The third question raised above can be answered only by experimentation. Which of the above methods is most advantageous when three problems are learned abreast? Three groups of animals entirely different from those just discussed were allowed to learn simultaneously boxes II and III and maze I (pp. 95 and 100).
Group I. 1 trial on box II. then immediately 1 trial on maze I, then 1 trial on box III. " II. 3 trials on box II, then immediately 3 trials on maze 1, then 3 trials on box III. " III. 5 trials on box II. then immediately 5 trials on maze I, then 5 trials on box III.
For our purpose we need to examine only the distribu- tion curves of the learning of box II. These curves are to STUDY OF HABIT FORMATION § H h^ w § > § 03 P 03 o W ^ H O o O {^ H y 0 a PQ S ^ Ew M OQ O PQ Q w o H Hi h te: ni^ .J Oh "I c CE 1^ E^ njTEt ce: Ha EFFICIENCY OF TRAINING METHODS 233 be compared with those on p. 229. Looking first at the num- ber of trials required to learn the problem we see (1) that by any of the methods the absolute number of trials re- quired to learn box II, when 2 other problems are learned simultaneously, is far greater than when box II is learned alone; (2) that the one-in-one method shows the greatest economy so far as the necessary trials are concerned when contrasted with the three-in-one and the five-in-one methods.
Average Time of Last Tbial on Average Time on First Regular Practice in Retention Test No. animals Minutes Minutes Av. Av. M.V. Av. Av. M.V.
The above table on retention, constructed as the table on p. 231, shows clearly that the smaller the number of trials given per day when several problems are learned abreast, the better the retention. The absolute loss during the 60 days non-practice period is greater when two problems are learned in conjunction with box II than when box II is learned alone. Again, when we compare the methods from the standpoint of days, we see that the situation is re- versed. The distribution curve (Fig. 47) shows again that within the limits of this experiment the larger the number of trials, the fewer the number of days required to learn.
Sensory habits. — A corresponding series of experiments has been made on sensory habits. Tests upon the dancing mouse indicate that for the white-black discrimination, the fewer the tests per day, within the limits of 2 and 100, the higher the efficiency of the method of training, as measured in terms of the total number of tests necessary for the establishment of a perfect habit, and the lower its efficiency as measured in terms of the number of series given. The table given below shows the total number of errors in each set of 10 trials for each of the four methods used.^^ No "^ Since sensory habits arise so very slowly, it can easily be seen that, in so far as the conservation of the experimenter's time is concerned, it has been the custom to effect a compromise by giving 10-20 trials per day. This is a purely practical matter and has no bearing upon the physiological problems involved in learning.
234 STUDY OF HABIT FORMATION TABLE NUMBEE OF EKRORS IN WHITE-BlACK SeEIES FOR DIFFERENT METHODS OF Training 2 or 5 tests 10 tests 20 tests 100 or more Sets per day per day per day tests per day Av. no.
Av. no.
Av. no.
Av. no.
tests errors errors errors errors 1 2 3 4 5 6 7 8 9 10 0 11 0 12 0 13 0 0 14 0 0 15 0 0 16 0 17 18 19 20 0 specific experiments have been made to test the efficiency of these methods upon retention, nor their effect when two or more sensory (or one sensory and one motor) habits are required simultaneously.
The delayed reaction. — Economy in learning the de- layed reaction has been tested only incidentally. The effect of giving 5 trials and 10 trials during the training period (i.e., establishing ordinary sensory habit) is shown in the following table taken from Hunter: TABLE 5 trials 10 trials Rat No. of trials Rat No. of trials on learning on learning It is clearly indicated here that the use of 5 trials favored rapid learning more than did the use of 10 trials.
EFFECT OF AGE ON LEARNING 235 No results are given for the effect on retention of the 5-trial method as over against the 10-trial method. It appears, however, that the number of trials given per day in the preliminary training does not affect the interval of delay.
VII. Other conditions which affect learning. — Some of the factors which must be considered with reference to learning are (1) the age of the animal, (2) possibly sex, (3) certain general bodily conditions — starvation, arrested growth, etc., (4) relative brain weight as compared with body weight, and (5) removal of parts of nervous system. Only a few of these have been considered with any degree of care; indeed, some have been left wholly untouched.
(i) Age. — From some experiments carried out by the author in 1902 he concluded that the young white rat forms motor habits, maze, box I, etc., much more rapidly than the adults. The young animals form the habit of manipu- lation much less rapidly than do the adults, the general conclusion being that any habit which requires for its learning excess running, climbing, etc., could be learned by any animal at the age of about 35 days more rapidly than by the adult. On the other hand, those problems which depend for their solution upon control of move- ment, can be learned by the adult more rapidly. This conclusion was probably not based upon a sufficiently large number of tests.
Since the above work was completed two other experi- ments bearing upon the same problem have been made. The first relates to sensory habits. In the dancing mouse Yerkes has established beyond question the fact that the dancer at 1 month of age can form a particular black-white habit in a much less number of trials than can an older individual. From the first to the seventh month there is a steadily marked increase in the capacity to form habits of this character. From the seventh to the tenth month there is retrogression in this capacity. Further- more, it would seem that when the difference in intensity between the stimuli is slight, the young animals respond to it more readily than do the adults. When, however, the 236 STUDY OF HABIT FORMATION difference is large, the old individuals respond as readily as the young (Weber law type of experiments).
The other experiment referred to is that of Slonaker upon the activity of the white rat at different ages. It would seem from this work that the most active period in the life of the rat is at the age of (approximately) 100 days. This was determined by allowing the animal to live in a cage which turned on a stationary axis each time the animal left the nest box to obtain water or food. It would thus seem that our early conclusions need confirmation. But it must be remembered that Yerkes worked upon a sensory habit and used punishment, whereas we worked upon a motor habit and without punishment. In regard to Slonaker 's work it may be stated that the period of greatest activity as obtained by his method need not at all coincide with the period of greatest activity of the kind involved in attacks upon problem boxes. ^^ (2) Sex. — No very sj^stematic studies have been made upon sex differences in learning. All through the litera- ture we find statements to the effect that possibly there was a sex difference, or that there was a slight difference, etc. No experiments which we consider crucial have been made which will enable us to state in what forms of be- havior sex differences appear. Personally, from long years of association with animal work, we are prepared to affirm our belief that there are differences everywhere between the males and the females, which may be seen clearly, as every one admits, during the periods of heat, in the female during pregnancy (or brooding, depending upon the phyla). In regard to the differences in the methods of attack upon problems there may be subtle differences which are very difficult to bring under experimental control. Yerkes has come out rather definitely in his statements con- cerning the sex differences in the dancing mouse. He states, on the basis of several experiments, that the young males 22 Miss Hubbert is making a thorough study of the effect of age upon the formation of maze habits in rats, using the maze with a camera lucida attachment, by means of which the movements may be recorded during the learning process.
LEARNING IN DEFECTIVE ANIMALS 237 acquire the habit of discrimination (black-white) more quickly than do the females, but that between the ages of 4 to 10 months the females acquire the habit more quickly; and that the female makes more mistakes early in training tests than the male, but that this condition gives place soon to greater accuracy of response on the part of the fe- male. This author remarks incidentally that the male dancer is somewhat more sensitive to punishment than is the female. Also, in his book on the dancing mouse, he states that the labyrinth habit (labyrinth C) was acquired by the female more quickly than by the male.
(3) Certain bodily conditions. — No definite experiments have yet been published upon the effects of starvation. Prom some unpublished experiments at the Wistar Institute of Anatomy it would appear that rats which had been stunted by underfeeding and then returned to a normal diet learned more rapidly than normal animals. ^^ Through the work of Osborne and Mendel it has been found possible to hold body growth in statu quo and then later to accelerate it at will. This method of controlling growth offers interesting possibilities in the study of habit formation.
(4) Learning in animals with less than normal brain weights. — Basset has carried out an extensive piece of work upon a strain of rats the brain weight of which was far below normal. This strain of rats (Mus noregicus albinus) was obtained from the Wistar Institute of Anatomy. The strain was strictly inbred for 6 or 7 generations. Through the effects of inbreeding, or some other cause (starting with a mutant in which the relative brain weight was small, etc.) not yet determined, this strain of rats had, relatively to the body length, a much smaller brain weight than the normal rat. The following curve (Fig. 48) shows the distribution of brain weights of the normal and inbred rats actually used in Basset 's work. It will be seen that the inbred curve shows the greatest frequency at.88^; the normal curve at -^ The experiments were made by Dr. J. W. Haves. See also the report of Langfeld. " Psychophysiological Tests During Prolonged Fast," Psijchological Bulletin, 1913, p. 83.
STUDY OF HABIT FORMATION .92^. The inbred distribution is from.70^ to.95^; that of the normals, from.84^ to 1.05^. The average relative brain weight with respect to body length of the 62 normal animals is.93351^; and that of the inbreds,.87335^, or 6.44^ less than that of the normals. Basset then tested 62 of the norfi rL Q Fig. 48. Distribution of Brain Weights in (Upper) Normal AND IN (Lower) Inbred Rats mal animals and 62 of the animals with the small brain weight (inbreds) with respect to the rapidity with which they acquire the maze habit and that of the inclined plane. It appears that the animals with the lesser brain weight re- quire on the average a larger number of days to learn these two problems. Furthermore, the time for relearning was far greater in the case of the inbreds than in that of the normals (see p. 246).
(5) Motor habits in animals with parts of the nervous system removed. — It has been shown (Franz) in both EFFECT OF PREVIOUS TRAINING 239 monkeys and cats that newly formed motor habits (prob- lem boxes) (1) are lost if a bilateral lesion in the frontal lobes is made. (2) Unilateral lesions produce only a slowing in the performance of the acts. (3) Habits once lost after the removal of the frontal lobes may be regained through training. (4) Habits of long standing are not lost when such lesions are made. It would seem from these experiments that the arcs involved in the acquisition of motor habits usually, but not necessarily, embrace the frontal lobes. From certain recent experiments we are led to believe that the decerebrate frog is incapable of forming habits. The use of operative technique in the study of sen- sory habits is discussed on p. 209 and its use in the study of localization of brain function (sensory projection cen- VIII. Effect of previous habits upon the formation of new habits: motor habits. — Neither in the study of human habits nor in that of animal habits have there been thor- ough systematic attempts to bring the mutual relations of habits under control nor to test the effect of previous train- ing upon the formation of new habits. There is an enormous field of work here which has just been touched upon. Re- search work would yield very rapid results and of such a kind as to be of very great importance in the study of human training methods. We are able to cite in the animal world the following particulars concerning the influence of previous habits upon the formation of new habits: the danc- ing mouse not previously trained on a simple labyrinth (C) makes the first correct trip on an average at the end of 19.7 trials. On the other hand, animals previously trained in another labyrinth (B) will make a successful trip in this labyrinth (C) on the 7th trial. Mastery of B by untrained animals requires 8.2 trials; by those previously trained on C (a rather difficult maze), 5 trials. (Yerkes: ^' Dancing Mouse.") Similar results have been obtained with the white rat. Fig. 49 shows the curve of untrained rats on box III and of trained (by previously learning boxes I and II) rats in learning the same problem. The differ- ence between the two curves is quite marked. Different STUDY OF HABIT FORMATION species seem to present characteristic differences in this re- spect (Richardson). In testing untrained pigeons on a given maze along with pigeons which had been previously Minij fes 1 j 1 ] 1 1 1 1 1 1 1 1 1 1 1 1 Trained Rats, four males and 1 1 three females.
1 1 - - - Untrained Rats, three males and one female.
1 1 Time of trials not shown: 1 Trained Rats Untrained Rats First trial 5.72 min 40.14 min.
Second trial 5.97 min.
Fourth trial 7.77 min.
1 fs 1 1 1 \ s \ 7 \ \ £ \ \ ^ y' \ \ / \ , , / \ s V '" \ y s c TRIALS Fig. 49. Showing the Effect of Previous Training in Acquiring New Habits ( From Richardson, Psych. Mons., Serial No. 48. ) trained on another maze, it was found that the untrained animals learned in as few trials as the trained. Further- more, the excess time and errors during the early trials are greater for the trained than for the untrained. Excess effortj however, is eliminated more rapidly (in a less num- EFFECT OF PREVIOUS TRAINING 241 ber of trials) in the case of the trained than the untrained. It would seem that the early habits acquired by the trained group persisted in the new work and interfered with the formation of another habit (Hunter).-* That interference of this kind may exist comes out still more clearly in work on the squirrels. It was found that squirrels trained to open box I and box II are at a disadvantage as compared with untrained animals when they come to learn box III. Boxes I and II require scraping, butting up of the latch with the nose, etc. These acts were carried over to box III; consequently the animals lost time in trying to open box III by persisting in movements organized to meet other conditions (Yoakum). It seems evident, although the above results do not wholly justify the view, that the type of results we shall get in such cases is dependent upon the similarity between the old habit and the habit w^e desire to instil; certainly also upon the differences in the use of receptors, — an animal in which position habits arise easily is very likely to carry these over into a new situa- tion, whereas an animal with keen vision will likely be attracted by the visual stimuli in the new situation, etc. It will be recalled that the problem under discussion is similar to that involved in the question of specific vs. gen- eral training. However, in the case of animals we shall not rest content in settling the matter by argumentation. We shall try to build up by experimentation a series of facts which will show the reciprocal effects of one motor habit apon another; or motor habits upon sensory, or vice versa; or sensory habits upon sensory, etc., and to analyze these facts in terms of the organization of reflex arc systems.
IX. Retention. — The term " memory " has been ill- advisedly used in the experiments upon the effect of inter- vals of disuse upon habit functions. The term " reten- ^* By interference nothing mysterious is meant. We mean nothing more than that old habits persisted in the new work. Consequently the group of random movements released by the new stimulus did not contain in it the successful movement. Not until the old habits had been exercised continuously but unavailingly, with resulting fatigue, did the stimulus call out the new group, which contained the needed movement.
242 STUDY OF HABIT FORMATION tion " has been employed in a static sense in this same connection, referring chiefly to the " persistence of modi- fications " in the nervous system.^° Both terms are ill- defined. It seems possible to keep the term retention and make its meaning more definite. In behavior the term re- tention covers this phenomenon; viz., that an object to which an animal has learned to respond in a definite way will for a more or less definite period in which the given response has been prevented (i.e., by not presenting the object) call forth in various degrees of perfection the old (or habitual) response. If the response is as definite at the end of the period of disuse as before we say that there has been no loss in retention or that retention was perfect. In most cases the response, after a period of disuse, is not perfect (i.e., there is excess effort). The effect of the period of disuse can be measured (in terms of time, distance, errors, etc.) by comparing the first trial after disuse with the last trial before disuse; or other- wise expressed, the last trial Z of regular training with the first trial a of retraining. If a certain length of time is overstepped the excess effort of trial a may be as great as that of A, the first trial in the training series. In this case the habit appears to be lost. Only complete retraining will tell us whether this is really the case. If the retraining series requires the same number of trials and the same amount of excess effort as the regular training series, the ^^ The neural picture, both during the formation of a habit and during the period of non-practice, has been interestingly presented by William James. Since his Principles appeared we have described habit by saying that the neural impulse aroused by the given stimulus in some way digs an ever-deeper trench for itself, and that it in some way modifies the conduction system (molecular rearrange- ment) over which it passes. Accordingly we would speak of path- ways in the nervous system and of their modifications. We would further assume that such modifications persist for definite intervals of time, possibly never wholly disappearing. So far as we know there is no longer any justification for assuming that a neural im- pulse, which is probably electrical, in any way modifies permanently the conductor over which it passes. While we are ignorant yet of what happens at the surfaces of separation of the conduction systems as regards permeability and the reverse, it seems quite unlikely that neuro-physiology will ever discover any structural modifications in the conductors themselves.
RETENTION 243 evidence is clear that the habit was really lost. Behavior has not fixed upon definite methods of making retention tests. Shall we compare only the last trial of the train- ing series with the first trial of retraining? Or shall we average the last three trials of training with the first three trials of retraining? Or shall we carry out a wholly new retraining series? In retraining this rather interesting question comes up. It often happens that in the training series we use different methods for the different groups, as we saw in section VII, where the number of trials given per day differed in the different groups. Now, in the retraining series shall we use a common method for all the animals, or shall we retrain with a method identi- cal with that of training? Furthermore, during the period of disuse some systematic method of caring for the animals should be adopted. In most retention tests we read that at the end of the training series the animals were put away until the time for the retention test. No effort is ever made to keep the animals in the " problem solving condition." It is quite obvious that retention tests made under such uncontrolled conditions are worthless from the standpoint of testing pure retentiveness and for purposes of comparison. In the tests which Ulrich and Basset made in the Hopkins Laboratory they adopted the following technique: A long run-way was constructed at the end of which food was placed. At the close of the training period the animal was forced each day at feeding time to travel to the end of the run-way in order to get food. By this method the ani- mal was kept in condition. The amount of food taken was also controlled, i.e., kept constant as in the training period (the animal eating for a certain definite time). ' Fur- thermore, the animal was kept from becoming wild and from putting on fat. We suggest the following as a pos- sible standard of procedure. Some standard of accuracy of mastery must be adopted. This will necessarily differ even upon the same problem with the different species of animal used. This standard, where possible, should be based upon more than one criterion. Time is the best single criterion in motor habits. The time allowed for a trial 244 STUDY OF HABIT FORMATION must be determined previously by averaging the results of at least 30 animals which have learned the problem just to the point where no errors appear and no excess distance is traversed. (Such records are gradually collected on all pieces of apparatus in general use in the laboratory.) Distance traversed, where it can be measured accurately (as in the maze, p. 100), is probably the next best criterion. Any distance greater than the measured shortest route is " excess." Mastery of the problem regardless of the time cannot be said to have been attained until there is no excess distance. Freedom from error for 30 consecutive trials has been Yerkes' criterion of mastery for sensory habits. This appears to work well enough. After the animal has mastered the problem it should be worked with and fed each day by forcing it to go for food to the end of a long run-way. It should be allowed to eat the full amount of food that the experimenter knows to be necessary in order to maintain metabolism. The animal should have been made familiar with this method of obtaining food before even the training period is begun. Furthermore, since neither in w^ork upon motor habits, nor in that on sensory habits are the animals usually allowed to eat their full quota of food in the experimental room, they should be fed the remainder in this run-way. At the end of the period of disuse retraining should begin. The fairest way seems to be to retrain all animals by the same method re- gardless of the training method. Furthermore, we suggest that one trial per day only be given in motor habits and that 10 trials per day be given in sensory habits. Retraining should be carried to the point of original mastery. The last trial of training should be separately compared with the first trial of retraining. The whole of the retraining series should then be averaged by successive trials and plotted (i.e., without telescoping and then averaging). It must be remembered that only the first trial appears " pure." Practice effects appear immediately.
In regard to the concrete data upon retention it must be confessed that there is not a single research in the whole field which throws particular emphasis upon retention.
RETENTION What results we have have been made incidentally in con- nection with studies on habit formation. Since no approach to uniformity in the method of making tests on retention has been observed by the different experimenters, we can do little more than summarize their findings.
Mammals. — Yerkes shows that in the dancer a perfectly acquired black-white habit will endure for a period of 2 V 6 6 4 S > • \^ f * \ \ \ \ \ s -^ V \ \ N \, s 1 \ N \ > V \ -- \, s Fig. 50. Error Curves Plotted from the Data Given by 10 Dancers in White-Black Discrimination Tests The solid line ( ) is the error curve of the original learning process; the broken line (----) is that of the relearning process, after an interval of 8 weeks. (After- Yerkes, Tlie Dancing Mouse, Macmillan, New York, 1907, p. 257.)
The abscissa shows the number of series of trials, each series con- sisting of 10 trials; the ordinate the average number of errors made in each series.
weeks at least. The results for 4-week intervals show ex- treme individual differences in retention. In the case of 2 animals retention tests were made after 3 different rest intervals. At the end of 2 weeks the habit was present in both individuals. After 6 and 4 weeks respectively it still persisted and was apparently improved as the result of ad- ditional training received at the end of the 2-weeks period. At the expiration of 10 weeks it had apparently disappeared. Fig. 50 shows the curve both of training and retraining. It would appear from this curve that the mice reacquired the white-blacl^ discrimination habit much more readily 246 STUDY OF HABIT FORMATION