In 2013, Leila Selimbegović & Armand Chatard observed that the mere presence of a mirror could make suicide thoughts more accessible.
Back in 2017, I was looking for supervisors for my PhD thesis. My personal interests were quite clear in my head back then: I was interested in attitudes, values, construal levels,… but most importantly, I wanted to study self-related stuff. So I had a list of potential authors in mind. Leila and Armand struck me as interesting authors with two papers I found particularly interesting: Chatard & Selimbegović, 2011, a study suggesting failure and self-awareness could interact to predict increased escape-related thoughts; and Selimbegović & Chatard, 2013, suggesting that every time I brushed my teeth in front of a mirror, suicide-related thoughts might be easier to retrieve from memory!
This study’s theoretical framework implied that self-awareness would make aversive self-to-standards comparisons more salient (Duval & Wicklund, 1972) and thus would ensue a motivation to escape from the state of self-focused attention. Suicide being one means among others to escape from the self, it followed that suicide-related thoughts could be made more accessible by a mere effect of goal-means activation (Baumeister, 1990; Kruglanski et al., 2002).
Something really strange appeared (or more precisely, failed to appear) in this study: self-discrepancies were manipulated but played no role in the mirror effect. Indeed, what the authors found was that making self-discrepancies salient (vs. not making these salient) dit not increase the effect. One would think self-discrepancies should be a mediator (this is what would be expected from the theoretical framework I mentioned). The study did not allow for a clear mediation analysis (which are quite complicated as suggested by Bullock et al., 2010), but at least, if the theory held, and a mediation existed, one would expect the effect of self-awareness to be equal or inferior to the effect of the direct cause of increased escape-thought accessibility – that is, self-discrepancies saliency. But that’s not what the data suggested. I hence came to Leila & Armand with a PhD project suggesting mechanisms to this effect. Long story short, it involved perceiving one as an object (objective self-awareness [“Me“], vs. subjective self-awareness [“I“], Duval & Wicklund, 1972) causing a feeling of shame leading to escape motivation.
I was aware of the replication crisis (and so did my new supervisors) back then, and hence, the first step of the thesis was quite straightfoward: replicating the mirror effect in a direct fashion, with a pre-registered analysis plan. It was my first steps in the realm of meta-science.
The mirror effect: Self-awareness alone increases suicide thought accessibility - ScienceDirect Leila and a mirror, 2013
Me, my hair loss and and mirror, 2020
METHODS SECTION
The study was registered here: https://osf.io/v6bhx
We conducted a registered direct replication of Selimbegović and Chatard (2013), but with the addition of measures of shame at the very end of the study (in order to not disrupt the direct aspect of the replication). We measured shame with a direct measure (SSGS, Marschall et al., 1994) and also with an indirect measure (an Implicit Association Task assessing the association between the self and shame).
The first issue we bumped into when analyzing the data collected in our replication was the fact that we failed to predict our data would be far from normal. Looking back, it seems obvious that reaction times would not be normally distributed. But, as stated in a popular tweet:
[insert Rodeo tweet here]
Following the registered analytical path, we failed to replicate the mirror effect. Moreover, we did not fund evidence for a moderation of shame and guilt (regardless of whether the measure was direct or indirect). At this point a serious revision of my expectations had to be made.
Despite this failure to directly replicate the original finding, the study was, however, a great experience to me. During the process, I discovered pre-registration, robust statistics, and equivalence testing. The effect size we found in our replication was not significantly different from the smallest effect size we deemed interesting. More substantially maybe, we explored the data and found that reducing the sample size by excluding participants deviating from the median by more than 2 MAD yielded a successful replication on one of the two expected hypothesis (negative RT as a covariant). Crucially, 2 MAD is one quite common cut-off suggested by Leys et al. (2013). Using a multiverse approach, we found that the more the sample was reduced, the larger the effect size (see figure 1).
Figure 1. Small multiverse analysis investigating the effect of outliers’ cut-off on effect sizes
However, these were exploratory analyses failing to give evidence for the effect. I revised my priors regarding the mirror effect accordingly (that might be understood from the tone in the paper published at Collabra: Psychology, a journal curating the world with open access articles and a very interesting business model).
Conducting studies allows for a constant revision of one’s priors. As such, failure to replicate constitute an important progress. It is perfectly okay to lose confidence, to the extent that it is the learning process that matters. As always, several aspects can be questioned to restore some consistency between the original study and the failed replication. Crucially, it is complicated to register a perfect analysis plan. However, preregistration is important to transparently indicate when one threads away from a theory-driven confirmatory approach toward a data-driven exploration.
References
-
Baumeister, R. F. (1990). Suicide as escape from self. Psychological Review, 97(1), 90–113. https://doi.org/10.1037/0033-295x.97.1.90
-
Bullock, J. G., Green, D. P., & Ha, S. E. (2010). Yes, but what’s the mechanism? (don’t expect an easy answer). Journal of Personality and Social Psychology, 98(4), 550–558. https://doi.org/10.1037/a0018933
-
Chatard, A., & Selimbegović, L. (2011). When self-destructive thoughts flash through the mind: Failure to meet standards affects the accessibility of suicide-related thoughts. Journal of Personality and Social Psychology, 100(4), 587–605. https://doi.org/10.1037/a0022461
-
Duval, S., & Wicklund, R. A. (1972). A theory of objective self awareness. Oxford, England: Academic Press
-
Kruglanski, A. W., Shah, J. Y., Fishbach, A., Friedman, R., Woo Young Chun, & Sleeth-Keppler, D. (2002). A theory of goal systems. In Advances in Experimental Social Psychology (Vol. 34, pp. 331–378). Academic Press. https://doi.org/10.1016/S0065-2601(02)80008-9
-
Leys, C., Ley, C., Klein, O., Bernard, P., & Licata, L. (2013). Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median. Journal of Experimental Social Psychology, 49(4), 764–766. https://doi.org/10.1016/j.jesp.2013.03.013
-
Monéger, J., Chatard, A., & Selimbegović, L. (2020). The Mirror Effect: A Preregistered Replication. Collabra: Psychology, 6(1), 18. https://doi.org/10.1525/collabra.321
-
Selimbegović, L., & Chatard, A. (2013). The mirror effect: self-awareness alone increases suicide thought accessibility. Consciousness and Cognition, 22(3), 756–764. https://doi.org/10.1016/j.concog.2013.04.014
P.S.: This article has been an opportunity to look back at my first year as PhD student. So I’ll add that I warmly thanks Leila & Armand for their support and for giving me the privilege to work under their supervision.
