The Development of Fore/Futurum Ut from Ovid to Festus: A Study in Semantic Change and Its Basis in Discourse Situation

by Laurence D. Stephens
The Development of Fore/Futurum Ut from Ovid to Festus: A Study in Semantic Change and Its Basis in Discourse Situation
Laurence D. Stephens
The American Journal of Philology
Start Page: 
End Page: 
Select license: 
Select License


1. Introduction. In two previous studies1 I provided the first systematic treatment of the syntax, meaning, and discourse functions of the so- called periphrasis of the Latin future infinitive, forelfuturum (esse) ut(i),*from its first attestations in Plautus and Terence through Livy. I showed that, contrary to prevalent opinion, the construction, at least in its free instances (i.e., with verbs which have FAP's and supines), is not merely a periphrasis for the FA1 or FPI. In fact I established over a score of constraints which differentiate it from the FI. After Livy the construction undergoes substantial changes. Before the construction disappears, the constraints on fore ut are gradually eroded, and it begins to merge semantically and pragmatically with the FI. These changes have never before been documented. I will do so here and indicate, where recoverable, the sorts of textual or discourse situations which promote them. I will also show that, in terms of historical seman- tics, the construction follows a gradual but coherent path of evolution, which is even, to a certain extent, predictable from theoretical consid- erations.

I trace these developments through the second century B.C. The construction is attested in the following authors: Ovid, Vitruvius, Seneca the Elder, Valerius Maximus, Cornelius Celsus, Seneca the

'L. D. Stephens, "The Latin Construction forelfuturum (esse) ut(i): Syntactic, Semantic, Pragmatic, and Diachronic Considerations," AJP 110 (1989) 595-627 and "On the Model Semantics of the Latin Construction forelfuturum (esse) ut(i)," Indogermani- sche Forschungen 96 (1991).

ZHenceforward, I use fore ut to indicate all of the possibilities of the formula in the text, unless it would be confusing to do so given the presence of futurum in an example under discussion. In fact the relative frequency of futurum increases considerably at the expense of fore after Livy. I will employ the following abbreviations: FA1 = future active infinitive, FPI = future passive infinitive (supine + iri), FI = future infinitive, either active or passive, FAP = future active participle, GV = governing verb. It is sometimes necessary to distinguish the cases in which the use of fore ut is unavoidable, if futurity is to be marked, because the verb in question lacks a FA1 or FPI, from cases in which one or the other FI was available. The former cases will be referred to as "necessary," the latter as "free."

Amencan Journal of Philology 111 (1990)513-542 D 1990 by The Johns Hopktns Unwers~tyPress

Younger, Petronius, Columela, Pliny the Elder, the Znstitutio of Quin- tilian, the major and minor declamations attributed to Quintilian, Fron- tinus, Tacitus, Pliny the Younger, Suetonius, Florus, and Festus. I have relied primarily on the list prepared by Sjostrand,3 but have checked and supplemented this material. Sjostrand missed a few cases, for ex- ample Petronius 78.3, which contains no departure from earlier usage, and Celsus 5.28.1D, which does (see section 3.15). In this material there are some 93 instances of fore ut or its finite counterpart. A rather larger number of subjunctive verbs are embedded under the construction, whether as at Quint. Znst. 3.8.17, with triple, asyndetic repetition of ut, or, in 15 cases, with co-ordinated verb phrases. (Such co-ordination attains its highest relative frequency in Tacitus and Suetonius; on the semantics see section 3.3.) Thus the material provides quite a substan- tial data base in which to detect innovations, particularly as they may be tested against the extensive material available for the earlier period from Plautus to Livy (cf., e.g., section 3.18).

This study is concerned with later changes in the fundamental semantic values and discourse functions characteristic of fore ut in the period from Plautus to Livy, but, as a prefatory observation, it is worth noting that the construction is affected by other, independent changes in Latin syntax. One of these is its appearance after non dubito, where, according to classical usage, quin and not the accusative and infinitive is expected. The A. c. I. dependent on non dubito is first attested at Lucr. 5.249.4 Its usage with non dubito develops through a stage in which it is generally restricted to a word order preceding non dubito,5 a serialization effect paralleled by, inter alia, the sequence of tenses used with the historical present in certain authors. This word order restric- tion has ceased to operate by the time forelfuturum ut is first observed with non dubito at Plin., ep. 6.21.7

neque enim dubito futurum ut non deponas .
Another development that overtakes fore ut is the replacement of ut (. . .) non by ne, first attested at Val. Max. 1.1.8

3Nils Sjostrand, De futuri injinitivi usu latinorum quaestiones duae (Lund 1892). 4See Cyril Bailey, Titi Lucreti Cari De Rerum Natura Libri Sex (Oxford 1947) vol.

3. 1358 ad loc. SA. Szantyr, Lateinische Syntax und Stilistik (Munich 1965) 357.

futurum enim, si quid prodigii in ea accidisset, ne dignosceretur, utri rem divinam fieri oporteret; . . .

and 6.4.3

continuo enim rex affirmavit fore ne amplius de se Ptolemaeus queretur.

It is well known that the ultimate source of ne for ut (. . .) non is provided by discourse situations in which jussive or optative force is salient even when a consecutive reading is not excluded, as in one of the earliest instances at Plaut., Poen. 1252

ne indigna indignis di darent, id ego evenisset vellem; . . .

where it is probably not chance that the ne clause precedes evenisset and is potentially analyzable as paratactic. The proximate source for fore ne is probably to be sought in the use of ne with other verbs that introduce happenings rather than doings in which the control of an agent is salient (see section 2 on these concepts). Examples would be accidere, e.g., ("in einem SonderfallV6) at Cic., div. 2.21

potest . . . accidere ne fiat

and evenire, e.g., Val. Max. 7.4.4

quo evenit ne . . . sciret . . .

As far as the decline of the construction after the middle of the second century A.D. is concerned, spot checks reveal that it certainly does not occur in Apuleius' Metamorphoses (probably after 160 A.D.), nor Historia Augusta (Probus 24.2 is merely a consecutive clause dependent on tantae . . . claritudinis [interestingly, in the context of a false prophecy by haruspices: cf. section 3.11]), nor the Vulgate (although its finite counterpart futurum est ut does, e.g., I Sam. 2:36, Matt. 2:13, Luke 9:44). A search for the construction in inscriptions remains to be done. The changes affecting fore ut documented below are neither isolated nor peculiar to this construction. Rather, they are part of the larger pattern of evolution of the Latin future tense system in general and, in

6Szantyr (note 5 above) 642.

particular, are in some ways converse complements to changes in the future periphrastic (FAP + forms of sum). Limitations of space, how- ever, require that I postpone treatment of this question to a future study.

2. The synchronic state of fore ut through Livy. I review here the more important constraints on fore ut established in my earlier studies. As there, the constraints are given here in topical order. Constraints 1-5 are primarily syntactic in their formulation, and 6-11 semantic and prag- matic,' and relate to the semantic parameters of [control] and [inten- tion]. A predication has the feature value [+control] if it expresses a situation in which "one of the entities involved has the power to deter- mine whether or not that situation will ~btain."~

In particular [+control] predications have agents. The feature [intention] cross-classifies [+control] predications and refers to whether the situation is or is not intended by the agent. In particular [fintention] situations exclude such qualifiers as invitus (of the agent), imprudenter; casu, fortuito, etc. Documentation of these constraints will be found in my 1989 article. Constraints 12-20 concern epistemic and deontic modality. Epistemic modality concerns notions such as certainty, probability, possibility, etc., as opposed to factuality. Deontic modality covers notions such as behavioral (as opposed to logical) necessity, obligation, permission, etc. Documentation of constraints 12-20 will be found in my 1991 article. Of course, the syntactic constraints have semantic and pragmatic motiva- tions.

All of the constraints derive from the basic value of the construc- tion, which is to portray future situations as happenings rather than doings. When the construction is used with verbs which normally ex- press actions, the agents' control over and/or intention to perform the action is either obviated in the discourse context, de-emphasized, or otherwise not salient. This function, of course, is suggested by the construction's literal translation, "that it will be (so) that." The effect of this basic function can be seen most immediately in constraint 5 on prolepsis and constraint 9 on the expression of realization, below. Pro- leptic agents do not occur with fore ut, because prolepsis is a promi- nence device which would emphasize the prototypical characteristics

7"Pragrnatics can be usefully defined as the study of how utterances have mean- ings in situations," G. N. Leach (Principles of Pragmatics [London 19831 x).

". C. Dik, Functional Grammar3 (Dordrecht 1981) 32-33.

of an agent such as control and intention and would, therefore, be incompatible with a construction that indicates happenings rather than doings. Similarly, if the realization or non-realization of the predication embedded under fore ut is commented on, it is formulated in terms of an impersonal construction, e.g., id quod evenit, since use of a nor- mally agentival construction would contradict the portrayal of the situa- tion in question as a happening. Reference to the future quite generally involves modal notions such as possibility, probability, necessity, cer- tainty, confidence, doubt, etc., rather than pure indication of a future temporal relation, so that it is not surprising that there should be a modal component to the meaning of fore ut as well. Since expressions of certainty, by their very nature, de-emphasize notions of control and intention and tend to portray agents of actions rather as participants in events, there is a clear affinity between fore ut and epistemic modal qualifications such as 'certainly' and 'surely,' as reflected in constraints 12-14. Similarly, the same de-emphasis of agentival control and inten- tion creates an affinity between fore ut and the compulsive (as opposed to the morally obligative) type of deontic modality, as reflected in con- straints 19 and 20. Furthermore, the simple presence of an element etymologically meaning ". . . it will be . . ." could readily have been interpreted as an explicit expression of the "neu~tic"~

or "I-say-s~"~~ component of utterance meaning specifically indicating a positive quali- fication of the principal subject's commitment to the likelihood of the embedded predication. Such an interpretation leads not only to a modal value in general, but to the specifically subjective epistemic modality reflected in constraints 15 and 16.

Word order. The GV and fore always precede ut, which always pre- cedes the verb embedded under it; the GV and fore may occur in either order relative to each other.

No constituent embedded under fore ut may be a relative pronoun, whether in prolepsis or not, whether coniunctio relativa or subor- dinating relative.

fore ut is not co-ordinated by simple conjunctions, such as et, atque, -que, sed, with the FI, unless the verb embedded under it lacks a FI or supine; i.e., only necessary instances of fore ut may be co-ordi- nated with the FI.

9R. M. Hare, "Meaning and Speech Acts," Practical Inference (London 1971)74

93. loJohn Lyons, Semantics (Cambridge 1977) vol. 2. 798-809.

Simple negation such that non or neque has scope directly over fore or futurum (esse) does not occur.

The agent of the predication embedded under fore ut, whether ex- pressed as nominative subject or ab plus ablative, never occurs in prolepsis. Other constituents, however, particularly temporal ad- juncts may be proleptic.

As compared to the FI, verba dicendi, in contrast to verba sentiendi ac cogitandi, are avoided as GV's of fore ut.

As compared to the FI, the subject of the main verb and the subject of the verb embedded under fore ut rarely refer to the same person or entity.

Adjuncts of psychological manner, such as aequo anirno, libenter; studiose, gravius (withferre) are not admitted in predications embed- ded under fore ut.

If the later realization or non-realization of the future situation ex- pressed by fore ut is explicitly commented on, it is always so done by an impersonal construction lacking an overt agent. In other words, the realization of a predication embedded under fore ut is always presented as a happening rather than as a doing under the control of an agent.

fore ut is never used to express an action under the control of and intended by the subject of the GV, unless it is part of a request or other directive illocutionary act.

As compared to the FI, fore ut occurs far more frequently in three types of discourse situations: 1) polite expressions, including direc- tive speech acts; 2) portrayal of past instances of foresight, predic- tion, or prophecy which have proved valid; 3) assertions of certainty or conviction and assurances given to an addressee to induce him to some course of action.

The following constraints relate to modal semantics (epistemic mean- ings such as 'surely,' 'certainly' and deontic meanings such as compul- sion and obligation):

12. The free instances of fore ut are never combined with clauses which serve to weaken or make more cautious the epistemic commitment of the principal subject to the probability of the predication embedded under it. Letting WEC stand for a clause of weakened epistemic com- mitment, and I$ for the predication embedded under fore ut, we may represent constraint 12 as

*[WEC + GV fore ut I$], where * means 'does not occurlis not well formed.'

fore ut does not combine with non-harmonic, epistemic adverbs, such as fortasse, which suggest doubt about or improbability of the predication embedded under it.

fore ut does not combine with harmonic, epistemic verbs, such as certe, profecto, nempe, etc., expressing certainty or a high degree of commitment to the embedded predication. Letting E Adv stand for any epistemic modal adverb, constraints 13 and 14 may be repre- sented as

*[E Adv + GV fore ut +].
fore ut is never governed by a modal adjective such as probabile, manifesturn, or certum. Letting E Adj G stand for any governing phrase consisting of such an adjective, constraint 15 may be repre- sented as

*[E Adj G fore ut +I.
The (free) instances of fore ut are subjective rather than objective in their epistemic modal value; i.e., they express a (positive) qualifica- tion of the principal subject's commitment to the probability of the occurrence of the situation referred to by his utterance or thought, more like English adverbs such as surely, certainly, etc., than like phrases such as it is certain.11

When forms of posse are embedded under fore ut, they never have epistemic meaning (e.g., 'it is possiblelpracticable') but always refer to dynamic modality (e.g., 'canlis able'). In other words they refer to the ability or capacity of agents to act in a certain way rather than to the possibility of certain situations occurring.

fore ut is not governed by any form of scire.

fore ut may independently express deontic modality, but its deontic meaning is restricted to the sense of compulsion arising from some outside force or lack of alternative; fore ut does not express such deontic notions as moral or legal obligation. Constraint 19 may be represented as

[+deontic]: (fore ut +) 3[+compulsive]: (fore ut +).

llAccording to J. Lyons (note 10 above) 799, "it seems clear that the main differ- ence between subjectively and objectively modalized utterances is that the latter, but not the former, contain an unqualified, or categorical, I-say-so component. The speaker is committed by the utterance of an objectively modalized utterance to the factuality of the information he is giving to the addressee: he is performing an act of telling . . . objective modalization differs from subjective modalization, the very essence of which is to express the speaker's reservations about giving or qualified, or categorical, "I-say-so" to the factuality of the proposition embedded in his utterance. Subjectively modalized state- ments . . . are statements of opinion, or hearsay, or tentative inference, rather than state- ments of fact; and they are reported as such." Cf. also E. C. Traugott, "On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change," Language 65 (1989) 31-35.

20. Accordingly, the gerundlgerundive, debere, necesse est, oportet, and other verb phrases expressing deontic modality are not embedded under fore ut. Letting DVP stand for a deontically rnodalized verb phrase, constraint 20 may be represented as

*[fore ut + (DVP)].

3. Erosion of the constraints on fore ut after Livy. In this section I proceed seriatim through the violations of the constraints listed above as they begin to occur after Livy.

3.1. Word order. fore is never found following ut. The only violation of the constraint that the GV must precede ut and the embedded verb occurs at Ovid. her 16.279-80:

1. hoc mihi-nam repeto-fore ut a caeleste sagitta
figar, erat verax vaticinata soror.

This violation is best regarded as arising from metrical requirements of the elegiac couplet, rather than as a change which has become accept- able in the literary register during the half-generation or so separating the births of Livy and Ovid: it is never found in prose.

It is interesting to note that the discourse situation of the violation in example 1remains that of the characteristic type 2, prophecy proved valid; cf. the immediately preceding verses 277-78. In fact, aside from this violation, Ovid observes all the syntactic, semantic, and pragmatic constraints on fore ut in force before Livy.

3.2. Relativization. This constraint is never violated. The preservation of this constraint is fairly readily explained. The relativization of any constituent of the predication embedded under fore ut would produce a species of relative Verschrankung as in

2. aberat omnis dolor, qui si adesset nec molliter ferret; . . . (Cic.,jn. 2.64)

Relative Verschrankung, however, like coniunctio relativa, declines in frequency in later Latin.

3.3. Non-harmonic co-ordination. This constraint is violated only at Suetonius 2.97:

3. sub idem tempus ictu fulminis ex inscriptione statuae eius prima nomi- nis littera effluxit; responsum est, centum solos dies posthac victurum, quem numerum C littera notaret, futurumque ut inter deos referre- tur, . . .

The constraint against non-harmonic conjunction arises from the se- mantic and pragmatic differences between fore ut and the FI: simple co-ordination is, in general, used for the connection of semantically more similar and pragmatically more closely parallel items, so that sim- ply co-ordinated predications are less likely to differ in those features that would motivate switching between fore ut and the FI. Accordingly, in 4 we can observe Livy changing from fore ut to the FI in predications linked by temporal inde, when the subject of the second becomes the same as that of the principal subject and the action referred to is under the control of and intended by the principal subject:

4. sperantes fore ut Romani equites abeuntium novissimum agmen aggre- derentur; inde . . . se, qui equitatu et levi armatura plus possent, con- versuros aciem. (Livy 42.57.11)

Thus, unnecessary non-harmonic co-ordination of fore ut with the FI is evidence for the effacement of the differences between fore ut and the FI. The violation in 3, however, is basically syntactic and does not involve the motivating differences in terms of the semantic parameters of [control] and [intention] of the principal subject. victurum is not under Augustus' control, and he is the patient of the passive verb referretur; not its controlling agent. Even so, it is probably not chance that fore ut is not found co-ordinated with the FI until after the appearance of considerable evidence for the semantic and pragmatic merger of fore ut with the FI, as represented by the effacements of constraints 4, 6, 9, 10, 11, 12, 15, 16, 17, and 18 (see the respective subsections).

3.4. Negation. This constraint is violated at Valerius Maximus 4.1.12:

5. idem filios suos monuit ut funebri eius lecto humeros subicerent, atque huic exequiarum illum honorem vocis adiecit, non fore ut postea id officium ab illis maiori viro praestari posset.

[Quint.] decl. maior 4.19:

6. mathematicus hoc non futurum dixit ut vellem. sed ut occiderem .

is not a violation, since the scope of non is over dixit, not futurum: "he did not say X, but Y."

The constraint against direct negation of fore is related to the difference between subjective and objective modality and is exactly parallel to the constraint against direct negation of the modal verbs debere and oportet in their epistemic meanings of 'surely, certainly' (but not in their deontic meanings of 'ought, should'). This more gen- eral constraint derives from the subjective nature of the epistemic mo- dality expressed by these forms. l2 In contrast, the objective epistemic modal necesse est may be directly negated. Similar constraints on the negation of subjective epistemic modal expressions can be found in other languages. For example, in English, the modal adverbs certainly, surely, probably, possibly, etc., cannot be directly negated, whereas the objective modal phrases of the corresponding adjectives plus forms of to be may be so negated. Thus violation of constraint 4 indicates that

fore ut is losing its primarily subjective character.

3.5. There is no violation of the constraint against prolepsis of the agent of the predication embedded under fore ut. The preservation of constraint 5 is not surprising, since prolepsis in general declines in the literary register, except for archaizing authors. (It remains frequent, of course, in registers closer to vulgar Latin, as in Chiron.) The general decline in prolepsis of all types of constituents of predications embed- ded under fore ut is striking. In the material from Plautus to Livy, obligatory constituents, such as the semantic patient of the embedded verb sometimes occur in prolepsis, e.g.,

7. ea comitia puto fore ut ducantur. (Cic., Att. 90.7)

Prolepsis of obligatory constituents does not occur in the material from Ovid to Festus (but prolepsis of a patient constituent is found in the Vulgate at Luke 9:44). Whereas in the earlier period approximately half of optional adverbial, prepositional, and nominal constituents having the semantic functions of means, manner, cause, result, attendant cir- cumstance, and the like occur in prolepsis, none are so found in the later material, except for ita (twice) and sic (once). (certe at Festus 464.1-32 will be discussed at 3.14 below.) Temporal adjuncts, such as

I2Cf. A. M. Bolkestein, Problems in the Description ofModal Verbs (Assen 1980) 72-73, 100-101, 116, 130-31.

tarde, protinus, brevi, etc., are the only semantic type which remains more or less admissible in prolepsis, but the proportion of temporal adjuncts in prolepsis declines significantly to about 21% from 55% in the earlier material.

3.6. verba dicendi as GV's. Verb phrases of saying, writing, predicting (i.e., verba dicendi in the wide sense), as opposed to verb phrases of thinking, believing, knowing, certainty, hoping, etc., (i.e., verba sen- tiendi ac cogitandi in the wide sense) as phrases governing fore ut increase in frequency after Livy. In the later period the overall odds for verba dicendi (55% to 45%) increase two and a third times over the odds from Plautus to Livy (34% to 66%), a statistically significant difference. In the earlier period, verba dicendi are more frequent governing neces- sary or passive instances of fore ut than governing the free, active in- stances. Proportionally, the greatest increase in frequency in the later period occurs in just the class of free, active instances. The restricted occurrence of verba dicendi, particularly with the free, active instances offore ut, in the earlier period is in part a consequence of constraint 10: if available, the FA1 is used instead of fore ut when a person asserts that he will intentionally perform an action which is under his control. The increase in frequency of the use of fore ut after verba dicendi indicates its spread to contexts which are likely to undermine constraint 10.

3.7. Co-reference of principal and embedded subjects. Among the free active instances of fore ut dependent on GV's with personal subjects, the rate of coreferentiality between the principal and embedded sub- jects does not change significantly.

3.8. Adjuncts of psychological manner. Optional constituents of this semantic class continue to be excluded from predications embedded under fore ut. The relative proportions of the class of temporal ad- juncts, on the one hand, and the more open class of manner adjuncts, means, cause, result, etc., on the other remain basically unchanged: respectively 54.55% and 45.45% from Ovid to Festus and 46.15% and 53.85% from Plautus to Livy. The one important development in regard to adjuncts is the appearance of the epistemic modal adverb certe at Festus 150.25L, which will be discussed at section 3.14.

3.9. Realization of the embedded predication. This constraint against the expression of the realization of a predication embedded under fore ut as an action under the control of the agent of the embedded predica- tion cannot be directly tested, since explicit authorial comments on the realization of the predications in question do not occur in the material from Ovid to Festus. It is, however, encroached upon in the reverse situation, in which a completed [+control] action is characterized as having been predicted. In the material from Plautus to Livy, this reverse situation always requires the FAI, not fore ut, e.g.,

8. cum illi ipsi venissent quos ego iam multis ac summis viris ad me id temporis venturos esse praedixeram. (Cic., Catil. 1.10.5)

Tacitus, however, uses fore ut to describe the prediction of the murder of Agrippina after the event.

9. hunc sui finem multos ante annos crediderat Agrippina contempserat- que. nam consulenti super Nerone responderunt Chaldaei fore ut im- peraret matremque occideret; . . . (Tac., ann. 14.9)

Example 9, nevertheless, is not a full violation, since, in its textual situation, the completed action is not described by a personal [+con- trol] verb, but referred to by the noun phrase hunc$nem.

3.10. Actions intended and controlled by the principal subject. To- gether with constraint 9, the constraint against using fore ut to express an action under the control of and intended by the subject of the GV is the semantically most revealing. It points to the general association of the construction with discourse/textual situations in which it is more appropriate to portray a situation as an event that will happen rather than an action that an agent will perform under his own control. Such an association is only to be expected from the etymological meaning of forelfuturum ut (in oratio obliqua) "that it will be (so) that. . . ." I have argued that the construction developed by gradual, analogical general- ization from a prototypical use for only [-control] situations. By the time of Cicero, the free use of the construction had spread to verbs which in normal textual situations express [ + control, + intention] ac- tions, but only in the very restricted pragmatic context of polite re- quests. To understand the processes which led to this state of the con- struction and which set the stage for the erosion of constraint 10, it is necessary to examine the pragmatic and textual situations in more de- tail than provided by my 1989 article. The earliest attestation of fore ut occurs in a request; it is a necessary instance since the embedded verb

has no FI; and furthermore that verb is fieri, so that the situation is explicitly portrayed as a [-control], and therefore [-intention], event which, in the context, is obviously undesired:

10. (Sirno) hoc ego nurnquarn ratu' sum fore me ut tibi fierern supplex . . . nonne audes, quaesso, aliquarn partern rnihi gratiarn facere hinc de argento? (Plaut., Pseud. 1318-19, 1322)

In the same pragmatic situation, with the identical governing phrase, the construction was generalized to free instances of [ -control] verbs as in

11. nurnquarn ego ratus sum fore ut rex rnaxirnus in hac terra et ornniurn, quos novi, privato hornini gratiarn deberern. (Sall.; lug. 110.1)

where the situation is also clearly not desired by the principal subject. Again, in the same pragmatic situation and with a nearly identical gov- erning phrase, the construction spread to verbs which are normally [+control, tintention], but in such a way as still to portray the action of the embedded predication as unintended, and in fact the result of circumstances beyond the agent's control (note patior), as in

12. nurnquarn putavi fore ut supplex ad te venirern; sed hercule facile patior datum ternpus in quo arnorern experirer tuurn . . . arnabo te, da rnihi et hoc . . . (Cic., Att. 407C.l)

The next stage is represented by the spread of the construction from the introduction to a request to its conclusion: thanks in advance, as it were, as in

13. etsi iarn sperabarn, cum has litteras accepisses, fore ut ea quae supe- rioribus litteris a te petissernus irnpetrata essent, tarnen non faciarn finern rogandi quoad nobis nuntiaturn erit te id fecisse quod rnagna cum spe expectamus. deinde enirn confido fore ut alio genere lit- terarurn utarnur tibique pro tuo surnrno beneficio gratias agarnus.

(Cic., Att. 407E.2)

In 13 the predication embedded under the second occurrence of fore ut is no longer portrayed as undesirable or unintended. Yet, since compli- ance with a request (impetrata) is a felicity condition for the speech act of thanking (gratias agarnus), the situation remains contingent, and not actually under Cicero's control; pragmatically, it is a polite way for Cicero to express his confidence that Atticus will comply with his re- quest.

It should also be noted that there is an additional constraint on the use of fore ut with normally [+control, +intention] verbs in request situations, which I did not note in my 1989article. Those situations are always direct, first person requests, or imperatives, not reports or por- trayals in the second or third person of such requests by someone other than the original speakerlwriter.

The synchronic state of the construction attained in Cicero is ob- viously unstable, and one would expect further generalization of the range of usage. A spread to third person, reported situations, which, moreover, while still directive in illocutionary force, lack the element of politeness and de-emphasis on intention, is found in the minor decla- mations attributed to Quintilian.

14. maior frater minorem in adulterio deprehendit. pro (eo) rogante patre et abdicationem eius promittente, dimisit . . . quid enim mihi pol- licitus est pater? si dimisissem, futurum ut abdicaret fratrem meum.

([Quint.] decl. minor 275.0-3W)

This degree of violation of constraint 10 will be denoted by V(lO.l). The next stage in the generalization of the construction that might be ex- pected would be the effacement of the restriction to contexts of direc- tive illocutionary force. Frontinus provides an instructive example of the processes by which this restriction was lost in the literary language. Consider

15. Ti. Gracchus, cum edixisset futurum ut ex volonum numero fortibus

libertatem daret, ignavos crucibus adfigeret . . . (Frontinus, strut. 4.7.24)

Here daret and adJigeret are both [+control] and [+intention]. More- over, there is no overt indication that the context involves directive illocutionary force. Exactly the same episode is treated by Livy in 16, and he uses the FI, not fore or futurum ut:

16. qui caput hostis rettulisset, eum se externplo liberum iussurum esse;

qui loco cessisset, in eum servili supplicio anirnadversururn; . . . (Livy 24.14.7)

From Livy's account, it is clear that Tiberius Gracchus' words are part of an exhortation before battle, and thus, in fact, situated in a context of directive illocutionary force. It is clear, therefore, that Livy had not generalized the use of fore ut to third person, reported directive con- texts as observed in the minor declamations attributed to Quintilian (14 above). Furthermore, the narrative at Livy 24.14 is an unlikely textual situation for such a generalization, so long as fore ut maintained its association with de-emphasis on the control and intention of the agent, for Livy explicitly remarks on Tiberius Gracchus' desire to grant free- dom to the volones (cf. 24.14.3-4). In contrast, Frontinus gives no indi- cation that the context of edixisset is adhortatory. Nevertheless, the departure from constraint 10 is minimal, since Frontinus and his audi- ence would either have known or could have conjectured the fuller context. Thus, in effect, we have the extension of a construction war- ranted in a specific type of pragmatic function, as in 14, to a textual situation in which that function is no longer salient, but at the same time, with which it is not incompatible. This degree of violation of constraint 10 will be denoted by V(10.2). It is well known that much semantic change takes place through gradual, contextually conditioned loss of salience of the distinctive values of semantic features, rather than by one step, categorical reversal of the values of those features: lingua non facit saltus. The complete loss of constraint 10 (denoted by V[10.3]) is also attested in Frontinus

17. Cyzicum cum oppugnaret Mithridates, captivos eius urbis produxit ostenditque obsessis, arbitratus futurum ut miseratione suorum com- pelleret ad deditionem oppidanos; . . . (Front., strat. 4.5.21)

In contrast to 16,17 does not contain any performative speech act at all. It is clear from the context and from the ablative of means miseratione (since adjuncts in the semantic function of means may occur only in [+control] predications) that the embedded predication must be [+control, +intention]. At this stage, forelfuturum ut has merged with the FI in terms of pragmatic function and of the semantic parameters of agent coreferentiality, [control] and [intention].

Since it is the tendency to associate fore ut with the portrayal of situations as events not under the control of agents that motivates con- straint 9, which requires the realization of predication embedded under

fore ut to be stated via impersonal, [-control] constructions, that asso- ciation would have to be weakened before constraint 9 was lost. Conse- quently, while there is evidence for only a partial loss of constraint 9, it is only to be expected that violation of constraint 10, which is direct evidence for the disappearance of the association, should precede the partial violation of constraint 9. Furthermore, since constraint 3 against co-ordination of fore ut with the FI rests on the semantic and pragmatic differences between them, it is not surprising that both constraints 10 and 9 should appear at least partially violated before constraint 3. Moreover, since it has been possible to distinguish three degrees of violation of the semantic substance underlying constraint 10, the follow- ing order of violation of the constraints in question may be taken as well motivated:

where the decimals refer to the degree of violation and > means 'is a necessary stage in the development toward and does not chronologi- cally follow.'

3.11. Discourse situations. In comparison to the material from Plautus to Livy, in the period from Ovid to Festus there is a decline, though small, in the proportion of instances of fore ut which occur in discourse situations of Type 1, polite expressions, especially requests; Type 2, portrayals of foresight, prediction, or prophecy which have proved valid; and Type 3, assertions of certainty or conviction and assurances given to an addressee to induce him to some course of action. More important is the change in the relative frequency of these three types. Type 1 decreases substantially, and Type 2 increases greatly, accounting for, respectively, 14% and 54% of these types. The increase in Type 2 is not surprising given the increased interest in omens and prophecy of the writers in the later period; in fact, two-thirds of the instances in Sue- tonius are of Type 2. Striking also is the emergence of two new, well- defined types of discourse situation. The first constitutes a generaliza- tion of Type 3 situations to threats. The use of forelfutururn ut to ex- press the content of a threat is found first in Valerius Maximus

quo aperte denuntiabat futurum ut spiritum poenae inpenderent, quem pugnae dubitassent. (Val. Max. 2.7.ext.2)

nec multo post in senatu Pompeio cuidam equiti Romano quiddam perneganti, dum vincula minatur, affirmavit fore ut ex Pompeio Pom- peianus fieret, acerba cavillatione simul hominis nomen incessens veteremque partium fortunam. (Suet. 3.57)

Note also the very clear case The use of the construction in threats is a genuine innovation, since in the period from Plautus to Livy only the FI is used, and not fore ut, e.g.,

19.a. huc nisi venirem Kalendis Septembribus, etiam fabros se missurum et domum meam disturbaturum esse dixit. (Cic., Phil. 5.19)

Cf. Cic., Phil. 1.12.

19.b. ex qua se instructum et paratum ad urbem venturum esse minitatur. (Cic., Phil. 3.1)

The second new situation is the presentation of the content of rumor: This usage occurs in the minor declamations attributed to Quintilian

20.a. rumor erat futurum ut pauperis filia sacerdos crearetur. (decl. rnin. 252.0W)

but is most widespread in Tacitus, e.g.,

20.b. ac tum rumor incesserat fore ut disiecti aliisque nationibus permixti diversas in terras traherentur. (Tac., ann. 4.46)

20.c. mox, quia rumor incedebat fore ut nuru ac nepoti conciliaretur Caesar, saevitiam quam paenitentiam maluit. (Tac., ann. 6.23)

Cf. also ann. 1.5, hist. 1.54. When rumor eratlincedebatl etc., governs fore ut, the verb of the embedded predication is always passive. The use offore ut after rumor is a genuine innovation, since the use of the supine

+iri in this function had been established since Terence:

21. quom interea rumor venit datum iri gladiatores, populu' convolat, . . . (Ter., Hec. 40)

3.12. Combination with clauses indicating weakened epistemic com- mitment of the principal subject. Constraint 12 arises from the fact that fore ut, when it is used in an epistemic modal meaning, implicates a

strengthened degree of commitment to the proposition embedded under

it, which the FI does not implicate.

3.12.1 Position on the modal epistemic scale. In the period from Plau- tus to Livy, the epistemic value of fore ut corresponds fairly closely to the degree of qualification indicated by English modal adverbs cer- tainly, surely. Thus violation of constraint 12 is evidence of the spread of the construction down the scale of epistemic modality to lower levels of (subjectively assessed) probability of occurrence of the situation in question. (For such a scale, see Ultan.13) The epistemic modal meaning is beginning to extend also to the degree of qualification indicated by English likely, probably. The first violation of constraint 12 occurs in Seneca the Younger:

22. non equidem reor
fore ut recuset ac meos spernat toros;
quod si impotenti pertinax animo abnuet . . .

(Sen., Herc. fur: 348-50)
3.12.2 Further epistemic demotion. Once usages such as 22 become established, the next stage in the epistemic weakening of the construc- tion is its ability to signal notions such as 'likely' independently of any additional, overt qualification. This stage can be observed in Quintilian, at least in an indirect question with futurum sit ut.

23. sed in iis quoque quae constabit posse fieri coniectura aliquando erit, si quaeretur an utique futurum sit ut Carthaginem superent Romani; ut redeat Hannibal si Scipio exercitum in Africam transtulerit; ut servent fidem Samnites si Romani arma deposuerint.

(Quint., inst. 3.8.17)
The larger context of 23 is extremely valuable for assessing the modal semantics of forelfuturum ut at this period, since in it Quintilian pro- ceeds systematically through the epistemic modal scale from possibility (posse Jieri) to probability (futurum sit ut).

'3R. Ultan, "The Nature of Future Tenses," Universalsof Human Language vol. 3, ed. J. H. Greenberg, et al. (Stanford 1987) 83-123.


24. rem de qua deliberatur aut certum est posse fieri aut incertum. si incertum . . .cum autem de hoc quaeritur, coniectura est: an Isth- mos intercidi, an siccari palus Pomptina, an portus fieri Ostiae possit, an Alexander terras ultra Oceanum sit inventurus.

(Quint., inst. 3.8.16)
It will be noticed that in these indirect questions, where it is possibility that is in question, Quintilian expresses the future temporal relation by the future periphrastic (sit inventurus), not by the subjunctive equiva- lent of our construction, futurum sit ut plus subjunctive verb. In clear contrast, in 23, where he treats situations, the possibility of which is certain (constabit possejeri), he uses futurum sit ut for the correspond- ing indirect questions, not the future periphrastic. Thus, as the Loeb translation with likely makes clear, futurum sit ut is distinguished from the FAP plus sit as probability is distinguished from possibility. Nev- ertheless, the entire context is concerned with coniectura and thus en- compasses a lower segment of the epistemic modal scale than that with which forelfuturum ut is associated in the period from Plautus to Livy (contrast, e.g., Cic., fam. 22.5.4 and acad. 2.100 with the FI).

It should be clearly understood, moreover, that the semantic change just documented is one of extension rather than replacement. At least in its corresponding finite form, the construction can still indepen- dently indicate the higher (or "certainty") end of the epistemic modal scale, as its use in parallel with epistemic necesse est in Cornelius Celsus shows:

25. quamvis autem non abscisus nervus est, tamen si circa tumor durus diu permanet, necesse est et diuturnum ulcus esse et sano quoque eo tumorem permanere; futurumque est ut tarde membrum id vel ex- tendatur vel contrahatur. (Celsus 5.26.28B)

3.13 Non-combination with non-harmonic modal adverbs is readily explained by the association of fore ut with the highest end of the epi- stemic modal scale and by the fact that modal adverbs and the construc- tion both qualify the principal subject's epistemic commitment and thus, both, have co-extensive scopes. Consequently an occurrence with non-harmonic modal adverbs such as fortasse would produce an incompatibility similar to English perhaps surely or even perhaps prob- ably. Even with the extension of fore ut to objective epistemic modal uses (see section 3.14 below), the constraint against combination with non-harmonic modal adverbs would be expected to remain in force, since, in general, as BolkesteinI4 remarks, "it is less acceptable to sub- jectively specify one's commitment to a probability (of some state of affairs being the case) as a weak one when the probability is 'objec- tively' stated to be a strong one . . . than it is the other way round: to subjectively specify one's strong commitment to a probability objec- tively stated to be weak. . . ." Accordingly, no violation of constraint 13 is found in the material from Ovid to Festus.

3.14. Combination with harmonic modal adverbs. The constraint against co-occurrence offore ut with harmonic epistemic adverbs such as certe and profecto, unlike constraint 13, is not motivated by strong and outright semantic incompatibility between the high and low ends of the epistemic modal scale. Rather, it may be the result either of the redundancy of such combination or, perhaps, of the fact that fore ut occupies a narrower portion of the scale so that such apparently har- monic modal adverbs do not actually reinforce it. fore ut shares con- straint 14 with oportet, in its epistemic meaning, but not with debere. Whatever form the ultimate, exact, explanation of constraint 14 may take, it does not seem to be related to the difference between objective and subjective modality. A difference in the scope of the modal ele- ments is not necessary for semantic compatibility in the case of har- monic reinforcement, since there is a single modality running through the whole predication, although expressed in more than one place. Fur- thermore, there is no evidence that, in combination with harmonic mo- dal adverbs, epistemic debere always requires an objective interpreta- tion.

In section 3.12 we have seen evidence for the extension of fore ut to a lower portion of the epistemic modal scale. When this extension becomes established, it ceases to be redundant or uninformative to combine the construction with harmonic adverbs such as certe, since they can now reinforce its modality. Such combination is indeed found, but, in accordance with the evidence from Celsus for the persistence of the association of fore ut with the uppermost end of the scale, it does not occur until Festus:

I4Bolkestein (note 12 above) 74.

26. rusum itaque consultus Apollo respondit, non esse persolutum ab his votum, quod homines immolati non essent: quos si expulissent, certe fore ut ea clade liberarentur. (Festus 150.23-26L)

3.15. Modal adjectives governing fore ut. Like English modal ad- jectives plus that-clauses (type it is certainlprobable that . . .), Latin (neuter) modal adjectives plus accusativus cum injinitivo are objective modals; i.e., they are used for the unqualified assertions of some degree of probability of a situation rather than for the assertions of some situa- tion, the speaker's commitment to which he qualifies to some degree. The fact that, from Plautus to Livy, they may not govern fore ut can be explained as a result of scope conflict. It is an apparently language universal rule that subjective epistemic modality must have wide scope, in particular scope over any objective modality. Structures such as cer- tum erat fore ut would violate this rule by placing the subjective modal element within the syntactic scope of the objective element. Thus the occurrence of fore ut governed by a (neuter) modal adjective is evi- dence that the construction has acquired objective as well as subjective senses. Such a structure occurs first in Valerius Maximus

27. nam cum ad excidium eius summo studio Alexander ferretur pro- gressumque extra moenia Anaximenen praeceptorem suum vidisset, quia manifestum erat futurum ut preces suas irae eius opponeret, non facturum se quod petisset iuravit. (Val. Max. 7.3.ext.4)

The next occurrence is in the minor declamations attributed to Quin- tilian.

28.a. cum interrogasses, certum erat fore ut protinus cupiditas aliqua in

animum tuum descenderet; et, cum concupisses, ut raperes. (decl. min.301.16W)

There is an important difference between the discourse situations in examples 27 and 28.a which suggests the way in which this change in usage developed. The textual situations such as that in 27 must have been an important factor promoting the generalization of fore ut to objective, epistemic modal uses. Grammatically impersonal expres- sions of evidentness or obviousness may be narratively salient because they implicate a subjective state of certain belief on the part of a partici- pant in that narrative, and thus overlap in discourse function with an explicit portrayal of that state via typically subjective GV's. Conse- quently, the use of forelfuturum ut in such situations would not consti- tute a complete violation of the semantic and pragmatic substance of constraint 15. This contextually mitigated degree of violation of con- straint 15 will be represented as V(15.1). The complete violation as seen in 28.a will be represented as V(15.2). Furthermore, a stage intermedi- ate between 27 and 28.a can be observed in Celsus:

28.b. certaque esse fiducia potest fore ut undique vitiosa car0 exci- dat . . . (Celsus 5.28.1D)

where there is no narrative participant involved and the modal adjective is predicative on$ducia. It is surely not accidental that the first occur- rence of forelfuturum ut embedded under a neuter adjective of epi- stemic character occurs in just such a situation of overlap as in 27, well before such embedding in a fully objective discourse situation as in


3.16. ~estriction'to subjective epistemic modality. We have seen al- ready in the preceding section the first stage in the extension of fore ut to objective modal sense, namely its use after neuter modal adjectives which in their textual situations implicate the subjective commitment of a narrative participant to the embedded predication (V[15.1]). Fully ob- jective usage was facilitated by the development of the finite counter- part of infinitival forelfuturum ut. l5 Examples have already been give at 23 and 25 (V[16]). As has also been noted above, constraint 4 against direct negation of fore ut is connected with its subjective character. Just as V(15.1) is evidence for a growing affinity for objective uses, so is V(4). While there is no reason to expect V(4) either to precede or follow V(15.1), both may be expected to precede V(16). Furthermore, since the development of an objective epistemic value of fore ut is necessary if it is to be governed without scope conflict by a neuter epistemic adjective in textual situations which do not implicate the subjectivity of a narra- tive participant (V[15.21), we may take the following partial ordering of the appearance of the violations to be well motivated:

15The first occurrence appears to be Cic., Flacc. 2: quod si esset aliquando fu- turum ut aliquis de L. Flacci pernicie cogitaret.

3.17. Dynamic posse.

It is obvious that by posset in this passage, Metellus is not referring to the ability or capacity of his sons to carry a funeral bier or perform funeral rites in the future. Rather, he is asserting that it is not possible for a situation to come about in which they will do so for a greater man. Consequently, the force of posset here is epistemic rather than dy- namic. Violation of constraint 17 constitutes a generalization of the epistemic modal uses of fore ut, since with it the construction can appear in utterances containing two modalities, an objective one ex- pressed by posse and a subjective one potentially expressed by fore ut.

3.18. scire governingfore ut. In its basic uses scire is a factive verb,16 in that by using it in normal contexts a speakerlwriter commits himself to the truth of the accusativus cum injinitivo predication embedded under it. Whether or not justifiable in epistemology, the future can be treated as known in Latin as well as English, so that the FI is perfectly accept- able after scire. Lyons17 characterizes 'know' as follows: "if [a speaker] employs 'know' rather than 'believe,' he is making a stronger commit- ment: by employing 'know' he is claiming that his belief in the truth ofp is well grounded and in his judgement at least unassailable, and by virtue of this fact, which he should be able to substantiate, if called upon to do so, by providing the evidence, he has the right to assert p and to authorize others to subscribe to its truth." Nevertheless, know or scire epistemically modalizes an utterance, since each introduces an explicit epistemic qualification. The fact that fore ut, in the period from Plautus to Livy, is never governed by scire, but is governed by verba sentiendi ac cogitandi in the "belief" range, such as conjidere, putare, reri, etc., provides an upper bound to that part of the epistemic modal scale to which the construction is applicable. It is associated with cer- tainty and high degrees of commitment, but not with factivity.

160n factivity see P. and C. Kiparsky, "Fact," Progress in Linguistics, ed. M. Bierwisch and K. Heidolph (The Hague 1970) 143-73, and reprinted in Semantics, ed. D. Steinberg and L. Jacobovits (Cambridge 1971) 345-69.

17J. Lyons (note 10 above) 794.
scire, like English know, however, is not factive (i.e., does not commit the speakerlwriter to belief in the truth of its complement) in all discourse or textual situations. It may be used ironically or in hyperbole (as commonly with English just know). Such pragmatically non-factive usages could provide the entry point for embedding forelfuturum ut under scire. Just such a usage is found in Seneca the Younger's anec- dote about a Pythagorean:

30. sutor ille, quem quaeris, elatus, combustus est; quod nobis fortasse molestum est, qui in aeternum nostros amittimus tibi minime, qui scis futurum ut renascatur. iocatus in Pythagoricum. (Sen., de ben. 7.21.1)

Embedding under scire, however, was already well established in his father's time, e.g.,

31. si hic desiero, scio futurum ut vos ill0 loco desinatis legere quo ego a scholasticis recessi; . . . (Sen., suas. 6.27)

It is important to realize just how significant an innovation Seneca the Elder's is in comparison to earlier usage. Taking into consideration all types of the construction, necessary, free, with active, and with pas- sivized verbs, I find no case of fore ut embedded under scire in 123 instances from Plautus to Livy. By contrast three of the nine cases in Seneca the Elder are embedded under scire. If in fact there had been no change, and scire plus fore ut were uniformly acceptable in the earlier period and in Seneca the Elder, there would be a chance of only one in five thousand (p = 129! 9!/132! 6!) that all three instances should at random be restricted to Seneca the Elder.

Violation of constraint 18 is evidence for the generalization of fore ut up the epistemic modal scale. Moreover, from theoretical considera- tions, the loss of constraint 18 likely established the conditions which led to the generalization of fore ut down to a lower portion of the epi- stemic scale, as seen in section 3.12. So long as fore ut was restricted to governing verba sentiendi in the "belief" range of the epistemic scale, it contrasted with the FI more or less as certainty versus probability (and lower points), but did not itself occur in saliently different epistemic ranges. When, however, it became established as a complement to scire, there arose a contrast between utterances of the types scio fore ut and puto fore ut, which did contrast in terms of a salient difference in epistemic range. Furthermore, the new contrast scio fore ut : puto fore ut stood in a proportional semantic analogy with scio FI : puto FI. At this stage, it was but a small step to extend the analogy to a further term involving contexts of weakened epistemic commitment (WEC). As noted in section 3.17, from occurrences in the context of overt WEC, the construction was further extended to independent use in the lower, 'likely' range of the scale. After that usage was established, it could then be informatively reinforced by epistemic modal adverbs in the 'surely' range of the scale. Thus the effacement of constraint 18 could have been the starting point for a whole series of modal semantic changes as evidenced by the violation of the following constraints, the degree of violation being indicated by the decimal:

I believe that it is quite likely that this was in fact the actual causal sequence of events. The data are certainly in accord with the hypoth- esis. The embedding of fore ut under scire is well established before the first evidence of the combination of fore ut with clauses of weakened epistemic commitment. fore ut continues to be associated with the higher end of the epistemic modal for some time. Finally, modal re- inforcement of fore ut does not occur until the end of the later period. Furthermore, it is a priori likely that the loss of constraint 18 was a smaller and easier transition than the loss of constraint 12, for, as has been observed above, there are fully warranted discourse situations in which embedding fore ut under scio would not have been an outright violation of constraint 18, whereas situations promoting loss of con- straint 12 seem less readily available.

3.19-20. Deontic modality. For a long time into the period after Livy, fore ut continues to be able to signal independently of any other overt

element the compulsive sense of deontic modality, as in

32. cumque vulgo fore praedicarent ut si privatus redisset, Milonis exem- plo circumpositis armatis causam apud iudices diceret. (Suet. 1.30)

Nevertheless, just as its original range along the epistemic modal scale is gradually widened, so too is its range along the deontic. Evidence for its deontic generalization, however, is not forthcoming until Festus:

33. sacrificium fit cap(ite aperto) . . .Metellus pontifex (maximus Clau- dium augurem iussis)set adesse[t], ut eum . . . (Su1)pici Ser. f. inaug(urati0) . . . ret se sacra sibi fam(i1iaria . . . sup)plicandum es- set capite . . . esset, futurum ut cum ap(erto capite) . . . facienda es- set, pontif(ex) . . . Claudius provocavit. (Festus, 462.28-361464.12)

Although the passage is badly mutilated, its content can be recon- structed,lS and it is absolutely certain that the gerundive facienda esset is embedded under futurum ut. Thus passage 33 constitutes a violation of constraint 20. Moreover, and more importantly for the semantics, the sense of the gerundive in 33 is not that of compulsion arising from some outside force, but rather of a religious obligation. Consequently 33 is also evidence of the erosion of constraint 19.

The violations of constraints 19 and 20 are unlikely to have been abrupt innovations by Festus. In the case of his violation of constraint 14, concerning epistemic modality, it was possible to document the in- tervening stages in the evolution of fore ut which led up to that point. We may suppose that there was a parallel, gradual evolution of the deontic modality of fore ut leading up to 33.

The restriction of fore ut to the compulsive sense of deontic mo- dality arises from the association of the construction with situations neither controlled nor intended by an agent. Certainly a special case of such situations are those in which an agent acts under outside compul- sion. Furthermore, it would be impossible for an agent to assert (or be portrayed as asserting) a moral obligation which he fully accepts and intends to fulfill, as Claudius does in 33, without seriously encroaching on the semantic basis of constraint 10. Therefore, violation of con- straint 19 or 20 would not be expected to occur until after the complete loss of 10. Since three degrees of violation of 10 have been differenti- ated, we may take the ordering:

to be motivated.

IsCf. the supplements of Theodor Mommsen, Romisches Staatsrecht3 2.Bd.l.Abt. (Leipzig 1887) 35, n. 1. See also F. Bona, "Aeterio Capitone et Fest. 462.28L, (Saturno) sacrificium fit cap(ite aperto)," SDHZ 29 (1963) 316-25; B. Gledigow, "Condictio und inauguratio. Ein Beitrag zur romischen Sakralverfassung," Hermes 98 (1970) 369-79; and

J. Linderski, The Augural Law (ANRW 11, 16.3 [I9861 2146-2312) 2219, n. 275.

4. Distribution of violations. In the 93 instances of fore ut or its finite counterpart, 14 of the 18 categorical constraints discussed above are found to be violated at least to some degree. Clear cases of these viola- tions total 33, or about 10 violations of some kind for every 28 occurrences of the construction (rate = .355). That none of these 33 instances of violation occur in the even more extensive data-set from Plautus to Livy is obviously statistically significant. These violations, moreover, are not randomly distributed over the authors and works considered or in relation to each other. Their distribution may be por- trayed by matrix in which each row corresponds to a degree of violation of a constraint and each column to an author or work with the number of violations of each type in each author or work given in the cell at the intersection of the corresponding row and column.

It is, unfortunately, not possible to arrange the authors and works according to a completely well-ordered, rigid chronology. The major uncertainty concerns the minor declamations attributed to Quintilian. Winterbottom19 believes that "in all probability they will postdate the Znstitutio (c. 90 A.D.)" and considers the lexical and syntactic evidence to be consistent with a dating to the second or even first century. The problem is exacerbated by the possibility that the Declamationes mi- nores may contain considerably later material. Consequently, if a viola- tion is first attested in them, it may be concluded only that it does not occur until after Quintilian but not that it occurs before another author. To emphasize this limitation, a vertical line has been inserted down the right hand side of the column corresponding to the Declamationes mi- nores. The ordering of the remaining authors in whom violations are attested is on the whole unproblematic except in the case of two pairs. The relative chronological positions of Valerius Maximus and Cornelius Celsus and of Quintilian and Frontinus (both born ca. 35 A.D.) are quite close respectively. In each of the two pairs, the author writing in the relatively more technical genre has been placed after the author writing in the relatively more literary one on the grounds that there is ample evidence from the historical study of other syntactic and semantic phe- nomena in Latin that the more technical genres admit innovations ear- lier and more readily than the more literary, and, thus, that given one of the former and one of the latter, both of approximately the same date, if an innovation has penetrated into the former, but not the latter, it has

I9M. W. Winterbottom, The Minor Declamations Ascribed to Quintilian (Berlin 1984) xv.

presumably had less time in which to do so. I, of course, do not wish to suggest that synchronic generic or register differences invariably imply diachronic differences. The constraints violated are ranged down the vertical axis of the table according to the position of their first occur- rence in the chronological/generic partial ordering described above. As an aid to reading, an asterisk (*) is used to mark first attestations. In certain cases it has been possible in the foregoing discussion to distin- guish between degrees of violation of the semantic substance underly- ing a given constraint. These degrees of violation are represented by a decimal following the constraint number. In particular, 10.1refers to the use of fore ut for [+control, +intention] predication in contexts of reported, rather than direct first person, directive illocutionary force and 10.2 and 10.3 to the extension to contexts not involving directive illocutionary force. 11.1 refers to the use of fore ut in threats and 11.2 to give the content of rumors (11.1and 11.2 do not necessarily correspond to a gradable difference in degree of innovation). 12.1 refers to the co- occurrence of fore ut with clauses of weakened epistemic commitment,

Sen. Val. Sen. decl. Ov. Rhet. Max. Cel. Phil. Quint. min. Front. Tac. Suet. Fest.


12.2to independent use to indicate a lower range of the epistemic modal scale. 15.1 refers to the embedding of fore ut under neuter adjectives of epistemic character in textual situations where they implicate epistemic subjectivity and 15.2 to such embeddings where they are fully objective. The matrix summarizing the distribution of the violations of the classi- cal constraints on fore ut in the material studied is given in the table on page 543.

In the table there are three noticeable concentrations of innova- tions: the first in Valerius Maximus, the second in the minor declama- tions attributed to Quintilian, and the third in Festus. More importantly, the order in which the innovations appear is surely not random. In section 3, it has been possible to motivate the chronological ordering or partial ordering of subsets of the entire set of violations according to degree, namely

34.a. Relating to the semantic features of [control] and [intention]: V(lO.l) > V(10.2) > V(10.3) > V(9) > V(3)

34.b. Relating to position on the epistemic modal scale:
V(l8) > V(12.1) > V(12.2) > V(14)

34.c. Relating to objective modality:
V(4), V(15.1) > V(16) > V(15.2)

34.d. Relating to deontic modality:
V(lO.l) > V(10.2) > V(10.3) > V(19), V(20)

The orderings in 34 are strongly confirmed by the high level of correla- tion with the chronological/generic partial ordering of the horizontal axis of the table; in fact there is no outright contradiction of any of them. A number of reservations, however, need to be taken into consid- eration. In regard to 34.a, V(lO.l) is attested only in the Declamationes minores, so it is possible that its occurrence may postdate the attesta- tions of V(10.2), V(10.3), V(9), and V(3). It should be noted, however, that the distinction in degree of violation between V(10.1), the use of fore ut with [+control, +intention] predications in the third person, reported situations of overt directive illocutionary force, and V(10.2), the same except the context is not overtly but only inferentially direc- tive, is the finest one drawn, and, perhaps, should not be expected to be reflected chronologically. In regard to 34.c, given the contemporaneity of Valerius Maximus and Celsus, V(16) is not chronologically separated from V(4) or V(15.1), but it is separated from them by a genreiregister- type of difference, since Celsus is obviously the more technical writer. In regard to 34.a and 34.d, V(10.2) does not appear before V(10.3); since, however, it does not appear after V(10.3), but at the same time there is no outright contradiction with the prediction. Of course the comparison of the orderings in 34 with the data in the table does not constitute scientific hypothesis testing in its strong sense. Obviously it is not possible to search continuously in time through all registers and varieties of Latin. Rather, only the cross-sectional, synchronic implica- tions of the orderings in 34 are in practice testable, such that for any violations of constraints x and y, if V(x) > V(y), then in any temporal cross-section at time ti after V(y), the occurrence of V(y) implies the occurrence of V(x): V(x) > V(y) j[V(y) +V(x)lti. Even so, the data cannot falsify such implicational predictions in any individual case, since there is always the possibility of chance non-attestation in any text or set of texts, and, furthermore, differences in subject matter and style cannot be controlled for as they might affect the probability of the occurrence of the various utterance types in which violations of the various constraints can be detected. Moreover, it is obvious that the very distinctions drawn between the degrees of violation of individual constraints could not have been made without prior examination of the data. Nevertheless, and granting these reservations, I believe that the pattern displayed in the table is a coherent one typical of gradual lin- guistic change developing over time.

  • Recommend Us