There’s something I’ve found interesting about the semantics of perfectives in different languages, but which I don’t think is mentioned very often or paid much attention in general writing on aspect. Specifically, it’s about the interaction of perfective aspect with telicity, and the extent to which the meaning of a perfective clause depends on the classification of verbs or events into inherently telic or atelic.

My main references from this blog post are A Grammar of Coastal Marind by Olsson, Telicity and Durativity by Wilhelm, and The Oxford Handbook of Event Structure, Chapter 18, Inner Aspect Crosslinguistically.

Types of Perfective

Based on the above, there seem to be at least the following kinds of perfective:

Telicity-Sensitive Perfectives

A perfective aspect whose semantics are strongly dependent on the predicate’s inherent aspect is the “default” from a European point of view. With telic predicates, the perfective aspect entails that the event ended successfully, even if the predicates describe durative events which could be interrupted during execution. Although English is not a good example of a language with a strongly grammaticalised perfective/imperfective, it suffices to demonstrate the point. The following are not grammatical in English:

* I built a house, but I didn’t finish it

* I built a house, and I’m still building it

With atelic predicates, which have no inherent endpoint, only termination is entailed, not completion:

I danced, but I didn’t finish

I danced, and I’m still dancing

A perfective atelic clause is compatible with both cancellation before some implicit typical endpoint, and continuation after termination.

Strong Perfectives

According to Wilhelm, Dënesųłinë́ has a perfective which always entails completion. Examples which add continuation are ungrammatical:

* yághelgus-ú ʔąłų yálgus

He.jump.PRF-and still he.jump.IMPFV

“He jumped and he’s still jumping”

(perhaps more literal: ?? “He finished jumping and he’s still jumping”)

The strong perfective of Dënesųłinë́ imparts a quasi-telic meaning to all predicates. With predicates which are atelic, it is often best translated into English with “finish”. Similarly, with stative predicates like “be sitting”, the imperfective is used while the state holds, and the perfective implies that the subject is no longer seated. Inherent telicity of predicates therefore matters less, because successful completion is a core meaning of the inflectional category of perfective.

Weak Perfectives

The final type is a perfective that only encodes termination, and not successful completion, regardless of the meaning of the predicate. For a punctual predicate, without inherent duration, this makes little difference, because starting is the same as finishing so failure is, in a sense, impossible. You can either knock or not knock on the door, but it requires a special context to start knocking and yet fail to knock.

But for predicates which are telic but also durative, i.e. the Vendlerian class of accomplishments, stopping is not the same as finishing. And in languages with weak perfectives, successful completion of accomplishments is a cancellable implicature of the perfective aspect, not an entailment.

One implication of such semantics is that weak perfectives look inception-oriented. While strong perfectives can be translated by “finish”, a perfective which can be used no matter how long the event took place for, or how it ended, can often be informally translated by “begin”.

Languages with such a weak perfective do not seem to be that uncommon. Coastal Marind is such a language. To quote from Olsson’s grammar, where he discusses the semantics of Marind verbs and their interaction with inflectional aspect:

An event such as running can be thought of as consisting of three parts: the onset or initial boundary (start to run, take off running), the ensuing activity phase (be running), and the endpoint or final boundary (stop running). We can refer to these three parts as the ingressive part, the durative part, and the completion of the event. The most central concept in discussions of aspect has been completion, because linguists traditionally described the major difference between verb forms in e.g. Ancient Greek, Slavic, and Romance languages as either implying completion (aorists or perfective forms) or not implying completion (imperfective forms).

In Coastal Marind, the notion of completion is largely irrelevant for the description of the tense-aspect morphology. Instead the main distinction is between situations that are presented as ingressive (i.e. referring to their initial boundary or starting point) and situations that are presented as durative. One facet of this is visible in lexicalisation of actionality distinctions (or 'Aktionsart'): Coastal Marind almost completely lacks state-denoting verbs such as 'be hungry' or 'be round', but there is a large number of verbs denoting the onset of such states, e.g. wahun 'become hungry', and ibayeb 'become round'. The corresponding states are expressed by derived Extended forms, e.g. wahut-a 'be hungry'. This pattern recurs with e.g. motion verbs, so there are verbs such as tapeb 'fly up, take off flying', but not verb simply meaning 'fly'. To refer to the durative part of such situations one uses the derived Extended form, e.g. tapeb-a 'be flying, be in flight'. 

A second concomitant of the importance of the initial boundary of events is that Coastal Marind appears to lack any grammaticalised resources that imply the completion of an activity. I made many attempts to probe this by discussing sentences ... with speakers, but I failed to identify any (affixal or non-affixal) material that narrows down the meaning of activities such as working, cooking, eating, etc. to include their final boundary. I suggest that telicity is not grammaticalised in Coastal Marind (cf. Smith's analysis of Navajo as lacking grammaticalised telicity; Smith 1997: 297).

While Olsson’s discussion focuses to a large extent on ingression, I suspect that the inceptive meaning for the verb forms not marked as durative or imperfective may be downstream of the de facto perfective simply lacking completion semantics for any aksionart class. Verbs can still have completion and subsequent state as an implicature:

kay k-a-w-a keway

road DIR-3SG.A-3SG.U-AUX break

“the road broke”

“the road is broken”

Multiple Options, or Grammaticalised Completion Marking Combined with (Im)perfective Marking

While Marind does not have a culmination affix or construction, or have any grammaticalised means to assert successful completion, some languages do. In The Oxford Handbook of Event Structure, Travis discusses the distinction between two such forms in Austronesian languages like Tagalog and Malagasy, and the well known control inflections of Salish, in the context of nonculminating accomplishments.

The first interesting thing here is that, in all the listed examples below, explicit culmination is the marked option. While there are languages that have affixes for unexpected non-culmination (frustrative marking, often translated as “in vain”), such markers tend to perhaps be semantically simpler and be less deeply entangled in the aspectual system. Oddly, Marind has such an inflectional frustrative marker to explicitly deny culmination, even though it lacks any inflectional way to mark culmination.

A second interesting this is that, if a language offers a inflectional choice between asserting completion and leaving it as an implicature, the form which encodes completion consistently seems to acquire additional meanings: limited control (“manage to…”), accidental action, and ability. This is maybe not surprising, because choosing to assert that an action terminated successfully implies that there was some doubt, and if there was doubt then there must be a reason for such doubt.

Finally, such inflectional pseudo-telicity may cross-cut a perfective-imperfective distinction. Clearly a progressive reading is not compatible with asserting culmination, but a habitual, potential, or iterated reading is. In Salish, combining culmination with the imperfective or irrealis seems to give an abilitative reading, while the limited control and accident readings are perhaps less likely in such a construction.

Languages which mark both separately in this way decouple marking an event as closed and marking it as culminating. The following represents my impression of what this decoupling may often end up meaning:

	Imperfective (open)	Weak Perfective (closed)
-Culmination Assertion	Default progressive, habitual etc.	Default in narrative for sequences of events, culmination as implicature
+Culmination Assertion	ability ?? habitual unexpected culmination	mirative / strong assertion manage to accidentally

Austronesian – Tagalog

For this reason, the culminating forms in Austronesian are labelled AIA, for Ability and Involuntary Action. Compare the normal perfective and AIA marked forms of the same verb in the following example:

Pumunta sa Maynila si Pedro pero naligaw siya, kaya hindi siya nakapunta

N-AT.PERF-go DAT Manila NOM Pedro but get.lost NOM.3SG, hence not NOM.3SG AIA.AT.PERF.go

“Pedro went to Manila, but got lost and didn’t get there”

The negated second verb, nakapunta, is AIA marked, and therefore asserts that the going event did not reach its successful conclusion. It could maybe be better translated as “manage to go (there)”.

Compare also the following, and the additional senses added by the completive AIA form:

Kinunan ni Ben ng litrato si Luisa

N.PERF-take GEN Ben GEN picture NOM Luisa

“Ben took a picture of Luisa”

Nakunan ni Ben ng litrato si Luisa

AIA.PERF-take GEN Ben GEN picture NOM Luisa

“Ben managed to take a picture of Luisa”

“Ben accidentally took a picture of Luisa”

Salish – St̓át̓imcets

St̓át̓imcets marks completion by the “out of control” circumfix ka- -a. By default, the culmination of durative events is cancellable:

máys-en-lhkan ti q’láxan-a, t’u7 cw7a7 t’u7 kw-s tsúkw-s-an

fix-TR-1SG.SU DET fence-DET, but NEG just DET-NOM finish-CAUS-1SG.ERG

“I fixed a fence, but I didn’t finish”

With ka- -a, the same readings familiar from the Austronesian AIA forms occur again:

wá7=lhkalh=t’u7 ka-nás-a ekw7úna Sát’=a snímulh

IMPFV=1PL.SUBJ=ADD KA-go-A right.over.there Lillooet=EXIS us

“We can go to Lillooet by ourselves”

ka-gwél-s=kan-a

CIRC-burn-CAUS=1SG.SUBJ-CIRC

“I managed to get it lit”

Summary

As Olsson said in his grammar of Marind, there’s a tendency in the European linguistic tradition to conflate perfectivity with culmination, because European perfectives and aspect forms are sensitive to lexical aspect, and because for telic verbs, European perfectives do entail culmination. It’s always interesting when extrapolation based on what’s familiar misses part of the picture.

I do suspect, though, that the framing in terms of inceptivity is maybe too strong for Marind, and the actual cross-linguistic distinction here is between weak (terminating) vs strong (culminating) perfectives, and the locus of telicity marking as a property of lexical items and open constructions, vs inflectional marking of pseudo-telicity.