Sense, reference and metalanguage in semantic comparisons
Since meanings are unobservable, we cannot examine their historical variation or cross-linguistic universality directly in the way that we might, for instance, explore the history or distribution of a particular sound. In principle, it is a fairly straightforward matter to answer the question of whether the nasal [n], for instance, is found universally, and of what diachronic developments it is likely to undergo. To do this, one examines every language in turn, or at least a large sample of languages, to see whether they all have sounds which meet the criteria for [n] – criteria established by the discipline of phonetics – and uses this as the basis for exploration of its historical developments. In semantics, things are not so simple. In this section, we will discuss two particular points:
• the difference between universality/variation of sense and universality/ variation of reference, and
• the problems of the appropriate metalanguage in undertaking cross linguistic or historical semantic study.
Sense and reference are both crucial aspects of meaning (see Chapter 3). But there is a big difference between them when we study meaning cross linguistically or historically: it is much easier to establish cross-linguistic identity of reference than of sense. To see why, imagine that we are con ducting an investigation into body-part terminology in the languages of the world. As part of this study, we want to test the hypothesis that all languages have at least one expression which has the meaning ‘skin’ (per haps among other, polysemous meanings). Questions like this are the stock-in-trade of the study of semantic typology, which we discuss in 11.4. In the course of this investigation, we discover an interesting situation in Warlpiri (Pama-Nyungan, Central Australia). In Warlpiri, the word for ‘skin’, pinti, is also used to refer to bark and peel, and these two other uses are just as literal as the ‘skin’ use itself. What conclusions should we draw from this about the sense of pinti? Pinti clearly refers to skin, but does it contain ‘skin’ as one of its senses? Perhaps Warlpiri doesn’t actually express the distinct meaning ‘skin’, but contains instead a single general meaning applying to all three types of referent simultaneously, along the lines of ‘outer layer of person, animal, tree or fruit’. In this case, we would have to claim that ‘skin’ isn’t a semantic universal, since it doesn’t independently exist in Warlpiri: pinti refers to skin, but ‘skin’ isn’t a separate sense of the word.
QUESTION Would this seem a reasonable conclusion? What are its advantages and problems?
The conclusion that pinti is general in meaning would also have consequences for historical study. For example, it would mean that there would be no point in asking questions about the diachronic origin of the ‘skin’ meaning, such as whether it developed from the ‘bark’ or ‘peel’ meanings, or vice versa. If pinti is general between these senses, these questions cannot be asked.
Another possibility, though, is that pinti is polysemous, with the three distinct meanings {skin, bark, peel}. In this situation, pinti not only refers to skin, it also contains ‘skin’ as one of its three separate senses. If this was the case, the status of ‘skin’ as a semantic universal wouldn’t be threatened: we could claim that the meaning ‘skin’ is found in Warlpiri, but that it is not individually lexicalized. This means that there is not a word which just expresses the meaning ‘skin’ on its own; ‘skin’ always comes along in a ‘package’ with other meanings included, even though each of the meanings is conceptually separate. Speakers of Warlpiri can obviously distinguish skin, bark and peel, as shown by the fact that they treat each in different ways. They can also distinguish between them linguistically at a phrasal level (e.g. a phrase like ‘pinti of animal’ can only mean ‘skin’). It’s just that in pinti, they’re all combined together. (See 5.3 for discussion of the problems that affect attempts to tell whether an expression is general or polysemous.)
Whether pinti is general or polysemous, its reference is determined by its sense. Determining its reference is, at least for practical purposes, easy enough: we can get a Warlpiri speaker to point, draw pictures, and so on. These activities have their own subtle ambiguities (see 2.3.2), but, at least for concrete nouns like pinti, they are usually straightforward enough for the purposes of practical linguistic description. As a result, we can talk with some certainty about cross-linguistic differences of reference for this kind of noun. But what this discussion has shown is that questions of cross-linguistic differences of sense are more complicated, and conclusions based on them accordingly harder to reach.
An associated problem is the question of the correct or optimal meta language for the description of meanings. Claims about the universality of given meanings necessitate a particular metalanguage in which the meanings can be described. Similarly, studying meaning change implies that we have a reliable metalanguage which can be used to represent historical sense developments accurately. But this immediately introduces complications since there is not yet any agreement about what the correct metalanguage for semantic description is. Semantic theories like those of Jackendoff, Wierzbicka and many others presuppose a universal set of primitive concepts lexicalized in all languages. In claiming that the meanings of all languages can be translated into a unique, universal metalanguage, these types of theory constitute strong hypotheses of semantic universalism. As illustrated elsewhere, however, these hypotheses are also highly controversial (2.5; 8.1.3). There is no agreement that a single universal metalanguage for semantic description is even possible, let alone agreement on what it should be like. The absence of an agreed standard for description complicates the process of achieving consensus in com parative or historical studies of meaning. Two investigators can always disagree about the details of a word’s meaning. But the prospects for agreement are obviously improved if they are at least working with the same descriptive metalanguage: if they are not, it may not be even clear whether they agree or not. We will see some examples of this sort of problem in 11.4.2 and 11.4.4 below.
The idea that the variation among languages conceals an identity of meaning at some deeper level has a lot to be said for it. Humans share the same perceptual and cognitive organs, and we regularly succeed in making ourselves understood, even across breathtaking cultural and linguistic divides. We also inhabit the same shared world to which we refer. Surely, one might ask, this means that the meanings we express are also the same, deep down? If they weren’t, we couldn’t accurately translate from one language to another, and we would have no guarantee that understanding was possible across linguistic divides (see Chapter 1).
Linguistics in general, and semantic theory in particular, certainly assume that languages are mutually translatable in a way that preserves important meaning components. If we abandoned this assumption, any cross-linguistic work involving meaning would be impossible. But it is one thing to presuppose a rough and ready translatability, and quite another to suppose that exactly the same meaning or concept can be captured by the words of different languages. Just as it seems obvious that we can convey the essence or gist of our thoughts in another language, so it is a commonplace that no two languages ever convey exactly the same ideas. The cross-linguistic and historical study of meaning has to weave a course between these two equally obvious positions.
