Background and Goals

Past research had shown that different kinds of everyday activities “fit” distinct types of motivation, and these studies have typically examined specific motive pairs. For instance:

• Prevention vs. Promotion:²
  • Prevention motivation—the desire to ensure against losses and carefully maintain what one currently has—fits with vigilant activities like stopping procrastination or avoiding being unprepared for work.³
  • Promotion motivation—the desire to make progress and avoid being stuck at one’s status quo—fits with eager activities like completing tasks promptly or preparing for work.
• Truth vs. Control:⁴
  • Truth motivation—the desire to establish what is real and right in one’s life—fits with evaluation-oriented activities like making detailed comparisons.⁵
  • Control motivation—the desire to manage what happens in one’s life—fits with action-oriented activities like effecting change.

However, research had not yet established which kinds of goal pursuit activities are prompted when these motivations work together in combination.

I conducted mixed-methods experimental research to validate the following framework, which I proposed based upon my knowledge of the underlying motivational theories and prior research.

2×2 Motivated Activity Framework:

Business Objective

While this research was conducted in an academic setting, and thus was not instigated by a particular business goal, one can imagine possible objectives that it might help to satisfy. For instance:

• Identify consumers most likely to be interested in product features that facilitate different types of activities
• Increase users’ app usage frequency by surfacing the most motivationally-relevant content on the home page

Research Questions

• Do people agree with our framework’s proposals about the motivations underlying different types of activities? And if so, does their behavior align with their explicit agreement?
• Are activities more likely to be “top of mind” for people when they “fit” the individuals’ strongest motives?

Hypotheses

• Study Set 1: [Pre-registration for Study 1B available here.]
  • Participants will consistently categorize goal pursuit process activities from the framework as reflecting the proposed motivation.
  • Participants will categorize activities from the proposed framework more quickly than more general goal pursuit activities unrelated to the framework.
• Study Set 2: [Pre-registration for Study 2B available here.]
  • Participants will remember activities that reflect their own motives earlier than activities that do not.

Study Set 1

Methods

I conducted an online experiment in which participants categorized goal pursuit activities and answered questions about their categorization decisions (Study 1A), followed by a pre-registered experiment aiming to replicate and extend the first study at Columbia University (Study 1B).

Reasoning

• I chose to conduct Study 1A online because it offered me the best chance to achieve a high level of age, education, and income diversity. Then, I chose to conduct Study 1B with a student sample to ensure that the results replicated with a more racially diverse group of participants.
• I proceeded with both attitudinal and behavioral dependent measures because people’s explicitly-stated beliefs do not always align with how they act on those beliefs (i.e., the “attitude-behavior gap“).
• To complement these quantitative dependent measures, I also asked participants open-ended questions about their categorization decisions to get a clearer qualitative understanding of participants’ feelings and beliefs about the motives underlying the activities tested.

Participants

• Study 1A: 61 U.S. adults recruited on Amazon Mechanical Turk (“M-Turk”) made 6,262 activity categorization decisions (M = 103 activities per participant)⁶
  • Age: ranging from 24 to 68 (M = 37)
  • Sex: 51% male; 48% female
  • Race/Ethnicity: 80% White; 8% Black; 7% Asian; 2% Hispanic; 2% multiracial
  • Education: 18% high school diploma; 16% some college; 10% associate’s degree; 48% bachelor’s degree; 8% master’s degree
• Study 1B: 67 undergraduate and post-baccalaureate students recruited from Columbia University’s psychology department participant pool made 2,703 activity categorization decisions (M = 40 activities per participant)
  • Age: ranging from 18 to 38 (M = 21)
  • Sex: 36% male; 61% female
  • Race/Ethnicity: 42% White; 24% Asian; 15% multiracial; 9% Hispanic; 7% Black
  • Education: 1% some high school; 33% high school diploma; 51% some college; 4% associate’s degree; 7% bachelor’s degree

Study Materials

First, participants completed self-report measures of prevention and promotion motivation⁷ as well as truth- and control-oriented motivational orientations called assessment and locomotion⁸. In Study 1B, participants also completed a behavioral measure of prevention and promotion, which uses their response speed to assess motivational strength, prior to the self-report measures.⁹

Next, they learned about two different types of motivation called promotion and prevention:

Then, participants completed a sorting task in which they used their keyboard to categorize activities as reflecting either promotion or prevention motivation:¹⁰

In Study 1B, participants then completed the same sorting task with truth and control motivation after learning about this second pair of motives:

Within these tasks, the survey software recorded two types of data for each activity that each participant categorized:

1. The selected motivational category, which told us which motive the participant thought best “fit” the activity (i.e., a self-report measure of beliefs)
2. The time it took participants to make their decision, which revealed how clearly the motive fit one category over the other (i.e., a behavioral measure)

Activity List:¹¹

	Prevention Motivation	Promotion Motivation	General
Truth Motivation	Assess, Examine, Judge, Review, Scrutinize, Verify	Discover, Explore, Imagine, Invent, Seek, Wonder	Consider, Inquire, Investigate
Control Motivation	Defend, Guard, Maintain, Preserve, Protect, Resist	Accelerate, Elevate, Launch, Lead, Progress, Propel	Control, Manage, Operate
General	Evade, Prevent, Secure	Achieve, Gain, Grow	Sleep, Shower, Communicate

Finally, after completing the sorting tasks, participants answered both open- and closed-ended questions about their categorization decisions:

Analytic Strategy:

• I conducted a Bayesian multilevel logistic regression analysis to test whether whether our predictions about the hypothesized motivation type from each motive pair aligned with participants’ categorization decisions for each activity.¹²
• I conducted a Bayesian multilevel linear regression analysis to test whether tested whether activities that I hypothesized “fit” one of the motive categories were sorted more quickly than general, everyday activities that I hypothesized were not associated with one of these types of motivation (e.g., sleep, shower).¹³
• I chose to proceed with multilevel analyses because modeling the goal pursuit activities that we tested as random stimuli reflected our thinking about these activities.¹⁴ (For what it’s worth, this is also a more conservative approach!)
• Additionally, both of these multilevel regression analyses were conducted using Bayesian (rather than frequentist) statistical methods because this approach provided me with distributions of possible values for each parameter in our models, and also ensured that the fairly complex multilevel analysis could actually be completed (which often isn’t the case when conducting similar multilevel analyses using frequentist methods).¹⁵
• Finally, I conducted a qualitative content analysis on participants’ responses to the open-ended questions about their categorization decision processes.
• To access the study materials, raw data, and R code used for Studies 1A and 1B, please visit the Open Science Framework page for this project.

Results

Quantitative Analyses

I found that participants consistently and quickly sorted activities in line with my hypothesized framework (plotted as red squares and green circles in the example graph below from Study 1B). By contrast, “general” activities hypothesized not to align with one of the motives in question (plotted as gray triangles below) were categorized less consistently and more slowly. [Click to enlarge]

Participants’ promotion vs. prevention motivation categorization decisions aligned with my hypotheses in Study 1A and Study 1B:¹⁶ [Click to enlarge]

As did participants’ control vs. truth motivation categorization decisions when this was tested in Study 1B:¹⁷

Additionally, participants categorized activities from my framework as promotion vs. prevention more quickly than they did for “general” everyday activities in both Studies 1A and 1B, providing behavioral evidence of the activities’ associations with the proposed underlying motives:¹⁸

And the same was true for control vs. truth categorization decisions when this task was included in Study 1B:¹⁹

Qualitative Analyses

My qualitative content analysis examining the activities that participants listed in their open-ended responses revealed that activities hypothesized not to fall within my final 2×2 motivational framework (shown in gray below) tended to be more frequently listed as difficult to categorize as compared to activities that were included in the 2×2 framework (shown in blue below):

Study Set 2

Methods

I conducted an online study in which participants’ motivations were measured, and then they were were surprised with a memory test for goal pursuit activities that they had rated in an earlier task (Study 2A). This study was followed by a pre-registered experiment aiming to replicate the first study at Columbia University with a manipulation (vs. measurement) of their motives (Study 2B). In both studies, I examined the order in which remembered activities were listed by participants.

Reasoning

• Similar to Study Set 1, I chose to conduct Study 2A online to attain age, education, and income diversity, and Study 2B with a university sample to attain more racial diversity.
• I replaced the motive measure from Study 2A with an experimental motivational manipulation in Study 2B, which allowed me to make causal inferences about the effect that participants’ own motives have on the activities are remembered earliest.
• I proceeded with a behavioral dependent measure that participants were highly unlikely to have anticipated because I was interested in minimizing bias in our results (e.g., from participants wanting their responses later in the study to be consistent with their responses in earlier tasks).

Participants

• Study 2A: 57 U.S. adults recruited on Amazon Mechanical Turk (“M-Turk”) remembered 254 activities from the 2×2 framework (M = 4.46 activities per participant)
  • Age: ranging from 21 to 64 (M = 36)
  • Sex: 53% male; 47% female
  • Race/Ethnicity: 72% White; 9% Asian; 7% Black; 8% Hispanic; 5% multiracial
  • Education: 2% some high school; 21% high school diploma; 19% some college; 18% associate’s degree; 35% bachelor’s degree; 5% master’s degree
• Study 2B: 119 members of the Columbia University community remembered 916 activities from the 2×2 framework (M = 7.70 activities per participant)
  • Age: ranging from 18 to 68 (M = 24)
  • Sex: 37% male; 60% female
  • Race/Ethnicity: 47% Asian; 24% White; 16% multiracial; 8% Hispanic; 4% Black
  • Education: 12% high school diploma; 31% some college; 2% associate’s degree; 23% bachelor’s degree; 27% master’s degree; 2% doctorate

Study Materials

First, I measured or manipulated participants’ promotion/prevention motives.

Next, in order to present the activities to participants a first time, I asked them to rate how important participants considered it to engage in each activity in their own lives:

Then, participants completed a brief distractor task to clear their working memory:²²

Finally, I surprised participants with an unaided recall task in which they were asked to remember as many of the activities as they could from the importance rating task and to list them in a series of empty text boxes on the page:

Within this tasks, the survey software recorded the “memory rank” for each activity that each participant remembered. For instance, the activity listed in the top text box was coded as having a memory rank of 1, the activity listed in the next text box was coded as having a memory rank of 2, and so on.²³

Analytic Strategy

• I conducted a Bayesian multilevel linear regression analysis to test whether participants remembered activities from the 2×2 framework earlier (vs. later) when the activity “fit” their promotion vs. prevention motivation.²⁴
• Additionally, because I hypothesized that the activities that “fit” participants’ motives would be recalled earliest in the memory task, we also examined a subset of the data that was recalled earliest in the studies. These analyses in Study 2A indicated that the interaction effect of interest is most strongly associated with memory rank when including only the first 13 activities that participants recalled within the unaided recall task, so I proceeded with an analysis of this same subset in Study 2B.
• I chose to proceed with Bayesian multilevel analyses for the same reasons as in Study Set 1.
• To access the study materials, raw data, and R code used for Studies 2A and 2B, please visit the Open Science Framework page for this project.

Results

Quantitative Analyses

I found that the activities relevant to participants’ stronger (vs. weaker) motivations tended to come to mind first. [Click to enlarge]

Interestingly, though, the effect emerged in the analyses of both the full dataset²⁵ and the subset of the 13 activities recalled earliest in Study 2A (and the former is shown in the figure above). In Study 2B, it only emerged in the latter case (shown in the figure above).²⁶

Recommendations and Impact

While this research was conducted in an academic context and thus did not result in any recommendations for a specific business, these findings have important implications for harnessing the motivations underlying people’s activities in different aspects of their lives. I highlight these implications in several publications that discuss how this framework can be applied to different contexts, as well as through follow-up research that has begun to test these recommendations.

Product Design and Marketing

Recommendations

Publication: “Marketplace solutions to motivational threats: Helping consumers with four distinct types of vulnerability” [PDF]

Businesses should first determine which combination of motivational needs (i.e., which of the four cells in the 2×2 framework) their product or service satisfies.

Impact

Follow-up research has begun to validate these recommendations using real-world stimuli.

For instance, one study has revealed that when people are presented with mobile apps that facilitate activities that do (vs. do not) “fit” their own motivations, they are more likely to choose to use them.

In this study, after first assessing participants’ promotion vs. prevention motivational predominance using the behavioral measure used above, I told participants that they would be testing a new mobile app within their computer’s browser window at the end of the survey.

I framed this task as a real test so participants were making an actual behavioral choice about which of the apps to use. Participants were presented with pairs of fictional iPhone apps on the same topic, but framed as facilitating activities from different cells in the 2×2 motivational framework:²⁷

I then presented all participants with bogus “loading” and “error” screens because in this first study, I was solely interested in their app usage choices, rather than what they did once they started using the selected app (another interesting question for a future study!):

Results revealed that the relative strength of participants’ own promotion vs. prevention motivation predicted the proportion of apps they facilitated activities “fitting” (vs. not “fitting”) their own motives:

I am currently extending this paradigm within my dissertation research to test for simultaneous motivational “fit” effects on both the promotion vs. prevention and truth vs. control dimensions.

Education

Recommendations

Publication: “Harnessing regulatory focus and regulatory fit to improve educational outcomes” [PDF]

Once educators have determined whether activities that do vs. do not “fit” students’ motivations will be most beneficial (see the article for more detail on this question), they will need to carefully consider how to facilitate such (non)fit. Note: These recommendations would also be relevant to managers looking to improve their employees’ outcomes in the workplace!

Impact

My follow-up research has begun to validate these recommendations with students using real study strategies.

For instance, one study has revealed that when university students are asked to rate how enjoyable they would find study activities that do (vs. do not) “fit” their own motivations, I detect simultaneous motivational fit effects on both the promotion vs. prevention and truth vs. control dimensions.

Results revealed that the extent to which each study activity was anticipated to be enjoyable was positively influenced by the strength of its motivational “fit” with participants’ own motives on both dimensions:

Scientific Community

This research has also been shared with the scientific community at psychology conferences and through an article preprint (which I am currently revising to include my subsequent dissertation research that builds on this foundational work).

Poster Presentation (SPSP 2020)

[Click to enlarge]

Article Preprint

[View on PsyArXiv]

Limitations and Learnings

While this research provided some clarity into our research questions, it had some important limitations:

• As described in more detail in the article, although I found that participants’ measured motives predicted how “top of mind” activities were in Study 2A, when experimentally manipulating participants’ motivation in Study 2B, I only found evidence for the hypothesized effect when examining a smaller group of activities remembered earliest by these participants. Potential explanations for this difference include:
  • Differences in the samples’ engagement in the two studies (and my data provide suggestive evidence for this proposal; see the full article for more information).
  • The interaction between participants’ motivation (measured or manipulated) and activities’ underlying motives only affect the earliest items that come to mind, but that I was not able to detect this difference in Study 2A because the number of activities participants remembered simply did not reach this threshold.
  • The activation of the motivation induced by the priming in Study 2B may have decreased throughout the course of the study, thus impacting more strongly the activities that were remembered earlier (vs. later).
• Additional research is required to determine if any of these possible explanations is correct. Importantly, if follow-up research indicates that the priming does, in fact, wear off during the course of the study, future work should be done to confirm this shift in accessibility and pinpoint its time course.
  • Understanding the duration during which such an induction is effective would be particularly important for people and businesses interested in developing interventions to promote specific types of activities.

The process also revealed a few unexpected learnings.

• In retrospect, it would have been useful to also conduct an alternate version of Study Set 2—that replicated the same “top of mind” effect with truth vs. control motivations.
  • My subsequent dissertation research expands on this idea, testing for simultaneous effects of motivational “fit” on both the promotion vs. prevention and truth vs. control dimensions.
• When I launched Study 1A, I was less familiar with common motivational measures of promotion vs. prevention and, thus, lacked nuanced insight into their strengths and weaknesses. For instance, with time, I realized that behavioral (vs. self-report) measures are better predictors of behavioral (vs. self-report) outcomes.
  • Thankfully, as this project progressed, I developed a deeper understanding of these differences. As a result, I added a behavioral measure in subsequent studies.

Footnotes

1. This pilot research was conducted by an undergraduate honors student working in our lab under my supervision. A correlational analysis indicated that participants’ own promotion (vs. prevention) motivational predominance positively predicted the proportion of apps framed as facilitating promotion (vs. prevention) activities that they chose to use (r(77)= 0.24, p = .031). Please note that this work has not yet been published; in my dissertation work, I am currently replicating and extending it to test for simultaneous effects of promotion-prevention and truth-control motivational “fit” on app usage choices. ↩︎
2. Higgins, E. T. (1997). Beyond pleasure and pain. American Psychologist, 52(12), 1280–1300. https://doi.org/10.1037/0003-066X.52.12.1280 ↩︎
3. Freitas, A. L., & Higgins, E. T. (2002). Enjoying goal-directed action: The role of regulatory fit. Psychological Science, 13(1), 1–6. https://doi.org/10.1111/1467-9280.00401 ↩︎
4. Higgins, E. T. (2018). Going in the right direction: Locomotion control and assessment truth, working together. In C. E. Kopetz & A. Fishbach (Eds.), The motivation-cognition interface: From the lab to the real world; A Festschrift in honor of Arie W. Kruglanski. Routledge. ↩︎
5. Avnet, T., & Higgins, E. T. (2003). Locomotion, assessment, and regulatory fit: Value transfer from “how” to “what.” Journal of Experimental Social Psychology, 39(5), 525–530. https://doi.org/10.1016/S0022-1031(03)00027-1 ↩︎
6. Across Studies 1a and 1b, as per common practice when working with response latency data (e.g., Carpenter et al., 2019), I excluded activity categorization decisions for which the response latency was too short (under 250 ms), thus indicating overly fast responding (i.e., “button mashing”), and too long (over 10,000 ms), indicating a lack of attention. ↩︎
7. Higgins, E. T., Friedman, R., Harlow, R. E., Idson, L. C., Ayduk, O. N., & Taylor, A. (2001). Achievement orientations from subjective histories of success: Promotion pride versus prevention pride. European Journal of Social Psychology, 31(1), 3–23. https://doi.org/10.1002/ejsp.27 ↩︎
8. Kruglanski, A. W., Thompson, E. P., Higgins, E. T., Atash, M. N., Pierro, A., Shah, J. Y., & Spiegel, S. (2000). To “do the right thing” or to “just do it”: Locomotion and assessment as distinct self-regulatory imperatives. Journal of Personality and Social Psychology, 79(5), 793–815. https://doi.org/10.1037/0022-3514.79.5.793 ↩︎
9. Higgins, E. T., Shah, J., & Friedman, R. (1997). Emotional responses to goal attainment: Strength of regulatory focus as moderator. Journal of Personality and Social Psychology, 72(3), 515–525. https://doi.org/10.1037/0022-3514.72.3.515 ↩︎
10. I developed and administered this sorting task using Javascript embedded within the Qualtrics survey. ↩︎
11. A slightly larger activity list was tested in Study 1a and refined during the analysis phase before proceeding with this final list for Study 1b. These lists are available in the Supplementary Material on the Open Science Framework project page. ↩︎
12. For each of these logistic regressions, I selected the subset of the data comprising only activities I hypothesized to reflect the pair of motives in question and analyzed these data within a Bayesian multilevel binomial logistic regression model. Participants’ categorization decisions (i.e., our primary dependent variable) were dummy-coded for this analysis (0=prevention, 1=promotion OR 0=truth; 1=control), and the same coding scheme was applied to the hypothesized motivation of each activity (i.e., the primary independent variable). The model also included motive predominance as a covariate, as I was also interested in testing if categorization decisions were associated with individual differences in participants’ own motivation. Other covariates included as fixed effects were the initial placement of the category names (effect-coded: –0.5=right; +0.5=left), as well as trial, frequency of use in the English language for each activity word, and length in characters for each activity word (all rescaled: M=0; SD=1). Random effects included in the model were random intercepts of participant and activity; random slopes of trial, activity frequency, and activity length by participant; and random slopes of promotion pride dominance by activity. ↩︎
13. To conduct this analysis, I analyzed all categorization decisions for all activities within a Bayesian multilevel linear regression model. Participants’ response latencies for each categorization decision were log-transformed to normalize the distribution. Then, these log-transformed response latencies were regressed on a dummy-coded variable capturing whether or not each activity was hypothesized to reflect a specific motivation from the category options (0=does not “fit” either of the motives; 1=”fits” one of the motives). This model included the same covariates and random effects as the logistic regression models described in Footnote 12. ↩︎
14. Although I believe that the list of activities that I tested reflects the different kinds processes driven by each motive, I do not claim that these activity lists are complete. For this reason, it is more accurate to treat these activities as a random sample among the larger set of activities associated with prevention, promotion, truth, and control motivation. Please note that I only report the fixed effects from these analyses here because, although I was interested in accounting for random effects in the models, I did not have any specific hypotheses about them (e.g., with respect to heterogeneity in the effects). However, the complete model output, including both fixed and random effects, are available in the Supplementary Material on the Open Science Framework project page. ↩︎
15. Because these studies were the first attempt to study the hypothesized associations between regulatory focus and goal pursuit activities, I took a conservative approach by proceeding with uninformative prior distributions reflecting no prior knowledge for all variables. All analyses were conducted using four Markov chains. I assessed model convergence by ensuring that all R-hat values fell at or below 1.01, and to ensure each model’s success in achieving this convergence benchmark, I used between 2,000 and 4,000 MCMC iterations per chain. In order to draw inferences from these Bayesian multilevel analyses, I followed a common procedure for Bayesian parameter estimation recommended by Kruschke (2011), which involves the examination the 95% credible intervals around the estimated regression coefficients. If this credible interval excluded 0, I concluded that the coefficient reflected an effect of interest. The posterior distributions for all effects of interest, including a visualization of these 95% credible intervals, are provided in the Supplementary Material. ↩︎
16. To test if my hypotheses about the activities’ promotion vs. prevention motivational underpinnings predicted participants’ categorization decisions, I examined the model output from the Bayesian multilevel logistic regression analyses. In Study 1A, the estimated effect of hypothesized activity motivation was 2.82 log odds units (OR=16.78), with a 95% credible interval that did not include zero (2.35, 3.32). In Study 1B, the estimated effect of hypothesized activity motivation was 7.76 log odds units (OR= 2,344.90), with a 95% credible interval that did not include zero (6.33, 9.46). These results indicate that there is an association between the hypothesized promotion vs. prevention motivation of goal pursuit activities and participants’ categorization decisions. ↩︎
17. Similar to Footnote 14, to test if my hypotheses about the activities’ control vs. truth motivational underpinnings predicted participants’ categorization decisions, I examined the model output from the Bayesian multilevel logistic regression analyses. The estimated effect of hypothesized activity motivation was 2.34 log odds units (OR= 2,344.90), with a 95% credible interval that did not include zero (1.74, 2.98). These results indicate that there is an association between the hypothesized control vs. truth motivation of goal pursuit activities and participants’ categorization decisions. ↩︎
18. To test if my hypotheses about the activities’ associations with promotion or prevention motivation (vs. lack thereof) predicted how quickly participants made their categorization decisions, I examined the model output from the Bayesian multilevel linear regression analyses. In Study 1A, the estimated effect of this hypothesized association was –0.05 log-transformed milliseconds, with a 95% credible interval that did not include zero (–0.09, –0.01). In Study 1B, the estimated effect of this hypothesized association was –0.28 log-transformed milliseconds, with a 95% credible interval that did not include zero (–0.40, –0.15). These results suggest that, as hypothesized, participants were faster in categorizing activities that I hypothesized were (vs. were not) associated with promotion or prevention motivation. ↩︎
19. Similar to Footnote 16, to test if my hypotheses about the activities’ associations with control or truth motivation (vs. lack thereof) predicted how quickly participants made their categorization decisions, I examined the model output from the Bayesian multilevel linear regression analyses. The estimated effect of this hypothesized association was –0.08 log-transformed milliseconds, with a 95% credible interval that did not include zero (–0.15, –0.02). These results suggest that, as hypothesized, participants were faster in categorizing activities that we hypothesized were (vs. were not) associated with control or truth motivation. ↩︎
20. Higgins, E. T., Shah, J., & Friedman, R. (1997). Emotional responses to goal attainment: Strength of regulatory focus as moderator. Journal of Personality and Social Psychology, 72(3), 515–525. https://doi.org/10.1037/0022-3514.72.3.515 ↩︎
21. Higgins, E. T., Roney, C. J. R., Crowe, E., & Hymes, C. (1994). Ideal versus ought predilections for approach and avoidance: Distinct self-regulatory systems. Journal of Personality and Social Psychology, 66(2), 276–286. https://doi.org/10.1037/0022-3514.66.2.276 ↩︎
22. These distractor items were drawn from visual pattern-completion items from Raven’s Progressive Matrices; Raven, J., & Raven, J. (2003). Raven Progressive Matrices. In R. S. McCallum (Ed.), Handbook of nonverbal assessment (pp. 223–237). Kluwer Academic / Plenum Publishers. https://doi.org/10.1007/978-1-4615-0153-4_11 ↩︎
23. These memory ranks were then log-transformed to normalize the distributions and then multiplied by –1 so that higher scores indicated that activities were more “top of mind” (i.e., accessible). ↩︎
24. To conduct these analyses, I analyzed the memory ranks for all of the 24 activities of interest from the 2×2 framework that each participant remembered within a Bayesian multilevel linear regression model. Participants’ reversed, log-transformed memory ranks for each remembered activity were regressed on the hypothesized activity motivation (effect-coded: 0.5=prevention; +0.5=promotion), the participant’s mean-centered promotion (vs. prevention) motivational predominance (in Study 2A) OR the participant’s motivational induction condition (in Study 2B; effect-coded: 0.5=prevention; +0.5=promotion), and the interaction of interest between activity motivation and participants’ motivational dominance. Other covariates included as fixed effects were memory total (i.e., the total number of activities the participant remembered) as well as activity frequency and activity length (both rescaled: M=0; SD=1). Random effects included in the model were random intercepts of participant and activity; random slopes of activity frequency and length by participant; and random slopes of participants’ motivational predominance by activity.  ↩︎
25. To test if my hypotheses about the “fit” between activities’ motivational underpinnings and participants’ own motivation predicted the order in which activities were remembered in Study 2A, I examined the model output from the Bayesian multilevel linear regression analyses. In the analysis of the full dataset, the effect size for the participant motivation x activity motivation interaction of interest was 0.27 units, with a 95% credible interval that did not include zero (0.05, 0.49). Similarly, in the analysis of the subset of the first 13 activities remembered, the effect size for the participant motivation x activity motivation interaction of interest was 0.28 units, with a 95% credible interval that did not include zero (0.06, 0.50). These results indicate that activities are more “top of mind” when they “fit” participants’ own motives. ↩︎
26. In the analysis of the full dataset in Study 2B, I did not find the hypothesized effect; the effect size for the participant motivation x activity motivation interaction of interest was 0.09 units, with a 95% credible interval that included zero (–0.10, 0.28). However, the effect of interest did emerge in the analysis of the subset of the first 13 activities remembered; the effect size for the participant motivation x activity motivation interaction was 0.20 units, with a 95% credible interval that did not include zero (0.002, 0.39). This latter result suggests that activities are more “top of mind” when they “fit” participants’ own motives, but that this may only be the case for the earliest activities remembered. ↩︎
27. Importantly, I counterbalanced the app titles, icons, and imagery across the stimuli to ensure that these design features were not responsible for the results. ↩︎