Comparison of Bayesian and empirical ranking approaches to visual perception

How can we apply decision-making theories to wild animal behavior? Predictions arising from dual process theory and Bayesian decision theory

Our understanding of decision-making processes and cognitive biases is ever increasing, thanks to an accumulation of testable models and a large body of research over the last several decades. The vast majority of this work has been done in humans and laboratory animals because these study subjects and situations allow for tightly controlled experiments. However, it raises questions about how this knowledge can be applied to wild animals in their complex environments. Here, we review two prominent decision-making theories, dual process theory and Bayesian decision theory, to assess the similarities in these approaches and consider how they may apply to wild animals living in heterogenous environments within complicated social groupings. In particular, we wanted to assess when wild animals are likely to respond to a situation with a quick heuristic decision and when they are likely to spend more time and energy on the decision-making process. Based on the literature and evidence from our multi-destination routing experiments on primates, we find that individuals are likely to make quick, heuristic decisions when they encounter routine situations, or signals/cues that accurately predict a certain outcome, or easy problems that experience or evolutionary history has prepared them for. Conversely, effortful decision-making is likely in novel or surprising situations, when signals and cues have unpredictable or uncertain relationships to an outcome, and when problems are computationally complex. Though if problems are overly complex, satisficing via heuristics is likely, to avoid costly mental effort. We present hypotheses for how animals with different socio-ecologies may have to distribute their cognitive effort. Finally, we examine the conservation implications and potential cognitive overload for animals experiencing increasingly novel situations caused by current human-induced rapid environmental change.

The Effects of Adding Pictorial Depth Cues to the Poggendorff Illusion

We tested if the misapplication of perceptual constancy mechanisms might explain the perceived misalignment of the oblique lines in the Poggendorff illusion. Specifically, whether these mechanisms might treat the rectangle in the middle portion of the Poggendorff stimulus as an occluder in front of one long line appearing on either side, causing an apparent decrease in the rectangle's width and an apparent increase in the misalignment of the oblique lines. The study aimed to examine these possibilities by examining the effects of adding pictorial depth cues. In experiments 1 and 2, we presented a central rectangle composed of either large or small bricks to determine if this manipulation would change the perceived alignment of the oblique lines and the perceived width of the central rectangle, respectively. The experiments demonstrated no changes that would support a misapplication of perceptual constancy in driving the illusion, despite some evidence of perceptual size rescaling of the central rectangle. In experiment 3, we presented Poggendorff stimuli in front and at the back of a corridor background rich in texture and linear perspective depth cues to determine if adding these cues would affect the Poggendorff illusion. The central rectangle was physically large and small when presented in front and at the back of the corridor, respectively. The strength of the Poggendorff illusion varied as a function of the physical size of the central rectangle, and, contrary to our predictions, the addition of pictorial depth cues in both the central rectangle and the background decreased rather than increased the strength of the illusion. The implications of these results with regards to different theories are discussed. It could be the case that the illusion depends on both low-level and cognitive mechanisms and that deleterious effects occur on the former when the latter ascribes more certainty to the oblique lines being the same line receding into the distance.

Why Does Rubin's Vase Differ Radically From Optical Illusions? Framing Effects Contra Cognitive Illusions

Many researchers use the term "context" loosely to denote diverse kinds of reference points. The issue is not about terminology but rather about the common conflation of one kind of reference points, such as rules of perception, which is responsible for optical illusions, with another kind, known as "context" or "frame," as exemplified in Rubin's vase. Many researchers regard Rubin's vase as a special kind of optical illusions. This paper rather argues that the two phenomena are radically different. Optical illusions are occasional mistakes that people quickly recognize and eagerly correct, while the different figures of Rubin's vase are not mistakes but, rather, the outcomes of different perspectives that do not need correction. The competing figures in Rubin's vase can, at best, in light of more information, be more warranted or unwarranted. This paper discusses at length one ramification of the proposed distinction. The framing effects, such as loss/gain frame, are the products of contexts and, hence, resemble greatly the figures in Rubin's vase. In contrast, cognitive illusions generated occasionally by the rules of thumb (heuristics) are mistakes and, hence, resemble optical illusions. The proposed distinction carries other ramifications regarding, e.g., happiness studies, moral judgments, and the new philosophy of science.

A review on various explanations of Ponzo-like illusions

This article reviews theoretical and empirical arguments for and against various theories that explain the classic Ponzo illusion and its variants from two different viewpoints concerning the role of perceived depth in size distortions. The first viewpoint argues that all Ponzo-like illusions are driven by perceived depth. The second viewpoint argues that the classic Ponzo illusion is unrelated to depth perception. This review will give special focus to the first viewpoint and consists of three sections. In the first section, the role of the number of pictorial depth cues and previous experience in the strength of all Ponzo-like illusions are discussed. In the second section, we contrast the first viewpoint against the theories that explain the classic Ponzo illusion with mechanisms that are unrelated to depth perception. In the last section, we propose a Bayesian-motivated reconceptualization of Richard Gregory's misapplied size constancy theory that explains Ponzo-variant illusions in terms of prior information and prediction errors. The new account explains why some studies have provided inconsistent evidence for misapplied size constancy theory.

Dynamic face mask enhances continuous flash suppression

In continuous flash suppression (CFS), an image presented to one eye is suppressed from awareness by a dynamic image masker presented to the other eye. Previous studies report that face stimuli break out of CFS more readily when they are oriented upright and contain ecologically relevant information such as facial expressions or direct eye gaze, potentially implicating face processing in the mechanisms of interocular competition. It is unknown, however, whether face content helps to drive interocular suppression when incorporated into the dynamic masker itself, either by engaging higher-level visual mechanisms that underlie face detection or due to lower-level image features that the faces happen to contain. To investigate this, we devised a dynamic mask composed of upright faces and tested how well it suppressed detection of face or grating targets presented to the other eye. Relative contributions of higher-level and lower-level features were compared by manipulating the image properties of the mask. Results show that the dynamic face mask is strikingly effective at suppressing sensory input presented to the opposing eye, but its effectiveness is largely attributable to image texture, which can be quantified in terms of image entropy and edge density. This is because strong suppression was still observed following phase-scrambling or spatial inversion of the face elements, and while a target-selective effect was observed for the face mask, inverting the face elements to interfere with configural processing did not significantly diminish this effect. Thus, visual properties of faces, such as their image entropy and complex phase structure, predominate in driving interocular suppression rather than face detection per se.

Bayesian Decision Models: A Primer

To understand decision-making behavior in simple, controlled environments, Bayesian models are often useful. First, optimal behavior is always Bayesian. Second, even when behavior deviates from optimality, the Bayesian approach offers candidate models to account for suboptimalities. Third, a realist interpretation of Bayesian models opens the door to studying the neural representation of uncertainty. In this tutorial, we review the principles of Bayesian models of decision making and then focus on five case studies with exercises. We conclude with reflections and future directions.

Representing Color Ensembles

Colors are rarely uniform, yet little is known about how people represent color distributions. We introduce a new method for studying color ensembles based on intertrial learning in visual search. Participants looked for an oddly colored diamond among diamonds with colors taken from either uniform or Gaussian color distributions. On test trials, the targets had various distances in feature space from the mean of the preceding distractor color distribution. Targets on test trials therefore served as probes into probabilistic representations of distractor colors. Test-trial response times revealed a striking similarity between the physical distribution of colors and their internal representations. The results demonstrate that the visual system represents color ensembles in a more detailed way than previously thought, coding not only mean and variance but, most surprisingly, the actual shape (uniform or Gaussian) of the distribution of colors in the environment.

Coral reef fish perceive lightness illusions

Visual illusions occur when information from images are perceived differently from the actual physical properties of the stimulus in terms of brightness, size, colour and/or motion. Illusions are therefore important tools for sensory perception research and from an ecological perspective, relevant for visually guided animals viewing signals in heterogeneous environments. Here, we tested whether fish perceived a lightness cube illusion in which identical coloured targets appear (for humans) to return different spectral outputs depending on the apparent amount of illumination they are perceived to be under. Triggerfish (Rhinecanthus aculeatus) were trained to peck at coloured targets to receive food rewards, and were shown to experience similar shifts in colour perception when targets were placed in illusory shadows. Fish therefore appear to experience similar simultaneous contrast mechanisms to humans, even when targets are embedded in complex, scene-type illusions. Studies such as these help unlock the fundamental principles of visual system mechanisms.

Schizophrenia, Subjectivity, and Mindreading

A number of recent articles, many appearing in Schizophrenia Bulletin, signal a renewed interest in phenomenological approaches to our understanding of schizophrenia. These approaches conceptualize schizophrenia as a disorder of altered self-awareness and decreased prereflective social attunement, which may manifest as an impaired understanding of self, others, and the physical world. Phenomenological approaches to psychopathology are sometimes construed as being incompatible with the reductionistic methodology of contemporary neuroscience. In this article, we re-examine findings from the phenomenological investigation of schizophrenia in light of an influential neurocomputational account of mindreading, which postulates that understanding of others is subserved by coherent internal self-models. We argue that the phenomenological approach to schizophrenia is not incompatible with a neurocomputational account of mindreading, and that the 2 approaches should instead be viewed as existing in a relationship of mutual constraint and enlightenment. Our hypothesis, while speculative, is an attempt to marry the phenomenological and neuronal realities of schizophrenia. Furthermore, it has implications for psychotherapeutic interventions and future research.

Whatever next? Predictive brains, situated agents, and the future of cognitive science

Brains, it has recently been argued, are essentially prediction machines. They are bundles of cells that support perception and action by constantly attempting to match incoming sensory inputs with top-down expectations or predictions. This is achieved using a hierarchical generative model that aims to minimize prediction error within a bidirectional cascade of cortical processing. Such accounts offer a unifying model of perception and action, illuminate the functional role of attention, and may neatly capture the special contribution of cortical processing to adaptive success. This target article critically examines this "hierarchical prediction machine" approach, concluding that it offers the best clue yet to the shape of a unified science of mind and action. Sections 1 and 2 lay out the key elements and implications of the approach. Section 3 explores a variety of pitfalls and challenges, spanning the evidential, the methodological, and the more properly conceptual. The paper ends (sections 4 and 5) by asking how such approaches might impact our more general vision of mind, experience, and agency.

Cortical long-range interactions embed statistical knowledge of natural sensory input: a voltage-sensitive dye imaging study

How is contextual processing as demonstrated with simplified stimuli, cortically enacted in response to ecologically relevant complex and dynamic stimuli? Using voltage-sensitive dye imaging, we captured mesoscopic population dynamics across several square millimeters of cat primary visual cortex. By presenting natural movies locally through either one or two adjacent apertures, we show that simultaneous presentation leads to mutual facilitation of activity. These synergistic effects were most effective when both movie patches originated from the same natural movie, thus forming a coherent stimulus in which the inherent spatio-temporal structure of natural movies were preserved in accord with Gestalt principles of perceptual organization. These results suggest that natural sensory input triggers cooperative mechanisms that are imprinted into the cortical functional architecture as early as in primary visual cortex.

Misperception of aspect ratio in binocularly viewed surfaces

The horizontal-vertical illusion, in which the vertical dimension is overestimated relative to the horizontal direction, has been explained in terms of the statistical relationship between the lengths of lines in the world, and the lengths of their projections onto the retina (Howe & Purves, 2002). The current study shows that this illusion affects the apparent aspect ratio of shapes, and investigates how it interacts with binocular cues to surface slant. One way in which statistical information could give rise to the horizontal-vertical illusion would be through prior assumptions about the distribution of slant. This prior would then be expected to interact with retinal cues to slant. We determined the aspect ratio of stereoscopically viewed ellipses that appeared circular. We show that observers' judgements of aspect ratio were affected by surface slant, but that the largest image vertical:horizontal aspect ratio that was considered to be a surface with a circular profile was always found for surfaces close to fronto-parallel. This is not consistent with a Bayesian model in which the horizontal-vertical illusion arises from a non-uniform prior probability distribution for slant. Rather, we suggest that assumptions about the slant of surfaces affect apparent aspect ratio in a manner that is more heuristic, and partially dissociated from apparent slant.

Manual matching of perceived surface orientation is affected by arm posture: evidence of calibration between proprioception and visual experience in near space

Proprioception of hand orientation (orientation production using the hand) is compared with manual matching of visual orientation (visual surface matching using the hand) in two experiments. In experiment 1, using self-selected arm postures, the proportions of wrist and elbow flexion spontaneously used to orient the pitch of the hand (20 and 80%, respectively) are relatively similar across both manual matching tasks and manual orientation production tasks for most participants. Proprioceptive error closely matched perceptual biases previously reported for visual orientation perception, suggesting calibration of proprioception to visual biases. A minority of participants, who attempted to use primarily wrist flexion while holding the forearm horizontal, performed poorly at the manual matching task, consistent with proprioceptive error caused by biomechanical constraints of their self-selected posture. In experiment 2, postural choices were constrained to primarily wrist or elbow flexion without imposing biomechanical constraints (using a raised forearm). Identical relative offsets were found between the two constraint groups in manual matching and manual orientation production. The results support two claims: (1) manual orientation matching to visual surfaces is based on manual proprioception and (2) calibration between visual and proprioceptive experiences guarantees relatively accurate manual matching for surfaces within reach, despite systematic visual biases in perceived surface orientation.

Chinese characters reveal impacts of prior experience on very early stages of perception

BACKGROUND

Visual perception is strongly determined by accumulated experience with the world, which has been shown for shape, color, and position perception, in the field of visuomotor learning, and in neural computation. In addition, visual perception is tuned to statistics of natural scenes. Such prior experience is modulated by neuronal top-down control the temporal properties of which had been subject to recent studies. Here, we deal with these temporal properties and address the question how early in time accumulated past experience can modulate visual perception.

RESULTS

We performed stimulus discrimination experiments and compared a group of Chinese participants with a German control group. The perception of our briefly presented visual objects (targets) was disturbed by masking stimuli which appeared in close spatiotemporal proximity. These masking stimuli were either intact or scrambled Chinese characters and did not overlap with the targets. In contrast to German controls, Chinese participants show substantial performance differences for real versus scrambled Chinese characters if these masking stimuli were presented as early as less than 100 milliseconds after the onset of the target. For Chinese observers, it even occured that meaningful masking stimuli enhanced target identification if they were shown at least 100 milliseconds after target onset while the same stimuli impaired recognition if presented in close temporal proximity to the target. The latter finding challenges interpretations of our data that solely rely on stimulus contours or geometric properties and emphasizes the impact of prior experience on the very early temporal dynamics of the visual system.

CONCLUSIONS

Our findings demonstrate that prior experience which had been accummulated long before the experiments can modulate the time course of perception intriguingly early, namely already immediately after the perceptual onset of a visual event. This modulation cannot solely operate as a feedback in response to the visual event but is rather a permanent effect.

Color constancy

A quarter of a century ago, the first systematic behavioral experiments were performed to clarify the nature of color constancy-the effect whereby the perceived color of a surface remains constant despite changes in the spectrum of the illumination. At about the same time, new models of color constancy appeared, along with physiological data on cortical mechanisms and photographic colorimetric measurements of natural scenes. Since then, as this review shows, there have been many advances. The theoretical requirements for constancy have been better delineated and the range of experimental techniques has been greatly expanded; novel invariant properties of images and a variety of neural mechanisms have been identified; and increasing recognition has been given to the relevance of natural surfaces and scenes as laboratory stimuli. Even so, there remain many theoretical and experimental challenges, not least to develop an account of color constancy that goes beyond deterministic and relatively simple laboratory stimuli and instead deals with the intrinsically variable nature of surfaces and illuminations present in the natural world.

Distributions of observed death tolls govern sensitivity to human fatalities

How we react to humanitarian crises, epidemics, and other tragic events involving the loss of human lives depends largely on the extent to which we are moved by the size of their associated death tolls. Many studies have demonstrated that people generally exhibit a diminishing sensitivity to the number of human fatalities and, equivalently, a preference for risky (vs. sure) alternatives in decisions under risk involving human losses. However, the reason for this tendency remains unknown. Here we show that the distributions of event-related death tolls that people observe govern their evaluations of, and risk preferences concerning, human fatalities. In particular, we show that our diminishing sensitivity to human fatalities follows from the fact that these death tolls are approximately power-law distributed. We further show that, by manipulating the distribution of mortality-related events that people observe, we can alter their risk preferences in decisions involving fatalities. Finally, we show that the tendency to be risk-seeking in mortality-related decisions is lower in countries in which high-mortality events are more frequently observed. Our results support a model of magnitude evaluation based on memory sampling and relative judgment. This model departs from the utility-based approaches typically encountered in psychology and economics in that it does not rely on stable, underlying value representations to explain valuation and choice, or on choice behavior to derive value functions. Instead, preferences concerning human fatalities emerge spontaneously from the distributions of sampled events and the relative nature of the evaluation process.

Visuomotor mental rotation: the reaction time advantage for anti-pointing is not influenced by perceptual experience with the cardinal axes

In the visuomotor mental rotation (VMR) paradigm, participants execute a center-out reaching movement to a location that deviates from a visual cue by a predetermined instruction angle. Previous work has demonstrated a linear increase in reaction time (RT) as a function of the amplitude of the instruction angle (Georgopoulos and Massey in Exp Brain Res 65:361-370, 1987). In contrast, we recently reported a RT advantage for an instruction angle of 180 degrees relative to a 90 degrees angle (Neely and Heath in Neurosci Lett 463:194-198, 2009). It is possible, however, that perceptual expertise with the cardinal axes, which are perceptually familiar reference frames, influenced the results of our previous investigation. To address this issue, we employed a VMR paradigm identical to that of our previous work, with the exception that the stimulus array was shifted 45 degrees from the horizontal and vertical meridians. Our results demonstrated that RTs were fastest and least variable when the instruction angle was 0 degrees, followed by 180 degrees, which in turn, was faster than 90 degrees. Such findings establish that the RT advantage for the 180 degrees instruction angle is not influenced by perceptual expertise with the cardinal axes. Moreover, the present results provide convergent evidence that RT is not determined by the angle of rotation; instead, they indicate that response latencies reflect computational differences in the complexity of response remapping.

An empirical explanation of the speed-distance effect

Understanding motion perception continues to be the subject of much debate, a central challenge being to account for why the speeds and directions seen accord with neither the physical movements of objects nor their projected movements on the retina. Here we investigate the varied perceptions of speed that occur when stimuli moving across the retina traverse different projected distances (the speed-distance effect). By analyzing a database of moving objects projected onto an image plane we show that this phenomenology can be quantitatively accounted for by the frequency of occurrence of image speeds generated by perspective transformation. These results indicate that speed-distance effects are determined empirically from accumulated past experience with the relationship between image speeds and moving objects.

An empirical explanation of the flash-lag effect

When a flash of light is presented in physical alignment with a moving object, the flash is perceived to lag behind the position of the object. This phenomenon, known as the flash-lag effect, has been of particular interest to vision scientists because of the challenge it presents to understanding how the visual system generates perceptions of objects in motion. Although various explanations have been offered, the significance of this effect remains a matter of debate. Here, we show that: (i) contrary to previous reports based on limited data, the flash-lag effect is an increasing nonlinear function of image speed; and (ii) this function is accurately predicted by the frequency of occurrence of image speeds generated by the perspective transformation of moving objects. These results support the conclusion that perceptions of the relative position of a moving object are determined by accumulated experience with image speeds, in this way allowing for visual behavior in response to real-world sources whose speeds and positions cannot be perceived directly.

What's next? New evidence for prediction in human vision

Everyday visual experience involves making implicit predictions, as revealed by our surprise when something disturbs our expectations. Many theories of vision have been premised on the central role played by prediction. Yet, implicit prediction in human vision has been difficult to assess in the laboratory, and many results have not distinguished between the indisputably important role of memory and the future-oriented aspect of prediction. Now, a new and unexpected finding - that humans can resume an interrupted visual search much faster than they can start a new search - offers new hope, because the rapid resumption of a search seems to depend on participants forming an implicit prediction of what they will see after the interruption. These findings combined with results of recent neurophysiology studies provide a framework for studying implicit prediction in perception.

Perceptual learning depends on perceptual constancy

Perceptual learning refers to experience-induced improvements in the pick-up of information. Perceptual constancy describes the fact that, despite variable sensory input, perceptual representations typically correspond to stable properties of objects. Here, we show evidence of a strong link between perceptual learning and perceptual constancy: Perceptual learning depends on constancy-based perceptual representations. Perceptual learning may involve changes in early sensory analyzers, but such changes may in general be constrained by categorical distinctions among the high-level perceptual representations to which they contribute. Using established relations of perceptual constancy and sensory inputs, we tested the ability to discover regularities in tasks that dissociated perceptual and sensory invariants. We found that human subjects could learn to classify based on a perceptual invariant that depended on an underlying sensory invariant but could not learn the identical sensory invariant when it did not correlate with a perceptual invariant. These results suggest that constancy-based representations, known to be important for thought and action, also guide learning and plasticity.

Evolution of visually guided behavior in artificial agents

Recent work on brightness, color, and form has suggested that human visual percepts represent the probable sources of retinal images rather than stimulus features as such. Here we investigate the plausibility of this empirical concept of vision by allowing autonomous agents to evolve in virtual environments based solely on the relative success of their behavior. The responses of evolved agents to visual stimuli indicate that fitness improves as the neural network control systems gradually incorporate the statistical relationship between projected images and behavior appropriate to the sources of the inherently ambiguous images. These results: (1) demonstrate the merits of a wholly empirical strategy of animal vision as a means of contending with the inverse optics problem; (2) argue that the information incorporated into biological visual processing circuitry is the relationship between images and their probable sources; and (3) suggest why human percepts do not map neatly onto physical reality.

Vision as Bayesian inference: analysis by synthesis?

We argue that the study of human vision should be aimed at determining how humans perform natural tasks with natural images. Attempts to understand the phenomenology of vision from artificial stimuli, although worthwhile as a starting point, can lead to faulty generalizations about visual systems, because of the enormous complexity of natural images. Dealing with this complexity is daunting, but Bayesian inference on structured probability distributions offers the ability to design theories of vision that can deal with the complexity of natural images, and that use 'analysis by synthesis' strategies with intriguing similarities to the brain. We examine these strategies using recent examples from computer vision, and outline some important implications for cognitive science.