ELSEVIER Lingua 102(1997)115-131 Aspectual scope and the difference between logical and semantic representation* Balthasar Bickel* Seminur fiir Allgemeine Spruchxkenschuft. Universitiit Ziirich. Pluttenstrusse 54, CH-8032 Ziirich, Snitzer-land Received 20 June 1996; revised version 12 November 1996 Abstract A common theory of aspect marking, which we might label ‘selection theory’, assumes that (viewpoint) aspect and Aktionsart (temporal schemes expressed by predicates or, compo- sitionally, by propositions) stand in an operator-operandum relationship, where aspect opera- tors select matching elements in an Aktionsart, thereby highlighting specific boundaries or phases. In this paper, I show that under the premise of a selection theory, we need to distin- guish between two types of nondefeasible (non-pragmatic) meaning representations: Whereas semantic representations are aspectologically relevant and inside the scope of aspect opera- tors, logical representations, i.e., representations including all truth conditions contributed by an expression, are outside the scope of aspect operators. This distinction allows a more prin- cipled notion of ‘aspectual pair’ in Slavic languages and has also important consequences for cognitive architecture in the domain of meaning representations. 1. Introduction Aspectology has long been concerned with the fact that grammatical (or ‘view- point’) aspect interacts in systematic ways with the temporal characteristics of pred- icates or whole propositions, i.e., with what is often referred to as lexical and sen- tential ‘Aktionsart’. Whereas traditionally the interaction between aspect and Aktionsart is often seen as a side-issue, this interaction has been taken as the core Some of the ideas of this paper were first presented at a workshop on the interaction between lexis and aspect in Constance, May 25-27, 1995. I thank the audience for the very stimulating discussion. I am especially indebted to Sabine Stall, who first convinced me how little semantics and how much prag- matics is involved in aspectual meaning, and who gave me good advice on Russian. Many thanks also go to Ronny Boogart, Kathrin Cooper and to the editors and reviewers of Lingua for helpful comments on earlier versions. I am also indebted to Steve Levinson for some input to a discussion on related issues long ago. All shortcomings and mistakes are of course my own responsibility. * E-mail: [email protected] 0024-3841/97/$17.00 0 1997 Elsevier Science B.V. All rights reserved PI1 SOO24-384 I (97)00004- 1II6 B. Bid4 i Liuguu 102 (IYWJ 11.5-131 concept in ‘selection theories’ of aspect (e.g., Johanson, 1971, 1996, forthcoming; Breu, 1984, 1985, 1994; Timberlake. 1985; Sasse, 1991a,b).’ The basic tenet of such theories is that, while having essentially the same representational format, aspect and Aktionsart are in an operator-operandum relation. Aspect operators can then be said to ‘select’ matching items in the Aktionsart, thereby locating specific event parts in the course of time. To give a classical example, we may briefly illustrate the theory by the truth-con- ditional effects of the French impallfclit discussed by Carey (1957): (1) a. 11 se baignait. + 11 s’est baigd. ‘He was bathing.’ ‘He bathed.’ b. I1 se noyait. * + I1 s’est noye. ‘He was drowning’ ‘He drowned.’ Although the same aspecto-temporal forms are used in (la) and (lb), the truth val- ues of the forms in the imparfait differ dramatically: whereas il se haignait ‘he was bathing’ implies that the subject referent bathed, il se noyait does not support an analogous inference. (1 b) merely suggests that the subject referent was in a stage that COULD have culminated in being drowned but did not necessarily do so. In a selec- tion theory this difference is explained in a straightforward manner: The imperfec- tive aspect contained in the French imparfafait is defined as a selector of ‘phases’, i.e., of temporal extensions in an Aktionsart. In both (la) and (1 b), the imparfait selects a phase, abbreviated here as ‘cp’ - but in different configurations. The Aktionsart of se haigner ‘to bathe’ in (la) is [cp], i.e., it has temporal extension and is not specified for any boundaries. By contrast, the Aktionsart of se noyei- ‘to drown’ in (lb) is ‘telic’, i.e., it contains not only a phase, but also a final transitional boundary towards which the situation develops over time. If we use ‘r’ to symbolize a bound- ary, this can be represented as [cp r]. Selection of v, in [cp z] implies that there is another part of the event, viz. the final r, that is not selected and is therefore not nec- essarily implied by the form. By contrast, selection of ~1 in the Aktionsart [cp] of se haigner- in (la) implies that the whole event took place. Notice that the difference between telic and atelic predicates is a semantic differ- ence, i.e., a difference that arises from language-specific lexical properties rather than from nonlinguistic conceptualisation. Languages differ in the specification of Aktionsarten. Whereas, for instance, French mourir- and Engl. to die are telic (‘[cpr]‘) and include a pre-lethal phase (cf. he was dying, ii mourait), their Chinese ‘equivalent’ si denotes a sudden transition and its subsequent state ([r cp]) (cf. Valin and LaPolla, forthcoming). Against this background, it seems that aspect markers clearly operate on semantic representations. There are some grammatical markers that can operate not only on semantic repre- sentations, but also on pragmatic, i.e., inferentially enriched representations. An ’ Alternatives to selection theories try to define aspect categories without