Schicknick, H., Hoffmann, H-J., Schneider R. & Crusio, W. E. (1993) Genetic Analysis of Isolation-induced Aggression. III. Classical Cross-breeding Analysis of Differences Between Two Closely-Related Inbred Mouse Strains. Behavioral and Neural Biology 59 242-248.

Genetic Analysis of Isolation-induced Aggression. III. Classical Cross-breeding Analysis of Differences Between Two Closely-Related Inbred Mouse Strains

Horst Schicknick1,3, Hans-Jürgen Hoffmann1, Regine Schneider1, and Wim E. Crusio2,4

  1. Institute of Brain Research, PF 1860, O-3010 Magdeburg, Germany.
  2. Génétique, Neurogénétique et Comportement, URA 1294 CNRS, UFR Biomédicale, Université René Descartes (Paris V), 45 rue des Saints-Pères, 75270 Paris Cedex 06, France
  3. To whom correspondence should be addressed.
  4. We thank Mr. J. Krüger (Institute of Human Genetics, University of Heidelberg, Germany) for developing a computer program for executing the maximum-likelihood procedure used in the Collins test and Antje Leddin (Magdeburg) for expertly conducting the aggression tests.

Send correspondence and proofs to: Dr. Horst Schicknick at the above address.

Tel: (INT)-49-391-674164.

Fax: (INT)-49-391-676160


In two preceding papers we reported on two closely-related inbred mouse strains, ABG and AB//Halle that display very large differences in isolation-induced intermale aggression. In the present article we investigated animals from a complete Mendelian cross between these strains to test the hypothesis that the behavioral difference is due to genetic variation at only few loci, possibly just one. In the quantitative-genetic analysis of generation means and variances for the behavioral variables analyzed, relatively simple models were found. As epistasis was present in some cases, the monogenic hypothesis could not be confirmed. Also, the analysis of the segregating generations by means of Collins' nonparametric method revealed significant deviations of observed from expected distributions. We conclude that differences at more than just one single locus are correlated with the behavioral difference.

Keywords: isolation-induced aggression; inbred strains of mice; Mendelian cross; quantitative-genetic analysis


Numerous studies of aggressive behavior in the mouse have shown a complex genetic regulation of this phenotype. Both Y-chromosomal (Selmanoff, Goldman, Maxson, & Ginsburg, 1977; Maxson, Didier-Erickson, & Ogawa, 1989; Maxson, Ginsburg, & Trattner, 1982; Carlier, Roubertoux, Kottler, & Degrelle, 1990) and autosomal (Elefteriou, Bailey, & Denenberg, 1974; Ciaranello, Lipsky, & Axelrod, 1974; Orenberg, Renson, Elliott, Barchas, & Kessler, 1975; Popova & Kulikov, 1986; Carlier et al., 1990) genes appear to be implicated. Unfortunately, the involvement of multiple genetic loci renders a detailed genetic analysis of any one of these genes very difficult. However, should two inbred strains carry identical alleles at all but one (or only few) of these loci, an analysis of the effects of the latter might become possible.

Few genetic differences might be expected between pairs of closely related inbred strains. In a preceding paper (Schneider, Hoffmann, Schicknick, & Moutier, 1992) we reported on two such strains, ABG and AB//Halle, that displayed opposite extremes for intermale aggression. Even isolation failed to induce aggression in strain ABG, although this treatment is almost always effective in other inbred mouse strains (Goldsmith, Brain, & Benton, 1976; Cairns, MacCombie, & Hood, 1983). Practically no overlap was found for the distributions of aggression-related variables between ABG an AB//Halle. A cross-fostering study (Hoffmann, Schneider, & Crusio, 1993) suggested the absence of postnatal maternal influences. In combination with the relatively small genetic distance between these two strains (Schneider et al., 1992), this leads us to put forward the hypothesis that the behavioral difference may be correlated with a difference at one single locus only. We have bred a complete Mendelian cross between these two strains and have analyzed the results with both parametric and nonparametric quantitative-genetic methods to test this hypothesis.



A detailed description of the albino strains ABG (G) and AB//Halle (H) has been presented before (Schneider et al., 1992). Briefly, both strains have been developed from the same, partially inbred stock, resulting in only a small genetic distance between the two. Both strains had been bred in our mouse colony for 4-6 generations at the start of the experiment. From these parental strains, we bred reciprocal F1's (GH and HG, the female parent being indicated first), an F2 generation (GH x GH), and backcrosses to both parents (B1: GH x H; B2: GH x G). All animals were kept under standard laboratory conditions with a light regime of 12:12 LD (light on at 6:00 a.m.). They were provided with food (Altromin 1320 rat and mouse maintenance diet) and water ad libitum. Soiled bedding was changed weekly. For matings, two or three females and one male were housed together. Females were removed and housed individually immediately after diagnosis of pregnancy. Pups were weaned three weeks after parturition.

Behavioral testing

The experiments were carried out in the breeding room between 9:00 a.m. and 1:00 p.m. After weaning males from each litter were housed in groups of 2-6 for two weeks (cage type I: 25 x 40 x 18 cm); thereafter they were isolated for two weeks in smaller plastic cages (type II: 10 x 40 x 10 cm). Next, the test male (aged 50 ± 2 days) was put directly into the type I test cage. The test male was allowed to habituate for 1 min; then a standard opponent male of the inbred strain C3H/Ola was put into the test cage. Uninjured standard opponents were used up to five times to minimize the influence of previous fighting experience. For each individual a clean test cage was used in order to avoid effects of residual odors. Aggressive behavior was assessed by means of the methods of Hahn and Haber (1982) and Selmanoff, Maxson, & Ginsburg (1976) during an observation period of 10 min. The frequencies per minute of tail rattling (TR), attacks (AT), and aggressive grooming (AG) were registered. In addition, the latency to the first attack was recorded (L1). We also analyzed the sum of the frequencies of aggressive acts (SAA = TR + AT + AG), since strain distributions for this composite variable overlapped only minimally (Schneider et al., 1992).

Quantitative-genetic analysis

Before any parametric quantitative-genetic analysis can be carried through, an adequate scale has to be found (Crusio, 1992). Briefly, the variances of nonsegregating generations should be homogeneous on such a scale. In addition, no systematic covariation should occur between measures of central tendency and variation. The HOMAL program (Crusio, 1990) was used to find suitable scale transformations.

A model comprising 7 parameters is necessary to describe all variation between the generations of our classical cross (Table 1; cf. Mather & Jinks, 1982): the general mean (m), additive-genetic variation due to differences between homozygous allele pairs ([d]), dominance ([h]), epistasis ([i]-, [j]-, and [l]-type), and a reciprocal effect ([dr]; cf. Crusio & van Abeelen, 1987). Usually, there is no need for a complex model containing all these parameters to fit the observed variation between the different generations. A more simple model, in which some of the possible parameters are omitted, will suffice in most cases. To analyze our data we searched for the best fitting and most parsimonious model using Kerbusch's procedure (Kerbusch, van der Staay, & Hendriks, 1981). With this method, weighted least-squares estimates for the parameters are obtained from the generation means for all possible models. By assuming m always to be present and by taking into account the interaction parameters only if the parameters of the corresponding main effect(s) are also part of the model (see Crusio & van Abeelen, 1987), the number of possible models is reduced. The analysis starts by fitting the simplest model, which is rejected only if expectations for the generation means based on it deviate significantly from the observed ones (as evaluated by a c2 test), or if the data are described significantly better by a more complex model. If rejected, the next more complex model is fitted and the same procedure applied.

The second step is the estimation of the genetic and environmental components of the phenotypic variance. Here, we use 4 parameters (Table 1; Mather & Jinks, 1982): E - the environmental part of the variance, D - the variance due to additive-genetic effects, H - the variance due to dominance deviations, and S(dh) - the covariance between the additive-genetic effects and dominance deviations of genes. As in the present experiment more groups than parameters are available, an iterative weighted least-squares estimation procedure, developed by Hayman (1960) and recommended by Mather and Jinks (1982), was applied, using the COMPVAR program (Crusio, 1991). If epistasis is absent, the ratio kD = [d]2/D estimates the number of effective genetic factors exerting additive-genetic influences on the character under investigation (Mather & Jinks, 1982; see also Crusio, 1992). In the case of a single-gene difference between two strains, this ratio should approach 1, but we should keep in mind that this estimate is very weak (Zeng, Houle, & Cockerham, 1990).

However, the monogenic hypothesis can be tested additionally by calculating expected generation variances from the estimates of the genetic components of means of the best-fitting model. This procedure is based on the fact that in the monogenic case the following equations will hold true: D = [d]2, H = [h]2, and S (hd) = [h] x [d]. Observed and expected generation variances are then compared by means of a c2-test. A significant c2 indicates rejection of the monogenic hypothesis.

Finally, we applied a nonparametric test, developed by Collins (1967; an illustrative worked example of this method may be found in Tully & Hirsch, 1982). This test is based on the principle that, if a monogenic difference exists between two parental strains, the distributions expected in the segregating generations (F2 and backcrosses) can be estimated from those observed in the parental and F1 generations using simple Mendelian segregation ratios. The distributions of the different generations are assessed by determining the frequencies in classes chosen in such a way that they contain about equal numbers of individuals (which is not a necessary condition, however). Expected distributions were estimated using a maximum-likelihood procedure and compared with the observed ones by a G-test.


Table 2 shows the untransformed results of the aggression test for all generations of the Mendelian cross between ABG and AB//Halle. Adequate scales could be found for all variables and are shown in Table 3, which also presents the results of the Kerbusch model-searching procedure. The results of the analysis of the generation variances are entered in Table 4, whereas Table 5 presents a comparison of the observed variances with those expected under a monogenic hypothesis. In what follows, only the more important findings are described and discussed; the details of analysis can be studied by perusing the tables.

Tail rattling (TR). The Kerbusch procedure rendered a relatively simple mdh model. However, such a simple model did not fit the observed generation variances. Parameter H was even estimated to be negative, a theoretical impossibility. Such a result indicates either absent or very low dominance (rendering a negative estimate because of sampling error), or violations of the assumptions underlying the analysis. The low estimate of kD should therefore be interpreted with caution. The observed generation variances differed significantly from those expected if only a single-gene difference were involved. This result was confirmed by the Collins analysis, which also rejected the monogenic hypothesis (G = 26.09, df = 3, p < 0.001).

Attacks (AT). An mdh model did fit the observed generation means (c2 = 0.6, df = 4, p > 0.86) but had to be rejected because a more complex model fitted the data significantly better. In consequence, a less parsimonious mdhdrj model had to be accepted for this variable, rendering the low value obtained for kD rather dubious. It should be noted, however, that both the reciprocal effect and the [j]­type epistasis were estimated to be quite small. A simple, non-epistatic model did fit the observed generation variances. Still, the presence of epistasis indicates a polygenic inheritance and, in fact, the generation variances expected with a single-gene inheritance deviated from the observed ones. Calculating the expected variances on the basis of estimates obtained with an mdh model rendered a similar outcome (c2 = 9.50, df = 3, p < 0.05). The results of Collins' nonparametric method confirmed a polygenic inheritance for this variable, too (G = 26.35, df = 5, p < 0.001).

Aggressive grooming (AG). Only a very simple md model was needed to explain the variation between the generation means. Indeed, a model including dominance did not fit the observed generation variances (Table 4). Yet, a reduced model comprising only the parameters E and D, did not explain the data any better (c2 = 17.74, df = 5, p < 0.01). As in both these analyses a negative estimate was obtained for D, kD could not be calculated. The above situation is without doubt due to the fact that the variances of the nonsegregating generations are larger than those of the segregating ones. Not surprisingly, both the parametric and the nonparametric (G = 18.25, df = 5, p < 0.01) tests rejected the hypothesis of a monogenic difference between strains G and H.

Sum of aggressive acts (SAA). The results of the Kerbusch analysis of this variable were similar to those obtained for AT: although an mdh model fitted the data, it had to be rejected in favor of a less parsimonious one, because a more complicated model showed a significantly better fit. This time, however, the estimate obtained for [j] was more substantial. Consequently, problems were encountered in the analysis of the variances, where a negative estimate was obtained for H. Therefore, the low estimate for kD should again be viewed with caution. Once more, generation variances expected under a monogenic hypothesis deviated significantly from the observed ones (Table 5), even when the expectations were derived from an mdh model (c2 = 16.46, df = 3, p < 0.001). This result was confirmed by the Collins test (9 phenotypical classes: G = 40.87, df = 8, p < 0.001). The same conclusion was reached when two classes were used, based on the 2.0 threshold proposed by Schneider et al. (1992): G = 22.46, df = 1, p < 0.001.

Latency to the first attack (L1). Simple models containing only additive-genetic effects and dominance fitted both the observed means and variances. In this case, the low kD may therefore be taken to indicate a possible monogenic difference. A negative estimate was obtained for H, which thus is probably only very small. The generation variances did not deviate significantly from those expected under a monogenic hypothesis. However, the latter hypothesis was rejected by the Collins test (G = 20.58, df = 6, p < 0.01).


In the present study we investigated the genetic bases of the aggressive behavior of male mice stemming from a complete Mendelian cross between the closely related inbred strains ABG and AB//Halle. These strains show opposite extremes for intermale aggression, but the genetic distance between them is relatively small (Schneider et al., 1992), leading to the hypothesis that the behavioral difference might be correlated with allelic variation at one single locus only.

Summarizing the results of our parametric quantitative-genetic analyses, and leaving L1 besides for the moment, this hypothesis has to be rejected even for those variables where the Kerbusch model-searching procedure rendered non-epistatic models, i.e., for tail rattling (TR) and aggressive grooming (AG). Accordingly, the nonparametric Collins test yielded significant differences of expected and observed distributions for all behavioral variables. The hypothesis of a single-gene difference underlying the behavioral differentiation of strains ABG and AB//Halle must therefore be rejected.

For the latency to first attack (L1) the obtained results are less equivocal. Although dominance had been found for this variable in the analysis of means, a negative estimate for the variance due to this effect was obtained in the analysis of the generation variances (hence, H was assumed to be zero). This may indicate violations of the underlying assumptions in this case, too. Alternatively, only small dominance variation may be present and estimated to be negative due to sampling error. Still, the observed distributions of the segregating generations deviated strongly and significantly from those expected under a monogenic hypothesis in the Collins test. We conclude that the monogenic hypothesis has to be rejected for this variable, too.

Finally, two more specific results of the present quantitative-genetic analyses merit discussion. First, the direction of the dominance found in the analysis of the generation means was towards higher levels of aggression. Such directional dominance for high aggression in an unfamiliar situation has been reported only rarely, except when the opponent is nonaggressive but otherwise normal (see Hewitt & Broadhurst, 1983, for a review), as was the case here. Thus, if we interpret this genetic architecture in terms of selection pressures that have been exerted in the evolutionary past (Broadhurst & Jinks, 1974), there are some indications for an adaptive value of higher levels of aggression in this situation. However, as has already been stated before (Crusio, 1992), extreme care should be exersized when generalizing the results of a classical cross which involves only two inbred strains that, in addition, in the present case are also closely related.

Second, only for AT did we obtain a (very small) reciprocal effect. Such an effect may be due to pre- and/or postnatal maternal influences (Roubertoux, Nosten-Bertrand, & Carlier, 1990), to mitochondrial factors (Carlier, Nosten-Bertrand, & Michard-Vanhée, 1992), or to Y-chromosomal genes (Maxson, 1992). Earlier, we showed that there is no effect of the postnatal environment on the development of isolation-induced aggression in our mouse strains (Hoffmann et al., 1993). Except for the rather unlikely possibility of precisely balanced opposing reciprocal effects, the present results may be taken to indicate that the observed differences in aggressive behavior between ABG and AB//Halle are most probably due to autosomal genetic differences at more than one locus.


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, and New York: American Elsevier.

Cairns, R. B., MacCombie, D. J., & Hood, K. E. (1983). A developmental-genetic analysis of aggressive behavior in mice: I. Behavioral outcomes. Journal of Comparative Psychology, 97, 69-89.

Carlier, M., Nosten-Bertrand, M., & Michard-Vanhée, C. (1992). Separating genetic effects from maternal environmental effects. 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 (pp. 111-126). Amsterdam: Elsevier.

Carlier, M., Roubertoux, P.L., Kottler, M.L., & Degrelle, H. (1990). Y chromosome and aggression in strains of laboratory mice. Behavior Genetics, 20, 137-156.

Ciaranello, R. D., Lipsky, A., & Axelrod, J. (1974). Association between fighting behavior and catecholamine biosynthetic enzyme activity in two inbred mouse sublines. Proceedings of the National Academy of Sciences USA, 71, 3006-3008.

Collins, R. L. (1967). A general nonparametric theory of genetic analysis I. Application to the classical cross. Genetics, 56, 551.

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

Crusio, W. E. (1991). COMPVAR: A computer program for iteratively estimating components of variance in cross-breeding experiments. Journal of Heredity, 82, 359

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 (pp. 231-250). Amsterdam: Elsevier.

Crusio, W. E., & van Abeelen, J. H. F. (1987). Zinc-induced peripheral anosmia and behavioral responses to novelty in mice: A quantitative-genetic analysis. Behavioral and Neural Biology, 48, 63-82.

Eleftheriou, B. E., Bailey, D. W., & Denenberg, V. H. (1974). Genetic analysis of fighting behavior in mice. Physiology and Behavior, 13, 773-777.

Goldsmith, J. F., Brain, P. F., & Benton D. (1976). Effects of age at differential housing and the duration of individual housing/grouping on intermale fighting behavior and adrenocortical activity in TO strain mice. Aggressive Behavior, 2, 307-323.

Hahn, M. E., & Haber, S. B. (1982). The inheritance of agonistic behavior in male mice: A diallel analysis. Aggressive Behavior, 8, 19-38.

Hayman, B. I. (1960). Maximum likelihood estimation of genetic components of variation. Biometrics 16, 369-381.

Hewitt, J. K., & Broadhurst, P. L. (1983). Genetic architecture and the evolution of aggressive behavior. In: E. C. Simmel, M. E. Hahn, & J. K. Walters (Eds.), Aggressive Behavior: Genetic and Neural Approaches (pp. 37-66). Hillsdale, New Jersey: Lawrence Erlbaum.

Hoffmann, H. J., Schneider, R., & Crusio, W. E. (1993). Genetic analysis of isolation-induced aggression. II. Postnatal environmental influences in AB mice. Behavior Genetics, 23, (in press).

Kerbusch, J. M. L., van der Staay, F. J., & Hendriks, N. (1981). A searching procedure for transformations and models in a classical Mendelian cross breeding study. Behavior Genetics, 11, 239-254.

Mather, K., & Jinks, J. L. (1982). Biometrical Genetics, 3rd ed. London: Chapman and Hall.

Maxson, S.C. (1992). Methodological issues in genetic analyses of an agonistic behavior (offense) in male 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 (pp. 349-373). Amsterdam: Elsevier.

Maxson, S. C., Didier-Erickson, A., & Ogawa, S. (1989). The Y chromosome, social signals, and offense in mice. Behavioral and Neural Biology, 52, 251-259.

Maxson, S. C., Ginsburg, B. E., & Trattner A. (1979). Interaction of Y-chromosomal and autosomal gene(s) in the development of intermale aggression in mice. Behavior Genetics, 9, 219-226.

Orenberg, E. K., Renson, J., Elliott, G. R., Barchas, J. D., & Kessler, S. (1975). Genetic determination of aggressive behavior and brain cyclic AMP. Psychopharmacological Communications, 1, 99-107.

Popova, N. K., & Kulikov, A. V. (1986). Genetic analysis of "spontaneous" intermale aggression in mice. Aggressive Behavior, 12, 425-431.

Roubertoux, P., Nosten-Bertrand, M., & Carlier, M. (1990). Additive and interactive effects of genotype and maternal environment. Adances in the Study of Behavior, 19, 205-247.

Schneider, R., Hoffmann, H.J., Schicknick, H., & Moutier, R. (1992). Genetic analysis of isolation-induced aggression. I. Comparison between closely related inbred mouse strains. Behavioral and Neural Biology 57, 198-204.

Selmanoff, M. K., Goldman, B. D., Maxson, S. C., & Ginsburg, B. E. (1977). Correlated effects of the Y chromosome of mice on developmental changes in testosterone levels and intermale aggression. Life Sciences, 20, 359-366.

Selmanoff, M. K., Maxson, S. C., & Ginsburg, B. E. (1976). Chromosomal determinants of intermale aggressive behavior in inbred mice. Behavior Genetics, 6, 53-69.

Tully, T., & Hirsch, J. (1982). Behaviour-genetic analysis of Phormia regina II. Detection of a single, major-gene effect from behavioural variation for central excitatory state (CES) using hybrid crosses. Animal Behaviour, 30, 1193-1202.

Zeng, Z.-B., Houle, D., & Cockerham, C. C. (1990). How informative is Wright's estimator of the number of genes affecting a quantitative character? Genetics, 126, 235-247.

Table 1: Components of the Means of Generations from a Classical Cross between Two Inbred Mouse Strains ABG (G) and AB//Halle (H)
Generation Genetical Parameters
Components of means Components of variances
m [d] [h] [dr] [i] [j] [l] E D H S(dh)
P1 (H) 1 1 0 1 1 0 0 1 0 0 0
P2 (G) 1 -1 0 -1 1 0 0 1 0 0 0
F1 (HG) 1 0 1 1 0 0 1 1 0 0 0
F1 (GH) 1 0 1 -1 0 0 1 1 0 0 0
F2 (GHxGH) 1 0 0.5 0 0 0 0.25 1 0.50 0.25 0
B1 (GHxH) 1 0.5 0.5 0 0.25 0.25 0.25 1 0.25 0.25 -0.5
B2 (GHxG) 1 -0.5 0.5 0 0.25 -0.25 0.25 1 0.25 0.25 0.5

See test for explanation

Table 2: Results of the Aggression Test for the Different Generations of the Classical Cross between the Inbred Mouse Strains ABG (G) and AB//Hallea
Generation Behavioural variable
P1 (H) 41 0.71 ± 0.10 2.42 ± 0.27 0.79 ± 0.06 3.92 ± 0.32 261.0 ± 28.2
P2 (G) 34 0.07 ± 0.02 0.41 ± 0.13 0.44 ± 0.05 0.93 ± 0.18 499.4 ± 24.8
F1 (HG) 20 0.34 ± 0.07 3.22 ± 0.55 0.80 ± 0.09 4.35 ± 0.58 284.5 ± 42.5
F1 (GH) 20 0.52 ± 0.09 2.90 ± 0.44 0.66 ± 0.07 4.07 ± 0.45 231.5 ± 40.5
F2 (GHxGH) 131 0.39 ± 0.05 2.19 ± 0.18 0.57 ± 0.02 3.15 ± 0.21 301.3 ± 17.9
B1 (GHxH) 60 0.51 ± 0.05 2.84 ± 0.26 0.73 ± 0.04 4.08 ± 0.27 234.5 ± 21.8
B2 (GHxG) 80 0.16 ± 0.03 1.25 ± 0.15 0.57 ± 0.04 1.98 ± 0.17 398.4 ± 20.1

aUntransformed means ± SEM; n = number of observations; TR = Tail rattling; AT = Attack; AG = Aggressive grooming; SAA = Sum of aggressive acts per minute; L1 = Latency to first attack (in sec).

Table 3: Estimates for the Genetical Parameters and Goodness of Fit for the Best-fitting Models Found with Kerbusch's Procedurea
Variable TR AT AG SAA L1
x1/3 x1/3 + ln(x+1) x2
m 0.45 0.80 2.02 1.02 182320
[d] 0.32 0.45 0.19 0.48 -91547
[h] 0.16 0.50 - 0.47 -91494
[dr] - -0.03 - - -
[i] - - - - -
[j] - 0.09 - 0.20 -
[l] - - - - -
c2 6.421 0.005 6.962 0.372 1.768
df 4 2 5 3 4
p < 0.169 0.637 0.222 0.838 0.818

a See Table 2 for abbreviations.

Table 4: Estimates for the Components of the Generation Variances, the Number of Genetic Factors Involved (kD) and Goodness of Fita
Behavioural parameter TR AT AG SAA L1 x 107
E 0.11 0.24 0.14 0.23 1394
D 0.27 0.44 0.00 0.52 1535
H 0.00 0.00 0.01 0.00 0
S(dh) 0.04 0.13 0.04 0.08 754
c2 24.34 11.30 17.46 19.51 3.45
df 3 3 5 3 3
p < 0.001 0.05 0.01 0.01 0.33
kD 0.39 0.46 b 0.45 0.55

a See Table 2 for abreviations.

b Not calculated.

Table 5: Observed and Expected Generation Variances Calculated Assuming a Single-gene Differencea
Behavioral variable TR AT AG SAA L1 x 107
Eb 0.11 0.24 0.14 0.23 1394
D 0.11 0.20 0.04 0.23 838
H 0.03 0.26 - 0.22 837
S(dh) 0.05 0.23 - 0.23 838
Generation obs. exp. obs. exp. obs. exp. obs. exp. obs. exp.
P1 (H) 0.14 0.11 0.20 0.24 0.12 0.14 0.17 0.23 1360 1394
P2 (G) 0.08 0.11 0.24 0.24 0.15 0.14 0.20 0.23 1420 1394
F1 (HG) 0.11 0.11 0.40 0.24 0.21 0.14 0.45 0.23 1710 1394
F1 (GH) 0.09 0.11 0.18 0.24 0.09 0.14 0.21 0.23 1100 1394
F2 (GHxGH) 0.17 0.17 0.38 0.41 0.08 0.16 0.38 0.41 1900 2022
B1 (GHxH) 0.08 0.12 0.20 0.25 0.09 0.15 0.21 0.23 1140 1394
B2 (GHxG) 0.12 0.17 0.34 0.47 0.13 0.15 0.30 0.46 1900 2232
c2 10.42 9.35 22.22 15.58 3.02
df 3 3 5 3 3
p < 0.05 0.05 0.001 0.01 0.39

a See Table II for abbreviations

b Components of generation variances calculated from estimates of components of means assuming a non-epistatic model (monogenic hypothesis).