Toward a unified understanding of climate anxiety: Examining measures of climate and eco-anxiety

PLOS One | https://doi.org/10.1371/journal.pone.0339562 December 29, 2025 7 / 19 present sample, the readiness factor was reliable, with ⍵ = .708. The specific factors had acceptable reliability for political participation (⍵ = .637) and low reliabilities for individual behavior (⍵ = .176) and policy acceptance (⍵ = .251). Given that specific factors in a bifactor model are orthogonal to all other factors and that shared variance is explained by the general readiness factor, this lower reliability is to be expected [41]. In addition, in latent level analyses low reliabilities do not limit the interpretability of factors. Climate anxiety. To measure climate anxiety, we used the German version of the Climate Anxiety Scale (CAS; [9]). After reading the instruction “Below you can see some statements about the emotional consequences of climate change. Please indicate to what extent the following statements apply to you personally.” participants rated their levels of impairment across various climate anxiety symptoms. For example, items from the German factor structure included behavioral symptoms (“I find myself crying because of climate change”) and cognitive consequences (“I think, ‘why can’t I handle climate change better?’”). Responses were provided on a seven-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree). Notably, the German response scale used differs from the five-point Likert-type scale used in Clayton and Karazsia’s [6] original version.

As discussed above, factor structure varies across languages and studies. Therefore, we will test competing mea- surement models based on the literature. The reliability of the best-fitting model of climate anxiety in this study, a g-factor model, was high (⍵ = .958). Eco-anxiety. The German version of the Hogg Eco-Anxiety Scale (HEAS [7]) was used to measure eco-anxiety. The instruction of the scale reads: “Over the last 2 weeks, how often have you been bothered by the following problems, when thinking about climate change and other global environmental conditions (e.g., global warming, ecological degradation, resource depletion, species extinction, ozone hole, pollution of the oceans, deforestation)?“ The scale consists of 13 items covering four symptom dimensions: affective symptoms (“Feeling nervous, anxious or on edge”), rumination (“Unable to stop thinking about future climate change and other global environmental problems”), behavioral symptoms (“Difficulty sleeping”), and anxiety about personal impact (“Feeling anxious about the impact of your personal behaviors on the earth”). Participants rated the frequency of these experiences over the past two weeks on a four-point scale (0 = not at all, 1 = several of the days, 2 = over half the days, 3 = nearly every day). The final higher-order measurement model for eco-anxiety with four first-order factors showed high reliability of the second-order factor, with ⍵ = .940. Neuroticism. Neuroticism was measured using nine items from the Trait-Self Description Inventory (TSDI [49]).

These items, a subset of neuroticism items from the TSDI, were heuristically selected to optimize construct validity [45]. Participants responded on a seven-point Likert-type scale ranging from 1 (strongly agree) to 7 (strongly disagree). An example item is “I’m often fearful that I will fail to reach my goals.” Due to extremely high facet loadings (up to λ = .826) on a general neuroticism factor, deviating from Olaru et al. [45], we modeled only a general factor of neuroticism indicated directly by the nine items—a model also used in other studies [e.g., 50]. This factor demonstrated high reliability (⍵ = .912). Statistical analysis For all constructs used in this manuscript, we tested measurement models and evaluated their fit. Model fit was deemed good with a comparative fit index (CFI) and Tucker–Lewis index (TLI) ≥.950, root mean square error of approximation (RMSEA) <.050, and standardized root mean square residual (SRMR) <.080, and acceptable with CFI and TLI ≥ .90, RMSEA < .08, and SRMR < .11 [51–53]. We first tested and compared competing measurement models for each construct separately. After identifying the best-fitting and most parsimonious model for the CAS and HEAS, we tested their overlap in joint measurement models. Given that our goal was to assess whether both scales capture the same overarching construct, we did not exclude poten- tially redundant items that measured similar content across the two scales beforehand.