not-yet-known not-yet-known not-yet-known unknown Methods Study system: The brown anole (A. sagrei ) is a small tropical lizard native to Cuba and The Bahamas and invasive throughout Florida, USA, and other parts of the world (Kolbe et al. 2014; Losos 2011, Figure 1a). They are considered a generalist and are successful biological invaders because they tolerate a broad range of thermal conditions (Kolbe et al. 2014), inhabit a range of structural environments, consume a diversity of prey (Chejanovski et al. 2022), and have a high reproductive rate (March to October; Lee et al. 1989; Hall et al. 2020). Females bury eggs in a variety of microhabitats and embryos successfully develop over a broad range of thermal and hydric conditions (Pruett et al. 2020; Hall & Warner 2021). Our study populations are from human-made islands in Cresent Beach, Florida, which were formed during the creation of the Atlantic Intracoastal Waterway between 1928 and 1970 (Baker 2014). The primary substrate is marine sediment (sand and crushed coquina), but trees (cedar trees, live oaks, and palm trees) and other vegetation have colonized the islands, forming varying densities of dark organic soil. Thus, some islands are mostly “shaded” with many tall trees that provide cool, moist nesting habitat. Conversely, others are “open” with little shade and relatively warm, dry nest environments. Importantly, the percentage open and shaded habitat differs across islands with conditions on open islands being more variable than shaded islands. Brown anole nests are found in both open and shaded microhabitats on each island, and eggs experience considerable variation in temperature, moisture, and substrate type within and among islands (Pruett et al. 2022). These factors influence many traits in reptiles (Warner et al. 2018) including developmental rate, egg hatching success, body size, and performance in A. sagrei (Warner et al. 2012; Pearson & Warner 2018; Pruett & Warner 2021). Moreover, the combined effects of several nest abiotic conditions can have important consequences for A. sagrei development (Hall et al., 2022). Breeding colony collection and husbandry: Lizards were collected from a shaded (Figure 1b) (nmales=17; nfemales=32) and open (Figure 1c) (nmales=21; nfemales=33) island on 29 June 2020. The open island total vegetated area was only 5.9%, whereas the shaded island was 98.3% covered by vegetation (analyzed in ImageJ, Schneider et al. 2012). Females were housed individually in our laboratory at Auburn University in plastic cages (29x26x39 cm), and were provided a nesting pot filled with moist peat moss (checked for eggs thrice weekly), two perching rods, artificial plants, and a 12 h:12 h light:dark cycle of UV light. A male lizard was swapped weekly between a block of two cages, but males from one population were never mixed with females from another. We fed lizards two crickets (dusted with calcium and vitamin powder) twice per week. Egg incubation: Eggs were allocated to one of two incubation treatments that mimicked multivariate nest conditions (temperature, moisture, substrate) on the shaded and open islands (see Pruett et al. 2020). Each female’s first egg was randomly allocated to one treatment, and subsequent eggs were alternated between the two treatments to create a split-clutch design. Thermal and hydric conditions for the shaded treatment were based on average values for three nests that had incubated in organic soil collected from the shaded island, dried to constant weight, and rehydrated to 50% moisture by mass (Figure 2a). Conditions for the open treatment were based on average values for three nests that had <40% shade cover. The eggs in this treatment were incubated in sand/crushed shell substrate collected from the open island, which was dried to constant weight and rehydrated to 5% moisture by mass (Figure 2b). We created water retention curves for these substrates using a Wescor microvolt meter equipped with a C-52 sample chamber (Figure 2c); based on these curves, incubation substrate in the open and shaded treatments were about -150 kPa and -30 kPa, respectively, which result in high hatching success (Hall et al. 2022). To mimic nest temperatures, we created two 16-day thermal profiles collected from nests using Thermochron iButtons (Pruett et al. 2020) (Figure 2d). These 16-day profiles were uploaded into four programmable incubators (Memmert IP55 Plus; two incubators per treatment) and looped continuously. Eggs were incubated in glass jars (59.14 ml) filled approximately half full of substrate. Plastic wrap was stretched across the top and secured with a rubber band to reduce evaporation but allow gas exchange. Jars were provided fresh substrate once per week to maintain moisture treatments throughout development. We checked incubators for hatchlings daily. Hatchling measurements and care: Hatchlings were measured for snout-vent length (SVL), tail length (nearest mm) and mass (nearest 0.0001 g) and placed individually in a clear plastic cup (946.3 ml, 11.4 cm diameter x 13.9 cm tall) that contained a bamboo perch and artificial plant and was sealed with a screen mesh top. Hatchlings were fed 10 fruit flies (dusted with calcium and vitamin powder) three times per week and experienced the same ambient conditions as adults. Desiccation rate varies with habitat in Anolis lizards (Gunderson et al. 2011) and can impact hydration in ways that influence performance (Walvoord 2003, Moeller et al. 2023). At 3-5 days post-hatch, desiccation rate was measured by placing hatchlings in small cylinder-shaped mesh containers (approximately 3 cm diameter x 10 cm long) that were laid inside a glass chamber filled with Dry Rite™ desiccant (Supplemental Figure 1). The glass chamber was placed in an incubator set to 28 °C for one hour. We measured the body mass of hatchlings before and after the trial to calculate water loss over the one-hour interval. Fecal matter was left in the chamber in 12 trials (3.87% of all trials), and these data were excluded from analysis. Due to slight changes in humidity, we measured the relative humidity of the chamber for analysis. Because locomotor performance influences survival in hatchling lizards (Miles 2004), we measured sprint speed (at 7-9 days post hatching) by chasing lizards along a 100 cm racetrack containing photo sensors every 25 cm. The track was placed at a 45º incline and composed of wood with grains of sand glued to the surface to enhance traction and stimulate natural running behavior (Supplemental Figure 2). Hatchlings were prompted to run by gently tapping the base of the tail with a paint brush. Each hatchling was raced three times, with a 5–10-minute rest between trials. Across all trials, we considered the fastest speed across a single 25 cm section of the track to be the sprint speed. Hatchlings that refused to run or jumped out of the track were given a rest period and ran again. Hatchlings with more than five failed attempts were excluded from analysis. Data analysis: All analyses were performed in R (R Core Team 2021). To consider how adult females differ between islands, potentially influencing reaction norms, we analyzed female SVL, mass, and fecundity using general (SVL, mass) and generalized (fecundity; Poisson distribution) linear models with island source as a fixed effect. To address our first objective which considers the effects of multivariate conditions on plasticity: we used linear mixed-effects models to analyze egg mass, egg survival, incubation period, hatchling body mass, SVL, sprint speed and water loss, as response variables and included source island, incubation treatment, and appropriate covariates (see below) as fixed effects. We compared models with different random effects structures to determine the best fit (see below). We initially included all two-way interactions but removed them from final models if they were not statistically significant (Engqvist 2005, Zuur et al 2009). We used package “lme4” (Bates et al. 2015) to conduct general and generalized linear mixed effects models for phenotypes and survival (binomial distribution), respectively. For hatchling phenotypes, data were scaled and centered using the “scale” function (Becker et. al. 1988). For analyses of hatchling size (i.e., SVL and mass), egg mass and oviposition date were included as covariates to account for these specific maternal effects; indeed, females produce increasingly larger eggs throughout the breeding season (Pearson & Warner 2018). For analysis of sprint speed, we included hatchling SVL, the number of stops by lizards during running trials and the ambient room temperature as covariates as all these variables can affect running speed (Logan et al 2018). For analysis of desiccation rate, relative humidity was a covariate to account for small variation in the desiccation chamber, and values of water loss were log transformed to reduce heteroskedasticity. For analysis of incubation period (number of days between oviposition and hatching), egg mass was included as a covariate because larger eggs often take longer to develop. Additionally, because eggs were collected three times per week some eggs remained longer in the nesting pots than others, and therefore had more time to absorb water; including egg size as a covariate allowed us to control for this effect. For egg and hatchling survival, we included week of oviposition as independent variables to account for possible differences in seasonal maternal investment (Hall et al. 2020). To address our second objective, which was to test the hypothesis that plasticity varies among family groups, we conducted three types of analyses: comparisons of mixed models, tests for gene by environment interaction, and assessment of family-level regressions. We used these analyses which were more (model comparisons) or less (family-level regressions) conservative because statistical and biological significance are not identical, and selection may still operate on reaction norms even when differences among genotypes (or families) are not statistically supported (Morrissey & Lieftin 2015). Thus, the mixed model comparisons provide the strongest support of family-level variation in reaction norms and the family-level regressions relatively weaker support. For model comparisons of hatching phenotype, we constructed four models that varied in random effects structure and evaluated their fits using an F-test. Maternal identification was used as the random effect, and the model structures were random intercept, random slope, random intercept and slope, and a null model with no random effect (Figure 3). A random intercept model indicates family-level differences in the mean trait value, but little among-individual plasticity (Figure 3a). This supports consistent past selection on plasticity and weak selection on the value of the phenotype (i.e., homogenous reaction norms) and indicates little present opportunity for plasticity evolution. A random slope model indicates little variation in the mean phenotype, but substantial among-individual plasticity (Figure 3b), indicating strong past selection on the phenotype and weak selection on plasticity, but presents opportunity for the evolution of plasticity. The random slope and intercept model indicates weak past selection on both the phenotype value and plasticity (Figure 3c.) and presents opportunity for selection to operate on both. Finally, the null model indicates strong past selection on both the phenotype value and plasticity (Figure 3c) and little, if any, current opportunity for natural selection to shape either. For our next set of analyses, we used linear models to explicitly test for genotype by environment interactions in all the egg and hatchling traits. These linear models included incubation treatment, maternal ID, and their interaction as fixed effects. A statistically significant interaction term would indicate family-level variation in reaction norms. Our final set of analyses used a less conservative approach; here, we analyzed each family group independently and visualized differences in plasticity within and between islands. We calculated family-level reaction norms using a linear model of sibling offspring (i.e., offspring produced by the same female) from both treatments. Only families that contained at least 5 total siblings and 2 siblings per treatment were used in this analysis. The slopes of the models were then grouped by source island. Slope values of these maternal reaction norms were compared using liner models for among and within island habitat types and analyzed using fishers F test and Fligner-Killing test. This allowed us to directly compare each family group (an index of genotype) and how they interacted across developmental environments both among and within populations. Because Fishers F test can be susceptible to outliers (Crawley 2013), Fligner-Killeen tests were also performed to test for variance to ensure robustness.