Psychological barriers moderate the attitude-behavior gap for climate change

Behavioral change has been increasingly recognized as a means for combating climate change. However, being concerned about climate problems and knowing the importance of individual actions in mitigating them is not enough for greater adherence to a more sustainable lifestyle. Psychological barriers such as (1) finding change unnecessary; (2) conflicting goals; (3) interpersonal relationships; (4) lack of knowledge; and (5) tokenism have been proposed as an explanation for the gap between environmental attitudes and actions. Yet, so far, this hypothesis has remained untested. This study aimed to assess if psychological barriers moderate the association between environmental attitudes and climate action. A sample of Portuguese individuals (N = 937) responded to a survey measuring climate change beliefs and environmental concerns as an index of environmental attitudes, a scale of self-reported frequency of environmental action, and finally, the dragons of inaction psychological barrier scale. Our participants revealed generally elevated positive environmental attitudes. These attitudes were positively and moderately related to greater self-reported frequency of environmental action in areas such as reusing materials, reduced consumption of animal products, water and energy saving, and airplane use, but not driving less. Critically, the association between attitudes and behavior was negatively moderated by psychological barriers for the reuse, food, and saving domains, but not for driving or flying. In conclusion, our results corroborate the assumption that psychological barriers can partly explain the attitude-behavior gap in the climate action domain.


Introduction
Global surface temperatures are rising at unprecedented levels due to an increase in the concentration of greenhouse gases produced by human activity [1,2]. This warming trend has propelled an increased frequency of extreme weather events, which has and will continue to have a significant impact on human physical [3,4] and mental health [5,6]. Despite the irreversibility of some of the consequences of climate change, drastically reducing anthropogenic greenhouse gas emissions is imperative to prevent global warming from reaching the 1.5˚C mark [7]. Reducing these emissions will require major changes in human behavior toward a more sustainable lifestyle [8][9][10][11]. The frequency with which people engage in pro-environmental behaviors is assumed to be related to their environmental attitudes [12][13][14][15] three possible valued objects: the self, other people, and all living beings. Believing that something valued is being threatened is an important aspect of developing favorable environmental attitudes, even when those concerns are at a personal level (egoistic) [41]. The present study assessed environmental attitudes by measuring climate change-related beliefs and environmental concerns. We predicted that believing in the negative impact of climate change paired with displaying environmental concerns for either the self, others, or nature should lead to favorable attitudes regarding climate change.

The attitude-behavior gap
Studies often report a moderate relationship between attitudes and pro-environmental behaviors [e.g., r = 0.36 in 12, r = 0.482 in 22]. Despite this relationship, a growing number of studies reports that being concerned or feeling responsible for acting in favor of the environment is not enough to increase pro-environmental behavior [12,14,18,[44][45][46][47]. This gap has been named the attitude-behavior gap or the value-action gap. Terms such as belief-action gap, knowledge-action gap, and attitude-behavior inconsistency have also been employed [18]. Critically, not all pro-environmental actions are similarly affected by attitudes [14]. The higher the perceived costs of an action, the lesser the effect attitudes have on actual behaviors [48]. This could be explained by environmentally concerned individuals feeling dissatisfied when their actions do not correspond to their values and attitudes, yet they also do not want to bear significant personal costs or discomfort in pursuing them. Accordingly, the low-cost hypothesis states that the lower the personal costs in a situation, the easier it is for actors to transform their attitudes into corresponding behavior [48,49]. The cost of an action usually depends on the efforts necessary to execute it; for example, recycling can increase in cost if no nearby structures allow for the correct waste disposal and people would need to walk or drive more to dispose of their waste [50]. Decreasing meat consumption [51][52][53] and reducing personal vehicle use [54][55][56], two of the behaviors with the highest carbon emissions, are usually perceived as highly costly behaviors because sustainable alternatives are seen as less convenient (e.g., unappealing vegetarian meal alternatives or uncomfortable public transportation). The weak association between attitudes and behavior could be partially explained by measurement inadequacies such as social desirability biases, temporal gaps, and faulty instruments [18]. The literature has also increasingly pointed to the relevance of assessing structural barriers (i.e., external systemic or infrastructure barriers) such as the accessibility and condition of sustainable alternatives, which can also hinder action-taking by aware and concerned individuals. For example, in the case of household carbon emissions, the cost of more sustainable items like organic and vegan food [57,58], a perceived lack of available time [45,59,60], and difficulty in accessing structures or products in some areas [61,62] have been pointed as barriers preventing concerned individuals from acting in accordance with their attitudes. However, even in situations where contextual and external factors do not heavily constrain individuals, the frequency of environmental action may still be low [14,16]. barriers were categorized into seven groups: Limited Cognition, Ideologies, Comparison with others, Sunk costs, Discredence, Perceived Risks, and Limited Behavior.
Gifford and Chen [17] evaluated the impact of some of these psychological barriers on intentions to engage in mitigative food choices (e.g., purchase of organically grown food). They observed that the dismissal of anthropogenic climate change evidence was the strongest perceived barrier to mitigation, the second being incompatible financial and time investments, followed by satisfaction with current mitigation behaviors. Gifford's psychological barriers were also negatively associated with self-reported energy conservation: the existence of other conflicting goals and aspirations had the strongest negative correlation with energy conservation behaviors, and interpersonal influences the weakest [64].
Each of these studies pointed to a different structure of psychological barriers. More recently, Lacroix et al. [19] developed a short scale to measure psychological barriers to proenvironmental behavior that could be applied across multiple domains of environmental behavior, the Dragons of Inaction Psychological Barriers (DIPB) scale. After a set of three studies, the final scale measured five barriers: (a) Change Unnecessary, that is, the denial of environmental problems and of the necessity to act; (b) Conflicting Goals and Aspirations, comprising limited time, sunk costs, and difficulty in changing habits; (c) Interpersonal Relationships, covering the fear of social disapproval or criticism; (d) Lacking Knowledge, representing not knowing how to change; and (e) Tokenism, the belief that no more personal investment is needed to change. The five barriers model was invariant for a wide range of behaviors that the authors considered as high-difficulty (e.g., eating less meat and driving less) and low-difficulty (e.g., recycling, saving water, saving energy, buying green); nonetheless, the impact of each barrier varied across behaviors. The interpersonal relationships barrier was the strongest for eating less meat, the conflicting goals and aspirations barrier was more substantial for driving less, and the lacking knowledge barrier was the strongest for buying green. Tokenism did not differ across behaviors. The DIPB scale was also adapted to the Colombian population, with a similar five-factor structure adequately describing the data [65]. Wang et al. [66] showed that the DIPB was associated with inaction and mediated the relation between positive emotions towards nature (awe) and climate change inaction. Here we adapted the DIPB scale for the Portuguese population to measure the five psychological barriers identified in these previous studies.

The current study
Psychological barriers have been pointed out as a possible explanation of inaction in individuals concerned about climate change but not effectively acting on their beliefs [16]. The DIPB scale was created to measure the effect of psychological barriers on concerned individuals under the assumption that they may help explain the attitude-behavior gap [19]. However, to the best of our knowledge, no study has assessed whether psychological barriers help explain this gap.
The current study aimed to test whether psychological barriers, measured by the DIPB scale, partially explain the gap between environmental attitudes and pro-environmental behaviors. Our goal was to test the model illustrated in Fig 1. Attitudes were measured by climate change beliefs and environmental values [12]. To include a wide range of pro-climatic behavior from more to less impactful, we measured the self-reported frequency of pro-environmental behaviors in the following domains: transportation, food, energy and water saving, waste management, and sustainable purchase. We expected to find that behaviors related to transportation and food would be less affected by attitudes and would be hindered more by psychological barriers than the lower-investment counterparts (waste management, purchase, and conservation).
We collected data in Portugal. Previous surveys have shown that the Portuguese are aware and concerned about climate change and the environment [30,67,68]. Nevertheless, the country has a low carbon-emission reduction performance [69], and the citizens report engaging mainly in low-impact behaviors [67]. As a member of the Organization for Economic Cooperation and Development (OECD) and the European Union, Portugal shares characteristics with similar countries [10], where citizens living in urban contexts have higher levels of education, more accessible public transportation systems and a better supply of low carbon products. These characteristics are usually associated with a reduction of structural barriers for environmental action, thus making Portugal a suitable context to study the attitude-behavior gap and the role of psychological barriers.

Participants
The study was distributed online between November 2021 and March 2022. We recruited 937 Portuguese-speaking individuals residing in Portugal via the University of Porto's mailing list, various online groups, and the promotion ad feature of Facebook and Instagram. Overall, 650 participants identified as women (69%), 280 as men (30%), and seven as other/no response (1%). Ages ranged from 18 to 80, with the average being 36 years (SD = 16). Most participants reported having some form of higher education degree (67%), 290 concluded high school (31%), and only 18 did not complete secondary education (2%). More than half (56.7%) resided in Portugal's two most populated and urbanized cities (Porto and Lisbon). Participants completed a written online informed consent form at the beginning of the study (i.e., they read and clicked on consent statements before moving on to the main study) and were debriefed at the end. If participants did not consent to the terms of participation, the study was ended before completion of the questionnaire. The Faculty of Psychology and Education Sciences ethics committee approved the research protocol (Ref.ª 2021/09-08).

Instruments
Environmental concerns. We used the Environmental Concern Items from Schultz [35], which has been referred to as a robust measurement of environmental concern [70]. It comprises 12 items, divided into four altruistic (e.g., "All people"), four egoistic (e.g., "My lifestyle"), and four biospheric (e.g., "Animals") domains of environmental concern. Participants rated the 12 items using a scale from 1 (not important) to 7 (supreme importance) in response to the question: "I am concerned about the environmental problems because of the consequences for. . ..". The instrument has good reliability and correlates with other measures of environmental attitude, self-reported pro-environmental behavior, and social value orientation [35,41,70]. Items are shown in Table 1.
Climate change beliefs. We measured climate change beliefs by adapting items from The Climate Change Attitude Survey scale [71]. The belief sub-scale comprises nine items rated on a Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The used items are presented in Table 1.
Frequency of pro-environmental behaviors. We built a questionnaire to measure the self-reported frequency of the most relevant pro-environmental behaviors in the following domains: transportation, food, energy and water saving, waste management, and sustainable purchase [11,67,72,73]. The set of 21 items (see Table 2) was rated regarding their frequency on a Likert scale from 1 (rarely) to 5 (very often). We asked participants to rate the frequency of each action at the moment of the scale completion (during the COVID-19 pandemic) and what they believe was the frequency of their behavior before the onset of the COVID-19 pandemic. We intended to assess if behaviors changed because of the ongoing pandemic situation (which could particularly affect transportation behaviors). Responses were similar for both questions; hence, we will only present the results regarding the current self-reported frequency of action. Psychological barriers. We used the 22-item DIPB scale [19] and two additional items that were removed from the original final scale (items 23 and 24) but were recommended to be assessed in contexts where the government plays an important role. We included these items to assess if they could bring relevant information to measuring psychological barriers in Portugal since the Portuguese tend to attribute to local and national government a high (larger than the European Union average) responsibility in fighting climate change [74]. Each item was answered on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree). All items are presented in Table 3. The DIPB scale assesses five psychological barriers: (a) Change Unnecessary, (b) Conflicting Goals and Aspirations, (c) Interpersonal Relationships, (d) Lacking knowledge, and (e) Tokenism. Before answering the scale, participants first selected one behavior from a list of five proenvironmental behaviors to serve as a reference while rating each item. The selected behavior should be one that participants believed was necessary to help the environment but which they were not currently doing or were not doing often enough. The behaviors were: (a) eating less meat; (b) taking public transport, cycling, or walking rather than driving; (c) reducing water use (e.g., taking shorter showers, repairing leaks); (d) making more eco-friendly purchases, and (e) recycling all items permitted by their local programs. Due to a programming error, one domain of environmental behavior formerly tested in the DIPB scale (i.e., wearing a sweater rather than turning up the heat in the winter) was not included in the list.

Procedure
We requested permission from the original authors of the scales to use and adapt them in the present study. Our first step was to translate all items into Portuguese. We focused our translation on capturing relevant concepts to measure our research questions [75,76]. All three authors translated all scales independently and then discussed the three translations to settle on a final version of each item. Afterward, an environmental science major assessed the translation and deemed the terminology suitable. The translated items used in the present study are available on our OSF page: https://osf.io/j9th2/ Completion of the questionnaire took, on average, 15-min. Participants completed the scales in the following order: (1 st ) environmental concern, (2 nd ) climate change beliefs, (3 rd ) frequency of ecological behaviors, (4 th ) the psychological barriers scale, and (5 th ) demographics questions.

Data analysis
We performed confirmatory factor analysis to evaluate the structure of the responses to the environmental concerns, climate change beliefs, and psychological barriers scales. For the frequency of pro-environmental behavior, we first ran an exploratory factor analysis to determine the structure of the measurement of these behaviors, followed by a confirmatory factor analysis with the final identified model. All analyses were performed using R. 4.1.3 [77]. We used the lavaan package [78] for the confirmatory factor analysis, and the package psych [79] for the exploratory factor analyses and the reliability of our measurements. Finally, we used the base R Stats package to compute the moderation analysis using the lm function [77].
We first tested the multivariate normality of our data using the Energy test and Mardia's multivariate skewness and kurtosis tests. Given that our data was ordinal, non-normal, and the sample size was large, we employed a diagonally weighted least squares (DWLS) estimation based on a polychoric correlation matrix. This method produces more accurate factor loadings estimates than the more commonly used maximum likelihood estimation method [80,81].
To assess model adequacy, we used the following model fit indices: chi-square goodness-offit statistic (χ2), the root-mean-square error of approximation (RMSEA), Bentler's comparative fit index (CFI), and the standardized root-mean-square residual (SRMR). For χ2, a small, non-significant value indicates a good fit. However, χ2 is known to be affected by sample size. RSMEA smaller than 0.06 indicates a good fit, and of 0.08 or less, a reasonable fit. For the CFI, values higher than 0.95 indicate a good fit, and values between 0.90 and 0.95 indicate an adequate fit. Finally, for the SRMR, a value below 0.08 indicates an acceptable fit. The cutoff criteria were based on Hu and Bertler [82]. Note that these cutoff criteria are not static and should be weighted considering the data since the index values can be influenced by sample size, the number of variables analyzed, and missing data [83].
Finally, we calculated the reliability of each of our measurements. Since our data is ordinal and non-normal, we report ordinal alpha (α), ordinal omega total (ωt), and Guttman's greatest lower bound (GLB) [84].
All the materials, data, and analysis scripts used are publicly available at: https://osf.io/j9th2/  Table 2 shows these values for the Frequency of Pro-Environmental Behavior, and Table 3 for the Psychological Barrier scales. The item numbering represents the order in which items were presented in the scale for rating. For the Psychological Barrier scale, we present the item order used in this study and the scale's original numbering. We described each item's mean (M) and standard deviation (SD). Ordinal alpha (α) for each initial factor (before item removal) is presented after the factor's name and a column with the alpha value after removing different items inside that factor. For the factors with only two items, we presented the item's correlation.

Environmental attitudes
Our goal was to create an environmental attitude second-order latent factor consisting of a factor of environmental concerns and a factor of climate change beliefs. We first tested each of these constructs independently, with one model for responses in the environmental concerns scale, and another for the climate change beliefs scale. Regarding environmental concerns, participants reported higher levels of biospheric concerns (M = 6.33, SD = 1.52), followed by altruistic (M = 6.26, SD = 1.11), and lastly egoistic concerns (M = 5.56, SD = 1.15), replicating prior results [35]. CFA revealed that the 3-factor solution proposed by Schultz [35] had an adequate fit, χ2 (51) [70], we retained this factor in our model since it positively correlated with the other concern factors and loaded on the second-order latent variable of Attitudes [35,41].

Environmental behaviors
We assessed the self-reported frequency of environmental behaviors in six domains: transportation, food, energy and water saving, waste management, and sustainable purchase. To describe the relationship between these behaviors, we performed an exploratory factor analysis. We used principal axis factor analysis to extract factors for its relative tolerance of nonnormality and a Promax rotation since the factors are assumed to be correlated [85]. A scree plot indicated that five factors should be retained [86]. Items were assigned to the factors where they had the strongest loadings. Items with loadings below 0.4 were removed, which led to the removal of the following items: 3,6,8,9,19, and 20. After removal of these items, no loadings were salient on more than one factor. Fig 3 presents the five-factor solution obtained. Factor 1 was named Reuse since items loading on it were related to sustainable waste management and investment in local products. Factor 2 was called Driving because items loading on this factor refer to reducing driving habits.

Psychological barriers
We tested if we could replicate in our sample the five-factor structure found in Lacroix et al. [19]. 066. Yet, there was still an item (item 21) that showed a standardized loading of 0.265 which is below the criterion of 0.3. Since we are still left with four items in the tokenism factor, we decided to remove this low loading item, and its removal further improved the Tokenism factor's reliability (α = 0.85). All We also tested a second-order factor model where each of the five barriers loaded on a general psychological barrier factor. The model slightly misfitted the data, as indicated by the larger SRMR index (> 0.08), χ2(184) = 659.171, CFI = 0.928, RMSEA = 0.053 [0.067, 0.77], SRMR = 0.085. The modification indices indicated that model fit could be vastly improved if we considered the covariance between the Change Unnecessary and the Tokenism factors. The high correlation values between these two factors suggest a strong association. The DIPB scale measurement of Tokenism implies that current efforts are sufficient, and change is not necessary, which overlaps with the content of the Change Unnecessary factor and may explain some specific variance related to these factors. By including the covariance between Change Unnecessary and Tokenism, model fit improved, χ2(183) = 516.954, CFI = 0.956,

Moderation analysis
Finally, our last aim was to assess if the psychological barriers moderated the influence of attitudes on the frequency of pro-environmental behaviors, thereby partly explaining the attitudebehavior gap. We performed a latent moderation analysis using a factor score procedure [87]. Although this approach neglects measurement error, given our sample size (N = 937) estimation error for the correlations in our sample should be low [88]. Simulations also show that this method performs relatively well compared to other methods even under suboptimal conditions, such as with lower reliabilities, null to medium correlations between the predictor and the moderator, and unequal factor loadings [87].
The latent moderation analysis involved three steps. First, we independently estimated the individual latent scores for each pro-environmental behavior factor separately using DWLS estimation. The models were saturated, and hence the fit was perfect. These latent scores served as our latent dependent variable in the moderation analysis. Second, we estimated individual factor scores for our latent predictors, namely Attitudes and Psychological Barriers, in a single model that allowed them to correlate. When estimating the model of Attitudes and Psychological Barriers together, the model did not converge if we included the correlation between the Change Unnecessary and Tokenism factors. Since the model without this We entered the factor scores in a linear regression model of the following form: YðBehaviorÞ ¼ XðConcernÞ þ ZðBarriersÞ þ XZ Table 4 contains the statistics for each regression model. The confidence intervals of 95% (CI) for each of the coefficients was obtained via bootstrapping with 1000 replications. F-statistic and adjusted R 2 for each of the regression models was as follows, Reuse: 003, R 2 = 0.01. Attitudes were a significant positive predictor of the frequency of all environmental behaviors except for Driving. Psychological Barriers were found to be a significant moderator for the relationship between Attitudes and the frequency of behaviors related to Reuse, Food, and Saving. Psychological Barriers did not significantly moderate the relationship between attitudes and Flying and Driving.
We graphicly illustrated the moderation effect of Psychological Barriers for each Pro-Environmental Behavior in Fig 6. Linear predictions between Attitudes (independent variable) and Pro-Environmental Behavior (dependent variable) were computed for those with low, average, and high levels of Psychological Barriers (i.e., one SD below the mean, the mean, and one SD above the mean, respectively). For Reuse and Food, the effect of psychological barriers mattered most at high levels of environmental attitudes. For Saving, the pattern with the effects of psychological barriers reversed for high and low levels of environmental attitudes.

Discussion
In this study, we evaluated whether psychological barriers could partly explain the gap between environmental attitudes and engagement in different domains of environmental action. We assessed environmental attitudes as the latent factor representing the covariation of environmental concerns and climate change beliefs. We used this model to assess if favorable environmental attitudes relate to a higher frequency of climate action. Finally, we examined if psychological barriers, as assessed by the DIPB scale, moderated the relation between attitudes and behavior frequency in each domain of action. Next, we will first discuss our findings regarding the relationship between attitudes and behavior, followed by how psychological barriers may moderate it.

Environmental attitudes and behaviors
Our sample of Portuguese individuals was highly concerned about climate change and believed in the human impact on this phenomenon. Climate change beliefs were very high on all levels, and environmental concerns were higher in the biospheric and altruistic domains than the egoistic one. Previous studies indicated that biospheric and altruistic values tend to be the most correlated with the intention to act [32,40], yet all concerns are usually correlated to each other [35,[41][42][43]. Accordingly, we were able to extract a single factor reflecting pro-environmental attitudes from the measures of climate change beliefs and environmental concerns. The frequency of pro-environmental behaviors showed some variability. Behaviors with lower carbon impact, such as reducing water use, conserving energy, and recycling or using less plastic, were done more frequently than reducing car use, airplane traveling, and eating less meat or animal products. These results align with the idea that high carbon impact behaviors require more investment to change [10,48].
Positive environmental attitudes predicted a higher frequency of environmental behavior across all our tested domains except for behaviors related to driving less. However, it is worth pointing out that the association between attitudes and self-reported frequency was not very high (estimates ranged between .025 and .381), which is consistent with previous reports of a modest association between environmental attitudes and behavior [12].

Psychological barriers
The DIPB scale was created to evaluate psychological barriers across different environmental behavior domains and was originally developed and tested in a Canadian sample [19]. We were able to replicate the original 5-factor structure in our Portuguese sample with some minor model adaptations. This pattern is similar to the one obtained in a Colombian sample [65]. The short format of the DIPB scale and the generality of the 5-factor solution extracted from it suggests that the scale is useful to, briefly and efficiently, measure psychological barriers to pro-environmental action in diverse cultural contexts.
In our sample, change unnecessary and interpersonal relationships were overall the lowestrated barriers, while tokenism, lacking knowledge, and conflicting goals were the strongest barriers. It is worth noting that the items concerning the perceived government's duty in facilitating action (items 23 and 24) were removed due to low model adequacy, but these were the highest-rated barriers to pro-environmental behavior in our sample. We believe the DIPB scale did not capture this phenomenon well, but the dismissal of personal responsibility might be a promising psychological barrier to be further explored in the future. An Australian study [89] showed that attributing greater responsibility to the government for environmental protection was related to more negative environmental intentions and behavior than attributing responsibility to the community. Although the measurement of this barrier through items 23 and 24 did not work well within the context of the DIPB scale, it does not mean that this barrier is not a relevant determinant of environmental action. Future studies should consider how to measure this psychological barrier effectively.
In our study, the psychological barriers moderated the effect of attitudes on behavior domains like waste management (reuse factor), eating less meat (food factor), and conservation behaviors (saving factor). Psychological barriers had a stronger influence on individuals with a higher environmental positive attitude, indicating that they may help explain the attitude-behavior gap that prevents concerned individuals from effective action. This relation was clear for the reuse and food factors but less so for the saving factor. Critically, this moderation was not significant in the transportation domains (car usage and plane traveling). The moderation effects observed here align with our predictions that psychological barriers may partly explain the attitude-behavior gap in some behaviors but not others. The transportation and food domains are considered of high difficulty to change [19]. Nevertheless, against our expectations, we only observed a moderation effect for food choices, but not driving or flying. This may suggest that although both behaviors are of high difficulty and have a high impact on the carbon footprint, they cannot be explained in the same way.
On the one hand, in our analysis, food habits were explained by attitudes with psychological barriers moderating that relationship. In the literature, meat-eating behavior interventions appear often associated with values and attitudes [90]. For example, self-transcending values effectively increase negative attitudes toward meat consumption [91], and attitudes about animals are important in choosing vegetarianism [92]. Furthermore, lack of knowledge (which is one of the assessed psychological barriers in our study) usually acts as a barrier to decreasing meat intake; and interventions focused on increasing awareness coupled with knowledge on how to replace meat with less impactful food choices increase positive attitudes and intentions towards meat reduction [90,93].
On the other hand, our analysis indicates that transportation behaviors (i.e., flying and driving) were less explained by attitudes, and no moderation effect was observed. The willingness to reduce flying was positively associated with attitudes, yet this behavior was not moderated by psychological barriers. Unfortunately, we recorded no information about the flying habits of our sample. It has been observed that habitual plane traveling is associated with a higher resistance to reducing flying in the future compared to those who air travel only occasionally [94] or prefer other means of transportation [95]. Flight reduction is also associated with how accessible and efficient the alternatives are (e.g., train) [94]. Hence it is possible that our results could be explained by the air travel habits of our sample. Future studies should control the frequency of air travel to assess if psychological barriers play different roles depending on how much people use this mode of transportation. Driving was the only domain not predicted by climate change attitudes. This result might point to other stronger motives behind driving habits that are not covered by our scales. For example, Innocenti et al. [96] found that people are biased against public transportation, preferring to travel by car even when the costs of this option are higher. A possible explanation is that cars and driving symbolize freedom, independence, status, and a pleasurable activity [97]. Other factors, such as travel time, which is longer for public transportation, are also crucial when choosing a traveling mode [98]. As such, people with complex activity patterns (e.g., taking children to school before going to work) might prefer to avoid the rigid and unreliable public transportation system, preferring to travel by personal transportation, which can be either bike or car. Note that bike traveling is also limited by the existence of safe and adequate bike lanes [99]. As for flying, we have not assessed the preferred transportation modes of our sample. Hence, we could not test whether psychological barriers vary depending on the driving habits of our participants. Future studies should consider measuring this variable.
A meta-analysis on behavioral interventions to promote household action on climate change pointed out that interventions on reducing meat eating tend to be more effective than reducing private car usage [100]. Interventions on meat reduction usually center around removing external barriers/facilitating mitigation behavior, while transportation interventions revolve mostly around information-based interventions, increasing the appeal of not using cars. These studies indicate that while food mitigation behaviors can be mostly linked with the variables assessed by our study, driving and flying habits might be associated with social status, previous behaviors, and practicality over alternative methods (i.e., public transport, walking, riding a bike), which were not evaluated here and may well explain why we could not effectively capture the motives behind driving and flying habits.

Limitations and future directions
Collecting data via a survey has its limitations. Self-reported measures of pro-environmental behavior might not entirely reflect authentic environmental action as it may depend on the participant's interpretation of the question and recollection of their activity [101] or be influenced by social desirability [102]. It remains, therefore, a challenge to objectively monitor the frequency of pro-environmental action. Nonetheless, we forward some promising alternative methods. Laboratory observations through experimental tasks do not rely as much on interpretation and offer experimenters higher degrees of control [101]. Laboratory tasks usually rely on assessing participants' decisions in different manipulated scenarios. For example, the Pro-Environmental Behavior Task [103] requires participants to make several trips. On each trip, they can choose between an environmentally friendly or unfriendly mode of transportation with manipulated waiting periods. Another way to curb recollection and interpretation restrictions is through momentary ecological assessment using mobile devices to register behaviors while they are happening [104].
There may be some limitations of generalizability related to our sample demographics. We collected data in Portugal from relatively young, mostly female, urban, and digitally active individuals (since this study was advertised via email and social media). Our results therefore may not generalize to individuals with different demographic characteristics. We note, however, that we replicated with our sample the main results observed in other studies with North-American and Latin-American samples [19,65]; hence our findings are likely to generalize across cultures.
This study showed that the DIPB scale has some degree of explanatory power toward the existence of a gap between environmental attitudes and (self-reported) ecological behavior. However, future studies could integrate these barriers into a broader model framework, such as the comprehensive action determination model [26]. The effects of habits and perceived behavioral control were shown to have a direct relationship with pro-environmental behavior [12], and their inclusion in this model framework might add explanatory power to the attitude-behavior gap.
Furthermore, we lacked a measure of structural barriers. Future studies should consider including a measure of structural and external barriers to better separate the influence of psychological barriers from other factors that may prevent people from taking more action. A lack of accessibility to infrastructures might be increasing behavior costs. For example, enhanced accessibility to recycling facilities encourages people to recycle [105]. Low availability of public transportation can be responsible for increased car use in rural areas [61], while product prices impact what people eat [57]. Hence, the accessibility of some structures or products may help explain the lack of action from concerned individuals beyond intrinsic barriers.
A final note regarding the measurement of psychological barriers falls upon the necessity to develop more robust ways to study people's tendency to place responsibility for action on the government and industry, which was not well assessed by the DIPB scale but had a strong presence in our sample.

Conclusions
The present study enhanced our understanding of the relationship between attitudes, psychological barriers, and pro-climate behaviors. We demonstrated that not all types of behaviors are equally influenced by attitudes and the perception of psychological barriers. The DIPB scale could, to some degree, be a helpful tool to explain the attitude-behavior gap. This knowledge will help to plan more focused and efficient interventions which focus on breaking psychological barriers. Behaviors related to food choices, waste management, or water and energy saving can benefit from an intervention related to increasing attitudes while not ignoring the potential effects of psychological barriers. Finally, driving seems to be more resistant to change, and merely stimulating positive environmental attitudes or dismantling psychological barriers may not be enough.