A risk of tautology

With respect to Pitfall 6 (sect. 4.6), F&S expose their thesis to the risk of being an unfalsifiable tautology. One could always preserve it from falsification by claiming that if any top-down factor influences x (any attribute provisionally taken as truly perceptual) then x does not belong to true “front-end” perception (Lyons Reference Lyons2011). This logical difficulty is overcome if perception is defined as an explanandum independent of its determinants (explanantia). By calling “perception (and, less formally, seeing)…both (typically unconscious) visual processing and the (conscious) percepts that result” (sect. 1, para. 3) and by taking front-end stimulus processing as the only truly perceptual stage, F&S mix the denial of the explanandum (any claimed top-down influence on perceptual experience is greatly exaggerated) and the rejection of not truly perceptual explanantia.

Categorical perception

The existence of categorical perception (Goldstone & Hendrickson Reference Goldstone and Hendrickson2010) seems to contradict F&S. This subfield is operationally defined by the warping of perceived similarity space, such that – once a categorical boundary is established – discrimination is better across categories than within categories. Categorical perception can be acquired during individual experience (Beale & Keil Reference Beale and Keil1995), and it constitutes a paradigmatic case of observer-dependent, familiarity-mediated sensitivity change.

Tertiary qualities

Equally problematic for the thesis F&S defend is the domain of tertiary qualities as defined in the Gestalt literature (Köhler Reference Köhler1938; Metzger Reference Metzger1941; Sinico Reference Sinico2015; Toccafondi Reference Toccafondi, Centi and Huemer2009). Consider the influence of observer's states on perceived facial expressions (not a revolutionary effect, given the naive idea that beauty is in the eye of the beholder). Despite being commonly conceived as subjective, expressive qualities are phenomenally objective (i.e., perceived as belonging to the object; Köhler Reference Köhler1929; Reference Köhler1938) and show a remarkable – though not exclusive – dependence on configural stimulus properties. Therefore, assessing observer-dependent effects on tertiary (in particular, expressive) qualities contributes significantly to perceptual science, which can tolerate, we believe, a circumscribed leakage of cognition into perceptual apartments, consistent with grounded cognition (Barsalou Reference Barsalou2010; Kiefer & Barsalou Reference Kiefer, Barsalou, Prinz, Beisert and Herwig2013).

Penetrability of perceived facial expressions

To remain “empirically anchored” (sect. 4, para. 1), consider our studies on the effects of comfort and discomfort of bodily actions on perceived facial expressions of emotion. Using a novel motor action mood induction procedure (MAMIP), Fantoni and Gerbino (Reference Fantoni and Gerbino2014; Gerbino et al. Reference Gerbino, Manzini, Rigutti and Fantoni2014) demonstrated a congruency effect in participants who performed a facial emotion identification task along a happy-to-angry morph continuum after a sequence of visually guided reaches: A face perceived as neutral in a baseline condition appeared slightly happy after comfortable actions and slightly angry after uncomfortable actions. In agreement with F&S (sect. 4.2.3), we considered such evidence insufficient to claim that bodily action influences affective perception in a within-cycle fashion (better than “top-down,” we deem): An action-induced transient mood might shift the point of subjective neutrality in identification by influencing only postperceptual (not perceptual) processing. In a subsequent study (Fantoni et al. Reference Fantoni, Rigutti and Gerbino2016), we corroborated the perceptual nature of bodily action effects using an emotion detection task. Rather than response bias changes attributable to cognitive set shifts, MAMIP produced systematic, mood-congruent sensitivity changes in the detection of both positive and negative target emotions, with constant 0.354 d′ increments (p = 0.000) in congruent (comfortable-happiness, uncomfortable-anger) over incongruent (uncomfortable-happiness, comfortable-anger) conditions at increasing percent emotion in the morph.

Referring to the final checklist F&S propose (sect. 5.2), our facilitation-by-congruency effect on emotion detection hold the following properties:

1. 1. It is robust (relative to the Uniquely Disconfirmatory Predictions criterion) and immune to the El Greco fallacy; the detection threshold for happiness decreased about 2.2% after a sequence of congruent-comfortable than incongruent-uncomfortable reaches (p = 0.001), and the detection threshold for anger (in a different group of observers, to control for carryover effects) decreased about 7.4% after a sequence of congruent-uncomfortable than incongruent-comfortable reaches (p = 0.003);

2. 2. It is operationalized by shifts of performance-based measures (d′, absolute threshold, just noticeable difference) consistent over different paradigms (identification and detection);

3. 3. It occurs even in the absence of response bias shifts, given that the response criterion c did not change significantly across congruency conditions for both happiness and anger detection;

4. 4. It is dependent on the observer's internal states induced by motor actions unrelated to targets to be detected;

5. 5. It is independent of peripheral attention; the performance improvement produced by uncomfortable actions, inducing high arousal, is selective (i.e., they improve anger, not happiness, detection);

6. 6. It is independent of memory in the sense that the detection of subtle signs of happiness or anger in unfamiliar faces likely involves a general perceptual ability, independent of episodic memories of specific individuals.

To summarize, our evidence challenges the bold claim that cognition does not affect perception, supports a circumscribed penetrability of affective perception, and suggests indirect affective priming as a candidate mechanism by which observer's states mold perceptual properties experienced as phenomenally objective and yet loaded with meaning. If brains resonate with the similarity of motor and affective states (Niedenthal et al. Reference Niedenthal, Mermillod, Maringer and Hess2010), bodily actions may behave as affectively polarized primes that preactivate the representations of emotionally related facial features, thus facilitating the encoding of features belonging to the same, rather than to a different, valence domain.