There is considerable debate about the occurrence of assortative mating between phenotypic traits measured within natural populations. Meta-analyses have implied that assortative mating occurs generally in natural populations but recent work indicates these conclusions largely result from biased data. Specifically, estimates of phenotypic correlations between mating partners do not solely result from non-random associations between individual-level traits of partners but also from other biological processes (joint phenotypic plasticity, indirect genetic effects), methodological practices (observer bias), and other unexplained residual correlations (e.g. correlated measurement error). This paper puts this critique to test. First, we estimated the overall phenotypic correlation between phenotypic traits of mating partners for a wild population of great tits. Second, we estimated various key variance components to reveal the extent to which phenotypic correlations between partners resulted from assortative mating, reversible plasticity, social partner effects, and methodological practices. We performed our analyses for a range of phenotypic traits (body mass, breathing rate, exploration behaviour, wing and tarsus length) to derive general conclusions not hinging on the specifics of the traits involved. Our analyses support the conclusion that patterns of assortative mating exist at first glance but occur because of the biasing effects of correlated residuals likely caused by a combination of phenotypic responses to unknown environmental factors or measurement error – not because of intrinsic patterns of assortative mating.

Between 2010 to 2019, we studied a great tit (Parus major) population breeding in 12 nest-box plots near the city of Starnberg in Southern Germany. Seven to 10 days after hatching, we caught the parents with a spring trap placed inside the nest-box and marked each individual, if not captured before, with a unique combination of one metal- and up to three colour rings. Immediately after ringing, birds were subjected to a cage-test to measure their exploration behaviour. After the test, we removed the focal bird from the cage and measured its breathing rate, tarsus and wing length, and body mass.

We conducted all data analyses in R v. 4.0.2 (R Core Team, 2020), and all models were fitted within a Bayesian framework using Markov chain Monte Carlo (MCMC) methods in the MCMCglmm package (Hadfield, 2010). Please also refer to the Methods section of the main manuscript for a detailed description of conducted analyses.