Bryn Mawr Classical Review 2002.06.36
Fredrik Oldsjö, Tense and Aspect in Caesar's Narrative. Studia Latina Upsaliensis 26. Uppsala: Acta Universitatis Upsaliensis, 2001. Pp. 544. ISBN 91-544-5076-8.
Reviewed by Gene Adam, Princeton Consultants, Inc. (email@example.com)
Word count: 4710 words
Few scholars would attempt to combine historical linguistics, syntax, semantics, and discourse linguistics in order to work over a segment of Latin grammar that, if it rates any reflection at all, seems a model of clarity, economy, and capability. As Varro wrote two millennia ago, Latin tempus simply amounts to two morphological verb stems, the perfectum and infectum, and two aspectual semantics: perfective and imperfective.1 Past, present, and future tenses are built on the stems. Thus the perfect system, amaverat, amavit, amaverit, and the imperfect system: amabat, amat, and amabit. To this add a few rules that constrain the sequence of tenses in certain clause structures, and one can describe any action and point to any location in time.
Elegant, yes. It has long been recognized, however, that Varro's schema accounts for a present perfect but not an aoristic perfect. Further study shows that the conjugated verbs are strangely incommensurable: imperfects mark tense (-ba-) but not aspect, perfects aspect (amav-) but not tense; the perfect therefore would seem to convey its tense-value through aspect, and the imperfect its aspectual value through tense. And the notion of aspect itself is much disputed. Not only is it hard to define and difficult to schematize; it is not clear to what precise element of the utterance it should be assigned. Latin pedagogical grammars employ a shorthand that rolls up contextual aspectual meanings into the morphology of the tense-aspect system: but from a linguistic, not pedagogic, standpoint, imperfect aspect entails conativity or iterations no more than the semantics of the genitive case require a weight or a crime. Once aspect is de-contextualized, we have grammatical aspect -- the slippery semantics that seem to inhere in the "aspect stems" and have a variegated proto-Indo-European history -- and lexical aspect, representing the total temporal semantics of the verb with other elements of the phrase or clause in which it sits. Grammatical aspect is what we refer to with the terms perfective and imperfective. Lexical aspect, sometimes called "situation", sometimes called Aktionsart, is a key part of what is missing from Varro's description of the Latin tense-aspect system. Consider castra fecerunt and iter fecerunt: they have the same grammatical tense and aspect fixed to the verb, but the temporal semantics of the one phrase are not interchangeable with that of the other. Lexical items such as adverbs and nouns contribute to the aspectual profile of the clause, which means that aspect cannot be reduced simply to a verb stem or a tense-aspect marker.
At least from a syntactic and semantic point of view, Latin tense and aspect studies indeed look very different today than they did even forty years ago. European scholars, predominantly but not exclusively Dutch, have expanded our analytical frameworks by tracking functional and discourse-centered developments in the field of linguistics. Outside of classical philology there is a rich and fascinating literature produced by logicians, analytical philosophers, and linguists of various stripes, ranging from the discourse-centered work of Paul Hopper2 to the logical semantics of Greg Carlson,3 from the computational linguistics of David Dowty4 to the cognitive and developmental psycholinguistics of Michael Bamberg.5 These and many other scholars have not only built valuable frameworks for handling long-standing, familiar questions, but have broadened the field by showing that tense and aspect play non-referential, non-temporal roles in such phenomena as text cohesion, backgrounding and foregrounding, subjective-objective posturing, distancing strategies, and so forth. Tense is multifunctional, and narrative is of course much more than a concatenation of past-tensed events. If there is doubt that the new analytics can be applied to a great literature in a dead language, the late Suzanne Fleischman's splendid book on the vernacular Romance verse of the Middle Ages lays it to rest.6
Which brings us to Frederick Oldsjö's epideixis on deixis, some 500 pages on Latin tense and aspect. O. doesn't announce a thesis at the start of his book, but promises solid empirical work as a check on all of the recent theoretical literature. The book slowly unfolds first as a theoretical dissection of Latin tense and aspect from historical-linguistic and syntactic viewpoints, then proceeds as an investigation of "how Caesar uses finite and non-finite forms...in the verb system of Classical Latin ... by describing the possibilities and discussing certain factors behind the choices made". By the end of the book, one is convinced that O.'s strength indeed lies in solid, precise empirical work on a large corpus of texts. With almost superhuman stamina, O. resurrects the method of counting grammatical forms and constructions: a large amount of data on Caesar (and additional data on other Roman historians) is consequently now available to scholars, be they interested in tense and aspect or simply in basic prose constructions. Apart from the number-crunching, one often admires the agility with which O. uncovers first one angle then another from which to view the data, and the tenacity with which he pursues his conclusions. Elsewhere, O. can be less tenacious than tendentious; the logic is at times exuberant, almost manic, and there is the occasional swipe at those with whom he disagrees (including this reviewer: but I really hadn't thought of aspect as a "sensitive and even emotionally charged issue" until I read this book). Yet for a number of reasons, including scope, new data, and the bibliography, this book is probably the best introduction to these matters available to the typical Latin scholar.
Drawing deeply from the secondary literature, O. builds a case over the course of the book that the speaker's choice of narrative tenses (aside from the historical present) is conditioned at several levels, from grammatical and lexical aspect through clause structure and information type. In the early chapters, however, O. argues strenuously that the lexical properties of Latin verbs put constraints on their combination with the tenses' aspectual values, much like the standard picture drawn of the proto-Indo-European verbal system. This premise casts the die for the remainder of the book. On the one hand there is the conviction that aspect stems and their Aktionsarten have an affinity for certain tenses, and in the other hand there is the supposition that choice of tense (and thus aspect) is conditioned by rules at the narrative level. Both are subjects that O. takes to a fair depth. Whether or not he successfully melds these two views is largely the story of this book.
Chapter 1 opens with some basics of tense and aspect as well as the temporality of verb roots and full situation types. In order to establish a definition of tense, O. revises Reichenbach's familiar schema of tenses.7 This section will test the dedication of anyone who intends to read the book in its entirety, and it comes early for that. Briefly, Reichenbach described a speaker as using a simple past tense as an index from the moment of speaking to a reference point in the past. That reference point locates the event. With a pluperfect tense, the event is located prior to the past reference point. With a present perfect, the event comes after the past reference point. By building out this schema of events, reference points, and the moment of speaking, each tense can be described in a single system. Note also that with these movable reference points, it is not necessary to add "perfective" aspect to the equation.
While O. has clearly applied special care to his examination of Reichenbach's model, his treatment is not entirely fair. Reichenbach looks undeservedly lax when O. supplies him with examples that he didn't use in his summary of English tenses (43). Similarly, while O.'s reform of the schema allows for conditional perfects through a secondary reference point, Reichenbach had in fact excluded modals from his definition of tense.8 After reaching almost Ptolemaic complexity, O.'s own model, or rather its intent, becomes clear nevertheless in Chapter 2, where aspect is treated. In O.'s revision, neither the Latin perfect nor the imperfect has a reference point. That's fine for the aoristic perfect, since even in Reichenbach it is more a matter of formal consistency that a simple tense should have a reference point. The present perfect meanwhile retains its prior reference point. It is the imperfect that O. is interested in. Some scholars, notably Harm Pinkster, have questioned the utility of analyzing Latin for grammatical aspect since the latter is quite difficult to pin down in any principled way (though easy to refer to casually).9 Better to apply Occam's razor and account for the imperfect's aspectual semantics as deriving from its "simultaneity" with a past reference point, typically established in discourse by the narrative perfect. Having removed the reference point, O. restores the aspect to the equation, arguing that the "ongoingness" of the Latin imperfect is a case of aspect. Perhaps the larger question is whether or not aspect can be resolved as a derivative of tense. O. argues the reverse (93).
In Chapter 3 O. takes his readers through a quick linguistic history of Greek and Latin tense-aspect formations. The exposition will be familiar here: perfect stems were once formed on stative roots, aorist on dynamic, punctual roots, and imperfect on dynamic, durative roots. In Latin, perfective and aoristic semantics fused in the perfect, leaving the perfect and imperfect stems. The former required markers to preserve its aspectual value, while the latter simply retained the imperfectivity of the present active stem, but required a tense marker. As a result, Latin perfects are marked for aspect and not tense (except in the pluperfect and future perfect), and Latin imperfects for tense but not aspect. However, as O. explains it, the functional picture is different: the perfect derives a deictic capacity from its aspect, and the imperfect an aspectual value from its systematic opposition to the perfect. In other words, by the time Latin has reached the classical period, both tenses have tense, and both have aspect.
In Chapter 4 O. comes to his last theoretical section. In order to grasp the full temporal profile of a clause, one must take into account the root temporality of the verb (i.e. notwithstanding its morphology) as well as other temporal contributions apart from the verb. These factors O. calls Aktionsart and situation type, respectively. The concept of Aktionsart is necessary to portray the interaction of a verb's root semantics and the aspectual marker attached to it, while situation-type sums up the temporal contribution of adverbs and other elements over the whole clause. All four factors that contribute to the temporality of the clause will now be in place: tense, grammaticalized aspect, Aktionsart, and situation-type.
O. introduces Aktionsart with the notions of telicity and atelicity. The action that a verb refers to may be telic or not: that is it has a natural endpoint, or it doesn't. Attingere and interficere are telic; pugnare and remanere are atelic. While a binary division of Aktionsart into telic/atelic comprehends the universe of Aktionsarten, it is insufficient to describe the variety of Aktionsarten found within that universe. Therefore O.'s implementation of Aktionsart comes to resemble that developed elsewhere for situation types: states and processes, with the latter divided into achievements and accomplishments. A severely limiting factor, as O. points out, is that only some verbs have an Aktionsart pronounced enough to emerge unscathed in any situation type (e.g. perficere); many others are more reactive to the situation type (e.g. contendere remis, contendere ire, contendere cotidie, contendere magno cursu: will the real Aktionsart please stand up?). O. concludes that the concept of Aktionsart is "rather vague and problematic....[and only useful where] distinct and stable" (156-7). This sound conclusion contrasts with the idea that verbs of particular Aktionsarten fit naturally with perfective or imperfective grammatical aspect. This latter opinion will be recapitulated, if somewhat disguised, in a later section on "information types".
Chapter 4 also treats situation types. As always, O. has looked into the secondary literature with a great deal of thoroughness. Here he starts with Vendler's well known inaugural article,10 adds Verkuyl,11 and swells the ranks with probably two dozen scholars more. There is a nice discussion of the compositional elements of the clause (or "Situation", a unit whose boundaries are not specified) that bear on its temporality; these include adverbs, subjects, objects, directional phrases, and negations.
Chapters 5-7 constitute the second of five main divisions of the book; we reach the primary material in the third section. In these tortuous chapters, O. first argues that some verbs will "normally" take the perfect because they refer to telic, punctual situations, and that some verbs "normally" take the imperfect because they refer to stative situations. This generalization is only lightly defended. The basic argument is that when switching such verbs out of their "normal" tense/aspects to non-normal ones, a conflict arises between the situation and the grammatical aspect, which O. labels a "pointed meaning". The elaborations on the argument involve certain verbs possessing a range of meanings, either literal versus metaphorical, or active versus passive, that release the constraints. In addition, "ontological status" of subject and object (situation elements) is considered, as well as negation. O. makes the first part of his argument with the example of relinquo, which in its transitive form is telic and punctual ("to leave"), and in an intransitive form is stative ("to remain").
There are two sticking points here. One, O. has just established in his sections on the compositionality of situation type that the aspectual profile of the situation does not determine the choice between perfective and imperfective aspect. Rather, the tense/aspect of the verb, along with other elements of the clause, determines the aspectual profile of the situation. In each of the passages he cites as a natural fit between grammatical aspect and Aktionsart, a switch of the verb tense/aspect may or may not produce a different situation type, but it is difficult to see how the grammatical aspect on the verb is constrained by Aktionsart. For example, relinquebatur una per Sequanos via (BG 1.9.1) as opposed to relicta est una per Sequanos via; or pilisque missis ad gladios redierunt (BC 3.93.2) versus pilisque missis ad gladios redibant. What O. does here is set up a return to roll-ups like the "iterative imperfect", despite having demonstrated how to properly decompose the aspectual profile of a clause.
Later in the same section (5.1), O. tries to establish that in some cases it is the logic of the proposition that determines the choice of tense/aspect. For example, Ipse Cicero, cum tenuissima valetudine esset, ne nocturnum quidem sibi tempus ad quietem relinquebat, ut ultro militum concursu ac vocibus sibi parcere cogeretur (BG 5.40.7). But: Ipse Cicero, cum tenuissima valetudine esset, ne nocturnum quidem sibi tempus ad quietem reliquit, ut ultro militum concursu ac vocibus sibi parcere cogeretur. If the proposition is that Quintus Cicero stayed up through the nights, then in both cases this proposition is expressed.
Finally, citing a set of passages containing claudere, O. argues that if Caesar had used the imperfect with the punctual situation portas claudere and the perfect with the stative situation novissimum agmen claudere, pointed aspectual meanings would have resulted. We'll never know, since Caesar didn't try either in the texts that we have. But does this mean that it was the situation containing the verb claudere that attracted one tense or the other? The legitimate point that O. ultimately makes is that a verb that typically has a single, punctual application can help to produce an iterative situation type if paired with the right elements. In English, "I closed the gates for several days" probably will be interpreted as an iterated closing of the gates, as will "I was closing the gates for several days". It doesn't follow that "close the gates" naturally takes one tense/aspect or another. In Latin, Expeditae cohortes novissimum agmen claudebant (BC 1.79.1) and Expeditae cohortes novissimum agmen clauserunt seem to express the same atelic lexical aspect, "close the rear of the column", in either form. However, once O.'s difficult premise is in place, he is compelled to plead special circumstances for exceptions (such as subsistebant at BC 1.79.1, called "strange" (n. 416)).
Other discomfiting instances of exception-handling extend through Chapters 5, 6, and 7. In Chapter 6, Imperfective Aspect and Telic Situation Type, what is improbable according to the rules O. has set up (e.g. an imperfect such as interficiebantur instead of interfecti sunt) is now also attributed to dramatic emphasis or to the narrator's subjective choice. Still further, a narrator chooses to build a telic situation with an imperfect because he perceives the action as iterative. Yet we know this not because we have access to his perceptions but rather because he has used the imperfect with a situation in which the telic action is repeated. (In the cited passages, once again the iterativity expressed by the proposition would survive with perfect: e.g. conpluresque milites in viis urbis omnibus partibus interficiebantur versus ... interfecti sunt (BC 3.106.5).) Against this criticism, however, it is indeed admirable that O. cites and attempts to deal with his counterexamples. Chapters 8 through 10 present the bulk of the analysis of Caesar. In 8, O. discusses how he arrived at his verb counts. Issues range from the state of the texts to clause punctuation, types of constructions excluded, participles versus adjectives, and the tense-ambiguity of 3rd conjugation verbs, that is, whether a particular verb is perfect or historical present. (It may be unclear why the last is an issue, since O. regards the two narrative forms as interchangeable, claims that the historical present functions as a perfective tense, and in the control language, English, translates them equally as simple past tenses. Nevertheless, O. rightly points to the consequences one view or another of this tense may have for an editor.)
In Chapter 9, O. begins to set out his actual data, providing comparisons of Caesar, Livy, Velleius Paterculus, Florus (epitome) and Eutropius. Since any narrative in Latin can be constructed entirely in the perfect tense, while no Latin narrative can be constructed entirely in historical presents, imperfects, or pluperfects, the burden of O.'s book is to explain the alternations from the baseline perfect to the other three tenses. O. starts with narrative pace, which he defines as the ratio of historical years narrated to pages written. Caesar has the slowest pace at 52 pages per year, while Velleius and Eutropius the quickest at .65 and.06 pages per year respectively. A first cut at analysis shows that Caesar has the highest percentage of imperfects, while Velleius and Eutropius the lowest. The "stylistic variation" of the historical present also tracks this pattern, meaning that the imperfect has an inverse frequency to the historical present. In broad terms, where the tenses are grouped as perfect and historical present versus imperfect and pluperfect, the alternation is driven by foreground and background concerns, and O. hints as much here. He develops interesting correlations between the higher-frequency imperfect in Livy's earlier books and the greater amount of background information found there; and notes a diminution of tense alternations in later books (30 and onward), by which point Livy's writing had become more "procedural". There are additional interesting discussions of Velleius, Eutropius, and Florus in this vein. Occasionally O. analyses individual tense-aspect usages, as for example a "conative" imperfect in Velleius, in order to recapitulate theories presented earlier in the book.
Chapter 10 has a plethora of tables containing verb counts and distributions in Caesar. The effort was clearly immense. O. steps through each tense's distribution in main and subordinate clauses, connecting the results to discourse functions of foregrounding and backgrounding. Following this discussion, O. examines the tenses' distributions in different types of subordinate clauses (30 types, temporal and non-temporal, atque through uti). Meanwhile, his earlier focus on the imperfect's temporal-aspectual value returns here. O. argues that if the imperfect were meant to signal simultaneity with a past reference point (temporal), then it should occur with simul(atque) and dum, which it doesn't, except once with dum. Rather, he states, because the perfect can also be used in overlapping situations, the imperfect must signal an ongoing, incomplete action (aspectual). I won't go through O.'s reasoning, but the passage at 341 on BG 7.17.2 is representative.
A good deal of time is then spent demonstrating that the perfect and historical present are functionally equivalent. As mentioned above, O. shows that the two tenses added together have an inverse frequency to the use of the imperfect, so that where either the perfect or historical present is used with greater frequency, the imperfect is used with lesser frequency, and vice versa. This correlation tracks the narrative-level opposition between the perfect and the imperfect, wherein the perfect is used to move the narrative forward while the imperfect to give background. O. concludes that these verb counts "prove on statistical grounds that the present of narration is generally used in the same function as the narrative perfect in Caesar's texts" (350). This is quite a leap, however; similar distribution does not prove equal, or even similar, function. In the same vein, he states that "the present of narration functions as a perfective tense form in Caesar" (ibid.). Later in the same chapter, some historical presents are said to have been substituted for imperfects. (Indeed, if Caesar gives the impression that the Latin historical present alternates only with the perfect, Livy shows that it's a reasonable alternative to the imperfect: 4.9.1 dum haec Romae geruntur, 10.36.16 dum haec in Apulia gerebantur.) Finally, note 750: that the present tense can be used both in narrative contexts and in oratio recta within narrative "is an argument against the view that the present tense form has an inherent value of 'presentness'...." Here O. is recalling his earlier morphological arguments on the tense's lack of aspect and tense markers; but with the help of the notion of "functioning as", the present tense can be said to do anything.
These views of the historical present raise the question of why Caesar used it at all; other tenses would probably have been easier. Or, alternatively, one could ask why Caesar used other tenses, since the present, unmarked for tense and aspect, may be found referring to the past, present, and future. O. has a number of answers to the first question. The perfect is unmarked in narrative -- so it is reasonable to use another form that is unmarked in general. Furthermore, one need not constantly establish the past temporality that the perfect provides, since such is the norm in narrative, and, in addition, iconic order takes care of sequentiality. And finally the present's "lighter and more flexible" nature may be stylistically preferable to that of the perfect. The second question O. answers by noting that narratives must begin with a perfect and that indications of result and completion will do better with the perfect. (The latter facts explain why similar distribution does not prove similar function.) All of these arguments finally wind up in the safe harbor of variatio, "between a morphologically marked and unmarked form [perfect and present respectively]".
We are back on dry land when O. turns his statistical evidence toward the question of composition. Here again, the real value of this book, in my opinion, is that through assiduous and precise effort applied to verb counts O. lets us see things that we have not been able to see before. O. is able to chart out the frequencies of the perfect and historical present not just in Caesar's work over time but also in Livy's. This section is original and valuable. Chapter 10 also contains discussions of the historical infinitive and the present perfect. The latter is neatly divided into groups representing comments on the discourse, on the source of knowledge, including nature, topology, and anthropology, and generic comments; and its occurrences are counted.
In Chapter 11 O. classifies participial constructions and all indicative verbs in the narrative tenses in BG 1 and BC 1 into twelve information types, such as "decision and command" and "transportation". The results establish (in part) that decision and command information-types predominate in main clauses, possession and control in subordinate clauses, and influence in participial constructions. A distribution of these information types by tense is also presented, with the result that decision and command, at the dynamic end of information-type spectrum, is mainly perfective (notwithstanding the high-count historical presents), while mental state, spatio-temporal location and quality and existence, at the non-dynamic end of the spectrum, are predominantly imperfective. Finally, O. observes that if a situation is dynamic and contains main line events it will be in a main clause, while attendant circumstances will be in subordinate clauses. As attendant circumstances have been equated with the non-dynamic information types noted above, and the latter have been counted as predominantly imperfective, subordinate clauses are rich in background and imperfectivity, while the converse holds for main clauses. This distribution holds when the clauses and opposing situation types are matrixed. When they are not, non-dynamic, imperfective situations interpreted as attendant circumstances are perfectly at home in main clauses. Since the historical present is lumped into the perfective category, no interesting data are developed here.
In 11 and 12, O. begins to sort out the rules of background and foreground, which combine tense and aspect, information type, situation type, finite and non-finite verbs, and their correlations with main and subordinate clauses. Extricating and analyzing these typologically dissimilar variables, bound together in each sentence of Caesar, takes real work; O. is methodical and original. Further development of this material will likely produce a valuable contribution to Latin syntax and information structure; here it allows O. and his readers a respite from the strict referentiality that dominated the semantics of tense found in the earlier chapters.
In 12, O. explains iconicity and sequentiality in the context of situation types and sequentiality as a necessary condition for narrative movement. Taking up the subject of grounding once more, O. is again adroit in his discussion of how the reader attaches degrees of importance to information encoded in main clauses, subordinate clauses, participial constructions, and voice, as well as the information types themselves.
In conclusion, O. has chosen to wrap discussions of tense-aspect morphology and semantics into a book on the predilections of Caesarian narrative. One will naturally juxtapose the arguments made in these two large divisions. Indeed, arriving at the book's final discussions on sequentiality and grounding, having been thoroughly tutored in the construction of Latin narrative, one may wonder about some of the proposals of the early chapters. Recall O.'s argument that telic situations "naturally" -- i.e. aspectually -- take the perfect, while atelic situations naturally take the imperfect (Ch. 5, passim). In light of all that we learn in the later chapters, the question becomes even more difficult: it is clear that narrative too conditions the distribution of the tenses and lexical items. The imperfect will largely produce non-sequential situations, and the perfect will largely produce sequential situations (leaving aside esse, and notwithstanding the input of other elements of the situation). This opposition is decisive for constructing narratives, and leaves the perfect carrying the indispensable main line of events while the imperfect elaborates on the main events. O. has also demonstrated that at least in Caesar's historiography, information structure has its own opposition: the main line of events consists of mostly telic verbs -- iubere, postulare, mittere, proficiscor, etc., and attendant circumstances consist of a high number of atelics, such as intellegere, sperare, arbitror, varieties of esse, and patere. Since the main line is necessarily in the perfect (leaving aside the historical present), then the high correlation between the telic Aktionsart of the verbs in Caesar and the perfective tense-form could well be a happenstance of a certain information structure piled on top of narrative-driven grammatical aspect. Similarly for the imperfect. In other words, using narrative to analyze aspectual structure of Latin may have skewed the analysis somewhat.
Readers may wish for a thoroughgoing summation at the end of this book rather than the summary of chapters that appears. O. has covered an immense amount of terrain, and in the end it is up to the reader to sort through the implications of all that he has presented. A list of tables would also have been handy, as they contain valuable data and there are many of them.
1. L.L.9.96, 101.
2. Hopper, Paul. 1979. "Aspect and foregrounding in discourse". In Syntax and Semantics Volume 12: Discourse and Syntax. ed. Givon, Talmy, 213-241. New York: Academic Press.
3. Carlson, Greg. 1982. "Generic terms and generic sentences". Journal of Philosophical Logic 11.145-181.
4. Dowty, David. 1986: "The effects of aspectual class on the temporal structure of discourse: semantics or pragmatics?" Linguistics and Philosophy 9.1.37-61.
5. Bamberg, Michael, and Virginia Marchman. 1991. "Binding and Unfolding: Towards the Linguistic Construction of Narrative Discourse". Discourse Processes 14.277-305.
6. Fleischman, Suzanne. 1990. Tense and Narrativity: From Medieval Performance to Modern Fiction.Austin: University of Texas Press.
7. Reichenbach, Hans. 1947. Elements of Symbolic Logic. New York: The Free Press. 287-298.
8. Ibid., 298.
9. Pinkster, Harm. 1990. Latin Syntax and Semantics.London and New York: Routledge. 222.
10. Vendler, Zeno. 1967. "Verbs and Times". In Linguistics in Philosophy. Ithaca: Cornell University Press. 97-121.
11. Verkuyl, Henk. 1972. On the Compositional Nature of the Aspects. Dordrecht: Reidel.