The COVID-19 pandemic has disrupted higher education. It began with the large scale shift to online instruction in early 2020 and continued into subsequent academic years with modifications to nearly all aspects of functioning including instructional models, residential offerings, student enrollment, and institutional finances (June and Elias, 2021; Smalley, 2021). There have been similar disruptions to the subjective academic experience of college students. With increasing reports of anxiety, struggles to sustain attention, and difficulties staying motivated, academic engagement and performance have suffered (Cao et al., 2020; Dennon, 2021; Hicks et al., 2021).

Confirmatory factor analyses indicated acceptable fit for the 6-factor model at each timepoint (RMSEA = 0.064 to 0.076, CFI = 0.935 to 0.952, TLI = 0.916 to 0.938; see Table 1 for specific results). Tests of measurement invariance supported strong, strong partial, or strict invariance over time for each of the six constructs (see Table 1).

Of the total 476 students in the data (Cohort 1 = 206, Cohort 2 = 270), 55–71% completed each survey and 70–82% had GPA data available; Supplementary Table 1 displays the participant flow by cohort and wave. Analyses comparing students with complete data to students with any missing data indicated that students in the two cohorts were equally likely to have missing data, χ² (1) = 0.052, p = 0.82, that students of color and white students were equally likely to have missing data, χ² (1) = 0.143, p = 0.71, and that levels of T1 variables did not differ across students with missing versus complete data, Wilks’ λ (5, 465) = 0.987, p = 0.313. However, as is typically observed in survey research, men were more likely to have missing data than women, χ² (1) = 6.54, p = 0.011, and students with missing data had significantly lower grades during Year 2 than students with complete data, t(378.46) = 3.188, p = 0.002. Comparisons of demographic characteristics across the two cohorts indicated the two cohorts were similar in terms of proportions of women, χ² (1) = 1.597, p = 0.206, and people of color, χ² (1) = 1.773, p = 0.183.

Correlations and descriptive statistics (Table 2) revealed the expected patterns of correlations between variables. The autonomous motives (intrinsic, identified) were more positively and strongly correlated with one another than with the controlled motives (introjected, external), and the degree of agreement decreased as motive types became more distally related on the SDT continuum. There was also substantial agreement between repeated measures over time, with means for each construct showing patterns of stability rather than increases or declines.

Latent change score models for the two cohorts combined (Table 1 and Figure 1) indicated patterns of stability for all constructs (see Supplementary Figure 1). Specifically, initial levels ranged from low (M = 1.84) for amotivation to high (M = 4.14) for identified regulation, and all change score estimates were not significantly different from zero for all six constructs (Supplementary Figure 1; ΔT2T1 = −0.28 to 0.01; ΔT3T2 = −0.10 to 0.12). Intrinsic and identified remained relatively high, both the approach and avoidance forms of introjection remained at moderate levels, with introjected avoidance motivation being slightly higher than introjected approach motivation, and external regulation and amotivation stayed relatively low.

When cohort and grades (year 2 and year 4) were added to the model as predictors and outcomes of initial levels of motivation and change scores (see Figure 1), cohort significantly predicted the ΔT3T2 estimates for identified regulation and intrinsic motivation (b = −0.23 to −0.16, p < 0.05), but not initial levels or any other change estimates. Thus, as expected, the two cohorts showed similar trajectories during their first year of college, but students who completed their fourth year of college during the pandemic reported steeper declines in intrinsic motivation and identified regulation between the end of their first year and their final year as compared to the pre-pandemic cohort (see Figure 2). The expected difference between cohorts in trajectories of amotivation, however, was not observed.

Changes in identified and intrinsic motivation did not predict year 4 grades when controlling for year 2 grades. Thus, although the COVID cohort reported steeper declines in some forms of motivation, the steeper declines in motivation did not appear to be detrimental to their achievement levels.³

The present study compared motivational change across two largely identical cohorts differing only in whether their fourth year of college began before or during the COVID-19 pandemic. Not surprisingly, there were no differences across cohorts in any of the motivation types from T1 to T2 when conditions for the two groups were quite similar. Although the cohorts appeared to show slightly different trajectories of identified regulation from T1 to T2 based on the model-implied trajectories in Figure 2, these differences were not statistically significant. Further, changes from T1 to T2 negatively predicted changes from T2 to T3, indicating that overall, declines early in college tended to buffer students from later declines.

Despite similarity between cohorts from T1 to T2, we found significant differences from T2 to T3 in identified regulation and intrinsic motivation, with the COVID cohort demonstrating motivational declines that were not observed in the pre-pandemic cohort. These findings suggest that the effects of the pandemic were motivationally specific, largely sapping enjoyment, curiosity-based engagement, and the sense that achievement is personally meaningful. Perhaps students in the COVID cohort struggled to sustain these more autonomous types of motivation because of diminished supports for basic psychological needs. Indeed, college students’ experiences of relatedness, competence, and autonomy were almost certainly diminished due to the social isolation, transition to emergency remote instruction, and constraints placed on their daily behaviors during the pandemic (e.g., Gonzalez-Ramirez et al., 2021; Tasso et al., 2021; Usher et al., 2021).

There was no effect of COVID-19 on the more controlled motives or amotivation. This is interesting to consider in light of Rahiem (2021) finding that autonomous motives were more prominent than controlled motives among students from Indonesia who remained motivated in spite of pandemic limitations. It is surprising, however, that there was no growth in amotivation among the COVID cohort given the general motivational difficulties that have been reported widely in the literature (e.g., Hicks et al., 2021; Tasso et al., 2021; Usher et al., 2021). Previous research has relied upon students’ subjective reports of declining “motivation,” framed as a global, unitary construct. By distinguishing among the different subtypes of motivation, the present study suggests that what appears to be an overall decline in motivation (and rise in amotivation) is actually specific to autonomous types of motivation. Students in the COVID cohort, therefore, appear not to have decreased in their overall investment in their schooling, but rather in the sense that the work is enjoyable and meaningful.

Although autonomous motivation typically predicts high academic achievement (Taylor et al., 2014; Brunet et al., 2015), motivational change did not predict Year 4 GPA in the present study when controlling for Year 2 GPA, nor were there achievement differences between the two cohorts. One might have expected lower GPAs for the COVID cohort given the reported difficulties with motivation, self-regulation, stress management, and attentional control in the broader literature (Hicks et al., 2021; Usher et al., 2021). Perhaps the relatively low and stable levels of amotivation for both cohorts was enough to maintain achievement, especially given that amotivation is typically a strong negative predictor of GPA (Taylor et al., 2014; Corpus et al., 2020; Howard et al., 2021). At the same time, it is difficult to compare GPAs for the two cohorts because standards for grading and assessment shifted so dramatically during the pandemic (Hartocollis, 2020). Future research would benefit from comparisons using less subjective indicators of learning and achievement, such as scores on particular instruments assessing conceptual understanding. It would also be fruitful to include other outcomes (e.g., well-being, persistence) that may be more responsive to changes in autonomous motives (Howard et al., 2021). At the same time, it is important to remember that motivation itself is a meaningful outcome.

The present study was also limited by the lack of motivational assessments during the second and third years of college. We cannot rule out the possibility that differences between the two cohorts from T2 to T3 were driven not by COVID-19 but rather by some other intervening event between the first and fourth years of college. Documenting comparable motivational responses throughout the fall of the first, second, and third years of college would help to show that diverging pathways in the fourth year aligned more precisely with the onset of the COVID-19 pandemic. Even then, it is likely that students’ declining motivational trajectories resulted not only from the pandemic but also from unprecedented sociocultural and regional stressors (e.g., racial protests, epic wildfires) occurring simultaneously. The correlational nature of the present study simply does not allow for causal inferences.

It is also important to acknowledge the results of our missing data analyses, which indicated that our data may underrepresent the experiences of men and lower-achieving students in particular. Our use of FIML estimation reflects best practices for minimizing bias in model estimates due to missing data. However, future research may perhaps benefit from a focus in particular on understanding the motivational experiences of men and lower-achieving students using more personalized incentives or non-survey methods.

Despite these limitations, the present study shows that previously reported motivational costs of COVID-19 cannot be explained by simple memory biases that come from retrospective reports. Instead, it appears that the motivational decline associated with the COVID-19 pandemic is real, can be observed nearly 1 year into the pandemic, and is specific to autonomous motives. Although these changes were relatively small, prior research indicates that even small changes in autonomous regulation types can have important implications for outcomes (Howard et al., 2021; Pap et al., 2021).

Remediation efforts, therefore, might focus on supporting the basic psychological needs that help to build such motives. Students’ sense of belonging can be enhanced when peers work collaboratively on tasks, instructors and students learn of shared preferences, and content is represented as consistent with valued goals (Walton and Brady, 2017). Such interventions to promote belonging increase students’ intrinsic motivation, sense of meaning, and academic achievement (Walton and Brady, 2017). Likewise, instructors can be effectively taught to use autonomy-supportive practices, such as providing students with meaningful choices, using informational language, and acknowledging negative emotions (Su and Reeve, 2011). These practices, in turn, enhance students’ intrinsic motivation, engagement, and academic achievement, while reducing their levels of amotivation (Patall et al., 2010; Cheon and Reeve, 2015). With broader implementation, such practices may forestall the decline in autonomous motives associated with the COVID-19 pandemic.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Reed College Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.

JC, KR, and ZL developed the study concept. JC and ZL collected the data. KR performed the data analysis. JC and KR wrote and revised the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.