The Effect of Sex-Hormones on the Performance of a Learned Response

The restoration of copulatory responses in gonadectomized mammals following the administration of sex-hormones (Allen & Doisy, 1923; Marrian & Parkes, 1930; Boling & Blandau, 1939), and the return of cyclical variations in running activity, a characteristic of normal oestrous behaviour in the white rat, when an oestrogen is administered daily (Hemmingsen, 1933; Young & Fish, 1945), exemplify the fact that automatic adjustments of an animal’s behaviour to its environment are governed by the co-ordinated action of its nervous and endocrine systems. In spite of the almost axiomatic quality of this major generalization, and of a considerable literature on the gross patterns of behaviour that are controlled by sex-hormones, few attempts have been made to analyse the character of the integration in specific instances. In view of the intricacy of the underlying processes, this is not surprising where learned responses are concerned. Differences that may be found in the rate or character of solution of a problem, after the administration of some drug or hormone, will merely show that some change in behaviour has taken place, without indicating what alteration in physiological function underlies the change.

The more immediate study of simpler acquired functional patterns of the central nervous system has, however, been eased by the use of the ‘Goal-Gradient Alley’ which Hull introduced in 1934 as an aid to the analysis of the dependent variables of maze learning. Hull’s (1932) hypothesis of a ‘goal-gradient’ was derived from the observation that during maze learning rats run more rapidly in the later than in the earlier part of a maze, and that blind alleys nearer the goal tend to be dropped out of the response pattern earlier than those near the start. He therefore postulated an ‘excitatory gradient’ in motivation and strength of conditioning, the gradient increasing in intensity from the beginning of the maze to the food box. This excitatory gradient arose, he thought, from certain characteristics in the mechanism of learning. At each stage in the performance of a learned response, the strength of excitation is a product of the interaction between the sensory input at that particular moment of the response, and the neural pattern built up during the earlier stages of the response, and during all stages of previous similar responses. Hence there is a progressive increase in the strength of excitation as performance proceeds. The increase in strength results from the gradual addition of stimuli associated with the goal, and the elimination of stimuli that are not so associated.

The general validity of the main hypothesis and of its postulates, in given conditions, has been established. Hull (1934), arguing from the premise that the stronger an excitatory tendency, the more rapid will be the actions which result from it, showed that in a goal-gradient alley, which essentially is a long straight tunnel with a food box at one end, there is a speed-of-locomotion gradient paralleling the assumed goal excitatory gradient. Thus he demonstrated that in the process of learning to run the alley, rats ran faster near the end of the run, i.e. nearer the food box, than at the beginning. When learning was complete, the speed was the same in all sections of the alley. On the other hand, inhibition of the learned response by satiation or non-reward caused a return of the excitatory gradient—as manifested by an increase in the time taken by the rats to cover the first part of the alley.

Drew (1939), however, showed that a gradient in running speed, with higher speed nearer the goal, did not occur if training was restricted to a single daily run, but was present during learning only if rats were given successive trials during each day’s training session. In his view the normal response lacked a gradient in running speed from the first to last section of the alley, and any improvement in performance during training was equally manifest in all sections. A gradient, he concluded, was a superimposition, and could occur in a learned response only in the presence of factors that tended to weaken and inhibit the connexion between stimulus and response at the moment of response. Increase in the gradient is therefore a measure of these factors, and may be found when trials are rapidly repeated, or following satiation or non-reward. When there is no inhibition, as for example, when one training run only is given a day, there is therefore no gradient. Drew restated these conclusions by suggesting that Hull’s postulated excitatory gradient should be considered as a gradient of inhibition, which lessened as the goal was approached. Drew’s conclusion finds support in the work of other experimenters (Buel, 1938; Buxton, 1940).

Miller & Miles (1935) had previously used the goal-gradient alley to analyse the effects of caffeine upon the speed-of-locomotion gradient caused by satiation and non-reward, and hence upon the strength of conditioning. Because there is available to-day a considerable body of knowledge on factors which may cause a change in performance in the goal-gradient alley, this technique seemed suitable for an analysis of the effects of sex-hormones on conditioning. It was accordingly used in the investigation described in the present paper.

There is little previous work with sex-hormones in this field of study. Herren & Haterius (1931) found that the central reflex time of the Achilles jerk is greater during oestrus than in dioestrus. Bilateral abdominal sympathectomy abolished this increase, which was therefore considered to be due to impulses arising from the hydration of the genital tract (Herren & Haterius, 1932). Later observations by Chauchard (1943) that the excitability of nervous tissue, as indicated chronaxi-metrically, is depressed by sex-hormones, is in accord with these earlier results. A still earlier observation of Kreps (1923) indicated that oestrus disturbs previously established conditioned responses in the dog, and more recently Vanderplank (1938) has shown that injections of oestriol will inhibit previously established conditioned reflexes in the rudd.

The apparatus consists of a straight wooden alley, divided into five 3 ft. sections. Sections are interchangeable, and, when joined together, provide a 15 ft. enclosed tunnel 4 in. × 6 in. in cross-section, the roof of which can also be removed in sections. At each end of the alley is a short extra section, 1 ft. 3 in. in length. These serve as starting and reward boxes. The inner walls of the sections and end boxes are painted black, and the whole alley is mounted 3 ft. 6 in. from the floor on an iron frame.

The starting box is closed by a lid, and access to the first section of the alley is by a sliding door which is lifted vertically by hand. There is no lid to the reward box. The five sections and reward box are divided from each other by light, wooden, oneway swing doors. The opening of the sliding door at the start, and of the swing doors at the end of each section, are recorded electrically on a chronograph run from a constant speed motor, giving a paper speed of 1·1 cm./sec. The chronograph is remotely controlled, and is not housed in the same room as the alley.

Fifty-two sexually mature female white rats, weighing between 150 and 200 g. were used. They were received from a breeding station, and were reputed to be 3–6 months old. The litter relationships were not known. After spaying, they were housed in eight separate cages, containing 9, 8, 7, 7, 6, 6, 5, 4 animals respectively.

Training did not begin until at least a week after spaying, and after animals had fallen to 85 % of their preoperative weight. The training technique is fairly simple.

On the first day animals are placed in the final section of the alley, with its roof off, and a feeding dish containing bread and milk is placed in the reward box. They are then gently pushed through the swing door between the last section and reward box, and allowed to feed for 30 sec. This is done three times consecutively, after which each animal is given 3 g. of bread in a separate cage. This procedure is repeated the second day, after which the animals are moved one section farther from the reward box each day, and the already learned section is roofed in. After the whole alley has been learned in this way, full training runs are given once every 12 hr., at 8.30 a.m. and 8.30 p.m. In this way complete learning, indicated by a small trial-to-trial variation in time for individual animals, is reached in 7–10 days.

The training method has been described in detail, because observation confirmed previous investigators’ experience (Hull, 1934; Drew, 1939) that the form of response in the alley is affected by both the occurrence of inhibition and the type of training. Inhibition, as well as producing the typical speed-of-locomotion gradient described by Hull, also restores to a conditioned-response sequence the type of behaviour found earlier in training. Thus rats trained by being placed at the beginning of the alley and pushed to the end will, when inhibited, tend to run more slowly where they experienced most difficulty before (i.e. in the first section). Rats trained as in the present experiment will, on the other hand, usually run slowly in the last section.

In the present experiment the aim was to train each animal to a mean running speed near the population mean, and this criterion, rather than identical treatment for each animal, governed the exact number of training trials given to each animal. When learning was complete, each animal fed for 30 sec. in the reward box at every 12-hourly trial, and was then removed manually to a separate cage and given 6 g. of bread soaked in milk. Trial and error showed that this amount was always finished within 20 min., at the end of which time animals were returned to their home cages, where they were given water ad lib. If less food is given at each trial, motivation is so high as to mask any alterations in response that may be influenced by hormonal stimulation. The régime described, although providing a low motive, gives consistent results for individual animals, and a consistent population mean.

The fifty-two trained animals were divided arbitrarily into four equal groups, because statistical analysis showed that in so far as they did not differ significantly in either the means or the variances of their running-times they could all be regarded as forming a single homogeneous population.

Running time was then recorded after the injection of:

Before injection, the running times of the animals in each group were recorded twice at 12-hourly intervals. The results for the groups were once again analysed to discover whether there were significant differences either in population means or variances as between different times and different groups. None was found. Running-time was then measured every 12 hr. from the 6th to the 114th hr. after injection.

The running times for the animals in each group were pooled, and the mean running-time and the variance of the groups were compared with control group I, which received an injection of oil only. Means were compared by Fisher’s ‘t’ test, and Fisher’s method of analysis of variance, as described by Snedecor (1946), was used.

Differences were regarded as statistically significant when there was less than one chance in fifty that such a difference would attain the observed figure if the samples were random ones drawn from the same population.

The results obtained are presented in Figs. 1–3, and Tables 1–3.

Fig. 1 shows the mean running-time for each group plotted against the time in hours after injection. At each of the ten 12-hourly intervals of the experiment at which recordings were taken, the mean running-time of groups II (oestradiol alone) and III (oestradiol and 200γ testosterone) was greater than that of control group I. On the other hand, group IV (oestradiol and 1 mg. testosterone) was very similar to group I.

An analysis of variance (Table 1) of the group-mean running-time at each 12-hourly interval shows that the variance between the four groups after treatment was significantly greater than that within the groups.

The differences between the group-mean running-times for groups I and II, and groups 1 and III are, however, statistically significant only at 78, and 18 and 78 hr. after injection respectively (Table 2). The mean differences at other times do not appear statistically significant because of the high variances of the individual observations.

Thus it will be seen from Fig. 2 that the variances of groups II and III are greater at every observed interval after injection than the variance of control group I. The variance of group IV, on the other hand, corresponds to that of group I. The records of each individual animal were analysed to discover why the variance was so high in groups II and III after injection, and hence why the mean differences at each 12-hourly interval were not always significant when compared to the mean running-time of the control group. It was found that although a significant behavioural change can be demonstrated in the majority of animals in groups II and III (ten out of thirteen in group II, and eleven out of thirteen in group III), this change did not take place at the same time after injection in each animal. Nor was the duration of the changed behaviour the same for all animals. Thus, whereas one animal exhibited a significant response decrement at one experimental session only, another continued to show it over three or four consecutive 12-hourly observations.

Table 3 compares the mean differences in the average running-times of each group for the whole alley over the total experimental period. The times taken to run each section of the alley are also compared separately in this table and in Fig. 3.

The differences in running-times for the whole alley between groups I and II, and between groups I and III, are highly significant.

Groups II and III also took longer than control group I in all sections of the alley. Group IV, however, behaved in the same way as the controls.

The difference in timing over sections was statistically significant in only the first section of the alley for group III, and the last section for group II. The difference approaches significance, however, for all sections except the third for group II, and the time spent by this group in the first two sections is significantly greater than that spent by control group I. (n=13, diff. = 0·80, S.E. diff. = ±0·29, t = 2·818, P < 0·02.)

All these differences were not due to the different amount of oil given as a vehicle for the hormones, for from this point of view group I acted as a control for group II, and group III for group IV. It therefore follows that the administration of 20γ oestradiol benzoate to spayed mature female rats causes a response decrement of a characteristic type in a previously learned reaction. The same result is observed after the administration of 200γ testosterone propionate together with 20γ oestradiol benzoate. But if the dose of testosterone propionate is raised to 1 mg., the response decrement does not occur.

There were no differences between the behaviour of groups II and III. Nor were there any between groups I and IV. It is therefore probable that the response decrement found in group III was due to the oestradiol alone, and that 200γ testosterone propionate was too little, and 1 mg. testosterone sufficient to neutralize the effect of the injected oestrogen.

Vaginal smears were taken from all animals during the experimental period. No cornification was noted in any animal of group I. Marked cornification was first observed between 30 and 42 hr. after injection in all animals of groups II and III, and persisted long after the 114th hour. All animals of group IV also showed some cornification. But although cornified cells appeared 30–42 hr. after injection, their number was much less than in groups II and III, and they disappeared 24–96 hr. after their first manifestation. This is further evidence that the 1 mg. of testosterone propionate neutralized the injected oestrogen.

These experiments show that oestrogens can cause a decrease in the speed of performance of a learned response. The increase in time taken to run the alley is due mainly to an increase in the time taken to traverse the first two and the last sections of the alley.^* This type of decrement in response has been observed by other investigators (Hull, 1934; Drew, 1939) and confirms the previously established principle that weakening of a learned response restores to that response behaviour which occurs in earlier stages of training,

The possible physiological mechanisms underlying the gradual weakening of a learned response may be considered seriatim. Loss of a conditioned response may result from the repeated rapid elicitation of the response (adaptation), from alteration in the strength of the conditioned stimulus—for example by satiation or repeated omission of a reward for performance (internal inhibition), or from inter-ference due to the interaction of other stimuli during the elicitation of the response (external inhibition) (Pavlov, 1927; Hilgard & Marquis, 1940).

That the response decrement did not represent adaptation in the sense in which the term is defined is indicated first by the fact that the conditioned responses showed little tendency to diminish with the passage of time, and secondly by the fact that the present experiment did not allow rapidly repeated elicitation of the response.

It also seems clear that the decrement in response was not due to internal inhibition, since the strength of the conditioned stimulus (hunger) was kept constant. After being run, all rats consumed the 6 g. of food they were given in a separate cage. Although motivation was never maximal, there is no indication of its alteration. Lack of a significant alteration in the mean body weight of each experimental group between the 6th and 114th hr. of the experiment supports this conclusion.

The explanation for the weakening of response presumably falls, therefore, into the category of external inhibition. This may be either from an alteration in the quality or in the quantity of the extraneous stimuli acting upon the organism during performance, or a direct effect of oestrogens upon the functional equilibrium of the central nervous system. Both may in fact be facets of the same process.

It is known that the water content of a wide variety of tissues is affected by oestrogenic stimulation (Zuckerman, Palmer & Bourne, 1939). That of the skin, uterus and vagina increases 6–18 hr. after injection of 0·1 γ oestradiol, and decreases at 24 hr. There is a further rise in water content from 30 to 60 hr., after which water is no longer retained in the tissues, and the water content drops to below the normal resting level. The changes in water balance in the brain, heart and gut are to a certain extent reciprocal to those observed in the uterus. There is a rise in water content 24 hr. after injection of 0·1γ oestradiol, followed by a fall at 30 hr., and a further rise between 48 and 66 hr., when the uterus is beginning to lose water.

That the uterus, which is but one of the tissues involved in these widespread changes in water balance (Astwood, 1938), may be a source of inhibitory impulses to the central nervous system has been suggested by Herren & Haterius (1931, 1932). It is also known, however, that the cyclical alterations in spontaneous activity that characterize the oestrous cycle of rats do not depend on impulses arising from the genital tract (Durrant, 1931) and that they therefore depend upon changes in other tissues. Direct alteration in the functional equilibrium of the central nervous system after oestrogenization is shown not only by the changes in water balance, but also by the alteration in threshold to some environmental patterns of stimulation (for example, sexual patterns) ; by the selective facilitation of certain motor responses relating to these patterns; by the alteration in certain autonomic reactions; and by changes in equilibrium within the autonomic nervous system. Hydration of the brain, which may cause a relative cerebral anoxia, might in itself lead to alterations in the function of the central nervous system, and hence in its ability to subserve the performance of a learned response. For example, Welsh (1943) has shown that the responsiveness of the cortex to stimulation, and the vigour of its spontaneous activity, decrease markedly under conditions of anoxia.

There is considerable evidence, however, of the specific effect of gonadal hormones on autonomic equilibrium. Anderson (1940a) has shown that female rats are less timid when sexually receptive, and that timidity increases after spaying (Anderson, 1940b). It has also been shown that the timidity of spayed rats decreases after oestrogenization (Anderson & Anderson, 1940). Heat periods can recur in the absence of afferent impulses from the genitalia (Sherrington, 1906; Bacq, 1932; Bard, 1935), and the results of experiments involving trauma to the central nervous system have led many workers to the conclusion that the motor patterns of female copulatory behaviour must result from facilitation by oestrogens of impulses from autonomic motor centres located between the midbrain and anterior hypothalamus (e.g. Brooks, 1937; Bard, 1940; Brookhart, Dey & Ranson, 1941). A specific affinity of sex hormones for cells in this region has therefore been postulated. Effkemann & Striiver (1939) have further shown that oestrogens have a sympatheticotropic action, which they measured by alterations in pupillary diameters. Effkemann (1939) also showed that testosterone had a parasympathetico-tropic action when measured by the same method, and thought that his results were also due to the specific action of gonadal hormones on diencephalic autonomic centres.

External inhibition may therefore result directly from impulses arising from tissues which retain water after oestrogenic stimulation, or indirectly from the effect of this hydration upon other functions of affected tissues (for example, the effect of hydration upon the blood supply to the brain), or indirectly from the action of oestrogens upon autonomic adjustment reactions. The importance of the last possibility lies in the fact that the strength of any superimposed and acquired adjustment involving learning must depend upon the integrity of the autonomic adjustment an animal makes to its environment. Alteration in autonomic adjustment may follow variation in the quality of sensory input, and such alteration may therefore be the antecedent to the presence of external inhibition to a learned response.