Experiment 1
Analytic Plan
Parents’ and children’s answers were scored using a four-point scale as follows: 1 as not smart or sporty, 2 as sort of smart or sporty, 3 as smart or sporty, and 4 as really smart or sporty. Next, linear mixed-effects regression models were built using R package lme4. Cohen’s d was calculated for effect sizes. 95% Confidence intervals were calculated using Wald method. In all models, random effects were modeled for each participant as (1|id). Post hoc simple effects tests were conducted with R package emmeans and p-values were Bonferroni corrected. For critical effects, post hoc sensitivity analyses were conducted using the R package simr. Children and adults participated in slightly different variants of the study, and data were analyzed separately.
Children’s responses
We first fit an omnibus mixed linear effects model including main effects of Age (centered, continuous in months), Learning Style (visual, hands-on), and Question Type (smart, sporty) and all interaction terms to test which factors predicted participant ratings. There was a significant Learning Style main effect, bLearningStyle = 0.22, 95% CI = [0.09, 0.34], p < 0.001, d = 0.47. The interaction between Question Type and Learning Style was also significant, bLearning Style*Question = −0.20, 95% CI = [−0.32,−0.08], p = 0.001, d = −0.44. A post hoc sensitivity analysis determined that the sample was adequately powered to detect this main interaction, 90.30% power with a 95% CI = [88.30%, 92.06%]. All the remaining terms were non-significant.
We next decompose the significant two-way Question Type by Learning Style interaction using follow-up simple effects contrasts (Bonferroni corrected for two tests). Results showed that children viewed visual learners (M(SD)Visual = 2.88 (0.99)) as significantly smarter than hands-on learners (M(SD)Hands-on = 2.26 (1.09), Fig. 1 left panel), t = 3.48, padjusted = 0.001, d = 0.58. In contrast, children rated visual and hands-on learners (M(SD)Visual = 2.04 (1.17)) as similarly sporty (M(SD)Hands-on = 2.23 (1.21), Fig. 1 left panel), t = 1.08, p = 0.540, d = 0.18. Scatterplots, provided in our supplementary materials, illustrate the absence of an age effect. These graphs plot each Question Type and Learning Style and are presented in Supplementary Fig. 1.
Left panel: ratings from the child sample. Right panel: ratings from the parent sample. Ratings were converted to 1-4 to indicate how smart/sporty the participants think about the student described as a visual/hands-on learner. 1 – not very smart/sporty, 2 – sort of smart/sporty, 3 – smart/sporty, 4 – really smart/sporty.
Parents’ responses
A similar set of analyses were then conducted with the parent sample but without age. Results showed that all main effects and interactions were significant. For the main effect of Question Type, bQuestion = 0.16, 95% CI = [0.10, 0.23], p < 0.001, d = 0.56. For the main effect of Learning Style, bLearning Style = 0.10, 95% CI = [0.03, 0.17], p = 0.007, d = 0.32. For the interaction between Learning Style and Question Type, bLearning Style*Question =–0.33, 95% CI = [–0.40, –0.26], p < 0.001, d = –1.11. A post hoc sensitivity analysis determined that the sample was adequately powered to detect this main interaction, 100% power with a 95% CI (99.63%, 100%).
To parse out the interaction effect, we again conducted follow-up simple effects contrasts (Bonferroni corrected for two tests). Results showed that parents rated visual learners (M(SD)Visual = 3.02 (0.49)) as significantly smarter than hands-on learners (M(SD)Hands-on = 2.55 (0.74), Fig. 1right panel), t = 4.66, padjusted < 0.001, d = 0.68. In contrast, parents viewed hands-on learners (M(SD)Hands-on = 2.88 (0.79)) as significantly sportier than visual learners (M(SD)Visual = 2.03 (0.99), Fig. 1 right panel), t = 8.47, padjusted < 0.001, d = 1.24.
Experiment 2
We fit a mixed binary logistic regression to examine whether question (smarter and sportier) and group membership (teacher and parent) predicted participants’ selections (the visual or hands-on learner). Identical to Experiment 1, random effects were again modeled for each participant. The same packages were used as in Experiment 1. The main difference was that due to the binary nature of the DV, odds ratios are reported instead of Cohen’s d.
Figure 2 displays the percentage of participants’ answers who selected each student type (visual or hands-on learner) divided by question (smarter and sportier) and participant group (teacher and parent). The main effect of the question type was significant, b = 1.61, 95% CI = [1.29, 1.92], z = 9.95, p < 0.001, odds ratio = 4.99. There was no main effect of group on judgments (parent or teacher), b = 0.25, 95% CI = [–0.07, 0.56], z = 1.53, p = 0.126, odds ratio = 1.28. But there was a significant interaction between the question and group, b = –0.43, 95% CI = [–0.74, –0.11], z = –2.63, p = 0.008, odds ratio = 0.65. As shown in Fig. 2, participants were more likely to characterize the visual learner as smarter than the hands-on learner, but the hands-on learner as sportier than the visual learner. In terms of interaction, teachers showed a larger effect than parents (again see Fig. 2).
a percentage to the question “Which student is smarter”. b percentage to the question “Which student is sportier” Each bar indicates the percentage that the sample selected the visual or hands-on learner as “smarter” or “sportier”, adding up to 100%.
For the exploratory open-ended question, we then drew word clouds for each learner style (Fig. 3) to help us to visualize the frequency with which different subjects were listed. As shown in Fig. 3 parents and teachers viewed each learner as having different educational strengths. Next, we break down those strengths using frequency data.
Color and word size indicate the frequency of the word (word stems used).
Visual learner
Teachers viewed visual learners as likely to excel at math/mathematics (22.65%), history (14.89%), English (9.39%), art (8.74%), and reading (5.83%). Similarly, the top five subjects listed by parents were math/mathematics (19.56%), English (9.96%), art (7.75%), history (7.75%), and reading (5.54%).
Hands-on learner
Teachers viewed hands-on learners as likely to excel in science (19.14%), art (16.83%), gym (11.55%), math (6.60%), and music (5.28%). Similarly, the top five subjects listed by parents were science (13.52%), art (11.03%), gym (6.76%), chemistry (5.34%), and physics (4.63%). We note that collapsing across subcategories of science (e.g., chemistry, physics) does not change the nature of the top three subjects.
Experiment 3
The same packages and analytic method were used as in Experiment 1. However, we did not use Bonferroni correction because of pre-registration.
Pre-registered Hypothesis 1
Hypothesis
Visual learners will be rated as having higher grades than hands-on learners for “traditional” core school subjects (e.g., math, science, social studies, language arts). In contrast, we predict that hands-on learners will be rated higher on non-core subjects than visual learners (e.g., art, music, gym).
Analysis
We ran a mixed linear effects model including main effects of Subject Type (core, non-core), Learning Style (visual, hands-on), and Sample (teacher, parent) and all their interaction terms to predict grade ratings. There were significant main effects of Learning Style, bLearning Style = 0.07, 95% CI = [0.004, 0.128], p = 0.036, d = 0.08, and Subject Type, bSubject Type = –0.69, 95% CI = [–0.75, –0.63], p < 0.001, d = –0.86, but not Sample, bSample = –0.05, 95% CI = [–0.17, 0.08], p = 0.456, d = –0.11. There was a significant interaction between Learning Style and Subject Type, bLearning Style*Subject Type = –0.60, 95% CI = [–0.66, –0.54], p < 0.001, d = –0.75. A post hoc sensitivity analysis determined that the sample was adequately powered to detect this interaction, 100% power with a 95% CI (99.63%, 100%).
We next decomposed the significant two-way Learning Style by Subject Type interaction using simple effects contrasts. As predicted, for core subjects, participants rated the visual learner as having higher grades than the hands-on learner, t = 12.89, p < 0.001, d = 0.65. In contrast, for non-core subjects, participants rated the hands-on learner as having higher grades than the visual learner, t = 13.94, p < 0.001, d = 0.81 (Fig. 4).
Grade ratings were on a 1–10 scale, representing letter grades from Below C-, C-, C, C+,… A+, respectively. Each bar indicates the mean grade rating with error bars showing s.e.m.s. Subject type is divided into core (math, science, language arts, social studies) and non-core (art, gym, music).
Pre-registered Hypothesis 2 A
Hypothesis
Visual learners will be rated as having higher grades for math, social studies, and language arts, but not science. For science, we predict an opposite trend wherein hands-on are rated as having higher grades than visual learners.
Analysis
We ran a mixed linear effects model which tested whether the main effects of Subject (math, social studies, language arts, science), Learning Style (visual, hands-on), and their interaction predicted grade ratings. There were significant main effects of Learning Style, bLearning Style = –0.53, 95% CI = [–0.60, –0.46], p < 0.001, d = –0.79, and Subject, bSubject = 0.97, 95% CI = [0.85, 1.09], p < 0.001, d = 0.83. The interaction between Learning Style and Subject also yielded significance, bLearning Style*Subject = 0.62, 95% CI = [0.50, 0.74], p < 0.001, d = 0.53. A post hoc sensitivity analysis determined that the sample was adequately powered to detect this interaction, 100% power with a 95% CI (99.63%, 100%).
We next decompose the significant two-way Learning Style by Subject interaction using simple effects contrasts. As predicted, participants rated the visual learner to score higher than the hands-on learner in language arts ((M(SD)Visual = 7.15(1.86), (M(SD)Hands-on = 5.67(1.69), t = 10.23, p < 0.001, d = 1.02), math ((M(SD)Visual = 7.36(1.79), (M(SD)Hands-on = 5.82(1.91), t = 10.65, p < 0.001, d = 1.07), social studies ((M(SD)Visual = 6.96(1.96), (M(SD)Hands-on = 5.55(1.72), t = 9.72, p < 0.001, d = 0.97), but not science (M(SD)Visual = 7.62(1.55), (M(SD)Hands-on = 7.80(1.72), t = 1.21, p = 0.226, d = 0.12, Fig. 5).
Grade ratings were on a 1–10 scale, representing letter grades from Below C-, C-, C, C+,… A+, respectively. Each bar indicates the mean grade rating with error bars showing s.e.m.
Pre-registered Hypothesis 2B
Hypothesis
Hands-on learners will be rated higher on all non-core subjects than visual learners.
Analysis
Similar to the analysis of Hypothesis 2 A, we first ran an omnibus analysis which examined whether the main effects of Subject (art, gym, music) and Learning Style (visual, hands-on) and their interaction predicted grade ratings. There were significant main effects of Learning Style, bLearning Style = 0.66, 95% CI = [0.59, 0.74], p < 0.001, d = 1.06, and Subject, bSubject = 0.73, 95% CI = [0.62, 0.84], p < 0.001, d = 0.83. The interaction between Learning Style and Subject was also significant, bLearning Style*Subject = –0.43, 95% CI = [–0.54, –0.32], p < 0.001, d = –0.48. A post hoc sensitivity analysis determined that the sample was adequately powered to detect this interaction, 100% power with a 95% CI (99.63%, 100%).
We next decomposed the significant two-way Learning Style by Subject interaction using simple effects contrast tests. As predicted, participants rated the hands-on learner to score higher than the visual learner in Art ((M(SD)Visual = 8.61(1.55), (M(SD)Hands-on = 9.09(1.10), t = 3.47, p < 0.001, d = 0.35), Gym ((M(SD)Visual = 7.07(2.04), (M(SD)Hands-on = 9.01(1.23), t = 14.16, p < 0.001, d = 1.42), and Music ((M(SD)Visual = 6.69(1.87), (M(SD)Hands-on = 8.25(1.56), t = 11.42, p < 0.001, d = 1.14, Fig. 6).
Grade ratings were on a 1–10 scale, representing letter grades from Below C-, C-, C, C+,… A+, respectively. Each bar indicates the mean grade rating with error bars showing s.e.m.
Pre-registered Hypothesis 3
Hypothesis/goal
Replication of Experiment 2 findings that visual learners are perceived to be smarter than hands-on learners, and that teachers show a larger effect.
Analysis
Figure 7 displayed the percentage of participants’ answers of student type (visual/hands-on learner) by question type (smarter and harder) and sample group (teacher and parent). We fit a mixed effects binary logistic regression to examine whether the question (smarter, works harder) and group (teacher, parent) predicted participants’ answers (the visual or hands-on learner). The main effect of the question type was significant, b = –0.83, 95% CI = [–1.04, –0.612], z = –7.59, p < 0.001, odds ratio = 0.44. Participants were more likely to pick the visual learner as smarter than the hands-on learner; in contrast (Fig. 7a), they were more likely to pick the hands-on learner as working harder than the visual learner (Fig. 7b). There was no main effect of the sample group (parent or teacher), b = –0.071, 95% CI = [–0.28, 0.14], z =–0.65, p =0.515, odds ratio = 0.93. The interaction between the question and group was not significant, b = –0.001, 95% CI = [–0.21, 0.21], z = –0.01, p = 0.990, odds ratio = 1.00. We did not replicate the differences between parents and teachers from Experiment 2 and so it is not discussed further.
a percentage to the question “Which student is smarter”. b percentage to the question “Which student is sportier”. Each bar indicates the percentage that the sample selected the visual or hands-on learner as “smarter” or “sportier”, adding up to 100%.
link