A popular theoretical model for explaining the evolution of complex life cycles was provided by Earl Werner. The theory predicts the size at which an individual should switch stages to maximise growth rate relative to mortality rate across the life history. 

Werner’s theory assumes that body size does not change during the transition from one phase to another (e.g. from larva to adult)—a key assumption that has not been tested systematically but could alter the predictions of the model. 

We quantified how growth rate and mass change across larval stages and metamorphosis for 105 species of fish, amphibians, insects, crustaceans and molluscs Across all taxonomic groups, we found support for Werner’s assumption that growth rates are maintained or increase around transitions. We found that changes in growth and mass were greatest during metamorphosis, and change in growth correlated with development time. Importantly, most species either gained or lost mass when switching to a new stage—a direct contradiction of Werner’s assumption. When we explored the consequences of energy loss and gain in a numerical model, we found that individuals should switch stages at a larger and smaller size, respectively, relative to what Werner’s standard theory predicts.

Our results suggest that while there is support for Werner’s assumption regarding growth rates, mass changes profoundly alter the timing of transitions that are predicted to maximise fitness, and therefore the original model omits an important component that may contribute to the evolution of complex life cycles. Future studies should test for conditions that alter the costs of transitions, so that we can have a better understanding of how mass loss or gain affects fitness.

We conducted a comparative analysis of larval growth rates to test Werner's theory on complex life cycles. For each study, we extracted body mass and age data across immature stages during development. We noted whether transitions were between stages within the larval phase (e.g. from one larval stage to another), or whether transitions were from the larval phase to the adult phase (i.e. metamorphosis). Finally, we recorded the total number immature stages for each species. 

To test Werner’s hypotheses across species, we first divided mass and age by the maximum size and age for that species, so that we were working with relative values (i.e. both mass and time bound between zero and one). We fit linear, quadratic and exponential models to the relative mass over time data for each larval stage and adult phase within species, and used Akaike Information Criterion (AIC) to select the most parsimonious model (Table S4; Akaike 1978, Quinn and Keough 2002). In cases where the most parsimonious model yielded body mass predictions that were biologically irrelevant (e.g. negative values), we selected the next most parsimonious model. Using the body mass predictions from the fitted curves, we calculated derivatives as a measure of instantaneous growth rate (Anger 1991; see Figure S2 for example). For our response variable, we found the derivative at the last time point of the initial stage/phase and the derivative at the first time point in the next stage/phase (Figure 1). Then, we calculated the difference between the two derivatives. If the change in growth rate is a positive value, growth rate after transitioning is greater than growth rate before, while a negative value indicates growth rate before is greatest. As a second response variable, we calculated the change in mass—initial mass after transition to a new stage/phase minus the final mass in the initial stage/phase (Figure 1). If the change in mass is a positive value, the organism grew during the transition and if the change in mass is a negative value, the organism lost mass. Finally, for species with multiple larval stages (i.e. crustaceans and insects) we tested how change in growth rate and change in mass correlated with development time.