2 Habituation, sensitization and stimulus learning
‘Habituation is an instance of the living system’s power of establishing a conservative equilibrium to change of external conditions.’
The Nature of Learning (Humphrey, 1933, p. 133)
The phenomena of habituation should make an appropriate beginning for the experimental study of the learning process, since in many ways they are the simplest. However, the main lesson to be learned from the study of habituation and this makes it an even more appropriate subject to start with — is that habituation is almost never as simple as it first seems.
Initial simplicity can be achieved by defining habituation as the reduction of response to a repeated stimulus (Humphrey, 1933; Harris, 1943; Thorpe, 1963; Hinde, 1970). Humphrey’s experimental example was obtained by using a snail, crawling on a board. An electrical attachment meant that the board could be given a standardized jerk at regular intervals, usually of 2 seconds. At the first such jerk, a snail would usually draw in its horns, but as the jerks were repeated, the extent of the withdrawal would gradually decrease, until the snail crawled around without appearing to notice the movements of its wooden platform. A sudden increase in the intensity of the jerks might then bring back the withdrawal response, but with extended experience some animals became indifferent to all mechanical vibration, and were unperturbed by violent banging (Humphrey, 1933). This can serve as a paradigm for research into habituation,
since a great deal of modern experimentation is performed on similar responses in another gastropod mollusc, the giant sea slug.
Habituation as a form of learning
Humphrey supposed that habituation was important as an example of learning — that is, being influenced by information received from the outside world — in the sense that his snail was learning not to respond to a stimulus that was not worth responding to, but there are difficulties in assuming that habituation is always a form of learning. There is often a close relation between the waning of response to repeated stimulation and fatigue — Sherrington (1906) referred to the fading of reflexes in spinal preparations as ‘reflex fatigue’, and one would wish to distinguish physical exhaustion, even of some temporary kind, from the retention of behavioural information. Thus although Thorpe (I 963), Hinde (1970) and Gray (1975) have full discussions of habituation as a form of learning, in Hull (1943) and Mackintosh (1974), it is ignored. As far as the snail is concerned, any learning, even so far as it is distinguished from reflex fatigue, is clearly of a rather limited kind. The progress of learning is assessed by the absence of an in-built response, so there is no question of the organism learning any new motor organization, or dramatically increasing the frequency of certain motor performances, as happens in instrumental learning (see chapter 5; similarly, since the result of experience is that the animal ceases to respond to a stimulus which did elicit responding to start with, there is no increase in the range of stimuli responded to, as happens in the acquisition phase of classical conditioning: see chapter 3).
But suppose that a 5—month-old infant is shown a photo- graph of a particular human face, for 10 seconds, six times, and loses interest in it, but perks up when a different face is presented (Miranda and Fanz, 1974; Cornell, 1974) — surely the decline in response to the first face could be classed as habituation, according to our definition, and surely there is no doubt that external information about the visual pattern of the photograph has in some way been retained? For this
reason alone, it would be unwise to exclude habituation from consideration as a form of learning. In addition, changes in responsiveness to repeated stimuli are often involved in the training procedures of operant and classical conditioning which are taken as the standard laboratory paradigms for learning.
Alternative mechanisms of habituation
Habituation, defined as the phenomenon of reduced response to repeated stimulation, can, according to Thorpe’s review (1963), be found in flatworms and sea- anemones as well as snails and slugs. When comparing these results to those obtained with mammals, the first reaction must be agreement with Hinde’s conclusion that descriptively similar phenomena . ‘may depend on mechanisms differing greatly in complexity’ (Hinde, 1970, p. 579). This is amply confirmed by considering the two extremely different kinds of research and theory which are often conjoined in discussions of habituation (e.g. Tighe and Leaton, 1976; Peeke and Petrinovitch, 1984). The first theory (Groves and. Thompson, 1970; Thompson and Spencer, 1966) is concerned to predict reactions obtained from isolated spinal cords; while the second (Sokolov, 1963, 1975) is based on experiments performed on normal human subjects. It is hardly surprising that these theories differ —but unfortunately there has been a tendency to see the theories as competitors, so that for instance Thon and Pauzie (1984) suggest that Sokolov’s theory is wrong because it fails to explain some results obtained with the habituation of the cardiac response of the blowfly, while Thompson and Glanzman (1976) say that perhaps Sokolov’s hypotheses can be interpreted as a special case of the spinal cord model. One of my themes in this and subsequent chapters will be support for the view, put forward by Thorpe (1963) and Hinde (1970) among others, that behavioural phenomena in different species, which may comply with the same descriptive definition, and possibly serve similar biological purposes, in the various species, may nevertheless be based on very different physiological and psychological mechanisms. As Tighe and Leaton (1976) emphasize in their comparison of habituation
in human infants and simple animal models, this means that although there may be areas of behavioural similarity, if one looks for them it is easy enough to find ways in which one .. species may differ from another — for instance, a human infant will be sensitive to photographs of faces in ways that are neither desirable nor possible for a snail.
Before examining the claims of the two different theories 1’ (Groves and Thompson, 1970; Sokolov, 1963, 1975), consider certain obvious possible results of exposure to a repeated stimulus.
Results of exposure to stimuli
(i) Sensory fatigue
It is conceivable, though not very likely, that the snail’s sense organs for detecting vibration become less sensitive with repeated use during a short time interval. When sniffing for the bouquet of a glass of wine, one should rely on the first few sniffs for this reason, and various tactile and other sense organs may respond mainly to changes in stimulation over time. Usually, response decrements which are attributable to this sort of sensory adaptation are not said to be habituation (Harris, 1943; Hinde, 1970). To show that sensory adaptation is not responsible for a reduced response, a number of tests can be made. For instance, the response decline may be shown to survive a change in sense organ — Dethier (1963) showed that reduced response in blowflies to a certain sugar solution detected by receptors on the left leg was retained when the solution was detected by the right leg.
(ii) Response fatigue
Similarly, fatigue in muscle systems could obviously cause reduced response to a repeated stimulus, without qualifying as a process of learning, and response fatigue is often distinguished from habituation. The experimental test to demonstrate that response fatigue is not responsible for a decline in response involves changing the stimulus — if a stronger (or in some cases just a weaker or different) stimulus
brings back the response, then simple response fatigue cannot have been the cause of the response decline (Humphrey, 1930, 1933). This is called ‘dishabituation’. Often giving a new stimulus will not only bring back response to the new stimulus, but will also mean that responses will start to be given again to the old stimulus. Thus Humphrey’s snails, having habituated to a gentle jerk, would draw in their horns for a large bang, and also go back to drawing in their horns for the next few gentle jerks.
(iii) Habituation in the S-R connection
This is the standard kind of habituation in the ‘synaptic theories of habituation’ discussed by Groves and Thompson (1970). The idea is that a change is taking place not in the sensory input or in the motor output devices, as in (i) and (ii) above, but in the connection between the two. It is thus sometimes said to be more ‘central’ than the changes in the sensory and motor periphery. However, within the dissected out ganglia of the sea-slug, or the dissected out spinal cord of the frog, it has been possible to locate the S-R habituation process in individual synapses. In discussing both habituation and dishabituation of the gill withdrawal reflex of Aplysia, Castellucci and Kandel (1976) reported, ‘These mechanisms share a common locus, the presynaptic terminals of the central neurons projecting on their central target cells. Habituation involves a homosynaptic depression of the terminal due to repeated activity of the sensory neurons’ (p. 31). By recording the electrical activity in single neurons of the spinal cord taken from a frog and maintained in an oxygenated solution, Thompson and Glanzman (1976) discovered that ‘this simplified monosynaptic system in the isolated frog spinal cord exhibits retention or “memory” of habituation, the critical parameter distinguishing habituation as a simple form of behavioural plasticity of learning from neuronal refractory phenomena’ (p. 72). It is clear therefore that synapses between individual sensory and motor neurons can habituate, and it is likely that this mechanism controls a substantial fraction of the habituation seen in the total behavioural repertoire of sea-slugs. It does not therefore follow
that this mechanism is the only one available to explain all reductions in response to repeated stimulation in all other animal species.
(iv) Sensitized states affecting the S-R connection
It is a matter of empirical fact that repeated stimulation sometimes has an effect which is the opposite of habituation — there is some kind of warm-up or sensitizing process so that later stimuli produce a bigger response than earlier ones. In Groves and Thompson’s theory, it was proposed that sensitizing synapses were anatomically distinct from the S-R pathway, thus comprising an external state, whose increasing arousal with repeated stimulation would facilitate the S-R connection, and thus have behavioural effects in the opposite direction to the habituating synapses mentioned above (1970, pp. 433—4).
(v) Familiarity via formation of a memory of the stimulus
This brings the problems of perception and memory into theories of habituation, which is appropriate in discussing the reactions given to novel or familiar stimuli by intact mammals. The main theory is due to Sokolov (1963, 1975), who works in the Pavlovian tradition, and therefore usually uses the phrase ‘extinction of the orienting reflex’ to describe habituation, a practice which I shall not follow here. Another phrase of Sokolov’s, which it is customary to retain, is ‘the formation of a neuronal model of the stimulus’ to describe the processes of categorization and memory which result from repeated experiences of an external event. The essentials of Sokolov’s theory are given in Figure 2.1. The function of the neuronal model is to distinguish novel from familiar stimuli, and also unexpected and surprising from expected and there-fore insignificant events (this being an extra function, since familiar or well-known stimuli may be surprising if they occur at an unexpected time or place). It is supposed in Sokolov’s theory that there is an active process of comparison between an incoming stimulus and the established neuronal model: if
there is a match between the two, the incoming stimulus can be ignored.
Figure 2.1Sokolov’s theory of habituation. After Sokolov (1963, 1975).
(vi) Decreased attention to familiar stimuli
Given a comparator mechanism, any response originally given to a stimulus, such as the drawing-in of a snail’s horns, may be suspended when the stimulus has become familiar and matches expectations. However, in Sokolov’s theory, emphasis is given to physiological reactions, measurable in human subjects, which are correlated to some degree with the attention and arousal generated by incoming stimuli. Such responses are selected partly for convenience, and partly because they appear to be useful indicators of stimulus novelty. They include, understandably, the turning of the head and eyes towards the source of a localized stimulus, and the desynchronization of the electroencephalograph (EEG) which is known to occur with subjective attention to external stimuli in normal human adults. Slightly less obviously, responses which are in practice correlated with stimulus novelty include dilation of blood vessels in the head, and the
‘galvanic skin response’ or ‘GSR’, which is a drop in the electrical resistance of the skin, usually measured in the hand. In the theory, these signs are taken to be indications of the activity of the amplifying systems in Figure 2.1. For our purposes it is convenient to refer to this as concerning attention — in intact vertebrates, and especially in birds and mammals (Mackintosh, 1975) we assume that there is some degree of gating for incoming stimuli, so that insignificant, and habituated, stimuli are not processed in the same way as novel or important ones (see chapter 8). Clearly this involves multi- stage and hierarchically organized perceptual systems, and we would not wish to invoke attention of this kind when discussing the synapses or an individual neuron in the frog spinal cord or in the abdominal ganglia of Aplysia, in (iii) above. In mammals there is ample evidence that the anatomical basis of attention includes the reticular activating system of the brain (Sokolov, 1963).
(vii) Increased capacities for discrimination and classification
Especially in the short term, increased familiarity with stimuli means that they are ignored. However, exposure to the stimulus in the first instance will have elicited increased attention to it, and according to (v) above, familiarity with a stimulus implies that the perceiver possesses a neuronal model, or memory of it, that ‘registers not only the elementary, but also the complex properties of the signal’ which include temporal relationships (Sokolov, 1975, p. 218). Thus, if an already familiar stimulus acquires new significance — for instance, by a change in context, or by changes in internal or external motivational factors — the existence of a preformed neuronal model may improve perceptual performance. This sort of effect was assumed in the theory of Hebb (1949) and confirmed in the experiment of Gibson and Walk (1956), in which rats reared in cages with bas-reliefs of circles and triangles on the walls were able later to distinguish circles from triangles in an experimental task which was failed by others without this previous experience. This is called perceptual learning, or exposure learning (Hall, 1980; Bateson,
1973). Over shorter time intervals something similar may be called ‘within-event learning’ (Rescorla and Durlach, 1981), or ‘latent learning’ (Tolman, 1948). The point, perhaps, is that the biological function of sense organs and the perceptual apparatus of the nervous system is not to enable animals to become indifferent to events thereby perceived, it is to acquire useful information about the outside world. Habituation is in one sense a secondary phenomenon, in that it represents a lessening of response to consistent stimulation; making the appropriate response to a novel stimulus in the first place, and storing representations Of familiar perceptions, are an at least equally important aspect of exposure to stimuli.
Habituation and sensitization in the spinal cord
Thompson and Spencer proposed in an influential theoretical paper in 1966 that the hindlimb flexion reflex of the acute spinal cat should be used ‘as a model system for analysis of the neuronal mechanisms involved in habituation and sensitization’. Following on from the pioneering studies of Sherrington (1906) on spinal reflexes, the recommended object of study was the twitching of a muscle in the hindlimb of a decerebrate cat with a complete transection of the spinal cord at the 12th thoracic vertebra, in response to electric shocks delivered to the skin of the limb every 10 seconds. As no neural information could pass from the hind leg to the brain, and the cerebral hemispheres had in any case been removed in this preparation, it was possible to claim with confidence that any behavioural changes which took place were due to neuronal mechanisms in the spinal cord itself, and Thompson and Spencer provided a useful list of phenomena which they observed under these conditions.
(1) ‘Given that a particular stimulus elicits a response, repeated applications of the stimulus result in decreased response.’ This is of course the basic phenomenon of interest although by itself it is indistinguishable from stimulus or response fatigue.
(2) ‘If the stimulus is withheld, the response tends to recover over time (spontaneous recovery).’ This is more problematical
than it looks, since recovery time may vary from seconds to weeks. It is likely that different mechanisms are involved in habituation which lasts for seconds on the one hand and days or more on the other (egg. Castellucci and Kandel, 1976).
(3) ‘If repeated series of habituation training and spontaneous recovery are given, habituation becomes successively more rapid.’ This also suggests a difference between short-term and long-term mechanisms.
(4) ‘Other things being equal, the more rapid the frequency of stimulation, the more rapid and/or pronounced is habituation.’
(5) ‘The weaker the stimulus, the more rapid and/or pronounced is habituation. Strong stimuli may yield no significant habituation.’ This is why, in Sokolov’s terms, habituation is confined to orienting responses — stronger stimuli giving rise to adaptive or defensive reflexes which do not habituate. Thus an animal may not habituate to the taste of its most valuable food, or the presence of its most dangerous predator.
(6) ‘The effects of habituation training may proceed beyond the zero or asymptotic response level’ This means that after an animal has stopped responding altogether, continued exposure to the stimulus will mean that when the series is ended, recovery of the response will be delayed.
(7) ‘Habituation of response to a given stimulus exhibits stimulus generalization to other stimuli’ (see chapter 8 for discussion of generalization).
(8) ‘Presentation of another (usually strong) stimulus results in recovery of the habituation response (dishabituation).’ This differentiates habituation proper from purely physical fatigue. For Sokolov, it provides evidence for the richness of the neuronal model, since slight changes in a complex stimulus may lead to dishabituation. This has also proved to be a convenient method for assessing the perceptual abilities of pre-verbal human infants (Olson, 1976).
(9) ‘Upon repeated application of the dishabituatory stimulus, the amount of dishabituation produced habituates’.
That is, the organism habituates to the dishabituatory stimulus as well.
(All the above points from Thompson and Spencer, 1966, pp. 18-19.)
Having found all these phenomena mediated by the spinal cord, Thompson and Spencer proposed that the lowering of responsiveness in habituation was caused by one kind of synaptic change, and that rapid dishabituation was caused by a separate synaptic process of sensitization, which increases responsiveness more generally (see (iv) above, p. 39) Groves and Thompson (1970) went on to elaborate this ‘dual process theory’ in a version which continues to receive support (Peeke, 1983; Thon and Pauzie, 1984; Peeke and Petrinovich, 1984). The two processes are simply the decrememental one identified with habituation, and the incremental one identified with sensitization, which are presumed to be to some extent anatomically independent in the nervous system.
Application of the dual process theory to the startle response of the rat
Although the dual process theory of Groves and Thompson (1970) arose from results obtained in spinal cats, in which the sensitization process was prominent, it was applied with some success to an experimental technique commonly used with ordinary laboratory rats. In this the rat is placed in a test chamber which allows for the measurement of overall activity (for instance, by transducers which pick up forces transmitted to the floor of the chamber) and a series of loud tones is sounded. The first few of these induce a distinctive jump from the animal, which gradually wanes in intensity as the series proceeds. This is a standard form of habituation (see point (1), page 42) but numerous other phenomena can be observed, and the effects of ageing, drug treatment and so on can be assessed. A most interesting result was obtained by Davis and Wagner (1969) and termed the ‘incremental stimulus intensity effect’ (see Figure 2.2). In their experiment rats were divided into four groups matched for initial strength of the startle response, and then given a session in which 750
tones were presented, 8 seconds apart. For one group, all the tones were I 20 dB (which is very loud indeed) but over the session there was slow but steady habituation. Another group received tones of 100 dB, and this group showed a much larger reduction in the response to these, but, at the end of the session, when all groups were tested with I 20 dB tones, as might be expected this group showed dishabituation (see point 8, page 43). A third group (labelled ‘gradual’ in Figure 2.2.) started off with 83 dB tones, and the loudness was increased in 2.5 dB steps after every 50 tones, the intensity thus ending up at 118 dB. The striking result was that not only did this gradual group show a low level of startle responses throughout the session, but that this low level was retained to the end, even with the final test using a 120 dB tone. Thus the gradually increasing series appeared to have produced more effective habituation to 120 dB tones than continuous exposure to the 120 dB tones themselves. Clearly this would not follow directly from the ‘neuronal model of the stimulus theory’ of Sokolov (1963, 1975; see point (v), p. 39 above). It would be possible to claim that the fact of the gradual increase was itself incorporated into expectations of the animals, especially since a control group which received all the same stimuli as the gradual group but in a randomized order, gave very different results (this group is not shown on Figure 2.2, but the results for it were very similar to those for the ‘constant 100’ group).
However, since, as Figure 2.2. shows, Groves and Thompson were able to obtain roughly analogous results with the hindlimb flexion reflex of the spinal cat, it seems plausible that, as they suggest, an interaction between a sensitizing and an habituating process brought about the low level of responsiveness in the gradual group. Other results (e.g. Davis, 1972) support the idea that 100 or 120 dB tones, when presented repeatedly, have a generally arousing effect. The argument is along the lines that the gradually increasing intensity allows for very pronounced habituation at low intensities, before the sensitizing effects of the 100 dB and above tones make themselves felt, and that the systematic ordering of the increases allows for the maximum generalization of habituation.
Figure 2.2 Habituation to gradual increases in (a) a rat and (b) a spinal cord.
At a test with a stimulus of high intensity, on the far right of the figures, strength of response is least if the test has been preceded by a series of gradual increases in intensity, by comparison with a series of exactly similar stimuli, or a series of medium intensity. This result has been obtained for (a) the startle reflex in rats and (b) for the spinal reflex of bending the leg in response to an electrical shock to its skin. After Groves and Thompson (1970).
The incremental series of stimuli results in pronounced relative response habituation at low stimulus intensities, and very little sensitization which may decay considerably within each block of trials. Both habituation and sensitization (including decay of sensitization) generalize substantially to each new series of stimuli. The result is a summation of the habituation occurring within all blocks of trials. (Groves and Thompson, 1970, p. 442)
This is less than wholly convincing as a theory, but it is instructive that such complicated results can be obtained with the habituation of a spinal reflex. That this is so does not, however, require us to suppose that all other perceptual processes operate like spinal reflexes in intact animals.
Habituation in a giant sea-s1ug
Aplysia californica is a foot-long slug that grazes on seaweed and weighs several pounds (its English relatives are only a fraction of the size) . Its popularity with experimental physiologists is due to the fact that its nerve cells are gigantic, some up to 1 mm across, and relatively few in number. The behavioural responses of its mantle-shelf, which contains gills and a siphon, are controlled by the abdominal ganglion which contains only about 2,000 neurons, several of which can be individually identified in every animal. This makes it an excellent subject for the plotting-out of which nerve cell does what, and the examination of how the electrical characteristics of neurons change as a function of experience (Kandel, 1976; Castellucci and Kandel, 1976). However, I do not think
I am alone in suspecting that its proponents may be over-rating its suitability for experiments on higher cognitive processes.
If Aplysia californica is kept in a tank of cooled and aerated seawater, its gill and; siphon will normally be extended, but if various areas on the top of the animal, in its ‘mantle-shelf’ where these extensibles are located, are poked or brushed, or stimulated with a strong jet of water, the gill and siphon are temporarily withdrawn. This allows the performance of experiments on habituation of exactly the same kind as that of Humphrey (1930) on a terrestrial snail withdrawing its horns. Kandel (1976) and Castellucci and Kandel (1976) reviewed experiments on habituation in Aplysia. Pinsker et al. (1970) reported rapid habituation of the gill-withdrawal reflex over the first 10 elicitations by a jet of seawater, separated by 3 minute intervals. The reflex recovered after a rest of two hours, but could also be brought back without this long rest if a long and strong tactile stimulus was applied to the neck region. In that report Pinkser et al. note that six of Thompson and Spencer’s nine characteristics of habituation (listed above, pp. 42—4) had been obtained in Aplysia, but that three others were absent: greater habituation with repeated periods of habituation and recovery(3); generalization of habituation to a stimulus in another part of the receptive field (7); and delayed recovery of the response when the habituation series is continued after the animal has stopped responding (6). Given the very specific nature of the neural circuits involved, and the limited body area which produces Aplysia’s withdrawal reflex, it seems inevitable that generalization will be limited: the other two missing characteristics suggest that habituation in Aplysia is only a short-term phenomenon, and does not include the longer-term mechanisms that obtain in even the spinal cords of vertebrates. However, Carew et al. (1972) and Carew and Kandel (1973) demonstrated that, with shorter inter-trial intervals, repeated periods of habituation and recovery do indeed produce faster habituation in the later blocks, in Aplysia, and that after this habituation is still very fast 24 hours later. Thus an appreciable range of behavioural phenomena characteristic of habituation is obtainable in Aplysia, and the essential features can be
observed even when the abdominal ganglion is dissected out from the animal for greater ease of electrical recording.
It is almost tautologous that, if response output to a stimulus input is changing, then there must be changes in the activity of synapses between sensory and motor neurons, but it is of great scientific interest to observe the actual decrements in the excitatory post-synaptic potentials (EPSPs) in the motor neuron, and to discover that this arises because the pre-synaptic terminals of the sensory neuron release progressively less of their normal neurotransmitter sub-stances, when the sensory neuron is repeatedly stimulated. It appears, then, that in Aplysia, synaptic habituation is a change in the sensory nerve — it is not that the motor nerve becomes less able to respond. Also, when there is dishabituation, or sensitization, this is because the synaptic activity of the sensory nerve is facilitated (Castellucci and Kandel, 1976, pp. 30—2).
There is no doubt that these are very important discoveries, arid that Aplysia californica, and researchers dedicated to its use, may make further contributions of fundamental importance to the understanding of the neural basis of learning. But there are two notes of caution that need to be sounded. One is that there are general differences between invertebrate and vertebrate neurons, and therefore results obtained with Aplysia may not be completely representative of all animal synapses. (One difference is that invertebrate neurons are almost all unipolar, that is they do not have separate input and output lines, and therefore do not really have any dendrites: Bullock, 1974.) This is probably not so important in terms of psychological analysis as the fact that mechanisms of synaptic change, while being of the utmost interest physiologically, are simply on too small a scale to be of direct relevance in explaining the phenomena of perception, learning and memory in higher vertebrates or in any other animal (such as the distant relatives of Aplysia, the octopus and the squid) where the behavioural phenomena involve the activity of a whole brain with millions of neurons, rather than a small and very specialized circuit involving only dozens.
On these grounds we may wish to disagree with Castellucci and Kandel (1976, p. 43) when they say that investigations
of the gill withdrawal reflex ‘could also be brought to bear on the analysis of several forms of complex learning’. Kandel (1976) went so far as to present a figure in which the top half is the reduction of a sea-slug’s gill withdrawal reflex during five successive days of training, and the bottom half is the reduction in errors made by a human subject over five days of practice at the task of depressing a Morse key for exactly 0.7 seconds. Since in both cases scores reduce as the days go by, but tend to start off rather high at the beginning of a day’s session, there are superficial similarities between the human and the slug data. But, with due respect to the value of physiological research, it is a form of behaviourist fallacy to believe that such similarities mean anything profound, or that because Aplysia’s habituation takes this form ‘Studies along these lines could specify how long term memory is established and how it relates to the short term process’ (Kandel, 1976).
Habituation in human infants and adults
John Locke (1689/1982) shared the view that the learning abilities of molluscs were in a fundamental sense equivalent to those of the human species, but suggested that the factor of perceptual complexity ought to be considered as supplying a measure of difference between mollusc and man (his examples were cockles and oysters — see p. 8 above) . No evidence obtained in the last three centuries, concerning either snails and slugs or human subjects, suggests that Locke was wrong in this respect. One of the main reasons why theories of habituation based on experiments with human subjects incorporate a relatively elaborate form of stimulus memory, such as Sokolov’s ‘neuronal model’, is that such experiments reveal a high degree of perceptual exactness. The typical form for these experiments is that a subject sits or lies quietly while a stimulus such as a tone of a certain pitch, loudness and duration is presented at regular intervals. Certain responses correlated with attention are measured, for instance the electroencephalogram or skin resistance (see p. 40 above), and these show a gradual decline over 10 to 20 presentations, after which the measurements give no sign that the subject has
detected the stimulus. That the stimulus is indeed being detected, even if ‘pre-attentively’ (e.g. Deutsch and Deutsch, 1963) can be quickly shown by altering it. Even a very slight change in pitch, loudness (either up or down), or duration leads to an immediate recovery of the attentive responses, and in fact the thresholds of perceptual exactness obtained by this method are often lower than when a much more active process of recognition is required of the subjects (Sokolov, 1963; Gray, 1975).
Stimulus representation and the missing stimulus effect
It is because very small changes in the stimulus can be picked up in this way that it is necessary to assume that the internal representation or memory of the stimulus is relatively complete as to all perceivable details. This would appear to include not only the physical characteristics of the stimulus, such as pitch in the case of a tone, but also the context and in particular the temporal sequence in which stimuli are given. If compound stimuli of, say tones and lights, usually occur, then missing out one of the elements of the compound will reinstate attentive responses and, generally speaking, the amount of attention engaged will be proportional to the amount of change in the parameters of the complex stimulus (Sokolov, 1975, p. 218). A very direct way of demonstrating the incorporation of time values into the representation of the stimulus is simply to miss out a stimulus occasionally in a normally regular sequence. This often brings back attentive responses at a high level (Sokolov, 1963). It is arguable that any theory which succeeds in explaining this result, even if it is simpler than Sokolov’s own theory, will nevertheless have to include devices of roughly similar power — for instance Horn’s (1967) theoretical sketch includes ‘extrapolatory’ and ‘comparator’ neurons (Gray, 1975, pp. 20—1). Thus the results of experiments on habituation require us to adopt a position which ought to be agreeable for other reasons — that perception of stimuli by human subjects involves pre-attentive comparisons with stored representations derived and extrapolated from previous experience.
Habituation in human infants
This has proved in the first place to be an extremely useful technique for assessing the infant’s perceptual abilities, but the theory of habituation is also relevant to accounts of the gradual development of the infant’s perceptual competence (Olson, 1976). The usefulness of the techniques arises because, by presenting a standard stimulus repeatedly followed by an altered version of it, the infant’s perceptual sensitivity to that particular form of alteration can be assessed (Jeffrey and Cohen, 1971). As the theory of human habituation demands the assumption that exposure to stimuli results in the building up of stored representations of the stimuli, it supplies a starting point for a general empiricist account of how the empty mind of the newborn gradually fills with knowledge derived from the outside world. This should not be lightly disregarded, even though, as Olson (1976, pp. 24—6) observes, the development of memory in human infants must quickly become elaborated in at least three other ways:
(a) an increasingly fine-grained repertoire of mental features or categories with which an experience can be represented; (b) an increasingly sophisticated repertoire of encoding and retrieval strategies, largely involving language, to aid in recovering memories; and (c) increasingly accurate knowledge about the nature of one’s own memory system, which in turn yields a more realistic selection of strategies and tasks.
Habituation, exploration and curiosity in animals
Such measures of human psychophysiology as the galvanic skin response and vasoconstriction of the limbs bear only a putative relation to the efficiency of perception; while another aspect of the human ‘orienting reflex’, the turning of the head towards a localized stimulus, has a very obvious functional role. In animals the functional aspects of orienting responses are often obvious, since the ears as well as the eyes may scan the environment, and sniffing in mammals is usually an active process. Frequently the whole posture of an animal betrays the detection of a novel stimulus, and many species have well-
defined routines of ‘inspective’ or ‘inquisitive’ exploratory behaviour (Berlyne, 1960; Hinde, 1970; Erlich, 1970; Menzel and Juno, 1985). In these, the initial reactions to a novel stimulus and contrasting indifference given to familiar objects, both suggest that the process of habituation involves the use of stored representations or ‘neuronal models’ of perceived events, rather than merely the depression of activity at certain synapses. There has in fact been an enormous amount of physiological work on mammals such as the rabbit and the rat, not least that performed by Sokolov and his co-workers (Sokolov and Vinogradova 1975) designed to pinpoint the interactions between the sensory cortex and the limbic system, most significantly the hippocampus, which are presumed to provide an overall system for attention, memory and habituation, with quite different cognitive capacities from those exhibited by the spinal cord (see Gray 1982, 1984; O’Keefe and Nadel, 1978). This work will not be reviewed here, but it is pertinent to mention at every opportunity that behavioural . i similarities in habituation observed across disparate animal species, do not imply that identical physical mechanisms are at work in all cases, or that the summary of the phenomena that suffices for the simplest cases is all that it necessary to understand and account for all the others.
Also, the involvement of the limbic system of the brain in the development of stimulus knowledge and memory suggests that, in addition to the dimension of the perceptual complexity of habituation in intact higher animals, it is also necessary to discuss the motivational significance of the reactions to novel and familiar stimuli. In one sense the importance of habituation is that it refers to a category of experiences which are independent of the imperatives of pain and need —habituation occurs in the experimental context without the addition of the usual ‘motivationally significant events’ (Dickinson, 1980) of desired food objects or unwanted discomfort. On the other hand, a great deal of behavioural evidence (Berlyne, 1960; Hinde, 1970) suggests that stimulus novelty and familiarity should be regarded as a motivational system of its own, even if not unrelated to choice in foraging, or to wariness and fear. To some extent all these must interact in natural patterns of behaviour, and the concept of neophobia
in feeding responses has been useful in descriptions of a number of aspects of food choice, especially in rats (Cowan, 1976; Rozin and Kalat, 1971) but in other species as well. As a species, rats are conservative about their food and prefer to eat what they are already used to, unless they are either (a) very hungry indeed, or (b) in poor health — in both cases a change in diet is advisable (Rozin and Kalat, 1971). Curiosity about novel but clearly inedible objects such as latches or manipulable puzzles is most marked in primates (e.g. Butler 1953; Harlow et al., 1950) for reasons which are poorly under-stood. However many species show some kind of fear reaction to a stimulus that is intense and vivid and very novel —investigative reactions may follow when it becomes slightly less of a novelty. Very many other species also show some degree of social curiosity, whose expression depends of course on the instinctive social pattern of each. Search for additional physical or social stimulation as an apparent end in itself merges with the specialized subject of play (Smith, 1986) which is most common in the young of co- operative mammalian species including carnivores and primates.
In all these cases species-specific patterns of behaviour are dominant, but over and above these the most plausible generalization is that of Berlyne (1960) that an internal level of arousal, correlated with stimulus novelty, is responsible for some of the motivational impetus of play and exploration. Familiar stimuli are of course uninteresting and unarousing, and the essence of Berlyne’s theory is that every individual has an optimal level of arousal. Therefore, for an under aroused or bored animal, novel stimuli may be rewarding, while once the optimal level has been exceeded, security and familiarity are sought. The alternation between these two states is some-times directly visible in young primates (Harlow, 1962; Mason, 1967).
Conclusion: habituation is not always the simplest form of learning
The description of habituation — that we start off with one stimulus, which elicits one response, and then cease to do so as it is repeated — is certainly about as simple a description
as one could give about a behavioural phenomenon which might qualify as learning from experience. Even the nine points listed by Thompson and Spencer (1966) to do with how much and when a response declines in habituation, and when it might recover, do not require a tremendously complex psychological theory for their explanation. That even the simplest of animals may exhibit responses which decline with repeated elicitation strongly suggests that relatively unelaborate biological solutions have been found to the problem of evolving a system which displays habituation, and the fact that all of the nine points can be obtained from vertebrate spinal cords, and most of them from the abdominal ganglion of a gastropod mollusc, is ample experimental confirmation that this is indeed the case. But, if we take exactly the same behavioural description, or set of behavioural descriptions, and apply it to the decline in attention which a human subject gives to a repeated stimulus of no great intrinsic interest, or even to the decline in exploratory behaviour elicited by a novel object placed in a rat’s cage, we are under no obligation to begin with the assumption, that the explanations which apply to the slug or spinal cord will continue to be appropriate. On the contrary, the results of such additional experiments, since they demonstrate the complexity of both the initial perceptions of novel stimuli and the comparisons with internal representations which change novel stimuli into familiar ones, mean that what is describable as a decline in responsiveness to a repeated stimulus in these cases requires an explanation of quite different order, involving both the perceptual and the motivational systems of the whole animal, rather than just a few synapses which intervene between stimulus and response. Similarity of behavioural description does not imply similarity of psychological explanation. I trust that the reader will become thoroughly bored and irritated by the repetition of this point during the next two chapters, since this will mean that he or she has begun to compare it with an internal representation which, while this may lower attentiveness to immediate repetitions, will also indicate the involvement of one of the highest and least reflexive forms of learning.