404 Animal Memory and Cognition
phenomena. Second, the theory is general and can be used, for
example, to explain the acquisition of concepts or categories,
as will be explained shortly.
Spence’s theory, as indicated, suggests that a variety of
stimuli simultaneously become excitatory when rewarded
and inhibitory when nonrewarded. Lashley (1929) opposed
Spence on this score, suggesting instead the more cognitive
view that animals selectively attend to stimuli. Discrimina-
tion learning, according to Lashley, consists of successively
eliminating, one by one, stimuli that fail to predict success-
fully, the animal ultimately fastening on the relevant stimulus
dimension. Lashley’s attentional-cognitive view was not sup-
ported by experiments showing that reversal of the positive
and negative cues (i.e., B+W–to W+B–) while animals
were still responding at a chance level (50% correct) on the
original problem (B+W–) produced serious retardation in
learning the reversal (W+B–). Such retardation should not
occur, according to Lashley, because animals responding at a
chance level have not (by definition) isolated the relevant
stimulus dimension; thus, reversing the S+and S–cues
should not influence the final solution, which is contrary to
fact. Attentional theories that assume animals can attend to
two or more stimuli simultaneously are better able to deal
with the reversal findings described previously (see, e.g.,
Sutherland & Mackintosh, 1971). Spence’s theory predicts,
of course, that reversing the S+and S–cues when the ani-
mal is responding at a 50% level will have a deleterious effect
on discriminative responding. This is because animals re-
sponding at a 50% (or chance) level have nevertheless
learned something about the S+and S–cues, enough to re-
tard reversal learning.
Gestalt psychology, which emphasized perception, ex-
plained discrimination learning in terms of learning rela-
tionships between stimuli. In a B+W–discrimination, for
example, the animal would not learn that B is excitatory and
W is inhibitory, as Spence suggested, but rather would learn
to select the darker of the two stimuli. Offered as support for
the Gestalt view was the phenomenon of transposition. For
example, an animal that learned to select medium gray (posi-
tive) over light gray (negative) might, in a subsequent test
phase, when given a choice between the medium gray and a
newly introduced dark gray, actually select the novel dark
gray because it is the darker of the two stimuli.
Spence’s arguments with Lashley and the Gestalt psychol-
ogists illustrate an important point suggested earlier: Discrim-
ination learning has been and continues to be an important
battleground for testing the adequacy of associative versus
various nonassociative approaches to animal cognition and
learning. Spence’s theory is able to explain transposition
in associative terms without appealing to the learning of
relationships. This is shown graphically in Figure 14.2. The
figure shows inhibition and its generalization associated with
the negative (S–) cue (dotted line) and excitation and its gen-
eralization associated with the positive (S+) cue (solid line).
Net excitation is shown by the length of the solid vertical
lines above various stimulus points. Note that greater net
excitation is associated with the S+cue rather than with the
S–cue, and so the animal will select the S+cue. However,
greater net excitation is associated with the cue to the right
ofthe S+cue, and so the animal will select that, novel
untrained stimulus in preference to the S+cue—the transpo-
sition phenomenon. In sum, Spence’s theory is able to ex-
plain transposition by employing rather orthodox associative
concepts.
More recently, individuals concerned with evaluating the
role of cognition have employed categorization experiments
that are, essentially, elaborate discrimination learning inves-
tigations. In these, pigeons might be shown numerous photo-
graphic slides containing (say) trees and numerous other
slides lacking trees (e.g., Herrnstein, 1979). The pigeon
might be rewarded for pecking the “tree” slides (S+cue) but
not reinforced for pecking the non-tree slides (S–cue). The
pigeons readily learn this sort of discrimination, which they
transfer well to novel stimuli. The meaning of results of this
sort remains unclear. For example, one might think it is eas-
ier to learn a category (trees vs. non-trees) than to learn 80
unrelated slides (40+and 40–). Vaughan and Greene
(1984), however, employing 160+slides and 160–slides,
uncategorized, showed that pigeons more or less easily came
to master the discrimination and even performed well after a
2-year rest. Although pigeons perform better with categorical
than with noncategorical grouping of stimuli, this may not
indicate that they learned a concept (see Watanabe, Lea, &
Dittrich, 1993). It is the case that category slides will have
more in common with each other than will noncategory
slides. Thus more excitatory stimulus generalization willFigure 14.2From Spence (1937). Excitation at S+(solid line) and inhibi-
tion at S–(dotted line) and their generalization to other stimuli. As ex-
plained in the text, more net excitation (excitation minus inhibition) exists at
S+than at S–, thus producing discriminative responding, and more excita-
tion exists at a stimulus to the right of S+(stimulus 409) than at S–itself,
producing transposition.10492.506.686.483.16 4.84
1.52
0.00256 409 655
()160
()39 62 100STIMULUS SIZE