François Mougeot, Centre for Ecology and Hydrology, Hill of Brathens, Banchory AB31 4BW, Scotland, UK. Tel.: 44 1330826333. Fax: 44 1330823303. E-mail: email@example.com
1According to the ‘territorial behaviour’ hypothesis, red grouse population cycles are caused by delayed density-dependent changes in male aggressiveness influencing recruitment. These lagged changes in aggressiveness might be caused by changes in the kin structuring of male populations and differential aggressive behaviour between kin and non-kin (‘kinship’ hypothesis).
2A population-level manipulation of male aggressiveness in autumn affected the kin structure of male populations, their subsequent aggressiveness, and recruitment of both sexes. On two moors, we implanted the old territorial cocks in autumn with testosterone on an experimental area (T-areas) and with sham implants on a control area (C-areas).
3Increased aggressiveness in autumn t reduced recruitment in autumns t and t+ 1, and breeding density of both sexes in springs t+ 1 and t+ 2, confirming previous studies elsewhere. A new observation was that cocks on T-areas had bigger combs (an ornament whose size is testosterone-dependent) than those on C-areas for at least 1·5 years after treatment, evidence that they remained more aggressive.
4Increased aggressiveness reduced not only subsequent density but also kin structuring among territorial cocks. This is consistent with the ‘kinship’ hypothesis that changes in the kin structure of male populations mediate year-to-year changes in male aggressiveness.
5Increased aggressiveness did not increase intensity of infection by the dominant intestinal nematode Trichostrongylus tenuis, which might have affected recruitment through reduced breeding success. Moreover, breeding success after treatment was no lower on the T- than on the C-areas.
6The results show for the first time that increased aggressiveness affects both kin structure and subsequent recruitment, supporting a key assumption of the kinship hypothesis for red grouse population cycles.
Red grouse populations commonly show cyclic dynamics with periods of 4–10 years (Moss & Watson 2001). These cycles might be caused by a nematode parasite (Trichostrongylus tenuis Eberth.) influencing breeding success and number of young available for recruitment (Hudson 1986; Hudson, Newborn & Dobson 1992; Hudson et al. 1998, 2002) or by changes in territorial behaviour influencing recruitment. According to the ‘territorial behaviour’ hypothesis, grouse cycles are caused by changes in the aggressiveness of territorial cocks (Watson 1985; Mountford et al. 1990; Moss, Parr & Lambin 1994; Watson et al. 1994; Moss, Watson & Parr 1996; MacColl et al. 2000; Moss & Watson 2001). Aggressiveness varies during a cycle and lags behind density (Watson et al. 1994; Moss et al. 1996). Recent experiments showed that autumn aggressiveness regulates red grouse populations: testosterone implants increased male aggressiveness for a short period in autumn t, reduced the breeding density of both sexes in spring t+ 1, and changed population growth rates from positive to negative (Mougeot et al. 2003a). Moreover, on two moors in northern England, implants reduced recruitment and density for more than a year after their contents were exhausted (Mougeot et al. 2003b). Based on this new finding that aggressiveness is influenced by previous territorial contests, deductive modelling highlighted how changes in aggressiveness from autumn to autumn, influenced by density, can cause cyclic population dynamics (Matthiopoulos et al. 2003).
Although interactions between aggressiveness and density can readily cause unstable population dynamics (Matthiopoulos et al. 2003), the behaviour mediating this interaction remains unclear. The ‘kinship’ hypothesis is based on kin facilitation and changes in the kin structure of male populations (Charnov & Finnerty 1980; Watson et al. 1994; Hendry et al. 1997; Matthiopoulos, Moss & Lambin 1998, 2000, 2002). Territorial cocks are less aggressive towards kin than to non-kin, and favour recruitment of kin close to their territories, so forming territory clusters of related cocks (Watson et al. 1994; Piertney et al. 1999). As population density increases space becomes limiting, tolerance among relatives decreases and curtails recruitment; without recruitment sufficient to compensate for natural mortality, kin clusters decay rapidly and reduced recruitment drives density down.
We report on the long-term effects of a population-level manipulation of cock aggressiveness on two grouse moors in north-east Scotland. We tested whether previous demographic results from northern England could be replicated in north-east Scotland. Moreover, we investigated in more detail possible mechanisms for the lagged effect of aggressiveness on recruitment. First, we checked whether autumn testosterone treatment had long-term effects on the size of cocks’ combs, which is testosterone-dependent and relates to aggressiveness (Moss et al. 1979). We expected cocks to be more aggressive a year after the testosterone implants were exhausted, and therefore to have bigger combs on the testosterone-treated than on the control areas. Secondly, we tested a prediction of the ‘kinship’ hypothesis, that increasing aggressiveness of old territorial cocks reduces kin structure in subsequent breeding populations. Thirdly, we tested whether the testosterone treatment increased intensity of infection by caecal threadworms (T. tenuis), thereby reducing subsequent breeding success (Hudson et al. 1992, 2002) and the number of young available for recruitment.
We conducted this experiment on two moors (replicates) in north-east Scotland (Aberdeenshire): Glen Dye (56°58′ N−2°34′ W) and Edinglassie (57°12′ N−3°07′ W) moors, hereafter moors 1 and 2, respectively. Bag records indicated that grouse populations on both moors were cyclic, increasing and close to the peak of the cycle when the experiment was done (authors, unpublished data). On each moor, two 1-km2 areas separated by 0·5 km were randomly assigned a treatment (testosterone or control). In September–November 2001, we caught and ringed 179 cocks within these areas, by lamping and netting them at night (Hudson 1986). Grouse were aged from plumage and morphology (Cramp & Simmons 1980), classified as ‘reared young’ in their summer of hatching, as ‘young adults’ until their second summer (young recruited the previous autumn) and as ‘old adults’ thereafter. We implanted 77 old cocks (16–23 cocks per area) but no young cock. Two silastic tube implants (each 20 mm long, 0·62 mm inner and 0·95 mm outer diameter) were inserted under the breast skin of each cock. Implants were either empty (control or C-areas) or filled with testosterone proprionate (testosterone or T-areas). Testosterone implants increased testosterone levels for up to 3 months (Mougeot et al. 2003a,b). From the number of cocks counted in autumn 2001 (see below), we estimated that 84% and 71% of cocks were caught and 70% and 78% of old cocks (31% and 33% of all cocks) were implanted on the T- and C-areas, respectively. We held Home Office licences for conducting all described procedures.
We counted grouse with trained dogs (Jenkins, Watson & Miller 1963). All areas were of 1-km2 so total counts reflected density. We counted grouse in spring (April), summer (late July) and autumn (late September–early October) from spring 2001 to spring 2003. Summer counts gave adult grouse density, breeding success (number of young per hen) and breeding production (the total number of young fledged). The number of cocks available for recruitment (before the experiment) was the number of adult cocks plus half the number of young counted in July (sex ratio at fledging averages 1 : 1; R. Moss, unpublished data). September counts gave the numbers of cocks competing for territories in autumn. Spring counts gave the numbers of cocks that successfully established a territory. Hunters shot grouse on both moors in August–September 2001 and 2002. Shooting is a normal cause of mortality and does not prevent population cycles (Hudson & Dobson 2001; Haydon et al. 2002). Autumn counts were conducted after harvesting, so changes in numbers between autumn and spring counts were unlikely to be affected by harvesting.
comb size measurements
Red grouse cocks have conspicuous supra-orbital red combs, their main sexual ornament, which function in intrasexual competition and mate choice (Moss et al. 1979; Mougeot & Redpath 2004; Mougeot et al. 2004). Comb size is testosterone-dependent and relates to aggressiveness: cocks with bigger combs are more aggressive, more likely to obtain a territory during autumn territorial contests, and hold larger territories (Moss et al. 1979; MacColl et al. 2000). We measured combs of cocks caught (autumn 2001, spring 2002, autumn 2002, spring 2003) and shot on study areas (autumn 2002). From the length and width of flattened combs measured with a ruler (to the nearest 1 mm), we calculated comb size (comb width × length, in mm2) as a proxy of aggressiveness (see Moss et al. 1979). In all, we measured combs of 457 cocks (5–77 cocks per area per season).
kin structuring of male populations
In autumn 2001, spring 2002 and spring 2003 (moor 2 only) we caught cocks, recorded their location to the nearest 7–10 m with a GPS (eTrex personal navigator, Garmin Olathe, Kansas, USA) and took body feathers as a source of DNA for genotyping. We used capture locations to map presumptive cock territories by tessellation, and resolved patterns of kin structure by identifying related individuals from genotype data. DNA was extracted from feathers using standard procedures, and each cock was genotyped at 13–17 highly variable microsatellite loci (Piertney & Dallas 1997; Piertney et al. 1999). Related individuals were identified using several approaches. Non-overlapping distributions of relatedness for non-relatives and first-order relatives allow correct identification of relationship from relatedness with minimal type I or type II errors (Piertney et al. 1999). First-order relatives were identified as father–son or brother–brother from their age. Father–son relationships were identified using cervus (Marshall et al. 1998), which by inference also identified some full-brother relationships. Brothers whose father did not have a territory on the study area were identified from relatedness estimates (Queller & Goodnight 1989). Uncle–nephew and grandson–grandfather dyads were inferred from common relatives.
We determined kin structure among territorial cocks on each area before (autumn 2001) and after treatment (springs 2002 and 2003). The kinship hypothesis is based on kin-facilitation by clusters of related cocks (Matthiopoulos et al. 1998, 2000; MacColl et al. 2000). Relevant measures of kin structure are thus based on relatedness among neighbouring cocks, so we used: (1) number of cocks with a neighbouring first-order relative (son or brother with adjacent territory); and (2) number of old cocks with a neighbouring son. To account for differences in density between areas, we analysed proportions of cocks rather than actual numbers. In autumn 2001, we determined levels of kin structure using only old cocks (old cocks were already territorial, while young cocks are trying to establish a territory), as an index of kin structure in the 2001 breeding season.
We measured intensity of T. tenuis infection on each area in autumn 2001 (prior to experiment), autumn 2002 and spring 2003 using direct worm counts (from guts of grouse shot in autumn on the study areas) or faecal egg counts (from faecal samples collected from grouse caught in autumn and spring). Details on the methods are given in Wilson (1983), Moss et al. (1990) and Seivwright et al. (2004). Faecal egg counts provide reliable estimates of parasite intensity in spring (Moss et al. 1990) and autumn (Seivwright et al. 2004). All analyses were based on worm intensity, the number of worms per host being calculated from faecal egg concentration where necessary. In all, we estimated worm intensity for 205 grouse (6–29 per area per season).
We used generalized linear models (genmod procedure, SAS 2001). Dependent variables were fitted using the normal error distribution for comb size, Poisson distribution for counts of cocks, hens or young and negative binomial distribution for worm intensity.
We tested whether changes over time in grouse density differed between treatment areas 1 year after treatment (in autumn 2001) by including moor (1 vs. 2), treatment area (C- vs. T-area), time of count (July 2002, September 2002 and April 2003) and all interactions between variables as fixed effects. We particularly tested whether changes in numbers over time differed between treatment areas (time × treatment interaction).
To test for differences between treatment areas in recruitment success of reared young cocks, we used the number of young cocks present in spring t as the dependent variable, with the number of reared young cocks in summer t-1 as an offset (genmod procedure, SAS 2001). Models included moor, treatment area and their interaction as fixed effects.
We first tested for differences between treatment areas in comb size and worm burden by season. Models included moor, treatment and their interaction as fixed effects. Before analyses, we standardized comb size and worm data for age and sampling date. We also checked whether changes in comb size from autumn 2001 (baseline, prior to treatment) to autumn 2002 differed between treatment areas (testing for a time × treatment interaction).
To test for differences in kin structure between treatment areas we used each kin structure measure (number of cocks or old cocks with a neighbouring relative or son, respectively) as dependent variables with the number of cocks genotyped as an offset, to account for differences in cock numbers between areas.
delayed effects of treatment on population change and recruitment
We tested for differences in population change between treatment areas in the year after implanting with testosterone by considering the number of cocks counted in July 2002 (young and old cocks available for recruitment), in September 2002 (cocks competing for territories in autumn, after harvesting) and in April 2003 (cocks that successfully established a territory). Comparing changes in numbers, rather than numbers in each season separately, controlled for any differences in initial numbers among areas. Over this period, cock density decreased significantly more on the T-areas than on the C-areas (Fig. 1a; Table 1: significant time × treatment interaction) on both moors (non-significant time × treatment × moor interaction). The significant treatment × moor interaction (Table 1) indicated that differences in cock numbers between T- and C-areas were greater on moor 2 than on moor 1 (Fig. 1a). When the effect of time was investigated further, cock density declined significantly more on the T- than on the C-areas between July 2002 and April 2003 (time × treatment interaction: F = 9·87; P < 0·01) and tended to decline more between September 2002 and April 2003 (F = 3·29; P = 0·07), but not between July and September 2002 (F = 2·36; P = 0·13).
Table 1. Delayed effects of testosterone implants on changes in cock and hen density between July 2002 and April 2003 (genmod procedure, type 1 analyses, SAS 2001)
We further tested whether the proportion of reared young cocks that recruited into the following spring territorial populations differed between treatment areas. In the year of implanting (summer 2001–spring 2002), the proportion of reared young cocks that recruited was significantly lower on the T- than on the C-areas on both moors and did not differ significantly between moors (Fig. 2; genmod: moor: = 0·18; P = 0·67; treatment; = 5·12; P < 0·05; moor × treatment: = 0·39; P = 0·53). In the following year (summer 2002–spring 2003), recruitment success of reared young cocks was again significantly lower on the T- than on the C-areas ( = 8·18; P < 0·01) on both moors (non-significant moor × treatment interaction: = 0·08; P = 0·77) and was also lower on moor 1 than on moor 2 (Fig. 2; = 5·09; P < 0·05).
Changes over time in hen density were similar overall to changes in cock density, both decreasing more on the T- than on the C-areas between July 2002 and April 2003 (Table 1). In spring 2003, 1·5 years after treatment, cock density was 30 and 57% lower, and hen density was 29 and 48% lower on the T- than on the C-areas on moors 1 and 2, respectively.
delayed effects of treatment on cock comb size
Cocks have bigger combs in spring than in autumn (Fig. 1c). Therefore we first analysed comb size by season. In spring, comb size differed between age groups (young cocks had smaller combs than old cocks; F1,183= 3·98; P < 0·05) and was a quadratic function of sampling date (date: F1,183 = 15·04; P < 0·001; date2: F1,183 = 14·62; P < 0·001). In autumn, comb size again differed between young and old (young cocks had smaller combs: F1,266 = 28·14; P < 0·001) and increased linearly with sampling date (date: F1,266 = 13·27; P < 0·001; date2: F1,265 = 0·21; P = 0·64). Before testing for differences between treatment areas, we standardized comb size data for age and sampling date, using residuals from the relationships with age and date in spring and autumn separately.
In September 2001, before treatment, standardized comb size did not differ significantly between treatment areas (GLM: F1,194 = 0·33; P = 0·57) on either moor (treatment × moor interaction: F1,194 = 0·02; P = 0·88). One month after implanting (October 2001), cocks on T-areas had significantly bigger combs than those on C-areas (Fig. 1c; treatment: F1,37 = 50·06; P < 0·001) on both moors (treatment × moor interaction: F1,37 = 0·27; P = 0·61). Cocks on T-areas still had bigger combs than those on the C-areas in spring 2002 (treatment effect: F1,31 = 5·01; P < 0·05), autumn 2002 (F1,26= 8·58; P < 0·01) and spring 2003 (F1,148 = 18·01; P < 0·001) on both moors (all moor × treatment interactions non-significant; Fig. 1c).
In addition, to control for any initial differences in comb size among areas, we tested whether changes over time in standardized comb size, between autumn 2001 (before implanting) and autumn 2002 (1 year after implanting), differed between treatment areas. A GLM model for standardized comb size included moor, time (autumn 2001 vs. autumn 2002), treatment and all their interactions as fixed effects. Between autumn 2001 and autumn 2002, comb size increased significantly more on the T- than on the C-areas (time × treatment interaction: F1,220 = 6·53; P < 0·05).
effect of treatment on the kin structure of male populations
Before treatment (autumn 2001), the proportion of cocks with a neighbouring relative did not differ between T- and C-areas, but was significantly lower on the T- than on the C-areas after treatment in spring 2002 (Table 2; Figs 1b and 3). Similar differences occurred when considering the proportion of cocks with a neighbouring son as a measure of kin structure (Table 2). Proportions were as follows: moor 1: 56% (n = 18 males) vs. 8% (n = 12); moor 2: 33% (n = 24) vs. 7% (n = 15) on the C- and T-areas, respectively. We further tested whether changes in kin structure between autumn 2001 and spring 2002 differed between treatment areas by including moor, treatment, time and all the interactions between these variables as fixed effects, and by testing for a time × treatment interaction. Changes over time in the proportion of cocks with a neighbouring relative decreased more on the T- than on the C-areas between 2001 and 2002 (time × treatment interaction: F1,1 = 5·57; P < 0·05; Table 2).
Table 2. Differences in kin structuring between treatment and control areas before (autumn 2001) and after (spring 2002 and 2003) the manipulation of aggressiveness (genmod procedure, SAS 2001)
On moor 2, we tested whether differences in kin structure between treatment areas changed between spring 2002 and spring 2003 (using GLMs with year, treatment and the year × treatment interaction as fixed effects). We found a significant treatment effect for both measures of kin structure (proportion of cocks with a neighbouring relative: F1,2 = 12·14; P < 0·001; proportion of cocks with a neighbouring son: F1,2 = 5·05; P < 0·05), but no significant year effect (both F-values < 1·81; both P > 0·10) or year × treatment effects (both F-values < 0·26; both P > 0·10). Hence differences in kin structure between treatment areas were maintained for two consecutive springs after treatment (Fig. 1b).
Male populations comprised few patrilines (cocks descended from a common male ancestor), but the size of the dominant patriline differed between treatment areas. Male populations on control areas were dominated by one large patriline (in spring 2002 the control areas comprised two patrilines of 26 and four related cocks on moor 1, and three patrilines of 29, five and three related cocks on moor 2; Fig. 3). However, male populations on the testosterone areas were not dominated by one large patriline (testosterone areas comprised two patrilines of 11 and five related cocks on moor 1, and five patrilines of two related cocks each on moor 2; Fig. 3). In spring 2003, we found similar differences between treatment areas on moor 2 (four patrilines of 32, seven, two and two related cocks on the control area, and two patrilines of five and three related cocks on the testosterone area).
differences in t. tenuis parasite intensity between treatment areas
To account for seasonal variations in T. tenuis intensity (grouse had more worms in autumn than in spring; Fig. 1d), we analysed parasite counts by season. In autumn, worm intensities differed significantly between young and old grouse (young had fewer worms; genmod: = 34·26; P < 0·001) and increased linearly with sampling date (date: = 27·95; P < 0·001). We thus standardized for age and date prior to testing for differences in worm intensities between treatment areas. In autumn 2001, before treatment, standardized worm intensities tended to be lower on moor 2 than on moor 1 (Fig. 1d; genmod: = 3·58; P = 0·06) but did not differ between T- and C-areas ( = 0·01; P = 0·92) on either moor. In autumn 2002, 1 year after treatment, worm intensities did not differ between moors (Fig. 1d; genmod: = 0·06; P = 0·81) or between treatment areas ( = 0·15; P = 0·69) on either moor (moor × treatment interaction: = 0·48; P = 0·49).
In spring 2003, variation in T. tenuis intensity was not explained by sampling date (genmod: = 0·56; P = 0·55). Worm intensities did not differ significantly between moors ( = 1·49; P = 0·22) but did differ between treatment areas ( = 5·67; P < 0·05), grouse having fewer worms on the T- than on the C-areas (Fig. 1d) on both moors (moor × treatment interaction: = 1·47; P = 0·23).
differences in breeding success between treatment areas
In July 2002, after manipulation of aggressiveness, breeding success did not differ between treatment areas on moor 2 [mean number of young per hen (95% confidence interval) of 4·5 (3·6–5·6) and 5·5 (4·5–6·7) on T- and C-areas, respectively; genmod: χ2 = 1·86; P = 0·17], but was significantly higher on the T- than on the C- area on moor 1 [7·2 (6·1–8·5) and 3·3 (2·5–4·3) young per hen on the T- and C-areas, respectively; genmod: χ2 = 26·07; P < 0·001].
An experimental increase in autumn aggressiveness reduced subsequent density and kin structure in territorial cocks. It also increased aggressiveness and depressed recruitment in the following autumn, but did not increase parasite burdens or reduce breeding success. Below we discuss these findings in relation to the territorial behaviour and kinship hypotheses for red grouse population cycles.
delayed effects of aggressiveness on recruitment, density and comb size
A brief (2–3 months) experimental increase in autumn aggressiveness of old adult cocks, caused by testosterone implants, reduced recruitment and density (Fig. 1, Table 1) in both sexes not only in the current season (autumn t and spring t+ 1), but also in the next one (autumn t+ 1 and spring t+ 2). On the T-area of moor 1, none the less, density in spring t+ 2 was not much lower than in spring t. This was because, in July 2002, higher breeding success caused higher density. Had breeding success on the T-area been as on the C-area, the experimentally depressed recruitment on the T-area would very probably have caused density in spring t+ 2 to have been lower there than in spring t, as on moor 2.
The fall in cock numbers between July 2002 and April 2003 was proportionately greater on the T- than on the C-areas, and occurred mainly between autumn 2002 (late September–early October) and spring 2003, after grouse shooting. This suggests that grouse shooting made little contribution to the observed differences in population change. Our results in north-east Scotland are thus consistent with previous findings in northern England (Mougeot et al. 2003b). They also confirm that cock numbers determine hen numbers (Moss et al. 1996; Mougeot et al. 2003a,b), as reductions in cock density were associated with similar reductions in hen density. A new finding is that comb size (a proxy of aggressiveness) remained larger on the T-areas, for more than a year after the implants were exhausted. Cocks on the T-areas had bigger combs than those on the C-areas in autumns t and t+ 1 and in springs t+ 1 and t+ 2. In previous experiments, comb size was not measured, although it was inferred that aggressiveness was higher on the T- than on the C-areas (Mougeot et al. 2003b). Comb size is testosterone-dependent (Moss et al. 1979) and correlates positively with circulating testosterone levels in both autumn and spring (Mougeot et al. 2005). Hence the lagged effects of the testosterone treatment upon recruitment and density were very probably due to cocks continuing to be more aggressive long after the implants were exhausted. This finding amplifies previous evidence (Mougeot et al. 2003b) for a territorial memory in cock red grouse populations: autumn aggressiveness is influenced by previous territorial contests. This discovery is the basis of recent population models, which deduce that population cycles are likely if changes in aggressiveness from one autumn to the next show an abrupt transition from tolerant to intolerant behaviour with increasing density (Matthiopoulos et al. 2003).
aggressiveness and population kin structure
A key assumption of the kinship hypothesis is that, just before the peak of a cycle, space becomes limiting and cocks become intolerant (more aggressive), which causes a breakdown of kin clusters. Reduced kin structuring, in turn, contributes to a further increase in aggressiveness (cocks are more aggressive towards non-kin than towards kin) and reduces recruitment, driving populations into decline (Mountford et al. 1990; Lock 2003). By increasing the aggressiveness of old (previously territorial) cocks in autumn during territory establishment, the testosterone treatment prevented young cocks from recruiting and depressed subsequent breeding density (Mougeot et al. 2003a). We have now shown that the treatment reduced not only density, but also increased aggressiveness and reduced kin structuring in subsequent territorial cock populations.
Consistent with previous studies (Piertney et al. 1998; Piertney et al. 1999), we found remarkable levels of kin structuring in male grouse populations, populations on control areas being dominated by few extended families, and organized in territory clusters of related cocks. Populations on both moors were increasing and close to a cyclic peak when the experiment was conducted. Such high levels of kin structuring were expected in increasing, high-density populations (Mountford et al. 1990; Lock 2003). In contrast, kin structuring was reduced on T-areas. Both the proportion of males with a neighbouring relative and the proportion of old males with a neighbouring son were significantly lower on T-areas than on C-areas after treatment. On moor 2, we also found that differences in kin structure between treatment areas were maintained for two consecutive springs after the initial manipulation of aggressiveness, showing that the effects of aggressiveness on kin structure lasted for more than a year.
Recent work showed that kin structure changed throughout an 8-year grouse population cycle, with changes in kinship preceding changes in cock density (Lock 2003). At equivalent density, kin structuring was greater and combs smaller (lower aggressiveness) during the increase than during the decline phase of the cycle (Lock 2003). Our experimental results show for the first time that experimentally increased aggressiveness reduces subsequent levels of kin structure in territorial male populations. Moreover, the finding that cocks continued to be more aggressive a year after the initial treatment is consistent with the idea that changes in kin structure of territorial male populations mediate changes in aggressiveness between autumns: the cause of males on the T-areas continuing to be more aggressive could well have been the reduced kin structuring in these populations. However, other possible mechanisms cannot be ruled out, for changes in male aggressiveness could be effected by factors other than, or operating with, changes in kin structure. For example, the aggressiveness of individual cocks in autumn t could be influenced by their individual memories of the territorial contests in autumn t − 1. We therefore conclude only that increased aggressiveness caused reductions in kin structure and increases in aggressiveness, consistent with the hypothesis that changes in kin structure mediate changes in aggressiveness.
differences in parasite burdens and breeding success between treatment areas
Past research has refuted trophic interactions with food or predators as causal mechanisms of red grouse population cycles (Moss & Watson 2001), but not the possibility that parasites (caecal threadworms T. tenuis) can be sufficient to cause cycles. According to the parasite hypothesis, unstable, cyclic population dynamics are caused by parasite induced reductions in host fecundity in combination with a low level of parasite aggregation in the host population (Hudson et al. 1992, 1998, 2002). Parasites were shown by experiment to reduce breeding success of grouse and affect their population dynamics (Hudson et al. 1992, 1998). Moreover, parasites might interact with behaviour in two ways: parasites can limit cock aggressiveness (Fox & Hudson 2001), and high testosterone levels might impair immune function (Folstad & Karter 1992; Mougeot et al. 2004), thereby increasing susceptibility to parasites. In our experiment, however, the testosterone treatment did not increase mean parasite burdens. This might be because we implanted only a small proportion of available cocks (old adult cocks, which accounted for ∼30% of all cocks). In fact, we found lower worm burdens on the T- than on the C-areas 1·5 years after treatment. This could be explained by the lower grouse densities on the T-areas, for parasite burdens follow changes in host density (Hudson et al. 1992) and the testosterone treatment reduced density.
Breeding success was no lower on the T- than on the C-areas after treatment. In fact, breeding success was higher on the T- than on the C-area on moor 1, as in another experiment in which cocks were implanted with testosterone in spring (Moss et al. 1994). Perhaps cocks on the T-areas attracted better quality females (because they held larger territories or were more attractive), thereby achieving better breeding success (see Moss et al. 1994). Recruitment depends upon breeding production and on the proportion of reared young cocks that establish into the territorial population (Moss & Watson 1991, 2001). The ‘parasite’ and ‘territorial behaviour’ hypotheses are thus not mutually exclusive, but our data show that the population declines following experimentally increased aggressiveness were not associated with reduced breeding success or increased parasite burdens.
Hypotheses of how intrinsic behavioural processes drive cyclic dynamics in red grouse predict a delayed density-dependent relationship between density and aggressiveness. The kinship hypothesis suggests that changes in kin structure mediate changes in aggressiveness, which moderate recruitment. Previous studies showed that aggressiveness affects recruitment and regulates grouse density (Mougeot et al. 2003a), that aggressiveness shows delayed density dependence (Watson et al. 1994; Moss et al. 1996), that cocks are less aggressive towards kin than non-kin (Watson et al. 1994), that young cocks born into larger kin clusters have a better chance of being recruited into the territorial population (MacColl et al. 2000) and that kin structure changes during the course of a population cycle in accordance with expectations (Lock 2003). The present study further supports the hypothesis in showing experimentally that increased aggressiveness simultaneously affects kin structure and subsequent recruitment. Although several risky predictions of the kinship hypothesis have proved correct, other behavioural mechanisms could also account for the observations made so far. The crucial experiment is now to test whether alterations of kin structure cause changes in aggressiveness, recruitment and population density.
We are grateful to the landowners and keepers of Edinglassie and Glen Dye estates for allowing us to conduct the work on their moors. Particular thanks are due to: D. Calder and A. Dykes for their help with organizing the fieldwork; R. Cox, J. Renet, J. Irvine and L. Monet for their help with fieldwork; Freda Marshall for assistance with genotyping; D. Elston for his help with the statistical analyses; and B. E. Arroyo and X. Lambin for their comments on the manuscript. This work was funded by an NERC grant (NER/A/S/1999/00074).