Abstract
In many species, individuals who experience harsh conditions during development often have poor health and fitness outcomes in adulthood relative to peers who do not. There are two classes of evolutionary hypotheses for the origins of these early life contributors to inequality in adulthood: developmental constraints (DC) models, which focus on the deleterious effects of low-quality early-life environments, and predictive adaptive response (PAR) hypotheses, which emphasize the cost of mismatches between early and adult environments. Distinguishing DC and PAR models empirically is difficult for both conceptual and analytical reasons. Here, we resolve this difficulty by providing explicit mathematical definitions for DC, PARs, and related concepts, and propose a novel, quadratic regression-based statistical test derived from these definitions. Simulations show that this approach improves the ability to discriminate between DC and PAR hypotheses relative to a common alternative based on testing for interaction effects between developmental and adult environments. Simulated data indicate that the interaction effects approach often conflates PARs with DC, while the quadratic regression approach yields high sensitivity and specificity for detecting PARs. Our results highlight the value of linking verbal and visual models to a formal mathematical treatment for understanding the developmental origins of inequitable adult outcomes.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
Text has been updated for greater clarity, including clarifications about the nature of our definitions of hypotheses. In addition, we further explain the 'real world' implications of hypothesis testing strategies for biologists studying early life effects.
↵6 The term immediate adaptive response is in use elsewhere in the literature [e.g., 18, 46]. It is sometimes used as a synonym for developmental constraints, so we deliberately avoid using it here.
↵7 It is true that a developmental adaptive response (DAR) could trigger adaptation. However, the problem there is that PAR and DAR are indistinguishable, not that the test for PAR is incorrect. Another way to put this is that the test for β < 0 is a test for both PAR and DAR, not that omitted variable bias means that this test is inaccurate. This is discussed in detail in Section A.2.
↵8 For example, a 4th order polynomial has 15 terms; because we consider 5 values of each coefficient in the polynomial, a 4th order polynomial requires consideration of 515 rather than 510 realities.
↵9 Note that the range of coefficients is not the same as the range of, e.g., e0. Instead it is the range of possible causal impact of powers of environmental quality on outcomes.
↵10 An affine function of a bounded, continuous variable can transform it into continuous variable with a range of [0,1]. Moreover, the test for PARs or DC can be adjusted to account for that transformation.