Three concerns have been raised by Sánchez-Tójar et al . (2020) about our meta-analysis of transgenerational effects (Yin et al . 2019): (1) the analysis did not fully account for the non-independence among data, (2) the study lacked transparency and the peer review was ineffective, and (3) the literature search had limited coverage. These concerns are addressed below.
First, Sánchez-Tójar et al . raised concerns regarding data non-independence, and in their reanalysis, they attempted to account for phylogenetic non-independence by assuming a Brownian motion model of trait evolution. The revised estimates for effect sizes were largely consistent with our own, and all remained significant except when accounting for phylogenetic non-independence (Fig. 1). However, applying this analysis to our data is problematic for three critical reasons. First, the phylogenetic non-independence assumes that traits diverge over time from the ancestral trait of a common ancestor (Felsenstein 1985). This assumption is violated by our data, as most traits are non-homologous (Fig. S1). Second, the Brownian motion model assumes a constant rate of evolution, but this assumption is often unrealistic for traits under selection (Felsenstein 1985). Third, to analyze the data with plants and animals, the Brownian motion model causes great loss of statistic power due to the high variance-covariance estimated (Fig. S2 and S3). If we use other models of trait evolution, such as assuming an accelerating rate of evolution (Blomberg et al . 2003) or assuming incomplete phylogenetic dependence of trait evolution (Pagel 1997), significant results are retained (Fig. 1). Importantly, whether the results are significant or not depends on different assumptions of evolutionary models rather than the data, and notwithstanding the effects of different assumed evolutionary models, the effect size remains consistent across models (Fig. 1). Thus, our main finding still stands that transgenerational effects do indeed seem to be adaptive.
Concerning the transparency of our study and suggestions that the peer-review was ineffective, both the data and R scripts of this study have been provided since the first submission. Given this level of transparency, the reviewers have already commented on most issues raised by Sánchez-Tójar et al ., and we have responded accordingly. These issues include searching in subject areas, dealing with missing keywords, combining data from full-factorial and non-full-factorial experimental designs, and further issues in the supplementary materials. The reviewers also provided constructive suggestions for dealing with difficult issues such as how to decide if a trait is negatively related to fitness. To give even more transparency, a point-by-point reply to reviewers from our initial study submission is provided as supplementary material.
Finally, we agree with Sánchez-Tójar et al . that studies may have been left out of our original meta-analysis. Indeed, transgenerational effects are found across the tree of life, indicating that they either have a deep evolutionary origin or have arisen many times (Salinaet al . 2013; Uller et al . 2013). An essential aim of our meta-analysis was to reveal the specific scenarios that give rise to transgenerational effects. Thus, we fully agree that meta-analyses incorporating more data will further our understanding of when and where transgenerational effects are adaptive.
We thank Sánchez-Tójar et al. for their comment, and we hope these discussions will promote open science practice and inspire future comprehensive meta-analyses of transgenerational effects.