Preliminary analyses
Preliminary analyses were conducted to assess differences between the three groups in terms of age and evaluation of vignette characters. A one-way ANOVA revealed no difference in age across the three groups, F(2, 155) = 1.43, p = .241. In addition there was no difference across groups in terms of how realistic or likeable participants perceived the vignette characters to be (realistic/Mark F(2, 153) = 1.72, p = .182, realistic/Paul F(2, 153) = 1.18, p = .308, realistic/Killian F(2, 153) = .498, p = .609, realistic/Simon F(2, 153) = 2.66, p = .073, likeable/Mark F(2, 153) = 1.16, p = .315, likeable/Paul F(2, 153) = .750, p = .474, likeable/Killian F(2, 153) = .450, p = .639, likeable/Simon F(2, 153) = .450, p = .639)
As a manipulation check at the debrief stage, participants were also asked for their agreement with a continuum statement: “How much do you agree with the statement “We are all sometimes like Simon or Killian, it’s just a question of how extreme that state is””. They responded on a scale from 1 (strongly disagree) to 6 (strongly agree). Descriptive statistics indicated that participants in the Continuum Group accepted/endorsed the validity of the continuum information (M = 4.61; SD = 1.38).
Data reduction
Raw data consisted of participants’ responses (MHE and emotional reactions: anger, sympathy, discomfort) to each of the four vignette characters at Time 1, Time 2, and Time 3. From the raw data, “between category accentuation” (BCA) and “within category assimilation” (WCA) scores were calculated, guided by Foroni and Rothbart’s [
19] data reduction analysis. The BCA scores comprised the absolute difference between responses to Killian and Paul. Only Killian and Paul were chosen for the BCA as their vignettes are directly on either side of the depression label boundary. We believe to add all four characters into this (i.e. add the difference between Mark and Killian, and Paul and Simon), would dilute the BCA result.
This produced a score ranging from 0 (no difference) to 5 (large difference). As there are two within-category pairs (i.e. Mark & Paul, and Killian & Simon), absolute difference scores for these pairs were averaged to create the WCA scores. This also produced a score ranging from 0 (no difference) to 5 (large difference). See Fig.
1 for a visual representation of between-category and within-category pairs. See Table
1 for descriptive statistics for BCA and WCA scores.
Table 1
Descriptive statistics for between category accentuation and within category assimilation scores
Mental health | Control | 1 | 0–4 | 1.55 | 1.01 | 0.00–3.00 | 1.28 | 0.58 |
2 | 0–4 | 1.75 | 1.08 | 0.00–3.00 | 1.25 | 0.62 |
3 | 0–4 | 1.90 | 1.03 | 0.00–3.00 | 1.14 | 0.67 |
Mental health | Label | 1 | 0–4 | 1.71 | 0.98 | 0.50–3.00 | 1.34 | 0.58 |
2 | 0–4 | 1.90 | 1.05 | 0.50–2.50 | 1.35 | 0.60 |
3 | 0–4 | 1.94 | 1.05 | 0.50–3.50 | 1.41 | 0.66 |
Mental health | Continuum | 1 | 0–4 | 1.59 | 0.91 | 0.00–2.50 | 1.24 | 0.54 |
2 | 0–5 | 1.90 | 1.10 | 0.00–3.50 | 1.16 | 0.67 |
3 | 0–5 | 1.82 | 1.11 | 0.00–3.00 | 1.25 | 0.69 |
Sympathy | Control | 1 | 0–4 | 1.50 | 1.03 | 0.00–3.00 | 1.20 | 0.67 |
2 | 0–5 | 1.59 | 1.25 | 0.00–3.50 | 1.17 | 0.70 |
3 | 0–5 | 1.40 | 1.27 | 0.00–4.00 | 1.19 | 0.77 |
Sympathy | Label | 1 | 0–4 | 1.76 | 1.04 | 0.00–4.00 | 1.27 | 0.88 |
2 | 0–5 | 1.50 | 1.22 | 0.00–3.50 | 1.15 | 0.73 |
3 | 0–5 | 1.50 | 1.21 | 0.00–3.00 | 1.22 | 0.74 |
Sympathy | Continuum | 1 | 0–5 | 1.55 | 1.12 | 0.00–2.50 | 1.23 | 0.69 |
2 | 0–5 | 1.50 | 1.40 | 0.00–4.50 | 1.21 | 0.93 |
3 | 0–5 | 1.48 | 1.32 | 0.00–3.00 | 1.23 | 0.85 |
Anger | Control | 1 | 0–2 | 0.63 | 0.79 | 0.00–2.50 | 0.61 | 0.75 |
2 | 0–4 | 0.62 | 0.95 | 0.00–2.50 | 0.47 | 0.64 |
3 | 0–3 | 0.56 | 0.83 | 0.00–3.00 | 0.65 | 0.75 |
Anger | Label | 1 | 0–3 | 0.81 | 1.01 | 0.00–2.50 | 0.60 | 0.61 |
2 | 0–5 | 0.63 | 0.95 | 0.00–2.50 | 0.58 | 0.62 |
3 | 0–4 | 0.60 | 0.89 | 0.00–4.00 | 0.52 | 0.75 |
Anger | Continuum | 1 | 0–3 | 0.79 | 0.89 | 0.00–4.00 | 0.93 | 0.83 |
2 | 0–5 | 1.11 | 1.27 | 0.00–5.00 | 0.89 | 1.06 |
3 | 0–5 | 0.88 | 1.14 | 0.00–4.00 | 0.97 | 1.01 |
Discomfort | Control | 1 | 0–4 | 1.38 | 1.14 | 0.00–3.50 | 0.76 | 0.80 |
2 | 0–4 | 1.17 | 1.09 | 0.00–2.50 | 0.60 | 0.56 |
3 | 0–4 | 0.92 | 1.20 | 0.00–2.50 | 0.78 | 0.63 |
Discomfort | Label | 1 | 0–4 | 1.32 | 1.04 | 0.00–3.50 | 1.15 | 0.97 |
2 | 0–4 | 1.13 | 1.15 | 0.00–3.50 | 1.05 | 0.82 |
3 | 0–4 | 1.17 | 1.07 | 0.00–2.50 | 0.83 | 0.73 |
Discomfort | Continuum | 1 | 0–5 | 1.38 | 1.41 | 0.00–2.50 | 1.09 | 0.85 |
2 | 0–5 | 1.59 | 1.53 | 0.00–5.00 | 1.10 | 0.99 |
3 | 0–5 | 1.46 | 1.48 | 0.00–3.50 | 0.88 | 0.82 |
ANOVA analyses
Eight 3 × 3 × 2 mixed between-within subjects ANOVAs were carried out to examine the impact of condition (between subjects, three levels: Control Group, Label Group, Continuum Group), time (within-subjects, three levels: Time 1, Time 2, and Time 3), and participant gender (between subjects, two levels: male female), on the four BCA scores, and the four WCA scores.
The initial focus of our analysis for each hypothesis was on the three-way interaction to determine, at the outset, whether a condition*time interaction varied as a function of gender on the basis of findings of previous research [
23,
24]. In relation to our first three hypotheses, our focus was on the two-way interaction between condition*time. The primary evidence for categorisation is present in the difference between BCA and WCA scores as we move from an uncategorised state (Time 1) to a categorised state (Time 2). It was hypothesised that a significant Condition*Time interaction would capture that this effect is most prominent in the Label Group, followed by the Continuum Group, and non-existent in the Control Group. It was hypothesised that a significant Condition*Time interaction would also indicate that this effect is reduced at Time 3 when labels are no longer present. The false discovery rate was again used to adjust significance levels [
32] yielding adjusted
p values of between < .006 and < .05). Only statistics that are significant at the adjusted levels are reported here.
One three-way interaction was significant (WCA sympathy), F(4, 300) = 3.52, p = .008. The dataset was thus split for male and female participants and the Condition*Time interaction was interpreted. However, there was not a significant Time*Condition interaction for either males or females, nor was there a significant main effect for Time, or Condition. No significant interactions were observed between Condition*Time, Condition*Gender, or Time*Gender on any of the eight dependent variables. Finally, on inspection of the main effects, three significant main effects were identified. For BCA mental health scores, there was a main effect for Time, F(2, 149) = 4.07, p = .019. Tukey post hoc comparison indicates that, regardless of gender and condition, all participants had marginally lower (p = .06) BCA Mental Health scores at Time 1 (M = 1.62) compared to Time 2 (M = 1.84) and significantly lower (p = .016) than Time 3 (M = 1.89). For WCA anger scores there was a significant main effect for condition, F(2, 150) = 5.44, p = .005. Tukey post hoc analysis indicated that regardless of Time and Gender, the Continuum Group had significantly higher scores (M = .936) than the Control Group (M = .577) and the Label Group (M = .567), p = .016 and .012 respectively. For WCA discomfort scores, there was a main effect for Condition, F(2, 150) = 4.29, p = .015. Tukey post hoc analysis indicated that the Control Group had significantly lower WCA discomfort scores (M = .710) than the Label Group (M = 1.01) or the Continuum Group (M = 1.02), p = .034 and .026 respectively.