What people see predicts what people do (Gibson, Reference Gibson1979). However, when researchers test connections between perception and action in contexts or when using paradigms that lack ecological validity, the utility of their conclusions is suspect. Aligning with Cesario's perspective, researchers across multiple fields that involve eye-tracking have highlighted discrepancies that emerge as a function of the presence – or rather, lack – of ecologically valid testing procedures. In the early 1900s perceivers wore contact lenses with attached pointers. In the 1930s, perceivers sat with their chin, forehead, and back of their head affixed to a metal frame attached to the desk. By the 1990s, perceivers wore bulky headgear with relatively large suspended cameras. These techniques thwart attempts to capture natural viewing experiences. Some developmental psychologists, aware of these limitations, created mobile eye-tracking, in which small cameras are affixed to baseball caps or eyeglasses that toddlers, children, and adults wear (Franchak, Kretch, Soska, & Adolph, Reference Franchak, Kretch, Soska and Adolph2011). Using this technology, researchers found that infants spent far less time looking at their mother's face during social interactions (Franchak et al., Reference Franchak, Kretch, Soska and Adolph2011; Jung, Zimmerman, & Pérez-Edgar, Reference Jung, Zimmerman and Pérez-Edgar2018), than previous research had concluded – research that used equipment requiring infants sit immobile at desk-mounted eye-trackers (Soska, Adolph, & Johnson, Reference Soska, Adolph and Johnson2010).

When industrial research adopted ecologically valid eye-tracking technology, they discovered previous conclusions had been wrong too. When eye-tracking allowed pilots to engage with the simulation screen freely without imposing restrictions on head and body movements, researchers found that expert pilots in cockpits allocated attention in ways that gathered critical information during tactical operations (Li, Chiu, Kuo, & Wu, Reference Li, Chiu, Kuo and Wu2013; Pérez-Edgar, MacNeill, & Fu, Reference Pérez-Edgar, MacNeill and Fu2020). Previously, the field had relied on restrictive eye-tracking technology with chinrests, and reported pilots failed to attend to necessary information (Sulzer & Skelton, Reference Sulzer and Skelton1976).

In our social cognition lab, we monitor eye movements using infrared sensors embedded into the frame of a typical-looking monitor, which records eye movements without awareness and as individuals freely move their heads and torsos within a 42° space. Under these naturalistic viewing experiences, we discovered that it was only among participants who frequently attended to an officer in a police–civilian altercation that pre-existing feelings of identification with police influenced legal decisions (Granot, Balcetis, Schneider, & Tyler, Reference Granot, Balcetis, Schneider and Tyler2014). These data reconciled discrepancies between empirical studies and real trial data in the courts that have been inconclusive as to whether people punish outgroup members more harshly (Anwar, Bayer, & Hjalmarsson, Reference Anwar, Bayer and Hjalmarsson2012), more leniently (McGowen & King, Reference McGowen and King1982), or without bias (Mazzella & Feingold, Reference Mazzella and Feingold1994).

We also agree with Cesario that removing social context undermines ecological validity. Context offers a metaphorically thicker rather than thinner slice of information that informs understanding. Social context comes in many forms including the dimension of time. When we incorporated time by presenting dynamic rather than static visual scenes of the police aggression, we discovered individual differences in attention patterns that predicted why and when bias in police punishment decisions emerged (Granot et al., Reference Granot, Balcetis, Schneider and Tyler2014).

However, we disagree with Cesario's conclusion that decision-makers including police and teachers respond without prejudice to the behaviors exhibited by individuals with whom they engage. In his “Missing Contingencies Flaw” tenet, Cesario argues that behaviors presented to decision-makers differ. This reflects an error of naïve realism. We argue people do respond with bias to the same stimulus because they do not perceive the same stimulus the same way in all cases. Ample evidence finds that individuals experience idiosyncratic perceptual experiences for at least two reasons. First, the demands on attention are greater than the attentional resources perceivers have available. As a result, attention is selective and directed (Broadbent, Reference Broadbent and Broadbent1958; John, Bartlett, Shimokochi, & Kleinman, Reference John, Bartlett, Shimokochi and Kleinman1973); given that attention drives visual experience, differences in attention produce differences in perception (Mack & Rock, Reference Mack, Rock and Wright1998). Moreover, differences in attention are systemic and vary as a function of characteristics of perceivers themselves. For instance, individuals attend to sources of threat, particularly when threats are members of a social outgroup (Koster, Crombez, Van Damme, Verschuere, & De Houwer, Reference Koster, Crombez, Van Damme, Verschuere and De Houwer2004). Indeed, White participants fixated longer on the civilian compared to an officer when viewing video evidence of both engaged in a physical altercation; though Asian participants did too, they showed significantly greater parity than did White participants (Sternisko, Granot, & Balcetis, Reference Sternisko, Granot and Balcetis2017). Moreover, greater visual attention on the officer increased the severity of punishment leveraged against him. Although the stimulus participants responded to was the same, the manner in which individuals engaged attention varied systematically, resulting in differences in perceptual interpretation.

Second, because foveal view where details are encoded with great clarity constitutes only a small subset of the field of vision, most information that enters the visual system is processed through peripheral vision which is specialized for detecting movement, but little else (Fairchild, Reference Fairchild2005). As a result, much of visual input is ambiguous and idiosyncratically interpreted. Indeed, even when presented with the same line drawing, individuals reached markedly different understandings of what they saw as a function of what they had previously been thinking about (Balcetis & Dunning, Reference Balcetis and Dunning2006). Moreover, when encouraged to reflect on their visual experience and reconstrue their percept in alternate ways, they could not. This suggests their initial biases affected encoding in such a way that visual cues relevant to reinterpretation were removed.

In other work, even when attention was experimentally held constant and participants attended to the same social target, prior attitudes biased perception of the attended stimulus (Granot et al., Reference Granot, Balcetis, Schneider and Tyler2014). Even when focusing attention to the officer, perceivers differed in the degree to which they believed they saw the officer initiate physical contact, search the civilian, display a weapon, and pursue the civilian. These are objective, discrete, observable behaviors, but perceivers differed in the degree to which they believed they saw them happen even when attending to the same social target.

While we agree with Cesario there are limitations in extrapolating real-world consequences from experimental findings given that differences in ecological validity impact outcomes, it is an error to assume that all individuals perceive the same stimulus the same way.