Behavior Genetics. Vol. 6. No. 1, 1976
Genetic :and Environmental Correlations Between Brain Weight and Maze Learning in Inbred Strains of Mice and Their F~ Hybrids B . P a d e h ~ a n d M . S o l l e r 2' ~ Received4 Apr. 1973 -Final 4 Mar. 1974
The relationships among body weight, cerebellum wetght, cerebrum weight, maze-learning ability in a double T-maze, and discrimination learning in a Y-maze were studied in six inbred strains of mice and some of their Ft hybrids. The subjects were 131 male albino mice from 14 genotypic groups: five inbred groups and nine groups of crossbred offspring. Intra- and intergroup correlations were computed between all possible pairs of the anatomical and behavioral traits. A significant difference between the mtragroup and intergroup correlations for any pair of variables was taken to indicate the presence Of a genetic correlation between the two variables. On this basis, positive genetic correlations were indicated between T-maze learning ability and Y-maze learning ability, between body weight and Tmaze learning ability, and possibly between body weight and both cerebellum and cerebrum weight and between cerebrum weight and T-maze learning ability. Negative genetic correlations were indicated between cerebellum weight and running time in both mazes and between total number of successes in the Y-maze and Y-maze running time. KEY WORDS: learning; brain: genettc correlations: mice.
INTRODUCTION The relationship between brain size and learning capacity has been explored using a number of procedures, among them phyletic comparisons Supported by a grant from the Bar-Ilan University Research Council. Department of psychology, Bar-llan University, Ramat Gan, Israel, Department of Biology, Bar-llan University, Ramat Gain Israel. Present address: Department of Genetics, Hebrew University, Jerusalem, Israel. 31 (c) 1976 Plenum Publishing Corporation, 227 West 17th Street, New York, N,Y. IOOII, No part of this public!ilion may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise, without written permission of the publisher.
32
Padeh and SoRer
(Rensch, 1956; Cobb, 1965), extirpation experiments (Lashley. as cited in Beach et al., 1960), transplantation experiments (Bresler and Bitterman, 1969), and selection experiments (Wimer and Prater, 1966: Wimer et al., !969; Elias, 1969). The results are generally consistent with the notion that brain size is related somewhat to learning capacity and perhaps to other behavioral traits (but see Bassett, 1914, and Heron, 1935, for negative results). In the present study, we present additional information on this question obtained from an analysis of six inbred strains of mice and crosses between them. MATERIALS
AND
METHODS
Subjects The experimental population consisted of 131 male albino mice. These were produced in two series of crosses; the first series involved four stratus of mice, the second three. The strains used in these crosses were as follows S W R and B A L B / c were obtained from the Weizmann Institute of Science. Rehovot, Israel. These strains came to the Weizmann Institute from The Jackson Laboratories, Bar Harbor, Maine, in t959 and were maintained thereafter by inbreeding. Swiss NZ and T O were obtained from the Israel Table I. Strains Involved in Each Series of Crosses "rod the Number of Male Offspring Ob(:tined from Each Type of Mating" [*'emttte t)arenl
Male parelll
l,'irst series of crosses
TO Swissll Swiss NZ Swiss i
T()
Swiss I1
Swiss NZ
Swiss I
2 3 (1) 7 (1) 9
3 4 (1) 3 (2) 8
~.} 0 5 (1) 6 (2)
0 3 5 0
Second series of crosses BALB/c BALB/c SWR Swiss NZ
0 3 12 (2)
SWR 0 ,S 10
Swiss NZ 0 8 (2) 9
" In parentheses: number (,f mi(.e for which only behavior,'d data were obi'tined.
Genetic and Environmental Correlations Between Brain V~eight and Mate Learning
33
Institute for Biological Research, Nes Ziona, Israel. Swiss NZ originated from a strain designated Swiss obtained from the Pasteur Institute in France in 1951. T O was obtained from the Medical Research Council, Millhill, England, in 1956. These strains were maintained by mass mating. SWR, BALB/c, Swiss NZ, and T O are listed under the appropriate Israel institutions in the International Index o/ Laboratory Animals (! 970). Swiss I and Swiss II were obtained from Beilinson Hospital, Petach Tikva, Israel. Both strains are sublines of Swiss NZ. Swiss I originated from animals obtained from the Israel Institute of Biological Research in 1958 and were maintained thereafter by inbreeding. Swiss II originated from animals obtained in 1961 and were maintained thereafter by mass mating. Table I shows the strains and the total number of male offspring for each cross. A number of crosses were not fertile; in particular, females of the BALB/c strain were infertile in all combinations. Offspring from reciprocal matings were considered to form one genotypic groups, so that in all there were 14 genotypic groups: five groups of inbred offspring and nine groups of crossbred offspring. Construction and learning to operate the mazes took more time than anticipated. As a result, mice of the first series were tested at the age of 410--620 days and those of the second series at 290-430 days.
Apparatus The m i c e were tested in two mazes: a double T-maze, in which the mouse had to traverse a given path from start to goal box, and a discrimination Y-maze, in which the mouse had to learn to associate a light signal with the arm of the maze leading to the goal box. In both cases, electrical shock of 80 V was applied via an electrified floor. The mazes were constructed of wood and were painted with lacquer. The ceiling was made of glass plates. Clear plastic doors of the guillotine type were mounted at the exit of tile start box and at the entrance of the goal box. Clear plastic doors were fixed at the end of the blind arms of the T-maze. Each stem of the T-maze was 5 cm wide, 10 cm high, and 80 cm long. Arm length in the Y-maze was 40 cm, the goal boxes were 20 by 15 cm, and the start box was rhomboid-shaped: 12 cm at the base and ! 5 cm at the entrance to the maze.
Procedure After weaning, the male mice of each cross were raised in the same cage until the age of 3 months. From that time, each mouse was kept in a separate cage until the end of the experiment. The experimental animals were chosen for testing in a random stratified manner in groups of 10-12 mice. Each group was tested over a
34
Padeh and Soller
period of 2 weeks, first in the T-maze and then in the Y-maze. At the end of this period, the animals were killed and their body weights and brain weights were recorded. Each series of trials for each maze involved three sets of runs carried out on alternate days. In the T-maze. the mouse was run six times each day, with 20-sec intertrial intervals, for a total of 18 runs. In the Y-maze. the mouse was run eight times each day, with 20-see intertrial intervals, for a total of 24 runs. A "white noise" generator was operated during the maze learning, and the runs were carried out under artificial light, with the place of the mazes fixed so as to maintain constant visual clues. The following measures of maze learning ability were taken: the number of runs without error (S, also henceforth termed "successes"), the average number of errors per run (E), and the average time per run (T), with a maximum of 3 rain allowed. Anatomical measurements were taken as follows: The animal was anesthetized with ether and weighed. It was then decapitated and the cerebrum was separated in situ from the olfactory lobes and. by a cut along the transverse fissure, from the cerebellum. The cerebellum was separated f r o m the brain stem by a cut along its posterior margin. The cerebrum and cerebellum were removed from the cranium and weighed on an analytical torsion balance. Thus the variables considered in this study were body weight, cerebellum weight, cerebrum weight. T-maze successes (TS), Tmaze errors (TE), T-maze time (TT), Y-maze successes (Y,S), Y-maze errors (YE), and Y-maze time (YT),
Statistical Analyses The significance of differences between offspring groups for the various anatomical and behavioral traits was tested by means of a one-way, unequal-number analysis of variance and covariance. Data from the first and second Series of crosses were analyzed separately. The primary statistics used in the analysis of these data were the intragenotypic-group and inter-genotypic-group correlations coefficients b e t w e e n the various traits. The intragroup correlations were calculated separately for each group and then combined into a single estimate by pooling variances and covariances. The intergroup correlations were based on the group means. Exact tests of significance for matrices of results as were obtained in this study require multiple comparison techniques and are difficult to calculate. Instead, we decided to treat as noteworthy any correlation coefficient, or difference between correlation coefficients, that would have been significant at the 5% level, using Fisher's Zr transformation. Such results will be termed "significant," without, however, specifying a particular type I error.
Genetic and Environmental Correlations Between Brain ~Aeight and Maze Learning
35
Two basic assumptions were made in interpreting the correlation coefficients: (1) that each group was isogenic and (2) that the only environmental differences between the genotypic groups were those due to sampling variations in the distribution of random environmental factors among the individual mice in the group. These assumptions seem reasonable in view of the origin of the inbred lines and the precautions taken to eliminate systematic environmental differences between the various offspring groups. On the first assumption, the intragroup correlation (rw) between two traits is an estimate of the environmental correlation (rE) between them. On the second assumption, it can readily be shown that the intergroup correlation (rt) is also an estimate of rE, uniess there are group-specific nonenvironmental factors causing the members of a genotypic group to vary in a similar manner. Thus a significant difference between rw and rx would indicate that there were such group-specific nonenvironmental factors affecting the traits under consideration. We shall assume such factors to be genetic in nature, although maternal effects are a possibility. In the presence of such group-specific genetic factors, the intergroup correlation would have the following composition: r, =
Coy A,s ~ Coy E,~lh (Vat A, + Var E~/h) '/2 (Vat AI + Vat E j / h ) ~'2
(1)
where Coy A~j is the covariance component due to genetic factors that differ with respect to the genotypic groups, Coy E o is the covariance component d u e to environmental factors that influence t r a i t s / a n d j on the same individual. Var A is the variance component due to genetic factors that differ with respect to the genotypic groups, Var E is the variance component due to environmental factors that are random with respect to genotyplc groups, and ~ is the average number of animals in the various genotypic groups. A somewhat different derivation, and more detailed discussion of this result, for the case n = I is given in McClearn et al. (1970). Examination of this expression shows that in the absence of any genetic covariance between the two traits (i.e., when Coy A,j = 0) the sign of r~ and rw should be the same, although r~ might be less than rw to an extent dependent on the magnitude of Var A, and Var Aj. Thus, on these assumptions, a genetic correlation between the two variables would be indicated whenever (1) rw and r~ have the Same sign, but r~ is significantly greater than rw, and (2) rw and rt have opposite signs, with either rx or rw significantly different from zero. In these cases, the sign of (r~ - rw) would give the sign of the genetic correlation. If, in addition, rw were close to zero, then Cov E J h would probably be rather small, and rl would be a lower estimate of the genetic correlation, ra (Falconer, 1960). When rw is significantly different from rl (in the manner described
36
Padeh and Soller
above), ra :~ 0; however, the converse is not true. Numerical substitution in equation (1) shows that there are a number of conditions in which ra = 0 and rw t: r~. The specific values of ra for which this will be true depend on the heritabilities of the two traits, and on the number of offspring in each genotypic group. Generally speaking, however, when genetic and environmental correlations have the same sign, and are of the same order of magnitude, rr will not be very different than rw, and the experimental approach described here will not be able to detect the presence of the genetic correlation. Thus this approach will be most effective when environmental correlations are close to zero. In addition to pleiotropy, genetic correlations calculated from data on strains and their crosses can also be generated (I) if the strains tested do not rePresent a random sample from a population in genetic equilibrium with respect to the traits under consideration and (2) if heterotic effects affect both traits simultaneously. We have no way of examining the possible lack of genetic equilibrium among the strains of this study. Heterotic effects, however, were not present for the behavioral traits (Padeh, 1966). Hence we would not expect them to contribute to the correlations involving these traits. R ESU LTS Table I1 shows the mean values for the various traits in both series of crosses separately, as well as the level of significance of the between-family T a b l e II. Mean Values of Anatomical and Behavioral Traits and Significance Level of Differences Between Genotypi(' Groups in the Two Series of Crosses Trail, Age ((lays) Body weight (g) Cerebellum weighl (rag) Cerebrmn weight (rag)
First series
Second series
510 32,4 " 139.0 283,2 b
360 26.8" 138.8 265.9 b
T-maze Successes (total number) Errors (number/run) T i m e (see/run)
7.15 b 1.73 ~ 55.98
5.25 2.01~ 68.05
Y-maze Successes (total number) Errors (number/run) Time (sec/run)
11.97 0.82 30.42
12.12 0.79 24.18
" P < 0.01. b p < 0.05.
Genetic and Environmental Correlations Between Brain Weight and Maze Learning
37
0
I
0~
I I
2
C~
1 ~
,
~
II
I
~
II
I
I
I
I
~
I
~ O
II
I
I
I
C)
9
r
I
I
I
I
0 c~_.~
oO
I
I I
I I
.=. -~
Ia
~
F~
F~
F~
T
38
Padeh and Soller
differences. Significant differences between the groups were found for all of the variables except cerebellum weight, TT, YS, YE, and YT. Both series of crosses were remarkably similar (except for age) in their mean values, and in the traits for which significant differences were found. For this reason, the correlation coefficients estimated separately from each series of crosses were pooled, and the subsequent results are based on the pooled estimates. These pooled intra- and intergroup correlations among the various anatomical and behavioral traits are presented in Table III. Anatomical Traits The intragroup correlations between body weight and both cerebrum and cerebellum weight were moderately high and significant. T h u s environmental factors that led to greater body weight also led to increased cerebellum and cerebrum weight. The partial correlations (body weight held constant) between cerebrum and cerebellum weight were 0.20 for the intragroup correlation and 0.43 for the intergroup correlation. The mtragroup correlation was highly significant and indicates that there are environmental factors affecting both cerebrum and cerebellum weight, independent Of their effect on body weight. Behavioral Traits There was only a small intragroup (i.e., environmental) correlation between TS and YS. In contrast, the intergroup correlation between TS and YS was significant and much higher than the corresponding intragroup correlation. This implies a high genetic correlation between successes in both mazes. Similar results were obtained for TE and YE. for TE and YS, and for Ts and YE. There were significant and moderately high negative intragroup correlations between TS and TT and between YS and YT. These correlations are probably due to the fact that the more successful a mouse is in negotiating the maze, the less time it takes. There was also a moderately high and significant intragroup correlation between TT and YT, indicating that some of the environmental factors affecting time in one maze affected time in a similar manner in the other maze. The intergroup correlation between YS and YT was positive and significant. Taken in conjunction with the significant negative intragroup correlation, this implies the existence of a strong positive genetic correlation between successes and time in the Y-maze. That is, strains that made fewer errors took more time. This paradoxical result may have been due to the structure of our Y-maze. It was observed that mice which took more time at the decision point to evaluate the relevant signal made fewer errors.
Genetic and Environmental Correlations Between Brain Weight and Maze Learning
39
Anatomical Traits and Behavioral Traits
The intragroup correlations between the anatomical and behavioral traits were all very small and nonsignificant, except for those between body weight and time in both mazes, which were low but significant. The most striking aspect of the intergroup correlations was the high and significant correlations between body weight and TS, TE and TT. These correlations remain high and significant even when cerebrum weight is held constant, and thus are not mediated via brain size. A rationale for these effects is not immediately apparent. For the T-maze, the intergroup correlations between cerebrum weight and TS, TE and TT were significant (at the 5-10% level) and significantly different from the very small intragroup correlations as well. This would indicate a moderate to strong genetic correlation between cerebrum weight and maze performance. However, the partial correlations (body weight held constant) between cerebrum weight and the T-maze running measures were 0.35, 0.27. and -0.31 for TE. TS, and TT, respectively, and were no longer significant. For the Y-maze. the intergroup correlations between cerebrum weight and YS, YE, and YT are small and become even smaller when the partial correlations, with body weight held constant, are calculated (0.13, -0.19. and 0.11, respectively). The intergroup correlations between cerebellum weight and TS, TE, YS, and YT are at best moderately high. When the partial correlations with body weight held Constant are calculated, they become small (-0.02, -0.05, -0.33, and -0.16, respectively). In contrast, the intergroup correlations between cerebellum weight and time in the various mazes (TT and YT) are high and significant, and this significance is retained even when body weight is held constant (partial correlation coefficients of -0.75 for T T and -0.48 for YT). Apparently, high cerebellum weight due to genetic factors was accompanied by decreased running time, independent of body weight.
DISCUSSION An interesting result of the maze-learning analyses was the high genetic correlation implied between maze-learning ability as measured by successes and errors in the two dissimilar mazes. This correlation may indicate the presence of genetic factors having a general effect on mazelearning ability. This result stands in some contrast to previous work along these lines. Searle (1949) tested Tryon's maze-bright and maze-dull lines and found that in three of five mazes the dulls performed better than the brights. Similarly, Myers (1959) found that the relative performance of two strains of mice in an avoidance conditioning experiment changed in accordance with the stimulus and motivation. Perhaps the fact that in this study
40
Padeh and Soller
the motivation (electrical shock) was the same in both mazes contributed to the genetic correlation between performance in both mazes. However, if differential sensitivity to electrical shock were the reason for the correlated performance of the genotypic groups in the two mazes, we would have expected a high intergroup correlation between TT and YT. particularly in view of the moderately high intragroup correlation. This was not found. The small intragroup correlations between cerebrum weight and mazelearning ability imply ~that within the limited range of environments provided in this experiment, environmental factors affecting cerebral weight had no effect on maze-learning ability. In contrast, the moderately high intergroup partial correlations between cerebrum weight and T-maze learning ability may be taken to provide some evidence indicating that increased cerebral weight produced by genetic factors is correlated with the simple maze-learning ability involved in a T-maze. However, causality may not be inferred. The larger brain may cause superior maze-learning ability, or, conversely, superior maze-learning ability may cause larger brain size (Bennett et al., 1964). In our results, cerebrum weight seems to be only marginally, if at all, related to the ability to handle more complex problems such as are involved in learning a discrimination type of Y-maze. In contrast, Wimer and Prater (1966), Wimer et al. (1969), and Elias (1969) working with the Roderick-Wimer brain weight lines (Fuller and Wimer, 1973) did find a positive association between brain weight and discrimination learning performance. The most positive indication of a relationship between brain size and a behavioral trait obtained in this experiment was the high intergroup and low intragroup correlation between cerebellum weight and time in both mazes. These results indicate that higher cerebellum weight in our genotypic groups was accompanied by shorter running time. ACKNOWLEDGMENT The authors acknowledge with thanks the assistance and advice of Dr. Y. Lewin in the design and implementation of the experiment. They also wish to thank Dr. A. Genizi for his assistance in organizing and carrying out the statistical analyses. REFERENCES Bassett, G. C. (1914). Habit formation in a strain of albino rats of less than normal brain weight. Behav. Monogr. 2:21-46. Cited by McClearn, G. E. (1962). The inheritance of
behavior. In Postman, L. (ed.), Psychology in the Making, Knopf,New York.
Genetic and Environmental Correlations Between Brain Weight and Maze Learning
41
Beach, F. A., Hebb, D. O., Morgan, C. T., and Nissen, H. W. (1960), The Neuroanatomy o[ Lashley, McGraw-Hill, New York. Bennett, E. L., Diamond, M., Krech, D., and Rosenzweig, M. R. (1964). Chemical and anatomical plasticity of the brain. Science 146:610 619. Bresler, D. E., and Bitterman, M. E. (1969). Learning in fish with transplanted brain tissue. Science 163:590 592. Cobb, S. (1965). Brain size. Arch. Neurol. 12:555 561~ Elias, M. F. (1969). Differences in spatial discrimination reversal learning for mice genetically selected for high brain weight and unselected controls. Percept. Motor Skills 28:707 712, Falconer, D. S. (1960). Introduction to Quantitative Genetics, Oliver and Boyd, Edinburgh. Fuller, J. L., and Wimer, R, E. (1973). Behavior Genetics. In Dewsbury, D. A., and Rethlingshafer, D. A. (eds.), Comparative Psychologw A Modern Survey. McGraw-Hill, New York. Heron, W. T. (1935). The inheritance of maze-learning ability in the rat. J. Comp. Psychol. 19:77-89. Cited by Fuller, J. L., and Thompson, W. R, (1960). Behavior Genetics, Wiley, New York. International Index ~[ Laboratory Animals (1970). Compiled and distributed by Medical Research Laboratory Animals Centre. Canshalton, Surrey, U.K. McClearn, G. E., Wilson, J. R., and Meredith, W. (1970). The use of isogenic and heterogenic mouse stocks in behavioral research. In Lindzey, G., and Thiessen, D. E. (eds.), Contributions to Behavior-Genetic Analysis: The Mouse as a ProtOtype. AppletonCentury-Crofts, New York. Myers, A. K. (1959). Avoidance learning as a function of several differences in rats. J. Comp. Physiol. Psychol. 52:381 386. Padeh, B. (1966). Genetic and environmental correlations between maze-learning and brain size and RNA content. M,Sc. Thesis, Bar-llan University, Ramat Gan, Israel (in Hebrew). Rensch, B. (1956). Increase of learning capability with increase of brain size. Am. Naturalist 90:81.-95. Searle, L. V. (1949). The organization of hereditary maze-brightness and maze-dullness, Genet. Psychol. Monogr. 39:279 325. Wimer, C., and Prater, L. (1966). Some behavioral differences in mice genetica'lly selected for high and low brain weight, Psychol. Rep, 19:675-681. Wimer, C., Roderick, T. H., and Wimer, R. E. (1969). Supplementary report: Behavioral differences in mice genetically selected for brain weight. Psychol. Rep. 25:363-368. [
Genetic :and Environmental Correlations Between Brain Weight and Maze Learning in Inbred Strains of Mice and Their F~ Hybrids B . P a d e h ~ a n d M . S o l l e r 2' ~ Received4 Apr. 1973 -Final 4 Mar. 1974
The relationships among body weight, cerebellum wetght, cerebrum weight, maze-learning ability in a double T-maze, and discrimination learning in a Y-maze were studied in six inbred strains of mice and some of their Ft hybrids. The subjects were 131 male albino mice from 14 genotypic groups: five inbred groups and nine groups of crossbred offspring. Intra- and intergroup correlations were computed between all possible pairs of the anatomical and behavioral traits. A significant difference between the mtragroup and intergroup correlations for any pair of variables was taken to indicate the presence Of a genetic correlation between the two variables. On this basis, positive genetic correlations were indicated between T-maze learning ability and Y-maze learning ability, between body weight and Tmaze learning ability, and possibly between body weight and both cerebellum and cerebrum weight and between cerebrum weight and T-maze learning ability. Negative genetic correlations were indicated between cerebellum weight and running time in both mazes and between total number of successes in the Y-maze and Y-maze running time. KEY WORDS: learning; brain: genettc correlations: mice.
INTRODUCTION The relationship between brain size and learning capacity has been explored using a number of procedures, among them phyletic comparisons Supported by a grant from the Bar-Ilan University Research Council. Department of psychology, Bar-llan University, Ramat Gan, Israel, Department of Biology, Bar-llan University, Ramat Gain Israel. Present address: Department of Genetics, Hebrew University, Jerusalem, Israel. 31 (c) 1976 Plenum Publishing Corporation, 227 West 17th Street, New York, N,Y. IOOII, No part of this public!ilion may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise, without written permission of the publisher.
32
Padeh and SoRer
(Rensch, 1956; Cobb, 1965), extirpation experiments (Lashley. as cited in Beach et al., 1960), transplantation experiments (Bresler and Bitterman, 1969), and selection experiments (Wimer and Prater, 1966: Wimer et al., !969; Elias, 1969). The results are generally consistent with the notion that brain size is related somewhat to learning capacity and perhaps to other behavioral traits (but see Bassett, 1914, and Heron, 1935, for negative results). In the present study, we present additional information on this question obtained from an analysis of six inbred strains of mice and crosses between them. MATERIALS
AND
METHODS
Subjects The experimental population consisted of 131 male albino mice. These were produced in two series of crosses; the first series involved four stratus of mice, the second three. The strains used in these crosses were as follows S W R and B A L B / c were obtained from the Weizmann Institute of Science. Rehovot, Israel. These strains came to the Weizmann Institute from The Jackson Laboratories, Bar Harbor, Maine, in t959 and were maintained thereafter by inbreeding. Swiss NZ and T O were obtained from the Israel Table I. Strains Involved in Each Series of Crosses "rod the Number of Male Offspring Ob(:tined from Each Type of Mating" [*'emttte t)arenl
Male parelll
l,'irst series of crosses
TO Swissll Swiss NZ Swiss i
T()
Swiss I1
Swiss NZ
Swiss I
2 3 (1) 7 (1) 9
3 4 (1) 3 (2) 8
~.} 0 5 (1) 6 (2)
0 3 5 0
Second series of crosses BALB/c BALB/c SWR Swiss NZ
0 3 12 (2)
SWR 0 ,S 10
Swiss NZ 0 8 (2) 9
" In parentheses: number (,f mi(.e for which only behavior,'d data were obi'tined.
Genetic and Environmental Correlations Between Brain V~eight and Mate Learning
33
Institute for Biological Research, Nes Ziona, Israel. Swiss NZ originated from a strain designated Swiss obtained from the Pasteur Institute in France in 1951. T O was obtained from the Medical Research Council, Millhill, England, in 1956. These strains were maintained by mass mating. SWR, BALB/c, Swiss NZ, and T O are listed under the appropriate Israel institutions in the International Index o/ Laboratory Animals (! 970). Swiss I and Swiss II were obtained from Beilinson Hospital, Petach Tikva, Israel. Both strains are sublines of Swiss NZ. Swiss I originated from animals obtained from the Israel Institute of Biological Research in 1958 and were maintained thereafter by inbreeding. Swiss II originated from animals obtained in 1961 and were maintained thereafter by mass mating. Table I shows the strains and the total number of male offspring for each cross. A number of crosses were not fertile; in particular, females of the BALB/c strain were infertile in all combinations. Offspring from reciprocal matings were considered to form one genotypic groups, so that in all there were 14 genotypic groups: five groups of inbred offspring and nine groups of crossbred offspring. Construction and learning to operate the mazes took more time than anticipated. As a result, mice of the first series were tested at the age of 410--620 days and those of the second series at 290-430 days.
Apparatus The m i c e were tested in two mazes: a double T-maze, in which the mouse had to traverse a given path from start to goal box, and a discrimination Y-maze, in which the mouse had to learn to associate a light signal with the arm of the maze leading to the goal box. In both cases, electrical shock of 80 V was applied via an electrified floor. The mazes were constructed of wood and were painted with lacquer. The ceiling was made of glass plates. Clear plastic doors of the guillotine type were mounted at the exit of tile start box and at the entrance of the goal box. Clear plastic doors were fixed at the end of the blind arms of the T-maze. Each stem of the T-maze was 5 cm wide, 10 cm high, and 80 cm long. Arm length in the Y-maze was 40 cm, the goal boxes were 20 by 15 cm, and the start box was rhomboid-shaped: 12 cm at the base and ! 5 cm at the entrance to the maze.
Procedure After weaning, the male mice of each cross were raised in the same cage until the age of 3 months. From that time, each mouse was kept in a separate cage until the end of the experiment. The experimental animals were chosen for testing in a random stratified manner in groups of 10-12 mice. Each group was tested over a
34
Padeh and Soller
period of 2 weeks, first in the T-maze and then in the Y-maze. At the end of this period, the animals were killed and their body weights and brain weights were recorded. Each series of trials for each maze involved three sets of runs carried out on alternate days. In the T-maze. the mouse was run six times each day, with 20-sec intertrial intervals, for a total of 18 runs. In the Y-maze. the mouse was run eight times each day, with 20-see intertrial intervals, for a total of 24 runs. A "white noise" generator was operated during the maze learning, and the runs were carried out under artificial light, with the place of the mazes fixed so as to maintain constant visual clues. The following measures of maze learning ability were taken: the number of runs without error (S, also henceforth termed "successes"), the average number of errors per run (E), and the average time per run (T), with a maximum of 3 rain allowed. Anatomical measurements were taken as follows: The animal was anesthetized with ether and weighed. It was then decapitated and the cerebrum was separated in situ from the olfactory lobes and. by a cut along the transverse fissure, from the cerebellum. The cerebellum was separated f r o m the brain stem by a cut along its posterior margin. The cerebrum and cerebellum were removed from the cranium and weighed on an analytical torsion balance. Thus the variables considered in this study were body weight, cerebellum weight, cerebrum weight. T-maze successes (TS), Tmaze errors (TE), T-maze time (TT), Y-maze successes (Y,S), Y-maze errors (YE), and Y-maze time (YT),
Statistical Analyses The significance of differences between offspring groups for the various anatomical and behavioral traits was tested by means of a one-way, unequal-number analysis of variance and covariance. Data from the first and second Series of crosses were analyzed separately. The primary statistics used in the analysis of these data were the intragenotypic-group and inter-genotypic-group correlations coefficients b e t w e e n the various traits. The intragroup correlations were calculated separately for each group and then combined into a single estimate by pooling variances and covariances. The intergroup correlations were based on the group means. Exact tests of significance for matrices of results as were obtained in this study require multiple comparison techniques and are difficult to calculate. Instead, we decided to treat as noteworthy any correlation coefficient, or difference between correlation coefficients, that would have been significant at the 5% level, using Fisher's Zr transformation. Such results will be termed "significant," without, however, specifying a particular type I error.
Genetic and Environmental Correlations Between Brain ~Aeight and Maze Learning
35
Two basic assumptions were made in interpreting the correlation coefficients: (1) that each group was isogenic and (2) that the only environmental differences between the genotypic groups were those due to sampling variations in the distribution of random environmental factors among the individual mice in the group. These assumptions seem reasonable in view of the origin of the inbred lines and the precautions taken to eliminate systematic environmental differences between the various offspring groups. On the first assumption, the intragroup correlation (rw) between two traits is an estimate of the environmental correlation (rE) between them. On the second assumption, it can readily be shown that the intergroup correlation (rt) is also an estimate of rE, uniess there are group-specific nonenvironmental factors causing the members of a genotypic group to vary in a similar manner. Thus a significant difference between rw and rx would indicate that there were such group-specific nonenvironmental factors affecting the traits under consideration. We shall assume such factors to be genetic in nature, although maternal effects are a possibility. In the presence of such group-specific genetic factors, the intergroup correlation would have the following composition: r, =
Coy A,s ~ Coy E,~lh (Vat A, + Var E~/h) '/2 (Vat AI + Vat E j / h ) ~'2
(1)
where Coy A~j is the covariance component due to genetic factors that differ with respect to the genotypic groups, Coy E o is the covariance component d u e to environmental factors that influence t r a i t s / a n d j on the same individual. Var A is the variance component due to genetic factors that differ with respect to the genotypic groups, Var E is the variance component due to environmental factors that are random with respect to genotyplc groups, and ~ is the average number of animals in the various genotypic groups. A somewhat different derivation, and more detailed discussion of this result, for the case n = I is given in McClearn et al. (1970). Examination of this expression shows that in the absence of any genetic covariance between the two traits (i.e., when Coy A,j = 0) the sign of r~ and rw should be the same, although r~ might be less than rw to an extent dependent on the magnitude of Var A, and Var Aj. Thus, on these assumptions, a genetic correlation between the two variables would be indicated whenever (1) rw and r~ have the Same sign, but r~ is significantly greater than rw, and (2) rw and rt have opposite signs, with either rx or rw significantly different from zero. In these cases, the sign of (r~ - rw) would give the sign of the genetic correlation. If, in addition, rw were close to zero, then Cov E J h would probably be rather small, and rl would be a lower estimate of the genetic correlation, ra (Falconer, 1960). When rw is significantly different from rl (in the manner described
36
Padeh and Soller
above), ra :~ 0; however, the converse is not true. Numerical substitution in equation (1) shows that there are a number of conditions in which ra = 0 and rw t: r~. The specific values of ra for which this will be true depend on the heritabilities of the two traits, and on the number of offspring in each genotypic group. Generally speaking, however, when genetic and environmental correlations have the same sign, and are of the same order of magnitude, rr will not be very different than rw, and the experimental approach described here will not be able to detect the presence of the genetic correlation. Thus this approach will be most effective when environmental correlations are close to zero. In addition to pleiotropy, genetic correlations calculated from data on strains and their crosses can also be generated (I) if the strains tested do not rePresent a random sample from a population in genetic equilibrium with respect to the traits under consideration and (2) if heterotic effects affect both traits simultaneously. We have no way of examining the possible lack of genetic equilibrium among the strains of this study. Heterotic effects, however, were not present for the behavioral traits (Padeh, 1966). Hence we would not expect them to contribute to the correlations involving these traits. R ESU LTS Table I1 shows the mean values for the various traits in both series of crosses separately, as well as the level of significance of the between-family T a b l e II. Mean Values of Anatomical and Behavioral Traits and Significance Level of Differences Between Genotypi(' Groups in the Two Series of Crosses Trail, Age ((lays) Body weight (g) Cerebellum weighl (rag) Cerebrmn weight (rag)
First series
Second series
510 32,4 " 139.0 283,2 b
360 26.8" 138.8 265.9 b
T-maze Successes (total number) Errors (number/run) T i m e (see/run)
7.15 b 1.73 ~ 55.98
5.25 2.01~ 68.05
Y-maze Successes (total number) Errors (number/run) Time (sec/run)
11.97 0.82 30.42
12.12 0.79 24.18
" P < 0.01. b p < 0.05.
Genetic and Environmental Correlations Between Brain Weight and Maze Learning
37
0
I
0~
I I
2
C~
1 ~
,
~
II
I
~
II
I
I
I
I
~
I
~ O
II
I
I
I
C)
9
r
I
I
I
I
0 c~_.~
oO
I
I I
I I
.=. -~
Ia
~
F~
F~
F~
T
38
Padeh and Soller
differences. Significant differences between the groups were found for all of the variables except cerebellum weight, TT, YS, YE, and YT. Both series of crosses were remarkably similar (except for age) in their mean values, and in the traits for which significant differences were found. For this reason, the correlation coefficients estimated separately from each series of crosses were pooled, and the subsequent results are based on the pooled estimates. These pooled intra- and intergroup correlations among the various anatomical and behavioral traits are presented in Table III. Anatomical Traits The intragroup correlations between body weight and both cerebrum and cerebellum weight were moderately high and significant. T h u s environmental factors that led to greater body weight also led to increased cerebellum and cerebrum weight. The partial correlations (body weight held constant) between cerebrum and cerebellum weight were 0.20 for the intragroup correlation and 0.43 for the intergroup correlation. The mtragroup correlation was highly significant and indicates that there are environmental factors affecting both cerebrum and cerebellum weight, independent Of their effect on body weight. Behavioral Traits There was only a small intragroup (i.e., environmental) correlation between TS and YS. In contrast, the intergroup correlation between TS and YS was significant and much higher than the corresponding intragroup correlation. This implies a high genetic correlation between successes in both mazes. Similar results were obtained for TE and YE. for TE and YS, and for Ts and YE. There were significant and moderately high negative intragroup correlations between TS and TT and between YS and YT. These correlations are probably due to the fact that the more successful a mouse is in negotiating the maze, the less time it takes. There was also a moderately high and significant intragroup correlation between TT and YT, indicating that some of the environmental factors affecting time in one maze affected time in a similar manner in the other maze. The intergroup correlation between YS and YT was positive and significant. Taken in conjunction with the significant negative intragroup correlation, this implies the existence of a strong positive genetic correlation between successes and time in the Y-maze. That is, strains that made fewer errors took more time. This paradoxical result may have been due to the structure of our Y-maze. It was observed that mice which took more time at the decision point to evaluate the relevant signal made fewer errors.
Genetic and Environmental Correlations Between Brain Weight and Maze Learning
39
Anatomical Traits and Behavioral Traits
The intragroup correlations between the anatomical and behavioral traits were all very small and nonsignificant, except for those between body weight and time in both mazes, which were low but significant. The most striking aspect of the intergroup correlations was the high and significant correlations between body weight and TS, TE and TT. These correlations remain high and significant even when cerebrum weight is held constant, and thus are not mediated via brain size. A rationale for these effects is not immediately apparent. For the T-maze, the intergroup correlations between cerebrum weight and TS, TE and TT were significant (at the 5-10% level) and significantly different from the very small intragroup correlations as well. This would indicate a moderate to strong genetic correlation between cerebrum weight and maze performance. However, the partial correlations (body weight held constant) between cerebrum weight and the T-maze running measures were 0.35, 0.27. and -0.31 for TE. TS, and TT, respectively, and were no longer significant. For the Y-maze. the intergroup correlations between cerebrum weight and YS, YE, and YT are small and become even smaller when the partial correlations, with body weight held constant, are calculated (0.13, -0.19. and 0.11, respectively). The intergroup correlations between cerebellum weight and TS, TE, YS, and YT are at best moderately high. When the partial correlations with body weight held Constant are calculated, they become small (-0.02, -0.05, -0.33, and -0.16, respectively). In contrast, the intergroup correlations between cerebellum weight and time in the various mazes (TT and YT) are high and significant, and this significance is retained even when body weight is held constant (partial correlation coefficients of -0.75 for T T and -0.48 for YT). Apparently, high cerebellum weight due to genetic factors was accompanied by decreased running time, independent of body weight.
DISCUSSION An interesting result of the maze-learning analyses was the high genetic correlation implied between maze-learning ability as measured by successes and errors in the two dissimilar mazes. This correlation may indicate the presence of genetic factors having a general effect on mazelearning ability. This result stands in some contrast to previous work along these lines. Searle (1949) tested Tryon's maze-bright and maze-dull lines and found that in three of five mazes the dulls performed better than the brights. Similarly, Myers (1959) found that the relative performance of two strains of mice in an avoidance conditioning experiment changed in accordance with the stimulus and motivation. Perhaps the fact that in this study
40
Padeh and Soller
the motivation (electrical shock) was the same in both mazes contributed to the genetic correlation between performance in both mazes. However, if differential sensitivity to electrical shock were the reason for the correlated performance of the genotypic groups in the two mazes, we would have expected a high intergroup correlation between TT and YT. particularly in view of the moderately high intragroup correlation. This was not found. The small intragroup correlations between cerebrum weight and mazelearning ability imply ~that within the limited range of environments provided in this experiment, environmental factors affecting cerebral weight had no effect on maze-learning ability. In contrast, the moderately high intergroup partial correlations between cerebrum weight and T-maze learning ability may be taken to provide some evidence indicating that increased cerebral weight produced by genetic factors is correlated with the simple maze-learning ability involved in a T-maze. However, causality may not be inferred. The larger brain may cause superior maze-learning ability, or, conversely, superior maze-learning ability may cause larger brain size (Bennett et al., 1964). In our results, cerebrum weight seems to be only marginally, if at all, related to the ability to handle more complex problems such as are involved in learning a discrimination type of Y-maze. In contrast, Wimer and Prater (1966), Wimer et al. (1969), and Elias (1969) working with the Roderick-Wimer brain weight lines (Fuller and Wimer, 1973) did find a positive association between brain weight and discrimination learning performance. The most positive indication of a relationship between brain size and a behavioral trait obtained in this experiment was the high intergroup and low intragroup correlation between cerebellum weight and time in both mazes. These results indicate that higher cerebellum weight in our genotypic groups was accompanied by shorter running time. ACKNOWLEDGMENT The authors acknowledge with thanks the assistance and advice of Dr. Y. Lewin in the design and implementation of the experiment. They also wish to thank Dr. A. Genizi for his assistance in organizing and carrying out the statistical analyses. REFERENCES Bassett, G. C. (1914). Habit formation in a strain of albino rats of less than normal brain weight. Behav. Monogr. 2:21-46. Cited by McClearn, G. E. (1962). The inheritance of
behavior. In Postman, L. (ed.), Psychology in the Making, Knopf,New York.
Genetic and Environmental Correlations Between Brain Weight and Maze Learning
41
Beach, F. A., Hebb, D. O., Morgan, C. T., and Nissen, H. W. (1960), The Neuroanatomy o[ Lashley, McGraw-Hill, New York. Bennett, E. L., Diamond, M., Krech, D., and Rosenzweig, M. R. (1964). Chemical and anatomical plasticity of the brain. Science 146:610 619. Bresler, D. E., and Bitterman, M. E. (1969). Learning in fish with transplanted brain tissue. Science 163:590 592. Cobb, S. (1965). Brain size. Arch. Neurol. 12:555 561~ Elias, M. F. (1969). Differences in spatial discrimination reversal learning for mice genetically selected for high brain weight and unselected controls. Percept. Motor Skills 28:707 712, Falconer, D. S. (1960). Introduction to Quantitative Genetics, Oliver and Boyd, Edinburgh. Fuller, J. L., and Wimer, R, E. (1973). Behavior Genetics. In Dewsbury, D. A., and Rethlingshafer, D. A. (eds.), Comparative Psychologw A Modern Survey. McGraw-Hill, New York. Heron, W. T. (1935). The inheritance of maze-learning ability in the rat. J. Comp. Psychol. 19:77-89. Cited by Fuller, J. L., and Thompson, W. R, (1960). Behavior Genetics, Wiley, New York. International Index ~[ Laboratory Animals (1970). Compiled and distributed by Medical Research Laboratory Animals Centre. Canshalton, Surrey, U.K. McClearn, G. E., Wilson, J. R., and Meredith, W. (1970). The use of isogenic and heterogenic mouse stocks in behavioral research. In Lindzey, G., and Thiessen, D. E. (eds.), Contributions to Behavior-Genetic Analysis: The Mouse as a ProtOtype. AppletonCentury-Crofts, New York. Myers, A. K. (1959). Avoidance learning as a function of several differences in rats. J. Comp. Physiol. Psychol. 52:381 386. Padeh, B. (1966). Genetic and environmental correlations between maze-learning and brain size and RNA content. M,Sc. Thesis, Bar-llan University, Ramat Gan, Israel (in Hebrew). Rensch, B. (1956). Increase of learning capability with increase of brain size. Am. Naturalist 90:81.-95. Searle, L. V. (1949). The organization of hereditary maze-brightness and maze-dullness, Genet. Psychol. Monogr. 39:279 325. Wimer, C., and Prater, L. (1966). Some behavioral differences in mice genetica'lly selected for high and low brain weight, Psychol. Rep, 19:675-681. Wimer, C., Roderick, T. H., and Wimer, R. E. (1969). Supplementary report: Behavioral differences in mice genetically selected for brain weight. Psychol. Rep. 25:363-368. [
