Theme: Weather and seasons

Measures
Emotional valence.
Participants continuously reported how they felt using a Response Meter (ADInstruments, New Zealand) with a scale ranging from 1 ("extremely negative") to 10 ("extremely positive"). Above the numeric scale, we provided a negative-positive valence graphical scale modeled after the self-assessment manikin. A similar approach was employed in previous studies of the impact of time rhythms and weather of affect . The data was recorded with Powerlab and processed with LabChart 8.19 software (ADInstruments, New Zealand). Participants continuously reported their affect, while waiting for the five minutes without doing any unnecessary actions (resting baseline). We calculated the mean affective valence from the last two minutes of baseline to account for the part of the baseline that was the most proximal to the study and to limit the influence of interaction with the experimenter on affect. Electronic rating scales collect reliable and valid emotion ratings
Weather data.
We used weather data from the weather station in Poznan, collected by the Polish Institute of Meteorology and Water Management. The weather variables were matched to the experimental data of the participants by date and hour. We examined the impact of the following indicators: ambient temperature, air pressure, humidity, cloud cover, precipitation, wind speed, sunshine duration. Presents means and standard deviations for weather conditions per season, day of the week, and part of the day.
Role of seasons, part of the day, and weather conditions in affective valence.
We found a positive direct effect of spring on valence and a negative indirect effect of spring on valence via ambient temperature β = -.20, 95% CI [-.31, -.094]. These two opposing effects canceled each other out, producing a non-significant total effect of spring on valence, β = .21, 95% CI [-.12, .54]. This decomposition of the total effect suggests that participants would feel generally better in spring than in winter if not adverse effects of higher temperatures in spring. Yet, given their joint influence, the effect of spring on valence was non-significant. We found no difference in affective valence between parts of the day.
Of the weather conditions, only the ambient temperature predicted the participants’ affective valence. Participants felt better when it was cooler outside. The unstandardized estimate was b = -.03, showing that a decrease of one degree Celsius predicted an increase in an individual’s affect by 0.03 points on the scale from one to ten, an equivalent of a 3.32% valence SD. To further support our findings, we run an exploratory analysis, in which we tested the model for each season separately. We found that people felt better when it was colder outside in spring β = -.27, 95% CI [-.40, -.14]. and in autumn β = -.12, 95% CI [-.24, -.01]. The relationship in winter was not significant β = -.005, 95% CI [-.09, .08].

Seasons and weather conditions.

R elative to winter, the ambient temperature was higher in spring but not in autumn. The wind speed was higher in spring and in autumn than in winter. The cloudiness was higher, and the sunshine duration was lower in autumn when compared to winter. In contrast, the cloudiness was lower, and the sunshine duration was higher in spring when compared to winter. We found no differences between the seasons in precipitation, air pressure, and humidity.

Part of the day and weather conditions.

The early mornings were more humid than late mornings, early afternoons, late afternoons, and early evenings. The precipitation was higher in early afternoons, late afternoons, and early evenings when compared to early mornings. The sunshine duration was shorter in the early evenings than in the early mornings. We found no differences in temperature, air pressure, cloudiness, and wind speed between the parts of the day.
Discussion
We examined whether individuals who start a laboratory experiment report different levels of affect depending on contextual factors such as season, part of a day, and weather conditions. We found that research participants felt better when it was colder outside. However, this effect had negligible practical meaning. Any differences in reported affect would vary within one standard deviation as long as the differences in the temperature between participants on different occasions were below 30 Celsius degrees. We found that participant’s baseline affect did not depend on any other conditions. Thus, we conclude that differences in season, weather, and time of day have little impact on baseline affect among participants for most laboratory research schedules. As our study had a reasonable sample size resulting in high statistical power, we believe that the null results are robust. Our work corresponds well with other large-scale integrative projects that indicated the non-significance of occasion-specific factors in their effect on research participants’ characteristics.
Our findings support other research indicating that high ambient temperature is associated with lower positive affect. Some studies suggested the opposite, yet they did not consider the seasonal variations. To address the fact that weather is often nested in seasons, we built a multilevel model that accounted for more variance. If we did not include seasons in our analysis, we found a positive correlation between temperature and affect, which might suggest that individuals feel better on warmer days or in warmer seasons. If we we examined the association between temperature and affect within each season, we found that people felt better when it was cooler outside. This finding suggests that it is important to control for seasons when examining the association between weather and affect. However, the effect should be interpreted as small. Thus, temperature differences of as much as 30 Celsius degrees would not be likely to cause deviations from the affect among research participants of more than one standard deviation of the mean valence.
We found that the relationship between seasonal variation and affect was complex. First, individuals felt more positive affect in spring than in winter. Yet, at the same time, springs were much warmer than winters, and participants felt worse on days that were warmer. Consequently, these two effects operated together in opposing directions canceling each other out. This effect is puzzling because simple correlations indicated that the ambient temperature and spring (vs. winter) were positively related to affect. We suggest that the outcomes are best interpreted as avoidance of thermal discomfort related to high temperatures in spring and autumn in our region and low temperatures in winter. Individuals might feel worse during spring and autumn heat, but they also might feel somewhat worse during the winter cold. Furthermore, our findings may suggest that other factors differentiate between the seasons that were not included in the analysis but might have influenced participants’ affect.
Unexpectedly, we did not find influences of daily cycles on affect, when controlled for the weather conditions. Previous studies indicated that the circadian rhythm of affect was consistent with the standard work-rest pattern. In our study, participants could schedule the lab visit before, in between, after the work, due to their own preferences, which may indicate a non-standard work-rest pattern. Future studies could replicate our result with a more homogenous participants pool to account for the work-rest cycles.