Development and validation of the pandemic fatigue scale

The existence and nature of pandemic fatigue–defined as a gradually emerging subjective state of weariness and exhaustion from, and a general demotivation towards, following recommended health-protective behaviors, including keeping oneself informed during a pandemic–has been debated. Herein, we introduce the Pandemic Fatigue Scale and show how pandemic fatigue evolved during the COVID-19 pandemic, using data from one panel survey and two repeated cross-sectional surveys in Denmark and Germany (overall N = 34,582). We map the correlates of pandemic fatigue and show that pandemic fatigue is negatively related to people’s self-reported adherence to recommended health-protective behaviors. Manipulating the (de)motivational aspect of pandemic fatigue in a preregistered online experiment (N = 1584), we further show that pandemic fatigue negatively affects people’s intention to adhere to recommended health-protective behaviors. Combined, these findings provide evidence not only for the existence of pandemic fatigue, but also its psychological and behavioral associations.


Development of Public Adherence to Recommended Health-Protective Behaviors over Time
As observed elsewhere 1-3 , we find a significant decrease in people's tendency to adhere to physical distancing measures (βstandardized = -.32, t (15,

Exploratory Factor Analysis based on Polychoric Correlations
Acknowledging that treating ordinal data as continuous may introduce bias 4-6 , we conducted another exploratory factor analysis based on polychoric correlations rather than Pearson productmoment correlations. As for the exploratory factor analysis based on Pearson product-moment correlations, we verified the sampling adequacy of the data using the Kaiser-Meyer-Olkin (KMO) test 7 and found it to be acceptable (overall KMO = .93; all KMO values for individual items are > .86).
Bartlett's test of sphericity 8 further indicated that the item correlations were sufficiently large for conducting an exploratory factor analysis (X 2 (45) = 5442.45, p < .001). To determine the number of factors to extract, we once again considered the scree test 9 , Glorfeld's modified parallel analysis 10,11 , the very simple structure criterion 12 , and the Velicer's minimum average partial criterion 13 , which combined indicated that either a one-, two-, or three-factors solution would best reflect the data ( Figure S21 and Table S19).
Considering a one-, two-, and three-factor solution, the exploratory factor analysis revealed that a two-factor model fit the data better (RMSR = .02, RMSEA = .07, TLI = .97) than a one-factor model (RMSR = .07, RMSEA = .15; TLI = .83), as indicated by the RMSR being closer to zero 14 , a difference in RMSEA > .015 15 , and a TLI above .95 16 . A three-factor model also fit the data well (RMSR = .01, RMSEA = .04, TLI = .99), but not much better than the two-factor model. Thus, among the three models considered, a two-factor model is arguably superior as it strikes a better balance between model fit and model parsimony 14 . Of note, the two-factor model explained 60.99% of the variance, with the first factor (information fatigue) accounting for 21.00% of the variance and the second factor (behavioral fatigue) 39.99%. of interest to be statistically significant, even after controlling for participants' age, gender, education, and cognitive risk perceptions regarding COVID-19. That is, controlling for these factors in an ordinary least square regression analysis, we still find that participants in the high pandemic fatigue condition expressed weaker intentions to adhere to the four health-protective behaviors in question as compared to participants in both the low pandemic fatigue condition (β = .29, t(1,550) = 4.31, ptwo-tailed Bonferroniadjusted < .001, 95% CI [. 16, .43]) and the control condition (β = . 19 Note. Figure S1 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares and mixedmodels regressions with data from the Danish and German repeated cross-sectional surveys and the Danish panel survey. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. Note. For all correlations n ≤ 15,318 and n ≥ 14,388. Gender (male = 1, female = 0); Education (10 years or more = 1, less than 10 years = 0); Employment (unemployed = 1, employed = 0); Chronic disease (no = 1, yes = 0). The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations.   Figure S3. Pairwise correlations for all variables considered in the German repeated cross-sectional survey.
Note. Figure S4 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S5 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S6 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S7 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys.  Figure S8. Self-reported public adherence to recommended health-protective behaviors over time in Denmark and Germany.
Note. Figure S8A, Figure S8B, and Figure S8C shows the mean level of self-reported public adherence to recommended health-protective behaviors for each wave of the Danish repeated cross-sectional survey, the German repeated cross-sectional survey, and the Danish panel survey, respectively, together with ordinary least square regression lines with 95% confidence intervals.
Mask wearing

Mask wearing
Information seeking Information seeking n = 360 380 400 420 Mask wearing Note. Figure S9 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the Danish and German repeated cross-sectional surveys. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations. Note. Figure S10 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the Danish and German repeated cross-sectional surveys. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations. Note. Figure S11 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), Time (survey wave) 2 , new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S12 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), Time (survey wave) 2 , new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S13 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the Danish and German repeated cross-sectional surveys. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations.   Figure S14. OLS regressions predicting hygiene in Denmark and Germany.

Mask wearing
Note. Figure S14 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the Danish and German repeated cross-sectional surveys. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations. Note. Figure S15 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the Danish and German repeated cross-sectional surveys. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations. Note. Figure S16 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the Danish and German repeated cross-sectional surveys. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations. Note. Figure S17 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S18 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S19 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S20 shows estimated β coefficients with 95% confidence intervals based on mixed-model regressions with data from the Danish panel survey. Continuous time-invariant predictors as well as continuous time-varying contextual predictors (i.e., Time (survey wave), new cases per million, new deaths per million, reproduction rate, and policy stringency index) have been mean-centered. All other time-varying predictors have been centered using the person-mean centering approach to disaggregate the within-(WSE) and between-subjects effects (BSE) of these factors 21,22 . The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the surveys. Note. Figure S21A shows the results from a scree test 9 based on Pearson's product-moment correlations. Figure S21B shows the results from Glorfeld's modified parallel analysis 10,11 based on Pearson's product-moment correlations. Figure S21C shows the results from a scree test 9 based on polychoric correlations. Figure S21D shows the results from Glorfeld's modified parallel analysis 10,11 based on polychoric correlations. Abbreviations: Eigenvalue (Ev).  Figure S22. OLS regressions predicting adherence to health-protective behaviors in the U.S.
Note. Figure S22 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the experiment. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the experiment. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations.  Figure S23. OLS regressions predicting adherence to health-protective behaviors in the U.S.
Note. Figure S23 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the experiment. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the experiment. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations.  Figure S24. Two-factor and second-order models of pandemic fatigue fitted with robust diagonally weighted least squares estimation in Denmark and Germany.
Note. Figure 24A and Figure 24B show the two-factor model of pandemic fatigue with fully standardized factor loadings and (residual) variances for Denmark and Germany, respectively. Figure 24C and Figure 24D show the second-order model of pandemic fatigue with fully standardized factor loadings and (residual) variances for Denmark and Germany, respectively. All models were estimated using robust diagonally weighted least squares estimation 18 Figure S25. Two-factor and second-order models of pandemic fatigue fitted with robust maximum likelihood estimation and robust diagonally weighted least squares estimation in the U.S.
Note. Figure 25A and Figure 25B show the two-factor and second-order model of pandemic fatigue with fully standardized factor loadings and (residual) variances for the U.S. experimental sample fitted using robust maximum likelihood estimation with robust standard errors and a Satorra-Bentler scaled test statistic 23 . Figure 25C and Figure 25D show the two-factor and second-order model of pandemic fatigue with fully standardized factor loadings and (residual) variances for the U.S. experimental sample fitted using robust diagonally weighted least squares estimation 18 Figure S26. OLS regressions predicting pandemic fatigue in the U.S.
Note. Figure S26 shows standardized β coefficients with 95% confidence intervals based on ordinary least squares regressions with data from the experiment. All continuous predictors have been mean-centered and scaled by 1 standard deviation. The p-values have not been adjusted for multiple comparisons and are presented as follows: *** ptwo-tailed < .001; ** ptwo-tailed < 0.01; * ptwo-tailed < 0.05. Exact p-values for all models are presented in the R-output which has been deposited on the Open Science Framework at: http://dx.doi.org/10.17605/OSF.IO/XD463. The gender variable refers to participants self-identified gender as presented to them in the experiment. Participants who did not identify as either male or female are not included in the analyses due to an insufficient number of observations.  Note. Standard deviation (SD). Participants' cognitive risk perception was estimated by taking the product of one item measuring participants' assessment of how serious it would be for them to get infected and one item asking them to judge their own likelihood of contracting the virus. Note. Standard deviation (SD). Participants' cognitive risk perception was estimated by taking the product of one item measuring participants' assessment of how serious it would be for them to get infected and one item asking them to judge their own likelihood of contracting the virus. Note. Standard deviation (SD). Participants' cognitive risk perception was estimated by taking the product of one item measuring participants' assessment of how serious it would be for them to get infected and one item asking them to judge their own likelihood of contracting the virus. I am tired of all the COVID-19 discussions in TV shows, newspapers, and radio programs, etc.
(1) Strongly disagree - (1) Never -(5) Always Information seeking Please indicate the extent to which you disagree or agree with the following statements.
I regularly seek out information on the current COVID-19 situation.
(1) Strongly disagree -(7) Strongly agree I try to stay updated on the current COVID-19 restrictions.
I often read, listen to, or watch news about COVID-19.
I closely follow the announcements from the government and/or the health authorities concerning COVID-19.
I spend a considerable amount of time learning more about COVID-19.
Negative affect Please answer the following questions.
(1) Not at all - (1) Never -(7) Very often (1) Never -(5) Always Information seeking Please indicate the extent to which you disagree or agree with the following statements.
I regularly seek out information on the current COVID-19 situation.
(1) Strongly disagree -(7) Strongly agree I try to stay updated on the current COVID-19 restrictions.
I often read, listen to, or watch news about COVID-19.
I closely follow the announcements from the government and/or the health authorities concerning COVID-19.
I spend a considerable amount of time learning more about COVID-19.
Negative affect Please answer the following questions.
(1) Not at all -   Note. Standard deviation (SD). Participants' cognitive risk perception was estimated by taking the product of one item measuring participants' assessment of how serious it would be for them to get infected and one item asking them to judge their own likelihood of contracting the virus.  How likely do you think it is that you will be infected with the novel coronavirus (COVID-19)?
(1) Very unlikely -(7) Very likely How serious would it be for you if you contracted the novel coronavirus (COVID-19)?
(1) Not serious at all -(7) Very serious Pandemic fatigue On this page we kindly ask you to indicate how much you disagree or agree with the following statements.
I am tired of all the COVID-19 discussions in TV shows, newspapers, and radio programs, etc.
(1) Strongly disagree - Physical distancing On this page we kindly ask you to indicate how much you disagree or agree with the following statements.
Over the next two weeks I will avoid physical contacts and keep a safe distance to people outside my own household (1) Strongly disagree -(7) Strongly agree Hygiene On this page we kindly ask you to indicate how much you disagree or agree with the following statements.
Over the next two weeks I will wash my hands very often and thoroughly and/or use hand disinfectant frequently (1) Strongly disagree -(7) Strongly agree Mask wearing On this page we kindly ask you to indicate how much you disagree or agree with the following statements.
Over the next two weeks I will wear a face mask whenever I am inside and cannot keep a safe physical distance to people outside my own household (1) Strongly disagree -(7) Strongly agree

Information seeking
On this page we kindly ask you to indicate how much you disagree or agree with the following statements.
Over the next two weeks I will do everything I can to keep myself updated about the development of the pandemic, and stay informed about the current COVID-19 restrictions (1) Strongly disagree -(7) Strongly agree  Note. Response scale: 1 = strongly disagree, 2 = disagree, 3 = somewhat disagree, 4 = neutral/neither disagree nor agree, 5 = somewhat agree, 6 = agree, 7 = strongly agree. Abbreviations: Information fatigue (IF), behavioral fatigue (BF).