Animal behaviour

ANIMAL BEHAVIOUR, 2002, 63, 000–000
doi:10.1006/anbe.2002.3046, available online at http://www.idealibrary.com on
Family, sex and testosterone effects on garter snake behaviour RICHARD B. KING
Department of Biological Sciences, Northern Illinois University, DeKalb (Received 7 August 2000; initial acceptance 17 October 2000; final acceptance 7 December 2001; MS. number: A8850R) To better understand how genes and hormones interact to affect behaviour in nature, I used a factorialdesign to test for effects of family membership, sex and testosterone level on activity and defensivebehaviour of the common garter snake, Thamnophis sirtalis. Behaviours (latency to move, defensivestrikes, response distance) were scored prior to hormone manipulation (when snakes were 39 days of age),while sham or testosterone-containing implants were in place (190 days), following implant removal andsimulated hibernation (284 days), and when snakes were 428 days of age. Family membership hadpervasive effects on all three behaviours and on their ontogenetic trajectories, suggesting strong geneticor maternal components to behavioural variation. Sex had a significant effect on the number of defensivestrikes; females struck more frequently than males, but ontogenetic trajectories were similar between thesexes. Testosterone manipulation also had an effect on strikes: snakes in the elevated-testosteronetreatment group struck less frequently than shams while implants were in place. Sex and treatment effectson latency to move and response distance were lacking. Family*treatment interaction effects were lackingfor latency to move and number of defensive strikes but were present for response distance. Possibly,genetic or maternally induced variation in strikes is mediated through variation in circulating hormonelevels, whereas variation in response distance is mediated through receptor-level phenomena.
 2002 The Association for the Study of Animal Behaviour The interplay between genes and hormones is likely to be ences are often investigated through hormone assays important in shaping behavioural evolution (e.g. Crews & and experimental manipulations (hormone implants, Moore 1986; Crews 1987; Moore 1991; Ketterson & Nolan ablation of endocrine glands) without regard to related- 1992). Evidence for such an interplay is accumulating ness among individuals (Nelson 1995). Fortunately, from twin studies and pedigree analyses of humans (e.g.
methods used to investigate genetic and hormonal influ- Mendlewicz et al. 1999; An et al. 2000) and strain com- ences in nature need not be mutually exclusive. By parisons and controlled breeding designs in poultry, live- incorporating hormonal assays and manipulations into stock and laboratory rodents (Bates et al. 1986; Scott & analyses of variation within and between groups of rela- Washburn 1988). Most frequently, these studies have tives, behavioural effects of genes and hormones can be demonstrated high heritability for circulating levels studied simultaneously. This paper reports the results of of hormones and hormone-binding proteins. Linkages such an investigation of behaviour in the common garter between heritable variation in hormone levels and phenotypic traits, such as morphology, life history or High levels of genetic variation have been found in behaviour, have been found less frequently (e.g. Bates Thamnophis and allied genera for a range of behavioural et al. 1986; Gupta & Brush 1998; Otremski et al. 2000).
traits (reviewed in Brodie & Garland 1993; Burghardt & Especially rare are studies integrating genetic and hormo- Schwartz 1999). Behaviours related to prey preference, nal influences on behaviour in undomesticated species predator avoidance and escape have been particularly (but see Fairbairn & Roff 1999; Zera 1999; Zera & Huang well studied, are repeatable between trials and show high 1999), perhaps due to differences in the methods of levels of constancy over time (Brodie 1993; Brodie & investigation typically used by behavioural geneticists Russell 1999; Herzog & Burghardt 1988). Quantitative and endocrinologists. Genetic effects are frequently genetic analysis provides heritability estimates of 0.37– assessed from patterns of variation within and between 0.42 for predator avoidance and escape behaviours groups of relatives or from controlled breeding designs (Arnold & Bennett 1984; Brodie 1989; Garland 1988).
(Falconer & MacKay 1996). In contrast, hormonal influ- Studies of hormonal influences on garter snake behaviourhave focused mostly on courtship behaviour. Steroid Correspondence: R. B. King, Department of Biological Sciences,Northern Illinois University, DeKalb, IL 60115, U.S.A. (email: hormones appear to play an organizational role in behavioural differentiation between males and females;  2002 The Association for the Study of Animal Behaviour ANIMAL BEHAVIOUR, 63, 0
testosterone manipulations early in life (before snakes Table 1. Year and site of collection of gravid female garter snakes
first enter hibernation) modify expression of sex-typical courtship behaviour later in life (Crews 1985; Moore & Lindzey 1992; Whittier & Tokarz 1992).
behaviours in garter snakes (e.g. behaviours for whichgenetic data are available) is unknown. However, in somespecies of garter snakes, predator avoidance behaviour (e.g. defensive strikes) of neonates differs between males and females (Herzog & Burghardt 1986; Scudder & Burghardt 1983), suggesting that gonadal hormones might be involved. In addition, male garter snakes emerge from hibernation earlier than females andactively search for potential mates, whereas females are *See Figure 2 in Lawson & King (1996) for locations of study sites.
less active upon emergence. If testosterone plays anorganizational role in mate-searching behaviour (as itdoes in other aspects of courtship, Crews et al. 1985), variation is made easier by dividing these events into two males and females might show more general differences groups: factors that influence the level of circulating in activity as well (e.g. propensity to flee, Shine et al.
hormones (e.g. responsiveness to environmental stimuli, 2000). Behaviours scored in this study include latency to responsiveness to releasing hormones, rate of secretion, move, a measure of activity level (Herzog & Burghardt androgen binding proteins in the plasma, hormone half- 1986) and defensive behaviour and response distance, life), and factors that influence an organism’s response to measures of response to an approaching threat (e.g.
a given level of hormone (receptor- and postreceptor- Burghardt 1983; Arnold & Bennett 1984; Formanowicz level phenomena, including receptor density, receptor et al. 1990). Thus, this investigation serves as a test for sex affinity, efficiency of conversion of hormone to an active differences in nonreproductive behaviours as well as for form and neuroendocrine interactions). Two research possible effects of family membership and testosterone strategies for detecting genetic variation in hormone effects are then evident. One strategy is to test for genetic Garter snakes have several logistical attributes that variation in levels of circulating hormones. A second make them well suited for the work described here.
strategy is to manipulate hormone levels and test for Among the most important of these is large family size genetic variation in behavioural responses.
(averaging 15.6 and ranging from 6 to 31 among 28 I used this second strategy in the research described captive-born families; R. B. King, unpublished data), here. Offspring born to wild-caught females were divided which allows for an experimental design in which sex, into three groups, a sham-manipulated group and hormone and family effects can be tested simultaneously.
two hormone-manipulated groups. In one hormone- Whereas previous studies have attempted to avoid the manipulated group, testosterone levels were elevated via potentially confounding effects of family membership on testosterone-containing implants. In the other, testoster- responses to hormone manipulation (Crews 1985; Crews one levels were functionally reduced via implants con- et al. 1985; Shine & Crews 1988), they have not explicitly taining flutamide, which binds competitively with tested for family effects. Another attribute of garter snakes testosterone receptors, thus blocking testosterone’s action is that hormonal manipulations can be carried out easily (Liao et al. 1974; Neumann et al. 1977; Alexandre & via hormone-containing implants (e.g. Crews 1985; Balthazart 1987). The goal of these manipulations was Crews et al. 1985). Finally, the social behaviour of garter to mask individual variation in testosterone levels so snakes consists primarily of aggregative and reproductive that differences in receptor-level phenomena (e.g.
behaviours; intraspecific aggression appears to be lacking receptor density, receptor affinity) might be detected.
(Burghardt 1983; Gillingham 1987; Ford & Burghardt Comparisons between sham- and hormone-manipulated 1993). Thus, androgen levels are unlikely to be modified siblings provide a test for an effect of testosterone by aggressive interactions as may occur in territorial manipulation, whereas comparisons of treatment effects among litters provide a test for possibly heritable The effects of hormones on behaviour typically result variation in response to hormone manipulation.
from a cascade of events and genetic variation mightoccur at any point in this cascade. The effect of testoster-one starts with an internal or external environmental stimulus triggering the production of releasing hormones by the hypothalamus, the releasing hormones triggerrelease of gonadotropins by the anterior pituitary I obtained study animals by collecting gravid females in and these subsequently trigger secretion of testosterone.
the wild (Ottawa Co., Ohio, U.S.A.) and maintaining Testosterone travels to receptors in the central nervous them in captivity until parturition. Females were col- system where it is involved in regulation of gene tran- lected in late May and early June 1994, 1995 and 1996 scription which ultimately affects behaviour and mor- (Table 1) and housed individually until parturition.
phology (Hadley 1984). The task of detecting genetic Following parturition in late July–early August, I classified KING: GARTER SNAKE BEHAVIOUR
Table 2. Timing of behavioural tests, hormone manipulation, simu-
ing testosterone levels and the time course of hormone lated hibernation and collection of blood samples from neonatal manipulation, I collected blood samples from neonates at 195 days of age (while implants were in place), 318 days of age and 437 days of age. I also scored neonates for a setof morphological characters that were analysed separately(unpublished data). Upon completion of the study, sur- viving animals were returned to the wild or maintained incaptivity for use in captive breeding.
Hormone Manipulation and Assay
Hormone levels were manipulated using subcutaneous implants. Implants consisted of 7-mm lengths of empty diameter, Baxter T5715-3), tubing filled with 4–5 mm of crystalline testosterone (Sigma T-1500), or tubing filledwith 4–5 mm of flutamide (Sigma F-9397). Implants were *Ages shown refer to the start of behavioural testing.
sealed with silicon adhesive and inserted at mid-bodythrough a small lateral incision 2–3 scale rows above theventer (following Crews 1985; Crews et al. 1985). Prior to neonates by sex by everting the hemipenes of males, then implantation, snakes were immobilized via hypothermia.
measured them and placed them in individual cages for Ethanol was used to clean the skin prior to implantation.
captive maintenance. Fresh water was available continu- ously and food (large earthworms for gravid females, Laboratories Abbott Park, Illinois, U.S.A.) and cloth first- small earthworms or earthworm pieces for neonates) was aid tape was used to seal incisions.
provided three times a week. The room in which snakes Blood samples (100–300 l) were collected from caudal were housed was maintained at about 26 C and 50% RH with a 12:12 h light:dark photic cycle. A heating cable heparinized syringe (Bush & Smeller 1978). Blood running under one end of the cages provided gravid samples were centrifuged and the plasma fraction was females with a thermal gradient ranging from room frozen for hormone analysis following completion of bleed collection from all snakes. Testosterone levels were I divided litters containing at least four males and four determined by radioimmunoassay as described by King females into sham- and hormone-manipulation treat- et al. (2000). Steroids were ether-extracted from plasma ments. Most families were divided into sham- and (50–200 l diluted to 1 ml with distilled water) and stan- elevated-testosterone treatment groups, or sham- and dards (1.95–500 pg testosterone in 100 l methanol) flutamide-treatment groups, but a few large families were added to a radioimmunoassay containing 3H testosterone divided into sham-, elevated-testosterone and flutamide- (ca. 5000 cpm; New England Nuclear NET-370) and treatment groups. Within families, at least two males and testosterone antibody (1:60 000 dilution). Samples and two females were assigned to each treatment. I scored standards were incubated overnight at 4 C, after which neonates for three behaviours at 39 days of age (Table 2).
bound and unbound steroid was separated using a Implants were inserted when snakes were 109 days of age charcoal-dextran suspension. After centrifugation, the and behaviours were scored again when snakes were 190 supernatant from each sample was added to vials contain- days of age, after implants had been in place for 81 days.
ing scintillation cocktail (Bio-Safe II, Research Products Implants were then removed and snakes were placed in International, Mt Prospect, Illinois, U.S.A.), and counted simulated hibernation (7 C, 0:24 h LD photic cycle) for in a Beckman liquid scintillation counter. Testosterone 70 days. Behaviours were scored again following removal levels were determined using a logit–log curve-fitting from hibernation when snakes were 284 days of age. A programme. Testosterone levels at 195, 318 and 437 days final set of behavioural scores was obtained when snakes were determined in separate assays. Intra-assay variability were 428 days of age. The schedule of hormone manipu- was 10%, interassay variabitity was 13%, and assay lation and hibernation used here approximately parallels sensitivity ranged from 3 to 3327 pg per sample at 195 that used by Crews (1985) in his analysis of garter snake days, 5–1275 pg per sample at 318 days, and 6–2065 pg courtship behaviour. The schedule of behavioural tests was designed to detect (1) family and sex effects on body used here (provided by G. Niswender, Colorado behaviour soon after birth (39 days), (2) activational State University), had high cross-reactivity with 5- effects of testosterone on behaviour (190 days), and (3) organizational effects of testosterone on behaviour fol- reactivity with androstenedione (2%). However, because lowing emergence from hibernation (284 days) or later in dihydrotestosterone accounts for a relatively small pro- life prior to adulthood (428 days) (in nature, garter snakes portion of total androgens in garter snakes (Crews reach adulthood at 2–3 years of age, Rossman et al. 1996).
et al. 1985; Mason & Crews, 1985), the assay provides a To document the effectiveness of implants in manipulat- reasonably accurate estimate of testosterone levels.
ANIMAL BEHAVIOUR, 63, 0
Behavioural Tests
age and repeated measures multivariate analysis of vari-ance (MANOVA) of behaviour over the entire course of I scored three different measures of garter snake behav- the experiment (O’Brien & Kaiser 1985; Potvin et al.
iour, latency to move, defensive behaviour and response 1990; von Ende 1993). In these analyses, scores for a distance. Each behaviour was scored on two consecutive given behaviour at different ages were dependent vari- days and averaged across days for analysis. Behavioural ables; sex, family and treatment were between-subjects tests were conducted in an environmental room main- factors; and time (MANOVA only) was a within-subjects tained at 22 C. Latency to move and defensive behaviour factor. MANOVA identifies two general sources of vari- (total number of strikes at a stationary and a moving ation: ‘between-subjects’ and ‘within-subjects’ effects.
stimulus) were scored sequentially in a carpeted 75-cm Between-subjects effects (main effects of and interactions diameter arena. I measured latency to move by placing a among the between-subjects factors) reflect differences in snake in the centre of the arena under an 8-cm diameter a given response variable among factor levels over the opaque cover for 2 min, raising the cover from behind entire course of an experiment and can arise during an one-way glass, and recording the time elapsed until the experiment or from pre-existing differences present at the snake moved its head outside an 11-cm diameter circle start of an experiment. In the present study, between- marked on the carpet. After 30 s, I recorded the number of subjects effects were useful in identifying overall effects of strikes at a stationary stimulus. I held the stimulus (my sex and family. In contrast, within-subjects effects (main finger) about 2 cm in front of the snake’s head and effects of within-subjects factors and interactions among recorded the number of strikes in a 1-min interval. After within-subjects and between-subjects factors) reflect dif- another 30 s, I recorded the number of strikes at a moving ferences in how the score of a given response variable stimulus. In this test I wiggled my finger rapidly from side changes over time. Within-subjects effects, if significant, to side, and again, recorded the number of strikes in provide unambiguous evidence for a treatment effect on 1 min (strikes at a stationary stimulus and strikes at a moving stimulus follow Herzog & Burghardt 1986; see I conducted analyses using SPSS 10.0 statistical soft- also King & Turmo 1997). I summed the number of strikes ware. All ANOVA and MANOVA made use of type III at the stationary stimulus and at the moving stimulus for sums of squares. Pillai’s trace was used in MANOVA significance testing. The assumption of equality of vari- I recorded response distance beginning on the day ance was tested using Levene’s test. Where this assump- following completion of the other behavioural measures.
For this test, I placed a snake under an 8-cm diameter examined to ensure that there was no systematic relation- ship between group means and variances. Examination of the frequency distribution of residuals revealed no paper silhouette of a bird’s head measuring 4 cm marked departures from normality. Each behaviour was 8 cm high with beak and eyes marked in black) tested separately (see Results for evidence of the absence was positioned at the other end of the arena. Plexiglas of any strong correlation among behaviours). Because and a removable partition separated the snake from the some mortality occurred over the course of the investiga- rest of the arena. After 2 min, I lifted the cover and tion, sample size was maximized over each time interval removed the opaque partition and then moved the pred- by conducting separate analyses of behaviour at 39–190 ator from side to side and towards the snake in 10-cm days, 39–284 days and 39–428 days. For analyses of increments using a long rod from behind one-way glass. I behaviour at 39–284 days and 39–428 days, sample size recorded the distance at which the snake first responded criteria were relaxed from two to one offspring per sex per (e.g. by orienting towards the threat or fleeing rearward).
treatment. This precluded testing the highest-order inter- Snakes that showed no change in behaviour were given a action in these analyses but this interaction was consist- 10. Thus, snakes that responded to a distant ently nonsignificant at 39 days of age (see Results) and stimulus had high response distance scores and snakes over 39–190 days. Comparison of analyses using the that failed to respond or responded only to a nearby original versus relaxed sample size criteria revealed no qualitative difference in the conclusions reached (analy- Behavioural tests were conducted blind to sex and ses not shown). For analyses of behaviour at 39–284 days treatment of snakes and to previous behavioural scores.
and 39–428 days, reverse Helmert contrasts were used to To meet assumptions of normality and equality of vari- identify over which time intervals significant changes in ance more closely in the analyses described below, I behaviour occurred. Except as noted below, results were transformed latency to move using natural logarithms, qualitatively similar for analyses of behaviour at 39–190 the number of strikes by adding one and computing the days, 39–284 days and 39–428 days and so for simplicity square root, and the response distance by adding 10 and only analyses of the entire 39–428 day interval are pre- dividing by 110 (converting response distance to a pro- sented here. Observed effect size ( 2, computed by SPSS as portion) and then computing the arcsine of the square of committing a type II error given the observed effectsize) were computed for all analyses. Observed power is Analysis
necessarily low when observed effect size is small. For this I tested for sex, family and treatment effects using reason, minimum detectable effect size ( 2 analysis of variance (ANOVA) of behaviour at 39 days of h*F/(dfh*F + dfe KING: GARTER SNAKE BEHAVIOUR
Table 3. P values from repeated measures multivariate analysis of variance of behaviour in sham- versus
flutamide-treated garter snakes
<0.001
<0.001
<0.001
Because the primary interest of this analysis is in treatment effects on changes in behaviour over time (i.e. thetime*treatment interaction), only within-subjects sources of variation are shown. P values less than 0.05 are shown to hypothesis and error degrees of freedom and F is the successive days. Repeatability was highest when snakes were tested at 39 days of age (r =0.62, 0.78 and 0.30 for page 177). Effect sizes 2=0.01, 0.06 and 0.14 correspond latency to move, strikes and response distance (N=377 to small, medium and large effect sizes (Cohen 1988; animals from 26 families). Repeatability at 190, 284 and 428 days of age ranged from 0.43 to 0.56 for latency to I initially compared sham- and flutamide-treated move, 0.65–0.73 for strikes and 0.09–0.19 for response snakes to determine whether these two groups might be distance (N=213, 176 and 165 sham-treated animals from pooled in order to increase sample size in subsequent 24 families at 190, 284 and 428 days, respectively).
analyses. The general absence of time*treatment interac-tion effects while implants were in place (see Results),together with the observation that testosterone levelswere uniformly low in all but the elevated-testosterone Flutamide-treatment Effects
treatment snakes, suggested that the use of flutamide tofunctionally reduce testosterone levels was unnecessary.
Therefore, I pooled sham- and flutamide-treated snakes Ninety-three offspring (43 males, 50 females) belonging and conducted two subsequent sets of analyses. The first to seven families were divided into sham- and flutamide- involved tests for family and sex effects on behaviour in treatment groups (N=50 and 43, respectively). MANOVA the pooled sham- and flutamide-treatment groups and revealed no significant within-subjects effects of treat- the second involved tests for treatment effects between ment (i.e. no time*treatment, time*family*treatment, the elevated-testosterone treatment group and the pooled time*sex*treatment, time*family*sex*treatment effects) sham- and flutamide-treatment groups. Comparison of on latency to move or response distance over any time analyses in which sham- and flutamide-treatment snakes interval (Table 3). MANOVA revealed a significant were not pooled with those in which these treatments time*treatment effect on strikes over the entire experi- were pooled revealed no qualitative difference in the ment (Table 3, 39–428 days; P=0.047). However, this conclusions reached (analyses not shown).
source of variation was far from significant over 39–190days (P=0.745) or 39–284 days (P=0.958). Further-more, time*treatment effect arose only during the 284–428 day Repeatability
Repeatability was computed as the intraclass corre- lation between measures of each behaviour across ANIMAL BEHAVIOUR, 63, 0
power of tests for within-subjects effects involving treat- (Table 4, Fig. 1) (power was sufficient to detect medium ment was sometimes low and minimum detectable effect size was large ( 2>0.14) over the entire experiment (Table3). However, analysis over 39–190 days had power suffi- Testosterone-treatment Effects
cient to detect a medium time*treatment effect size( 2=0.058). Analysis of covariance, with family and treat- Two hundred and sixty-nine offspring (132 males, 137 ment as factors and ln(snout–vent length) as a covariate, females) belonging to 17 families were used to test for revealed that ln(testosterone) did not differ between differences in behaviour between testosterone-treatment males belonging to sham- and flutamide-treatment group and the pooled sham- and flutamide-treatment groups while implants were in place (195 days: groups (N=113 and 156, respectively). Crystalline testos- terone was still present in implants upon their removal.
minimum 2=0.098) or following implant removal (318 Radioimmunoassay of blood samples collected when snakes were 195 days old (after completion of the second set of behavioural tests and before implants were removed) revealed that implants had a marked effect on 2 =0.115). Given these results, sham- and circulating testosterone levels. Testosterone levels aver- flutamide-treatment groups were pooled for the analyses aged 87.3 pg/ml (95% confidence interval=65.5, 116.4) among 66 sham- and flutamide-treatment males com-pared testosterone-treatment males and 8636.0 pg/ml (6840.4,10 903.1) in 49 testosterone-treatment females (means Family and Sex Effects
and confidence intervals backtransformed from naturallogarithms). Testosterone levels in elevated-testosterone Three hundred and seventy-seven offspring (188 males, treatment animals approached those reported in young 187 females) belonging to 26 families were used to test for male garter snakes during a pulse of high testosterone sex and family effects on behaviour prior to hormone that occurs shortly after birth (24 540–122 490 pg/ml, manipulation. ANOVA revealed highly significant family Table 1 in Crews et al. 1985). However, this pulse was not effects on all three behaviours (Table 4). In addition, evident in another natricine snake (King et al. 2000).
there was a highly significant effect of sex on strikes; Testosterone levels in elevated-testosterone treatment females struck more frequently than males. No significant animals were similar to those seen in adult male family*sex interactions were present. Power was sufficient garter snakes (1400–72 000 pg/ml, Table 1 in Weil 1985).
to detect small sex effects and medium to large family*sex Radioimmunoassay of blood samples collected when effects (Table 4). Consistent with ANOVA results, snakes where 318 days and 436 days of age revealed MANOVA revealed significant between-subjects effects of hormone manipulations had only temporary effects family on all three behaviours and of sex on strikes over on testosterone levels. At 318 days of age, testosterone the entire experiment (Table 4). There were significant levels averaged 78.1 pg/ml (95% confidence inter- within-subjects effects of time and of family (i.e. a val=62.0, 98.4) among 76 sham- and flutamide-treatment time*family interaction) on all three behaviours, indicat- males compared with 82.1 pg/ml (59.8, 112.8) in 53 ing that behavioural scores changed over time and that testosterone-treatment males. At 436 days of age, testos- the pattern of change differed among families (Table 4, terone levels averaged 1037.4 pg/ml (95% confidence Fig. 1). Reverse-Helmert contrasts indicated that signifi- interval=691.1, 1557.3) among 69 sham- and flutamide- cant time effects were present over all intervals except treatment males compared with 1192.1 pg/ml (730.8, when contrasting latency at 428 days with previous times 1944.8) in 50 testosterone-treatment males. Analysis of and when contrasting strikes at 190 days with 39 days covariance, with family and treatment as factors and (Table 4). Latency decreased slightly between 39 and 190 ln(snout–vent length) as a covariate, revealed no effect of days and increased following emergence from simulated treatment on ln(testosterone) at 318 days (F hibernation (Fig. 1). Strikes remained constant from 39 to 190 days, decreased following emergence from simulated hibernation (284 days) and then increased by 428 days observed power=0.353, minimum 2=0.035).
(Fig. 1). Response distance decreased from 39 to 190 to MANOVA over the full 428-day experiment revealed 284 days and then increased by 428 days (Fig. 1). Reverse- Helmert contrasts revealed that, with one exception, treatment (time*treatment, time*family*treatment, time* time*family effects were significant for all three behav- sex*treatment) were consistently nonsignificant (Table iours over all time intervals (Table 4). The one exception 5). For strikes, within-subjects effects involving treat- was the contrast of response distance at 428 days with ment were also nonsignificant over the full 428-day previous times. Time*family effects are evident in profile experiment. However, the time*treatment interaction plots (Fig. 1): some families showed increases while other approached significance over the full experiment (Table families showed decreases in a given behaviour over a 5, P=0.076) and was significant in analyses of strikes at given time interval. The time*sex interaction was consist- ently nonsignificant, indicating that changes in the behaviour of males and females occurred in parallel reverse-Helmert contrasts indicated that a significant KING: GARTER SNAKE BEHAVIOUR
ANIMAL BEHAVIOUR, 63, 0
Figure 1. Profile plots showing variation in estimated marginal means for latency to move (ln transformed seconds), number of strikes
(square-root transformed) and response distance (arcsine square-root transformed proportions) at 39 days, 190 days, 284 days and 428 days.
(a) Variation among families, with families represented by separate lines. (b) Variation between male (—"—) and female (– –C– –) garter
snakes (bars represent ±1 standard error of estimated mean). The between-subjects effect of family and the within-subjects time-by-family
interaction were significant for all three behaviours; the between-subjects effect of sex was significant for strikes. Snakes underwent simulated
hibernation between 213 and 283 days ( ).
KING: GARTER SNAKE BEHAVIOUR
Table 5. P values from repeated measures multivariate analysis of variance of behaviour in testosterone-treated versus pooled sham- and
flutamide-treated garter snakes
3,125 11.57 <0.001 0.218 0.060 0.999
<0.001
4.19 <0.001 0.283 0.124 1.000
<0.001
<0.001
<0.001
3,125 86.09 <0.001 0.674 0.060 1.000
<0.001
4.19 <0.001 0.484 0.124 1.000
<0.001
<0.001
3,125 43.81 <0.001 0.513 0.060 1.000
<0.001
<0.001
<0.001
2.28 <0.001 0.177 0.124 1.000
0.043 0.122 0.124 0.991
0.037 0.124 0.124 0.992
Because the primary interest of this analysis is in treatment effects on changes in behaviour over time (i.e. the time*treatment interaction), only within-subjects sources of variation are shown. P values less than 0.05 are shown in bold.
treatment effect occurred while testosterone-containingimplants were in place: the contrast between 190 days and 39 days was significant but other contrasts were nonsignificant (Table 5). The direction of the treatmenteffect paralleled the difference seen between males and females in that males (Fig. 1) and testosterone-treatmentanimals (Fig. 2) showed lower frequencies of strikes than did females and sham-treatment animals, respectively.
For response distance, the time*family*treatment interac- tion was significant over the full 428 day experiment but (time*treatment, time*sex*treatment) were nonsignifi-cant (Table 5). Again, reverse-Helmert contrasts indicated testosterone-containing implants were in place: only the contrast between 190 days and 39 days was significant(Table 5). However, the time*family*treatment effect was nonsignificant in analyses of response distance measured and at 39, 190 and 284 days (MANOVA: F Figure 2. Profile plot showing variation in estimated marginal means
P=0.091). Other within-subjects effects on response dis- testosterone-treated (—"—) and pooled sham- and flutamide- tance involving treatment (time*treatment, time*sex* treated garter snakes (– –x– –) (bars represent ±1 standard error of treatment) were consistently nonsignificant. Power was estimated mean). The within-subjects time-by-treatment interaction was significant over the 39–190-day interval. Implants were in place between 109 and 196 days ( ) and snakes underwent simulated time*family*treatement effects (Table 5).
hibernation between 213 and 283 days ( ).
ANIMAL BEHAVIOUR, 63, 0
Table 6. Correlations among measures of the same behaviour over time and among different behaviours
−0.251
−0.265
Entries represent Pearson product-moment correlations among residuals from MANOVA analysis of behaviour over the full 428-dayexperiment. N=167. Statistically significant correlations (following adjustment for multiple tests by use of α=0.05/48 for correlations amongdifferent behaviours and α=0.05/6 for correlations among measures of the same behaviour over time) are shown in bold. Correlationsobtained from analyses of behaviour at 39 days, 39–190 days and 39–284 days were similar in magnitude.
Correlations among Behaviours
analysis (Brodie & Garland 1993) to behaviours measuredat 39 days of age yields heritability estimates of 0.65 Pearson correlations among residuals of the three (approximate 95% confidence interval: 0.37, 0.95) for behavioural scores at 39 days were nonsignificant latency, 0.56 (0.30, 0.86) for strikes, and 0.33 (0.13, 0.56) for response distance (heritability was computed as =0.090). Significant positive correlations from a two-factor ANOVA with family and sex included were typically present between measures of the same as factors; confidence intervals were computed as in behaviour over time (Table 6). These correlations were Becker 1992, following modification for the inclusion of strongest for strikes (range 0.457–0.590), intermediate for sex as a factor). These estimates assume that litters consist latency to move (0.197–0.385), and weakest (and some- of full siblings and that maternal effects are negligible.
times nonsignificant) for response distance ( Evidence is mounting that in some natricines, including 0.286). Small but significant negative correlations were T. sirtalis, multiple paternity within litters is common- present between strikes at 190 days and response distance place (Barry et al. 1992; Garner 1998; Gibson & Falls at 190 days and between strikes at 284 days and response 1975; McCracken et al. 1999; Prosser 1999; Schwartz et al.
distance at 190 days (Table 6). Use of residuals removes 1989). By itself, multiple paternity should lead to under- the effects of sex, family membership and treatment on estimates of h2 using full sibling analysis. King et al.
(2001) have taken advantage of the occurrence ofmultiple paternity within litters to explore the possibility DISCUSSION
that maternal effects also contribute to between-familyvariation. Based on an analysis of four litters each sired by Perhaps the clearest result of this investigation is the two males (eight sireships total), it appears that maternal pervasive effect family membership had on garter snake effects may markedly inflate estimates of h2 in natricines behaviour. Such family effects have been found consist- obtained using full sibling analysis. The nature of these ently in studies of the behaviour of natricine snakes maternal effects remains unexplored and may include (garter snakes and their allies) (reviewed by Brodie & both maternal environmental effects (effects of the com- Garland 1993; Burghardt & Schwartz 1999; see also mon uterine environment shared by littermates) and Burghardt et al. 2000). However, this study goes further in maternal genetic effects (effects of maternal genotype on documenting the presence of significant time*family interactions, indicating that ontogenetic trajectories also In the present study, variation between families may vary among families (Fig. 1). Patterns of variation within have been inflated by year and site effects: gravid females and between families have been used previously to esti- were collected in 3 years and from five sites (Table 1).
mate the heritability of defensive behaviour in natricine Efforts were made to minimize year effects by using snakes (Brodie & Garland 1993; Burghardt & Schwartz uniform rearing and testing conditions. Furthermore, site 1999; Burghardt et al. 2000). Applying a full sibling effects are likely to be small because sites were separated KING: GARTER SNAKE BEHAVIOUR
by less than 27 km and molecular genetic analyses sug- in testosterone level between treatment groups were non- gest that gene flow among them is common (Bittner significant. This suggests that testosterone has an activa- 1999; Lawson & King 1996). Examination of family tional effect on strikes. In contrast, testosterone has an means revealed no consistent year or site effects on the organizational effect on garter snake courtship behaviour: results presented here. However, small but significant elevating testosterone levels early in life elicits courtship differences in behaviour have been observed in wild- behaviour later in life, well after effects on circulating caught adult garter snakes from these sites (R. B. King & T.
hormone levels have passed (Crews 1985).
Manipulation of testosterone level had a similar effect A second clear result of this investigation is the pres- on males and females (time*sex*treatment effects were ence of sex differences in strikes but not in latency or consistently nonsignificant). However, it is unknown response distance. For all three behaviours, my analyses whether this effect was a direct result of testosterone were of sufficient power to detect even small differences manipulation or an effect mediated through the conver- between males and females using Cohen’s (1988) effect sion of testosterone to dihydrotestosterone (DHT) or size criteria (Table 4). Thus, the absence of sex effects on oestrogen. Significantly, cells that concentrate testoster- latency and response distance appears to be biologically one and oestrogen are equally common in many areas of meaningful and not simply a result of type II error. The the brains of male and female garter snakes, including effect of sex on strikes differs from the family effect areas such as the amygdala, which is associated with described above in that it is smaller in magnitude and limbic responses like defensive behaviour (Halpern et al.
does not include an ontogenetic component: ontogenetic 1982). Further experiments that directly manipulate DHT changes in strikes among males parallel those among or oestrogen levels or that manipulate testosterone levels females (Fig. 1). Sex differences in behaviour have been in the presence of inhibitors such as 5- -reductase (the demonstrated in laboratory studies of some neonatal enzyme responsible for conversion of testosterone to natricines but not others. Male Butler’s garter snakes, T. DHT) or aromatase (the enzyme responsible for conver- butleri, and southern water snakes, Nerodia fasciata, strike sion of testosterone to oestrogen) would be useful (Moore more frequently than do females (Scudder & Burghardt 1983; Herzog & Burghardt 1986). In contrast, among the A fourth result of this investigation is the presence of a common garter snakes tested here, females struck more time*family*treatment interaction effect on response dis- frequently than males. Sex differences are apparently tance but not on latency to move and strikes: families lacking in other natricine snakes (e.g. T. melanogaster, responded differently to hormone manipulation for T. ordinoides, T. radix, N. cyclopion, N. rhombifera, N. sipe- response distance. This result suggests that variation in don; Scudder & Burghardt 1983; Herzog & Burghardt response distance among families may be mediated 1986; E. D. Brodie III, unpublished data; A. Queral-Regil & through receptor- or postreceptor-level phenomena. In R. B. King, unpublished data; R. B. King & D. Anderson, contrast, families responded similarly (or not at all) to unpublished data). However, detection of sex differences hormone manipulation for latency to move and strikes may require moderately large sample sizes. In contrast to although admittedly, my analyses were only of sufficient the results presented here, an earlier study of 90 common power to detect large effects using Cohen’s (1988) effect garter snakes born to seven females failed to reveal any size criteria (Table 5). Taken at face value, the lack of difference between males and females (Herzog & time*family*treatment interaction effects on latency and Burghardt 1986). Sex differences in diet, foraging behav- strikes suggests that variation between families in these iour, spatial patterns and movement have been observed in snakes in the wild (Gregory et al. 1987; Shine 1991; Reinert 1993). Whether the sex difference in strikes docu- among males, families differ in circulating testosterone mented here and in the studies cited above is related to levels (R. B. King, J. H. Cline & C. J. Hubbard, unpub- behavioural differences in nature is unknown.
lished data; see also King et al. 2001), it is possible that A third result of this investigation is an apparent effect genetic or maternally induced variation in some garter of testosterone manipulation on strikes. Although this snake behaviour (e.g. strikes) is mediated through pro- effect was relatively small, the difference between treat- cesses that influence circulating hormone levels (e.g.
ment groups parallels the difference seen between males responsiveness to environmental stimuli, responsive- and females. Males (the sex with intrinsically higher ness to releasing hormones, rate of secretion, androgen testosterone levels) struck less frequently than did binding proteins in the plasma, hormone half-life).
females and snakes that received implants containing Detecting an effect of individual variation in circulat- testosterone showed a decrease in strike frequency ing testosterone level on behaviour using the data relative to sham-manipulated animals. Sexually mono- gathered in this study is complicated by the fact that morphic behaviours (latency, response distance) were testosterone level changes over time, precluding its use as a covariate in the repeated measures analyses used here to although, admittedly, small treatment effects may have test for family, sex and treatment effects. In addition, gone undetected because my analyses were only of suffi- testosterone was not assayed in all individuals (e.g. pilot cient power to detect medium effects using Cohen’s assays indicated only marginally detectable levels in (1988) effect size criteria (Table 5). The effect of treatment sham- and flutamide-treatment females, and insufficient on strikes was only evident while implants were in place.
blood volume was obtained from some individuals). To Following implant removal, differences in behaviour and avoid these problems, I used analysis of covariance ANIMAL BEHAVIOUR, 63, 0
(ANOVA) to test for a possible effect of testosterone level, the course of this investigation (Fig. 1, Table 4). One sex and family on behaviour among 91 testosterone- interpretation is that larger (older) snakes are less vulner- treatment animals from 15 families at 190 days of age able to predators and thus are less responsive to an (while implants were in place). I also used ANCOVA to approaching threat. Temporal changes in latency to move test for an effect of testosterone level and family on and strikes may be associated with emergence from simu- behaviour using data on 73 males from 10 families at 284 lated hibernation. Snakes were slower to move and struck and 428 days (after implants had been removed and less frequency at 284 days (1–2 days after emergence) testosterone had returned to baseline). Covariation than on earlier test dates (Fig. 1, Table 4). Changes in between testosterone level and strikes was consistently behaviour associated with recovery of physiological func- nonsignificant despite the fact that strikes was the one tions following emergence from hibernation have been behaviour that most clearly showed a testosterone treat- reported in natural populations of garter snakes as well ment effect. Covariation between testosterone level and behaviour did achieve statistical significance for response North American natricine snakes have become a model distance among testosterone-treatment animals at 190 system in evolutionary quantitative genetics (Brodie & days of age (P=0.049) and for latency to move among Garland 1993; King & Lawson 1995; Arnold & Phillips males at 284 days (P=0.041) but not at other times. Thus, 1999). Much previous work with these snakes has focused evidence that individual variation in testosterone levels on estimating heritability, genetic correlation and selec- has detectable effects on behaviour is equivocal at best.
tion on quantitative characters in an effort to assess However, the results of thus study suggest that future potential for evolutionary change. These studies have experiments designed specifically to test for such effects been groundbreaking in demonstrating how genetic cor- relations among traits may constrain evolutionary Garter snakes grew markedly over the course of this change (Arnold 1988), how combinations of traits (e.g.
investigation, from a mean of 1.6 g at birth to a mean of behaviour and morphology) can be the target of correla- 19.7 g at 438 days of age. Furthermore, females exceed tional selection (Brodie 1989), and how gene flow can males in body size as adults, a difference thought to be slow adaptive evolution (King & Lawson 1995). In con- mediated by an inhibitory effect of testosterone on trast, the work reported here and other recent studies of growth of males (Crews et al. 1985). Thus, it is possible natricines have focused on how proximate mechanisms that the differences in garter snake behaviour among (e.g. hormonal pathways, maternal effects, experience) families, sexes and treatments reported here are attribu- mediate the expression of quantitative traits (Burghardt table to differences in body size. Several lines of evidence et al. 2000; King et al. 2001). This emphasis on proximate suggest that this is not the case (a more detailed analysis mechanisms is complimentary to an evolutionary quan- of family, sex and testosterone effects on garter snake titative genetic approach. One interest in evolutionary morphology will be presented elsewhere). (1) Neither quantitative genetics is the relative constancy of genetic mass nor snout–vent length (SVL) covaried significantly with behaviour at 39, 190 or 284 days of age. SVL did covariance matrix, G) (Arnold & Phillips 1999). A con-
covary significantly with latency to move at 428 days of stant G simplifies prediction of long-term evolutionary
age (ANOVA with sex, family and treatment as factors, change but implies that such change is constrained by the P=0.016). (2) Size residuals and behaviour residuals gen- specific form of G. Tests for constancy have typically
erated from mutivariate analyses that included latency, involved comparisons of G among populations or closely
strikes, response distance and mass or SVL as dependent related taxa (e.g. Arnold & Phillips 1999). Alternatively, variables, and sex, family and treatment as factors were insight into the constancy of G can come from investiga-
uncorrelated at 39, 190 and 284 days of age. Residual SVL tions of proximate mechanisms. Studies such as this one was positively correlated with residual latency at 428 days can reveal the degree to which suites of traits are influ- enced by common pathways (e.g. a single hormonal by this age, larger snakes were slower to move than were control mechanism) and the ease with which expression smaller snakes. (3) Differences in body size between males of such traits might become uncoupled. The observation and females and between treatment groups, although that number of strikes is influenced by testosterone levels, statistically significant, were small. At 438 days, males whereas latency to move and response distance are exceeded females in SVL by just 1.4% (358 versus not, suggests that these behaviours may evolve relatively 353 mm) and females exceeded males in mass by 2.6% (20.1 versus 19.6 g). At 195 days (immediately following More generally, this study demonstrates the utility of implant removal), testosterone-treated animals exceeded combining quantitative genetic and endocrinological sham- and flutamide-treated animals in SVL by less than approaches to the study of behaviour. Evidence is 1% (276 versus 278 mm) and sham- and flutamide- accumulating for heritable variation in circulating levels treated animals exceeded testosterone-treated animals in of a range of hormones and hormone-binding proteins (e.g. Jaquish et al. 0000; Meikle et al. 1986, 1988a, b; Although variation in body size does not appear to Zarazaga et al. 1998). Furthermore, a number of studies explain the family, sex or treatment effects on behaviour provide evidence for correlations between hormonal vari- reported here, increasing body size may have contributed ation and variation in other phenotypic traits, including to temporal changes in behaviour. This is especially true behaviours related to activity, avoidance and aggression for response distance, which decreased consistently over (e.g. Glowa et al. 1992; Compaan et al. 1993; Gupta & KING: GARTER SNAKE BEHAVIOUR
Brush 1998). These studies lend credence to the proposal Becker, W. A. 1992. Manual of Quantitative Genetics. 5th edn.
that genetic variation in phenotype may be mediated Pullman, Washington: Academic Enterprises.
through hormonal pathways. However, studies of gene– Bittner, T. D. 1999. The evolutionary significance of melanism in the
hormone–behaviour interactions outside of humans and common garter snake, Thamnophis sirtalis. Ph.D. thesis, Northern domesticated animals remain rare. A notable exception Brodie, E. D., III 1989. Genetic correlations between morphology
involves the evolutionary endocrinology of crickets, in and antipredator behaviour in natural populations of the garter which juvenile hormone esterase, wing polymorphism snake Thamnophis ordinoides. Nature, 342, 542–543.
and migratory traits appear to be functionally interrelated Brodie, E. D., III 1993. Consistency of individual differences in
(Fairbairn & Roff 1999; Zera 1999; Zera & Huang 1999).
antipredator behaviour and colour pattern in the garter snake The widespread occurrence of geographical variation in Thamnophis ordinoides. Animal Behaviour, 45, 851–861.
behaviour (Foster & Endler 1999) as well as behavioural Brodie, E. D., III & Garland, T. Jr 1993. Quantitative genetics of
differences among species and higher taxonomic levels snake populations. In: Snakes: Ecology and Evolution (Ed. by R. A.
suggests that studies of gene–hormone–behaviour inter- Seigel & J. T. Collins), pp. 315–362. New York: McGraw-Hill.
actions might also benefit from geographical and Brodie, E. D., III & Russell, N. H. 1999. The consistency of individual
differences in behaviour: temperature effects on antipredator
behaviour in garter snakes. Animal Behaviour, 57, 445–451.
Burghardt, G. M. 1983. Aggregation and species discrimination in
Acknowledgments
newborn snakes. Zeitschrift fu¨r Tierpsychologie, 61, 89–101.
Burghardt, G. M. & Schwartz, J. M. 1999. Geographic variations on
This material is based on work supported by the National methodological themes in comparative ethology: a natricine Science Foundation under Grant No. 9409464. I thank T.
snake perspective. In: Geographic Variation in Behavior: Perspectives Bittner, J. Cline, R. Dickenson, C. Hubbard, C. Pedota, A.
on Evolutionary Mechanisms (Ed. by S. A. Foster & J. A. Endler), pp.
Queral-Regil and J. Turmo and for help collecting gravid 69–94. New York: Oxford University Press.
females, maintaining captives and conducting hormone Burghardt, G. M., Layne, D. G. & Konigsberg, L. 2000. The
assays. Snakes were collected under a permit from the genetics of dietary experience in a restricted natural population.
Ohio Division of Parks and Recreation; procedures involv- Psychological Science, 11, 69–72.
ing live animals were approved by the Northern Illinois Bush, M. & Smeller, J. 1978. Blood collection and injection tech-
niques in snakes. Veterinary. Medicine Small Animal Clinician, 73,
University Institutional Animal Care and Use Committee (Protocol No. 092). Comments from three anonymous Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences.
referees significantly improved the manuscript.
2nd edn. Hillsdale, New Jersey: L. Ehrlbaum.
Compaan, J. C., van Wattum, G., De Ruiter, A. J., van
Oortmerssen, G. A., Koolhaas, J. M. & Bohus, B. 1993. Genetic
References
differences in female house mice in aggressive response to
sex steroid hormone treatment. Physiology & Behavior, 54, 899–
Alexandre, C. & Balthazart, J. 1987. Inhibition of testosterone
metabolism in the brain and cloacal gland of the quail by specific Crews, D. 1985. Effects of early sex steroid hormone treatment on
inhibitors and antihormones. Journal of Endocrinology, 112, 189–
courtship behavior and sexual attractivity in the red-sided garter snake, Thamnophis sirtalis parietalis. Physiology & Behavior, 35,
An, P., Rice, T., Gagnon, J., Hong, Y. L., Leon, A. S., Skinner, J. S.,
Wilmore, J. H., Bouchard, C. & Rao, D. C. 2000. A genetic study
Crews, D. 1987. Diversity and evolution of behavioral controlling
of sex hormone-binding globulin measured before and after a mechanisms. In: Psychobiology of Reproductive Behavior: an 20-week endurance exercise training program: the HERITAGE Evolutionary Perspective (Ed. by D. Crews), pp. 88–119. Englewood family study. Metabolism: Clinical and Experimental, 49, 1014–
Crews, D. & Moore, M. C. 1986. Evolution of mechanisms
Arnold, S. J. 1988. Quantitative genetics and selection in natural
controlling mating behavior. Science, 231, 121–125.
populations: microevolution of vertebral numbers in the garter Crews, D., Diamond, M. A., Whittier, J. & Mason, R. 1985. Small
snake Thamnophis elegans. In: Proceedings of the Second male body size in garter snakes depends on testes. American International Conference on Quantitative Genetics (Ed. by B. S. Weir, Journal of Physiology, 249, R62–R66.
E. J. Eisen, M. M. Goodman & G. Namkoong), pp. 619–636.
von Ende, C. 1993. Repeated-measures analysis: growth and other
time-dependent measures. In: Design and Analysis of Ecological Arnold, S. J. & Bennett, A. F. 1984. Behavioral variation in
Experiments (Ed. by S. M. Scheiner & J. Gurevitch), pp. 113–137.
natural populations. III. Antipredator displays in the garter snake Thamnophis radix. Animal Behaviour, 32, 1108–1118.
Fairbairn, D. J. & Roff, D. A. 1999. The endocrine genetics of wing
Arnold, S. J. & Phillips, P. C. 1999. Hierarchical comparison of
polymorphism in Gryllus. A response to Zera. Evolution, 53, 977–
genetic variance-covariance matrices. II. Coastal-inland diver- gence in the garter snake Thamnorphis elegans. Evolution, 53,
Falconer, D. S. & MacKay, T. F. C. 1996. Introduction to Quantitative
Genetics. 4th edn. Essex: Addison-Wesley-Longman.
Barry, R. E., Weatherhead, P. J. & Phillipp, D. P. 1992. Multiple
Ford, N. B. & Burghardt, G. M. 1993. Perceptual mechanisms and
paternity in a wild population of northern water snakes, Nerodia the behavioral ecology of snakes. In: Snakes: Ecology and Evolution sipedon. Behavioral Ecology and Sociobiology, 30, 193–199.
(Ed. by R. A. Seigel & J. T. Collins), pp. 117–164. New York: Bates, R. O., Buchanan, D. S., Johnson, R. K., Wetteman, R. P.,
Fent, R. W. & Hutchens, L. K. 1986. Genetic parameter estimates
Formanowicz, D. R. Jr, Brodie, E. D. Jr & Bradley, P. J. 1990.
for reproductive traits of male and female littermate swine. Journal Behavioral compensation for tail loss in the ground skink, Scincella of Animal Science, 63, 377–385.
lateralis. Animal Behaviour, 40, 782–784.
ANIMAL BEHAVIOUR, 63, 0
Foster, S. A. & Endler, J. A. 1999. Geographic Variation in Behavior:
nuclear retention of 5-dihydrotestosterone in rat ventral prostate.
Endocrinology, 94, 1205–1209.
McCracken, G. F., Burghardt, G. M. & Houts, S. E. 1999.
Garland, T. Jr 1988. Genetic basis of activity metabolism. I. Inherit-
Microsatellite markers and multiple paternity in the garter snake ance of speed, stamina, and antipredator displays in the garter Thamnophis sirtalis. Molecular Ecology, 8, 1475–1480.
snake Thamnophis sirtalis. Evolution, 42, 335–350.
Mason, R. T. & Crews, D. 1985. Female mimicry in garter snakes.
Garner, T. W. J. 1998. A molecular investigation of population
Nature, 316, 59–60.
structure and paternity in the common garter snake, Thamnophis Meikle, A. W., Bishop, D. T., Stringham, J. D. & West, D. W. 1986.
sirtalis. M.S. thesis, University of Victoria.
Quantitative genetic and nongenetic factors that determine Gibson, R. A. & Falls, J. B. 1975. Evidence for multiple insemination
plasma sex steroid variation in normal male twins. Metabolism: in the common garter snake, Thamnophis sirtalis. Canadian Journal Clinical and Experimental, 35, 1090–1095.
of Zoology, 53, 1362–1368.
Meikle, A. W., Stringham, J. D., Woodward, M. G. & Nelson, J. C.
Gillingham, J. C. 1987. Social behavior. In: Snakes: Ecology and
1988a. Hereditary and environmental influences on the variation Evolutionary Biology (Ed. by R. A. Seigel, J. T. Collins & S. S. Novak), of thyroid hormones in normal male twins. Journal of Clinical Endocrinology and Metabolism, 66, 588–592.
Glowa, J. R., Sternberg, E. M. & Gold, P. W. 1992. Differential
Meikle, A. W., Stingham, J. D., Woodward, M. G. & Bishop,
behavioral response in LEW/N and F344/N rats: effects of cortico- D. T. 1988b. Heritability of variation in plasma cortisol levels.
tropin releasing hormone. Progress in Neuropsychopharmacology Metabolism: Clinical and Experimental, 37, 514–517.
and Biological Psychiatry, 16, 549–560.
Mendlewicz, J., Linkowski, P., Kerkhofs, M., Leproult, R.,
Gregory, P. T., Macartney, J. M. & Larsen, K. W. 1987. Spatial
Copinschi, G. & Van Cauter, E. 1999. Genetic control of 24-hour
patterns and movements. In: Snakes: Ecology and Evolutionary growth hormone secretion in man: a twin study. Journal of Clinical Biology (Ed. by R. A. Seigel, J. T. Collins & S. S. Novak), pp.
Endocrinology and Metabolism, 84, 856–862.
Moore, M. C. 1991. Application of organization-activation theory to
Gupta, P. & Brush, F. R. 1998. Differential behavioral and endo-
alternative male reproductive strategies: a review. Hormones and crinological effects of corticotropin-releasing hormon (CRH) in the Behavior, 25, 154–179.
Syracuse high- and low-avoidance rats. Hormones and Behavior, Moore, M. & Lindzey, J. 1992. The physiological basis of sexual
34, 262–267.
behavior in male reptiles. In: Biology of the Reptilia. Vol. 18:,Physiology E: Hormones, Brain, and Behavior (Ed. by C. Gans & D.
Hadley, M. E. 1984. Endocrinology. Englewood Cliffs, New Jersey:
Crews), pp. 70–113. Chicago: University of Chicago Press.
Nelson, R. J. 1995. An Introduction to Behavioral Endocrinology.
Halpern, M., Morrell, J. I. & Pfaff, D. W. 1982. Cellular
(3H)estradiol and (3H)testosterone localization in the brains Neumann, F., Graf, K.-J., Hasan, S. H., Schenck, B. & Steinbeck, H.
of garter snakes: an autoradiographic study. General and 1977. Central actions of antiandrogens. In: Androgens and Comparative Endocrinology, 46, 211–224.
Antiandrogens (Ed. by L. Martini & M. Motta), pp. 163–177. New Herzog, H. A. Jr & Burghardt, G. M. 1986. Development of
antipredator responses in snakes. I. Defensive and open-field O’Brien, R. G. & Kaiser, M. K. 1985. MANOVA method for
behaviors in newborns and adults of three species of garter analyzing repeated measures designs: an extensive primer.
snakes (Thamnophis melanogaster, T. sirtalis, T. butleri). Journal of Psychological Bulletin, 97, 316–333.
Comparative Psychology, 100, 372–379.
Otremski, I., Karasik, D. & Lishits, G. 2000. Genetic variation and
Herzog, H. A. Jr & Burghardt, G. M. 1988. Development of
covariation of parathyroid hormone levels and bond density in the antipredator responses in snakes. III. Stability of individual and human population. Calcified Tissue International, 66, 168–175.
litter differences over the first years of life. Ethology, 77, 250–258.
Potvin, C., Lechowicz, M. J. & Tardif, S. 1990. The statistical
Jaquish, C. E., Blangero, J., Haffner, S. M., Stern, M. P. &
analysis of ecophysiological response curves obtained from CacCluer, J. W. 0000. Quantitative genetics of serum sex
experiments involving repeated measures. Ecology, 71, 1389–
hormone-binding globulin levels in participants in the San Antonio Family Heart Study. Metabolism: Clinical and Experimental, Prosser, M. R. 1999. Sexual selection in northern water snakes,
46, 988–991.
Nerodia sipedon sipedon: examination of mating system and Ketterson, E. D. & Nolan, V. Jr 1992. Hormones and life histories:
correlates of male reproductive success using microsatellite DNA an integrative approach. American Naturalist, 140, S33–S62.
markers. Ph.D. thesis, McMaster Unviersity.
King, R. B. & Lawson, R. 1995. Color-pattern variation in Lake Erie
Reinert, H. K. 1993. Habitat selection in snakes. In: Snakes: Ecology
water snakes: the role of gene flow. Evolution, 49, 885–896.
and Evolution (Ed. by R. A. Seigel & J. T. Collins), pp. 201–240.
King, R. B. & Turmo, J. 1997. The effects of ecdysis on feeding
frequency and behavior of the common garter snake (Thamnophis Rossman, D. A., Ford, N. B. & Seigel, R. A. 1996. The Garter Snakes:
sirtalis). Journal of Herpetology, 31, 310–312.
Evolution and Ecology. Norman: University of Oklahoma Press.
King, R. B., Cline, J. H. & Hubbard, C. J. 2000. Age and litter effects
Schwartz, J. M., McCracken, G. F. & Burghardt, G. M. 1989.
on testosterone levels in young water snakes. Copeia, 2000,
Multiple paternity in wild populations of the garter snake, Thamnophis sirtalis. Behavioral Ecology and Sociobiology, 25, 269–
King, R. B., Milstead, W. B., Gibbs, H. L., Prosser, M. R.,
Burghardt, G. M. & McCracken, G. F. 2001. Application of
Scott, T. R. & Washburn, K. W. 1988. Genetic variation of plasma
microsatellite DNA markers to discriminate between maternal and growth hormone and its genetic association with growth traits in genetic effects on scalation and behavior in muliply-sired garter young chickens. Poultry Science, 67, 1781–1782.
snake litters. Canadian Journal of Zoology, 79, 121–128.
Scudder, R. M. & Burghardt, G. M. 1983. A comparative study of
Lawson, R. & King, R. B. 1996. Gene flow and melanism in Lake Erie
defensive behavior of three sympatric species of water snake garter snake populations. Biological Journal of Linnean Society, 59,
(Nerodia). Zeitschrift fu¨r Tierpsychologie, 63, 17–26.
Shine, R. 1991. Intersexual dietary divergence and the evolution
Liao, S., Howell, D. K. & Chang, T.-M. 1974. Action of a non-
of sexual dimorphism in snakes. American Naturalist, 138, 103–
steroidal antiandrogen, flutamide, on the receptor binding and KING: GARTER SNAKE BEHAVIOUR
Shine, R. & Crews, D. 1988. Why male garter snakes have small
Whittier, J. M. & Tokarz, R. R. 1992. Physiological regulation of
heads: the evolution and endocrine control of sexual dimorphism.
sexual behavior in female reptiles. In: Biology of the Reptilia. Vol. Evolution, 42, 1105–1110.
18: Physiology E: Hormones, Brain, and Behavior (Ed. by C. Gans & Shine, R., Harloow, P., Lemaster, M. P., Moore, I. T. & Mason, R.
D. Crews), pp. 24–69. Chicago: University of Chicago Press.
T. 2000. The transvestite serpent: why do male garter snakes court
Zarazaga, L. A., Malpaux, B., Guillaume, D., Bodin, L. &
(some) other males? Animal Behaviour, 59, 349–359.
Chemineau, P. 1998. Genetic variability in melatonin concen-
Shine, R., Olsson, M. M., Lemaster, M. P., Moore, I. T. & Mason,
trations in ewes originates in its synthesis, not in its catabolism.
R. T. 2000. Effects of sex, body size, temperature, and location on
American Journal of Physiology, 274, E1086–E1090.
the antipredator tactics of free-ranging gartersnakes (Thamnophis Zera, A. J. 1999. The endocrine genetics of wing polymorphism in
sirtalis, Colubridae). Behavioral Ecology, 11, 239–245.
Gryllus: critique of recent studies and state of the art. Evolution, 53,
Stevens, J. 1992. Applied Multivariate Statistics for the Social Sciences.
2nd edn. Hillsdale, New Jersey: L. Ehrlbaum.
Weil, M. R. 1985. Comparison of plasma and testicular testosterone
Zera, A. J. & Huang, Y. 1999. Evolutionary endocrinology of
levels during the active season in the common garter snake, juvenile hormone esterase: functional relationship with wing Thamnophis sirtalis (L.). Comparative Biochemistry and Physiology A, polymorphism in the cricket, Gryllus firmus. Evolution, 53, 837–
81, 585–587.

Source: http://www.cnah.org/pdf_files/12.pdf