Patterns and processes of dispersal behaviour in arvicoline rodents



    1. CNRS – UMR 7625, Laboratoire Ecologie-Evolution, Université Pierre et Marie Curie, Case 237, 7 Quai St Bernard, 75005 Paris, France
    2. CNRS/ENS UMS 3194, CEREEP – Ecotron IleDeFrance, École Normale Supérieure, 78 rue du Château, 77140 St-Pierre-lès-Nemours, France
    Search for more papers by this author

    1. Faculty of Applied Ecology and Agricultural Sciences, Hedmark University College, Campus Evenstad, Anne Evenstadsvei 80, NO-2480 Koppang, Norway
    2. Centre for Ecological and Evolutionary Synthesis, Department of Biology, University of Oslo, PO Box 1066, Blindern, Oslo NO-0316, Norway
    Search for more papers by this author
  • ROLF A. IMS,

    1. Institutt for Arctic and Marine Biologi, Universitetty of Tromsø, 9073 Tromsø, Norway
    Search for more papers by this author

    1. School of Biological Sciences, University of Aberdeen, AB24 2TZ Aberdeen, UK
    Search for more papers by this author

Jean-François Le Galliard, Fax: +33-(0)1-44273516; E-mail:


A good understanding of mammalian societies requires measuring patterns and comprehending processes of dispersal in each sex. We investigated dispersal behaviour in arvicoline rodents, a subfamily of mammals widespread in northern temperate environments and characterized by a multivoltine life cycle. In arvicoline rodents, variation in life history strategies occurs along a continuum from precocial to delayed maturation that reflects seasonal and ecological fluctuations. We compared dispersal across and within species focusing on the effects of external (condition-dependent) and internal (phenotype-dependent) factors. Our data revealed substantial, unexplained variation between species for dispersal distances and a strong variation within species for both dispersal distance and fraction. Some methodological aspects explained variation across studies, which cautions against comparisons that do not control for them. Overall, the species under consideration display frequent short-distance dispersal events and extremely flexible dispersal strategies, but they also have hitherto unexpected capacity to disperse long distances. Female arvicolines are predominantly philopatric relative to males, but we found no clear association between the mating system and the degree of sex bias in dispersal across species. Dispersal is a response to both various proximate and ultimate factors, including competition, inbreeding avoidance, mate searching and habitat quality. In particular, our review suggests that costs and benefits experienced during transience and settlement are prime determinants of condition dependence. Patterns of phenotype-dependent dispersal are idiosyncratic, except for a widespread association between an exploration/activity syndrome and natal dispersal. Consequences for population dynamics and genetic structures are discussed.


The diverse social systems of mammals result from differences in behavioural and demographic processes such as dispersal behaviour (Wolff & Sherman 2007). For instance, the spatial overlap of males and females, and the temporal dynamics and genetic structure of family groups are all importantly influenced by the dispersal behaviour of each sex (Arnaud et al. 2011; Holekamp et al. 2011; Schradin et al. 2011). Dispersal, in turn, depends on social organization, and previous analyses of mammalian dispersal suggested that territorial and mating systems in mammals are key factors that determine the sexual biases in dispersal (reviewed in Lawson Handley & Perrin 2007; Clutton-Brock & Lukas 2011). Hence, a good understanding of mammalian societies requires measuring patterns and comprehending processes of dispersal in each sex.

Here, we used a comparative approach to investigate dispersal behaviour in male and female arvicoline rodents. These rodents are small mammals from the subfamily Arvicolinae that includes approximately 150 species of voles, lemmings and muskrats (Wilson & Reeder 2005). This subfamily is especially widespread in northern temperate and boreal environments, where it has been the focus of dispersal studies for almost four decades (Lidicker 1985; Anderson 1989; Stenseth & Lidicker 1992; Andreassen et al. 2002). In general, arvicolines are relatively small-bodied (<100 g), live above ground where they feed on green vegetation, seeds and roots and have a short (less than annual) life cycle. Population dynamics patterns range from rather stable populations, through violent but regular cycles, to various forms of irregular dynamics (Ims et al. 2008). Some species such as muskrats and water voles adapted to aquatic life may have a larger body and a slower life cycle (longer than annual). Our comparative analysis differs from most of the previous studies of mammalian dispersal (reviewed in Lawson Handley & Perrin 2007; Clutton-Brock & Lukas 2011), in that we focus on a smaller taxonomic group of closely related mammals, with relatively similar morphologies, lifestyles and life cycles. Our analysis is also the first attempt, to our knowledge, to quantify patterns and processes of intraspecific variation in a group of closely related species. Intraspecific variation in dispersal has until now been neglected because most approaches of social systems tend to assume that dispersal is a fixed trait. Yet, several studies have pointed out that intraspecific variation can be important in vertebrates (reviewed in Clobert et al. 2004; Bowler & Benton 2005), and evidence of social and dispersal plasticity is widespread in small mammals (e.g. Madison 1990; Andreassen et al. 2002; Schradin et al. 2011).

Dispersal behaviour depends upon external factors (condition dependence) and internal factors (phenotype dependence, see Ims & Hjermann 2001; Bowler & Benton 2005; Clobert et al. 2009). Within this framework, Ims & Hjermann (2001) highlighted the importance of the life cycle and developmental mechanisms as determinants of dispersal. In arvicolines, natal dispersal (movement out of the natal area) can occur early in life just after weaning. But natal dispersal can also be delayed for members of seasonal cohorts that do not mature in their year of birth, and breeding dispersal (movement out of the breeding area) can also happen at any time during adult life. Two basic pathways of condition-dependent and phenotype-dependent dispersal have been recognized (see Fig. 1). First, condition dependence could act primarily in a direct and immediate manner because animals respond flexibly to changes in environmental conditions prior to and during dispersal. Individuals with different phenotypes or belonging to different age, sex classes or cohorts could respond differently to changes in environmental conditions because they may not value the costs and benefits of dispersal in the same way (Clobert et al. 2009). Second, environmental conditions may have lagged developmental effects on dispersal behaviour. In arvicolines, we expect in particular that environmental conditions experienced by lactating or gravid mothers translate into specific dispersal behaviours in offspring who make dispersal decision ahead of precocious reproduction (maternal effects, see Ims & Hjermann 2001). In particular, environmental conditions experienced early in life could induce delayed and sustained changes in some phenotypic and life history traits (e.g. body size or behavioural traits) that are strongly correlated with dispersal behaviour. Such dispersal syndromes, i.e. consistent suites of morphological, behavioural and life history traits associated with dispersal, have been found in some birds, fishes and reptiles (reviewed in Clobert et al. 2009; Cote et al. 2010), but their prevalence in small mammals has not been reviewed before now.

Figure 1.

 (A) The life cycle of an arvicoline rodent involves first a gestation and lactation period, followed by a fast growth and sexual maturation period and then by reproduction and growth cessation. In seasonal environments, maturation can occur as early as few days old for most species but can also be delayed until the next breeding season upon condition that individuals survive the harsh winter conditions (see also Box 1). (B) This life cycle implies that the natal and breeding dispersal behaviour can be influenced by both direct (plain arrows) and indirect (dashed arrows) effects of external conditions (condition dependence) and by dependence upon the phenotype (phenotype dependence). Current external conditions during emigration, transience and immigration can have direct effects on natal dispersal behaviour, which usually happens around sexual maturation (age 30–50 days old) and on breeding dispersal (adult life averages a few weeks in most voles and lemmings but can reach several years in muskrats). It is also possible that conditions experienced before conception, during gestation or during lactation have indirect, delayed effects on natal and breeding dispersal. For example, exposure of females to male androgens in utero may explain the relationship between natal dispersal and litter sex ratio in two vole species (Lambin 1994a). A relationship between the phenotype (e.g. morphology or life history dispersal syndromes) could also be induced early in life when external conditions have joint effects on dispersal and other phenotypic traits.

We recorded all published data on natal and breeding dispersal in arvicolines including studies of social dispersal (dispersal out of a natal or breeding group), population dispersal (dispersal out of a natal or breeding patch of habitat) and genetic dispersal (effective dispersal leading to genetic exchanges between two demes). These three types of dispersal all involved the three fundamental stages of departure, transience and settlement (Ims & Yoccoz 1997; Clobert et al. 2004, 2009). We gathered information on the methods and context of each study, quantitative information on dispersal fraction (emigration and immigration fractions) and distance (distance travelled between departure and settlement), as well as qualitative data on condition-dependent and phenotype-dependent dispersal. Based on this data set, we critically evaluated current knowledge of patterns and processes of dispersal behaviour during its three different stages, and we identified methodological and conceptual gaps. Our analysis is structured around three major issues. First, we summarize basic information about the ontogeny and behavioural mechanisms of dispersal, and we quantify variation in dispersal within and between species, including information for each sex and age class whenever available and potential correlations with the social system and body mass of the species. Second, we report on all external conditions that trigger dispersal to identify most relevant proximate factors at each dispersal stage and attempt to separate indirect effects (e.g. maternal effects) from direct effects of each factor (see Fig. 1 for definitions of both effects). We also search for common patterns in phenotype-dependent dispersal to identify the most relevant morphological, behavioural and life history traits associated with dispersal, as well as interactive effects of condition and phenotype on dispersal behaviour. Third, we discuss the potential consequences of reported patterns of dispersal for the population dynamics and genetic structure of arvicoline rodents.

Literature survey

We searched for the literature on dispersal in arvicolines using Web of Knowledge. We separated studies by paper, species and populations as some articles provided information on multiple species and/or study sites (n = 223). For each item, we recorded the species name, the study setting (laboratory, semi-natural or natural), the study design (observational or experimental), the method used to quantify dispersal, the spatial scale (maximum distance from edge to edge within the study area), landscape geometry (continuous, fragmented or mixed landscapes) and the geographic location (continent and country identity, see Appendix S1, Supporting Information for more details). Unavoidably, these studies encompass different definitions and estimates of dispersal (e.g. social or population dispersal). Clutton-Brock & Lukas (2011) discuss some of the difficulties associated with these differences. Here, we did not attempt to separate studies of social and population dispersal because the distinction is not necessarily clear cut in studies of arvicoline rodents.

In addition, quantitative data were extracted from each paper to estimate the emigration fraction (percentage of individuals from a given cohort leaving a patch or a natal environment within a given time duration, calculated as number of emigrants over the total number of individuals conditional on survival, n = 231) and immigration fraction (percentage of individuals entering a patch within a given time duration, calculated as number of immigrants over the total number of individuals conditional on survival, n = 24). In addition, we extracted data about the mean dispersal distance (n = 100), the shape of the dispersal kernels (n = 20, probability density function of the dispersal distance) and the daily movement capacity (estimated in metres per day, n = 47). Real-time population genetic approaches can measure instantaneous dispersal by means of parentage or immigration assignment techniques (Paetkau et al. 2004). Only five of our direct, quantitative estimates of dispersal were obtained from real-time population genetic approaches (Telfer et al. 2003b; Aars et al. 2006; Schweizer et al. 2007; Gauffre et al. 2009; Guivier et al. 2011).

We further gathered quantitative data on the standardized genetic differentiation (n = 63) and the isolation-by-distance (IBD) patterns (n = 13). We noted which molecular marker was used, and we extracted or calculated the standardized genetic differentiation inline image, from the main text, tables or raw data of each study (Hedrick 2005). We calculated for each population inline image according to the method and formula in the study of Heller & Siegismund (2009, see Table S1, Supporting information). We also noted whether a significant IBD was found. Genetic dissimilarity and IBD patterns depend importantly on the effective dispersal (i.e. movement of individuals to another group in which they successfully breed) but also on the mutation rate, the effective population size (Rousset 1997; Whitlock 2011) and the demographic and social structure (e.g. for arvicoline rodents, see Aars et al. 2006). Hence, these statistics measure only indirectly dispersal; in particular, the high and fluctuating population sizes and the extinction-colonization dynamics of arvicoline rodents greatly influence genetic drift and consequently genetic differentiation, irrespective of dispersal (Aars et al. 2006; Berthier et al. 2006; Gauffre et al. 2008). We therefore did not attempt to estimate dispersal from these indirect genetic approaches; rather, we used them to compare the genetic structure of arvicoline rodents and discuss how dispersal could shape observed patterns. For each quantitative estimate of dispersal and genetic structure, we recorded the mean, the age (adult or juvenile) and sex class, the spatial scale over which the estimate was calculated (in metres), the sample size, the number of demes and, when relevant, the time duration during which the estimate was computed (in days).

These quantitative data were supplemented with qualitative data (n = 467), summarizing effects of all external and internal factors of dispersal tested in each study, as well as information on all interactions between factors tested in each study (see the framework in Fig. 1). In most instances, effect sizes were not available, and we therefore decided to note only the sign of the relationship (nonsignificant, negative or positive) between a change in the factor and a change in the ‘dispersal metric’ (emigration, immigration, dispersal distance, movement capacity or genetic dissimilarity). For each main effect of an external factor, we noted whether direct and indirect pathways were involved. Effects on emigration and immigration may inform us on factors that affect departure and settlement behaviours, respectively. Effects on the dispersal distances may inform us on factors that affect the transience behaviour. However, most estimates of dispersal were conditional on survival during the transience and settlement stage, except for a small number of direct measurements based on telemetry techniques.

We also gathered information on the mean body mass and dominant social mating system of each species. Body masses were obtained from a database of the late quaternary mammals by Smith et al. (2003), except for eight species for which we checked the primary literature. Adult body mass was averaged across males and females and across locations. We further described the dominant social mating systems from spacing patterns in natural populations (Boonstra et al. 1987; Heske & Ostfeld 1990) and complemented this with information on the genetic mating system whenever available (n = 9 species). We identified three social mating systems: (i) polygyny, (ii) promiscuity and (iii) monogamy or facultative monogamy. Difficulties with such species typology are discussed in Appendix S1 (Supporting Information) and at the end of this review.

We compared patterns of dispersal behaviour across studies and species after having classified species by their mating system and body mass, and studies by their method, temporal and spatial scales and sample size. Analyses were conducted in two different ways. First, we present tallies of qualitative data to compare the strength of evidence for some qualitative patterns (e.g. male-biased dispersal). Second, we analysed quantitative estimates of dispersal with standard linear models. We first used the complete data set to compare emigration fraction and dispersal distance across all species controlling for species-specific and study-specific covariates. These models included a random species effect to control for the fact that multiple data were available for some species and to calculate variation within as well as between species. We implemented these models with the lme package in R 2.10.0 following guidelines in Pinheiro & Bates (2000) and tested fixed effects with the anova procedure. We then tested for intraspecific variation in emigration fraction and dispersal distance according to age and sex, using the same statistical procedures as above. For this purpose, we included only data where dispersal was quantified within a given age and/or sex class. Fixed effects of age and sex were tested with the anova procedure. We also compared measures of standardized genetic differentiation (inline image) across species and studies. We lastly tested for the effects of study-specific factors (type of molecular markers, landscape geometry, number of demes and spatial scale) and of species-specific factor (body mass and social mating system) using the same statistical approach as outlined above.

General patterns of dispersal in arvicolines

Dispersal behaviour: ontogeny, transience and settlement

Quite surprisingly, we still know little about how dispersal behaviour changes throughout an individual’s life. In arvicoline rodents, a relationship between the timing of natal departure and the onset of reproduction has been found, suggesting that natal dispersal is part of a life history decision associated with sexual maturation (see Box 1 for an overview). In studies that investigated cohorts of offspring born early in the breeding season, it was found that the age of natal dispersal matches the age of sexual maturation for meadow voles and root voles (as young as approximately 25–45 days old, see Bollinger et al. 1993; Andreassen & Ims 2001; Le Galliard et al. 2007), that natal dispersal is concomitant with mating and first reproduction in female Microtus arvalis (Boyce & Boyce 1988) and that individuals engage in longer movements typical of dispersal around sexual maturity in grey-sided voles (Saitoh 1995). A tentative proximate explanation could be that neuroendocrine mechanisms controlling the development of sexual behaviours, such as maturation and pair bonding, also mediate variation in natal dispersal behaviour (e.g. Solomon et al. 2009). Unfortunately, the neuroendocrine basis of natal dispersal has not been investigated, and field studies have yet to identify developmental mechanisms controlling dispersal behaviour.

Table Box 1.   Life history plasticity and natal dispersal strategies in arvicoline rodents
Arvicoline rodents display a very well-documented life history continuum from precocial sexual maturation (as early as 15–25 days old) and fast body growth to delayed maturation until the next breeding season and slower body growth (Lambin & Yoccoz 2001). In addition, arvicolines have a multivoltine life cycle and can produce up to 8 litters per year (Ergon et al. 2001). Variation along the precocial-delayed axis of sexual maturation is observed among muskrats and aquatic water voles characterized by maturation at the age of several months, and among most precocial species of voles and lemmings. The same life history variation occurs between early cohorts of offspring born at the beginning of the breeding season that mature during their birth year and late cohorts of offspring that postpone reproduction until the next breeding season (Lambin & Yoccoz 2001). This life history variation may contribute to variation in dispersal behaviour between and within species. In precocial species, the season of birth determines the benefits of early maturation and therefore of natal dispersal. Thus, natal dispersal decreases seasonally (Table 1), and conditions experienced during the breeding season are likely to be important for the natal dispersal decisions of early cohorts, while late cohorts postpone their dispersal until the next year (e.g. Lambin 1994b). In addition, natal dispersal is more constrained by time and competition with adults in fast- than in slow-maturing cohorts and species (Ims & Hjermann 2001). Fast-maturing dispersers may cue preferentially on conditions experienced by their mother or early in life because these conditions predict well conditions experienced as adults. Slow-maturing individuals have more time to sample their environment and could disperse longer distances but also experience substantial mortality costs. Hence, age at sexual maturation may contribute to explain both intraspecific and interspecific patterns of natal dispersal in microtine rodents and other mammals.

Events of breeding dispersal by females and by males may also be associated with reproductive activities (Stoddart 1970; Kawata 1989; Lambin 1997; Rajska-Jurgiel 2000), but the mechanisms and timing of breeding movements have been generally very poorly investigated. Using our qualitative data, we found that studies most often reported that dispersal is less frequent in adults than in juveniles or subadults [age effect, stronger dispersal in adults: six studies, stronger dispersal in juveniles or subadults: 30, balanced dispersal (no detectable age difference): 10] and suggested that adults may disperse longer distances than juveniles. Our quantitative estimates of emigration fractions and dispersal distances found no detectable differences between age classes (emigration fraction: F1,130 = 2.72, P = 0.10, n = 148; dispersal distance: F1,52 = 0.0002, P = 0.99, n = 69). Thus, age differences in dispersal are ambiguous.

In actively dispersing animals, the transience and settlement stages involve a variety of behavioural strategies leading to habitat selection by dispersers. These stages are the least studied processes of dispersal in arvicolines, and the required continuous monitoring of individuals in their landscape has been carried out satisfactorily so far in only two vole species. In root voles, the transient behaviour of natal dispersers involves 1–3 weeks of active exploration where individuals move between patches on a daily basis (Le Galliard & Rémy, personal observation), and settlement in one patch is often the end result of multiple unsuccessful moves into other, more attractive patches (Le Galliard et al. 2007; Rémy et al. 2011). In the same species, adult males use a very different exploration strategy that involves brief but long-distance movements followed directly by settlement (Steen 1994). Tracking water voles over much larger spatial scales in their natural environment, Fisher et al. (2009) observed that the transient dispersal behaviour of dispersers lasted for several days and involved long-distance movements of several hundreds of metres. Several animals used a ‘stepping-stone’ trajectory where voles successively stopped for several days in one patch and then moved to another patch until they settled down (Fisher et al. 2009).

Dispersal distances

One of the most consistent patterns was a nonuniform dispersal kernel with more short-distance than long-distance dispersal events (34 significant tests of 42 studies of distance-limited dispersal). A total of 13 (of 13) demographic studies reported a decreasing fraction of emigrants with distance to the settlement patch, and eight (of 8) demographic studies found support for a decreasing fraction of immigrants with dispersal distance from the source patch. Despite this observation, only 16 of 24 genetic studies found significant genetic IBD patterns. Yet, genetic IBD patterns may not necessarily be observed even if dispersal is actually restricted in space (Rousset 1997; Leblois et al. 2004; Guillot et al. 2009). In microtine rodents, the cause of no observation of a genetic IBD is not always clear but is attributed to migration–drift disequilibrium in two instances (Ehrich & Stenseth 2001; Francl et al. 2008), to dispersal barriers in one instance (Ratkiewicz & Borkowska 2006), to large effective population size in one case (Guivier et al. 2011) and to substantial, long-distance dispersal in three other studies (Ehrich et al. 2001; Redeker et al. 2006; Schweizer et al. 2007). In general, this demonstrates the difficulty of inferring directly distance-limited dispersal from spatial patterns of genetic variation.

Direct estimates of dispersal distances provided strong evidence for a fat-tailed, right-skewed dispersal kernel (17 of 19 studies of the shape of the kernel). However, a clear limitation of current studies is that only two corrected for lower capture rates of dispersers with distance (Watts 1970; Telfer et al. 2003b) and only two reported formal statistical tests of the shape of the dispersal kernel. Lambin (1994b) found that dispersal distances were skewed to the right in juvenile Townsend voles but did not follow a geometric distribution. Telfer et al. (2003a) found that dispersal distances of juvenile water voles could be approximated by a negative exponential distribution (Fig. 2A). Thus, more studies are needed if empirical kernels are to be compared or used in simulation studies. Long-distance movements are understudied, while determinants of short-distance dispersal have been thoroughly investigated (see next section).

Figure 2.

 General patterns of dispersal in arvicoline rodents. (A) Dispersal kernel in the water vole Arvicola terrestris from Scotland metapopulations. One of the best available estimate of a dispersal kernel was obtained by Telfer et al. (2003b). Dispersal distances were obtained by a combination of mark–recapture and genetic assignment methods accounting for unequal capture rates at the edges of the populations. The observed kernel fitted well to a negative exponential distribution with a mean of 1.8 km. (B) Average dispersal distances reported in the whole literature for arvicoline rodents (n = 100). Most estimates suggest relatively short-distance dispersal in these species of the order of a few 10 m although a tail of long-distance dispersal events was observed across species. Panel A was modified from Telfer et al. 2003b©Molecular Ecology.

Estimates of the average dispersal distance ranged from a few metres in an enclosure study to more than 2 km in a metapopulation study (mean = 163 m, median = 31 m, see Fig. 2B). Dispersal distances were of the same order of magnitude as daily movement capacities calculated for the same species (mean = 68.2 m/day, median = 21 m/day). Thus, arvicolines are generally reported to be dispersing short distances well below what they are capable of. For instance, Steen (1995) recorded movements averaging 700–800 m per night in male root voles equipped with radio-transmitters in a natural population, while the mean dispersal distance for this species is only 167 m (note however that many of the distance estimates were from small-scale enclosure studies where the maximum possible dispersal distance was below 100 m). We compared estimates of dispersal distances across studies and species, which demonstrated significant fixed effects of the method and spatial scale of the study and a significant random effect of the species identity (mixed-effect model after log transformation to ensure normality, method: F3,79 = 3.01, P = 0.03; spatial scale: F1,79 = 34.6, P < 0.0001; species identity(random): χ2 = 46.4, d.f. = 1, P < 0.0001). Not surprisingly, the mean dispersal distance in a study increases with the spatial scale of the study, and, after correcting for this effect, distance estimates were higher when dispersal was assessed by telemetry than by mark–recapture, removal grid or real-time genetic methods, which were undistinguishable. Interspecific differences in dispersal distances could not be explained by body mass (F1,12 = 0.31, P = 0.59) or by the social mating system (F2,11 = 1.15, P = 0.35).

Sex differences in dispersal

Previous reviews of mammalian dispersal have emphasized the strong link between sex-biased dispersal and mating systems (Greenwood 1980; Boonstra et al. 1987; Lawson Handley & Perrin 2007; Clutton-Brock & Lukas 2011). A high male bias in dispersal is expected to evolve in female-defence polygynous mating systems, which stems from the combined effects of inbreeding avoidance and of sexual differences in local mate and resource competition (Lawson Handley & Perrin 2007). Preferential cooperation among female kin and female mate choice for immigrants could strengthen the male bias in dispersal in social mammals. In monogamous species, which is observed less frequently in mammals, sexual asymmetries are weaker and dispersal is predicted to be more balanced (Greenwood 1980). To test this scenario, Boonstra et al. (1987) examined sex-biased natal philopatry and natal dispersal distances in five species of voles ranging from monogamous to polygynous mating systems and all studied through the same mark–recapture protocol (including grid size). They found female-biased philopatry in all species but male-biased dispersal distances only in polygynous species, supporting the contention that the mating system is a determinant of sex-biased dispersal in voles.

Our overview of sexual differences in dispersal in a larger sample of studies confirmed the male-biased pattern identified by Boonstra et al. (1987) but failed to support their relationship with the mating system. Dispersal was male-biased in most species, especially for emigration and immigration fractions, although as many as 45% of the statistical comparisons between sexes were not significant (sex effect on emigration, significant male bias: 44 tests, balanced: 32 tests; immigration, male-biased: five tests, balanced: seven tests; dispersal distance, male-biased: five tests, balanced: 12 tests, female-biased: one test; movement capacity, male-biased: nine tests, balanced: one test). A significant female-biased dispersal was found only for natal dispersal distances in the water vole in two different habitat types (Telfer et al. 2003b; Aars et al. 2006). In addition, and contrary to standard predictions, male-biased dispersal was not more often detected in polygynous or promiscuous than in monogamous mating systems (Fisher’s exact test of independence, P = 0.41). The absence of a relationship between male-biased dispersal and the mating system must, however, be interpreted with caution because of the small number of species included in the comparison and the difficulties with classifying mating systems (see Appendix S1, Supporting Information).

Our quantitative estimates of sex-biased dispersal yielded a rather similar picture. Dispersal distances corrected for differences between studies and species (see section on ‘Dispersal distances’ above) were significantly longer in males (F1,57 = 7.42, P = 0.008, n = 75; females: mean = 114.3 m, males: mean = 149.5 m). Emigration fractions (median = 0.21, n = 231) was influenced by the study method (more emigration was recorded with the removal grid method, see Appendix S1, Supporting Information) and by a negative effect of the time duration during which emigration was calculated, but not by the spatial scale of the study (mixed-effect model after square root transformation to ensure normality; method: F4,184 = 2.38, P = 0.05; duration: F1,184 = 14.4, P < 0.001). After controlling for these methodological effects, emigration fraction varied more within than between species [species identity (random): χ2 = 1.95, d.f. = 1, P = 0.16], and when comparing sexes, emigration fraction was significantly higher for males (F1,111 = 11.75, P < 0.001, n = 128). However, we could not find effects of mating system on these sexual differences in dispersal distances and emigration fractions (interaction sex × social mating system, dispersal distance: F2,55 = 0.62, P = 0.54; emigration fraction: F2,109 = 1.67, P = 0.19).

On the other hand, several studies suggested that sex-biased dispersal could depend on other factors, such as age class or spatial scale. For example, a stronger male bias was found for adult dispersal compared to dispersal of juveniles (13 of 15 studies). This means that breeding dispersal appears to be more strongly male-biased than natal dispersal in several arvicolines. Gauffre et al. (2009) further reported male-biased dispersal at short spatial scales between neighbouring colonies (around 15–600 m) but no sex bias in dispersal between patches of habitat (around 2–23 km). Other field studies have, on the contrary, emphasized that long-distance dispersal is typical of males in polygynous species (Steen 1994).

Processes of dispersal in arvicolines

Condition-dependent dispersal

Strong evidence of condition dependence in dispersal behaviour is apparent in arvicolines (see Table 1). Clear effects of landscape features are seen, including substantial increases in dispersal fractions with landscape connectivity (presence of habitat connection between patches) and matrix quality. Movement capacities are also influenced by small-scale features of the environment, such as habitat gaps and boundaries (e.g. Andreassen et al. 1998a and references therein). Effects of habitat fragmentation, defined as the process during which a large expanse of habitat is transformed into a number of smaller patches, were not consistent (Table 1). Fine-grained habitat fragmentation (i.e. around the scale of individual home ranges) typically increases interpatch movements within home ranges. Indeed, fragmentation of a large habitat patch into several small patches may lead to an expansion of space use because of breeding and foraging requirements, and thus fragmentation can increase interpatch movements at short spatial scales (approximately 10–20 m), although this is not dispersal sensu stricto (Ims et al. 1993; Andreassen et al. 1998b). In landscapes including larger patches and longer interpatch distances (>50–100 m) where interpatch movements could elicit dispersal, the departure stage of dispersal is generally negatively affected in more fragmented habitats (e.g. Diffendorfer et al. 1995). Interpatch distances and landscape geometry are also important determinants of a successful transience because dispersers are at risks of avian and mammalian predation (Ims & Andreassen 2000; Andreassen & Ims 2001; Lambin et al. 2004; Smith & Batzli 2006).

Table 1.   Patterns of condition-dependent dispersal in arvicoline rodents
Factor categoryExternal factorSign of the relationship with movement capacity and dispersal fraction (emigration or immigration) or distance
  1. Sample size refers to the number of studies.

Landscape featuresConnectivityPositive effect on movement capacity (n = 2), emigration fraction (n = 9) and dispersal distance (n = 1)
FragmentationNo effect on movement capacity (n = 1)
Positive effect (n = 2) and negative effect (n = 6) on emigration fraction
Positive effect on dispersal distance (n = 1)
Corridor widthIntermediate corridor width increases movement capacity (n = 2)
Landscape matrixLow matrix quality and presence of dispersal barriers decrease emigration fraction (n = 4)
Patch areaNegative effect on emigration fraction (n = 1)
Patch shapeNo effect of the edge-to-surface ratio on movement capacity (n = 1) and emigration fraction (n = 3)
Habitat quality featuresVegetation qualityDecreases (n = 13), increases (n = 2) or has no effect (n = 2) on emigration fraction
Increases (n = 7) or has no effect (n = 2) on immigration fraction
Habitat destructionIncreases emigration fraction (n = 1)
Predation riskNo effect of simulated predation risk on movement capacity (n = 1) and emigration fraction (n = 1)
Food availabilityDecreases emigration fraction (n = 4), increases immigration fraction (n = 1), no effect on emigration (n = 2)
Prolonged infant malnutrition increases emigration fraction later in life (n = 1)
Site familiarityDecreases movement capacity (n = 1) and increases immigration fraction (n = 1)
Social featuresPopulation densityPositive density dependence (n = 5), negative density dependence (n = 23), no effect (n = 16) on emigration
Negative density dependence (n = 5), no effect (n = 7) on immigration
Population sex ratioNo effect on emigration fraction (n = 4)
Population growthEmigration fraction stronger in growing populations (n = 2)
Population stressMore immigration into populations of testosterone implanted females (n = 1)
Conspecifics and relativesPresence of adult female increases natal emigration and decreases natal immigration by females (n = 2), no effect (n = 1)
Presence of opposite sex increases immigration (n = 3), presence of same sex decreases immigration (n = 3)
Presence of relatives increases emigration (n = 6), decreases emigration (n = 4), no effect (n = 6)
Maternal effectsMaternal factorsSignificant family effects on emigration fraction (n = 2) and movement capacity (n = 4)
No significant family effects on movement capacity (n = 2)
No effect of maternal body mass and size on emigration fraction (n = 2)
Food availabilityNo effect of maternal malnutrition during lactation on emigration fraction (n = 2)
Maternal stressNo effect of maternal stress on emigration fraction (n = 1) and dispersal distance (n = 1)
Litter characteristicsLitter sex ratio (percentage males) increases emigration fraction (n = 3)
Litter sex ratio has no effect on emigration fraction (n = 3) and dispersal distance (n = 2)
Litter size reduces emigration fraction (n = 1) or has no effect on emigration (n = 4) and dispersal distance (n = 1)

Habitat quality and familiarity with the home site are strong determinants of dispersal (Table 1). Features associated with high habitat quality include high vegetation cover and vegetation height, high plant forage quality or the availability of seeds. In most instances, individuals emigrate less from and immigrate more into high-quality habitat patches, which leads to an unequal spatial distribution along habitat quality gradients that can approach an ideal free distribution (Morris & MacEachern 2010). However, in two separate studies, individuals dispersed preferentially from high-quality to low-quality patches (Mazurkiewicz & Rajska 1975; Peles & Barrett 1996), which may be due to competitive exclusion of dispersers by dominant residents. The interaction between habitat quality and competitive exclusion was investigated in detail by Lin and Batzli in field experiments with prairie and meadow voles (Lin & Batzli 2001a,b, 2002). Meadow and prairie voles both preferred high-quality patches but meadow voles were more sensitive to habitat quality and dispersed more frequently than prairie voles. In the presence of meadow voles, the subdominant prairie voles dispersed into lower quality habitats, which resulted in some degree of habitat segregation between the two species. These experiments exemplify how differences in dispersal behaviour can lead to differential habitat choice and competitive exclusion between two sympatric rodents.

A wide range of social factors also influence dispersal fractions (Table 1). Competition with conspecifics should result in enhanced emigration, but most studies indicate that emigration is negatively density-dependent (23 out of 44 tests) or density-independent (16 tests). Negative density-dependent emigration was found in numerous comparisons between low and high densities within or across seasons. Such analyses confound density variation with temporal variation in food or climate conditions and prospects for reproduction early in life (Box 1, but see Andreassen & Ims 2001). A higher dispersal is often observed during the onset of breeding season and the early summer (e.g. Beacham 1979; Rajska-Jurgiel 1992; McGuire et al. 1993), and some controlled experiments suggest that this is partly related to seasonal cohort effects, independent from density fluctuations (Le Galliard et al. 2007; Hoset et al. 2008). Yet, negative density-dependent emigration was also detected using well-designed experiments (e.g. Fortier & Tamarin 1998; Lucia et al. 2008), and five studies of 12 have reached the conclusion that immigration is severely reduced at high population densities (e.g. Gundersen et al. 2002; Jacquot & Solomon 2004). On the other hand, some convincing evidence of positive density-dependent dispersal does exist (Saitoh 1995; Aars & Ims 2000; Gundersen et al. 2001). One way to reconcile these contradictory observations is to assume that selection promotes dispersal from locally high-density to low-density patches, but that the costs of transience and settlement are often higher at high regional population densities, leading to a strong reduction in dispersal movements in dense populations (Solomon 2003). This idea was first proposed under the framework of the ‘social fence hypothesis’ (Hestbeck 1982). Experimental studies with root voles found that density-dependent dispersal proceeds in this manner at small spatial scales (Aars & Ims 2000; Andreassen & Ims 2001; Gundersen et al. 2001).

In addition to the avoidance of competition with conspecifics, dispersal may be involved in mate searching, inbreeding avoidance and the avoidance of kin competition. Studies that have examined social factors in more detail support this multideterminism of natal dispersal, especially for emigration at small spatial scales (Table 1). Young female root voles and Townsend’s voles are likely to emigrate through competition with adult females and are attracted by unrelated males (Lambin 1994b; Gundersen & Andreassen 1998; Gundersen et al. 1999; Le Galliard et al. 2003), while mate searching is a component of breeding dispersal in prairie voles and water voles (McShea 1990; Lambin et al. in press). Avoidance of same-sex individuals is also observed, mainly because of resource competition among females and mate competition among males (Table 1). The effects of kin selection on emigration vary between species. In the monogamous and communal breeding prairie voles, good evidence exists for joint effects of inbreeding avoidance and kin cooperation. Emigration from a natal group is more common when mates within a communal group are related and opposite sex siblings tend to avoid each other (McGuire et al. 1993; Getz et al. 1994). On the other hand, natal philopatry is reduced when parents disappear or when groups are composed of unrelated individuals, and same-sex siblings tend to affiliate with each other (Getz et al. 1994; McGuire & Lowell 1995). In Townsend’s voles, inbreeding avoidance is also involved because males are more likely to emigrate and disperse longer distances when a mother or a littermate is present in the natal range (Lambin 1994b), which is not the case in grey-sided voles (Ims & Andreassen 1991; Ishibashi & Saitoh 2008) and root voles (Le Galliard et al. 2006, 2007). We generally lack evidence that local competition among relatives is an important determinant of natal dispersal (Table 1).

Local kin associations are seen regularly among philopatric individuals, especially matrilineal clusters of females, and kin facilitation is considered as a potential benefit of philopatry because related females breeding in close proximity may benefit in terms of territory acquisition and juvenile survival (see Lambin 1994c; Lambin & Yoccoz 1998; Lambin et al. 2001). However, apart for communally breeding prairie voles, there is no firm evidence that siblings choose preferentially to associate with relatives rather than to emigrate from the natal group when opportunities for independent reproduction are available (Kawata 1987; Lambin 1994c; Le Galliard et al. 2006, 2007). One explanation for the occurrence of female kin clusters in noncommunal species is instead that young related females associate preferentially when strong constraints on dispersal make kin clusters of females a ‘best of a bad job’ solution (facultative philopatry sensu Solomon 2003). This hypothesis remains to be tested.

Condition dependence can occur through a direct behavioural response to current conditions or may involve delayed responses via the life history (see Fig. 1). Most studies conducted so far have tended to support the hypothesis of a direct pathway (277 of 297 tests). An indirect, delayed pathway was assumed only in two different contexts. First, maternal effects on emigration and movement capacities were hypothesized but found in less than one-third of the tests (10 of 29 tests, see Table 1). Maternal effects are still poorly understood, and some manipulative studies did not find effects of conditions experienced in utero or during gestation or lactation (Bondrup-Nielsen 1992; Lambin 1994a; Rémy et al. 2011). Second, it has been demonstrated once that a prolonged malnutrition early in life can increase natal dispersal later in life (Bondrup-Nielsen 1993). This indirect pathway involving scope for delayed life history effects of early-life conditions should be investigated in more detail.

Phenotype-dependent dispersal and dispersal syndromes

The relationship between phenotypic traits and dispersal abilities was examined in a few vole species. Many studies suggest that dispersers are often not a random subset of the population, yet the relationship between phenotypic traits and dispersal varies between studies (see Table 2). Regarding behavioural traits, dispersers are often more active and more thorough explorers than residents in arvicolines reviewed here (nine positive tests of 11). In addition, some inconsistent life history differences between dispersers and residents were detected (Table 2). Ebenhard (1990) predicted from life history theory and observed in an island population that a colonization strategy is associated with a faster growth and larger reproductive effort, but this prediction has not been held up by more recent studies. The support for a genetic basis of observed differences between dispersers and residents is also generally weak (Table 2). Earlier studies have found differences in allozyme diversity between dispersers and residents that could reflect functional differences. These results are difficult to interpret (Pugh & Tamarin 1991) but could fruitfully be revised in the light of recent work showing association between dispersal and enzyme variants in butterflies (Hanski et al. 2004). We therefore urge for the use of modern genetic tools to investigate in more detail the genetic architecture of dispersal and dispersal-related traits in arvicolines. These studies should identify whether genetic differences are associated with behavioural differences in dispersal behaviour rather than ability to survive dispersal (Selonen & Hanski 2010).

Table 2.   Patterns of phenotype-dependent dispersal in arvicoline rodents. For each category of phenotypic trait, we list the identity of the trait and describe the difference observed between dispersing and nondispersing (resident) individuals
Trait categoryTrait identityDifferences between dispersers and residents
  1. Sample size refers to the number of studies by species combination reporting a given relationship.

MorphologyBody conditionNo differences in body condition (n = 2)
Body sizeLarger body size in dispersers (n = 6)
Smaller body size in dispersers (n = 4)
No differences (n = 18)
BehaviourActivityDispersers more active than residents (n = 4)
No differences in activity (n = 2)
ExplorationDispersers more explorer than residents (n = 5)
No differences in exploration (n = 1)
Social interactionsDispersers more aggressive/dominant than residents (n = 4)
Dispersers less dominant than residents (n = 1)
Dispersers less social than residents (n = 1)
Dispersers more social than residents (n = 1)
Life historySize at birthNo differences (n = 1)
Body growthDispersers have faster juvenile growth (n = 1)
Dispersers have slower adult growth (n = 1)
Age and size at maturationDispersers mature earlier (n = 2)
Dispersers mature later (n = 1)
Dispersers mature at a smaller body size (n = 1)
Reproductive effortDispersers make a stronger reproductive effort (n = 1)
No differences (n = 1)
GeneticAllozymeDispersers differ from residents for allozymes (n = 7)
ColorationDispersers exhibit different forehead blaze patterns (n = 1)
Genetic strainA strain disperses longer distances than another (n = 1)
No differences (n = 4)

Interaction between phenotype- and condition-dependent dispersal

Of 157 studies that examined the joint effects of external and internal factors, 118 tested for interactive effects and 63 reported a significant interaction between external and internal factors. Most of these involved sexual differences in condition dependence (45 cases). For instance, females are more sensitive to habitat fragmentation and population density than males (Fortier & Tamarin 1998; Aars et al. 1999; Aars & Ims 2000), and sexes often differ in their sensitivity to other social factors (Table 1). A further five studies showed differences in condition dependence between age classes, in particular stronger effects of social factors and habitat quality on natal dispersal by juveniles (Lofgren et al. 1996; Andreassen & Ims 2001). These results are quite expected because sexes and age classes differ in their competitive ability and limiting resources. Other cases involved individual differences in behaviour (n = 4 of 4 tests), body size or condition (n = 4 of 8 tests) and genotypes (n = 5 of 7 tests) but could not provide clear explanations. Interactions involving body mass and condition are important to consider because size is relevant to social dominance and territorial defence in small rodents. Rémy et al. (2011) manipulated jointly habitat quality and the body size of juveniles to test the idea that dispersers in better physical condition are more successful at settling in high-quality habitats but found no interactive effects on natal dispersal.

Consequences for population dynamics and genetic structure

Dispersal and population dynamics in arvicolines

Arvicolines have violent and often distinctly cyclic population dynamics. Some early hypotheses regarded dispersal as potential cycle driving mechanism (reviewed in Stenseth & Lidicker 1992). Lessons from analytical models have shown that cycle arises in the presence of both direct and delayed density dependence (Stenseth 1999). The lack of strong evidence for delayed pathways for condition-dependent dispersal in the literature indicates that dispersal as such is not likely to be responsible for multiannual cycles. It remains that dispersal is a key demographic process because it is involved in the metapopulation dynamics.

A first major issue is distance-limited dispersal and colonization. Many studies indicate that interpatch distances and habitat quality between patches can severely constrain dispersal and colonization within distances of hundreds of metres and several kilometres (Crone et al. 2001; Witt & Huntly 2001; Pita et al. 2007; do Rosario & Mathias 2007). For example, colonization of patches separated by <15 km depends on relative isolation and the presence of forest in muskrats (Schooley & Branch 2009). An exception to this rule is the existence of extensive natal dispersal over several hundreds of metres in tightly connected fragmented populations of water voles (reviewed in Lambin et al. 2004). Despite good evidence for distance-limited dispersal in all species at spatial scales ranging from 100 to 1000 m (see Fig. 2), demographic effects of habitat fragmentation were modest in five independent experimental studies (La Polla & Barrett 1993; Diffendorfer et al. 1995; Wolff et al. 1997; Davis-Born & Wolff 2000; Andreassen & Ims 2001). This result must however be interpreted with caution because most experiments were conducted at short spatial scales (approximately 50–100 m) relative to the dispersal capacity of microtine rodents and lasted only for one breeding season, while some demographic effects are only seen after several years of fragmentation (Robinson et al. 1992). Distance-limited dispersal has also consequences for spatial population synchrony and range expansion. Getz et al. (1978) and Lambin et al. (1998) reported demographic expansions moving at a mean speed of several kilometres per year in two vole populations. Interestingly, this fast spread at the demographic level is compatible with a relatively poor dispersal capacity at the individual level because demographic processes arise by contagion from nearest neighbour movements and several generations disperse each year (Sherratt et al. 2000).

A second major issue in arvicolines is density dependence given that population densities vary importantly between years and seasons in voles and lemmings and that density-dependent dispersal can contribute to these variations. Experimental studies conducted at short spatial scales with root voles indicate that negative density-dependent dispersal can quickly smooth out spatial differences in population densities (Aars & Ims 2000) and precludes the synchronizing effects of dispersal (Ims & Andreassen 2005). Density-dependent dispersal is also likely to interact with other demographic parameters, and such interactive effects are usually not included in population models. Specifically, dispersal may enhance mortality rates owing to increased exposure to avian predators (Aars et al. 1999). Ims & Andreassen (2000) showed that negative density- and population growth-dependent dispersal acted to enhance summer declines in fragmented root vole populations. Thus, although dispersal does not drive population cycles, it may affect the trajectory of the cycles by deepening the crash phase and binding low-phase populations into more connected networks (Lambin et al. 2004).

The last issue is whether some aspects of condition- and phenotype-dependent dispersal can influence population dynamics. Only a handful of empirical studies have addressed this issue. A first interesting possibility is that condition-dependent dispersal can increase demographic resilience to habitat fragmentation. For example, Lambin et al. (2004) observed that juvenile water voles disperse over long distances to target mates and pointed out that this form of dispersal is efficient at buffering sex ratio variation among patches. On the other hand, nonrandom dispersal could destabilize population dynamics. In small fragmented patches of habitats, Andreassen & Ims (2001) observed that emigration contributed disproportionately more to extinction risk than birth and death. Similarly, Crone et al. (2001) detected increased dispersal in response to habitat degradation, which can precipitate local extinction. Immigration by the territorial sex can also have destabilizing effects on the social organization of residents. Frequent turnover of adult male root voles was associated with higher female mortality, especially when these females were overlapping, but also with higher rates of infanticide and therefore a lower population growth (Andreassen & Gundersen 2006).

Dispersal and genetic structure in arvicolines

The spatial genetic structure and relationship with dispersal are summarized in Table 3 for the best studied arvicoline species. Common patterns observed across these species include a significant genetic differentiation even at small spatial scales, cases of genetic isolation by distance even at spatial scales of a few hundred metres or kilometres, genetic clusters that extend typically over a few kilometres and the existence of local kin structures among females in two species. There are, however, also some intraspecific variation in genetic structure (Table 3). In particular, the fluctuating population dynamics of some arvicoline rodents may influence genetic differentiation. In fossorial water voles, Berthier et al. (2006) documented a change from a high genetic differentiation and low isolation by distance during the low year to a low genetic differentiation but stronger isolation by distance during the peak year. The existence of contrasted genetic differentiation within the same population owing to age or sex structure or to fluctuating population dynamics altogether suggests that sampling consideration could account for the observed variability, which calls for caution when interpreting differences across studies.

Table 3.   Spatial genetic structure and inferences on dispersal patterns within wild populations of arvicoline rodents
Species identityStudy location and scaleGenetic markerReported patternsReferences
  1. We report data from species for which several studies were available. For each study, we indicate the study location and scale (continent or country, number of sampling sites or populations), the genetic marker (mtDNA = mitochondrial DNA, microsatellites = nuclear microsatellite markers). Raw data are provided in Table S1 (Supporting information).

Microtus arvalisEurope (1600 km, 8 pops)
Switzerland (200 km, 5 pops)
High genetic differentiation for mtDNA
Moderate genetic differentiation for microsatellites
Significant genetic isolation by distance for both markers
Male-biased dispersal
Heckel et al. (2005)
Hamilton et al. (2005)
Poland (18 km, 8 sites)MicrosatellitesLow, significant genetic differentiation
No significant genetic isolation by distance
Genetic maps reveal a potential barrier to dispersal
Ratkiewicz & Borkowska (2006)
France (30 km, 193 sites)MicrosatellitesSignificant genetic isolation by distance
Significant genetic correlations below 2–3 km
Indirect dispersal distance of 88 m
Gauffre et al. (2008)
France (24 km, 10 sites)MicrosatellitesSignificant relatedness in females below 100–200 m
Significant male-biased dispersal at a local scale
No sex-biased dispersal at a global scale
Gauffre et al. (2009)
Microtus californicusUSA (5 km, 7 sites)Microsatellites
Significant genetic isolation by distance
Weak genetic correlations
Indirect dispersal distance between 10 and 24 m
Adams & Hadly (2010)
USA (5 km, 11 sites)Microsatellites
Moderate genetic differentiation
Significant genetic isolation by distance
Neuwald (2010)
Myodes glareolusNorway (256 km, 31 sites)mtDNAModerate genetic differentiation
Significant genetic isolation by distance
Significant genetic correlations below 5 km
Stacy et al. (1997)
Norway (10 km, 31 sites)mtDNASignificant genetic isolation by distance
Significant genetic correlations below 2 km
Aars et al. (1998)
Denmark (10 km, 5 sites)MicrosatellitesLow, significant genetic differentiation
No significant genetic isolation by distance
Redeker et al. (2006)
France (80 km, 10 sites)MicrosatellitesLow, significant genetic differentiation
Significant isolation by distance in hedge networks
No significant isolation by distance in forests
Significant relatedness at scales below 100–200 m
Guivier et al. (2011)
Myodes rufocanusJapan (100 m, 1 pop)mtDNAMatrilineal kin structures in spring
Genetic isolation by distance in females
Ishibashi et al. (1997)
Japan (300 m, 1 pop)MicrosatellitesMaternal families in winterIshibashi et al. (1998)
Japan (300 m, 1 pop)MicrosatellitesMale-biased social dispersalIshibashi & Saitoh (2008)
Arvicola terrestrisScotland (25 km, 9 inland sites)MicrosatellitesLow-to-moderate genetic differentiation
Significant genetic isolation by distance
Stewart et al. (1999)
Scotland (6 km, 15 islands)MicrosatellitesHigh genetic differentiationTelfer et al. (2003a)
Scotland (10 km, 5 metapopulations)MicrosatellitesModerate genetic differentiation in juveniles
Low genetic differentiation in adults
Significant genetic correlations below 3 km
Aars et al. (2006)
France (163 km, 23 sites)MicrosatellitesLow, significant genetic differentiation
Significant genetic isolation by distance
Significant genetic correlations below 20–30 km
Berthier et al. (2005)
France (13 km, 7 sites)MicrosatellitesLow, significant genetic differentiation
Significant genetic isolation by distance
Weaker genetic differentiation during peak years
Stronger isolation by distance during peak years
Berthier et al. (2006)

In addition, differences in the genetic structure of a given species could be attributed to variation in demographic processes and flexible dispersal strategies. In particular, we reported above on clear demonstrations of negative density-dependent dispersal, which suggests that differences in population densities or dynamics should influence genetic differentiation. Berthier et al. (2005) revealed genetic discontinuities between areas with contrasting density in fossorial water voles, but only one study compared spatial genetic differentiation across different population dynamics. Ehrich et al. (2009) found for cyclically fluctuating Myodes populations that spatial genetic differentiation declined with the amplitude of the population cycle (i.e. temporal population variability). It was suggested that increased dispersal owing to more extinction–colonization dynamics was the underlying cause.

Our computation of standardized genetic dissimilarity values (inline image) for all populations and molecular markers [allozyme: three studies, mitochondrial DNA (mtDNA) control region: 17 studies, mtDNA RFLP: two studies, nucDNA AFLP: one study and microsatellites: 40 studies] indicated a mean of 0.41 (median = 0.27) at a median spatial scale of investigation of 10 km (see Table S1 for raw data, Supporting information). We compared estimates of inline image across studies using only mtDNA control region and nuclear microsatellite data. inline image was influenced by the genetic marker, spatial scale and number of demes, but not by landscape geometry (mixed-effect linear model after square-root transformation to ensure normality, genetic marker: F1,42 = 25.9, P < 0.0001, spatial scale: F1,42 = 11.1, P = 0.002, number of demes: F1,42 = 8.67, P = 0.005, landscape geometry: F2,42 = 0.52, P = 0.6). inline image increased with the spatial scale of the study and with the number of demes and was stronger for maternally inherited mitochondrial markers than for nuclear markers. When controlling for these factors, inline image varied little and not significantly between species (random, χ2 = 0.25, d.f. = 1, P = 0.62). These results are consequences of the relatively modest dispersal abilities of these species, of the predominant female philopatry (Hamilton et al. 2005) and smaller effective population of mtDNA markers and of the significant sensitivity of dispersal to landscapes’ features in microtine rodents.

Synthesis and conclusion

Our statistical analyses revealed substantial, unexplained variation between species for dispersal distances and a strong variation within species for both dispersal distance and emigration fraction. However, we must acknowledge that our comparison is subject to strong caveats because estimates were collected with different methods and different samples in separate locations for the same species. These methodological aspects consistently explained much of the variation in dispersal fractions, dispersal distances or genetic dissimilarity across studies, which cautions against interspecific comparisons that do not control for them. In addition, serious methodological limitations could bias systematically our quantitative estimates of dispersal. To address this issue, we suggest that future approaches of dispersal should be conducted at larger spatial scales, should try to incorporate data on all age classes and cohorts (see also Box 1), and should also account for unequal capture rates with distance (Telfer et al. 2003b). One promising way to accurately measure dispersal could be to combine mark–recapture and genetic assignment methods where a sufficiently portion of individuals can be sampled (Paetkau et al. 2004). These approaches may, however, be difficult to perform in spatially continuous and high-density populations of arvicoline rodents because of the magnitude of the sampling effort required. In these cases, direct observations of dispersal by means of telemetry could still be an option.

Our comparative analysis portrays species with frequent short-distance dispersal events (of the order of hundreds of metres) and extremely flexible dispersal strategies at these short distances. However, most species are also capable of dispersing long distances (on the order of kilometres) under some circumstances. Female arvicolines are more philopatric relative to males, but we could not find a significant association between the mating system and the degree of male-biased dispersal across species. The absence of a relationship with the mating system may be a consequence of the small number of species included and of a substantial variation in social organization within some species. Future studies of sex-biased dispersal would therefore benefit from including more species and/or from quantifying social and mating systems of each study population independently. An interesting result was the occurrence of sexual differences in condition-dependent dispersal. Sexes thus differed not only in their dispersal abilities but also in their dispersal plasticity, a fact that has rarely been accounted for in previous reviews of sex-biased dispersal.

The analysis of condition dependence shows that dispersal is a response to various proximate and ultimate factors, including competition, inbreeding avoidance, mate searching, habitat quality and the costs of transience and settlement. This result is in line with the dispersal patterns reported for mammals and birds investigated so far (Clobert et al. 2004, 2009). In particular, our review suggests that costs and benefits experienced during transience and during settlement are prime determinants of condition dependence; for example, dispersal is influenced by landscape geometry and habitat quality while prospecting, by predation risks during exploration and by the social conditions experienced during settlement. This indicates that our understanding of dispersal would benefit from more studies of the transience and settlement stages. Condition dependence results primarily from direct effects of current environmental conditions rather than delayed life history effects or indirect inductions through other phenotypic changes (see Fig. 1). We found no common patterns of phenotype-dependent dispersal in the few studies published so far, except for a widespread association between an exploration-activity syndrome and natal dispersal. Clobert et al. (2009) also raised the issue of idiosyncratic dispersal syndromes. They suggested that it could be explained by interactions between condition-dependent and phenotype-dependent dispersal such that different individuals dispersed in different contexts, but this explanation was not supported by our comparative analysis. We instead suggest that it might be more worthwhile to pursue molecular studies to unravel the genetic architecture of condition-dependent behaviour.


We thank Nicolas Perrin, Nelly Ménard, Eric Petit and Jean-Sébastien Pierre for their patience during the slow writing of this manuscript after the conference ‘Social systems: demographic and genetic issues’ organized in Rennes. We thank two anonymous reviewers and Eric Petit for comments that substantially improved a previous version of the manuscript. This study was funded by a grant from the Norwegian Research Council (NFR project 182612). J.-F. L. G was supported by the Région Ile-de-France R2DS program (grant 2007-06) and the Agence Nationale de la Recherche (ANR grant 07-JCJC-0120). XL was supported in part by a Leverhulme Trust Research Fellowship (RF-2011-304).

J.-F. is an evolutionary ecologist with broad interests in dispersal behaviour, life history strategies and population dynamics. This study forms part of a collaborative project during the doctoral research of A.R. at Hedmark University College. R.I. studies processes affecting the dynamics and structure of boreal communities, with a special emphasis on small mammals. X.L. is an ecologist interested in complex temporal and spatial population dynamics and host-parasite interactions.

Data accessibility

All data used in the analyses are available in the Supporting information and on Dryad at doi: 10.5061/dryad.71p127f8.