Chapter 5 of Aubrey’s thesis returns to the stratosphere, as he takes the analysis done on the perfect in Koine Greek and then considers what this has to say/contribute to how RRG works as an approach for linguistics. Particularly, he wants to consider how the Bybee-Dahl approach of look at grams without prejudging meta-categories of aspect/tense/mood, can mesh with Bhat’s typology of looking at languages in terms of A/T/M prominence. An initial diagrammatic relaisation of that is given on p138. This realisation itself has problems, in that it is an idealisation. A more concrete, though still rather abstract, diagram of Koine Greek comes on p140.

p141 offers a diagram that goes back to typology of verbal operators in RRG. Can the approach to categories presented in Aubrey’s synthesis of Bybee-Dahl-Bhat be realised within this typology? Aubrey says perhaps not, and perhaps it’s not necessary anyway.

If you’re after some key pay-off for understanding Greek, p143 has some good material. Aubrey says that the perfect, like the future, is a peripheral category, and that as such they “are not inherently tense or aspect”. As far as this analysis goes, the perfect functions within the aspect system, because Greek is aspect prominent. The future appears ambiguous, because it almost doesn’t know whether it’s aspectual or temporal.

Chapter 6 is the Conclusion chapter, and if you’ve been lost, here is a chance to catch up. I’d like to emphasise the first sentence here, “The goal of this thesis was not so much to solve a problem, but to fill in a gap.” (p145) Aubrey’s thesis doesn’t ‘solve’ the Greek perfect, it seeks to expand the usefulness of RRG for certain areas of language description. But obviously this thesis has important implications for understanding Koine Greek. Indeed, note the sentence on p150 under “Possibilities for future research”, where he says, “If we evaluate the grammars of the past century in terms of Chomsky’s types of linguistic adequacy, they fail to even meet the standard of descriptive adequacy, much less explanatory adequacy.” ouch.

I won’t offer a read-through of the two appendices. Appendix A lists of all the verbs examined and their predicate classes. It is worth reading, because of the way it helps you understand that typology of predicate classes and how it relates to actual verbs and the actual research Aubrey has done.

Appendix B is probably of more interest to my general (minute) readership. It is an overview of the Greek verbal system. Why do you need to read this? Because probably all you have read is those inadequate grammars we mentioned above! The overview here is “independent of the traditional grammatical tradition in terms of categories and terminology” (p187) and so well worth your reading.

And that’s a wrap. Take home message: Greek perfects are resultative/completive grams operating in the aspect system. Or, to uncover Aubrey’s view of aspect from Appendix B (p199):

The Koine Greek verbal system has three aspects: perfective, imperfective, and perfect. The perfective aspect makes no reference to internal temporal structure and is contextually bounded in its interpretation. The imperfective refers to temporal internal structure that is incomplete. It is contextually unbounded. If it appears in a clause with a goal periphery, there is nothing to suggest the goal was achieved. Lastly the perfect aspect refers to internal temporal structure that is either completed (completive) or exists as achieved state (resultative). It is inherently telic and will either assign an endpoint to a situation or event or denote the resultant state of that situation or event. As such, the perfect is both telic and bounded.

Apologies that these come out slowly, I am mostly busy trying to get a paper written by month’s end. Here are links to the earlier posts: 1, 2, 3, 4.

Today we’re picking up at 4.2.3, which deals with “Gram-specific tests part II: Predicate classes”. Probably you’ve had a mental blank and have no idea what Aubrey is talking about. Basically what Aubrey does in this section is take the idea of “anteriors” (indicating a thing that happened prior to another thing, “resultatives” (indicating that an action with a goal (a telic predicate) has created a new state (result), and “completives” (indicating that an action has, you guessed it, come to its inherent end point).

In this section Aubrey goes through Greek perfects and looks at how they match up with these gram types: First he looks at State Predicates (verbs that indicate a state of being); he finds that generally they do not appear in the perfect. The main contrast he does find is Type IV grammatical contrasts (i.e. the difference in forms means that the event is conceptualised differently). In which the perfect indicates entrance into the state. he also finds some involve an “intensive usage” indicating that the participant “experience[es] the state to the highest degree” (p99). Verbs with a meaning of strong emotion are also found in the perfect with such an intensive sense. Generally speaking, these findings correlated with resultative and completive ideas.

The other set of atelic predicates (semelfactives and activity), are easier to talk about. Semelfactives basically do not appear in the perfect. Activity predicates generally only appear in the perfect as active achievements with an endpoint. I.e. you don’t get “marched” in the perfect, but you do get “marched to the city”. This is followed by some stuff about negative clauses which we will skip today.

The next section (4.2.3.2) deals with telic predicates. i.e. verbs which involve or end with a change of state. Following the RRG typology, Aubrey looks through achievements, active achievements, and accomplishments. If you forgot what those are, you can revise them in my first post in this series.

Active achievements involve an activity that has a duration, and then a change fo state (the achievement) that is instantaneous. You find Greek perfects with this idea. You do not find Greek perfects (usually) with Activity predicates (the activity with a duration, but no change of state at the end of it).

I walked or I was walking are activities.

I walked to the park is an active achievement.

Aubrey finds that only when an activity is given a specific (object) or end point is it likely to appear in the perfect rather than the perfective. He also finds the perfect used for “exhaustive” completion of an activity, and in conjunction with voice that the perfect middle is used “to refer to the achieved state of an active achievement”, a role the imperfective middle will not do.

Overall this section also supports the idea that resultatives and completives are dominant in comparison to anteriors.

Aubrey then goes on to look at achievements and accomplishments. This too supports the above conclusion. Finally this chapter looks at causatives along the same lines.

I have skimmed over a lot of examples and argumentation in this chapter to basically give an overview of what is discussed and what is found. In the conclusion of this chapter, Aubrey affirms that the Greek perfect is not anterior but is split between resultatives and completives. His tests have some relevance for adjusting the tests themselves, and how this typology may fit with RRG as well.

What does this mean for how you understand the perfect in Greek?

“In sum, the Greek perfect is a synthetic verbal morpheme that patterns with other aspect morphemes and thus functions in the nuclear layer of the operator projection. Semantically, the morpheme conveys both resultative and completive meaning, depending on both the predicate being used and also the context of the predication. On many occasions, particularly with perfects derived from accomplishments and achievements, it is almost impossible to choose between resultative and completive readings.” (p131).

Okay, if you’re coming at this from a Koine Greek angle like me, we’re now getting to the good stuff. In chapter 4 Aubrey takes his methodological work and begins to apply it specifically to the Greek perfect. Firstly, he reviews some of the disagreements about the perfect, and this is actually really helpful to. For example, he talks about the Perfect as historically a derivation of a PIE stative verb that is generally agreed in the Homeric period to be a resultative type gram. There is less agreement about the shift into Classical Greek, with most agreeing that it came to be a kind of anterior gram, but little consensus about when this took place, McKay puts it as late as Byzantine Greek! As for the Koine period, there’s no consensus at all. Fanning sees it as a state-predicate, Porter treats it as stative but then goes out on a limb and says that stativity is its own aspect. Evans and Campbell are the most extreme, claiming that it is in fact imperfective in aspect. What’s a poor Greek scholar to do!

Well, Aubrey says let’s go ahead and apply the methodology and tests previsouly outlined. p83 has a useful table that talks about Greek morphemes in a way maybe you hadn’t thought about before. Put the root in the middle: λυ, now what can come immediately after that? 3 options: nothing, σ, or κ, so this is a 3-way contrast between imperfective, perfective, and perfect. Note that the aorist and the future are marked by sigma for perfective aspect. Immediately before the root there is a slot for reduplication, and then further ‘out’ the front you have the tense marker (the ε augment), while on the end slot you have “everything else”. If you’re reading Aubrey’s thesis with me, the table really makes this clear.

Anyway, what’s the upshot? Perfect marking goes next to the root, and in an aspect slot. You can’t mark perfect and also mark perfective or imperfective. You can mark perfect and then mark past time. So this supports both aspect prominence for Greek, and aspect for understanding the perfect.

Moving on to grammaticalisation, paradigmaticity, obligatoriness, and pervasiveness (remember those long words!), aspect is more pervasive (p84) because aspect is marked in both indicative and non-indicate forms; tense is only marked in the indicative.

The next step is to test the perfect with quantification. We talked about this back in my second post, and it’s around page 50 or so of the thesis. Basically, we want to see whether the perfect takes certain types of adverbs and not others. Cardinal counting (once, twice, thrice) should occur with perfectives, frequency counting (always, twice a day, five times a year, etc..) should occur with imperfectives. You probably need to read this section (p85-88) to get a full scope of it. Aubrey talks about how the different types of quantifier adverbs appear with different semantic types of perfects (resultatives, completives, anteriors), and talks about how ἀεὶ appears to function with perfects to indicate persistence rather than frequency. Anyway, the conclusion is that this test too confirms that the perfect should be understood primarily in aspectual terms.

The next metatest (p89-91) looks at discourse function, with a passage from Josephus. The conclusion is that the perfect do not provide mainline information that moves the narrative forward, which is what the perfective aspect does. It provides background information. This does not mean, as Campbell concluded, that the perfect is in fact imperfective though. “They both do convey temporal internal structure (Comrie 1976), but the type of temporal internal structure is distinct. The imperfective has no inherent endpoint, while the perfect at times places the focus on the completion of a backgrounded event and other times presents an event as a persistent state that existed concurrently with the foregrounded narrative” (p92).

The next section looks at gram-specific tests (anteriors, resultatives, and completives). In this post I round out the shorter section, looking at adverb tests, and leave the second half of chapter 4 for the next post. Specifically, we should expect adverbs such as “just” and “already” with anteriors, and “still” with resultatives. The conclusion is somewhat inconclusive, because the data splits. But what Aubrey does claim is that the Greek perfect allows for resultative-like grams. However the adverb test does not help us with completives, so the status of completives and anteriors requires further testing. Which we shall see in the second half of chapter 4.

Are you having fun yet? You can read parts 1 and 2 first.

Today we pick up at section 3.2.1, in which Aubrey talks about 2 approaches to categorising up how the perfect ‘fits in’ to languages. But I’m going to flip things up a bit by talking about the second approach (3.2.1.2) first. The second approach Aubrey discusses is by Bhat, who suggests an idealised model in which languages give more prominence to either aspect, tense, or mood. So imagine there are these 3 ‘ideal’ (almost Platonic) languages, one which privileges Aspect above all else, the other raises Tense to the same degree, the third treats Mood with this importance. Each actual language doesn’t match on of these ideals, but it will triangulate and lean towards one of these more than the others.

How do you know which element is more prominent? Features like: grammaticalisation (i.e. is the element embedded into the grammar of the language?), obligation (i.e. are speakers given no choice about whether to express the element or not. e.g. in some languages one must indicate whether something is perfective or imperfective in aspect, there is no non-aspect-marked form), systematic and pervasive (i.e. is the element featured throughout the language’s [verbal] system without significant gaps).

So a language that forces and features past vs. non-past may be tense prominent, a language that obliges perfective vs. imperfective is more likely to be aspect prominent, etc..

Aubrey discusses this model second for certain reasons, that Bhat manages to encapsulate ideas from Bybee-Dahl in a way useful for RRG. But in my view Bhat is a lot easier to understand than Bybee-Dhal and thus to explain. Bhat also (p59-60) suggests that some elements may be ‘re-explained’ within an language from one type of system to another. This can be seen in the Future. Is the future ‘tense’ modal (expressing ideas like desire, obligation, ability) or temporal in its operation? This differs from language to language, but it also differs diachronically, and this is certainly true for Ancient Greek. Some have analysed the future as deriving from a PIE desiderative suffix (p61), while offers argue that it derives from a perfective aspect suffix. So, the question then becomes, ‘Historically, did the future tense-form arise from a modal form or an aspect one?’ Regardless, with reference to Dionysius Thrax, Aubrey points out that Ancient Greek grammarians analysed it as aspectual.

Similarly with the perfect. For tense-prominent languages (still p61), perfects can be analysed temporally, a “past even that has current (present) relevance” (quoting Bhat). This certainly works quite well for English. On the other hand (p63) an aspect-prominent language has a perfect form that is aspectual, not temporal, in both its formation and its general significance.

Okay, now that we’ve got our heads around Bhat, let’s step back to 3.2.1.1, Bybee-Dahl. These two sideline categories like tense and aspect, and instead look at “grams”. What are “grams”? Imagine you just came up with a word that signified varies tense-forms or aspect-forms or mood-forms, but you wanted it not to mean any of them in particular, just all those varieties of putting a verb into a particular ‘form’. That’s a gram (I think!), or a “grammatical morpheme” – it’s a tag or structure that identifies a particular grammatical feature, but remember we’re dealing with verbal systems. So it’s when we analyse these grams that we go looking for their tense/aspect/mood in relation to their semantic content.

This frees us to look at grams more broadly, and consider how a range of notions: aspectual, temporal, modal, aktionsart-ish, and so on, can or could be realised in a language’s systems. There’s a great, difficult to summarise, illustrative diagram on p55.

One of the advantages of this model is that it sidesteps ‘beginning’ with the Perfect. instead, they present ‘gram-types’ such as “resultatives, completives, and anteriors” which have “similar paths of grammaticalization” that typically result in “perfect” forms. But it allows both synchronic and diachronic analysis of what that perfect means, its semiosis.

By dispensing with the tense-aspect-mood categories as a starting point, it allows a better approach to the individual systematisation of languages, however the disadvantages are a lack of engagement with some other linguistic theoretical elements (p57) such as that grammaticalisation of tense is usually distinct from grammaticalisation of aspect as a system. The failure is largely a result of tending to look at things from a more micro rather than macro perspective.

So Aubrey moves on in 3.2.2 to talk about how using these approaches, we must discuss mortphosyntactic and semantic tests (p64). How does the perfect line up with tense grams and aspect grams? How do we categorise actual grams in a particular language? We need some tests! Tests for grammatical prominence are something I talked a little bit earlier in this post: grammaticalisation, systematic organisation, obligatoriness, pervasiveness. The second step to is test for tense and aspect in particular. We must consider whether a particular gram functions within the paradigm of the tense system or aspect system, unambiguously or not. In some languages tense/aspect is always ambiguous, or better yet intertwined.

Aubrey talks through a number of other types of tests, which are summarised in a table on p70. Finally he talks about tests suitable for perfect-like grams (p71), which is the focus of the study. Bybee et alii set out two morphosyntactic tests for anterior and reultative grams: anteriors will take temporal adverbs likejust or already, but not still (prior state-of-affairs vs. persistent situation). Anterior grams are likely to be tense-oriented.

Resultative grams function in reverse: they should take still but not just and already, and resultatives are aspect-oriented. Sadly, Aubrey notes, Bybee et alii don’t have any more good tests for these 3 grams. Instead, some test need to be derived. Resultatives, by their semantic content, “require a normally telic verb”, so there’s a test waiting to happen. Completive grams tend to “correlate closely with change of state predicates”; these, and a few others, result in 7 tests (p74) for Aubrey to apply to perfect-like grams in Greek.

Great! In our next post we’ll get to work on understanding how Aubrey went with applying this to the Greek perfect

For part 1 click here.

Today I’m going to talk my way through Chapter 3 of Aubrey’s thesis. What did Aubrey do in Chapter 2? He told you all about Role and Reference Grammar and then said, “hey, why doesn’t it do a good typology/classification of tense and aspect? Wouldn’t that be handy-dandy?” Chapter 3 sketches out what an approach to tense and aspect in RRG would/could/should look like.

Aubrey starts off by talking about some of the meta-issues; for example how categories relate to a particular language and to universal grammar, or at least a universalising concept of language (p29). Similarly, simply choosing what terminology to use, what distinctions or categories to draw up, is itself a step of interpretation, before we even get to the empirical questions (p30). There is, almost inevitably, a back and forth between the more theoretical and the more ‘on the ground’ levels of analysis here (p31). This issue of the “bidirectional relationship between” the two levels, quoting Bache, then gets applied to Porter’s analysis by way of critique. Porter introduces the label ‘remote’ to refere to the past/non-past distinction in Greek. But this choice of label, and especially how Porter uses it, do not match with more generalised applications of the term in linguistics. That is, remoteness is generally used to refer to temporal remoteness in languages, to temporal distance. But Porter repurposes the term to use it precisely for non-temporal remoteness that may be used temporally. This is idiosyncratic at best.

In the following section Aubrey considers a very interesting element of methodology. It has to do with example sentences, replacement, and minimal pairs. A typical way of teasing out linguistic differences is to use example sentences, and change a single element, to demonstrate the difference this makes in utterance. Drawing on Bache, Aubrey gives us a typology of 4 types: (I) Replacement is not possible due to a gap in the grammar, (II) replacement results in an ungrammatical sentence, (III) replacement results in a distinct change of meaning (non-synonymy), (IV) replacement involves slight to zero change of meaning (near synonymy). Aubrey runs through examples, with reference to English and to Koine, and demonstrates how all 4 categories need to be examined for linguistic typology, but also how the boundaries of these types are sometimes unclear. Further on (p.45f) he points out that generally Type II and III are the categories that best demonstrate claims about “morphological and syntactic forms”, but a responsible use of example sentences will be one that considers what type of contrast is in play, not merely what contrast is set up.

Careful attention to this typology of contrast sentences is important because misplaced emphasis on the wrong categories can lead to unjustifiable claims. This is seen in the Koine tense/aspect debates from the late 80s (p.47). Undue emphasis on type 4 contrasts supported claims that Greek was tenseless, whereas alternate understandings of the contrast pairs would support a different explanation that saw these as extensional differences. What’s the upshot at this point of Aubrey’s argument? That Porter argued for something on the basis of Type IV examples, when he should have given more argument on the basis of Types II and III, and that this undermines the strength of Porter’s analysis and conclusion.

Now Aubrey moves on (p.50) to a select examination of work on typology of tense and aspect. Here’s some real meat for you. First up, Comrie’s “definition of aspect as involving the grammaticalization of the ‘internal temporal constituency,’ of a situation” (p.51), then Bhat’s 3 classes: (1) perfective vs. imperfective, (2) phasal (distinctions of temporal constituency: e.g. ingressive, resultative, progressive, etc., (3) quantificational (distinctions of countability).

Almost everyone who has heard about aspect in Greek is familiar with the first distinction here. The second is a set of terms that many will be familiar with, and generally in Greek assign as ‘usages’, or ‘senses’ of particular verbs in context (so-called “ingressive imperfect” for instance). The third similarly so, “habitual present”, “iterative imperfect”). This has really just been a way for typological Intermediate grammars to explain/classify usages of Greek tense-forms that don’t match grammar-translationists default glosses.

This 3rd category of Bhat’s is also “difficult…in descriptive practice”. Languages just don’t neatly have/define/contrast their aspectual markings along lines that one would like.

When it comes to tense, things are more simply (p53), I mean you can only really have three deictic tense references: past, present, future, in relation to the point of speech. Any variation on this is only going to do one of two things: mark degrees of temporal remoteness from the point of speech, or using non-deictic reference, e.g. indicating past/present/future from a different point of time. Of course, one can combing deictic and non-deictic reference systems, to indicate combinations of temporal reference in relation to both time of speech, time of event, time of relative event, etc..

I’m going to stop here for today. In the next half of chapter 3 Aubrey goes on to talk about typology and the Perfect more specifically.

I’ve started reading Mike Aubrey’s masters thesis, which I suggest you do too. To get the statement of bias out of the way, I have vague sympathetic feelings towards Aubrey based entirely on internet interactions, and recommend his updated-just-enough-to-keep-you-subscribed blog. Apart from that I have no vested interest in talking him up.

Why should you read his masters thesis? 1. It’s about Greek. 2. It applies linguistics, not just grrrammmar (I didn’t mean to write it like that, but my keyboard spat it out thus and I thought it worth keeping!), 3. It’s about the perfect, kind of. And everyone loves to talk about the perfect these days. 4. If nothing else, by the end of it you will have absorbed so much technical jargon that you will be able to bamboozle your interlocutors into silence on any topic, especially unrelated ones.

You should know, too, that I am not formally trained in linguistics. Studying 5.5 languages and having a perverse interest in the field turned me into a paralinguist, and leeching off my wife’s textbooks while she did an MA in applied linguistics have given me all the hubris of someone who thinks they know linguistics but probably would fail the intro course.

What’s Aubrey’s thesis doing? You can understand it as basically testing two things going in different directions: on the one hand, it’s testing some of the limits of Role and Reference Grammar on its ability to deal with “detailed semantics of distinctions” (p2, quoting Butler 2003), especially related to temporality and aspectuality (tense and aspect). So it’s a thesis putting RRG to the test. How? by applying RRG to the Greek perfect. In the other direction, it’s testing a hypothesis about the Greek perfect, so the outcome ought to be a better notion of RRG in the area of semantic distinction, and conversely a better understanding of the Greek perfect. Great!

In chapter 2 of his thesis, Aubrey sets to work to talk about what RRG is and what it does. RRG breaks things down around 2 ‘representations’, Syntactic and Semantic. What I’m going to do is try and give you a dummies guide to Aubrey’s thesis. So, syntactic has to do with how you arrange the relationships between words (and word-y things, linguistic units, but let’s just stick with words in this version), and semantics has to do with meaning. Good? Great. RRG has a bunch of neat ways of talking about different states, events, and processes, and a few ways to carve up the space of syntax.

RRG is basically realist about language. Language talks about stuff. To do that you must have ‘predication’ and you must have ‘reference’. I.e, you must talk about something (a referent), and you must say something about it (predication). Language is a tool to achieve communication: to say Xs about Ys.

So onwards, RRG treats semantics as the basis for syntax. How one talks about the structure of language utterances is determined by the ‘what’, the meaning, of that utterance. This is why the predicate is the syntactic nucleus. It’s the heart of the heart of what’s going on in a clause! Then you add in an argument or two, and voila! it’s a core. What’s an argument? it’s an X or a Y you substitute into sentence to make it work. e.g. John eats violins. ‘John’ is an argument. ‘violins’ is an argument. They are variables that we use with a nucleus, with a predicate, with ‘eats’, to make a core.

I’m going to skip some stuff about operators. Let’s go to 2.1.2 Semantic representation. In this section RRG breaks down “what verbs do” into different classes. Aubrey presents 6

1. States
2. Achievements
3. Accomplishments
4. Activities
5. Semelfactives
6. Active-achievements

A state just describes what is. “I am dead”. There is nothing dynamic about a state. It doesn’t have to be an unchanging state. “I am hungry” is a state, but not an eternal one. All the rest are dynamic, they involve action, movement, change. Achievements and accomplishments sound like they are the same thing, but they’re not. “The window shattered” is an achievement: something happened and a new state was achieved, but the action was ‘all at once’, or ‘punctual’. “The window froze” is an accomplishment: it didn’t all at once freeze, it underwent a process of getting colder until it reached a state of being frozen. Activities are dynamic, but they don’t necessarily have an end point. “I swam” describes an activity.

The last two are expansions, as you can see from semel-factive’s breakdown, it is an achievement that is simultaneous but doesn’t involve a change of state. Whereas an active achievement takes an Activity and adds an endpoint. “I swam to work” for example.

There’s a really great table at 2.4 which shows how all these 6 classes are divided up by whether they are +/i Static, Dynamic, Telic, Punctual. Telic, if it’s unclear, is whether it has an end point.

Not only this, but just after 2.5, Aubrey fills you in on various ways to ‘test’ to see whether a predicate can be classified into each of the 6 categories.

What does all this have to do with Aspect? Good thing we’re reading a thesis on it. In section 2.2 Aubrey notes that there is a pretty big ‘gap’ between RRG’s description of different types of predicates, and the question of aspect. It’s the problem that Aspect really lies in the realm of operators, things that ‘do stuff’, neat stuff like, ‘mark and show you the aspect of the verb’.

When you come to tense and aspect, these are part of “the difficult issues of peripheral and complex grammatical categories”. So what’s a poor Greek linguist to do? Aubrey says, that what RRG needs is some better categorisation and tools for classifying what’s going on with tense and aspect. And that’s chapter 3.