Modularity and Constraints in Language and Cognition

The Minnesota Symposia on Child Psychology, Volume 25

Megan R. Gunnar, Michael Maratsos

• Publisher: Psychology Press
• Year: 2014
• Topic: Psychology
• eBook ISBN:
  9781317782209
• Subtopic: Developmental Psychology
• Index: Psychology

One of the central problems in the study of modern cognition is the degree to which higher cognition is modularized: that is, how much are higher functions carried out by domain-specific, specialized, cognitive subsystems, rather than a highly general cognitive learning and inferring device? To date, ideas and proposals about modularity have been best developed in the study of vision and grammar. In the present volume, the usefulness of approaches employing modularity and domain specificity are further explored in papers on the development of biological thought, word meaning, symbols, and emotional development, as well as in the core area of grammar itself, by leading researchers in these fields. The volume also contains an introduction to some basic ideas and concepts in the study of modularity and domain-specificity, and some critical discussion of the overall problems of the modularity constraints approach to analyzing development.

1	Constraints, Modules, and Domain Specificity: An Introduction

Michael Maratsos

University of Minnesota

The general topic of the 1990 Minnesota Symposium was modules, constraints, and domain specificity. This chapter is an introduction to some general ideas and problems in this area.

A GENERAL BACKDROP

At its broadest, the problem of modern developmental psychology does not just reflect differences in investigator’s beliefs about particular hypotheses regarding psychology and the development of children. It also reflects a clash between two major approaches or attitudes to scientific investigation.

A first goal is one of the oldest goals of science: the uniting of a wide range of apparently diverse phenomena by the positing of a relatively small set of general underlying principles. Indeed, achievement of this goal has been one of the time-honored closest-to-the-heart characteristics of science since ancient times. With the inception of Newtonian physics, it achieved such successes as to become a paradigm goal. To explain why the phenomena that the planets do not fall into the sun really comes from the same reasons that objects do fall onto the earth is the kind of triumph toward which all sciences have since aimed.

This goal is reflected by the attempts of behaviorists to find a few elementary principles and processes by which all behavior could be explained. In more modern times, it is reflected in a search to explain the mind by highly general principles that can be posited to underlie all the diverse areas of cognition. In developmental psychology, Piaget’s theory provides a general model of this kind. More recently, most work of the information-processing school (e.g., Anderson, 1976) seeks a set of explanatory principles that apply equally well to all domains of thought, including language, problem solving, arithmetic, and any other subspecies of the cognitive processes.

On the other hand, in recent years, another kind of idea about cognition has emerged. Perhaps a great deal of cognition consists of special-purpose systems each designed for particular tasks, such as learning language, or dealing with quantity, systems that reflect the diverse sets of tools that given species evolve in order to find, or create, a particular niche. This kind of thought stresses its continuity with evolutionary biology, in which the idea of species-specific, particular systemic properties of organisms is a natural one. As Mayr (1980) said, much of current evolutionary thought stresses the diversity present within species; the diversity provides the food of evolution.

According to Mayr, the two ways of thought about organisms—underlying generality versus underlying diversity of organization—have indeed clashed before in the history of biological thought. He said that attempts to gain an analysis like that of physics was one of the hindrances to a development of evolutionary theory. For development of evolutionary theory requires taking as basis the very great diversity of behaviors and systems within species; a concentration upon achieving description in terms of a few relatively simple, abstract underlying elements and principles thus impeded at different times the development of evolutionary thought.

In some ways, of course, biology itself has come full circle. At the molecular level, it now seems possible to understand all of the different genetic messages as being arrangements in triplets of four relatively simple amino acids. Yet this analysis does not remove the usefulness of dealing with the behavioral diversity of organisms at the level of problems such as accounting for speciation, or niche-filling.

Generally speaking, it seems to me that psychologists have tended to look fairly exclusively to physics as a model of the goals and methods of scientific investigation. What the history of science seems to show is that no particular mode of analysis has always been the correct one for all problems. Indeed, as a digressive remark, it seems likely that for many problems that presently occupy psychology, the general model of scientific thought provided by biological evolutionary theories before the development of molecular genetics may well provide a more profitable mode.

To return to the original problem: It is unlikely that investigators will drop their search for general underlying principles to explain apparently diverse phenomena; nor should they. But the possibility that much of human cognitive and social organization is diverse at important underlying levels as well also is a serious competitor as a model of psychological organization. The differences between these two approaches and beliefs are real; they are not separated simply by misunderstandings of concepts and terminology, but by real differences in goals and ideal aims.

In this volume, the concentration is on current work pursuing possibilities and problems in the approach of more domain-specific, species-specific underlying cognitive systems. Work in this kind commonly turns up concepts like “constraint,” “module,” or “domain specific.” Indeed, in language, the prototypical examples of such systems in higher cognitive functioning, these concepts seem to arrive as a package.

As is often the case, a prototype may put things together that analytically may be separated. Thus, attention to the prototype alone may serve to confuse understanding of the separable nature of what are indeed partly individual concepts. Thus, a primary goal of this chapter is to give an idea of what such concepts mean; but an equally important secondary goal is to show how they are potentially partly independent.

DOMAIN SPECIFICITY

Probably the key concept of all these three is domain specificity. This concept refers to a claim that the manner in which information is processed in one domain may be different from the manner in which it is processed in another. This stands in opposition to systems like behaviorist associationism, for example, in which association of stimulus to stimulus, or stimulus to response, occurs in the same fashion for any kind of stimulus or response.

Domain specificity by itself, however, does not simply mean there is something specific about some domain of functioning in the behavior of the organism. An organism might learn to make specific types of responses in one domain, such as arithmetic, and in ordinary language terms, this would correspond to some degree of domain specificity in the organism’s responses.

But in current usage, domain specificity means there is something innate in the organism that causes responses in that domain to be different in some important way. Such systems of response are now easy to document in animals. For example, a general rule of conditioning is that the animal will associate two stimuli with each other only if they are temporally contiguous, that is very close in time. It is accordingly a classic finding of modern psychology that in at least one domain, this rule does not hold. For Garcia (Garcia & Revusky, 1970) has shown that a rat will associate a novel taste with induced sickness (induced by the experimenter) if the novel taste precedes the sickness by 30 minutes; if the taste occurs right before the rat sickens, the rat does not make any association. Only in the first circumstance of a delay of some time does the rat avoid the novel taste later on.

Evolutionarily, this makes sense. It takes food some time to make one sick. So a system for learning food avoidance would do far better to associate a novel taste from a while ago with illness than a novel taste that occurs just before. But obviously this special rule of conditioning violates the general rules. Thus, there is something specific to the domain of association of taste with illness in the rat, and this domain specificity is innate.

Garcia’s studies of illness and taste association in the rat thus provide prime examples of what is meant by a domain-specific learning system. In the study of higher cognition, it has been customary to think that the laws of learning and theory construction by the child are essentially the same in all domains of thought, such as language or logic (e.g., Piaget, 1970). Domain specificity refers to the hypothesis that there is instead something different among different domains in how they deal with information; they may have restrictions on what kinds of information are dealt with, or in what conclusions can be drawn from the information.

What counts as a domain? Essentially, any discernible body of functions that can be discriminated from another in some aspect of its content is a potential domain. Nor should it be expected that domains that are convenient for us to think of as separate will turn out to be different; or conversely, domains we think of as unified might turn out to be made up of separable domains upon further study. For example, arithmetic and logic are seen in the work of many 20th-century logicians and mathematicians as being essentially part of the same domain; but it might be that psychologically, aspects of the mind have developed to treat materials in the two fields in characteristically different ways. So a “domain” means any organized content area in which one might imagine or propose there being innate, specific distinctive mechanisms or ideas.

CONSTRAINTS AND DOMAIN SPECIFICITY

Constraint is a term that has an ordinary language meaning, but is developing a more specialized meaning in its current uses. Again it is obvious that in its widest sense, constraint would apply very broadly to what organisms do, and how they think. Organisms are constrained in their behaviors by systems of social rewards, punishments, and motives. Humans are constrained in how they can think by the presence or absence of education in some subject.

But constraint in current discourse in modern psychology is intended to mean something more specific than this. It is intended to mean that because of innate structural specification, the organism does not process information in some fashion that one might logically imagine it could. Thus, some limitation in one’s behavior that arises from education or training (or the lack of it) would not comprise a constraint.

Even adding innateness, however, does not capture the modern use of constraint. In broadest terms, no account of behavior or learning system has ever proposed that organisms are totally unconstrained in how they can process or construe incoming data. Any proposed processing system has proposed to specify what the organism does; it must thereby, by implication, specify things the organism does not do.

Behaviorist views of human behavior thus were, and were meant to be, highly constrained: For example, it was claimed that much of human functioning may and should be captured simply by analyzing the associations made between stimuli and responses under various conditions; all such stimuli and responses and conditions were supposed to be clearly physically observable; stimuli and their associated responses were stipulated as having to be temporally contiguous. These stipulations are very strong constraints indeed, and the ferocity of the constraints was customarily portrayed as a distinct advantage of the theory.

Important aspects of other, more current theories of human cognition also clearly comprise some kind of constraint. For example, modern information-processing theories generally claim that there is a short-term memory and working space distinct from long-term memory. This short-term working space is typically said to be able to hold only a limited number of items (Miller, 1956), or to have only a limited amount of attentional focus available to it.

But in current usage, it is very doubtful that either behavioristic constraints upon the association of stimulus and response, or information-processing constraints on short-term memory, are the kind of thing that is meant when someone uses the word “constraint” in discussing cognition. But, if “constraint” does not mean any possible innate constraint upon the organism’s processing of information, then, what does it mean? In most current uses, “constraint” refers to something the organism does not do that it might because of something innately specific to a particular domain. Thus, in Garcia’s experiments with rats and the conditioning of novel tastes and sickness, one might expect the rat to associate a novel taste and an immediately following illness, given that the rat typically (or at least often) associates temporally contiguous stimuli. But the rat does not. Therefore, something must constrain the rat from doing so; this is the specific mechanism that, in the domain of taste and illness, associates novel tastes with illness that follows some time later.

In current work in cognition, “constraint” refers to some constraint placed by a domain-specific structure that affects the organism’s processing of cognitive information. The notion of a constraint in modern cognitive work developed with certain central connotations derived to a large degree from some of its initial uses in the study of linguistics. I thus attempt to develop here at some length some of the logic of one of these initial usages. I then present more briefly some of the further, much wider uses “constraint” came to have in linguistic work.

The central initial example of a constraint in linguistics emerged in the analysis of a rule often called wh-movement, one of the transformational rules of Chomskyan grammar (Chomsky, 1957, 1965). This rule is the central operation in the transformational machinery for producing wh-questions like “what will he eat?” or “when can he come?” In the transformational description of such sentences, the wh-term appears initially in a declarative-like position in the sentence. Thus, the initial form of “what will he eat?” is “he will eat what?” in which “what” appears as the direct object of the verb “eat.” Indeed, in this sentence, “what” is a constituent that asks for information about that which was eaten, and so this initial form put it in a sentence position that indicates this direct object function more clearly. The wh-movement then moves “what” to the front of the sentence:

(A second rule is responsible for the movement of the auxiliary verb will from after the grammatical subject to in front of it; this rule is not important for discussion here.)

So in the transformational description, wh-terms frequently begin in one position in the sentence, then move to the front. In the following examples, I put “3/M” in a fully formed wh-question, in order to indicate the position from which the wh-term was moved by the transformation. Thus “what will he eat 3/M” indicates that “what” was moved from the position after “eat.”

A central property of English and many other languages is that wh-terms can be moved from a very great variety of positions to the front. Basically, any noun phrase constituent can be questioned by replacing it with a wh-term or phrase and moving the wh-constituent to the front. Indeed, in theory, wh-terms can be moved from an infinite variety of contexts. For one of the central properties of language is that sentences may be embedded inside other sentences. For example, the sentence “John thinks that Mary likes Tom” has a full sentence “Mary likes Tom” embedded as the grammatical direct object of the verb “think.” (In this sentence, the word “that” is an introducer for the embedded sentence.) The structural rules of natural languages allow such embedding to go on forever, or in practical terms, for as long as a speaker’s memory allows. One can say “John thinks (Mary convinced Sam [that Hillary likes Tom]),” which has two sentence embeddings; or one can say “John thinks (Mary convinced {Sam that Hillary said [that Margaret likes Tom]}), ...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Preface
• 1. Constraints, Modules, and Domain Specificity: An Introduction
• 2. Modularity and Constraints in Early Lexical Acquisition: Evidence From Children’s Early Language and Gesture
• 3. Constraints on Word Learning: Speculations About Their Nature, Origins, and Domain Specificity
• 4. The Origins of an Autonomous Biology
• 5. Language, Affect, and Social Order
• 6. The Logical and Extrinsic Sources of Modularity
• 7. Beyond Modules
• 8. What Do Developmental Psychologists Really Want?
• Author Index
• Subject Index