Skip to main content

Patterns, predictors, and consequence of dominance in hybrids

Cite this dataset

Thompson, Kenneth et al. (2020). Patterns, predictors, and consequence of dominance in hybrids [Dataset]. Dryad.


Compared to those of their parents, are the traits of first-generation (F1) hybrids typically intermediate, biased toward one parent, or mismatched for alternative parental phenotypes? And how does hybrid trait expression affect fitness? To address this empirical gap, we compiled data from 198 studies in which traits were measured in a common environment for two parent taxa and their F1 hybrids. We find that individual traits in F1s are, on average, halfway between the parental midpoint and one parental value (e.g., hybrid trait values are 0.75 if parents'values are 0 &1). When considering pairs of traits together, a hybrid's bivariate phenotype tends to resemble one parent (pairwise parent-bias) about 50 % more than the other while also exhibiting a similar magnitude of mismatch due to different traits having dominance in conflicting directions. We detect no phylogenetic signal nor an effect of parental genetic distance on dominance or mismatch. Using data from an experimental field planting of recombinant hybrid sunflowers, we illustrate that pairwise parent-bias improves fitness whereas pairwise mismatch reduces fitness. In sum, our study has three major conclusions. First, hybrids between ecologically divergent natural populations are not phenotypically intermediate but rather exhibit substantial mismatch while also resembling one parent more than the other. Second, dominance and mismatch do not seem to be governed by general rules but rather by the idiosyncratic evolutionary trajectories of individual traits in individual populations or species. Finally, selection against hybrids likely results from selection against both intermediate and mismatched phenotypes.


The systematic review data were compiled from the literature (many datasets came from emailing authors). The sunflower data were collected by Ken Whitney & co. All details given in text. Data are raw, as collected, in the 'raw_data.'

Usage notes

Everything should work well if directories are established using the .Rproj file in RStudio. See meta-data and readme file for further instructions. Email me ( with any queries. Data collection for follow-up projects is ongoing, and if you are interested in the most current dataset please contact me. I'd be happy to collaborate and/or share the updated dataset.