Multi-stable perception balances stability and sensitivity

Abstract

We report that multi-stable perception operates in a consistent, dynamical regime, balancing the conflicting goals of stability and sensitivity. When a multi-stable visual display is viewed continuously, its phenomenal appearance reverses spontaneously at irregular intervals. We characterized the perceptual dynamics of individual observers in terms of four statistical measures: the distribution of dominance times (mean and variance) and the novel, subtle dependence on prior history (correlation and time-constant). The dynamics of multi-stable perception is known to reflect several stabilizing and destabilizing factors. Phenomenologically, its main aspects are captured by a simplistic computational model with competition, adaptation, and noise. We identified small parameter volumes (~3% of the possible volume) in which the model reproduced both dominance distribution and history-dependence of each observer. For 21 of 24 data sets, the identified volumes clustered tightly (~15% of the possible volume), revealing a consistent “operating regime” of multi-stable perception. The “operating regime” turned out to be marginally stable or, equivalently, near the brink of an oscillatory instability. The chance probability of the observed clustering was <0.02. To understand the functional significance of this empirical “operating regime,” we compared it to the theoretical “sweet spot” of the model. We computed this “sweet spot” as the intersection of the parameter volumes in which the model produced stable perceptual outcomes and in which it was sensitive to input modulations. Remarkably, the empirical “operating regime” proved to be largely coextensive with the theoretical “sweet spot.” This demonstrated that perceptual dynamics was not merely consistent but also functionally optimized (in that it balances stability with sensitivity). Our results imply that multi-stable perception is not a laboratory curiosity, but reflects a functional optimization of perceptual dynamics for visual inference.

The ‘laws’ of binocular rivalry: 50 years of Levelt’s propositions

“…It has been shown to be rather similar for humans and non-human primates and together with the degree of serial correlation between dominance times it has been an important element in the debate on the relative roles of neural adaptation and neural noise in driving perceptual alternations (Fox & Herrmann, 1967;Moreno-Bote et al, 2007;Pastukhov et al, 2013;Shpiro et al, 2009;van Ee, 2009). On a general level, the distribution's peaked, rather than monotonically decreasing, shape indicates that the probability of a perceptual switch is low during the time period immediately following the previous switch, a temporary stability that may reflect a lack of accumulated adaptation or some other stabilizing mechanism (Kloosterman et al, 2014;Moreno-Bote et al, 2007;Pastukhov et al, 2013).…”

“…It has been shown to be rather similar for humans and non-human primates and together with the degree of serial correlation between dominance times it has been an important element in the debate on the relative roles of neural adaptation and neural noise in driving perceptual alternations (Fox & Herrmann, 1967;Moreno-Bote et al, 2007;Pastukhov et al, 2013;Shpiro et al, 2009;van Ee, 2009). On a general level, the distribution's peaked, rather than monotonically decreasing, shape indicates that the probability of a perceptual switch is low during the time period immediately following the previous switch, a temporary stability that may reflect a lack of accumulated adaptation or some other stabilizing mechanism (Kloosterman et al, 2014;Moreno-Bote et al, 2007;Pastukhov et al, 2013). In addition, while dominance time distributions can be quite variable between individual observers, they do correlate within observers from one paradigm to the next (Carter & Pettigrew, 2003;Shannon et al, 2011;Sheppard & Pettigrew, 2006; see also Baker & Graf, 2009b; but see Gallagher & Arnold, 2014), which is consistent with the recent identification of genetic (Shannon et al, 2011) and neurochemical factors (van Loon et al, 2013) that influence dominance durations across paradigms.…”

“…We therefore see these propositions as consistent with the notion that binocular rivalry alternations depend on a form of adaptation. Note that the above line of reasoning stays essentially unaltered if the role of adaptation is not to make the system actually reach a critical point, but to bring the system gradually closer to a critical point so that random fluctuations become more likely to provide a final push Huguet, Rinzel, & Hupé, 2014;Kang & Blake, 2011;Kim, Grabowecky, & Suzuki, 2006;Moreno-Bote et al, 2007;Pastukhov et al, 2013;Shpiro et al, 2009).…”

Transformation priming helps to disambiguate sudden changes of sensory inputs

“…For example, perceptual adaptation is thought to prioritize perceptual sensitivity (Kohn, 2007;Pastukhov, García-Rodríguez, et al, 2013;Theodoni, Kovács, Greenlee, & Deco, 2011;Webster, 2011). Conversely, neural persistence (Coltheart, 1980;Loftus & Irwin, 1998;Pastukhov & Braun, 2013b) and a rolling average over a longer period of perceptual history are thought to be used to minimize influence of neural noise and ensure perceptual constancy Cicchini et al, 2014;Fischer & Whitney, 2014).…”

“…It is the ubiquity of these events that raises the question of how the visual system resolves their dynamic ambiguity. The general answer to the problem is to gather and exploit prior knowledge (Friston, Breakspear, & Deco, 2012;Gregory, 2009;Metzger, 2009;Yuille & Kersten, 2006), and this process has been well studied from both behavioral (Kristjánsson http & Campana, 2010;Pastukhov & Braun, 2011, 2013b and theoretical (Friston et al, 2012;Pastukhov, García-Rodríguez, et al, 2013) perspectives, even though the neural implementations are still poorly understood (Daw, O'Doherty, Dayan, Seymour, & Dolan, 2006). The main focus of prior research was the knowledge about physical states (Hansen, Olkkonen, Walter, & Gegenfurtner, 2006;Weiss, Simoncelli, & Adelson, 2002;Yang & Purves, 2003;Yuille & Kersten, 2006), however this type of knowledge serves only as a weak constraint because the number of transformations by far outstrips the number of states.…”