Perceptual learning is specific beyond vision and decision making

Perceptual learning is usually assumed to occur within sensory areas or when sensory evidence is mapped onto decisions. Subsequent procedural and motor processes, involved in most perceptual learning experiments, are thought to play no role in the learning process. Here, we show that this is not the case. Observers trained with a standard three-line bisection task and indicated the offset direction of the central line by pressing either a left or right push button. Before and after training, observers adjusted the central line of the same bisection stimulus using a computer mouse. As expected, performance improved through training. Surprisingly, learning did not transfer to the untrained mouse adjustment condition. The same was true for the opposite, i.e., training with mouse adjustments did not transfer to the push button condition. We found partial transfer when observers adjusted the central line with two different adjustment procedures. We suggest that perceptual learning is specific to procedural motor aspects beyond visual processing. Our results support theories were visual stimuli are coded together with their corresponding actions.

What is new in perceptual learning?

What is new in perceptual learning? In the early days of research, specificity was the hallmark of perceptual learning; that is, improvements following training were limited to the trained stimulus features. For example, training with a stimulus improves performance for this stimulus but not for the same stimulus when rotated by 90° (Ball & Sekuler, 1987; Spang, Grimsen, Herzog, & Fahle, 2010). Because of this specificity, learning was thought to be mediated by neural changes at the early stages of vision. In the last decade, many procedures were discovered in which transfer occurs from trained to untrained conditions under certain conditions. The location of learning is now often thought to occur in higher stage of vision and decision-making. This special issue shows how the field has progressed along these lines.

Neural dynamics of grouping and segmentation explain properties of visual crowding

Investigations of visual crowding, where a target is difficult to identify because of flanking elements, has largely used a theoretical perspective based on local interactions where flanking elements pool with or substitute for properties of the target. This successful theoretical approach has motivated a wide variety of empirical investigations to identify mechanisms that cause crowding, and it has suggested practical applications to mitigate crowding effects. However, this theoretical approach has been unable to account for a parallel set of findings that crowding is influenced by long-range perceptual grouping effects. When the target and flankers are perceived as part of separate visual groups, crowding tends to be quite weak. Here, we describe how theoretical mechanisms for grouping and segmentation in cortical neural circuits can account for a wide variety of these long-range grouping effects. Building on previous work, we explain how crowding occurs in the model and explain how grouping in the model involves connected boundary signals that represent a key aspect of visual information. We then introduce new circuits that allow nonspecific top-down selection signals to flow along connected boundaries or within a surface contained by boundaries and thereby induce a segmentation that can separate the visual information corresponding to the flankers from the visual information corresponding to the target. When such segmentation occurs, crowding is shown to be weak. We compare the model's behavior to 5 sets of experimental findings on visual crowding and show that the model does a good job explaining the key empirical findings. (PsycINFO Database Record

What to Choose Next? A Paradigm for Testing Human Sequential Decision Making

Many of the decisions we make in our everyday lives are sequential and entail sparse rewards. While sequential decision-making has been extensively investigated in theory (e.g., by reinforcement learning models) there is no systematic experimental paradigm to test it. Here, we developed such a paradigm and investigated key components of reinforcement learning models: the eligibility trace (i.e., the memory trace of previous decision steps), the external reward, and the ability to exploit the statistics of the environment's structure (model-free vs. model-based mechanisms). We show that the eligibility trace decays not with sheer time, but rather with the number of discrete decision steps made by the participants. We further show that, unexpectedly, neither monetary rewards nor the environment's spatial regularity significantly modulate behavioral performance. Finally, we found that model-free learning algorithms describe human performance better than model-based algorithms.

Local versus global and retinotopic versus non-retinotopic motion processing in schizophrenia patients

Schizophrenia impairs cognitive functions as much as perception. For example, patients perceive global motion in random dot kinematograms less strongly, because, as it is argued, the integration of the dots into a single Gestalt is complex and therefore deteriorated. Similarly, the perception of apparent motion is impaired, because filling-in of the illusory trajectory requires complex processing. Here, we investigated very complex motion processing using the Ternus-Pikler display. First, we tested whether the perception of global apparent motion is impaired in schizophrenia patients compared to healthy controls. The task requires both the grouping of multiple elements into a coherent Gestalt and the filling-in of its illusory motion trajectory. Second, we tested the perception of rotation in the same stimulus, which in addition requires the computation of non-retinotopic motion. Contrary to earlier studies, patients were not impaired in either task and even tended to perform better than controls. The results suggest that complex visual processing itself is not impaired in schizophrenia patients.

The Honeycomb illusion: Uniform textures not perceived as such

We present a series of patterns, in which texture is perceived differently at fixation in comparison to the periphery, such that a physically uniform stimulus yields a nonuniform percept. We call this the Honeycomb illusion, and we discuss it in relation to the similar Extinction illusion (Ninio & Stevens, 2000). The effect remains strong despite multiple fixations, dynamic changes, and manipulations of the size of texture elements. We discuss the phenomenon in relation to how vision achieves a detailed and stable representation of the environment despite changes in retinal spatial resolution and dramatic changes across saccades. The Honeycomb illusion complements previous related observations in suggesting that this representation is not necessarily based on multiple fixations (i.e., memory) or on extrapolation from information available to central vision.

An overview of quantitative approaches in Gestalt perception

Gestalt psychology is often criticized as lacking quantitative measurements and precise mathematical models. While this is true of the early Gestalt school, today there are many quantitative approaches in Gestalt perception and the special issue of Vision Research "Quantitative Approaches in Gestalt Perception" showcases the current state-of-the-art. In this article we give an overview of these current approaches. For example, ideal observer models are one of the standard quantitative tools in vision research and there is a clear trend to try and apply this tool to Gestalt perception and thereby integrate Gestalt perception into mainstream vision research. More generally, Bayesian models, long popular in other areas of vision research, are increasingly being employed to model perceptual grouping as well. Thus, although experimental and theoretical approaches to Gestalt perception remain quite diverse, we are hopeful that these quantitative trends will pave the way for a unified theory.

What crowding can tell us about object representations

In crowding, perception of a target usually deteriorates when flanking elements are presented next to the target. Surprisingly, adding further flankers can lead to a release from crowding. In previous work we showed that, for example, vernier offset discrimination at 9° of eccentricity deteriorated when a vernier was embedded in a square. Adding further squares improved performance. The more squares presented, the better the performance, extending across 20° of the visual field. Here, we show that very similar results hold true for shapes other than squares, including unfamiliar, irregular shapes. Hence, uncrowding is not restricted to simple and familiar shapes. Our results provoke the question of whether any type of shape is represented at any location in the visual field. Moreover, small changes in the orientation of the flanking shapes led to strong increases in crowding strength. Hence, highly specific shape-specific interactions across large parts of the visual field determine vernier acuity.

Retinotopic encoding of the Ternus-Pikler display reflected in the early visual areas

The visual representation of the world is often assumed to be retinotopic, and many visual brain areas are indeed organized retinotopically. Visual perception, however, is not based on a reference frame anchored in retinotopic coordinates. For example, when an object moves, motion of its constituent parts is perceived relative to the object rather than in retinotopic coordinates. The moving object thus serves as a nonretinotopic reference system for computing the properties of its parts. It is largely unknown how the brain accomplishes this feat. Here, we used the Ternus-Pikler display to pit retinotopic processing in a stationary reference system against nonretinotopic processing in a moving one. Using 7T fMRI, we found that the average blood-oxygen-level dependent activations in V1, V2, and V3 reflected the retinotopic properties, but not the nonretinotopic percepts, of the Ternus-Pikler display. In the human motion processing complex (hMT+), activations were compatible with both retinotopic and nonretinotopic encoding. Thus, hMT+ may be the first visual area encoding the nonretinotopic percepts of the Ternus-Pikler display.

A New Conceptualization of Human Visual Sensory-Memory

Memory is an essential component of cognition and disorders of memory have significant individual and societal costs. The Atkinson–Shiffrin “modal model” forms the foundation of our understanding of human memory. It consists of three stores: Sensory Memory (SM), whose visual component is called iconic memory, Short-Term Memory (STM; also called working memory, WM), and Long-Term Memory (LTM). Since its inception, shortcomings of all three components of the modal model have been identified. While the theories of STM and LTM underwent significant modifications to address these shortcomings, models of the iconic memory remained largely unchanged: A high capacity but rapidly decaying store whose contents are encoded in retinotopic coordinates, i.e., according to how the stimulus is projected on the retina. The fundamental shortcoming of iconic memory models is that, because contents are encoded in retinotopic coordinates, the iconic memory cannot hold any useful information under normal viewing conditions when objects or the subject are in motion. Hence, half-century after its formulation, it remains an unresolved problem whether and how the first stage of the modal model serves any useful function and how subsequent stages of the modal model receive inputs from the environment. Here, we propose a new conceptualization of human visual sensory memory by introducing an additional component whose reference-frame consists of motion-grouping based coordinates rather than retinotopic coordinates. We review data supporting this new model and discuss how it offers solutions to the paradoxes of the traditional model of sensory memory.

Does sensitivity in binary choice tasks depend on response modality?

In most models of vision, a stimulus is processed in a series of dedicated visual areas, leading to categorization of this stimulus, and possible decision, which subsequently may be mapped onto a motor-response. In these models, stimulus processing is thought to be independent of the response modality. However, in theories of event coding, common coding, and sensorimotor contingency, stimuli may be very specifically mapped onto certain motor-responses. Here, we compared performance in a shape localization task and used three different response modalities: manual, saccadic, and verbal. Meta-contrast masking was employed at various inter-stimulus intervals (ISI) to manipulate target visibility. Although we found major differences in reaction times for the three response modalities, accuracy remained at the same level for each response modality (and all ISIs). Our results support the view that stimulus-response (S-R) associations exist only for specific instances, such as reflexes or skills, but not for arbitrary S-R pairings.

Motion-based nearest vector metric for reference frame selection in the perception of motion

We investigated how the visual system selects a reference frame for the perception of motion. Two concentric arcs underwent circular motion around the center of the display, where observers fixated. The outer (target) arc's angular velocity profile was modulated by a sine wave midflight whereas the inner (reference) arc moved at a constant angular speed. The task was to report whether the target reversed its direction of motion at any point during its motion. We investigated the effects of spatial and figural factors by systematically varying the radial and angular distances between the arcs, and their relative sizes. We found that the effectiveness of the reference frame decreases with increasing radial- and angular-distance measures. Drastic changes in the relative sizes of the arcs did not influence motion reversal thresholds, suggesting no influence of stimulus form on perceived motion. We also investigated the effect of common velocity by introducing velocity fluctuations to the reference arc as well. We found no effect of whether or not a reference frame has a constant motion. We examined several form- and motion-based metrics, which could potentially unify our findings. We found that a motion-based nearest vector metric can fully account for all the data reported here. These findings suggest that the selection of reference frames for motion processing does not result from a winner-take-all process, but instead, can be explained by a field whose strength decreases with the distance between the nearest motion vectors regardless of the form of the moving objects.

Time Slices: What Is the Duration of a Percept?

We experience the world as a seamless stream of percepts. However, intriguing illusions and recent experiments suggest that the world is not continuously translated into conscious perception. Instead, perception seems to operate in a discrete manner, just like movies appear continuous although they consist of discrete images. To explain how the temporal resolution of human vision can be fast compared to sluggish conscious perception, we propose a novel conceptual framework in which features of objects, such as their color, are quasi-continuously and unconsciously analyzed with high temporal resolution. Like other features, temporal features, such as duration, are coded as quantitative labels. When unconscious processing is "completed," all features are simultaneously rendered conscious at discrete moments in time, sometimes even hundreds of milliseconds after stimuli were presented.

EEG Correlates of Relative Motion Encoding

A large portion of the visual cortex is organized retinotopically, but perception is usually non-retinotopic. For example, a reflector on the spoke of a bicycle wheel appears to move on a circular or prolate cycloidal orbit as the bicycle moves forward, while in fact it traces out a curtate cycloidal trajectory. The moving bicycle serves as a non-retinotopic reference system to which the motion of the reflector is anchored. To study the neural correlates of non-retinotopic motion processing, we used the Ternus-Pikler display, where retinotopic processing in a stationary reference system is contrasted against non-retinotopic processing in a moving one. Using high-density EEG, we found similar brain responses for both retinotopic and non-retinotopic rotational apparent motion from the earliest evoked peak (around 120 ms) and throughout the rest of the visual processing, but only minor correlates of the motion of the reference system itself (mainly around 100-120 ms). We suggest that the visual system efficiently discounts the motion of the reference system from early on, allowing a largely reference system independent encoding of the motion of object parts.

Linking perceptual learning with identical stimuli to imagery perceptual learning

Perceptual learning is usually thought to be exclusively driven by the stimuli presented during training (and the underlying synaptic learning rules). In some way, we are slaves of our visual experiences. However, learning can occur even when no stimuli are presented at all. For example, Gabor contrast detection improves when only a blank screen is presented and observers are asked to imagine Gabor patches. Likewise, performance improves when observers are asked to imagine the nonexisting central line of a bisection stimulus to be offset either to the right or left. Hence, performance can improve without stimulus presentation. As shown in the auditory domain, performance can also improve when the very same stimulus is presented in all learning trials and observers were asked to discriminate differences which do not exist (observers were not told about the set up). Classic models of perceptual learning cannot handle these situations since they need proper stimulus presentation, i.e., variance in the stimuli, such as a left versus right offset in the bisection stimulus. Here, we show that perceptual learning with identical stimuli occurs in the visual domain, too. Second, we linked the two paradigms by telling observers that only the very same bisection stimulus was presented in all trials and asked them to imagine the central line to be offset either to the left or right. As in imagery learning, performance improved.

Putting low-level vision into global context: Why vision cannot be reduced to basic circuits

To cope with the complexity of vision, most models in neuroscience and computer vision are of hierarchical and feedforward nature. Low-level vision, such as edge and motion detection, is explained by basic low-level neural circuits, whose outputs serve as building blocks for more complex circuits computing higher level features such as shape and entire objects. There is an isomorphism between states of the outer world, neural circuits, and perception, inspired by the positivistic philosophy of the mind. Here, we show that although such an approach is conceptually and mathematically appealing, it fails to explain many phenomena including crowding, visual masking, and non-retinotopic processing.