Mukherji on Language Reference Vision

15 December 2009

Nirmalangshu Mukherji (Delhi University)

Language Reference Vision

Abstract:

Assume the following as a framework:

Grammaticality: By ‘language’ we (very narrowly) mean the grammatical system of human languages that generate <PF, LF>, where ‘PF’ is the phonetic form and ‘LF’ is the linguistic logical form (not to be confused with philosophical logical form a la Russell).
Internalism: Scientific/biological accounts of organisms do not mention factors external to the organism.
No-Semantics: Classical study of semantics as the study of language-world relations does not have explanatory value since it assumes what needs to be explained.
Grammar and Use: Beyond LF, issues of meaning, content, external significance, intentionality, etc. reduce to issues of how language is put to use by the organism in an environment (without violating (2)).
How do we explain the phenomenon of reference—the ability of humans to talk about the world—within this framework?

It follows from (1) that reference can be studied only as a post-LF phenomenon. Hence, the grammatical system must be interfacing with other systems to give rise to the phenomenon.

By (2), it follows that we can only describe the (sequence) of post-grammatical mental/internal systems in terms of a series of representations. If the resultant representation ‘matches’ the world, the reference is achieved, otherwise it fails. Throughout, we are trying to figure out what the organism ‘knows’.

By (3), we cannot postulate a reference/denotation relation to formulate post-grammatical representations; these representations need to be formulated so as to ‘take’ the representation to the world, as it were.

By (4), the familiar tool-metaphor emergres. Language/grammar is viewed as a tool that is put to a variety of uses. One of these is the achievement of reference.

Restricting the domain

Even with the preceding restrictions, the issue is too vast for proper controls. Some further restrictions may be obtained by narrowing the domain of inquiry on some principled basis.

In some sense, the general notion of reference could be viewed as a normative one: we find it useful to adopt the norm that what we say (typically) does pick out aspects of the world—the use of language is typically world-bound. The norm is most explicit in scientific discourse, although it is implicit in varying degrees in ‘common’ discourse as well (more explicit in legal discourse, less in fiction). The scientist typically engages in elaborate—often formal—discourse because she is interested in identifying real joints of nature. Notice it is just a norm/hope that, after rigorous empirical investigation, ‘electron’ picks out something in the world, ‘phlogiston’ unfortunately didn’t.

It seems that, insofar as reference is broadly viewed as a norm of discourse, the language promoting that discourse need not have the entire structure of human languages. Thus, Quine proposed a regimented (first-order, formal) language to show that all scientific talk can be conducted in the language. Most notably, this regimented language does not contain singular terms such as proper names, definite descriptions in the singular, demonstrative phrases in the singular; these are eliminable. Its only ‘referential device’ is the bound variable which ‘refers’ with systematic ambiguity.

Perhaps this is all that is required for a ‘language of science’. However, it is clear by now that these singular terms are not eliminable from natural languages: that is, natural languages are not regimented languages. So, if the notion of a regimented language suffices to capture the norm of reference, what are these singular terms specifically doing in natural languages? By asking this question, we align our concern with some classical philosophical concerns on names, descriptions and indexicality. We focus exclusively on these.

Tools and Instructions

Supposing these singular terms to be specific tools for (achieving) reference, we note that tools have at least a design and constrained uses based on that design (Wittgenstein). In order to put a tool to use, there are either overt or covert instructions about how to use them. The following questions arise.

(A)   What are the design features of these tools such that each of them can be discriminated on that basis to show that these are different tools?
(B)    Which part of the design is grammatical and whch part(s) is/are endowed by other systems of the mind?
(C)    What are the specific instructions encoded with respect to each of the tools in conformity with its design?
(D)   What are the instructions to? Just as PF is viewed as an instruction to the acoustic systems, can we view LF at least in part as instructions to the visual system (among other things)?
(E)    What is the evidence/literature for discriminating between these items in terms of instructions to the visual system (if at all)?
Some of these questions will be discussed.

Valeria on Pinker’s Theory of Graph

24 November – 1 December

Valeria Giardino (Institut Jean Nicod, CNRS-EHESS-ENS)

Discussing A Theory of Graph Comprehension by Pinker

Reference: Steven Pinker (1990), ‘A Theory of Graph Comprehension’ in R. Feedle (Ed.), Artificial Intelligence and the future of testing (pp.73-126)

From Pinker’s article:

A striking fact about human cognition is that we like to process quantitative information in graphic form. One only has to look at the number of ways in which information is depicted in pictorial form–line, bar, and pie graphs, Venn diagrams, flow charts, tree structures, node networks, to name just a few–or to the great lengths that computer companies go to advertise the graphic capabilities of their products, to see that charts and graphs have enormous appeal to people. All of this is true despite the fact that in virtually every case, the same information can be communicated by nonpictorial means: tables of numbers, lists of propositions cross-referenced by global variables, labeled bracketings, and so on. Perhaps pictorial displays are simply pleasing to the eye, but both introspection and experimental evidence (Carter, 1947; Culbertson & Powers, 1959; Schutz, 1961a, 1961b; Washburne, 1927) suggest that, in fact, graphic formats present information in a way that is easier for people to perceive and reason about. However, it is hard to think of a theory or principle in contemporary cognitive science that explains why this should be so; why, for example, people should differ so strikingly from computers in regard to the optimal input format for quantitative information.

The goal of this chapter is to address this unexplained phenomenon in a systematic way. In particular, I propose a theory of what a person knows when he or she knows how to read a graph, and which cognitive operations a person executes in the actual process of reading the graph. This theory will be used to generate predictions about what makes a person better or worse at reading graphs, and what makes a graph better or worse at conveying a given type of information to a reader. In pursuing these goals, one must recognize a very pervasive constraint. Comprehending a graph (unlike, say, seeing in depth, uttering a sentence, or reaching for a target) is not something that anyone could argue is accomplished by a special-purpose mental faculty. Graphs are a recent invention and if they are an especially effective method of communication, it must be because they exploit general cognitive and perceptual mechanisms effectively. Any theory that hopes to explain the process of graph comprehension will have to identify the psychological mechanisms used in interpreting a graph, and which operating principles of each mechanism contribute to the overall ease or difficulty of the graph-reading process. Thus, any theory of graph comprehension will draw heavily on general cognitive and perceptual theory, and where our knowledge of cognitive and perceptual mechanisms is sketchy, we can expect corresponding gaps in our ability to explain the understanding of graphs.