Access to perceptual ambiguity: Behavioral, metacognitive, and physiological measures of perception under conditions of uncertainty – AMBISENSE

The perception of our environment usually comes with a sense of how uncertain or certain we are about what we perceive. In some cases, however, stimuli that are objectively ambiguous are not acknowledged as such, raising the question of whether uncertainty is not processed at all in these cases or whether instead, uncertainty signals might be computed but inaccessible to conscious awareness. Our study tests participants hearing intervals of two consecutive Shepard tones, including a tritone interval which can be heard as going up or down depending on the context. Participants had to categorize the interval heard, report judgments of confidence, and they were equipped with an eye-tracker measuring pupil size.

In two experiments, behavioral responses showed that listeners could not explicitly access the ambiguity in this stimulus, even though their responses varied from trial to trial. However, pupil dilation was larger for the more ambiguous cases. The ambiguity of the stimulus for each listener was indexed by the entropy of behavioral responses, and this entropy was also a significant predictor of pupil size. In particular, entropy explained additional variation in pupil size independent of the explicit judgment of confidence in the specific situation that we investigated, in which the two measures were decoupled. Our data thus suggest that stimulus ambiguity is implicitly represented in the brain even without explicit awareness of this ambiguity.

Previous authors have pointed out that instead of revealing mechanisms involved in resolving ambiguity, pupil dilation may rather reflect the consequence of this resolution, such as awareness of a perceptual change or subsequent motor responses. Since our participants were neither told of the potential ambiguity of the stimulus, nor asked to wait for a switch in their percept, and since they were unaware of the stimulus uncertainty, our paradigm suggests that rather than reflecting the consequence of ambiguity resolution, our results may help characterize its neural underpinnings. It remains an open question how stimulus ambiguity becomes accessible to consciousness and why it sometimes remains altogether hidden from awareness.

«An implicit representation of stimulus ambiguity in pupil size«. PNAS 118 (48). November 2021.

www.pnas.org/doi/10.1073/pnas.2107997118

Perception is an ill-defined problem: at any moment in time, there is not enough sensory information to uniquely and completely specify the state of the outside world. Much of what we sense is noisy, partial, and susceptible to rival interpretations. Well-known cases of sensory ambiguity lead to visual illusions such as the Necker Cube and auditory illusions such as the Shepard illusion. More recent examples, which became viral on social media, include “natural” stimuli like the Dress for the visual domain (see Brainard and Hurlbert 2015), which was a photograph of a dress that could be perceived as either black-blue or white-gold, and Laurel-Yanny in the auditory domain (see Pressnitzer et al. 2018), which was a speech utterance that could be heard as either Laurel or Yanny. In both cases, a poor-quality photograph or a poor-quality recording gave rise to very distinct percepts across participants, with apparently little awareness of the profoundly ambiguous nature of the available sensory information.

While perceptual multistability is a classic issue in the philosophy of perception and in cognitive science, less attention has been given to the processes that can make ambiguity either accessible or inaccessible to consciousness. The recent cases provide striking examples of inaccessibility: while most subjects are quickly aware of the ambiguity of the Necker cube, and can exert voluntary control on their percept to some extent (Long & Toppino 1998), the reason the Dress and Laurel-Yanny became so popular appears largely due to the inability of most observers to access the alternative interpretation of other observers (Pressnitzer et al. 2018).

In the current project, we propose to investigate the nature and conditions of access to stimulus ambiguity. We will examine three main factors: the characteristics of the stimulus; its context; the expertise of observers. We will also collect different measures of access to ambiguity: behavioral performance; explicit judgments of confidence; implicit physiological correlates of confidence such as pupil dilation and reaction times.

This deeply interdisciplinary project has wide ranging implications for the vast psychophysical literature on perceptual multistability, as we believe that some but not all multistable stimuli block access to ambiguity (Schwartz et al. 2012). It is also relevant to long-standing philosophical debates on the nature of perception and vague judgments (see Egré 2013, 2017). Finally, practical applications exist, as we predict that we may be able to train people to gain access to stimulus ambiguity, thus improving their metacognitive performance and possibly their first-order performance on difficult perceptual tasks.