**Planned sample** We plan to collect 500 American residents or citizens from the Project Implicit research website (https://implicit.harvard.edu). 500 participants will afford 80% power to detect r = .124 and 95% power to detect r = .160. Participants will be excluded from all analyses if more than 10% of the trials on the ST-IAT were faster than 400 milliseconds. **Confirmatory hypotheses** Each hypothesis is followed directly by the planned statistical analyses. *H1*: Explicit and implicit environmentalist identification will be weakly or moderately positively correlated (as in S1). The relationship between explicit and implicit identification will be tested using Pearson's r correlation. *H2a*: Explicit and implicit identification will each uniquely positively predict civic environmental behaviors and environmental policy opinions. Using two multiple linear regressions, explicit and implicit identification will predict (1) civic behaviors and (2) policy opinions. *H2b*: The effect of explicit identification on policy opinions will be steeper for social conservatives because they may perceive more intergroup conflict between environmentalists and social (vs. economic) conservatism. Interaction term will be created by multiplying the standardized variables for social conservatism and explicit identification, and then predicting policy opinions with both main effects and the interaction in regression. *H2c*: Implicit identification will positively predict behaviors and policy opinions above and beyond self-nature overlap. Two additional regressions for behaviors and opinions predicted from implicit and explicit identification will now also include self-nature overlap. *H3*: Implicit identification will uniquely and positively predict policy attitudes better when explicit identification with environmentalism is moderate compared to high or low (Hawkins & Nosek 2012). This is a weak prediction because we did not find support for this hypothesis in Study 1. Policy opinions will be regressed simultaneously onto dummy explicit identification and implicit identification. *H4a*: Implicit identification will be positively predicted from contact with environmentalists (friends, family, co-workers). Implicit environmentalism will be regressed onto contact with environmentalists. *H4b*: Implicit identification will be positively predicted from early nature exposure (museum, summer camp, pets, etc.). Implicit environmentalism will be regressed onto childhood nature exposure. *H4c*: if either H4 holds, then mediation is expected (H4 var) → I → (civic behavior and separately policy attitudes). Four mediation models will be tested through multiple regression for direct, indirect, and total paths for the effect of (peer environmentalism OR nature exposure) on (civic behaviors OR policy attitudes) through implicit identification with environmentalists. **Analysis Plan - Exploratory analyses** Correlate rural/urban location with identification (3 vars) and political orientation (2 vars). Does rural/urban location add unique variance to the prediction of civic behavior or policy attitudes using regression above and beyond the other predictors? If the three composites in H4 for peer environmentalism (friend, co-worker, family) are not reliable (Cronbach's alpha >.60), exploratory analyses will be used on various combinations to work towards scoring a better index of peer environmentalism. Correlate self-nature overlap and self-environmentalist overlap with explicit and implicit identification, and use various combination to predict civic behavior and policy attitudes, to learn more about shared method variance and the underlying constructs. Demographics such as age, gender, and ethnicity will also be correlated with study variables. *Analytic note*: Individuals who have ambivalent explicit identification with environmentalists may show more implicit-explicit divergence. When an explicit attitude is less important and less ingrained, there can be more noise in implicit attitudes, leading to less correspondence with explicit reports. We assume this will generalize to the current measures of identity. Tests for heteroscedasticity (the circumstance in which the variability of a variable is unequal across the range of values of a second variable that predicts it) will reveal if this emerges in the current study for identity. Heteroscedasticity is not expected because it was not seen in Study 1. **Analysis Plan - General Notes, Scoring, and Composites.** Alpha = .05. Continuous predictors will be standardized (z-score) before analyses. *Note*. The ST-IAT is counterbalanced by random assignment: half of the participants get Self + Environmentalist first, and the other half will get Other + Environmentalist first. This random condition will be used as a covariate in the regressions for H2ab, H3, and H4abc to reduce noise in the implicit ratings of identification. Explicit identification with environmentalism (4 items) will be combined into a composite. Moderate explicit identification will be defined as 3.5-4.5 on the composite of the (1-7) 4-item scale. Moderate vs. non-moderate explicit identifiers will be separated using a dummy variable (0,1). The ST-IAT will be scored with the IAT D algorithm adapted for the ST-IAT (Greenwald, Nosek, & Banaji, 2003). Civic environmental behaviors will be combined into a composite for frequency. If Cronbach's alpha < .60, only the final item will be used as the DV for the hypotheses ("Consciously made time..."). Environmental policy opinions will be combined into a composite for support. If Cronbach's alpha < .60, the "carbon pollution tax" item will be used alone as a DV for the hypotheses. Childhood nature exposure will be combined into a composite if Cronbach's alpha across all items > .60. Otherwise, only the summer camp item will be used. Peer environmentalism will be calculated by z-scoring the six variables and then multiplying within each of the three categories by exposure: (friend environmentalism X contact with these friends), etc., to yield three environmentalismXexposure composites. Then, these composites will be combined into a superordinate composite if Cronbach's alpha > .60, or analyzed separately if not. We will conduct a t-test to examine whether the counterbalancing of explicit identification and the IOS questions influences Identification, IOS-env, or IOS-nat. We will also conduct regressions predicting each of the three variables from counterbalancing order, one of the other two variables, and their interaction. These regressions will tell us whether counterbalancing order influenced the relationship between these three variables. If any of these tests produce significant results, we will include counterbalancing order in all analyses where it is relevant to control for that variability.