Microsoft word - chapter10site.doc

Conclusions and issues for future research 10.1.1 Conclusions about the description of genericity in the literature The first part of this dissertation discussed the description of genericity in the (formal) semantic literature and in some traditional grammars of Dutch. The following main • Habitual sentences cannot be treated on a par with characterizing sentences. One contrast between habitual and characterizing sentences is that indefinite singulars in object position of habituals necessarily take a wide scope reading. This can be accounted for by assuming that the habitual quantifier HAB deviates from the generic quantifier GEN in not having (potentially) sentential scope (cf. also Rimell 2004). This was made more concrete by assuming that HAB takes scope from its position in the specifier of a projection we referred to as F2P. Indefinite singulars are treated as elements with quantificational force of their own: they undergo quantifier raising to adjoin to specifier positions higher in the tree than F2P. From this position, they take • Cohen (2001) observes that sentences with indefinite singulars such as A man doesn’t cry attribute definitional properties to indefinite singulars. We have argued in favour of Greenberg’s (2002) account of the subtle semantic contrasts between characterizing sentences with indefinite singulars and sentences with bare plurals (like Men don’t cry). Greenberg assumes that such sentences have the same basic semantic structure, but differ in the choice of accessible worlds where the generalizations are evaluated. In the case of sentences like A man doesn’t cry, the accessible worlds are restricted to worlds in which every man has the “in virtue of” property, i.e. being tough. • We have made the assumption that kinds and objects play the same semantic role (namely they both correspond to entities). This only makes sense if kinds and objects are similar from the perspective of the syntax-semantics interface and if the same strategies can be used to attribute properties to both kinds and objects. Sentences such as Every rhino is threatened with extinction illustrate that quantificational elements can range over sets of kinds in the same way that they can range over sets of objects. We showed that such sentences can be represented by assuming that it is possible for common nouns like rhino to refer to sets of kinds. Under this assumption, we expect determiners and other (quantificational) elements to interact with taxonomic common nouns in the same way as with common nouns referring to sets of objects. • In episodic sentences such as In Africa, I filmed the African elephant, there is a mismatch between the predicate, which does not normally accept kinds and a direct object noun phrase, which denotes a kind. Such mismatches can be resolved by assuming a realization relation. Cohen (1999: 46) argues that the sets of instances which are relevant for the interpretation of a sentence vary from case to case. Cohen refers to each realization relation as a coordinate C. Cohen does not discuss episodic sentences such as In Africa, I filmed the African elephant. We have made the assumption that in such sentences C stands for the representative object coordinate. This means that C(↑afrikaanse-olifant)(x) holds just in case x is a representative object of the kind corresponding to Afrikaanse olifant (‘African elephant’). • The crucial distinction made in traditional grammars is the one between categorial and generic nominal constituents. There are three important criteria on the basis of which it can be determined whether a nominal constituent is categorial or generic. A first criterion is that categorial constituents semantically correspond to “every member of a class/category”, while generic constituents correspond to “kinds”. A second point is that categorial nominal constituents cannot be used with predicates like uitgestorven zijn (‘be extinct’), which express properties of kinds. Only generic constituents can be used with such predicates. A third criterion is that categorial nominal constituents are introduced by een (‘a/an’) or by the null article, while generic constituents are introduced by the definite article de/het (‘the’). This distinction and the corresponding criteria are not unproblematic. According to a number of speakers, predicates such as met uitsterven bedreigd zijn (‘be threatened with extinction’) can be combined with bare plurals. This is not in agreement with the assumption that there is a one-to-one relation between indefinite articles (i.e. een and the null article) and categoriality. A second point is that traditional grammars do not assume genericity or one of the relevant phenomena to be a property of whole sentences. As a consequence, the widely shared intuition that genericity is to a considerable extent a property of sentences is not reflected in the traditional terminology. 10.1.2 Conclusions of the corpus and questionnaire studies In the second part of this dissertation, a number of corpus- and questionnaire-based studies were presented and discussed. The following main conclusions were drawn: • The first corpus study presented in chapter 6 has shown that there are significant correlations between corpus frequencies and taxonomic hierarchies: higher taxonomic levels correspond to a lower frequency of definite singulars (and, correspondingly, to a higher frequency of bare plurals). This is in agreement with observations made in the literature (cf. Heny 1972 and Carlson 1977). • Another corpus study has shown that different lexical semantic classes lead to different frequencies of types of DPs. Definite and indefinite singulars are used more frequently with animal names than with nationality names; definite plurals appear • A third study has revealed some significant differences between three highly frequent common nouns. Definite and indefinite singulars appear more frequently with mens (‘human/man’) than with man (‘man’) and vrouw (‘woman’). The opposite result was found for bare plurals. The largest difference between mens on the one hand and man and vrouw on the other was found for definite singulars. This result was related to the distinction between natural kind terms and nominal kind terms. While mens falls in the category of natural kind terms, man and vrouw are nominal kind terms. • The observation that lexical semantic classes are highly relevant to the study of generics was confirmed by the questionnaire results. A number of informants judge definite plurals more acceptable with nationality names than with animal names. This is in agreement with the fact that in corpora definite plurals are used more frequently with nationality names than with animal names. • Another result from the questionnaire study is that according to a considerable number of informants definite mass DPs are more acceptable in kind predicate sentences than in characterizing sentences. A similar, but less strong result was found for definite 10.1.3 Conclusions about issues on the syntax-semantics interface The third part of this dissertation was devoted to issues on the syntax-semantics interface. The main conclusions are summarized in the following points: • A number of arguments have been given in favour of the conclusion that Dutch (as well as English) bare plurals do not unambiguously refer to kinds, but are ambiguous between a kind-referential and a non-kind-referential, variable-introducing reading. This accounts for the ambivalent behaviour of bare plurals: On the one hand, kind predicates can take bare plurals; on the other hand, there are speakers who judge such sentences to be unacceptable. A second important observation is that bare plurals are ill-formed in object position of kind predicates, but they can be used in object position under characterizing interpretations. Thirdly, bare plurals do not necessarily refer to entities that are well-established as kinds. A last argument is that bare plurals do not receive representative object interpretations in episodic sentences. • Some of the argumentation in favour of the claim that bare plurals are ambiguous could not be applied in a straightforward way to bare mass arguments. Yet, the most plausible interpretation of the evidence is that bare mass arguments are ambiguous as well. Crucially, there are speakers who do not accept kind-referential readings of bare mass nouns in direct object position or in the complement position of a preposition. At the same time, sentences in which bare mass nouns receive a characterizing interpretation are grammatical beyond doubt. As subtle as the relevant observations may be, they cannot be accounted for if we assume that bare mass nouns • The influential theory of Longobardi (1994) leads to a number of problems and open questions. A first problem has to do with Longobardi’s assumption that expletive articles are licensed only if they express grammatical features, or as a last resort. He assumes that this is a “principle of Universal Grammar”. Such a principle does not predict that in a number of varieties of Dutch the acceptability of DPs with definite articles is influenced by the lexical semantic class of the common noun. We have shown that this cannot be accounted for straightforwardly by claiming that definite articles are acceptable only if they spell out morphological features or as a last resort. Other difficulties have to do with Longobardi’s account of subject-object asymmetries in terms of Lexical Government and his assumption that Germanic bare arguments do not have to fulfil this condition, as a consequence of raising from N0 to D0 at LF. One problem for the latter hypothesis is that there is no (substantial) empirical or other evidence that Dutch or English has overt raising from N0 to D0 triggered by the constraints suggested by Longobardi (1994). If there had been such evidence, it would have been reasonable to assume that, under some well-defined conditions, N-to-D raising can take place at LF. Furthermore, English and Dutch bare plurals are (relatively) unacceptable in object position of kind predicates (cf. ??The Sumerians invented pottery wheels). We have shown that Longobardi’s (1994) approach cannot account for these data in a straightforward way. • Some arguments were presented against Chierchia’s idea that Germanic languages differ from Romance languages in that Germanic arguments are not necessarily introduced by (empty or non-empty) determiners. Chierchia accounts for the semantic ill-formedness of characterizing sentences such as #The beavers build dams by assuming the economy principle Avoid Structure. The crucial point is that sentences like Beavers build dams (can) receive a characterizing interpretation. Chierchia claims that when the latter option is available, it must be chosen over one which involves projecting D. The problem is that a number of speakers of Dutch (and German) accept the use of definite articles in characterizing sentences such as De bevers bouwen dammen (‘The beavers build dams’). Such data contradict Avoid Structure. Possible assumptions to account for the German and English data are that Avoid Structure is active only in some varieties of Dutch and German or that some varieties have the same setting of the Nominal Mapping Parameter as Romance languages, while other varieties are similar to English. However, we argued that these suggestions have their • We have offered a detailed inventory of the distributional and selectional properties of empty determiners and definite articles, based on the assumption that in the languages under consideration, argument noun phrases are headed by a determiner (empty or non-empty). On the basis of the semantic and selectional properties of determiners, nine types of empty determiners can be distinguished. A language does not necessarily possess each of these determiners. Standard Dutch does, for example, not have an empty singular count determiner with a kind-referential interpretation (i.e. a phonologically empty determiner of the type [+R,+count,-pl]). Thus, we can deal with the fact that singular counts without overt determiners do not receive this reading. • The selectional properties of definite articles are more complicated. We have found that in many local varieties of Dutch kind-referential definite articles are only compatible with a restricted set of common nouns. It is, for example, possible, that they are acceptable only with nationality names. This can be dealt with by assuming that definite articles have semantic selectional features. For example, an article with a feature corresponding to [+human] cannot be combined with an animal name. Another observation about kind-referential DPs with definite articles is that there are a number of varieties in which they cannot be used in characterizing sentences, but are judged acceptable in argument positions of kind predicates. We have proposed that there is a relation between this observation and the fact that empty determiners are in many cases problematic in such positions. This suggests that the relevant varieties do possess definite articles introducing kind-referential DPs, but they have to fulfil an additional condition: they must be the ‘optimal’ output, i.e. competing structures with empty determiners have to be problematic. This dissertation has provided answers to a number of questions, but leaves a number of other questions for subsequent study. The following issues are left for future research: • Definite plurals, such as de bevers (‘the beavers’) and definite mass DPs, such as de melk (‘the milk’) are relatively marked in kind predicate and characterizing sentences. We have found that although a number of speakers judge characterizing sentences such as De bevers bouwen dammen (‘The beavers build dams’) acceptable, there are also speakers who do not accept them. Definite mass DPs are subject to less inter- speaker variation than definite plurals: characterizing sentences like De melk is wit (‘The milk is white’) are judged unacceptable by almost every speaker of Dutch. However, there are contexts/sentences in which such DPs are acceptable. For example, a sentence like De witte wijn is duur dit jaar (‘The white wine is expensive this year’) is acceptable. The conditions under which definite mass and plural DPs are acceptable in characterizing and kind predicate sentences have not been (fully) clarified in this study. More research is needed to clear up this point. • A related aspect is that sentences such as De bevers bouwen dammen (cf. above) are ambiguous between two interpretations, which can be extremely difficult to distinguish. In the kind-referential interpretation, de bevers refers to a kind. In a second interpretation, de bevers refers to a contextually determined group of beavers. However, a speaker can assume that this group is the group of all beavers living in the actual world. Note that there is an important semantic difference between the two readings in that the latter is a purely extensional interpretation. The question is, however, whether or not language users can make such subtle distinctions. An interesting question is whether this problem can be neutralized by using more sophisticated verbal and non-verbal questionnaire techniques. More research needs to • Future research could extend this study to include other noun phrase types, like brand names, plant names, substance names, names for inhabitants of cities or villages and names of political groups in futures studies. Another suggestion is that it would be interesting to incorporate other types of predicates/sentences (for example quantificational predicates like zeldzaam (‘rare’) or representative object sentences such as In Alaska filmden we de grizzly beer (‘In Alaska, we filmed the grizzly’)). Furthermore, future research could include more syntactic positions and differentiate between different types of Prepositional Phrases. • The questionnaire study presented in this dissertation focuses on intuitions about the acceptability of sentence types in local/regional varieties. More research needs to be done to shed light on the variation among Standard Dutch speakers (and in general among speakers of the same language) in their judgements about the acceptability of different types of characterizing and kind predicate sentences.

Source: http://www.albertoosterhof.be/chapter10.pdf