Thesis Prefatory Material: A Narrative Account

The purpose of this second post (for the first post, see here: Challenges in language analysis: thesis prefatory material) is to provide some introduction to my analysis and provide some background for the conclusions at which I arrived. In a sense, this is the narrative of the inductive process by which I came to my conclusions about the Greek perfect. This narrative isn’t in my thesis because the meaning of the Greek perfect is not a part of my central research question. It’s also not in my thesis because it is incredibly casually written.

I should also note that the fact that my thesis chooses to use a particular set of linguistic terminology and a particular metalanguage for describing the perfect does not mean that I view all other descriptions of the perfect with suspicion. I am certainly interested in debates about the semantic substance of the Greek perfect. However, semantic substance and metalanguage are separate issues.

Lastly, I should mention that the concept of telicity is one that is pretty essential for understanding both this post and also my thesis. I would encourage you take a moment to skim the Wikipedia article on the topic: Telicity.

Some important books of note:

Bybee, Perkins, & Pagliuca. 1994. The evolution of grammar: tense, aspect, and modality in the languages of the world. University of Chicago Press.

Bhat. 1999. The prominence of tense, aspect, and mood. John Benjamins.

Van Valin & LaPolla. 1997. Syntax. Cambridge

Van Valin. 2005. Exploring the Syntax-Semantics Interface. Cambridge

1. Working through the data

When I began working through the data on the perfect, I had not yet begun digging through the secondary literature on perfect-like categories across languages. To some extent I had some expectation as to what I would find having already read the majority of the work on the perfect from scholars like Haspelmath (1992) and Haug (2004, 2008).* The standard view of the perfect and its historical development is of its shift from referring to a resultant state, where the state denoted by the perfect is the culmination of a change of state (these are typically called resultatives and in Greek verbs like the ἵστημι/ἕστηκα are representative) in the earlier eras of the language to a slow shift toward usage more akin to the standard English perfect, which really doesn’t care about things like telicity very much at all (Comrie’s [1976—Amazon] description of the perfect generally is primarily about perfects that are more English-like…even though he talks about Greek). In the literature, the English perfect tends to be called either simply perfect or anterior. This sort of development parallels, for the most part, what is discussed in the standard grammars of the language: the Greek perfect over time shifted in meaning and merged with the aorist.

Now then, independent of the ongoing debates in NT studies about aspect and tense (which aren’t at all relevant here), I encountered two major problems with this normal view. One was a problem with the data. The other was a problem with theory and methodology. We can discuss them in that order.

Because both Haspelmath and Haug had made telicity a central factor in their discussions, I had resolved that it would be wise to use telicity as an organizing principle as I worked through instances of the Greek perfect.

1.1 Rabbit Trail on theory & method

Telicity is an important feature in Role and Reference Grammar’s approach to types of predicates (derived from the Vendler/Dowty typology).

a. States: be sick, be dead, know, believe
b. Achievements: pop, explode, shatter (the intransitive versions)
c. Accomplishments: melt, freeze, dry (the intransitive versions); learn
d. Activities: march. walk, roll (the intransitive versions); swim, think, write

To these four classes, RRG makes a few adaptions, adding semelfactives (instantaneous events with no change of state): pop, flash, flicker, etc. and also active achievements, which are related to both activities (which involve duration, but no change-of-state) and achievement (which involve a change of state, but no duration). Active achievements are events where half the situation is activity-like duration with with no change -of-state, followed by an instantaneous change of state, as we see in (2) below.

a. The soldiers marched in the park. Activity
a’. The soldiers marched to the park. Active achievement
b. Dana ate fish. Activity
b’. Dana ate the fish. Active achievement
c. Leslie painted for several hours. Activity
c’. Leslie painted Mary’s portrait. Active achievement

In Vendler/Dowty, these sorts of predicates are labeled as accomplishments. However, over the decades, this categorization of such predicates was regularly criticized since other accomplishments a distinctly different. Consider example (3).

a. The soldiers learned how to disassemble their guns.
b. The soldiers marched to the park.

While both sentences here involve duration and a change of state. The sentence in (3a) involves the duration and the change of state happening at the same time, while the sentence in (3b) involves a situation where the duration and the change of state exist separately. If the soldiers started at their barracks, when they are half way to the park, they are not also half way in the park. The change from not being in the park to being in the park exists as a single step of their feet. The change is, for the purposes of the discourse, instantaneous. Conversely, when the soldiers are halfway through learning to clean their guns they are also halfway through the change of state: they know how to disassemble half of their gun. It is because of this criticism of Vendler/Dowty that RRG introduced the category of the active achievement.** All of this gives us a set of semantic features that practically summarize the semantic content of each category:

a. State: [+static], [−dynamic], [−telic], [−punctual]
b. Activity [−static], [+dynamic], [−telic], [−punctual]
c. Accomplishment [−static], [−dynamic], [+telic], [−punctual]
d. Semelfactive [−static], [±dynamic], [−telic], [+punctual]
e. Achievement [−static], [−dynamic], [+telic], [+punctual]
f. Active achievement [−static], [+dynamic], [+telic], [−punctual]

Lastly, RRG makes the methodological choice of treating causativity as an additional parameter for analysis, so that each predicate type has a causative counterpart, as in (4).

a. Tucker was terrified. State
a’. Pierre terrifies Tucker. Causative state
b. Dave walked around the park. Activity
b’. Dave walked his dog in the park. Causative activity
c. The door opened abruptly. Accomplishment
c’. Rachel opened the door slowly. Causative accomplishment
d. The car crashed into the barrier. Achievement
d’. Dave crashed the car into the barrier. Causative achievement
e. The soldiers marched to the park. Active achievement
e’. The captain marched the soldiers to camp. Causative active achievement
f. The lightning flashed in the night. Semelfactive
f’. Henry flashed his headlights at another car. Causative semelfactive

Anyway, this is the framework of categories I used for looking at telicity and the perfect. Only states and activities are atelic. Everything else (including all causatives) are telic. So when I began actually examining verb, I had two questions in mind. First, what role does telicity play in the usage of the perfect? And secondly, are there any other factors that are relevant to the perfect’s usage?

In terms of the data itself, I choose not to limit myself to a corpus like the NT or NT and LXX. Rather I started with a selection of verbs (just over 400—another 300 were examined as I came across them) and then proceeded to examine all instances of those verbs in the New Testament, Septuagint, Apostolic Fathers, Josephus, Philo, Old Testament Greek Pseudepigrapha, and New Testament Apocrypha. Moreover, when verbs only appeared rarely in those texts (less than 50 times), I expanded my corpus to the entirety of the Perseus database of Greek texts. The use of Perseus was essential especially for verbs that never appeared in the perfect in my corpus since it helped confirm with slightly more confidence the acceptability of the perfect with that verb. This was particularly relevant for the atelic predicate types (states and activities).

1.2 Back to the data problem

And that brings us back to our data problem. The secondary literature had me expecting a shift from the use of the perfect with telic predicates to atelic predicates. That is how their discussions go. However, when we actually examine the data, finding instances of perfect with true atelic predicates was complicated than I anticipated. The perfect with state predicates was a fairly simple affair. While there are plenty of state predicates that do not allow the perfect to be formed at all (e.g. ἀρκέω appears 855 times in my corpus and Perseus, but never once appears in the perfect). There is also a significant set that do. Those states that do collocate with perfects fall into specific well-defined classes that are predictable from very specific semantic factors, including the semantics of resultatives—the category that has defined the use of the perfect in the Homeric period according to the secondary literature.***

So what about perfects with activity predicates and semelfactive, the other two kinds of atelic predicates? This is where the data got interesting. In my data, I found not a single semelfactive verb that allows the formation of the perfect. Moreover, there are only handful of clauses with activity predicates that could conceivably be understood as atelic among thousands upon thousands of clauses. This discovery was both perplexing and surprising. It was surprising because nothing in the secondary literature had prepared me for it—though perhaps that was my own fault. For example, when Haug (2004) talks about “atelic VP’s,” I think it would have been better understanding him as talking about VP’s that had they not being the perfect would have been atelic. That’s my best guess.**** But either way, we are left with an odd situation: among quite literally thousands upon thousands of Greek perfects in the Koine and Early Roman periods, the number of possible activity predicates in the perfect can be counted on a single hand. This doesn’t make sense considering that the standard view of the historical development of the Greek perfect moved from resultative to anterior to perfective (i.e. eventual merger with the aorist).

That’s problem #1. What about problem #2?

1.3 D. N. S. Bhat’s typology of grammatical prominence

Problem #2 is only a problem from a certain point of view. It started out as a problem for me, but by the end of the ordeal, it also provided me my solution.

All the morphosyntactic data says to us that Greek is an aspect prominent language and that English is a tense prominent language. But what does that mean from the perspective of semantics? To some extent we can simply say that the Greek perfect is simply a combination of the imperfective and perfective like it is in other aspect prominent languages (e.g. Supyire, Carlson 1994).***** But what does that mean in practice? Again, the secondary literature views the history as a change from resultative to anterior.

Wait a second though.

The English perfect is an anterior. If English is tense prominent and Greek is aspect prominent, then what does that mean for the semantics of the Greek perfect…especially considering that the vast majority of perfects in the Hellenistic and Early Roman periods are more certainly not resultative? From the perspective of theory and metalanguage, it feels like we need a third category: a category that is still aspect prominent and allows us to distinguish the non-resultative Greek perfects from the anterior perfect of English.

1.4 The third option: the answer to both problems

It turned out, for me, that I had too been myopic in my work. I was working through the typological literature on perfects and swimming in activity predicates that never formed perfects. But I was doing those two things totally independently from the other. I knew two things.

· I knew that there was a large gap in the data with so many atelic activity predicates not forming perfects at all.

· I knew there was something I didn’t like about the standard resultative-anterior development in conjunction with my understanding of Bhat’s typology of grammatical prominence in terms of the difference between tense (English) and aspect (Greek) prominent languages.

Anyone who has actually read this far through this narrative can probably see the answer already: telicity is still the defining factor, even for non-resultative perfects. It seems so obvious now. But at the time I didn’t have the luxury of those two sentences juxtaposed so directly like that. Telicity is precisely the difference between the English perfect and the Koine Greek perfect. And that makes sense: an endpoint, as a conceptual entity, is semantically related to aspect, but it is not semantically related to tense. That gives us a true, data motivated distinction between the English perfect and tense prominent and the Greek perfect as aspect prominent. The Greek perfect interacts with telic predictions in a manner distinct from its interaction with atelic predicates. Conversely, English shows no favoritism one way or the other.

For me, the epiphany came while I was reading Bybee, Perkins and Pagliuca’s (1994; hence forth BPP) third chapter: “Anterior, Perfective, and Related Senses.” At the time, I think it was probably my second or third read through. And it’s a little embarrassing that I had not noticed it before. It’s the related senses that are key here. Resultatives fall into that group and I spent a significant amount of time digging through, but because of my initial assumptions about what I would find in the data based on the secondary literature, I ignored another related sense: Completives. When I actually too the time to examine how completives behave according to BPP, I also found the missing link that not only explained the functions of non-resultative Greek perfects, but also motivated their usage as well. Moreover, it fit the descriptions of many grammars, both new and old. Completives are exactly what they sound like. They refer to grammatical morphemes that present an event from the perspective of having been totally completed. This is precisely what Goodwin (1897, 31-32; his emphasis) says: “As the perfect indicative represents an act as finished at the present time, so the perfect of any of the dependent moods properly represents an act as finished at the time (present, past, or future) at which the present of that mood would represent it as going on.” Jelf, in his translation of Raphael Kühner’s magisterial grammar similarly states, “The perfect expresses a complete action, whether it be not completed till the very moment of speaking, as γέγραφα, I have (just) written; or has been completed a long time before as ἡ πόλις ἔκισται, it has been built” (1866, 63).

In fact, essentially all grammars, with a few exceptions, make some reference to the completion of an event or actions as a central characteristic of the Greek perfect, going back thousands of years. Consider the Stoic’s view of the Greek verb (adapted from Robins 1993, 27):

Past Present Future
Incomplete παρῳχειμένος παρατατικός ἐνεστώς παρατατικός
Complete παρῳχειμένος συντελικός ἐνεστώς συντελικός
Undefined ἀόριστος μέλλων

The term συντελικός effectively means completed or completive. Not only is this the term used by the Stoic grammarians (according to Robins), LSJM in its entry for this word lists a reference from the scholia to Dionysius Thrax as stating, “ὁ παρακείμενος καλεῖται ἐνεστὼς συντελικός” (the perfect is called the completed present).

For me, this was the ultimate confirmation: independently collaborated statements about the meaning of a grammatical category from Ancient grammar, traditional grammar, and contemporary linguistic analysis, with all of them fitting quite clearly with what I saw in the data.

2. Conclusions: the meaning of the perfect

The perfect demonstrates a basic polysemic bipartite structure, organized around resultative semantics and completive semantics. In both cases, the perfect shows a clear preference for lexemes that are inherently telic. For example, with the historically older resultative perfects the causative state ἵστημι ‘I cause to stand’ in the imperfective aspect becomes ἕστηκα ‘I am standing.’ The final expression of the resultative perfect itself is not telic, but denotes the resultant state that persists following the telic change of state. With the completive semantics, the perfect refers to a change of state that has been totally completed or finished. Transitive active achievements tend to only have completives semantics, as in example (6)

ἀπεκρίθη ὁ Πιλᾶτος· Ὃ γέγραφα γέγραφα.
Pilate replied, “What I have written, I have written” (John 19:22).

There is no ongoing result expressed explicitly or implicitly, rather Pilate is stating that an event has been completed and that’s that.

While neither are technically inclined to be used with states, both resultative perfects and completive perfects demonstrate their own particular metaphoric extension with state predicates. Resultative perfects since they are supposed to refer to a state that persists as a result of a change of state, as in example (7).

διʼ οὗ καὶ τὴν προσαγωγὴν ἐσχήκαμεν τῇ πίστει εἰς τὴν χάριν ταύτην
Through whom we have gained possession by faith into this grace (Rom 5:2).

With completive semantics, state predicates gain a degree of intensification. This is only possible with certain lexical classes of states: states where the affectedness of the state can be expressed as having degree. Since Completives involve a participant being totally or completely affected by a change of state, the participant in the state predicate is presented as experiencing that state to the highest degree—i.e. completely affected. Verbs expressing mental states (anger, sorrow, etc.), expressing physiological states (weariness, hunger, etc.) readily allow this usage. We see this in examples (8-9).

Ἰσραὴλ μεμεθυσμένος οὐχὶ νοήσει
Israel, completely drunk, is unable to think (Sibylline Oracles 1.360).

He was exhausted (Josephus, Antiquities 14.462).

Returning to the telic predicate types, most intransitive, telic predicates conceivably allow for either resultative or completive semantics. Thus in examples (10-11), either the resultant sense ‘know’ or the completive sense ‘has learned’ could arguably be understood here.

μεμάθηκεν ἡ γλῶσσα αὐτῶν λαλεῖν ψευδῆ
Their tongue has learned/knows to speak lies (LXX Jer 9:4).

πατὴρ μοι Βαθουῆλος ἦν· ἀλλ ̓ ὁ ἤδη τέθνηκε
My father was Bethuel, but he has died/is dead already (Josephs, Antiquities 1.248).

But some perfects will only allow a completive readings, as in example (12).

τοῖς δὲ ἄλλοις ἅπασιν ἠπείλησα μείζω τιμωρίαν ἐπιθήσειν μὴ κομίσασιν εἰς τοὐμφανὲς ὅσα ἡρπάκεισαν
I threatened everyone else that I would inflict a greater punishment upon them, unless they produced before us everything they had seized (Life 335).

With causative predicates, perfects have more tendency toward completive readings, as in example (13).

τοὺς φίλους αὐτοῦ πάντας ἀπεκτόνασιν
They have killed all his friends (Josephus, Antiquities 12.391).

This clause demonstrates a common correlation between completive semantics and exhaustive plurals, described by BPP in their discussion of completives.

When a speaker/author wants to use resultative semantics with a normally causative verb, they tend to use the perfect middle to express that. In fact, in general, the perfect middle in the Koine fills the role that the perfect by itself used to fill in the older set of verbs like the ἵστημι ‘I cause to stand’ vs. ἕστηκα ‘I am standing’ distinction. We see this in example (14).

τὰ σπλάγχνα τῶν ἁγίων ἀναπέπαυται διὰ σοῦ
The hearts of the saints are refreshed because of you (Phlm 7).

From a historical, diachronic perspective, this makes a lot of sense. The perfect and the middle in Proto-Greek (and even in early Greek) were not fully formed. But they had both developed from the Proto-Indo-European Non-eventive/stative verb class. As their functions become more well-defined and grammaticalized, the middle took over part of that semantic space and the perfect took over the other part. Both sacrificing some functions to the other. As a result, verbs like ἵστημι never became the dominate expression of perfect semantics. This in turn lead to the perfects demise, as well. Completive perfects over the centuries after the Hellenistic and Early Roman era tended to be replaced in usage by the perfective aspect, while resultative perfectives tended to be replaced in usage by the middle voice. Eventually, at some point during Late Byzantine or early Medieval Greek, the perfect simply stopped being used, only to eventually to appear again periphrastically in Modern Greek by the European areal periphrastic perfect derived from the possessive construction.

3. Bibliography

Bhat, D. N. S. The Prominence of Tense, Aspect and Mood. Amsterdam: John Benjamins, 1999.

Bybee, Joan, Revere Perkins, and William Pagliuca. The evolution of grammar: Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press, 1994.

Carlson, Robert. A grammar of Supyire. Berlin: Mouton de Gryuter, 1994.

Clackson, James. Indo-European Linguistics: An Introduction. Cambridge: Cambridge University Press, 2007.

Comrie, Bernard. Aspect. Cambridge: Cambridge University Press, 1976.

Gero, Eva-Carin, and Armin von Stechow. “Tense in time: The Greek perfect.” In Words in time, edited by Regine Eckardt, Heusinger, Klaus von, & Christoph Schwarze, 251-293. Berline: Mouton de Gruyter, 2003.

Goodwin, W. W. Syntax of the Moods and Tenses of the Greek Verb. 4th. Boston: Ginn & Co., 1897.

Haspelmath, Martin. “From Resultative to Perfect in Ancient Greek.” In Nuevos Estudios Sobre Construcciones Resultativos, edited by Leza Iturrioz, & Luis José, 187-224. Guadalajara: Universidad de Guadalajara, 1992.

Haug, Dag. “Aristotle’s kinesis/energeia-test and the semantics of the Greek perfect.” Linguistics 42, no. 2 (2004): 387-418.

Haug, Dag. “From resulatives to anteriors in Ancient Greek: On the role of paradigmaticity in semantic change.” In Grammatical change and linguistic theory: The Rosendal papers, edited by Thórhallur Eythórsson, 285-305. Amsterdam: John Benjamins, 2008.

Horrocks, Geoffrey. Greek: A history of the language and its speakers. 2nd. Oxford: Wiley Blackwell, 2010.

Janasoff, Jay H. Hittite and the Indo-European Verb. Oxford: Oxford University Press, 2005.

Jelf, William. A Grammar of the Greek Language. 4th ed. Oxford: James Parker, 1866.

Kühner, Raphael. Ausführliche Grammatik der greicheischen Sprache. Hannover: Hahnschen Hofbuchhandlung, 1835.

Lehmann, Winfred P. Theoretical bases of Indo-European linguistics. New York: Routledge, 1993.

Rijksbaron, Albert. The syntax and semantics of the verb in Classical Greek: An introduction. 3rd. Chicago: University of Chicago Press, 2007.

Sihler, Andrew. New Comparative Grammar of Greek and Latin. Oxford: Oxford University Press, 1995.

Van Valin, Robert D. Exploring the syntax-semantics interface. Cambridge: Cambridge University Press, 2005.

Van Valin, Robert D., and Randy J. LaPolla. Syntax: Structure, meaning, and function. Cambridge: Cambridge University Press, 1997.

Wackernagel, Jacob. Lectures on syntax: With special reference to Greek, Latin, and Germanic. Translated by David Langslow. Oxford: Oxford University Press, 2009.

*Of course, I read much more than these, including essentially everything I could get my hands on that discussed both the perfect and Greek from any historical period. But for the purposes of my little narrative, that’s less relevant.

**In some of the RRG literature, the term active accomplishment is used, but for our purposes which terms is used is immaterial to the point. So don’t get hung up on that.

***See chapter 4…and to some extent, below.

****The caveat here is that I have not examined the Classical data. Haug‘s other article (Haug 2008) provides a single instance of what is justifiably be an atelic perfect from the Classical period, but it would contradict everything we know about processes of grammaticalization if truly atelic predicates were more prevalent in the Classical period than they are in the Koine. It seems probable that like my handful of clauses, this instance, of an activity predicate (Lysias 25.12.5) is equally unusual (but still diachronically and theoretically significant).There is after all ergative syntax in Jane Austen’s Pride and Prejudice, which should be impossible in English both then and now: “Bingley instantly prevailed on Miss Bennet to avoid the confinement of such an intrusion, and walk away with him into the shrubbery.”

*****Two things:

#1. No, that doesn’t mean that the Greek perfect is both imperfective and perfective at the same time. It means that imperfective aspect and perfective aspect got together and had a kid.

#2. Regardless what anyone says: Bhat’s is not talking about Aktionsart when he uses the word aspect. He is most certainly and indisputably talking about aspect.