Lashley, Karl Spencer Studies of cerebral function in learning IS 5 L3-? V'9 Reprinted from The Journal of Comparative Neurc logy Vol. 57, No. 1, February, 1933 DEPARTMENT ( UNIVERSITY OF 13*OtJT& STUDIES OF CEREBRAL FUNCTION IN LEARNING1 IX. MASS ACTION IN" RELATION TO THE NUMBER OF ELEMENTS IN THE PROBLEM TO BE LEARNED K. S. LASHLEY AND L. E. WILEY Department of Psychology, The University of Chicago THREE TEXT FIGURES AND FIVE PLATES INTRODUCTION Measurements of the influence of extent of cerebral lesion upon efficiency in various functions have given quite different results, according to the functions studied. In some cases there is clearly an all-or-nothing relation between some func- tional area and the capacity for performance; in others, a close relationship between surface extent of injury and degree of lowering of efficiency. The data thus far accumulated in quantitative studies are summarized in table 1. The constants given are based on error scores, where these are available, otherwise on trials for learning or relearning. The table shows a definite bimodal distribution of the constants. Six fall below 0.10, thirteen are above 0.50, and only three fall between these limits. Of these three (double platform box, difference threshold for two lights, and learning of a 1 cul de sac maze) the first reduces to zero when corrected for motor disorders, the second is approximately 0.50, and the third is based on a maze which is known on other grounds to be an unreliable measure of performance. 1 This work was supported by a grant from the Otho S. A. Sprague Memorial Institute. We are indebted to Prof. L. L. Thurstone and to Mrs. Annette M. Wiley for advice and assistance in the statistical treatment of the data and to Dr. Margaret Frank for assistance in the training of animals and the preparation of material for histological study. 4 K. S. LASHLEY AND L. E. WILEY The bimodality of distribution of these constants is good evidence that the positive correlations are not merely the TABLE 1 The relation between extent of lesion and efficiency of performance in previous studies of the effects of cerebral injury. The constants are for error scores where these are available, otherwise for trials TASK locus of LESION COEFFICIENT OF CORRELATION REFERENCE Simple maze, P. E. Frontal o.oo1 Lashley and Franz ('17) Lashley ( '20) Double platform box, L. Frontal 0.24 ± 0.15 Double platform box, L., corrected for Frontal 0.001 Lashley ( '20) motor disorder Delayed alternation, L. Frontal -0.02 ± 0.19 Loucks ('31) Delayed alternation, P. E. Frontal 0.54 ± 0.12 Loucks ('31) Maze habit, 8 culs de sac, L. Frontal 0.64 ± 0.08 Maier ( '32 a) Eeasoning Frontal 0.54 ± 0.09 Maier ( '32 a) Light-darkness discrimination, L. Visual 0.08 ± 0.14 Lashley ( '26) Light-darkness discrimination, P. E. Visual 0.72 ± 0.05 Lashley ('26) Light-darkness discrimination, P. E., Visual 0.73 ± 0.08 Lashley ('32) corrected for critical area Light-darkness discrimination, P. E. Visual 0.64 ± 0.10 Lashley ('30) Discrimination two lights, L. Visual 0.58 ± 0.10 Lashley ('30) Discrimination two lights, P. E. Visual 0.65 ± 0.10 Lashley ('30) Discrimination two lights, difference Visual 0.49 ± 0.09 Lashley ( '31 b) threshold Visual acuity and pattern vision Visual o.oo1 Lashley ( '31) Eeasoning Visual 0.75 ± 0.05 Maier ( '32 b) Discrimination, noise, P. E. Auditory 0.61 ± 0.11 Wiley ('32) Maze habit, 8 culs de sac, L. All parts 0.86 ± 0.03 Lashley ( '29) Maze habit, 3 culs de sac, L. All parts 0.65 ± 0.07 Lashley ( '29) Maze habit, 1 cul de sac, L. All parts 0.30 ± 0.16 Lashley ( '29) Maze habit, 8 culs de sac, P. E. All parts 0.51 ± 0.11 i Lashley ( '29) Maze habit, 1 cul de sac, P. B. All parts 0.00 ± 0.08 Lashley ( '29) L. = initial learning, P. E. = postoperative retention. 1 Correlations not computed, but no suggestion of a correlation from inspection of the data. result of some error in sampling, in which case they would show a normal distribution around zero, but express a genuine relation between the variables studied. All of the constants, however, have been determined from relatively small numbers STUDIES OF CEREBRAL FUNCTION. IX 0 of cases, and, although they do attest the importance of the mass relationship, they give little further information con- cerning it. Several purely statistical questions concerning the relationship remain to be answered before we can make -much progress in interpreting the functional significance of the correlations. These questions call for a larger mass of data than has been hitherto available and the primary aim of the present study is to obtain a large series of cases, which can be analyzed with some assurance of reliability. The continuity of the mass relationship The available evidence is not conclusive as to whether there is a continuous progression in the effects of cerebral lesions from the least to the greatest or whether there may be a critical amount of destruction below which injuries are rela- tively ineffective in producing deterioration. Several studies (Lashley, '26, '29, table 11; Maier, '32 a and b) have given indication of a sharp increase in the effectiveness of lesions at about 15 to 20 per cent destruction. This may be evidence for lower limit of extent of lesion necessary to produce signifi- cant symptoms, but the appearance of a sharp rise in effective- ness might also result, if the relation between extent of lesion and deterioration had a logarithmic or other accelerated form. Lashley ('26) found that the correlation ratio gave a higher value (0.84) for the relation between lesion and amnesia than did the correlation coefficient (0.72), indicating that the rela- tionship is curvilinear, but with the small number of cases the difference between these constants was not statistically reliable. Thurstone ('33) has analyzed the data for maze III of Lashley 's study ('29) and finds that the rate of learn- ing is for this maze a function of the sixth power of the intact cortex. Maier ( '32 a, b) has reported a very pro- nounced drop in efficiency in reasoning tests with lesions exceeding 18 to 20 per cent of the cortex. None of these studies provides sufficient material to deter- mine the reliability of the form of the function. Our first problem, therefore, has been to collect data on maze learning b K. S. LASHLEY AND L. E. WILEY after cerebral lesion for a larger number of cases, as a basis for a more reliable determination of the continuity and form of the relation between extent of lesion and efficiency in learning. The equivalence of different cortical areas for maze learning It seems quite clear from the results of Cameron ('28), Lashley ( '29), Maier ( '32), and Jacobsen2 that injuries in any part of the cortex result in some retardation in the rate of learning the maze. The relative effects of injuries in dif- ferent cytoarchitectural areas remain uncertain. Lashley ('29) attempted to measure the influence of injuries in dif- ferent areas by comparing the numbers of errors made by animals with lesions in different parts of the cortex, and concluded that, within the limits of accuracy of the experi- ment, the same amount of destruction within any area pro- duced the same amount of retardation in learning. The number of cases in his experiment was small, however, and the average deviation of 13 per cent from equality between the groups may have been really significant. Moreover, his method of grouping cases resulted in a considerable overlap of the areas compared and this may have tended to equalize the averages of the arbitrary groupings. We have attempted to deal more adequately with this problem of the relative effects of lesions in different areas. The relative effects of lesions symmetrical and asymmetrical in the two hemispheres All of the studies dealing with quantitative effects of cere- bral lesions have dealt with lesions which were made as nearly as possible symmetrical on the two hemispheres. Data on the critical areas for pattern vision (Lashley, '31) and unpublished data on maze retention after removal of one hemisphere indicate that somewhat different results are to be anticipated when the lesions are markedly asymmetrical. 'Unpublished experiments on reversal of training. STUDIES OF CEREBRAL FUNCTION. IX 7 Since there is never an exact duplication of the fields in- volved in the two hemispheres, it is desirable to determine whether this lack of symmetry affects the correlations between extent of lesion and learning records. Organic dementia in relation to the complexity of the task to be performed Clinical studies of organic dementia frequently suggest that a nearly normal ability in the execution of simple acts may accompany a marked deterioration in the performance of more difficult tasks. Comparing the rate of learning for three mazes of 1, 3, and 8 culs de sac by partially decerebrate rats, Lashley ('29) found that whereas the ratios of difficulty of the mazes for normal animals were as 1 to 2.2 to 6.5, for animals with cerebral lesions the ratios were as 1 to 3.5 to 20.6. This seemed to conform to the clinical evidence in show- ing that a given amount of cerebral destruction resulted in a much greater retardation in the performance of complex than of simple tasks. There was some indication also that large lesions produced a disproportionately greater retarda- tion in the more difficult mazes than in the simpler ones, al- though the evidence for this was not statistically valid (Lashley, '29, table 11). The meaning of these data is by no means clear. If we attempt to define difficulty merely in terms of the number of trials required for learning, then certain tasks, the learn- ing of which is unaffected by cerebral lesions, appear more difficult than maze learning. The difficulty of a task may de- pend upon the number of similar elements which must be integrated in order to give an efficient performance. In logi- cal processes the ability to keep in mind a number of separate elements and at the same time manipulate them in thought is essential, and, according to Boumann and Grimbaum ('25), this capacity primarily suffers in organic dementia. In learning relatively meaningless material the difficulty, as measured by practice necessary to secure perfect reproduc- tion, increases as the 3/2 power of the length of the series 8 K. S. LASHLEY AND L. E. WILEY (Thur stone, '30). In such situations the difficulty of the task corresponds to its objective complexity and may be stated quantitatively in terms of the number of items included in the test series. In addition to this quantitative aspect of ' difficulty' there are unquestionably qualitative elements. If the situation is unfamiliar, if the stimuli are nearly indistinguishable, if the integrations required are foreign to the native organization of the animal, or if the relations to be recognized are obscure, the difficulty of the task is increased. The maze studies reported by Lashley ( '29) did not distinguish clearly between number of similar elements and qualitative diversities. The mazes differed in number of culs de sac, but also presented fundamental differences in general plan, so that it is impos- sible to determine which of the possible sources of difficulty was predominant. In the present study we have attempted to devise a series of mazes differing only in the number of similar elements which must be learned and to test their relative difficulty for normal and operated animals. Problems We may summarize the chief problems with which the present study is concerned in the following questions : 1. Is there a significant correlation between extent of cere- bral lesion and degree of retardation in maze learning when tested with an extensive series of cases? Confirmation of this leads to the further problems : 2. Is retardation a continuous function of the size of the lesion, or is there a critical amount of injury below which little effect can be observed? 3. Are the effects of equal amounts of injury in different cytoarchitectural fields the same for the learning of enclosed mazes or are there characteristic differences in the degree of retardation resulting from injury within different loci? 4. With tasks differing in complexity (number of objectively similar elements), what is the relation of the performance STUDIES OF CEREBRAL FUNCTION. IX 9 of animals with cerebral lesions to the relative difficulty of the tasks for normal animals ? PROGRAM OF EXPERIMENTS Since a primary object of our work was to test the influence of extent of lesion upon the rate of learning tasks of different complexity we planned to train animals with equal lesions in four mazes with 4, 8, 12, and 16 culs de sac. In such ex- periments on learning with animals we are confronted with two variables which cannot be controlled simultaneously; chance individual variations in capacity and transfer of train- ing. Thus, if we wish to compare learning records on two different problems we may either train the same animals on both problems, in which case we control individual varia- tion but ignore transfer, or we may compare the records of different groups of animals, each trained on a single problem, in which case transfer effects are eliminated but individual differences are uncontrolled except statistically. In the first case transfer effects may be controlled by subdividing the subjects and reversing the order of training with a part, but where four problems are to be compared the arrangement of tests for transfer becomes impossibly cumbersome. We de- cided, therefore, to use separate groups of animals, one with each of the four mazes, attempting to get comparable groups with respect to extent of lesion and to control individual dif- ferences to some extent by subsequent training of all groups on the same task (maze V). The validity of this control is somewhat lessened by the possibility of differential transfer from the different mazes and the evidence for equality of the groups is not as clear as we could wish, but the differences are small in comparison with other differences revealed by the experiments. For comparison with the operated cases and for standardi- zation of the mazes we have trained groups of normal animals under parallel conditions with the operated ones. The vari- ous groups and the number of animals included in each are shown in figure 1, together with the ground plans of the 10 K. S. LASHLEY AND L. E. WILEY different mazes used. The groups are designated Normal and Operated I, II, III, and IV, corresponding to the number of the maze in which the animals of the group were first trained. St II II II NI5 Op 30 NI5 - 0P 32 ^ V II NI5 0P3I IV S NI5 On 34 / N60 0P 125 Fig. 1 General plan of the experiments. The four comparison mazes (actually constructed by subdividing the largest) are shown at the left, the control maze at the right. The number of animals trained in each maze is indicated. F, food compartment; S, starting compartment; N, normal group; Op., operated group. METHODS Mazes For comparative tests, mazes of the general plan of Lashley's maze III ('29) were used. This plan has the ad- vantage that the number of blind alleys can be increased in- definitely without introducing any fundamental change in the general plan. It has the possible disadvantage that the cor- rect path is one of simple alternation, capable of easy STUDIES OF CEREBRAL FUNCTION. IX 11 generalization. If the animals made such generalizations, mazes of this plan should all be of about equal difficulty, whether they have few or many culs de sac. We believe, however, that this objection is not serious since the forma- tion of the alternation habit seems to require many more trials than are required for complete mastery of our longest maze (Carr, T7; Hunter, '20; Loucks, '31). A maze with 16 culs de sac was so constructed that seg- ments could be closed off to give mazes with 4, 8, or 12 culs de sac. The ground plans of the resultant mazes are shown in figure 1. The maze was constructed of f-inch pine with 4-inch partitions between the alleys. The top was covered with fine wire mesh. Electrical contacts on counterbalanced sections in the floor permitted automatic recording of errors. Four separate starting boxes gave access to the segments of the maze and were closed off: by one-way doors of sheet metal. In later discussion the four mazes which can be arranged by blocking or opening the doorways between the culs de sac will be referred to as mazes I, II, III, and IV, as shown in figure 1. As a control of the training records with these four mazes, a fifth having 8 culs de sac was used. The ground plan of this maze is also shown in figure 1. Its construction was similar to that of the other, except that the order of alterna- tion of turns was reversed and the relative position of food box and starting box altered. Its orientation in the room during use was also different from that of the other mazes, as indicated in the figure. In later discussion this will be called maze V. Training methods Before controlled training in the mazes was begun, the animals were given preliminary training in traversing a straight runway, 10 feet in length, until they came through promptly and were not disturbed by handling. They were then fed for 3 days in succession only in the food box of the maze, without access to the alleys. In training a ' trial ' was 12 K. S. LASHLEY AND L. E. WILEY counted as a complete trip from starting box to food box.3 One trial was given on the first day and five trials per day thereafter. After each trial the animal was allowed to eat a bite of food. At the termination of each day's training it was fed to repletion. A rigid control of the incentive cannot be employed with operated animals, since those with larger lesions require con- stant care and special feeding to keep them in good condi- tion. We must therefore recognize the possibility of unequal motivation in different animals. The only test of the ex- istence of such differences that we have is the apparent eager- ness of the animals for food, and on this basis the operated animals must be judged more strongly motivated than normals. There is significant evidence from other sources that the differences between operated and normal animals in learning cannot be ascribed to differences in motivation. In two types of experiment (light-darkness discrimination, Lashley, '29, and delayed alternation, Loucks, '31) the in- feriority of operated animals has appeared only in post- operative relearning, although the motivation used was always the same. In studies employing the double platform (Lashley, '20) the same incentive (hunger) was used as in the maze studies with no evidence of inferiority on the part of the operated animals. Criteria of learning With mazes I to IV training, after one trial on the first day, was continued with 5 trials per day until the animal made a record of 10 consecutive errorless runs, or for 150 trials (100 trials with maze V), in case the criterion of 10 errorless trials was not reached earlier. Time and errors per trial and total trials to reach the criterion were recorded. Of these, errors are probably the most reliable criterion of progress in maze •It is not always possible to obtain a complete trial in one day because of limitations of the experimenter's time. In such cases the animals were removed from the maze, fed a limited amount, and returned to the maze on the following day, the accumulated time and errors being counted as a single trial. STUDIES OF CEREBRAL FUNCTION. IX 13 learning, and constants computed from the error records are emphasized in our treatment of the data. Constants com- puted from time and trials are also included as contributory, but less reliable evidence. The early trials, especially the first, are scarcely compar- able with later trials in the maze after the preliminary explor- atory period is passed. Lashley ( '17) has shown that in early trials there is a strong tendency to explore all parts of the maze. Various means have been suggested to correct the learning records for exploratory errors in early trials (Thur- stone, '33). We have not yet sufficient data for evaluation of the methods or to judge which of the errors are signifi- cant for learning. In the first trial the animal has had no opportunity to associate the maze with food and hence the motivation in this trial differs from that in all later trials. For this reason we have omitted the records of the first trial in the computation of all constants reported. Among the animals with cerebral lesions many failed to reach the criterion within 150 trials. This affects the error scores very little; inspection of the records reveals that 80 per cent of the total errors are made in the first 50 trials, and that by 150 trials the animals have settled down to rather stereotyped runs with at most 2 or 3 errors per trial. In computing correlations based on trials for learning, we have used total trials minus the number of errorless runs made before the termination of training, to avoid throwing all of these cases into one rank. None of the conclusions which we have drawn is dependent upon any of these special methods of grouping the data. That is, the group differences and correlations are essentially the same, within the limits' of reliability which we have required, whether we use total trials, trials less errorless runs, total errors, or total errors less errors in the first trial, total time, or total time less time for the first trial. In view of the large number of animals which failed to reach the criterion of learning within the training given, some method of extrapolating the learning curves and predicting 14 K. S. LASHLEY AND L. E. WILEY ultimate achievement might give more valid results than the mere averaging of the crude data. Uncertainty as to the legitimacy of the available methods, however, together with the enormous labor involved in such computations has led us to employ the crude scores. Animals which failed to run A few of the operated cases failed to reach the food com- partment after 12 to 18 hours in the maze and became inactive in the maze situation. Since such records of failure cannot be interpreted in terms of capacity to learn, these cases were simply discarded. Such behavior appeared most frequently TABLE 2 Comparison of the four groups of operated animals with respect to extent of lesion as a test of the selective effect of excluding records of animals which failed to get through the maze GROUP AVERAGE PERCENTAGE OP LESION a UPPER LIMIT OP RANGE NUMBER WITH MORE THAN 40 PER CENT DESTRUCTION I II III IV 20.7 25.1 25.7 24.6 13.4 13.8 14.1 12.4 65.3 56.4 60.2 49.7 2 4 6 5 among animals with extensive lesions and more often in the longer than in the shorter mazes, so that there was probably some selection exerted in this way. Table 2 summarizes the distribution of lesions in the four groups compared. The average destruction in the groups is essentially the same, the variation within the groups is not greatly different. The upper range indicates possible selection only for maze IV. It is doubtful, therefore, whether the failure of cases to get through the maze has had any appreciable effect upon the results. "We have attempted to control this possible selection by computing constants for animals with smaller lesions only and comparing these with similar constants for the entire group. STUDIES OF CEREBRAL FUNCTION. IX 15 Surgical and anatomical methods The methods of destroying the cerebral cortex were those previously described by Lashley ('29), with operation in two stages for the more extensive lesions. Since we wished to obtain four groups of animals with similar lesions, we oper- ated on not less than four animals at one time, attempting to duplicate the lesions in all and distributing them later to the four groups. At the time of operation a sketch was made of the type of lesion, and in case a member of the group died, it was replaced by another with duplicate operation. The method of reconstruction of the lesions was that de- scribed earlier: graphic reconstruction of serial sections (Lashley, '29 ).4 We have paid especial attention to the sub- cortical lesions, which cannot be avoided with larger destruc- tions of the cortex. Analyses of the data to determine the influence of subcortical injuries upon the results are reported on page 32. Graphic and statistical analysis The diagrams prepared in reconstruction of the serial sec- tions represent approximately the surface distribution of the *Loucks ('32) has recently advocated a method which differs from the above in three particulars: the use of a fiber stain, the arbitrary limitation of the boundary of the lesion at the point where total destruction of tissue cuts the pyramidal layer of the cortex, and the measurement of the lesion along the perimeter of each section, instead of surface area of the reconstructed diagram. These differences in method do not seem to us advantageous. Although in old lesions there is usually a clean-cut destruction of tissue, the cortex bounding the area of completed destruction often shows pathological changes which cer- tainly render it non-functional. Fiber stains do not reveal this and the arbitrary criterion of complete destruction disregards it. In determining the extent of lesion we have drawn the boundaries at the point where the cortex assumes a normal appearance, providing against personal bias by having the observer in ignorance of the animal's training record when the reconstruction of the lesion is made. The method of measuring the lesion along the perimeter of the section is doubtless somewhat more accurate than the determination from the graphic recon- struction. Some years ago, the senior author made a number of determinations by both methods. The difference in results by the two methods was about 5 per cent, which was within the limits of accuracy in remeasurement by either method. 16 K. S. LASHLEY AND L. E. WILEY lesions, as determined in relation to internal landmarks. The distribution of cytoarchitectural fields is variable and in a large series of cases it is quite impossible to study the cyto- architecture of the cortex in sufficient detail to determine the limits of the remaining fields in each case. At best we can only compare the diagrams of the lesions with the somewhat conventionalized diagram of cytoarchitectural fields adapted from Fortuyn's studies ('14). The classification of cases by areas destroyed has been made by superimposing a trans- parent diagram of the cytoarchitectural fields upon the dia- grams of lesions and measuring the area of the lesion within each field with a planimeter. These measurements were then expressed as percentage of the total neocortex and used as a basis for estimation of the effects of injury to different fields. The question of the relative effectiveness of lesions re- stricted to one hemisphere and of lesions of equal magnitude distributed symmetrically on both hemispheres in reducing learning ability has arisen continually in experimental work of this sort. It has been difficult to test the question by direct experiment because of the impossibility of distinguishing with certainty in maze studies between genuine reduction of learning capacity and possible disturbances of orientation produced by asymmetrical motor defects which frequently result from unilateral lesions. The recent study by Loucks ('32) of habits involving alternation of turns to right and left, where the turns were recorded separately has shown that even a strong motor tendency to rotation does not affect the rate of formation of the alternation habit, so that with a series of bilateral lesions like the present one, we may proceed with assurance upon the assumption that bilateral asymmetries have not influenced the training records through their motor effects. We have therefore attempted to determine whether or not the bilateral destruction of corresponding areas is more effective in retarding learning than unilateral destruction. To this end the lesion in one hemisphere was traced on trans- parent paper which was then inverted and superimposed upon STUDIES OF CEKEBKAL FUNCTION. IX 17 the other hemisphere. The area of overlap between the two lesions was then outlined, measured, and expressed as per- centage of the total area of the neocortex. These percentages have been used as a basis for computing the constants for comparison with those obtained by consideration of the total extent of lesion. In later discussions the measurements ob- tained in this way are referred to as 'lesions common to both hemispheres. ' In computing correlations we have used the method of rank order,5 in preference to the Pearson r, since our data do not follow a random distribution. All methods of measuring the degree of association between variables are based upon as- sumptions concerning random sampling which are not ful- filled by our data on brain lesions, and the use of correlation methods in such cases can be justified only as a crude method of expressing the presence or absence of a significant associa- tion. The relation between extent of lesion and retardation is probably not rectilinear, so that the correlation ratio would be a more suitable measure of the association. In most cases it would give a somewhat higher figure than the correlation coefficients reported, but in the present state of our knowledge slight differences in the magnitude of the coefficients have no significance. We are dealing with differences between indi- viduals and groups which are many times greater than the range of normal variation, so that refinement of statistical treatment is of relatively less importance than if we were trying to measure smaller differences. Special controls There is little doubt that the experimenter may influence the maze records of his animals by slight differences in pro- cedure of which he is scarcely aware. Somewhat more gentle handling of one than of another animal, deviations in the al- lowance of food, and personal variations in the criteria of errors may influence the data and are likely to do so in a 5p = 1 -~-d~,) P-E.p =l~f 0.7063. n(nM)' p Vn THE JOURNAL OV COMPARATIVE NEUROLOGY, VOL. 57, NO. 1 18 K. S. LASHLEY AND L. E. WILEY constant direction where the experimenter has definite pre- conceptions. We have tried to control such influences of the personal equation as follows. The recording of errors was automatic, leaving no room for personal judgment. The training was done, as far as possible, in ignorance of the character of the operation to which the animal had been subjected, although such ignorance can be only partial, since it is impossible to mistake an animal with extensive cerebral lesion. In all cases the lesions were reconstructed and the re- constructions checked without knowledge of the experimental records of the animals. TABLE 3 Correlations between the scores of the same animals in learning two mazes NUMBER CASES TOTAL ERRORS LESS FIRST TRIAL TOTAL TRIALS TOTAL TRIALS LESS CORRECT RUNS TOTAL TIME LESS FIRST RUN Normal animals I with V 15 —0.18 :t 0.18 0.09 ± 0.18 0.18 ± 0.18 0.50 ± 0.14 II with V 15 0.20 ± 0.18 0.18 ± 0.18 —0.30 ± 0.17 0.46 ± 0.14 III with V 15 0.22 ± 0.17 0.29 ± 0.17 0.44 ± 0.15 0.38 ± 0.16 IV with V 15 0.49 ± 0.14 0.06 ± 0.18 0.17 ± 0.18 0.42 ±0.15 Operated animals I with V 30 0.84 ± 0.04 0.78 ± 0.05 0.89 ± 0.03 0.72 ± 0.06 II with V 31 0.91 ± 0.02 0.78 ± 0.05 0.84 ± 0.04 0.63 ± 0.08 III with V 30 0.92 ± 0.02 0.84 jfc 0.04 0.90 ± 0.03 0.54 ± 0.09 IV with V 34 0.88 ± 0.02 0.80 ± 0.04 0.84 ± 0.04 0.69 ± 0.06 VALIDITY OF THE MAZE TECHNIQUE The consistency of performance of the animals and the validity of the methods can best be tested by a comparison of the scores made in different mazes. The scores by the various criteria for the same animals on the two mazes in which each was trained have been correlated and the results are shown in table 3. Except for time, the correlations obtained from normal animals are low. There is a tendency for them to increase in magnitude with increase in the number of culs de sac in the mazes, which suggests that we increase the reliability of maze studies by increasing the complexity STUDIES OF CEREBRAL FUNCTION. IX 19 of the mazes, but in general they indicate that our mazes do not provide a trustworthy index of individual differences among normal animals. In contrast to this, the correlations for the operated cases are uniformly high and significantly greater than their prob- able errors. The most consistent results are obtained from errors and from total trials less errorless runs, with cor- relation coefficients ranging from 0.84 to 0.91. They indicate that there is some common factor involved in the learning of all five mazes by operated animals which is reliably measured by the criteria which we have adopted. The nature of this factor is not clearly indicated. The lower correlations for time show that it is not mere activity or speed of running. Whether it is motivation, ekphorie, or some sort of insight into the problem is not revealed by these figures. The previous results of Lashley ('29) for learning in the double platform box and in the maze show that there is no correlation in the learning of these two problems, al- though the same incentives are used, so differences in motiva- tion seem improbable as a cause of the correlations. This throws us back upon some mechanism directly in- volved in the learning process itself as the function measured in our study. The divergent results with the double platform box (Lashley, '20), the learning of which was unaffected by any cortical lesion, indicate that mere fixation in memory or ekphorie is not the factor involved. Our knowledge of the actual factors responsible for maze learning, such as the in- fluence of thwarting in blind alleys, the formation of associa- tions with specific cues in the maze, maintenance of the sense of direction, symbolization of the maze pattern and the like, is too slight to justify any further conclusions concerning the nature of the function which is being measured. Lashley ('29) has attempted to relate it to general intelligence, but such speculation can be'justined only by showing a high cor- relation of maze learning with tests known to involve some general capacity which can be termed intelligence by common consent. All that we can justifiably conclude is that our mazes 20 K. S. LASHLEY AND L. E. WILEY reliably measure some function which is common to the learn- ing of different mazes. ANALYSIS OF EXPEEIMENTAL DATA Deterioration after cerebral lesion The training records of all operated cases are summarized in tables 4, 5, 6, and 7 and the details of the lesions are shown in plates 1 to 5. The numbers and arrangement of the figures in the plates correspond to the experimental numbers of the animals in the tables, for ready reference. Similar data for normal animals trained under parallel conditions are given in tables 8, 9, 10, and 11. The mean scores for the various criteria, with probable errors of the means and standard deviations of the distribu- tions, are summarized in table 12. With average destructions of 20 to 25 per cent, the operated animals require from 2 to 17 times as much practice to reach the criterion of learning as do normals. The greatest differences are in the numbers of errors, the least in the numbers of trials. Since training was discontinued after 150 trials, the average of trials for the operated cases does not express the actual retardation. Fifty-one of the operated animals failed to reach the criterion in mazes I to IV, and 47 in maze V, and the scores for trials would have been very much higher, if these animals had been trained to errorless running. The error scores, therefore, probably represent most truly the difference between the normal and operated groups. Judged by this criterion, the latter require from 7 to 17 times as much practice as the former. A comparison of table 12 with table 15 reveals that there is retardation, even for the smallest amounts of injury. Animals with lesions of less than 10 per cent of the neocortex required 153 per cent as much practice, measured in terms of errors, as did normals. All these differences between normal and operated groups are statistically reliable. They demonstrate that cerebral lesions produce a significant reduction in the capacity for maze performance, even when the lesions are quite small. STUDIES OF CEREBRAL FUNCTION. IX 21 The relation between extent of lesion and rate of learning The coefficients of correlation between extent of lesion and maze learning for four criteria with each of the mazes are given in tables 13 and 14. For all mazes by each criterion TABLE 4 Learning scores for operated group I on mazes I and V. The score for time omits the record of the first trial. Scores for all errors and for all except those of the first trial are given PER CENT DESTRUC- TION MAZE I MAZE V NO. .5 m£ fl Eh ■ E o E E B 1 B oo O o ^r- §2? Eh 00 M I !g5g Eh E S3 "2 I E as O U 09 ** ■8-5 u Eh 00 "3 I 1 3.0 561 21 18 11 471 61 27 16 2 3.0 681 42 35 46 223 33 25 36 3 3.3 229 34 23 31 956 141 106 41 4 4.9 981 36 35 11 2,284 109 97 16 5 5.7 682 51 46 61 524 80 50 31 6 9.8 850 69 60 31 1,594 220 165 61 7 9.9 168 17 13 11 754 104 61 91 8 10.9 246 51 31 26 216 39 33 26 9 13.4 793 44 37 21 1,335 146 129 36 10 14.1 394 41 40 46 351 108 71 36 11 14.3 683 48 44 66 4,351 368 275 69 12 15.2 3,599 579 561 150 1,810 787 549 100 13 17.9 3,447 142 130 66 1,991 258 253 100 14 18.4 2,974 116 100 66 1,176 156 108 56 15 19.1 498 109 88 56 314 95 64 41 16 19.3 532 37 33 41 256 82 72 31 17 19.6 8,884 62 57 63 479 72 30 31 18 20.7 1,122 282 280 150 2,195 689 648 100 19 21.0 1,431 85 61 54 534 75 70 41 20 21.9 12,373 263 249 150 2,343 420 324 100 31 24.6 2,598 396 367 150 1,307 362 353 100 22 25.2 3,077 41 38 31 10,417 295 272 100 23 28.3 3,901 591 566 150 5,387 899 831 100 24 31.4 11,104 272 258 84 5,735 533 478 100 25 31.9 2,580 153 147 107 2,055 247 220 86 26 34.7 2,894 602 563 121 2,758 554 451 100 27 36.6 28,043 616 609 150 24,932 1160 1044 100 28 36.8 3,130 269 226 81 1,082 319 300 100 29 41.7 14,356 1341 1330 150 9,596 1496 1448 100 30 65.3 14,147 1362 1356 150 10,247 1581 1462 100 22 K. S. LASHLEY AND L. E. WILEY the correlations are quite high and significantly greater than their probable errors. For the most significant of the criteria, errors exclusive of the first trial, the coefficients range from 0.57 ± 0.08 for group 4 on maze V to 0.80 ± 0.05, with an TABLE 5 Learning scores for operated group II in mazes II and V. Arranged as table 4 PER CENT DESTRUC- MAZE II MAZE V NO. .5 CO f* c co E o u 02 as p •h 5-1 — £ QJ.S N B P .5 oo »- C t o E E E ■J. o b cc »- TION S * tc 4) Eh o co "a 3 .9 * « o g«- E-i • 3 1 15 c co ot "3 I 31 2.1 23,984 100 93 35 505 49 44 36 32 2.1 7,882 166 119 47 934 92 70 61 33 5.3 6,247 115 87 31 808 69 54 81 34 6.2 1,636 112 60 31 131 37 8 11 35 12.4 6,293 91 88 71 319 27 20 21 36 13.5 7,680 222 182 61 772 74 56 46 37 14.7 3,254 387 308 136 562 202 143 41 38 15.6 1,313 102 57 150 465 90 62 100 39 16.4 14,959 387 367 150 3,421 170 135 76 40 18.6 1,440 107 93 36 79 25 5 11 41 19.2 22,365 318 314 77 538 107 96 31 42 20.6 6,697 139 133 50 323 49 38 26 43 20.7 2,040 188 149 71 817 168 153 81 44 21.3 3,926 338 285 150 2,054 294 201 85 45 21.3 88,476 442 432 30 1,685 313 226 46 46 21.5 2,276 183 160 136 557 178 104 71 47 22.5 19,579 858 838 150 13,060 998 695 100 48 22.9 528 85 77 56 428 167 89 30 49 24.9 17,295 307 283 98 1,508 129 125 100 50 26.1 5,235 917 900 150 3,232 2714 1187 100 51 31.9 9,671 1,708 1695 150 4,573 1087 1046 100 52 32.1 2,466 409 374 146 506 125 120 66 53 32.5 1,747 267 241 106 367 148 93 31 54 33.9 35,515 837 751 150 2,143 285 225 100 55 36.4 3,814 475 439 120 456 146 112 36 .-.<; 37.2 10,612 2,230 1509 66 1,678 498 437 36 57 39.8 15,500 1,973 1819 150 1,569 340 314 100 58 39.8 9,621 2,594 2120 150 8,646 1716 1712 100 41.2 6,352 717 703 150 2,237 802 469 100 60 46.3 92,267 2,485 2391 150 22,636 3770 3700 100 61 49.1 70,006 3,196 2971 150 21,437 2257 2253 100 82' .-<;.! 34,722 10,204 9437 1191 ■ Di< .1 after 119 trial 3. STUDIES OF CEREBRAL FUNCTION. IX 23 average of 0.72. These figures demonstrate a significant rela- tionship between extent of lesion and degree of retardation in maze learning, and substantiate the earlier results of Lashley ('29) with similar mazes. TABLE 6 Learning scores for group III in mazes III and V. Arranged as table 4 PES CENT DESTRUC- TION MAZE III MAZE V NO. Total time minus time first trial (seconds) to u o i o OS E oo © U U_ I6- ■ .1? Total time minus time first trial (seconds) 05 U o u u 0> Is 1 oo U 03 O U 03.2 ggd cEcc 03 .Is 'u 63 2.1 1,337 125 103 31 216 28 14 21 64 2.8 17,141 166 98 24 263 28 16 16 65 3.8 9,335 189 143 94 397 56 40 26 66 10.0 4,084 388 371 126 8093 530 437 48 67 10.5 3,176 216 145 61 185 29 21 21 68 13.5 981 115 80 41 153 40 32 11 69 15.6 4,645 188 137 77 365 47 36 31 70 16.9 8,934 127 85 51 210 197 26 16 71 18.4 9,460 1106 1095 150 756 256 143 46 72 19.0 13,445 1766 1735 150 2117 845 427 100 73 19.1 6,178 749 642 150 2896 438 437 100 74 19.4 1,119 144 99 58 124 21 14 26 75 19.9 1,949 118 96 66 384 72 63 30 76 20.6 56,316 723 655 150 1136 172 151 66 77 23.2 3,795 335 295 91 78 24.1 11,270 402 302 85 2922 408 223 61 79 24.5 10,614 605 528 146 666 94 88 41 80 25.2 14,857 410 348 150 466 127 78 26 81 26.8 12,435 4323 2411 150 4495 453 420 100 82 28.3 23,014 1043 995 124 3047 767 237 51 83 29.8 2,046 174 154 101 303 100 75 21 84 29.8 8,250 764 662 127 1904 615 471 100 85 30.4 9,074 453 440 150 1301 283 275 100 86 36.0 45,877 6947 6480 150 2085 441 394 100 87 40.0 7,594 671 634 150 1650 299 262 100 88 40.2 17,427 2133 1946 150 7170 1021 927 100 89 41.9 31,690 3898 3866 150 4028 1376 1144 100 90 44.2 11,713 666 619 150 1318 169 99 66 91 47.6 18,649 3488 3321 150 1767 535 511 100 92 1 52.3 5,891 1205 1102 150 1226 445 380 100 931 60.2 12,405 2791 2624 150 3754 1094 1008 921 Refused to run after 92 trials. 24 K. S. LASHLEY AND L. E. WILEY So far as the magnitude of the correlation is concerned, no definite conclusions are justified. Lashley reported a cor- relation of 0.72 between errors and extent of lesion — a figure TABLE 7 Learning scores for operated group IF in mazes IV and V. Arranged as table 4 PEE CENT DESTRUC- MAZE IV MAZE V NO. Total time minus time first trial (seconds) 03 E o i ■a S 8 S3 u w u ime 3 time rial nds) OJ E 2 CO E m O H og h TION 0» 1 cj fl ec OS* jS Eh Total t minui first t (seco 0> "3 1 a-gj an I 94 1.3 15,440 448 379 105 568 73 49 21 95 1.5 4,035 189 112 16 75 14 5 6 96 3.4 7,476 180 149 60 1,013 37 32 11 97 6.6 4,321 183 173 51 1,902 89 83 33 98 11.5 6,136 145 98 31 446 34 21 21 99 12.3 5,869 810 737 150 1,664 389 347 100 100 13.4 1,600 306 232 46 391 87 65 16 101 15.6 4,747 313 285 76 184 57 21 16 102 18.4 3,521 257 194 89 812 80 75 41 103 18.8 3,492 645 530 71 738 163 115 71 104 19.9 33,967 6,050 6,011 150 1,525 478 468 100 105 21.3 8,920 418 389 98 151 19 9 11 106 21.5 1,596 197 157 71 195 48 42 31 107 21.5 5,898 644 607 57 386 136 63 51 108 22.8 7,384 1,276 954 150 1,043 618 362 100 109 23.5 2,248 161 127 61 175 30 23 11 110 23.5 6,536 561 523 150 244 40 35 21 111 24.9 93,181 2,063 1,855 150 12,651 623 479 100 112 25.3 2,157 188 131 66 236 29 17 21 113 25.4 5,954 1,233 1,221 150 997 419 407 100 114 25.4 7,661 1,495 1,386 150 3,116 984 874 100 115 25.6 1,731 360 230 51 423 138 96 41 116 26.8 4,665 461 428 136 597 156 119 46 117 29.4 4,038 714 501 106 367 61 38 16 118 29.4 79,522 12,844 7,163 150 6,939 1217 1192 100 119 32.6 16,241 1,774 801 150 3,983 505 378 100 120 34.7 19,262 497 445 119 828 159 76 43 121 37.7 29,654 2,659 1,888 150 5,565 738 544 100 122 37.7 27,971 2,748 2,734 150 1,730 320 312 100 123 42.1 13,177 924 812 150 940 203 143 61 124 43.2 5,858 1,725 1,660 150 1,922 764 691 100 125 44.2 6,253 667 547 150 1,095 423 216 51 126 46.5 12,920 1,013 741 150 4,073 272 266 100 127 49.7 137,964 12,978 12,818 150 19,320 1504 1364 90 STUDIES OF CEREBRAL FUNCTION. IX 25 TABLE 8 Training records of normal group I in mazes I and V. Arranged as table 4 MAZE I MAZE V NO. ■ s O .si g«H sj-g 00 E o K u M ■ ■ § ■ ui E I 00 IE 00 £_ m £ 1 2833 1756 34 31 21 1243 488 45 21 21 2 187 164 17 14 16 494 363 69 54 41 3 206 179 12 10 6 1049 581 88 52 26 4 285 114 23 11 11 413 360 51 40 36 5 4138 3998 34 31 15 5873 5428 79 68 17 6 4331 4201 55 49 71 865 534 38 29 21 7 2084 1824 50 44 41 615 385 36 26 31 8 567 481 21 17 31 466 381 36 29 26 9 414 387 13 11 16 629 359 47 32 51 10 764 668 29 22 41 284 230 22 17 26 11 1565 1510 33 31 20 61 48 2 2 11 12 264 222 5 4 11 844 677 47 29 21 13 713 404 13 8 21 3319 2604 172 147 46 14 171 103 4 2 6 428 243 25 16 21 15 2436 2389 25 24 35 4715 2870 167 104 36 TABLE 9 Training records of normal group II in mazes II and V. Arranged as table 4 MAZE II MAZE V NO. 00 •a ?! 03 2 u g«H 111 00 E 2 E A oo Is 00 u *3 00 si 00 .5 2 a* §J| oo E £ E m 00 I? III J2 "3 I 16 513 426 48 40 21 241 101 29 11 11 17 10781 2475 112 52 26 1333 654 72 46 32 18 6586 5654 88 49 26 1069 326 39 21 26 19 4389 3129 100 69 41 5998 898 123 18 36 20 18885 17513 127 104 45 9337 1002 182 20 31 21 2618 1981 90 73 26 536 300 44 24 31 22 12927 11336 199 156 18 405 287 45 37 26 23 3012 2647 116 98 54 8324 4628 172 88 40 24 4748 1838 124 77 37 346 304 20 13 26 25 2846 2616 80 73 50 344 189 21 14 11 26 2536 2391 90 82 30 235 100 14 8 21 27 5570 4293 113 95 35 896 329 33 17 16 28 2235 1071 81 58 46 198 149 13 10 16 29 1595 1222 58 46 21 1301 1166 64 58 21 30 829 707 33 27 31 432 205 22 11 31 26 K. S. LASHLEY AND L. E. WILEY TABLE 10 Training records of normal group HI in mazes III and V. Arranged as table 4 MAZE III MAZE V NO. Time (seconds) Time minus time first trial m h 6 E go .2 2 \u Is at c Time minus time first trial 05 E £ u H .2 2 £ 2 03 * 31 750 656 31 25 21 312 124 21 8 16 32 2163 2030 89 82 41 215 63 13 4 11 33 2064 1504 82 65 29 441 299 56 47 43 34 3287 2816 175 142 56 853 349 35 13 21 35 7210 5145 189 171 46 591 372 46 35 26 36 2071 1814 90 76 31 879 720 46 35 40 37 4849 1669 191 125 31 627 262 48 26 16 38 9620 2420 169 98 52 414 147 31 13 16 39 2518 1258 92 61 51 406 355 23 20 41 40 9555 8790 153 132 43 457 155 24 3 6 41 1462 527 89 43 26 219 141 36 25 11 42 5802 4537 146 114 57 1508 793 98 82 100 43 7241 2155 267 64 101 1158 602 40 21 41 44 1360 1114 50 39 11 341 182 19 7 31 45 2442 1542 82 49 36 143 82 7 4 16 TABLE 11 Training records of normal group IV in mazes IV and V. Arranged as table 4 MAZE IV MAZE V NO. S3 o si Time minus time first trial Errors M Errors minus o errors first 1-1 trial 35 u Time (seconds) Time minus time first trial Errors Errors minus errors first (rial ■a u 46 10382 3481 | 245 76 938 698 46 38 61 47 3558 2633 113 73 101 625 461 27 17 31 48 2327 1718 \ 180 130 101 364 355 50 49 56 49 3113 2941 465 448 36 2398 1612 417 349 51 50 9681 6391 113 87 51 2928 499 71 12 11 51 11100 2858 209 106 51 876 30 33 1 6 52 10562 7861 97 82 41 407 210 17 12 6 53 2291 1041 72 41 81 817 149 22 12 16 54 3761 2736 149 82 41 144 80 11 5 11 55 6389 2461 221 91 26 305 224 22 16 26 56 5376 2428 186 60 31 451 159 21 5 21 57 3979 1279 219 47 51 229 147 16 11 26 58 11539 11013 253 230 126 405 310 17 13 11 59 15035 4235 64 50 27 854 458 27 14 23 60 30241 22141 22 D 194 79 1956 407 72 28 16 i t E* <3s > c H* >~ «J 01 >- CC t» © H N N CO OJ l>- r-l OS "^ HI OS CO CO cm t-i © d OO rl O0 LO -* © oq h< l>" OO H< CO cm n co OS O HI b» © oq hi h<# d d cm* h* H N ri (O n io oo to co* hi d cm" iH HI CO CO t>- 00 OS CO IO CO rl © " HI HI t)I (O N S rl rH rl CO b- t- Hi CO t- IO HI CM* ri 00* CM CO CO rl " a N rl O IO O Oj C> IO M ci ri co id OS OS rH O 00 HI OS © r-i r-i CO CO OS t^ 00 CO rl IO rH rl d IO IO IO IO HI HI O os oq «m cm CO CO HI HI m CO CO W (O t>- os co oo d d o d rl rl N CO d d CM Hi" IO HI HI H< OS CM rl rl O t>- Hi CO Tl — " Hi" d CO CO CO CO ff ^ in t- O CO t^ N X H N ■* CO M ri M CO rf CC CO O o o !>; © © 00 00* IO © H< CJ CM CM CM O HI rl CM t* OO CO rl t>° CO HI d t- O rl O rl rl rl s to co to h sq »q o. oo* id d d CD CO CO IO ' M ft H T|( © H © CM -* OS OS t- O CM W t> HI 00 rH O r-i r-i t- CM rl OO © OS CD CO d io* io* id CM CO CD CO O Oj rl CO H^ CO H^ H1 m O OS O HI oo © hj os IO t^ 00 o rl OS rl CO ^ CO o HI t^ rji oo" d rH 00 CO CM CO CM 00 CO rH d oo* d d HI HI HI HI IO O CO OS CD "* CM 00 ; CM CM* CO* d '< CO CO CO CO TRIALS MINUS CORRECT RUNS o o CO o HI © OS CM l—l I—I r-i OS rH CM CM (M N CO N CO « H CO rl CO ri ri N rl rl rl rl CO OS 00 HI CO CO CD HI IO* © rH 00 IO t- OS 00 CO t~ CM CO ' OS !>; b«. CO d oo -t >t IO ■* HI -r)! a m ft O CM CM 00 H} iq cm io M N l> S H ■* in co n CO l>-# IO CD d co' CO -hh rl CO OS CO OS IO Hi CM t>; CM* CO d 00* hi o co oo OJ H IT IO OS t— CD ' !>. CM CD CO t-' id d oo HI OS CO CO 4 IO 00 CO CO N rl •* ffl CO CO r" © H -f ■* O IO rl 00 CM HI IO CM OS d ri d CO CO CM CM 00 CD rl CO IO CO CO OS OS id d d d HI O OS 00 CO t^ CO o rl n \r. H1 CO 00 OS S W N O ' N CO S ri 00 00 OS CO . CO l» N « ERRORS MINUS ERRORS FIRST TRIAL °. t~ b- HI £i "O »0 ri w CD X N rH O o t~ O •* hi oq oq HI CC N 00 •HH CM CO CO O OS CO 00 i>- © © oq CO ri OJ oi H- CI CO 00 CM OS o CO rl rl t^ CO CO o oq co c© io CO ri ri IO ' rc 1- 00 CO : CO -f CM CM ' ) AVERAGE PER CENT LESION o o o o o o © o t>; CO t>- HI H ri H ri + 1 +| +1 +1 N H N CD d it' ui •+ : 1 -ri -i ?i l~ 1- t- HI ri ri ri ri + 1 +1 +1 il , N H N CO -' o io io io i- i- IO IO O CM rH HI CO CO CO CO O rH O HI CO CO CO CO MAZE M M M > M W * > > > > rH hH M t> *-< — —. > > > > ft & o H 9 rH CM CO ** rl CM CO -* — CI M H" d- eL. Oh &, 5 £ z z — '.I i i d d 5 fi z c c n c c o o c : fc x x fe = g a a M E E m o o o o Z fc fc X 5 ^7 28 K. S. LASHLEY AND L. E. WILEY which corresponds exactly to our average, but this can scarcely be considered more than a chance correspondence. If all conditions of random sampling are fulfilled, the coeffi- cient of correlation gives a measure of the relative effective- ness of a common factor and other causes of deviation in determining the distribution of two series of variables. Our data, however, do not fulfill the necessary conditions, and the magnitude of the correlations only justifies the conclusion TABLE 13 Correlations between total extent of lesion and the various criteria of maze learning for the five mazes studied Maze I II III IV V Number of cases 30 32 31 34 125 Errors less first trial Time less first trial Total trials Trials less correct runs 0.80 ± 0.05 0.75 ± 0.06 0.75 ± 0.06 0.79 ± 0.05 0.80 ± 0.05 0.32 ± 0.11 0.69 ± 0.07 0.79 ± 0.05 0.70 ± 0.07 0.43 ± 0.10 0.71 ± 0.06 0.74 ± 0.06 0.60 ± 0.08 0.39 ± 0.10 0.66 ± 0.07 0.56 ± 0.08 0.64 ± 0.04 TABLE 14 Correlations between total extent of lesions and the various criteria of maze learning for the comparison maze (maze V) ERRORS LESS FIRST TRIAL TIME LESS FIRST TRIAL TOTAL TRIALS TRIALS MINUS CORRECT RUNS Group I 0.74 ± 0.06 0.60 ± 0.08 0.74 ± 0.06 0.79 ± 0.05 Group II 0.75 ± 0.06 0.51 ± 0.09 0.63 ± 0.08 0.71 ± 0.06 Group III 0.66 ± 0.07 0.52 ± 0.09 0.70 ± 0.07 0.73 ± 0.06 Group IV All combined 0.57 ± 0.08 0.64 ± 0.04 0.46 ± 0.10 0.54 ± 0.09 0.52 ± 0.09 that there is a significant relationship between the extent of cerebral lesion and the amount of retardation in maze learn- ing. In fact, the use of the correlation coefficient with our data is justified only as a rough test of the existence of a relation- ship. Data presented below show that the relationship is not rectilinear and that the correlation ratio would more accu- rately express it, as Lashley ('26) found for the effect of lesions upon brightness discrimination. Since the distribu- STUDIES OF CEREBKAL FUNCTION. IX 29 tion of cases is not normal either with respect to extent of lesion or learning scores, and since the slight increase in the magnitude of the measure of association given by the cor- relation ratio would be meaningless at present, we have not computed these constants. Comparison of table 3 with tables 13 and 14 shows that for operated animals the intermaze correlations are significantly higher than the correlations between extent of lesion and the criteria of learning. Since the intermaze correlations for normal animals are very low, the intermaze correlations for operated animals must be ascribed to some effect of the lesion, and it appears that the effective agent in cerebral lesion is more accurately measured by maze performance than by measurement of the surface area of the lesion. This may be due either to the failure of our methods of measurement of the lesions to express all of the significant characteristics of the lesion or to the fact that the correlation coefficient is a better expression of the relationship in one case than in the other. The latter possibility has certainly played some part in the matter, for the intermaze relationship is rectilinear, whereas the cortex-learning relationship is not. However, since the distribution of lesions is not normal, no correlation method can be depended upon to give a certain picture of the relationship and there is no present method of finding the causes of the discrepancy. Continuity of the mass relationship As a further test of the validity of our conclusions based on correlations, and to determine whether the correlations represent a continuous relationship or are due to the destruc- tion of some critical amount of tissue, we have divided the cases by class intervals of 10 per cent destruction and com- puted the average practice for learning required by the ani- mals in each class interval. The results of this analysis are presented in table 15. The last two intervals (40 to 49, and 50 per cent) are based upon too few cases to have significance. In the remaining four intervals there are only six inversions 30 K. S. LASHLEY AND L. E. WILEY of order in the fifteen series of constants. This is conclusive evidence of a continuous relationship between the extent of injury and the degree of retardation. Previous data (Lashley, '29; Maier, '32) have suggested that there may be a limit of size below which lesions are rela- tively ineffective and above which there is marked defect. In these more adequate data there is no indication of a con- sistent, marked flexion point between any two of the class TABLE 15 Average practice for learning re quired oy animals with various amounts of cereoral destruction grouped in class intervals of 10 per cent MAZE NORMAL 0-9 10-19 20-29 30-39 40-49 50+ Errors I 22.0 32.9 112.1 260.2 360.6 1330.0 1356.0 minus II 66.7 89.8 201.3 361.9 1118.5 2015.0 9437.0 errors III 85.7 114.7 448.5 705.6 3460.0 2075.2 1863.0 first trial IV 121.5 203.3 1155.3 1119.4 1467.0 3323.6 V 33.2 52.6 142.6 292.9 449.0 963.8 950.0 I 24.3 28.9 60.1 114.2 108.6 150.0 150.0 Total II 33.8 36.0 97.3 99.0 129.8 150.0 119.0* trials III 42.1 49.7 93.0 124.9 150.0 150.0 150.0 IV 61.2 58.0 87.6 110.4 142.3 150.0 V 26.8 34.2 48.4 64.7 84.1 90.6 97.3 Time I 1227 593 2205 4084 9550 14356 14147 minus II 3953 9937 8186 16228 11118 56208 34722 time III 2532 9284 5397 15844 27476 17415 9148 of first IV 5014 7818 8476 16535 23282 35234 trial V 610 757 1161 2387 3789 7085 5076 Mied after 119 trials. intervals. It is very probable that where such a condition has appeared in earlier data, it has been due to chance varia- tion in inadequate samples. With the limited data heretofore available, it has not been possible to define the form of the relationship between extent of lesion and degree of retardation. Lashley 's data ('26, '29) indicate that extensive lesions produce a disproportion- ately great effect. Thurstone ('33) has concluded from an analysis of the original data that the formula k = aC6, where STUDIES OF CEREBRAL FUNCTION IX 31 h is the efficiency in learning, a a constant, and C the amount of cortex intact, best expresses the form of the relation. With the 125 cases on maze V we have a more reliable basis for analyzing this relationship. Figure 2 shows the data for extent of lesions, grouped by class intervals of 5 per cent ERRORS 5 • ' 1 »1 i 1 IOO0 1 / j f ! 500 y • J / 1 , /. • -/ / • - N 10 20 30 40 50 60 PER CENT LESION 70 Fig. 2 The relation between extent of lesion and errors made in learning in maze V (8 culs de sac). The data on 60 normal and 127 operated itnimals ;ir. averaged by class intervals of 5 per cent destruction. The smooth curve is the best fitting one of logarithmic form. The numerals in the figure Indicate tin- number of cases on which the more unreliable points are based. destruction and plotted against average errors tor each class. The first point in the curve, zero destruction, is the average of 60 normal animals. The other points are determined by smaller numbers of cases. The continuous curve has been derived from the data by the method of least squares and is the best fitting logarithmic form. It is given l>y the n\ nation 32 K. S. LASHLEY AND L. E. WILEY E s== (38.39)e(0-0698)L, in which E is the error score, L the per- centage lesion, and e the Naperian base. "Where there is an adequate number of cases, the data con- form quite closely to the regular curve and suggest that the retardation from cortical destruction follows some definite law by which learning ability for the maze rapidly approaches zero with larger lesions. For the lower amounts of destruc- tion, where the number of cases is large enough to give relia- bility to the averages, the experimental data conform quite closely to the derived curve and there is no indication of any critical amount of destruction resulting in a sharp rise in the error scores. From this extensive series of cases and from indications given by less adequate earlier series, it seems certain that with increasing size of lesion, learning ability for the maze decreases at a steadily accelerating rate. The exact form of the curve has no significance at present. The influence of subcortical lesions upon the mass relationship It is impossible to obtain an extensive series of cases with large cerebral injuries without some lesions in the thalamus and archipallium. Lashley ('29) attempted to estimate the influence of such subcortical lesions in two ways: first, for control of thalamic lesions, by computing correlations sepa- rately for all cases with and for all cases without thalamic lesions, as well as for these two classes combined; secondly, by assigning an arbitrary value to each type of subcortical lesion and correcting the ranking for correlation obtained from cortical lesions by these arbitrary values. These analyses gave the following results. Correlations between errors and cortical destruction: For all cases, p = 0.86 For cases without thalamic lesion, p = 0.83 For cases with thalamic lesion, p = 0.86 With arbitrary correction for all types of subcortical lesions the correlations were the following between total destruction and the learning constants : Uncorrected Corrected 0.62 0.67 0.86 0.87 0.77 0.80 STUDIES OF CEREBRAL FUNCTION. IX 33 Time, Errors, Trials, The results of this analysis indicated that the inclusion in the series of animals with lesions in the archipallium or thalamus did not significantly alter the results from those which would have been obtained had only cases with lesions in the neocortex been included in all computation of constants. Correction for subcortical injury slightly raised the correla- tions with extent of destruction and indicated that "if lesions to internal structures could be accurately evaluated, the method would most probably reveal a still closer corre- spondence between learning ability and amount of functional tissue.' ' "We have instituted similar controls for the present data. The possibly significant subcortical structures, septum, caudate and lenticular nuclei, fornix, hippocampal lobes, col- liculi, habenulae, optic paths to the thalamus, anterior, median, and lateral thalamic nuclei, lateral and median geniculate bodies were represented on diagrams and three arbitrary grades of severity assigned to each. All brains were reexamined carefully for subcortical injuries and the amount of destruction in each of the above structures was graded and listed. All cases in which there was no subcortical injury or at most slight degeneration in the dorsal convexities of the hippocampal lobes were selected. With these the correlations between extent of cortical destruction and errors in learn- ing were computed. These constants are listed in table 16, in comparison with the constants computed from all cases. The elimination of the animals with significant subcortical lesions reduces the correlations very slightly, but since the cases with most extensive subcortical destruction have also the more severe cortical injuries, this procedure reduces the range of variation and the reduction in correlation is no more than would be expected from the reduction in range alone. THE JOURNAL OF COMPARATIVE NEUROLOGY, VOL. 57, NO. 1 34 K. S. LASHLEY AND L. E. WILEY We have attempted to analyze the data on the basis of specific subcortical injuries, but have not been able to dis- cover any significant relations. In our data it is usually possible to match any case with a subcortical injury with another having only a similar cortical destruction. It seems entirely a matter of chance as to which member of such pairs has the worse training record. We have not been able to find indications of specific effects upon maze learning of any lesions in the corpora striata or thalamus within our series of cases.6 It seems quite certain that the correlation between extent of cortical injury and the degree of retarda- tion is not due to the inclusion in the series of animals with subcortical lesions. TABLE 16 Comparison of the correlations between extent of lesion and errors in learning for all cases and for cases without significant subcortical lesions. Maze V GROUP ALL CASES P NO SUBCORTIAL LESIONS P I II III IV 0.80 ± 0.05 0.80 ± 0.05 0.70 ± 0.07 0.60 ± 0.08 0.76 ± 0.06 0.72 ± 0.07 0.69 ± 0.08 0.66 ± 0.08 Average 0.72 0.71 The relative effects of symmetrical and asymmetrical lesions in producing retardation of learning In order to test the influence of corresponding areas of the two hemispheres, we have determined the areas common to the destruction in both hemispheres, as described on page 16, and from these have computed the correlation between extent of symmetrical destruction and errors made during training, using only the data on mazes I and IV, as a sample. The results of this analysis were given in table 17. The cor- relation for the symmetrical portions alone is the same as for the total extent of lesion. That for the asymmetrical a The subcortical lesions are for the most part slight and unilateral. In no case is there an extensive bilateral injury to any thalamic center. STUDIES OF CEREBRAL FUNCTION. IX 35 portions with error scores is less (0.49 and 0.35), but is as great as is to be expected, considering the reduction in range of lesion when computed on this basis alone. There is a still smaller correlation between the extent of symmetrical and asymmetrical lesion. This may be responsible for the ap- parent correlation between the latter and error scores. We have attempted to partial out its influence, with the results shown in the last column of the table. Not much weight can be ascribed to such statistical analyses, however, since the relationships are not rectilinear and the data have not normal distribution. The partial correlations (asymmetrical lesion with errors, with the influence of symmetrical lesions held constant) seem still significantly large and suggest that the asymmetrical portions of the lesions contribute to the deterioration as do the symmetrical portions. TABLE 17 Comparison of the effects of lesions common to the two hemispheres and of the asymmetrical portions of the lesions. The correlations are between extent of lesion and scores for errors less first trial in learning maze V GROUP TOTAL LESION PART COMMON TO BOTH HEMISPHERES ASYMMETRICAL SYMMK™I0AL PABT ASYMMETRICAL PARTIAL CORRELATION I IV 0.74 0.57 0.79 0.60 0.49 0.39 0.35 0.24 0.32 0.39 The relative influence of lesions within different cyto architectural fields Lashley ('29) attempted to estimate the effectiveness of injuries within each of the chief functional fields by grouping his cases according to the field most seriously involved and computing constants for each group. The small number of animals which he had available made it necessary to include ambiguous cases, and inspection of his figures shows that there was an extensive overlap between the groups. This considerably reduces the validity of the evidence presented for equal effects of equal lesions in different areas. We have sought a more conclusive test of the matter by selecting nearly unequivocal cases from our data. Since there 36 K. S. LASHLEY AND L. E. WILEY is a good bit of individual variation in the positions of the boundaries of the fields and probably even more distortion resultant from our methods of plotting the lesions and, finally, a good bit of uncertainty in determining the exact boundaries of the fields by direct histological methods,7 there is no ac- curate means available for determining the amount of destruc- tion in the different functional areas. As an approximate measure, a transparent diagram of Lashley's modification of Fortuyn's diagram of cytoarchitectural fields was superim- posed upon the diagram of the lesions for each case, and the percentage of the total cortex destroyed within each field was measured with a planimeter. On the basis of these percentage measurements, we selected cases in which the extent of lesion in one field was three or more times as great as in any other field and in which the total extent of lesion in any except the primary field did not exceed 5 per cent of the total cortex. The cases assigned to the four principal areas were the following: ff'n (motor) : 17, 18, 32, 41, 64, 65, 70, 75, 76, 80, 111. j (somesthetic) : 2, 4, 7, 9, 20, 34, 47, 73, 74, 78, 85, 94, 102, 106. p (auditory) : 1, 3, 6, 11, 19, 39, 40, 52, 79, 98, 100, 103, 105. w (visual) : 10, 14, 35, 36, 44, 66, 69, 71, 72, 97, 99, 104, 113. For each of these groups the average extent of lesion and the average of errors (less first trial) in learning maze V were computed. These constants are given in table 18 and graphic- ally in figure 3. The average destruction for the four groups is so nearly the same that the differences may be disregarded. In comparison with the average for normal animals in maze V (33.1 ± 4.2 errors) all of the groups were markedly retarded. For the motor, visual, and somesthetic fields the training records are essentially equal, the differences between the 7 A survey of the literature on the cytoarchiteeture of the brains of rodents by the senior author reveals that the disagreements among investigators in this field are so great as to cast doubt upon the significance of most of the areas differentiated. No two investigators have used the same criteria for distinguish- ing the areas and such criteria as have been defined are purely relative. The confusion is particularly striking within the areas called auditory and somesthetic. STUDIES OF CEREBRAL FUNCTION. IX 37 groups being less than their probable errors, whereas all make approximately five times as many errors as do normals. The record for cases with lesions in field p is not consistent with the others. The score of 92.3 errors is markedly less than that of the next lowest group, but is also significantly TABLE 18 Analysis of effects of lesions largely confined to single cytoarchitectural fields FIELD NUMBER OF CASES AVERAGE LESION a AVERAGE ERRORS a 8TANDARD COEFFICIENT OF VARIATION ff'n (motor) j (somesthetic) p (auditory) w (visual) Normals 11 14 13 13 60 15.9 ± 1.7 15.4 ± 1.3 16.0 ± 1.5 15.9 ± 0.9 0 8.5 7.4 7.8 5.0 154.2 ±l 40.5 173.1 ± 32.3 92.3 ± 21.1 156.9 ± 41.5 33.1 ± 4.2 199.5 185.4 112.7 222.3 48.8 1.29 1.07 1.21 1.42 1.45 M izn p M I 1 ff'n M Fig. 3 A comparison of the effects of equal amounts of destruction in differ- ent cortical fields upon maze learning of maze V. The lines represent the relative magnitudes of the mean scores (M) with their probable errors for normal animals and for four groups of operated cases trained in maze V. N, normal; p, audi- tory; w, visual; ff'n, motor; j, somesthetic. higher than the score of normals. The reliability of the differences in both cases is low (ff'n — p = 80.8 ± 36.6; p — normal = 59.2 ± 21.5). The indications from this analysis of the data are that equal lesions within the motor, somesthe- tic, and visual areas produce approximately equal effects upon the capacity to learn the maze and that lesions within the 38 K. S. LASHLEY AND L. E. WILEY auditory area are relatively less effective, although they also produce a significant deterioration of the function.8 We have computed the correlation between extent of lesion and error scores for each of the above four groups with lesions largely restricted to single architectural fields. The constants obtained were the following : Motor, p = 0.66 ± 0.12 ; somesthetic, p = 0.58 ± 0.12 ; auditory, p = 0.07 ± 0.19 ; visual, p = 0.36 ± 0.17. This selection of cases greatly reduces the range of varia- tion and consequently affects the magnitude of the correla- tions. The mass relation seems to hold within the motor, somesthetic, and visual areas. As in the comparison of aver- ages, the auditory area does not conform to the trend of the others. For the question of the exact equivalence of the various parts of the cortex for maze learning our data must be re- garded as inconclusive. Clearly, lesions in any part of the cortex produce marked retardation. For the motor, somes- thetic, and visual areas this retardation is approximately equal, at least more nearly so than the demonstrated retarda- tions from peripheral sensory defect. The lesser effects of 8 We have attempted to use another method for comparison qf lesions in the different fields. The percentage of the cortex included within the lesion in each field was listed for each animal. On the assumption that the injury within each field contributes to the deterioration of the animal according to some determinant (D) which is constant for that field, the total error score may be expressed as the sum of the products of the percentage destruction within each field by the determinant for that field. This gives, for example, for animals nos. 16 and 17 in maze I the equations — 13.6 Df/n + 7.1 Dj = 57 10.9 Df/n + 7.8 Dj = 33. We thus obtained 127 equations for maze V. These were solved by the method of Doolittle to give an average value for D for each of the four chief fields. The results were Dff/n =7; Dp = 8; Dj = 18; Dw = 27. Using these values, the expected errors were computed for each animal and plotted against the experimental errors. This comparison gave theoretical values consistently too high for the lesser lesions and too low for the greater, showing that what had seemed a promising method is inapplicable to our data, because of the non- linear relation between extent of lesion and performance. We are indebted to Prof. L. L. Thurstone for the test of this method. STUDIES OF CEREBRAL FUNCTION. IX 39 injuries in the auditory area are probably significant, but even with so extensive a series of cases as are included in our study, the statistical reliability of our data is low. INFLUENCE OF THE COMPLEXITY OF THE PROBLEM To test the question whether an increasing number of culs de sac in the maze offers a progressively greater difficulty for animals with cerebral lesions than for normals, we have made a comparative analysis of the learning records of normal and operated animals trained on mazes I, II, III, and IV. The individual records of the normal animals have been presented in tables 4 to 11. Equation of groups Since, as was pointed out in our discussion of methods, it is necessary to use a separate group of animals with each of the mazes in the tests of complexity, some method of equat- ing the groups is desirable. Members of the groups were taken at random from the stock colony and were presumably a random sample of the colony. An attempt was made to obtain four identical series of animals with destructions of from 5 to 60 per cent of the cortex, by the method described on page 15. Comparative data on the extent of lesion have been presented in table 2. Group I shows the smallest aver- age percentage destruction, but also the greatest range. The averages for the other three groups are practically identical, with a somewhat smaller range in group IV than in the others. In the loci of the lesions the groups seem sufficiently alike to permit of direct comparison. As a test of the equality of the groups, all were trained on maze V after completion of training on the comparison mazes. The average scores for the eight groups in maze V are pre- sented in table 19. Measurement of the equality of the groups by the use of a second maze is complicated by the possibility of differential transfer from the different mazes used in the initial training. The table gives some indication that such a transfer has taken place. Converting all the constants of 40 K. S. LASHLEY AND L. E. WILEY the table into percentages of the scores of group I and aver- aging these, we obtain the following figures: group I, 100; group II, 89.9 ; group III, 72.9 ; group IV, 77.1 per cent. These figures indicate that the animals previously trained on the TABLE 19 Constants for normal and operated animals on maze V: to test the equality of the groups previously trained on mazes I to IV ERRORS LESS FIRST RUN TRIALS LES CORRECT RUN- TIME LESS FIRST TRIAL (SECONDS) Normals I 44.4 ± 6.3 36.4 28.7 ± 1.9 10.8 13.9 ± 1.3 7.4 1037 ± 254 1425 II 26.4 ± 3.7 21.5 25.0 ± 1.5 8.5 11.1 ± 0.8 4.4 709 ± 197 1098 III 22.9 ± 3.5 20.3 29.0 ± 3.9 22.4 11.7 ± 1.4 8.0 310 ± 39 222 IV 38.8 ± 14.6 83.9 24.8 ± 3.0 17.2 12.1 ± 1.8 10.4 387 ± 66 371 Operated I IT III IV 333.9 ± 47.7 387.7 451.4 ± 95.3 786.6 281.6 ± 36.7 297.8 265.5 ± 38.4 331.6 68.2 ± 4.0 32.0 65.3 ± 3.8 31.7 60.5 ± 4.2 34.4 56.8 ± 4.2 36.3 50.9 ± 4.0 48.7 ± 3.9 44.7 ± 4.2 44.4 ± 4.3 32.6 32.4 33.2 36.8 3263 ± 609 3177 ± 680 1857 ± 244 2244 ± 447 4948 5610 1983 3865 TABLE 20 Beliability of the differences for error scores among the normal and among the operated groups in maze V DIFFERENCE d/p.e.d Normal groups I- II I-III I- IV 18.0 ± 7.3 21.5 ± 7.2 5.6 ± 16.3 2.4 2.9 0.3 Operated groups II- I II-III II- IV 117.5 ± 106.0 169.8 ± 60.0 185.9 ± 61.0 1.1 2.8 3.0 mazes with few culs de sac learned maze V less rapidly than did those previously trained on mazes III and IV. The relia- bility of the differences between the constants in table 19 is not great. Table 20 gives the ratios of the greatest differ- ences between the error scores to their probable errors. Al- STUDIES OF CEREBRAL FUNCTION. IX 41 lowing for the differential transfer indicated above, these differences will be still further reduced. Groups III and IV tend to be somewhat better than groups I and II in both the normal and operated series, so the inequalities may be ex- pected to influence the results for both series in the same direction. Although absolutely large, the differences with maze V are relatively small in comparison with those obtain- ing between the records of the groups in the comparison TABLE 21 Average learning records of normal animals on the four mazes used for test of the influence of complexity MAZE ERRORS LESS FIRST TRIAL a TRIALS TRIALS LESS j 100.9 24.3 ± 2.9 33.8 ± 1.9 42.1 ± 3.5 61.3 ± 5.2 16.7 11.4 ± 1.0 10.9 21.0 ±1.2 20.4 21.9 ± 1.5 29.8 29.2 ±1.9 5.8 1227 ±214 1331 7.1 3953 ± 691 4452 8.4 2532 ±449 2081 10.9 5014 ±874 5266 TABLE 22 Average learning records of operated animals on the four mazes used for test of the influence of complexity MAZE ERRORS LESS FIRST TRIAL a TRIALS a TRIALS LESS CORRECT RUNS a TIME LESS FIRST TRIAL a I II III IV 246.7 ± 42.5 921.1 ± 203.5 1039.1 ± 169.2 1382.9 ± 588.8 345.4 1706.6 1396.9 5089.9 77.7 ± 6.2 103.8 ± 5.6 114.6 ± 5.2 109.0 ± 5.1 50.2 46.7 42.7 44.4 55.6 ± 6.1 79.4 ± 5.9 91.7 ± 5.6 88.4 ± 5.7 49.3 48.9 ! 46.3 49.1 4257 ± 784 16738 ± 2801 12419 ± 1488 17403 ± 3345 6074 23489 12283 28919 mazes. They seem unlikely to have produced a constant error in relation to the complexity of the mazes. They do restrict us to a consideration only of large differences as evidence of any genuine influence of the complexity of the problems upon rate of learning. Comparison of normal and operated animals in learnvm mazes of different complexities Tables 21 and 22 summarize the learning scores for normal and operated animals in mazes I to IV. For error scores there is a regular progression in practice required for learn- 42 K. S. LASHLEY AND L. E. WILEY ing from the simplest to the most complex. The scores by other criteria are less consistent, but show in general the same trend. The averages of all scores, expressed as percentages of the scores on maze I are given at the right. The ratios of practice required for learning by all criteria for mazes II, III, and IV on maze I are given in table 23. The objective complexities of the mazes, expressed in numbers of culs de sac are in the proportions of 1:2:3:4. The rela- tive difficulty for normal animals, in terms of the most con- sistent criterion, error scores, is as 1:3:4: 5.5. The ratio of difficulty for the operated cases (1 : 3.7 : 4.2 : 5.6) is not signifi- cantly different from that of the normal animals. The indica- TABLE 23 Comparison of scores in learning tests for normal and operated animals in the four comparison mazes, expressed as ratios on scores in maze I. N = normal group, Op. = operated group, 0p.<40% = cases with lesions of less than 40 per cent, included as a control of the effects of exclusion of cases which failed to get through the maze on the first trial RATIO OF OULS DE SAC ERRORS TRIALS TIME AVERAOE MAZE N. Op. Op.<40% N. Op. N. Op. N. Op. I 1 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 II 2 3.03 3.70 2.97 1.39 1.33 3.24 3.92 2.55 2.98 III 3 3.90 4.20 4.59 1.73 1.47 2.11 2.92 2.58 3.81 IV 4 5.50 5.60 6.23 2.52 1.40 4.12 4.07 4.06 4.19 tions from time and trials are essentially the same. There is no evidence that the longer mazes are disproportionately more difficult for the operated animals than for normals. The indication in table 19 that operated groups III and IV are superior to the others, even allowing for differential trans- fer, suggests that the selection of cases brought about by failure of some cases to get through a single trial may have favored those groups. We have therefore computed the aver- age errors for the operated animals exclusive of cases having more than 40 per cent destruction, thus eliminating the most badly deteriorated cases from groups I and II. The ratios of these averages on the scores for maze I for the four groups were the following : STUDIES OF CEREBRAL FUNCTION. IX 43 Ratio, Maze I 1.00 Maze II 2.97 Maze III 4.59 Maze IT 6.23 With this correction the operated animals do somewhat worse proportionately on the longer mazes than do normals, but the differences are still too small to be regarded as significant. As a further test we have computed the ratios for errors for groups taken by increments of 10 per cent lesion as given in table 15. These ratios are given in table 24. The figures are more variable, owing to the smaller number of cases, but there is no indication that any extent of lesion produces dis- proportionately poorer records in the more complex mazes. TABLE 24 The ratio of errors made during training for mazes II, HI, and IV to maze I, for animals classed according to extent of lesion MAZE i II in rv 0 1.00 3.03 3.90 5.50 1-10 1.00 2.72 3.46 6.10 10-20 1.00 1.79 4.00 10.30 20-30 1.00 1.39 2.72 4.31 30-40 1.00 3.12 9.63 4.07 40-50 1.00 1.52 1.56 2.50 As by other methods of treating the data, there is no evidence of any influence of cerebral lesion upon the proportionate difficulty of the different mazes. DISCUSSION Our primary object in these experiments was to test the influence of various amounts of cerebral destruction upon the capacity to form habits involving different numbers of similar tasks. We have failed to confirm Lashley 's finding ( '29) that increasing the number of culs de sac disproportionately in- creases the difficulty of the mazes for animals with brain lesions. His experiments differed from the present ones in the following respects: 1) All animals were trained in all mazes, thus permitting of transfer or interference effects. 2) His animals were first 44 K. S. L ASHLEY AND L. E. WILEY trained on the most complex maze, thus including in the records of this maze any greater difficulty which the operated animals may have had in adapting to the general training situation. 3) His mazes differed markedly in the relations of the culs de sac to the true path and so presented qualita- tively different situations. 4) His cases included a greater proportion of animals with extensive lesions and a few cases with lesions greater than our present maximum. 5) Our mazes all present the same general plan of simple right-left alterna- tion and so admit the possibility of learning by a simple generalization. The first of these differences in the experimental situations does not seem likely to have produced the differences in results, unless we assume that operated animals differ in capacity for transfer of training from normals. Such an assumption has been made by Melton ('31) concerning re- troactive inhibition, but the evidence available indicates that perseveration and consequently interference of habits is most likely in the operated animals, and this would have tended to make the simplest of Lashley's mazes (second in the series) relatively more difficult for the operated animals. The common plan of our mazes might permit of learning by simple generalization. But to account for our results on this basis, we should have to assume that the capacity to general- ize is less affected by extensive lesions than the capacity to form associations between unrelated elements — an assump- tion not in accord with Maier's experimental results ('31, '32). The available data are not adequate to decide among the remaining three possibilities. A large part of the inferiority of the operated animals on Lashley's most complex maze was evidently due to the inclusion of cases with lesions greater than 50 per cent. We have few comparable cases and it may be that the disproportionate retardation holds only for the most severely deteriorated animals. However, the influence of qualitative differences in the tasks cannot be disregarded. STUDIES OF CEREBRAL FUNCTION. IX 45 The evidence from studies of the normal growth of in- telligence in man and from clinical studies of dementia goes far toward proving that differences in capacity are in some measure qualitative. Tasks where a mere reduplication of elements is involved, as in span of attention or memorizing of nonsense syllables, do not reveal differences in intelligence brought out by batteries of qualitatively different tasks. Clinical studies, such as those of Head ('26) on semantic aphasia (interpreted as dementia by Henschen, '27), indicate that limitations to performance are set by qualitative char- acters of the tasks. We should therefore expect to find that a battery of qualitatively different tasks, graded in difficulty and making up an apparently continuous series for normal individuals, would show disproportionate difficulty at the higher levels for demented individuals in accord with the severity of the dementia. Comparison of the results with the qualitatively different tasks used by Cameron, Lashiey, and Maier suggests that this is actually the case for animals with cerebral lesions. Extensive experiments will be necessary, however, to establish the point and to reveal the nature of the effective qualitative differences, if they exist. For tasks where difficulty is determined chiefly by the re- duplication of similar elements our data seem conclusive. The difficulty of such tasks for animals with cerebral lesions increases at an accelerating rate with the extent of lesion, but the relative difficulty of the several tasks remains the same. Difficulties of interpretation arising from individual variations One of the most difficult problems in the study of cerebral functions is that of accounting for the widely divergent symptoms following similar lesions in different individuals. In our present series, as in other similar studies, we find a number of animals with extensive lesions and yet with train- ing records which are far better than the averages of other animals with equal amounts of destruction, and which may approach the records of normal animals. Such cases are nos. 46 K. S. LASHLEY AND L. E. WILEY 15, 16, 17, 19, 22, 48, 53, 73, 74, 82, 90, 112, 117, and 120. Many of these show markedly asymmetrical lesions in the two hemi- spheres, but it is possible to match almost every case with another from our series having practically identical lesions and a poor training record. Neither locus nor depth of lesion nor injuries to subcortical structures provides any apparent basis for the differences in scores. The more obvious ex- planations possible for the differences are : 1. Anatomical variation. Studies of variation in locus of cytoarchitectural areas by the senior author now in progress do not reveal such individual differences as would be required to account for the data on behavior. 2. Chance success in solving of the problems. Maze-learn- ing scores are certainly influenced to a large extent by chance factors which would seriously influence the number of errors or trials in the final score, but the very high intermaze cor- relations and the fact that many of the above animals showed superior ability in two different mazes means that, if chance determined the low scores, it was a chance discovery of some general principle of maze running, and this, although pos- sible, is difficult to fit into our present conceptions of maze learning. 3. Different animals employ differently localized cerebral mechanisms in learning the maze. One animal might be primarily dependent upon visual, another upon kinesthetic cues and the like, and a lesion in the striate area might in consequence markedly affect the former and leave the latter unaffected. Such an hypothesis is contradicted by the rela- tively slight effect of sensory privation on maze learning in comparison with the effects of lesions in cortical sensory fields (Lashley, '31 a). An alternative would be the assump- tion that the different sensory fields contribute differently to maze learning in different animals in other ways than by direct mediation of peripheral impulses. This hypothesis ap- proaches the doctrine of image types and studies of the latter have given no conclusive evidence that the image type in any way correlates with the mode of learning. It is doubtful STUDIES OF CEREBRAL FUNCTION. IX 47 that even the most extreme visual type, the eidetic, employs visual mechanisms in routine learning to a greater extent than do noneidetics. The assumption that differently local- ized functions predominate in maze learning by different animals, although possible, is not supported by any direct evidence. The difficulty of the problem is increased by the clinical evidence, especially in the field of aphasia. The conflicting evidence on localization bespeaks a condition in man like that which we find in our series of animals. Except for the pri- mary projection areas, negative cases have been reported for practically every cortical region. (Compare Monakow, '14, p. 768, for a summary of the situation on motor aphasia and Broca's area.) Negative cases in motor aphasia and similar non-sensory functions cannot be explained plausibly in terms of individual differences in the imagery used in speech. A significant point for the problem, perhaps, comes from the repeated observation that the severity and duration of symptoms from brain lesions are less in young than in old and less in intelligent than in low-grade individuals. If true, this can only mean that the severity of symptoms is dependent not only upon the locus and extent of lesion, but also upon the general level of dynamic functioning of the organism. In spite of the marked individual variation, the consistency of the results on various functions presented in table 1 and the uniform trend of the data summarized in figure 2 sug- gest that there must be some constant causal factor in con- sistency of maze performance dependent upon the mere quantity of cerebral tissue and not an artifact arising from the limitation of this or that special function. SUMMARY One hundred twenty-seven rats with cerebral lesions and 60 normal controls were trained in a maze of 8 culs de sac. The learning scores in this task have been analyzed with the following results : 48 K. S. LASHLEY AND L. E. WILEY 1. The relation between extent of lesion and retardation in maze learning is curvilinear and the error scores appear to be a logarithmic function of the extent of lesion. 2. It has been impossible to detect any influence of small injuries in the archipallium or thalamus (when combined with extensive cortical destruction) upon the maze scores. 3. The portions of the lesions, asymmetrical with respect- to the two hemispheres seem to contribute to the retardation, as well as do the symmetrical portions. 4. Lesions in all parts of the neocortex produce a marked deterioration. Our data are inconclusive with respect to the exact equipotentiality of the areas. The cases were divided into four equal groups, each of which was trained on one of four mazes with 4, 8, 12, and 16 culs de sac. The normal and operated groups were com- pared with respect to the relative difficulty of the four mazes. The operated animals were markedly retarded in all mazes, but the relative difficulty of the simple and complex mazes was the same for them as for the normal controls. This con- clusion applies to mazes in which difficulty is increased by duplicating identical elements. The problem of relative difficulty where qualitative differences are introduced remains unsettled. LITEEATUEE CITED Boumann, L., and A. A. Grunbaum 1925 Experimentelle-psychologische Unter- suchungen zur Aphasie und Paraphasie. Zschr. f. d. ges. Neurol, u. Psychiat., Bd. 96, S. 481-538. Cameron N. 1928 Cerebral destruction in its relation to maze learning. Psychol. Monogr., vol. 39, no. 1, pp. 1-68. Carr, H. A. 1917 The alternation problem: a preliminary study. J. Animal Behav., vol. 7, pp. 365-384. Fortuyn, A. E. B. D. 1914 Cortical cell-lamination of the brains of some rodents. Arch. Neurol. Psychiat., London, vol. 6, pp. 221-354. Head, H. 1926 Aphasia and kindred disorders of speech. New York. Henschen, S. E. 1927 Aphasiesysteme. Monatschr. f. Psychiat. u. Neurol., Bd. 65, S. 87-137. Hunter, W. S. 1920 The temporal maze and kinesthetic sensory processes in the white rat. Psychobiol., vol. 2, pp. 1-17. Lashley, K. S. 1917 A causal factor in the relation between the distribution of practice and the rate of learning. J. Animal Behav., vol. 7, pp. 139- 142. STUDIES OF CEREBRAL FUNCTION. IX 49 Lashley, K. S. 1920 Studies of cerebral function in learning. Psychobiol., vol. 2, pp. 55-135. 1926 Studies, etc. VII. The relation between cerebral learning, and retention. J. Coinp. Neur., vol. 41, pp. i-58. 1929 Brain mechanisms and intelligence. Chicago. 1930 The mechanism of vision. II. The influence of cerebral lesions upon the threshold of discrimination for brightness in the rat. J. Genet. Psychol., vol. 37, pp. 461-480. 1931a Cerebral control versus reflexology. J. General Psychol., vol. 5, pp. 3-20. 1931 b The mechanism of vision. IV. The cerebral areas necessary for pattern vision. J. Comp. Neur., vol. 53, pp. 419-478. 1932 Studies, etc. VIII. A reanalysis of data on mass action in the visual cortex. J. Comp. Neur., vol. 54, pp. 77-84. Lashley, K. S., and S. I. Franz 1917 The effects of cerebral destruction upon habit formation and retention in the albino rat. Psychobiol., vol. 1, pp. 71-139. Loucks, R. B. 1931 The efficacy of the rat's motor cortex in delayed alterna- tion. J. Comp. Neur., vol. 53, pp. 511-567. Maier, N. R. F. 1932 The effect of cerebral destruction on reasoning and learning in rats. J. Comp. Neur., vol. 54, pp. 45-7."). 1932 Cortical destruction of the posterior part of the brain and its effect on reasoning in rats. J. Comp. Neur., vol. 56, pp. 179-214. Melton, A. W. 1931 A note on the relation betAveen brain injury and re- tentiveness. Psychol. Bull., vol. 28, pp. 175-180. Pavlov, I. P. 1932 'The reply of a physiologist to psychologists. Psychol. K< v.. vol. 39, pp. 91-127. Thurstone, L. L. 1930 The relation between learning time and length of task Psychol. Rev., vol. 37, pp. 44-53. 1933 The error function in maze learning. Psychol. Rev. (In press.) Wiley, L. E. 1932 The function of the brain in audition. J. Comp. Neur., vol. 54, pp. 109-141. THE JOURNAL OF COMPARATIVE NEUROLOGY, VOL. 57, NO. PLATES 1 TO 5 Diagrams of the extent and locus of lesion in each of the animals reported in the present study. The diagrams are numbered to correspond to the numbering of the records in tables 4 to 7. Groups I to IV were trained, respectively, in mazes I to IV, then all cases on maze V. 50 STUDIES OF CEREBRAL FUNCTION. IX K. S. LASHL.EY AND L. E. WILEY PLATE 1 51 STUDIES OF CEREBRAL FUNCTION. IX K. S. LASHLEY AND L. E. WII^EY PLATE 2 52 STUDIES OF CEREBRAL FUNCTION. IX K. S. LASHLEY AND L. E. WIL.EY PLATE 3 n 6i- l ^ 1 ^s flfl "* jC Wm^K^ ^c_^^ »» ^- j£i^ ' STUDIES OP CEREBRAL FUNCTION. IX K. S. LASHL.EY AND Im. E. WILEY PLATE 4 54 STUDIES OF CEREBRAL FUNCTION. IX K. S. LASHLEY AND L. E. WILEY PLATE 5 1 A 55 I & Mescal QL Lashley, Karl Spencer 785 Studies of cerebral L37 function in learning v.9 PLEASE DO NOT REMOVE CARDS OR SLIPS FROM THIS POCKET UNIVERSITY OF TORONTO LIBRARY