Inter-subject representational similarity analysis reveals individual variations in affective experience when watching erotic movies

Inter-subject representational similarity analysis reveals individual variations in affective experience when watching erotic movies. Pin-Hao A. Chen, Eshin Jolly, Jin Hyun Cheong, Luke J. Chang. bioRxiv, Aug 6  2019. https://doi.org/10.1101/726570

Abstract: We spend much of our life pursuing or avoiding affective experiences. However, surprisingly little is known about how these experiences are represented in the brain and if they are shared across individuals. Here, we explore variations in the construction of an affective experience during a naturalistic viewing paradigm based on subjective preferences in sociosexual desire and self-control using intersubject representational similarity analysis (IS-RSA). We found that when watching erotic movies, intersubject variations in sociosexual desire preferences of 26 heterosexual males were associated with similarly structured fluctuations in the cortico-striatal reward, default mode, and mentalizing networks. In contrast, variations in the self-control preferences were associated with shared dynamics in the fronto-parietal executive control and cingulo-insula salience networks. Importantly, these results were specific to the affective experience, as we did not observe any relationship with variation in preferences when individuals watched neutral movies. Moreover, these results appear to require multivariate representations of preferences as we did not observe any significant results using single summary scores. Our findings demonstrate that multidimensional variations in individual preferences can be used to uncover unique dimensions of an affective experience, and that IS-RSA can provide new insights into the neural processes underlying psychological experiences elicited through naturalistic experimental designs.


---
Discussion

In this study, we used brain imaging to explore the affective experience of heterosexual male participants. We used a naturalistic experimental design (Hasson et al., 2004; Haxby et al., 2011), in which participants watched short clips of erotic and neutral movies in the MRI scanner. These types of designs are ideal for eliciting powerful psychological experiences and creating strong variation in brain activity underlying the experience (Jolly and Chang, 2019). In contrast to the standard practices in emotion research, we did not examine affective experience using self-reported feelings or videos selected to elicit specific emotional states (Coan et al., 2007; Lench et al., 2011; Lindquist et al., 2012; Quigley et al., 2014). Instead, we explored how variation in two distinct preferences (i.e., sociosexual desires and self-control) mapped onto individual variation in brain dynamics while watching the videos using intersubject representational similarity analysis (IS-RSA).

Consistent with our predictions, we found that when individuals watched erotic movies, individuals with similar sociosexual desire preferences showed higher similarities in patterns of neural dynamics in brain regions within the cortico-striatal reward and default mode and mentalizing networks than those with different preferences. In contrast, as individuals became closer in their self-control preferences, we observed greater similarities in patterns of neural dynamics in brain regions within the fronto-parietal executive control and cingulo-insula salience networks. Importantly, when individuals watched neutral movies, inter-subject similarities in sociosexual desire and self-control preferences played no prominent role in accounting for the similarities in patterns of neural dynamics. We used meta-analytic decoding to provide a crude reverse inference of the possible psychological states contributing to the affective experience. Consistent with our expectations, variations in sociosexual desire preferences revealed stronger associations with social and social judgment topics, whereas variations in self-control preferences revealed stronger associations with the executive control, cognitive control, error monitoring and conflict. Together, our results support our hypothesis that variation in individual preferences can be used to explore affective experiences. Though we only specifically examined preferences for sociosexual desire and self-control, we do not believe this to be an exhaustive list of possible preferences and speculate that many other potential measures might also provide insight into this experience. This study provides important conceptual and methodological advances to the investigation of affective experiences. Because there currently exists no objective measure of affective experiences (Chang et al., 2015), the field of emotion has a long history of grappling with measurement issues and has largely relied on self-report (Larsen and Fredrickson, 1999). One issue with trying to have participants map an experience into a high dimensional space of self-reported feelings is that this process requires both introspection (Nisbett and Wilson, 1977) and verbally labeling feelings using shared concepts (Lindquist et al., 2015). It’s possible that this verbal labeling process necessarily reduces the dimensionality of the representational space of the experience by filtering out processes that cannot be measured using this approach, which is why many studies find that 2-5 dimensions can explain the majority of the emotion rating variance (Chikazoe et al., 2014; Kragel and LaBar, 2015; Skerry and Saxe, 2015). In addition, most studies select a few stimuli to elicit a finite set of emotional states. However, this approach assumes that all participants will have a similar experience (Chang et al. 2018) and limits the variation in the emotional experiences (Cowen and Keltner, 2017), which can provide a statistical bias towards a low dimensional representation (Jolly and Chang, 2019). Our study provides an alternative approach to exploring affective experiences. Rather than assuming that participants will have the same response, which provides the basic premise of intersubject correlation (Hasson et al., 2004) and also functional alignment techniques (Guntupalli et al., 2016; Haxby et al., 2011), we assume that participants will have strong variations in their experience, which should correspond to structured variations in measures related to the experience. Importantly, we do not attempt to reduce the dimensionality of these measures to a single summary score, instead, we represent each item from the measure as a separate axis in a multiple-dimensional space and calculated the pairwise distance of each participant in this high-dimensional space. We believe that preserving the richness and complexity of all features in a high-dimensional space is important as there are many ways to answer a questionnaire that will produce an identical single summary score. Consistent with this intuition, we find that our IS-RSA results only hold when using a high dimensional representation and are not present when mapping participants’ distances using summary scores.

Though we believe this IS-RSA approach to be promising, there are several important limitations that should be acknowledged. First, we are using all of the features of each preference measure as an axis to map each participant and weighting the contribution of each feature equally. This means that we currently are unable to determine which features are specifically contributing to the experience. In addition, some of these features could be reflecting pure noise, which would be weighted equally as features that contain pure signal. It’s possible that this might be addressed with future work using multivariate regression techniques (e.g., partial least squares). Second, we are mapping individual position in this multidimensional space to intersubject dynamics of brain activity. This means that we do not know when in time processes specific to the experience occurred. It is possible to use similarity in spatial representations (van Baar et al., 2019), which might provide a way to extend this to which time points show a similar intersubject structure (Chang et al. 2018). However, this will also require accounting for multiple comparisons as well as non-independence in the time-series signals resulting from autocorrelation.

In summary, we have provided a demonstration of how variations in participants’ preferences can be used to uncover unique dimensions of an affective experience based on similarity in the intersubject structure of brain dynamics measured during the experience. This technique has the potential to provide a new approach to studying the neural processes underlying psychological experiences elicited through naturalistic experimental designs. Though this study provides a simple proof of concept, we hope that this work will inspire future innovations in analyzing naturalistic experimental designs, affective science, and psychological experiences.