## Thursday, 28 March 2013

### "Only confusion comes of mixing these two topics"

David Lewis, in his "General Semantics" (1970) wrote:
My proposals will also not conform to the expectations of those who, in analyzing meaning, turn immediately to the psychology and sociology of language users: to intentions, sense-experience, and mental ideas, or to social rules, conventions, and regularities. I distinguish two topics: first, the description of possible languages or grammars as abstract semantic systems whereby symbols are associated with aspects of the world; and second, the description of the psychological and sociological facts whereby a particular one of these abstract semantic systems is the one used by a person or population. Only confusion comes of mixing these two topics.
The idea is that there is a distinction between:
• the description of possible languages or grammars as abstract semantic systems whereby symbols are associated with aspects of the world.
• the description of the psychological and sociological facts whereby a particular one of these abstract semantic systems is the one used by a person or population.
For example, in semantic theory, when one describes an interpreted language $L$, one studies a semantic function $\| . \|_{L}$ from $L$-strings to semantic values (meanings, things, extensions, etc.). This is not the same topic as the topic of studying how an agent "grasps", or "implements", that meaning function; or how an agent "uses" or "speaks" or "cognizes" the language $L$.

One might believe that these are the same topic. But they aren't. Or one might believe that, though separate, one can relate them. Or one might believe that the second is, in some sense, "primary", and the first is "derivative". In each case, the view involved is one that needs to established, not merely asserted, in order to avoid Lewis's accusation that one is simply confusing these two topics.

1. This is spot on for all invented languages, i.e. planned and formal languages (Lewis was surely more familiar with the latter than the former). These are, after all, made to serve as tools one can use at leisure for particular purposes, and stop to use should they turn out not, or no longer, to fulfil that purpose. Being invented, they possess certain unchanging properties and functionalities, and there may well be an interest to study them (if only in order to determine what the language in question can and cannot be used for).

The situation is radically different for natural languages. These don’t exist as established systems at all, much rather describing them requires (i) delimiting them geographically (e.g. is Swiss German still a part of what we want to call German or has it evolved into a separate language?); (ii) delimiting them in terms what part of human behaviour counts as language and what doesn’t (Is a groan part of language? Is “ouch”? Is “ts-ts-ts”? Are gestures, nods? If not – why is sign language a language?); (iii) observing the speakers thus selected and abstracting rules of grammar and the meanings of the words/signs they say/make from what is regarded as relevant behaviour; (iv) establishing criteria according to which some such behaviour enters the description, while some other doesn’t (because it is considered “non-standard” or incorrect), etc.

The result is necessarily an artificial abstraction, and moreover – since the behaviour even of such a select group of speakers constantly changes – a snapshot picture of an ongoing joint activity. So it seems to me that for natural languages, the onus is clearly on the Lewisians to justify abstracting some aspects of the interaction between some people into a system.

That said, there are many good reasons for doing so, in particular for much of the highly specialised work done in linguistics. But it seems quite absurd to waste a single thought about the relation between the speaker and the language she speaks – you don’t think about the relation between a walker and the way she walks either, do you? It is equally absurd to first abstract part of some people’s behaviour into a Language, study that artificial abstract’s ‘semantics’ and then wonder how the speakers can ‘use’ the language in order to refer to things; or to decide that some part of what people do is to be considered semantics and another pragmatics, and that the two are not to be confounded. I strongly suspect that people who do this overlook that the name ‘language’ applied to formal structures is metaphorical and it makes as little sense to make claims about natural languages based on what we find in formal ‘languages’ as it would make sense to make claims about trees based on what we observe in logical trees or Porphyrian trees. Only confusion comes of mixing these two topics.

2. Naomi,

Thanks for all that!

"So it seems to me that for natural languages, the onus is clearly on the Lewisians to justify abstracting some aspects of the interaction between some people into a system."

I think this requires accepting your assumption about "abstraction", that a language is somehow obtained from the speech behaviour by "abstraction".

But there need not be any mention of "people" in describing a language. In describing a language, we have some alphabet, some syntax and perhaps some semantic and pragmatic functions. There is no "abstraction" involved.

In a way, this view about abstraction seems a bit like MIll's theory of natural numbers: they are "abstracted" from physical aggregates. But Frege explained the problems with this. Natural numbers are not properties of, or "abstracted from", physical aggregates. Rather, they are cardinalities of finite *sets* (well, for Frege, concepts; but here it makes no difference); this explains their universal applicability, and fits with the governing principle is

$c(X) = c(Y) \leftrightarrow X \sim Y$

Here, $X, Y$ must be sets or concepts (not physical aggregates).

Jeff

1. Thank you for these replies.

You say "In describing a language, we have some alphabet, some syntax and perhaps some semantic and pragmatic functions." - but what I am trying to point out is precisely that in *natural* languages (even if conceived as idiolects!) we do *not* HAVE any of that. I do not want to accept the comparison to structures or shapes in physics because, as I won't tire to insist, natural languages are essentially dynamic. So if you compare them to anything in physics, please pick a process, not an object. How about coffee being poured into a cup and spilling over?

I am interested to see that you refer to Basic Law V, of all things. But, if memory serves, C. Wright defends it by alluding to... abstraction! ;-)

2. Hi Naomi,

"I am trying to point out is precisely that in *natural* languages (even if conceived as idiolects!) we do *not* HAVE any of that."

In linguistics, one has an alphabet of phonemes which form strings; these strings have syntactic structure and semantic properties. The branches of linguistics are called: (i) phonology; (ii) syntax; (iii) morphology; (iii) semantics; (iv) pragmatics; and so on.

So, we have all of this and a lot more!

Jeff

3. Naomi,

"But it seems quite absurd to waste a single thought about the relation between the speaker and the language she speaks – you don’t think about the relation between a walker and the way she walks either, do you? It is equally absurd to first abstract part of some people’s behaviour into a Language ..."

We do, in physics, talk about the structure that a physical system exhibits; or the shape that a physical object has; or the cardinal number that a set of physical objects has.

But your view now seems to be that a claim like,

$A$ speaks $L$

is to be regarded as a fictional claim; or perhaps entirely reducible to various claims about $A$'s speech behaviour.

But how does this work?

Jeff

4. Lewis seems to be taking it as obvious that for every natural language there is an abstract language which is implemented by it. This needs justifying!

In fact, vagueness in natural language poses a serious threat to the possibility of doing so. It seems as though vagueness, which is an essential asect of natural language, has no analogue in formal language.

At the very least, it is certainly not the case that a natural language assigns to each sentence a unique, determinate proposition. We know context is important.

5. Hi,

In semantic theory, one is studying interpreted language, not "formal" languages.

There are various analyses of vagueness in semantics. One approach to have a single language with supervaluational semantics; another is to use fuzzy semantics; another is to say that it's indeterminate what language the speaker speak/cognizes; another is to insist that the issue is epistemic, rather than semantic.

It's a standard view in the literature on semantics for propositional content to be context-dependent (indexical-dependent, etc.).

Jeff