Crusio, W E., & Schmitt, A. (1998 In Press) A Multivariate Quantitative-Genetic Analysis of Behavioral Development in Mice. Developmental Psychobiology.

A Multivariate Quantitative-Genetic Analysis of Behavioral Development in Mice

Key words: reflexes; behavioral development; diallel cross; quantitative-genetics; multivariate analysis; genetic architecture; adaptive value; evolutionary history; genetic correlation; mouse

Wim E. Crusio

Andrea Schmitt

Send correspondence and proofs to:

Tel: 33 2 3825 7974

Fax: 33 2 3825 7979

E-Mail: crusio@citi2.fr

The present experiment attempted a behavior-genetic dissection of early behavioral development in laboratory mice. To this end, we used a full, replicated diallel cross to uncover the genetical architecture as well as the multivariate genetic structure underlying early behavioral ontogeny. A number of standard sensory-motor tests were administered on postnatal days 3, 5, 8, 10, 13, 17, and 22 to a total of 622 pups from 120 litters (4-6 pups per litter) from a 4 times replicated complete diallel cross between five inbred mouse strains. The first day on which an animal showed adult performance was taken as its score on that test. MANOVA did not show any effects of the pup's sex on the speed of development. Hayman's analysis of variance for diallel tables indicated no or only weak additive-genetic effects. Dominance was absent in almost all cases, except for the auricular startle response, where weak directional dominance for fast development was found. These results are in accordance with an evolutionary past of directional selection for well-canalized development. Factor analyses of the phenotypic and additive-genetic correlation matrices indicate that at least two factors are necessary to describe the behavioral variation.

The causation of a behavioral trait has two aspects: the phenogenetic and the phylogenetic ones (van Abeelen, 1979). Both concern the genetic correlates of a phenotype, the first in a gene-physiological, the latter in an evolutionary sense. Stated otherwise, neurobehavioral geneticists attempt to uncover the physiological pathways underlying the expression of a trait and to provide an answer to the question of what exactly is the adaptive value of this trait for the organism. With this knowledge it becomes possible to explain the similarities and differences between individuals belonging to the same species. As has been argued before (Crusio, 1992, 1995), quantitative-genetic methods may be employed with profit to address problems related to both aspects of causation. In the present article, we use these methods to arrive at a behavior-genetic dissection of early behavioral development in laboratory mice (Mus musculus domesticus).

The Genetic Architecture of Early Behavioral Development in Mice

Selection pressures mold a species' genetic make-up which consequently will show traces of this past selection. Therefore, information about the genetic architecture of neurobehavioral traits might permit us to make inferences about the evolutionary history of these traits (Broadhurst & Jinks, 1974). In its broadest sense, knowledge about that genetic architecture implies an understanding of the effects of genes governing a particular phenotype in a given population at a given time and includes information concerning the presence and size of certain genetic effects, the number of genetic units involved, etc. Generally, however, information about the presence and nature of dominance suffices (Crusio, 1992).

With very few exceptions, natural, non-pathological variation in neurobehavioral phenotypes is polygenically regulated. If dominance is present, we may then envisage two different situations: either (uni)directional or ambidirectional dominance. In the first case, dominance acts in the same direction for all genes involved (e.g., for high expression of the trait), whereas in the latter case it acts in one direction for some genes and in the opposite one for others. In its most extreme form, ambidirectional dominance may lead to situations where an F1 hybrid is completely intermediate between its parents, despite the presence of strong dominance effects.

Mather (1973) distinguished between three kinds of selection: stabilizing, directional, and disruptive. Stabilizing selection favors intermediate expression of the phenotype, directional selection favors either high or low expression, whereas with disruptive selection more than one phenotypic optimum exists. Disruptive selection will lead to di- or polymorphisms, which may be in stable equilibrium or may even lead to breeding isolation and incipient speciation (Thoday, 1972). The commonest example of the former is the existence of two sexes, whereas a possible example of the latter are the explosive adaptive radiation and speciation found among fish species belonging to the family Cichlidae in the great East-African lakes (Fryer & Iles, 1972).

Selection acts in favor of those genotypes that not only produce the phenotype selected for, but are also capable of producing progeny that differs little from this phenotype. In the long run this results in a population whose mean practically coincides with the optimum. Thus, stabilizing and directional selection have predictably different consequences for the genetic architecture of a trait (Broadhurst and Jinks, 1974; Fisher, 1958). If a population undergoes directional selection, genes for which a dominant allele is correlated with a phenotypic expression opposite to the favored direction will quickly become fixed for the recessive allele. The same applies to genes for which dominance is absent, i.e., where the heterozygote is intermediate between the two homozygotes. In contrast, selection against recessive alleles is much slower. The result will be that the first type of genes are not contributing to the genetic variation within the population anymore, whereas those genes where the allele correlated with the favored phenotypical expression is dominant remain genetically polymorphic for a much longer time, conserving genetic variance. It may easily be seen that directional selection leads to situations where dominance is directional, in the same direction as the selection. Stabilizing selection leads to situations where dominance is either absent or ambidirectional. Furthermore, directional selection generally results in lower levels of genetic variation than ambidirectional selection does.

The genetic architecture of a trait may be uncovered by using quantitative genetic methods. For a brief technical presentation of the most pertinent ones, we refer the reader to Crusio (1992). In the present study, we investigate the genetic architecture underlying early behavioral ontogeny, employing a full diallel cross (Jinks & Hayman, 1953). To diminish error variance and avoid problems with interpretability, and also to avoid ecologically nonsensical situations (Henderson, 1979, 1986), it is necessary to adapt the test situation to the species-typical characteristics of one's experimental subjects (Gerlai, 1996). Fox's (1965) battery of sensory-motor tests appeared to fulfill these conditions.

In his seminal 1965 paper, Fox developed a battery of sensory motor tests to assess behavioral ontogeny of house mice. These tests have been widely employed to assess, for example, genetic effects (e.g., Fuller & Geils, 1973; Fuller & Herman, 1974; van Abeelen & Schoones, 1977) or effects of nutrition (Wainwright, Huang, Bulman-Fleming, Mills, Redden & McCutcheon, 1991; Wainwright, Pelkman, & Wahlsten, 1989). Fox (1965) reported that strain differences were apparent, but also that environmental factors such as litter size might influence behavioral development.

Based on his initial findings, Fox (1965) proposed a division into 5 well-defined stages of mouse development: 1. Perinatal (from birth to 3 days of age): characterized by the presence of a few weak reflexes only. 2. Neonatal (3-9 days): strong and stereotyped reflexes start to occur. 3. Postnatal or Transition (9-15 days): primitive reflexes like rooting disappear, adult locomotor activities start to appear, and overgeneralized sensory responses are seen. 4. Postnatal infantile or Pre-juvenile (15-26 days): refinement of locomotor abilities and sensory capacities, overgeneralized responses disappear. 5. Juvenile (26 days onward until sexual maturity): further refinement, manipulative abilities start becoming adult-like. This subdivision, too, has found general acceptance (van Abeelen and Schoones, 1977).

As behavior during the juvenile stage starts becoming very adult-like, we decided to limit our behavioral observations to the first four stages. In addition, as it has been shown by Fuller and Herman (1974; see also Roubertoux, Nosten-Bertrand, Cohen-Salmon, & L'Hotellier, 1992), that practice may exert effects on behavioral development and, even more important, that these effects may differ according to genotype, we limited such practice effects by following van Abeelen and Schoones (1977): observations were restricted to 7 time points distributed over the first four stages of behavioral development.

Methods

Subjects

A full diallel cross consists of n strains and all possible F1 crosses between them, including reciprocals. As parents for the diallel cross, we used mice from the 4 highly inbred strains C57BL/6J, C57BR/cdJ, BALB/cJ, and DBA/2J (cf. Staats, 1985), bred for at least 10 generations in our own colony. As 5th parent we used the highly inbred strain BA//C, for which breeders were kindly provided by Dr. J.H.F. van Abeelen (Nijmegen, The Netherlands). The latter strain was inbred in Nijmegen from a random-bred obese stock originating from Dr. D. S. Falconer (Edinburgh, UK). It is now non-obese and characterized by good reproductive performance (Staats, 1985). In addition, it is not directly related to any standard laboratory inbred strain.

From all 25 possible crosses, one (hybrids) or two litters (inbreds) were raised simultaneously. Overbreeding was practiced to ensure that the required number of litters would be available for each cross and in each replication. The 30 litters obtained in this way constituted one replication or block. In all, four such replications were bred consecutively. For the second and fourth replications we used only the second litter of a particular dam. Third litters were taken for the first and third blocks. Males were removed when females became visibly pregnant and replaced 0-4 weeks after females had given birth. The latter variable period was necessary to synchronize births of animals of different genotypes within a particular block as much as possible.

Cages were inspected twice a day for births (8.00 am and 4.00 pm) and the day a litter was first seen was considered to be day one. To randomize possible postnatal effects, all newborn pups were routinely fostered to lactating mothers from a random-bred NMRI stock. Litters smaller than 5 and those without male pups were discarded. Litters were culled to a maximum of 8 pups. To standardize the received quantity of milk and maternal pup care, litters smaller than 8 were brought up to 8 with pups with another coat color. From each litter, one (litters with 5 pups) or two male pups (litters with 6 to 8 pups) were not tested but raised to adulthood for other experiments (see Crusio, Genthner-Grimm, & Schwegler, 1986; Crusio, Schwegler, & van Abeelen, 1989), so that finally between 4 and 6 pups were tested per litter. In total, 622 animals (233 males and 389 females) from 120 litters were tested, resulting in 37-42 (inbreds) and 17-24 subjects (hybrids) per cross. The animals were housed in plastic breeding cages (165 x 220 x 140 mm) with a metal cover and a bedding of wood shavings. Cages were cleaned once a week. Food pellets (Altromin Standarddiät, Altromin GmbH) and tap water were available ad libidum. The animals were maintained in an air-conditioned mouse room (22 ± 2^o C; humidity 55 ± 5 %) with a 12h light/dark schedule (lights on 6.00h 225 Lux).

Procedure

Thirteen sensory-motor responses and body weight were recorded for each pup on days 3, 5, and 8 (neonatal period), days 10 and 13 (postnatal transition), and days 17 and 22 (postnatal infantile; day of birth = day 1; see Fox, 1965, and van Abeelen & Schoones, 1977). The same individuals were tested at all ages. For identification purposes, pups were daily marked with a non-toxic purple dye. Tests took place between 8.30 am and 2.00 pm (occasionally 4.00 pm) in a sequence that varied randomly over groups.

Tests were administered in the following order: 1) Rooting (ROO): bilateral stimulation of the face induces the pup to crawl forwards, pushing the head in a rooting fashion. 2) Righting (RIG): when the pup is placed on its back, it turns over and rests with all four feet on the ground. 3) Postural flexion and extension (FLE): when suspended by the scruff of the neck, the pup either flexes or extends its limbs. 4) Vibrissae placing (VIB): the pup is suspended by its tail. When its vibrissae are stimulated with a pencil it raises its head and performs a placing response with the extended forelimb. 5) Visual placing (VIS): the pup is suspended by its tail and lowered toward a solid object (e.g., a pencil or table top), without its vibrissae being stimulated. It will raise its head and perform a placing response. 6) Forelimb grasping (FOR): occurs when the inside of the forepaw is stimulated with an object. 7) Hindlimb grasping (HIN): as previous, with the hindlimb. 8) Vertical screen test (VER): the pup is placed on a vertical wire-mesh screen and the time until the animal falls down is registered. 9) Bar holding (BAR): the pup must be able to hold onto a bar (1.5 cm diameter) for more than 5 sec. 10) Negative geotaxis (GEO): when the animal is placed on a 45^o incline with its head pointing down it will turn around and crawl up the slope. 11) Cliff-drop aversion (CLI): When the pup is placed on the edge of a cliff or table top, it will turn around and crawl away from the cliff drop. 12) Popcorn behavior (POP): exaggerated jumping and running behavior in response to a gentle puff of the experimenter's breath. 13) Auditory startle response (STA): a loud snap of the fingers close to the ears (but without tactile stimulation) causes an immediate startle response.

As was also done by Nosten (1989), an animal's score was the number of the first postnatal day on which the adult response of normal mice (Fox, 1965) was displayed for the first time. If an animal had not yet displayed the adult response by day 22, it was arbitrarily assigned a score of 25. Scores were averaged over litters. Therefore, in what follows, litters constituted the unit of analysis.

Analysis

Before any quantitative-genetic analysis can be carried out, an adequate scale has to be found (Kerbusch, van der Staay, & Hendriks, 1981). Briefly, on such a scale the data should be normally distributed, the variances of the 25 nonsegregating populations should be homogeneous, and no systematic covariations should occur between measures of central tendency and variation. The HOMAL program (Crusio, 1990) was used to find suitable transformations, or transformations that violated the requirements least. Next, a MANOVA with Sex, Cross, and Replication as main factors was carried out employing the SAS procedure GLM (SAS Institute, Inc. 1987). As no significant effect of Sex was obtained, neither alone (F(13,420) = 1.62, p > .076), nor in interaction with the other two main effects (Sex x Cross: F(312,5616) = 1.08, p > .162; Sex x Replication: F(39,1266) = 0.73, p > .893; Sex x Cross x Replication: F(793,5616) = 1.06, p > .118), results of both sexes were pooled.

The univariate analyses of variance and of variance-covariance of the diallel cross followed the methods of Hayman (1954a and b) described in detail by Crusio, Kerbusch, & van Abeelen (1984). Briefly, we first applied Hayman's ANOVA of diallel crosses. In this procedure, the a item tests primarily for additive variation, whereas the b item tests for allelic interaction stemming from dominance effects. The latter may be divided into three different sources of variation: b1 tests for directional dominance, b2 for additional dominance effects that can be accounted for by genes having unequal allele distributions over strains, and, finally, b3 for residual dominance effects. Items c and d test for general and specific reciprocal effects (Crusio, 1987), respectively. In the present case, because of the fostering procedure followed, significant c or d items will reflect prenatal effects, or possibly also postnatal effects exerted during the first few hours after birth.

In the next step, a variance-covariance analysis was performed. Here, for each strain, the variance of its F1 crosses is calculated (Vr) as well as the covariance of these hybrids with their non-recurrent parent (Wr) so that one pair of Wr,Vr values is obtained for each array. Whether the assumptions underlying a diallel-cross analysis are fulfilled (i.e., no epistasis, no multiple allelism, and independent distribution of alleles among parents) is tested by the linear regression of Wr on Vr that should have a slope of 1. Further, if a strain carries many dominant alleles, then its F1 hybrids tend to resemble it and Wr and Vr will be smaller for such a strain than for one carrying many recessive alleles. Thus, a significant correlation of Wr+Vr with the phenotypical value of the rth strain (Pr) indicates the presence of directional dominance. The sign of such a correlation will be opposite to the direction of the dominance.

Estimates of several components of the variance were obtained from this analysis: E, the environmental part of the phenotypic variance; D, the additive-genetic part; H1 and H2, the dominance contributions. From these components one can calculate the average degree of dominance () and the heritabilities in the narrow sense (the proportion of the phenotypic variance caused by additive-genetic effects) and in the broad sense (the proportion of the phenotypic variance caused by all genetic effects together).

Results

The results of the behavioral tests are arranged in Table 1 (untransformed values). Satisfactory transformations could be found for each variable and are presented in Table 2, together with the results of the Hayman diallel-cross ANOVA and some results of the variance-covariance analysis.

No violations of the assumptions underlying the analyses were indicated for any variable. Generally, only low levels of genetic variation were found. No differences between the 25 crosses were obtained for rooting, righting, postural flexion and extension, visual placing, and popcorn behavior. The between-cell item was significant for vibrissae placing, which appeared to be due to reciprocal differences, only. Relatively strong additive-genetic variation was indicated for forelimb and hindlimb placing, the vertical screen test, bar holding, negative geotaxis, and the auditory startle reflex. Dominance was generally absent or only very low. For forelimb placing, the vertical screen test, and bar holding, the b1 item was significant, but the b item and the correlation between W+V and the parental scores were not. We conclude that dominance was not an important feature in the genetic architectures of these behaviors. Only for the auditory startle reflex was there some indication of directional dominance (both b1 and the correlation between W+V and the parental scores significant), which was in the direction of lower scores in this case. For cliff-drop aversion, the between-cell item was not significant, but the b and b2 items were, possibly indicating very low levels of ambidirectional dominance. Both narrow and broad sense heritabilities were rather low, generally less than 20% and never exceeding 34%. Violations of the assumptions proved to be non-significant in all cases.

Rather strong differences were found between the four blocks for postural flexion and extension, vibrissae placing, visual placing, forelimb grasping, bar holding, negative geotaxis, and popcorn behavior.

Discussion

The results of the quantitative-genetic analyses reveal three different types of genetic control of early behavioral development: 1/ mainly additive genetic variation, supplemented with directional dominance for low scores (auditory startle reflex); 2/ low levels of additive genetic variation, without dominance or with only very low levels of ambidirectional dominance (forelimb and hindlimb placing, vertical screen test, bar holding, negative geotaxis); 3/ no genetic variation present at all (rooting, righting, postural flexion and extension, visual placing, and popcorn behavior).

The genetic architecture of the first group (auditory startle response) is diagnostic (Broadhurst & Jinks, 1974) for an evolutionary history of directional selection. The directional dominance for low scores found for this variable indicates that fast development of this reflex is adaptive. This appears a reasonable interpretation given the obvious advantages of this reflex for the individual in case of predator attack.

The genetic architecture found for the variables of the second group points to an evolutionary history of stabilizing selection. The levels of additive-genetic variation were very low, even more so when considering that the present analyses were carried out on litter means, effectively removing any within-litter environmental variation and thus "boosting" heritability. This finding points to a rather intense level of selection, any deviation of the optimum being fairly heavily selected against.

As was the case for the variables of the second group, the behavioral variables of the third one (no detectable levels of variation) must have been under very strong past selection, which has removed all genetical variation from the population. Nothing can be said about the nature of the selection that has been exerted on the variables of this third group in the evolutionary past, especially so since for the other variables we have found evidence for both stabilizing and directional selection.

It should perhaps be noted here that our experimental protocol may have limited the statistical power of our design. Firstly, because of time constraints, pups had to be tested over a 7.5 hour time period in the day, probably adding some error variance to the date. Secondly, because we wanted to minimize handling of the pups, testing took place at only 7 time points during a span of 22 days. Had more time points been chosen, the precision of our measurements might have been greater. However, these considerations obviously do not change the above conclusions regarding low levels of genetic variation, as our design clearly was able to detect significant genetic effects for a number of variables. In addition, the above noted possible increase in error variance was offset at least in part by the facts that all pups were fostered to random-bred females, that only pups from second or third litters were used, and that the analyses were carried out on litter means, procedures that may be expected to have appreciatively reduced environmentally-induced error variance. Of course, our efforts to limit variation in the pup's environment will also have limited the risk that genotype-environment interactions would be confounded our analysis.

Despite our efforts to minimize environmental differences, large block effects were obtained, indicating that even small environmental variations may have considerable effects on developmental speed. Part of these effects appear to be caused by prenatal effects of the mother's parity, which are very small, however (Crusio and Schmitt, 1996). Fortunately, the replicated design of the diallel cross allows a genetic analysis unbiased by the environmental differences between replications.

We may compare our results to those of Roubertoux, Baumann, Ragueneau, & Semal (1987), who investigated rooting in two of the strains that were also used here: C57BL/6 and BALB/c. They found that this response develops faster in C57BL/6 mice and attributed this difference to one single gene, located on chromosome 4. However, in the present experiment the difference between these two strains was only marginal and no genetic variation at all was detected for this variable. There are a number of possible explanations for this seeming discrepancy. First, Roubertoux et al. (1987) used the Bailey substrains C57BL/6By and BALB/cBy, whereas the Jackson substrains C57BL/6J and BALB/cJ (Staats, 1985) were used here. Second, Roubertoux et al. (1987) used the proportion of pups still showing the rooting response on day nine as a phenotypical measure, which differs from the one used here. (But note, however, that in our experiment pups developed the adult response - loss of the righting reflex - several days earlier than in Roubertoux et al.'s experiment. Fox's 1965 data appear to be intermediate). Third, apparently the difference between C57BL/6 and BALB/c is attributable to one gene solely. Perhaps only the Bailey, but not the Jackson substrains exhibit this difference. Alternatively, if in our diallel cross the other three strains were fixed for the same allele as either C57BL/6 or BALB/c, the resulting level of genetic variation may have been too low to allow detection in the present case. However, even if the latter possibilities were true, our conclusion of low levels of heritable variation for the speed of development of the rooting reflex would not change.

Finally, we may also compare our results with those of Henderson (1981), who investigated the genetic architecture of popcorn behavior employing an 8 x 8 diallel cross. Henderson (1981) reported directional dominance for this phenotype. Heritability was low, however (in the narrow sense: 12.5%; in the broad sense: 18.4%). Although we did not find any genetic variation at all in our sample, both studies clearly agree that genetic variation is very small. In addition, we did find directional dominance for the rather similar auditory startle response, too. It should be noted that Henderson's (1981) phenotypical measure was quite different from ours: the time required for all pups in a litter to hop out of their home cage at 15 days of age. Henderson (1981) therefore analyzed the intensity of response at a particular age rather than the speed of individual development, as was done here.

The Multivariate Structure of Early Behavioral Development

A weakness inherent in correlational studies is that a phenotypical correlation between characters does not necessarily reflect a functional relationship. On the other hand, if two independent processes, one causing a positive relationship, the other causing a negative relationship, act simultaneously upon two characters, the effects may cancel each other so that no detectable correlation can emerge. These problems can to a large extent be avoided by looking at the genetic correlations, that is, at correlations between the genetic effects that influence certain characters. These are the products of either genes with pleiotropic effects or of linkage disequilibrium. By using inbred strains that are only distantly related, the probability that a linkage disequilibrium occurs may be minimized so that a possible genetic correlation will most probably be caused by pleiotropy, indicating the existence of a (set of) gene(s) influencing both characters simultaneously. Thus, for these characters, at least part of the physiological pathways leading from genotype to phenotype must be shared and a causal, perhaps also functional, relationship must exist (Crusio, 1992, 1995). It is this special property that renders the genetic-correlational approach so uniquely valuable.

In the present situation genetic correlations may help us to discern whether different aspects of early behavioral development are the resultant of the concerted action of one single or of multiple independent developmental processes.

Method

Analysis

Following the univariate analyses described above, bivariate analyses permitted the estimation of phenotypic, genetic, and environmental correlations. Full computational details have been presented elsewhere (Crusio, 1993). Briefly, we have to obtain the bivariate equivalents of E, D, and H2. Then, rE = Exy/, rD = Dxy/, and rH = H2xy/ provide estimates of correlations between the environmental effects, additive-genetic effects, and dominance deviations, respectively, as exerted on characters x and y. Exy, Dxy, and H2xy stand for those portions of the total phenotypic covariance that are induced by the aforementioned effects. Briefly, bivariate equivalents of the monovariate estimates of E, D, and H2 were calculated by replacing the monovariate variance components with corresponding bivariate covariance components. For example, instead of Vr (the variances of the arrays) we calculated Wr,xy (the covariances between variables x and y of arrays).

In the monovariate case, standard errors for the estimates of E, D, and H2 may be calculated according to the method of Hayman (1954b). The extension to the bivariate case is straightforward (see also Crusio, 1993) and this method was used here to evaluate the statistical significance of Exy, Dxy, and H2xy and, by extension, of the respective correlations. The foregoing analysis, however, may only be applied if the assumptions underlying diallel-cross analyses (see above) are met in the bivariate case, too. This may be tested by examining the Wr:Wr(F-P) graph, which should have a slope of -1 in this case if dominance is present (Crusio, 1993).

To aid in the interpretation of the results, we performed principal factor analyses (PFA) on the obtained phenotypical, environmental, and additive-genetic correlation matrices. We used the SAS procedure FACTOR (SAS Institute Inc., 1987) and retained only principal components with an eigenvalue > 1, that were subsequently rotated using an orthoblique Harris-Kaiser rotation (SAS Institute Inc., 1987).

Results and Discussion

Phenotypical correlations are presented in Table 3, whereas the environmental and additive-genetic ones can be found in Table 4. As dominance was absent for most variables, dominance correlations were not estimated. Because after a reciprocal transformation of a variable the signs of correlations with this variable reverse, these correlations were multiplied by -1 to facilitate interpretation. In what follows, we will only consider correlations between variables for which significant additive-genetic variation was detected in the univariate analyses (forelimb and hindlimb placing, vertical screen test, bar holding, negative geotaxis, and auditory startle reflex).

In 5 of the 15 bivariate analyses did the slope of the Wr:Wr(F-P) graph deviate significantly from -1. However, this should be not too surprising as significant levels of dominance were absent for almost all variables. Thus we proceeded with the subsequent analyses. The PFA of the phenotypical correlation matrix revealed the existence of two almost completely orthogonal factors (interfactor correlation 0.01). For the environmental correlations, the PFA indicated the existence of 2 moderately correlated factors. However, these factors explained only about half of the total variation, indicating the fact that residual (variable-specific) factors are relatively important. This means that many influences of the environment work on only one of the variables at a time. Due to sampling error, some additive-genetic correlations were estimated > |1|. These values were put to ±1 for the subsequent PFA, which again indicated the existence of at least two factors influencing early behavioral ontogeny. In the latter case, however, a rather large interfactor correlation of 0.62 was obtained. This may be interpreted as indicating that much genetic variation, and therefore also the underlying developmental processes, are common for the phenotypes investigated. In addition, some genes influence only a subset of these characters. On the phenotypical level, this complex genetic regulation, in combination with the effects of environmental variation, results in two seemingly independent factors of development. It is also worthwhile to note that, in contrast to the phenotypical PFA, some loadings in the environmental and additive-genetic PFAs have opposite signs, implying that some environmental or genetic effects may increase the speed of development of one variable but decrease that of another.

These results should not be taken as irrefutable proof of the invalidity of developmental time scales, such as the one proposed by Wahlsten (1974), that combine the results from different sensorimotor tests. It has been shown repeatedly (Wainwright et al., 1989, 1991) that such a developmental timescale may constitute a useful shortcut to evaluate the more general effects of particular treatments. However, when using such methods, one should keep in mind that the processes underlying early behavioral development are not unitary, which may have important consequences for the interpretation of results obtained with such time scales.

Conclusion

The results of the present experiment indicate that early behavioral development is not a unitary process and has become strongly canalized (Lerner, 1954; Waddington, 1942) as a result of an evolutionary past of strong selection, reflecting the obvious importance of a well-tuned regulation of these processes for an individual's fitness.

Notes

The experimental parts of the work described here were carried out while the first author was working as a guest at the Institute of Human Genetics and Anthropology, University of Heidelberg, Germany, supported by a NATO Science Fellowship, awarded by the Netherlands Organization of Pure Research (ZWO; Den Haag, The Netherlands), and an Alexander-von-Humboldt stipend, awarded by the Alexander-von-Humboldt Foundation (Bonn, Germany). The analysis of the data and the writing-up profited from support of the Centre National de la Recherche Scientifique (CNRS UPR 9074, Orléans, France), the Ministry for Research and Technology, Région Centre, and Préfecture de la Région Centre. UPR 9074 is affiliated with INSERM and the University of Orléans. We gratefully acknowledge the generous hospitality of Profs. F. Vogel and W. Buselmaier (Heidelberg), who provided all materials necessary to carry out this study. Profs. Pierre L. Roubertoux and Michèle Carlier (Orléans, France) critically read the manuscript.

References

Broadhurst, P. L., & Jinks, J. L. (1974). What genetical architecture can tell us about the natural selection of behavioural traits. In J. H. F. van Abeelen (Ed.), The Genetics of Behaviour (pp. 43-63). Amsterdam: North-Holland.

Crusio, W.E. (1987). A note on the analysis of reciprocal effects in diallel crosses. Journal of Genetics 66, 177-185.

Crusio, W. E. (1990). HOMAL: a computer program for selecting adequate data transformations. Journal of Heredity 81, 173.

Crusio, W. E. (1992). Quantitative Genetics. In D. Goldowitz, D. Wahlsten, & R. Wimer (Eds.), Techniques for the Genetic Analysis of Brain and Behavior: Focus on the Mouse. Techniques in the Behavioral and Neural Sciences, Volume 8, Elsevier, Amsterdam, Pays Bas, pp. 231-250, 1992.

Crusio, W. E. (1993). Bi- and multivariate analyses of diallel crosses: A tool for the dissection of neurobehavioral phenotypes. Behavior Genetics 23, 59-67.

Crusio, W. E. (1995). Natural selection on hippocampal circuitry underlying exploratory behaviour in mice: Quantitative-genetic analysis. In E. Alleva, A. Fasolo, H.-P. Lipp, L. Nadel, & L. Ricceri (Eds.), Behavioural Brain Research in Naturalistic and Seminaturalistic Settings. NATO Advanced Study Institutes Series D, Behavioural and Social Sciences, Kluwer Academic Press, Dordrecht, pp. 323-342.

Crusio, W. E., Genthner-Grimm, G., & Schwegler, H. (1986). A quantitative-genetic analysis of hippocampal variation in the mouse. Journal of Neurogenetics 3, 203-214.

Crusio, W. E., Kerbusch, J. M. L., & van Abeelen, J. H. F. (1984). The replicated diallel cross: A generalized method of analysis. Behavior Genetics 14, 81-104.

Crusio, W. E., & Schmitt, A. (1996). Prenatal effects of parity on behavioral ontogeny in mice. Physiology and Behavior 59: 1171-1174.

Crusio, W. E., Schwegler, H., & van Abeelen, J. H. F. (1989). Behavioral responses to novelty and structural variation of the hippocampus in mice. II. Multivariate genetic analysis. Behavioural Brain Research 32, 81-88.

Fisher, R.A. (1958). The Genetical Theory of Natural Selection. New York: Dover Publications.

Fox, W. M. (1965). Reflex-ontogeny and behavioural development of the mouse. Animal Behaviour 13, 234-241.

Fryer, G. and Iles, T.D. (1972) The Cichlid Fishes of the Great Lakes of Africa: Their Biology and Evolution. Edinburgh: Oliver and Boyd.

Fuller, J. L., & Geils, H. D. (1973). Behavioral development in mice selected for differences in brain weight. Developmental Psychobiology 5, 307-318.

Fuller, J. L., & Herman, B. H. (1974). Effect of genotype and practice upon behavioral development in mice. Developmental Psychobiology 7, 21-30.

Gerlai, R. (1996). Molecular genetic analysis of mammalian behavior and brain processes: Caveats and perspectives. Seminars in the Neurosciences 8, 153-161.

Hayman, B. I. (1954a). The analysis of variance of diallel tables. Biometrics 10, 235-244.

Hayman, B. I. (1954b). The theory and analysis of diallel crosses. Genetics 39, 789-809.

Henderson, N. D. (1979). Adaptive significance adn animal behavior: The role of genotype-environment intercation. In J. R. Royce, & L. P. Mos (Eds.), Theoretical Advances in Behavior Genetics (pp. 243-287). NATO Advanced Study Institutes Series D, Behavioural and Social Sciences. Alphen aan den Rijn: Sijthoff and Noordhoff.

Henderson, N. D. (1981). A fit mouse is a hoppy mouse: Jumping behavior in 15-day-old Mus musculus. Developmental Psychobiology 14, 459-472.

Henderson, N. D. (1986). Predicting relationships between psychological constructs and genetic characters: An analysis of changing genetic influences on activity in mice. Behavior Genetics 16, 201-220.

Jinks, J. L., & Hayman, B. I. (1953). The analysis of diallel crosses. Maize Genetics News Letter 27, 48-54.

Lerner, I. M. (1954). Genetic Homeostasis. Edinburgh: Oliver and Boyd.

Mather, K. (1973). Genetical Structure of Populations. London: Chapman and Hall.

Nosten, M. (1989). Early development in mice VI: Additive and interactive effects of offspring genotype and maternal environments. Physiology and Behavior 45, 955-961.

Roubertoux, P. L., Bauman, L., Ragueneau, S., & Semal, C. (1987). Early development in mice. IV. Age at disappearance of the rooting response: Genetic analysis in newborn mice. Behavior Genetics 17, 453-464.

Roubertoux, P. L., Nosten-Bertrand, M., Cohen-Salmon, C., & L'Hotellier, L. (1992). Behavioral development: A tool for genetic analysis in mice. In D. Goldowitz, D. Wahlsten, & R. Wimer (Eds.), Techniques for the Genetic Analysis of Brain and Behavior: Focus on the Mouse. Techniques in the Behavioral and Neural Sciences, Volume 8, Elsevier, Amsterdam, Pays Bas, pp. 423-441, 1992.

SAS Institute Inc. (1987). SAS/STAT Guide for Personal Computers, Version 6 Edition. Cary, NC: SAS Institute Inc.

Staats, J. (1985). Standardized nomenclature for inbred strains of mice: Eight listing. Cancer Research 45, 945-977.

Thoday, J.M. (1972) Disruptive selection. Proceedings of the Royal Society of London, series B, 182, 109-143.

van Abeelen, J. H. F. (1979). Ethology and the genetic foundations of animal behavior. In J. R. Royce, & L. P. Mos (Eds.), Theoretical Advances in Behavior Genetics (pp. 101-112). NATO Advanced Study Institutes Series D, Behavioural and Social Sciences. Alphen aan den Rijn: Sijthoff and Noordhoff.

van Abeelen, J. H. F., & Schoones, A. H. (1977). Ontogeny of behavior in two inbred lines of selected mice. Developmental Psychobiology 10, 17-23.

Waddington, C. H. (1942). Canalization of development and the inheritance of acquired characters. Nature 150, 563-565.

Wahlsten, D. (1974). A developmental time scale for postnatal changes in brain and behavior of B6D2F₂ mice. Brain Research 72, 251-264.

Wainwright, P. E., Huang, Y. S., Bulman-Fleming, B., Mills, D. E., Redden, P., & McCutcheon, D. (1991). The role of n-3 essential fatty acids in brain and behavioral development: A cross-fostering study in the mouse. Lipids 26, 37-45.

Wainwright, P. E., Pelkman, C., & Wahlsten, D. (1989). The quantitative relationship between nutritional effects on preweaning growth and behavioral development in mice. Developmental Psychobiology 22, 183-195.

^aSee text for abbreviations of sensory-motor tests
^bEntries for the a, b, c, and d items are F values.
* p<0.05;
** p<0.01;
*** p<0.001.

^aSee text for abbreviations of sensory-motor tests
* p<0.05; ** p<0.01; *** p<0.001.

^aSee text for abbreviations of sensory-motor tests
^bDue to sampling error, estimates may exceed |1|
* p<0.05; ** p<0.01; *** p<0.001.

Only loadings larger than |0.30| are indicated.
Inter-factor correlations: phenotypical, r = -0.01; environmental, r = 0.31; additive genetic, r = 0.62