On the Compositional Nature of the Aspects
Verkuyl Henk

Abstract:
This study aims to make for a better understanding of the term ‘Aspects’ in linguistic theory. Its most current application is found in studies on Slavonic languages. In the abundant literature on the contrast between the Durative (or Imperfective) Aspect and the Nondurative (or Perfective) Aspect, their occurrence has been taken to be restricted to Slavonic and some other languages, generally speaking to languages whose Verbal systems are morphologically characterized with regard to this opposition.
The central hypothesis of transformational-generative theory that a distinction should be made between the deep structure and the surface structure of a language, entails the possibility for morphological systematicity to be nothing more than a manifestation of a general or even universal regularity expressed, for example, in the syntactic component of grammers of other languages. It will be shown in this study that the opposition between the two Aspects is present in Dutch, and as can be seen from the translated material, also in English, and that it should be described as the expression of regularities of a primarily syntactic-semantic nature.
In Chapter One I shall discuss the plausibility of the view that in non-Slavonic languages generalizations can be made pertaining to phenomena which in Slavonic grammars are generally accounted for in terms of the-contrast between the Durative and Nondurative Aspects. As this opposition should be related to the possibility for Durational Adverbials to occur in some sentences and to their exclusion from others, it follows that we cannot consider the Aspects free from selectional relations between these Adverbials and the constituents to which the term ‘Aspects’ applies.
In Chomsky’s Aspects of the Theory of Syntax (1965) we meet the claim that Durational Adverbials relate selectionally to Verbs, an opinion which in fact links up with the traditional assumption mentioned above that the Aspects should be assigned to Verbs. Even if we take into consideration recent modifications made by Chomsky  in Remarks on Nominalization (1968a) many inadequacies inherent to the model proposed in Aspects make it rather senseless to pursue other ways of remedying its inherent limitations, notably with respect to the status of Prepositional Phrases, to the organization and the function of the lexicon and to the specification of Verbs and Nouns.
Gruber’s abandonment of the principle of monocategorial lexical attachment which states that for every lexical item there is just one categorial node in deep structure to which it can be attached, as well as his basic assumption that no essential difference should be made between syntax and semantics at the level of deep structural representation, should be seen as a basic contribution to what in the work of Lakoff, Ross, McCawley, Postal and others has been developed as an alternative to Chomsky’s original and modified conception of transformational-generative theory.
Since Gruber’s work is relatively explicit as to the application of the principle of polycategorial lexical attachment, which allows lexical items to be attached to more than one category of the base, and since in his grammar Verbs can inherently be specified in terms of semantic primitives, his framework opens up perspectives for an adequate account of the Aspects in grammar. A brief survey of Gruber’s basic assumptions applied to the main theme of this study will follow the refutation of the Aspects-model and its extrapolations.
In Chapter 2 the compositional nature of the Aspects will be demonstrated with the help of a number of outwardly diverse sentences, all of which allow for the same generalization regarding the position of Durational Adverbials. The Durative and the Nondurative Aspects in these sentences appear to be composed of a Verbal subcategory on the one hand and a configuration of categories of a nominal nature on the other. Accordingly they can be represented by schemata. Sentences like "*Hij speelde een uur-lang het celloconcert van Schumann" (lit: He played Schumann’s cello concerto for an hour), "*Greetje wandelde urenlang een kilometer" (lit: Greetje walked a kilometre for hours), "*De muis at een weeklang de kaas" (lit: The mouse ate the cheese for a week), "*Hij hoorde urenlang dat De Gaulle was overleden" (lit: He heard for hours that De Gaulle had died), "*De chirurg genas een maandlang een hartpatiént" (lit: The surgeon cured a heart patient for a month), can all be analyzed on the basis of the following Nondurative scheme: V[VERB]V +NP[SPECIFIED QUANTITY OF X]NP. The asterisk is used to indicate that these sentences are ungrammatical in their single-event reading; they can, however, be interpreted as expressing frequency, though somewhat unnaturally. The category VERB represents one of the categories MOVEMENT, PERFORM, TAKE, ADD TO, TRANSITION, DO, etc., whereas the categories SPECIFIED QUANTITY OF X pertain to the countability, finiteness, or delimitation of X. Thus a cello concerto when performed should be conceived of as a finite piece of musical information, a kilometre when covered as a limited quantity of distance measuring units, a slice of bread when eaten as a specified portion of bread, and so on.
Sentences like "Hij speelde een uurlang cellomuziek" (He played cello music for an hour), "Greetje wandelde urenlang" (Greetje walked for hours), "De muis at een weeklang van de kaas" (The mouse ate from the cheese for a week), "De muis at een weeklang kaas" (The mouse ate cheese for a week), "De chirurg genas een maandlang hartpatiénten" (For a month the surgeon cured heart patients) and "Hij hoorde urenlang praten over De Gaulle's dood" (He heard people talking about De Gaulle’s death for hours) all fit into the Durative scheme V[VERB]V + NP[UNSPECIFIED QUANTITY OF X]NP. Eating from the cheese is to take (separate) an unspecified quantity of cheese from a (specified) quantity of cheese; music can be played for a virtually infinite time, and given the eternal life of Greetje she can walk infinitely. The surgeon restricted himself for a month to curing only members of the set of heart patients. In all these cases the termination of the events is dependent on the length of the stretch of time given in the Durational Adverbials.
In Chapter 3 it will be shown that the Indirect Object, as well as the Subject of the sentence, is also involved in the composition of the Aspects. Thus, in the latter case there is an opposition between sentences like " Er stroomt urenlang een liter water uit de rots" (lit: There is streaming a litre of water out of that rock for hours) and a sentence like "Er stroomt urenlang water uit de rots" (Water is streaming out of that rock for hours), which can be explained in terms of the schemata above. That is to say, the term ‘Aspects’ applies to configurations of categories of the following form NP[(UN)SPECIFIED QUANTITY OF X]NP + VP[V[VERB]V (+ NP[(UN)SPECIFIED QUANTITY OF X]NP)]VP. This very fact invites the question of where the upper bound of the Aspects should be located. Since there are selectional restrictions between the categories fitting into the Nondurative scheme and Durational Adverbials, it can safely be assumed that those constituents which are located “higher” than Durational Adverbials are not involved in the composition of the Aspects. Therefore attention will be given to the status of Durational Adverbials in underlying structure.
In Klooster and Verkuyl (1971) it is argued that sentences containing the Verb 'duren' (last) and those containing Durational Adverbials having an Indefinite Determiner are transformationally derived from one underlying source. Both 'duren' plus its Specifying Complement and the Durational Adverbial can be considered the Verb Phrase of a sentential structure whose Subject is an embedded S referring to events. In other words, these Durational Adverbials can be taken as predications over events. The rule transforming the underlying Predicate into a Durational Adverbial is called Adverbialization.
The Nondurative scheme can be used as a constraint on this transformational rule. This constraint states that if the sentential Subject of a sentence is specified as referring to one single event and if this Subject S dominates categories fitting into the Nondurative scheme, Adverbialization cannot take place. This amounts to saying that the VP containing the categories which constitute the input to this rule cannot be transformed into a Prepositional Phrase. Consequently we can block the sentences with the asterisks above in their single-event reading. To account for the tendency to interpret them in a frequency reading we could say that Adverbialization may take place if S_{n>1}. That is, if the S referring to Nondurative events is pluralized, Durational Adverbials may be derived. This requires that we specify what a single event is, i.e. which unit it is that can be pluralized.
In Lakoff and Ross (1966) attention was given to the question of how to determine the mutual relations between the constituents dominated by the then deep structural category Predicate Phrase, among which Durational Adverbials. They proposed a transformational rule, called Do so-replacement, which was intended to make explicit a criterion determining which constituents are to be located inside and which outside the Verb Phrase. Their proposal can be taken as an implicit attempt to determine the notion ‘event’: constituents derived via Do so-replacement and necessarily occurring inside the VP seem to make a necessary contribution to the reference to events. Therefore their do so-analysis will be closely investigated.
According to Lakoff and Ross Durational Adverbials were, in contrast to Chomsky’s position, generated outside the VP. As far as I can see, recent literature offers no serious proposal stating that the node Predicate Phrase is a deep structural entity. At best we can consider it a category developed during the transformational derivation. This renders the Do so-rule inadequate and consequently we need to restate the generalizations made by Lakoff and Ross. '
Starting from a proposal by Staal (1967) the underlying structure of an “action sentence” like "Arie at een haring" (Arie ate a herring) is described as Arie + DO + S[x EAT A HERRING]S, where S represents a propositional function and where DO + EAT form the Verb 'eten' (eat). At a further stage of our analysis a more  abstract  representation  of this structure results in a proposal linking up with proposals made by logicians like Reichenbach (1966) and Davidson (1967), the crucial point being that we need to introduce temporal variables into our description. That is, a sentence like "Arie at een haring" (Arie ate a herring) should be analyzed in terms of an existential quantification over events. This point as well as a formalization and extension of the Bach-McCawley proposals about the underlying structure of Noun Phrases in terms of Gruber’s framework enables us to relate the sentence under analysis transformationally to such sentences as "Wat Arie deed was een haring eten" (What Arie did was to eat a herring), "Wat Arie at was een haring" (What Arie ate was a hérring) and "Degene die een haring at was Arie" (The one who ate a herring was Arie). The structure underlying all four sentences (apart from possible topicalization elements) is roughly of the form NP[the x — who is Arie]NP VP[DO NP[PRO[SOME v]PRO REL[WH+v BE S[x EAT a y — which is a herring]S]REL]NP]VP, where v is a variable ranging over temporal entities, notably over events. The paraphrase related to this structure is something like ‘The one who is Arie occurs as the agens with respect to some event which is his eating something which is a herring’. In this structure DO expresses the relationship of ‘agency’ between Arie and some event.
By introducing event-variables into our description we obtain the equipment necessary to account for the pluralization of events, for frequency. The term ‘frequency’ applies to a series of similar events and therefore we ought to know the unit of quantification, the event-unit.
However, as Reichenbach pointed out, the determination of individuals is a matter of convention: we can expand events. This insight is supported by the flexibility of the scope of reference of dat (so) in a sentence like Arie at een haring en Piet deed dat ook (Arie ate a herring and Piet did so too) and of wat (what) in the pseudo-cleft sentence Wat Arie deed was een haring eten (What Arie did was to eat a herring). The latter sentence provides a paradigm that can be applied to determine the minimal structure some entity must have to be an event. It is exactly the S referring to minimal events which turns out to be the upper bound of the Aspects.