« PreviousContinue »
4. POINT ONE: PENETRATION INTO EXPERIMENTAL PSYCHOLOGY
The first point is that the penetration of information processing theories into experimental psychology is very substantial. To see this, one must take the broader view I have just emphasized. Information processing, not artificial intelligence, is the critical issue, simply because most tasks investigated in
psychology are not problem solving or complex learning.
Furthermore, the total range of work that now operates within an information processing framework by no means derives from a single source. More precisely, the wider domain, which we labeled cybernetic systems in Figure 2, has been the common source of all the work (especially if we understand it to include developments in operational mathematics, such as decision theory and game theory). But this broad development has permitted many parallel developments in psychology, all converging on the class of information processing systems. Let me briefly identify these main lines of development.
Perhaps the most important one in terms of number of investigators is that concerned with the study of immediate memory. In terms familiar to this audience, the basic problem is to discover the logical design of the short term memory. Actually, there appear to be several such memories, some of the order of hundreds
of milliseconds half life, at least one of the order of several seconds. Since no
anatomical or physiological data exist on these memories, their existence and characteristics must be inferred entirely from behavior. Thus, there is even
controversy over what memories exist (Melton, 1962).
Now the concern with the logical design of a system does not necessarily imply concern with a symbolic system. And, indeed, the genesis of this work goes back to communications engineering and information theory. The book by Broadbent on Perception and Communication (1958), which was one of the milestones in this area, shows this very well: signal processing, but not symbol processing.
What changed this was the discovery that the human immediate memory appears to hold symbols chunks, to use the term introduced by George Miller in his wellknown paper on the magic number seven (Miller, 1956). This established that one
should consider the human as an information processing system with a short term
memory of constant capacity, measured in number of symbols. By now, this view permeates all work, as can be seen in the numerous models of short term memory that are now available (many of them summarized in Norman (1969)).
A second development is in psycholinguistics, where the work of Chomsky has had a very large impact. First, observe that Chomskian linguistics implies a symbolic system. One can emphasize, as have the linguistics, that performance
should be distinguished from competence, so that a model of the linguistic ability (1..e., the set of syntactical rules) does not imply that language is in fact
processed in a person by a machine that takes the rule system as input. However, if one wants to draw any inspiration from linguistics for psychology, then it will still be a system of this kind -- 1.e., some kind of a system that deals with discrete symbols with rules and transformations on those symbols.
This is exactly what has happened in psycholinguistics, where many studies are being performed, taking seriously the notions of linguistic transformation and the encoding of meaning (semantics) in the so-called deep structure (Chomsky, 1967). The attempt to characterize the development of children's grammars, which thereby attributes to them a (simple) system of rule following behavior on symbol structures (language utterances), is part of the same picture (Smith and Miller, 1966).
Problem solving. A third development is the simulation of cognitive processes in problem solving by means of computer programs. This is the development associated with (intimately entwined with, would be a better phrase) artificial intelligence. The problem solver is viewed as a symbolic system, capable of following strategies
search, applying heuristics, calculating results, both symbolic and (on occasion) numeric, and evaluating partial results. The efforts referred to here are those one would also consider psychology (in line with the choices with respect to Figure 1), namely, those where direct comparison is made between the symbolic system and data from human behavior. Good representatives of this work can be found in the well-known collection by Feigenbaum and Feldman (1963) (see also
Concept formation. A fourth area of development is in the study of concept formation. Work in this area, of course, goes back many years (e.g., to Hull, 1920). A major turning point is symbolized by the book by Bruner, Goodnow and Austin (1956), which made extensive use of the notions of strategy and hypothesis formation, as well as cognitive strain (being essentially the dynamic memory load needed to carry out various strategies). The system implied there was very much a symbolic system, though its inspiration came out of decision and game theory, rather than computer science.
However, though there has been substantial work in artificial intelligence on concept formation (inspired in large part by the Bruner, Goodnow and Austin analysis) and even on information processing models for its psychology (e.g., Hunt, 1962; Hunt, Marin and Stone, 1966), most of the upsurge of work that followed in the late fifties and early sixties could not reasonably be seen as working within an information processing framework. It would be better characterized as a straightforward experimental investigation of psychological phenomena, in which
various limited questions were posed and investigated without any deep commitments
to the type of processing system implied. For example, studies were done to show that there was a systematic effect of the number of relevant versus irrelevant dimensions in the stimulus configurations; and to show the effect of the availa
bility of past information (Bourne, 1966).
However, gradually more explicit assumptions have been made about the nature
of the subject's processing first in terms of hypothesis testing (Restle, 1962), more recently in terms of general rule-following behavior (Haggard and Bourne, 1965). These shifts imply a symbolic processing system.
Summary. My purpose in quickly going over these lines of development is not
to establish them in any detail -- for this I have hardly done. It is to call your attention to the use of symbolic models in many places throughout experimental psychology. It suggests (and I maintain) that a shift in the Zeitgeist in psychology has taken place toward a view of man as an information processor.
In fact, I have left out several additional lines of development, for example
the work in verbal learning. Although the non-psychologist can be pardoned for
thinking that this is coextensive with psycholinguistics, in fact it is a separate experimental tradition going back to Ebbinghaus and his use of nonsense syllable learning (1885). Work on the learning of verbal materials serial lists, paired associates, and free recall -- have been one of the bastions of S-R psychology,
since the phenomena lend themselves well to explanation in terms of the formation
Let me quote a paragraph from a recent study by a psychologist who has long worked in this area. The study is entitled, "Image as a mediator in one-trial paired-associated learning." It seeks to investigate the use of mneumonic devices in memorization. It has long been known that if you want to memorize a list of, say, ten items, then a good way, to proceed is by having an already learned list of associations, say, 1-bun, 2-shoe, 3-tree, 4-door, ... 10-hen, and then (to memorize the new material) forming a bizarre visual scene involving each of the items and the word in the permanent list. That is, if the first item to be memorized was whale, then visualize the whale with a bun in its mouth; if the second was a bicycle, then visualize the bike riding down the toe of the shoe, and so on. It will then be found (so goes the lore) that the k item can be reliably recalled by going from the number, say 4, to its word, say door, and then to the visual scene, from which the object can be recalled. (The "l-bun,..." list is memorized once and can be used for a lifetime.)
The present study is a preliminary effort to make some
visual image with no immediate intent to assert the reality
of such a phenomenon. In the present study s's were
this report the term "image" will be used to refer to the
processes S's said they followed when instructed to "picture"
the 10 articles mentioned in connection with the previously
learned list of 10 words that rhyme with the first 10 numbers.
(B.R. Bugelski, E. Kidd and J. Segmen, 1968, p. 70.) This quotation accurately reflects the present state of verbal learning research. It is stiil much enmeshed in a behavioristic stance, which views with alarm attempts to deal with internal processes (symbolic or otherwise). But they are being driven to such attempts, in large part because of a shift in view to an
organism as an active, symbol manipulating system.
In the decades prior to the
current one, such notions as imagemediated paired associate learning simply did not call for investigation. The current attempts testify to the shift in the Zeitgeist.
A final comment: if one looks at where the excitement has been over the last ten years in psychology the places where rapid growth is taking place and which people talk about when asked "what's new" -- a substantial fraction of these turn out to be connected to this shift towards information processing models. The
work on immediate memory is one; the rise of a linguistically oriented psycholinguistics is another; the study of children's grammar (within psycholinguistics) is a third. (Possibly the work on problem solving is yet another, but that is more difficult for me to assess, since I am so involved in it.)
5. POINT TWO: FROM IMMEDIATE MEMORY TO IMMEDIATE PROCESSOR In the discussion of the possible relationships of information processing models to psychology we opted for the use of such models as detailed theories of behavior, rather than, say, metaphors or exercises in the discipline of operationalism. Even taking for granted the extent of the activity discussed above, there is still the question of its nature. Does the work on immediate memory use the notions of information processing only as a metaphor, rather than theory? After all, in a primarily experimental discipline (such as psychology still remains), one can play fast and loose with many verbal formulations and many metaphors, so long as they lead to asking interesting experimental questions.