, to assume a lying position’ in their base meaning as widely as they did for sedeo (only to some extent in Ibero-Romance, e.g. Portuguese jazer, Galician xacer, Spanish yacer). For possideo, doleo, pendeo or lateo, to the best of my knowledge, none of the Romance descendants have included ‘to obtain’, ‘to get hurt’, ‘to enter into a hanging position’ or ‘to go into hiding’ in their base meanings.^[27] For teneo, the verb with the most attestations in our corpora and an occurrence rate of ingressivity of 28.6 % in our sample, only Italian tenere seems to have included a meaning of ‘to take’ next to ‘to have, to hold’. Likewise, Italian is the only Romance language which clearly has both ‘to be silent’ and ‘to stop talking’ included in the base meaning of their direct descendant of taceo. Interestingly, affixed corradicals with a basic ingressive meaning existed in Latin for many such verbs that are shown to have lower occurrence rates for (grammatical) ingressivity (i.e. condolesco, latesco, conticesco).^[28]

Apart from the role played by Italian, we could therefore tentatively conclude that (a) part of the attestations of ingressivity with verbs that feature this function in the perfectum stem forms more often than other verbs (sedeo, and to some extent iaceo), could be related to their individual development towards including an ingressive reading within the semantic potential of the lexeme, which most likely started at some point early enough to account for the wide distribution of this lexical potential among Romance descendant forms, and that (b) stative verbs for which affixed corradicals were already thriving in Latin, were expressed in the perfectum stem tense forms less often since the ingressive use with statives was likely perceived as a much more marked (and possibly archaic) use than the anterior or perfective past use of their telic corradical counterparts.

Importantly, these data and results reveal the importance of including more lexemes in corpus queries, based on the relation in the language system between stative Aktionsart (i.e. ideally all stative verbs) and grammatical aspect (i.e. perfectum stem forms), rather than searching for forms of individual pairs of corradical verbs (i.e. stative and telic verbs such as tacuit and conticuit) and other fairly isolated lexemes such as habeo (filium).^[29] While Haverling’s (2003: 127–129, 2018]: 251–253) conclusions about the changes in the actional system in Late Latin (e.g. the conflation between tacuit and conticuit as having both stative and telic uses) seem to be supported by the tenses they both are attested in (pf. ind. vs. impf. ind.), these conclusions are problematic for two reasons. First, the discussion of examples of tacui(t) that do not express ingressivity completely disregards the semantic potential of the pf. ind. to convey a ‘complexive perfect’, as well attested in both Greek and Latin; with this use of the perfect tense, no assumptions about any diffuse Aktionsart or actionality of taceo is necessary.^[30] Second, as these conclusions are based on research on a (lexically) selective and rather restrictive number of target observations, they should not in themselves lead to conclusions about grammatical systems such as the pairing of ingressivity with the perfect tense or the functions of aspect and the perfect and imperfect tenses. At best, we can observe from some attestations that the semantic potential of the lexeme conticesco is expanded to include the state of being silent rather than just the event of becoming silent (e.g. Haverling 2013: 126).

Returning, in conclusion, to our insight based on Table 4 that direct Romance descendants of Latin statives with a significantly higher occurrence rate of ingressivity include both the stative and the ingressive variant in their base meaning (e.g. sedeo > It. sedere, Fr. seoir (obsolete) and iaceo > Port. Jazer, Gal. xacer, Sp. yacer), we can observe in Romance that these individual languages feature a lexical innovation that specializes in the ingressive variant, in much the same way as Latin featured innovations such as conticesco and latesco to complement taceo and lateo: in Italian, sedersi is used exclusively as ‘to take a seat, to sit down’ whereas sedere can be used for both meanings (i.e. also ‘to be seated’); in French, however, seoir also had both meanings, but asseoir was specialized in ‘to sit down’ and most commonly used in the reflexive version s’asseoir, of which a periphrasis être assis has been the default form for ‘to be seated’ since seoir became obsolete. Concerning iaceo, the semantics of Port. jazer, Gal. xacer and Sp. yacer still include both the stative ‘to be lying down’ and the telic ‘to recline’, although the verb is generally perceived as archaic and usually replaced by a reflexive version of a verb of another root for the ingressive reading (Port. and Gal. deitarse, Sp. acostarse and tumbarse), of which, again, a periphrasis conveys the corresponding state (Port. and Gal. estar deitado and Sp. estar acostado, estar tumbado).

As such, it would seem that those Latin statives that had no corradical and were specialized in the ingressive reading, (a) tended, to some extent, to naturally occur more often with ingressivity in Latin, as there was no (corradical) alternative, (b) subsumed the ingressive reading in their (lexical) base meaning because of that extensive use with grammatical ingressivity, (c) came to be perceived as inefficient or archaic, or otherwise highly specialized (e.g. here lies … on tombstones for Ibero-Romance descendants of iaceo), possibly because of the resulting ambiguity, and (d) were either supplemented by (d.1) a derived, specialized form (e.g. sedersi with ingressive reading only) or (d.2) replaced by a different lexeme (e.g. deitarse or tumbarse) for both readings altogether, often involving a reflexive clitic that renders explicit the inherent reflexive meaning of e.g. ‘sitting down’ or ‘lying down’, since these alternative lexemes are transitives themselves (e.g. deitar ‘to lay someone down’); the stative reading is then expressed (grammatically) by means of a periphrasis of these lexemes, e.g. estar deitado or estar tumbado. All in all, such developments are not entirely dissimilar to Haverling’s observations on the gradual fading of actionality with lexical ingressives such as conticesco. In any case, they strongly suggest a continuous exchange and mutual influencing between the Latin systems of (perfectum) tenses and of actionality.

Another cluster of language-internal influences on the occurrence rate of ingressivity is presented in Table 5. As illustrated in (21), and contrary to what many of the examples presented above might have suggested, grammatical ingressivity is not restricted to the indicative or even the pf. ind. tense: in fact, 43.8 % of attestations of the pf. inf. of the targeted stative lexemes feature ingressivity as part of their meaning, against 31.7 % of our attestations of the indicative and 29.6 % of attestations of the subjunctive. In the passage from Ovid’s 1st-century Epistulae (i.e. Heroides) below, the infinitives iacuisse and tenuisse convey situations (‘to lie on the couch’, ‘to hold a girl’) that follow from the initiation of the same states; again, therefore, the present infinitive would have constituted a less marked choice for more or less the same meaning.^[31]

One possible explanation for such a discrepancy between indicatives and, especially, infinitives may be related to the occurrence rates of ingressivity in narrative and in non-narrative discourse: while 41.7 % of our data points in narrative discourse feature ingressivity, these almost exclusively concern indicative tenses (24 out of 25); infinitives and subjunctives occurred far more often in non-narrative discourse (38 out of 43 infinitives and subjunctives), alongside many indicatives (121 out of 159 data points in non-narrative discourse). As such, particularly prone to feature ingressivity are, on the one hand, indicatives in narrative discourse (43.6 %) and, on the other, subjunctive and infinitive forms occurring in non-narrative discourse (14 out of 38, or 36.8 %). In narrative discourse, the possible reasons to use the pf. ind. with stative situations are rather limited (apart from ingressivity, these are mainly the complexive perfect and an external perspective; see Aerts 2021a: 55–62).^[32] In non-narrative discourse, more options with the indicative of states are available, especially ‘anterior situations’ or ‘absolute past situations’ (see Aerts 2021a: 59, 2021b: 178), resulting in a lower occurrence rate of ingressivity with indicative tense forms than in narrative discourse (24.8 % vs. 43.6 %). That diversity is more limited for infinitives and subjunctives, which express relative tense and absolute-relative tense, respectively, and therefore only anteriority in the case of perfectum stem forms.^[33] The added value of ingressivity, based on the perfectum stem’s aspectual inheritance from the aorist, is often combined with that anteriority in these attestations.^[34]

One final interesting observation can be made with regard to the conspicuous fronting of the analysed perfectum stem forms. One can imagine that, at a time when a certain function of a certain form is perceived as a ‘marked’ use of that form, that special meaning could be highlighted by conspicuously fronting the predicate.^[35] In that sense, the data in Table 6 seem to support the hypothesis that ingressivity was originally an archaic function of perfectum stem tenses that was revived in later texts: in the earliest texts (BCE 240–90), all fronted target observations did not occur with ingressive meaning and the only two attestations of ingressives (out of 24) did not display predicate fronting, in line with the idea that ingressivity was a less marked meaning then compared to later times, when ingressives occurred in fronted position about 30 % of the time. However, due to the potential influence of metric considerations on this correlation, Table 7 was added with the same information but with the exclusion of all data points from verse texts; an unfortunate result is, of course, the decimation of data points especially from the Early period (i.e. the archaic comedies).

Based on the data presented in Table 7, a number of observations can be made. First, in the prose texts of our sample, our target observations never occurred in fronted position before the 1st century CE; 3 out of 15 (i.e. 20 %) data points in these two time periods nevertheless occurred with ingressive meaning, confirming our previous observation that ingressivity might not have been perceived as a particularly marked function of the perfectum stem tenses in Early and Classical Latin. In later periods, when more data points survived the selection for our sample, ingressive data points increasingly occurred with verb fronting, seemingly highlighting the realization that ingressivity was a marked, perhaps unusual function of perfectum stem tenses but nevertheless a potential one that came in handy in increasingly intensive linguistic exchanges with Greek, especially in the translation of Christian texts. However, observe also that verb fronting became more frequent, albeit to a lesser extent, with non-ingressive meanings as well, that is to say, in general. Obviously, these observations are to be regarded with caution: on the one hand, the absolute frequencies in many of the cells in Tables 6 and 7 are very low and percentages could easily change in either direction with the addition of one or two extra data points; on the other hand, as shown by our data on verb fronting in general (i.e. without distinction between functions) and in line with observations made in the literature on the diachrony of word order patterns in Latin in general, we cannot at this point make any substantial claims about perceived markedness of the ingressive function on the basis of verb fronting.^[36]