Developing organisms typically mature earlier and at larger sizes in favorable growth conditions, while in rarer cases maturity is delayed. The rarer reaction norm is easily accommodated by general life history models, whereas the common pattern is not.  Theory suggests that a solution to this paradox lies in the existence of developmental thresholds that define the minimum size at which maturation or metamorphosis can commence, and in the evolution of these threshold sizes in response to environmental variation. For example, more ephemeral environments might favor the evolution of smaller thresholds, enabling earlier maturation. The threshold model makes two unique and untested predictions; first, reaction norms for age and size should steepen, and even change sign, with decreases in the threshold size, and second, food reductions at sizes below the threshold should delay maturation, while those occurring after the threshold should accelerate maturation. We test these predictions by performing targeted food manipulations in a suite of five damselfly species, which theory suggest should differ in threshold size. The results provide strong support for the threshold model’s predictions. In all species, early food reductions delayed maturation, while late reductions accelerated maturation. Reaction norms were steeper, and the effect of food reductions changed from decelerating to accelerating at a much smaller size in fast-developing species from ephemeral habitats. These results support the view that developmental thresholds can account for the widespread observation of negative correlations between age and size at maturity. Moreover, evolution of the threshold appears to be both central to the observed diversity of reaction norms for age and size at maturity, and is predictable.

We conducted a common garden experiment using five species of damselflies from Eastern North America: Enallagma doubledayi, E. pollutum, Ischnura ramburii, I. hastata and I. posita (Odonata: Coenagrionidae). Larvae used in the experiment were hatched from eggs laid by field-collected females sampled at sites in North Carolina, Georgia and Florida, USA, in June and July, 2014. We reared multiple family groups of each species from egg to adult in the laboratory to test for the effect of food reductions within species and for divergence in developmental thresholds and reaction norms among species. One subset of larvae from each family experienced high food levels during the entire experiment, whereas random subsets of individuals from each family were switched from high to low food levels at different body sizes. From the beginning of the experiment, larvae were fed Artemia nauplii ad libitum six days a week (174±34 nauplii per feeding, mean ± 1SD, n=43 random food doses), which represents our high food level treatment. At set intervals, random subsets of individuals from each family were switched from high to low food levels. After the switch, low food individuals were fed three days a week (Monday, Wednesday, Friday) for the remainder of the experiment using the same amount of nauplii as high food larvae. 

To estimate larval size, we measured the head width (maximum distance between the distal parts of the eyes) of living larvae. We converted larval head widths to dry weights by killing ~4 individuals per week and species, drying them for 48h, and weighing them to the nearest 0.001mg. With this data, we established allometric relationships between head width and dry weight for each species. Adult size was estimated by drying and weighing individuals that had emerged and hardened for 24h. For individuals that did not fully emerge from their exuvium, we weighed the adults together with their exuvia. To enable direct comparisons, we weighed the exuviae of successfully emerged individuals and estimated the relationship between the total dry weight (adult + exuvium) and adult dry weight (excluding the exuvium) for each species. Using these relationships, we converted total dry weights into adult dry weights for individuals that were measured with their exuviae. The dataset presented here only include data on adult dry weight (excluding exuviae) estimated in this way.

R code_The evolution of developmental thresholds.R
Computer code for replicating the analyses and figures in the manuscript. The code is fully annotated to allow the user to replicate the analysis.

hatch.rds
Data on the hatching size (hw, head width in mm) of the five species of damselflies studied. Each of the five species (species) was represented by individuals (ind) from multiple full-sib family groups (fam).

morph.rds
Data on the head width (hw.mm, in mm) and dry weight (drywt, in mg) of individual damselfly larvae that were sacrificed during the experiment to establish the allometric scaling between head width and dry weight of each species. Due to a handling error, one batch of individuals were poorly preserved and deemed unsuitable for estimating dry weights. We scored the preservation quality of each individual by visual inspection (quality, range 0 to 3). The analysis was only performed on individuals with no visible signs of deterioration (quality score = 3). For convenience, the natural log of dry weights and head widths is also included (logwt, loghw).

combined.rds
This dataset contain data on the head width (hw, in mm) and dry weight (drywt, in mg; estimated based on the allometric scaling of head width and dry weight of sacrificed individuals) of individual larvae measured at different points in time during the experiment. The age of each individual at the time of the measurement is given in the 'age' variable (days since hatching). Individuals belong to four different treatment groups (treatm.group). They were either continously fed high food levels (treatm.group=HI) or switched to low food levels early, intermediate or late in ontogeny (treatm.group=LO1, LO2 or LO3). Size measurements were taken during ontogenetic growth at constant high food levels (measure.group = "hi"), at the point in time when individuals were switched from high to low food levels (measure.group = "lo.start"), during ontogenetic growth after being switched to low food (measure.group = "post.lo"), when individuals were sacrificed and dried to estimate dry weights (measure.group = "essayed") and after completion of metamorphosis (measure.group = "adult"). The feeding regimen at the time of each measurement is given in the 'treatm' variable (HI or LO). Emergence success was scored on a scale from 0 to 4 by counting the groups of body part (legs, head, thorax and abdomen) that remained inside the exuvium (stuck.score, range = 0 to 4).