In their target article, Savage et al. gather numerous cross-disciplinary findings supporting the hypothesis that musicality evolved for the overarching purpose of social bonding. While acknowledging the large interconnectedness between musicality and events of social bonding, we propose a different conceptualization of the evolutionary mechanisms that led humans to develop a preference for music. Given the high opportunity costs and resources necessary for the evolution and practice of a complex trait-like musicality, one would not only have to argue that music is an effective tool for social bonding, but that its specific features are necessary and most efficient in building cohesive bonds, in the presence of feasible and less resource intensive alternatives. We argue that natural selection favored individuals who tend toward building social bonds in general, and that, therefore, the formation of social groups has revolved around numerous trivial aspects of human life. This becomes apparent in looking at studies of the so-called minimal group paradigm (Tajfel, Reference Tajfel1978), in which such arbitrary rules as a coin toss determining group affiliation were sufficient to create in-group favoritism (Tajfel, Reference Tajfel1978). Following that line of reasoning, the question of why musicality evolved broadens to the question of why there is esthetic appreciation at all. As Savage et al. point out, reactions in the dopaminergic reward system are elicited by sensory information of every domain allowing for the build-up, validation, and violation of expectations, thereby referring to predictive processing models of human perception and action (Clark, Reference Clark2013).

Being able to interact with, adapt to and predict future changes in the environment will have constituted the crucial advantage of running and maintaining a perceptual system. For that, a cognitive agent has to acquire an internal model of one's surroundings that is constantly generating new predictions and matching those predictions with current sensory evidence. In order to achieve that in an energy efficient fashion, stimuli that allow for confirmations of prior predictions as well as deviations from them are necessary for optimal learning (Sarasso, Neppi-Modona, Sacco, & Ronga, Reference Sarasso, Neppi-Modona, Sacco and Ronga2020), and are, therefore, preferred over highly predictable or unpredictable stimuli (Van de Cruys, Reference Van de Cruys2017). We can find this particular relationship between stimulus complexity and likability across multiple sensory domains (Essick et al., Reference Essick, McGlone, Dancer, Fabricant, Ragin, Phillips and Guest2010; Gold, Pearce, Mas-Herrero, Dagher, & Zatorre, Reference Gold, Pearce, Mas-Herrero, Dagher and Zatorre2019; Van de Cruys & Wagemans, Reference Van de Cruys and Wagemans2011). Perception in this regard can be considered as a form of statistical inference learning, in which intermediate stimulus complexity allows for optimal learning rates and is, therefore, valued (Erle, Reber, & Topolinski, Reference Erle, Reber and Topolinski2017).

Further support for this conceptualization comes from neuroimaging studies. Here, activity in regions such as the orbitofrontal cortex (OFC) has been associated with the attribution of salience to a presented stimulus, and activity in striatal regions such as the caudate, putamen, and nucleus accumbens has been associated with the discrepancy between expected and observed outcomes (reward prediction error). Elevated levels of activity in both those regions can be found during probabilistic reward learning tasks as well as during listening to pleasurable music (Koelsch, Reference Koelsch2020; Werlen et al., Reference Werlen, Shin, Gastambide, Francois, Tricklebank, Marston and Walton2020). Furthermore, temporally more fine-grained measures of neural activity such as the electroencephalogram (EEG) suggest that the process of listening to music is characterized by a constant build-up of musical expectations based on recent harmonic context and integration of currently perceived acoustic inputs into this context (Koelsch, Gunter, Friederici, & Schröger, Reference Koelsch, Gunter, Friederici and Schröger2000). During this process, the experienced tension and esthetic pleasure derived from music is not only dependent on the validation, but also the violation of formed expectations (Salimpoor, Zald, Zatorre, Dagher, & McIntosh, Reference Salimpoor, Zald, Zatorre, Dagher and McIntosh2015). The latter would be implausible if the syntactical structure of music had evolved mainly for the purpose of synchronizing and coordinating musical group performances, as stated by Savage et al. in their target article.

Moreover, this conceptualization of musicality is not limited to one's emotional status being informed by current rates of prediction error in an act of passive listening. It is rather embedded in an action–perception cycle in which not only the generation of, but also the consumption of music is an enactive process. This process is characterized by the constant generation of new predictions, attentional shifts selecting for the currently most informative aspects of music as well as the attempt to reduce prediction error rates via actions such as dancing (Koelsch, Vuust, & Friston, Reference Koelsch, Vuust and Friston2019). Furthermore, the emotional status derived from current prediction error rates could in turn modulate subsequent perceptual processing, that is, by how much weight is ascribed to prior predictions and current sensory inputs, respectively (Kraus, Niedeggen, & Hesselmann, Reference Kraus, Niedeggen and Hesselmann2021).

According to this conceptualization of musicality, deriving pleasure from mere perception of sensory stimuli would not be a “useless by-product of other capacities” but a direct implication of a perceptual system that has evolved to run an internal model of its environment which needs to be updated constantly (Friston & Friston, Reference Friston and Friston2013). This principle is also mirrored in phenomena such as the Pandora effect (Hsee & Ruan, Reference Hsee and Ruan2016), where participants are willing to engage in self-harming behavior solely for the purpose of uncertainty reduction. Note that this framework derives not only a plausible explanation for musicality (i.e., the tendency to intentionally produce and perceive specific auditory stimuli), but also for esthetic appreciation in other sensory domains (e.g., visual art or fine dining) from basic evolutionary demands that were placed on cognitive agents. This conceptualization of why humans evolved to create, consume, and value music is still compatible with all evidence put forward linking social bonding to musicality, but it questions the implied causality.