Holt has formulated (1931, pp. 37–40) an ingenious theory to account for repetitive and imitative cooing and babbling in the small child. His theory has been so widely noticed and utilized by social psychologists that it seems worth while to address special attention to it. Since the associationist position leads to different predictions of behaviour than does a consistent reinforcement position, it is desirable to make clear the logic behind these predictions.
Holt advocates the reflex-circle hypothesis of iterative behaviour. His argument (1931, pp. 37–40), condensed and sometimes paraphrased, runs substantially as follows:
Nervous excitations, seeking some outlet of least resistance, find their way purely fortuitously into the motor neurons of the muscles involved in making a given sound. The sound is made as a random response. But then a response occurs which is not random. The sound stimulates the child’s ears which send a distinctive excitation along the auditory nerves to the central nervous system. But this excitation arrives only a second or two after the above-mentioned random impulses have found, or while they are still finding, outlet from the central nervous system into the muscles which produce the sound. Therefore by simultaneous association the incoming excitation from the ears will find outlet along the tracts just used by the random impulses, that is, will go back to, and will further contract the exact muscles used in producing the sound. After a few repetitions of this process, the specific impulses which the sound elicits in the ears will acquire a synaptic connection with the motor nerves going out to the same muscles used in producing that sound. A reflex-circle will be established. Hence the infant persistently repeats any sound which he hears himself make. As soon as this process is firmly established, he repeats also those same articulate sounds which other persons utter to him. When the child is stimulating himself, the behaviour is called iteration; when a second person provides the stimulus, it is called imitation.
The writers, like so many others, are indebted to Holt for focusing attention on the problem of iteration and for taking the first steps in working out an explanation of it. Holt offers valuable clues in the observation that the child can hear himself talk while talking and the inference that this fact might bear on
the problem of imitation. The wide attention which Holt’s hypothesis has received indicates the great demand which exists for a basic approach to the problems of imitation, especially in relation to the acquisition of speech. It is particularly necessary to be explicit about this debt to Holt since this appendix will be an exploratory modification and amplification of his hypothesis. The position of the writers is occasionally presented in dramatic contrast to his in order to bring out succinctly the different implications of a simple associationist and a reinforcement position.
Reinforcement theorists agree with associationists that infants do make vocal responses and that these responses in turn stimulate the child. It is likewise agreed that children may respond to speech cues from others by making sounds similar to the stimulus. The query arises on a quite different score. Once a child has begun to repeat his own sounds according to this mechanism, how does he ever stop?1
Since each response is automatically associated with its own stimulus, and it is assumed that each association strengthens the connection, the iterative behaviour should continue indefinitely. Likewise, when stimulated to imitate by another person in the manner above described, the iterative behaviour should then continue without ceasing. This expectation from the Holt theory is not verified in the facts of children’s behaviour, and some modification of the Holt hypothesis is therefore imperative. Some hypothesis must be found which explains how the child can be induced to stop crying, babbling, or cooing.
The weakness in the Holt theory seems to be that he assumes that the connection between the vocal response and the auditory stimulus will be strengthened by mere temporal contiguity of these two events. This single assumption does not adequately interpret the facts. A temporal relationship is undoubtedly one condition of establishing a connection between a stimulus and a response—a response to a cue must occur before it can be learned—but this does not seem to be the only condition essential to learning. It is urged in this instance, as elsewhere throughout this book, that reward is essential to the strengthening of a connection
between response and stimulus, and that, without reward, extinction will occur regardless of temporal contiguity of stimulus and response.
The discussion which follows is an attempt to work out the implications of a reinforcement theory for the iterative and imitative vocal behaviour of small children. It is put into the appendix of this book because it is tentative and exploratory. The detailed history of even one child cannot be presented to verify or disprove the reinforcement hypothesis. The baby discussed is, like Holt’s a hypothetical baby; but this application of the reinforcement hypothesis may facilitate further discussion and research. At least such casual observation of infants’ behaviour as comes to hand seems to support a reinforcement interpretation. Investigations have not indicated that children always vocalize at the sound of similar vocal stimuli from other children.2
What is maintained here is that when they do babble or coo they should do so when rewarded and not by mere association.
The application of a reinforcement theory to the vocal behaviour of children will be presented with the aid of several examples. Responses selected for special exploration will be those of babbling and crying. Babbling behaviour will be discussed first, and very briefly.
Since the mother talks to the child while administering primary rewards such as food, the sound of the human voice should acquire secondary reward value. Because the child’s voice produces sounds similar to that of his mother’s, some of this acquired reward value generalizes to it. Thus circular connections of the type described by Holt may be strengthened by the rewarding value of the child hearing himself produce a sound similar to that produced by his mother. Whether the acquired reward value would be strong enough to produce a marked tendency to imitative babbling is in doubt; it is also doubtful that children actually have a strong tendency to imitative babbling. If the same sound is repeated in rapid succession without support from some primary reward, the acquired reward value of this sound is extinguished and the babbling response should be weakened until other responses occur. Thus the child should shift to babbling with a different sound and eventually
stop babbling altogether. After the passage of time, both the extinguished response of uttering the given sound and the acquired reward value of hearing that sound may undergo spontaneous recovery so that the child will return to uttering the sound. If the sound is again associated with primary reward, still more of its acquired reward value will be restored. As the child learns a better discrimination between the sounds which he utters and those which he hears others utter, the acquired reward value of his own sounds will be reduced. From this hypothesis it may be deduced that children talked to while being fed and otherwise cared for should exhibit more iterative and imitative babbling than children not talked to while being rewarded. An experiment to test a related deduction is suggested in Chapter V
, page 68, footnote 1.
The inferences to be drawn from a reinforcement position will be worked out in more detail in the case of crying behaviour. The drives for crying are often strong primary drives; those for cooing and babbling are usually weaker and more obscure. If a child babbles iteratively, so ought he to cry. The focus of the discussion will be an infant crying for his bottle. The following questions will be discussed: (1) Why does the child cry at all? (2) Does he cry harder upon hearing himself cry? (3) When ought he to stop crying, if at all, before he gets his bottle? (4) Should he cry harder upon hearing another child cry?
WHY DOES THE CHILD CRY AT ALL?
This question is neglected in the associationist account. Holt’s baby is vocalizing as a result of “fortuitous nervous excitation”. The “reinforcement” baby vocalizes only when the common drives are in operation.3
In a typical case, the motivation is the hunger drive. It should be noted that this drive is itself a stimulus, and a stimulus with both the properties possible. It has intensity and cue value. It provides a motive power sufficient to get the child into action and can acquire the distinctive function of determining which response will be made. At the very first appearance of the hunger stimulus, a large number of
responses will be made. Crying is high in the innate hierarchy of such responses. It is a response very likely to be rewarded because of its efficiency in acting as a stimulus for the adults around the child. As compared with wriggling, for instance, it can attract attention when wriggling will not because the adult does not have to be within sight of the child to respond to crying. Cries carry around comers. Furthermore, adults cannot close their ears to crying, whereas they may, as when asleep, close their eyes to exclude a visual stimulus. Crying is therefore a very efficient response, from the standpoint of the child, which is likely to bring the mother near and to set her in search of the drive which needs to be reduced. The assumption that a crying child is hungry is general among mothers in our society, so hunger-tension reduction in the child is rapidly effectuated. The result of such reduction is to reinforce the immediately preceding responses and, among them, crying. In this way, the crying response rapidly becomes dominant over other responses to the hunger-drive stimulus. Stated another way, the hunger drive has acquired cue value. It impels a specific response, i.e., crying.
This behavourial sequence of drive—crying—being-fed occurs repeatedly in the early weeks and months of the child’s life. If standard feeding routines are followed, there will be at least eight hundred trials in the first six months. If the child is fed whenever he is hungry and cries for food, the mass practice will be even more effective. The result is a powerful reinforcement of vocal behaviour in the infant. He learns to cry when the hunger drive, or any other drive, mounts.
It is well known that crying can be wholly or partly extinguished in the small child. If the parents are instructed not to feed the child when he is hungry and cries, the crying response will tend to drop out to the hunger-drive stimulus. Its innate dominance as a response to a strong drive is, however, so high that it may never drop out altogether. The trace of a cry may remain in spite of repeated extinction practice, especially when the drive stimulus rises to great intensity.
The associationist theory—at least not without elaboration—apparently cannot predict with certainty when the child will start crying (since the stimulus is fortuitous), when he will continue crying (since there is no fixation of response by reward), nor when he will stop crying (since there is no weakening of bond between stimulus and response by unrewarded trials).
To this the associationist would answer, “Yes” according to the reflex-circle theory outlined above. The reinforcement theorist would answer, “It depends”. The question then becomes: On what does it depend? Certainly not on the reflex-circle theory, because children are still hearing themselves cry just before they start to taper off their crying, and this excitation should tend to perpetuate or increase crying. The inferences which must be made from a reinforcement theory seem more valid.
The first time the child cries to the hunger-drive stimulus there is no reason to expect a difference in the strength of the crying response at any place along the continuum of crying responses. The second time he cries, however, there should be a difference. In the first trial, the crying responses to stimuli at the end of the series have been more strongly reinforced than to those earlier in the series, according to the principle of the gradient of reinforcement. As practice continues, the child should then cry yet harder at the end of the series than at the beginning. Another variable is present, however, to complicate the situation. When the child cries, he stimulates himself in two ways: First, the response of tensing the larynx gives off characteristic stimuli, and, second, the vocal response sets up an auditory stimulus, as Holt has observed. These stimuli are very much alike along the whole continuum of crying responses, presuming that the child cries for a few minutes before being fed. The child stimulates himself in very much the same way at the beginning
of the sequence as at the end when these responses are much stronger
. When he begins, say, the twentieth trial of crying to get a bottle, a new variable has appeared. The child cries to the stimuli of hunger drive, room, light, and so forth, and also
to the stimuli given off by his first cry. The earlier stimuli, therefore (by the principle of generalization), tend to evoke the response characteristic of the latter part of the series, i.e., the stronger crying response. One would expect, for this reason, that after the first cry, on the twentieth trial, the child would tend to emit the stronger cries characteristic of the end of the sequence; that is, the end response would become anticipatory in the time sequence and become attached to the earlier proprioceptive and auditory stimuli. In this way, the child would
cry harder when he hears himself cry after practice than he had done on the first or second trial of the sequence.4
If the anticipatory strengthening of the crying response is reinforced by earlier reward, the child will tend to cry harder on succeeding trials when he hears himself cry. Strong crying is, in fact, often rewarded. If, however, the stronger cries earlier in the sequence are not reinforced, one would not expect this type of strengthening. In the latter case, the vehemence of the whole series of cries would be expected to be reduced, although the characteristic curve should be maintained.
WHEN OUGHT HE TO STOP CRYING, IF AT ALL, BEFORE HE GETS HIS BOTTLE?
Holt’s hypothesis can scarcely be adapted to answer this question because, as has been pointed out, he offers no mechanism by which the child can stop responding. According to the hypothesis offered here, the child ought to learn finally to stop crying before actually getting the bottle. This would be expected to occur in the following way: In the feeding example under discussion, the child cries vehemently up to the point of reinforcement. During the crying sequence, a fatigue drive has been mounting. Each cry produces its own tension in the throat muscles. The longer the crying persists, the stronger will this new drive be. Two rewards occur at the point of feeding: first, the child gets the bottle and reduces his hunger drive; second, the child stops crying and reduces muscle tension in his throat (the fatigue drive). It is the first response, that of sucking, which actually stops crying in this case because sucking on a bottle is incompatible with crying and is more strongly rewarded. The fatigue drive, however, is simultaneously reduced, as just pointed out, and therefore the last crying response is doubly reinforced.
The last crying response in the series will be examined from the standpoint of the stimuli which it evokes. These stimuli are of three types. First, proprioceptive, that is, feeling himself cry; second, auditory, that is, hearing himself cry; and finally environmental, that is, seeing mother and bottle and hearing her voice or her footstep. Feeling himself cry and hearing himself cry have been present during the entire crying sequence and the
end stimuli are apparently very much like earlier stimuli in the series. This fact suggests that the goal responses or fractions of them will become anticipatory and move forward in the sequence. This is probably what occurs. The two responses in question are swallowing and stopping crying. Swallowing is not likely to move forward into the series of crying responses because it is not an efficient response in an anticipatory position, e.g., it does not shorten the time of bearing drive but may actually lengthen it.
The case is otherwise with stopping crying. This response may also become anticipatory and move forward from the point of reinforcement. If crying can be stopped before reinforcement occurs without delaying or reducing it, stopping will tend to occur as a response before the point of reinforcement. The hunger reward is not delayed, but the fatigue drive is reduced more quickly.
How far forward may the stopping response move? Suppose the relaxing response has generalized forward until it occurs at the time point when the drive stimulus passes threshold, a point where the crying response would normally be evoked. In this case, the child does not cry, the mother does not come, and hunger drive continues to mount. Net reinforcement over time is reduced.5
The relaxation response at this point is not rewarded and is therefore extinguished. Consequently, the response next in dominance occurs, and, eventually, crying.
Since there is a gradient of generalization of extinction, as of reward, the relaxing response will not be tried out on the very next stimulus unit to the drive stimulus. It will move forward along the series towards the point of reinforcement and will be tried out at, say, some distinctive stimulus such as the fifth (of
an imaginary twenty) in the series. If it is extinguished here, it will move still farther towards...