The measures *H*_{O} and *HL* were highly correlated (*r* = −0.974, *P* < 0.001) and for this reason statistical analyses are only presented for *H*_{O} analyses. *HL* data provided analogous results and are available from the authors. Heterozygosity was positively correlated between social mates during 2007 (*r* = 0.335, *P* = 0.006, *n* = 67) and 2006 (*r* = 0.181, *P* = 0.041, *n* = 94) breeding seasons, but no significant correlation was found in the previous years (2005: *r* = 0.101, *P* = 0.323, *n* = 97; 2004: *r* = −0.031, *P* = 0.738, *n* = 118; 2003: *r* = −0.129, *P* = 0.466, *n* = 34; 2002: *r* = 0.040, *P* = 0.869, *n* = 19). The strength of the correlation (estimated by the correlation coefficients, *r*) between mate heterozygosities has increased over the 6 years of study (WLS linear regression analysis; *r* = 0.913, *P* = 0.011, *n* = 6). We analysed assortative mating within colonies only considering those colonies with five or more genotyped pairs. No correlation was significant in any colony and year (all *P* > 0.1), probably because sample sizes (mean ± SE = 10.52 ± 5.16 pairs per colony; range = 5–21 pairs per colony) are relatively small when each colony is analysed separately. However, we found that correlation coefficients were significantly different from zero in 2006 (student *t*-test: *t* = 4.703, *P* = 0.009, *n* = 5 colonies) and 2007 (student *t*-test: *t* = 2.906, *P* = 0.034, *n* = 6 colonies) but not in 2005 (*t* = 0.224, *P* = 0.831, *n* = 6 colonies) and 2004 (*t *= −0.031, *P *=* *0.976, *n *=* *10 colonies) years. The number of colonies with five or more analysed pairs in previous years was too small as to be adequately analysed (2003: *n *=* *2; 2002: *n *=* *2). A general linear model weighted for the number of analysed individuals within each colony and controlling for colony identity (included as random effect; *F*_{1, 29 }=_{ }1.41, *P *=* *0.248) showed that correlations coefficients of intra-pair male and female heterozygosities correlations have progressively increased over the study years (year included as covariate; *F*_{1, 29 }=_{ }11.18, *P *=* *0.003). However, we found no significant interaction between colony identity and study year, indicating that the observed temporal pattern was similar among the studied colonies (*F*_{10, 19 }=_{ }1.41, *P *=* *0.249). We explored the possibility that the observed correlation of intra-pair male and female heterozygosities is a consequence of differences in genetic diversity between the study colonies which may have resulted in a passive heterozygosity-assortative mating independent of individual preferences (*sensu*Ferrer & Penteriani, 2003; see García-Navas *et al.*, 2009). We found that individual genetic diversity did not differ between colonies in any study year (one-way anovas, 2007: *F*_{4, 107 }=_{ }0.98; *P *=* *0.422; 2006: *F*_{6, 149 }=_{ }1.07; *P *=* *0.384; 2005: *F*_{1, 172 }=_{ }0.62; *P *=* *0.813; 2004: *F*_{12, 209 }=_{ }0.62; *P *=* *0.820; 2003: *F*_{6, 56}=_{ }1.72; *P *=* *0.134; 2002: *F*_{3, 27 }=_{ }0.847; *P *=* *0.480). The age structure neither differed among the studied colonies in any year (*P *>* *0.1 in all cases). Pair-wise *F*_{ST} values calculated using the program Fstat 2.9.3.2 (Goudet, 2001) showed a complete absence of genetic differentiation among the studied colonies in all study years (all *P *>* *0.1 in all cases).

Size (tarsus and wing length) and pectoral thickness were not correlated between social mates in any study year (all *P *>* *0.1). However, individuals showed an age-based assortative mating in some study years (2007: *r *=* *0.309, *n *=* *67, *P *=* *0.011; 2006: *r *=* *0.131, *n *=* *94, *P *=* *0.207; 2005: *r *=* *0.361, *n *=* *97, *P *<* *0.001; 2004: *r *=* *0.367, *n *=* *118, *P *<* *0.001; 2003: *r *=* *0.249, *n *=* *34, *P *=* *0.149; 2002: *r *= −0.119, *n *=* *19, *P *=* *0.616). Linear regression analyses revealed that age, individual size and pectoral thickness were not associated with heterozygosity in any year and sex (all *P *>* *0.1).