Conjunction and disjunction from a typological perspective

Our project investigates the meaning of and, or and their counterparts in other languages. These two classes of expressions (henceforth referred to as AND and OR) seem to have a central status in natural language grammars. Almost all languages express at least one of them, and in many languages, they can be used to combine expressions of almost all grammatical categories. For this reason, linguists are interested in developing unified theories of their meaning that let them combine with expressions of any kind. There are two conflicting research traditions in this field. One of them assigns a so-called Boolean meaning to AND. Informally speaking, the meaning of any expression involving AND is reduced to the meaning of an expression in which AND combines with full sentences. For instance, (1-a), in which AND combines with two verbs, is reduced to (1-b). (1) a. Jane drinks and smokes. b. Jane drinks and Jane smokes. However, there are examples which cannot straightforwardly be reduced to conjunctions of sentences. For instance, (2-b) sounds odd and cannot express the meaning of (2-a). For this reason, many linguists assume an additional, non-Boolean meaning for AND. (2) a. Jane and Mary met in the bar. b. Jane met in the bar and Mary met in the bar. If AND had two unrelated meanings, one would expect that some languages systematically use different expressions for Boolean and non-Boolean AND. We suspect that this is not the case and conclude that one of the two meanings should be reduced to the other. For theory- internal reasons, we think that the non-Boolean meaning is probably the basic one. However, these hypotheses cannot be tested on the basis of the data currently available. Detailed studies of the meaning of AND only exist for a small number of languages. Therefore, our main goal is to collect the necessary data to examine the predictions of our hypotheses. This will be done using SSWL, an online platform which allows linguists to contact informants for a wide range of languages and ask them questions about their language. The informants enter their answers into a freely searchable database. In addition to collecting data on the grammar and meaning of AND in various languages, we aim to improve the existing theories of non-Boolean AND. If our non-Boolean approach to AND holds up, it also raises interesting problems for some current theories of the meaning of OR, which suggests that these theories need to be revised.

One core question concerning human languages is whether certain abstract semantic building blocks are common to all of them and, if so, what these abstract meanings are. Investigating this issue is directly relevant for our understanding of the human linguistic ability. Set within this enterprise, our project focussed on connectives -- elements like English `and` or `or`. They are particularly interesting because they are found in probably all languages and tend to have a broad distribution within individual languages, as they can often `connect` expressions of different syntactic and semantic categories (cf. English `Ann and Bert`, `sang and dance` etc.). Our core question was whether all languages assign the same meaning to them and, if so, which. We concentrated on conjunctive elements, i.e., elements corresponding to English `and`, investigating whether their basic meaning is a `group-forming` operation (i.e., the operation that underlies the meanings of plurals like `the girls`) or an operation that essentially connects sentential meanings (requiring that both sentences are true). The sentence in (1a), for instance, has two meanings. The meaning in (1b) is the basic one on the `group` hypothesis, whereas the meaning in (1c) is basic according to the `sentential` hypothesis. However, each hypothesis could posit additional rules or abstract parts of the structure of the sentence that would allow it to derive the `other` meaning. 1. a. `Sue and Ann ate exactly three bananas.`b. Sue and Ann ate exactly three bananas between them. (E.g., Sue ate one, Ann two).b. Sue ate exactly three bananas and Ann ate exactly three bananas. Using the Terraling database - a new online platform where native-speaker linguists answer data questions and which thus allows researchers to directly access the implicit knowledge speakers have about their languages - we formulated queries based on the following rationale: Based on languages like English, we cannot distinguish between the hypotheses mentioned above, because both are compatible with the data. Yet, they differ in their predictions concerning the cross-linguistic situation: The `group` hypothesis leads us to expect that the extra elements needed to derive the meaning in (1c) correspond to additional words/endings in some languages; and the reverse holds for the `sentential` hypothesis. In light of these predictions, our data set yields a very clear pattern that strongly supports the group hypothesis. Interestingly, this seems to extend to conjunctions of elements with more `abstract` meanings, like verbs or sentences, which indicates that `group-formation` is a very basic property of the grammar of human languages. Since the `sentential` hypothesis could be considered the traditional one in linguistic research, our results are surprising and have far reaching consequences for semantic theory and for the broader question sketched above.