• capture–mark–recapture statistical modelling;
  • density dependence;
  • Mastomys natalensis;
  • state-dependent life history;
  • tropical rodent demography

1. By identifying ecological factors specific to functional categories of individuals, it may be possible to understand the mechanisms underlying life-history evolution and population dynamics. While empirical analyses within the field of population biology have focused on changes in population size, theoretical models assuming differential sensitivities of population growth rate or fitness to demographic parameters have mostly been untested, particularly against data on small mammals.

2. Statistical modelling of capture–mark–recapture data on the multimammate rat (Mastomys natalensis) from Tanzania shows that: (i) females survive slightly better than males and subadults survive much better than adults; (ii) temporal variation of survival of all individuals is similarly related to the rainfall of the month; (iii) subadults exhibit a strongly density-dependent low persistence rate in the population immediately after their first capture; (iv) subadults survival in later months is, however, positively related to density; and (v) adult survival shows negative density-dependence.

3. Both density-dependent and density-independent factors simultaneously determine stage-dependent survival variation of the multimammate rat. Whereas environmental factors in this population seem to affect survival rates of all individuals in a similar manner, density-dependent relationships are more complex.

4. The patterns of survival variation in small mammals may be different from those observed in large mammals.

5. Further studies of demography in small mammals should aim at understanding how much of the variability in population growth rate is accounted for by the variability of the demographic rates resulting from limiting (density-independent) and regulating (density-dependent) factors, respectively. This study emphasizes the use of robust and accurate statistical methods as well as stage- or age-structured population modelling.