Children vary in the extent to which their development is shaped by particular experiences (e.g. maltreatment, social support). This variation raises a question: Is there no single level of plasticity that maximizes biological fitness? One influential hypothesis states that when different levels of plasticity are optimal in different environmental states and the environment fluctuates unpredictably, natural selection may favor parents producing offspring with varying levels of plasticity. The current article presents a mathematical model assessing the logic of this hypothesis--specifically, it examines what conditions are required for natural selection to favor parents to bet-hedge by varying their offspring's plasticity. Consistent with existing theory from biology, results show that between-individual variation in plasticity cannot evolve when the environment only varies across space. If, however, the environment varies across time, selection can favor differential plasticity, provided fitness effects are large (i.e. variation in individuals' plasticity is correlated with substantial variation in fitness). Our model also generates a novel restriction: Differential plasticity only evolves when the cost of being mismatched to the environment exceeds the benefits of being well matched. Based on mechanistic considerations, we argue that bet-hedging by varying offspring plasticity, if it were to evolve, would be more likely instantiated via epigenetic mechanisms (e.g. pre- or postnatal developmental programming) than genetic ones (e.g. mating with genetically diverse partners). Our model suggests novel avenues for testing the bet-hedging hypothesis of differential plasticity, including empirical predictions and relevant measures. We also discuss several ways in which future work might extend our model.

A casual look at the literature in social cognition reveals a vast collection of biases, errors, violations of rational choice, and failures to maximize utility. One is tempted to draw the conclusion that the human mind is woefully muddled. We present a three-category evolutionary taxonomy of evidence of biases: biases are (a) heuristics, (b) error management effects, or (c) experimental artifacts. We conclude that much of the research on cognitive biases can be profitably reframed and understood in evolutionary terms. An adaptationist perspective suggests that the mind is remarkably well designed for important problems of survival and reproduction, and not fundamentally irrational. Our analysis is not an apologia intended to place the rational mind on a pedestal for admiration. Rather, it promises practical outcomes including a clearer view of the architecture of systems for judgment and decision making, and exposure of clashes between adaptations designed for the ancestral past and the demands of the present.

Children learn a lot from others, but the effectiveness of their social learning depends on the reliability of others' help. How do children adapt their future learning decisions based on the past reliability of receiving help? In two experiments, 4- to 6-year-olds (N = 60 each) interacted with a researcher who either followed through on promised help (Reliable condition) or failed to do so (Unreliable condition). Experiment 1 was inconclusive. However, with an improved design, Experiment 2 found that children in the Unreliable condition were more likely to forego a harder but more rewarding puzzle as their next task and choose an easier, less rewarding puzzle instead compared to those in the Reliable condition. Such decisions, while seemingly maladaptive at face value, likely reflect an adaptive response to the low likelihood of receiving help. These results extend our understanding of social learning across diverse ecological contexts.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

We present a consensus-based checklist to improve and document the transparency of research reports in social and behavioural research. An accompanying online application allows users to complete the form and generate a report that they can submit with their manuscript or post to a public repository.