Predicting the motion after-effect from sensitivity loss

Individual differences in direction-selective motion adaptation revealed by change-detection performance

The motion aftereffect (MAE) and motion adaptation in general are usually considered to be universal phenomena. However, in a preliminary study using a bias-free measure of the MAE we found some individuals who showed at best a weak effect of adaptation. These same individuals also performed poorly in a "change detection" test of motion adaptation based on visual search, leading to the conjecture that there is a bimodality in the population with respect to motion adaptation. The present study tested this possibility by screening 102 participants on two versions of the change-detection task while also considering potential confounding factors including eye movements, practice-based improvements, and deficits in visual search ability. The 5 strongest and the 5 weakest change detectors were selected for further testing of motion detection and contrast detection after adaptation. Data showed an inverse association between change-detection ability and performance in the motion-detection task. We extend previous findings by also showing i) the weakest change detectors exhibit less direction selectivity in their contrast thresholds after adapting to drifting gratings and ii) the ability to detect change in motion direction correlates with the ability to detect change in spatial orientation. Group differences between the strongest and weakest change detectors cannot be attributed to a lack of practice, nor can they be explained by poor fixation ability. Our results suggest genuine individual differences in the degree to which adaptation is specific to stimulus orientation and direction of motion.

Contrast sensitivity functions in autoencoders

Three decades ago, Atick et al. suggested that human frequency sensitivity may emerge from the enhancement required for a more efficient analysis of retinal images. Here we reassess the relevance of low-level vision tasks in the explanation of the contrast sensitivity functions (CSFs) in light of 1) the current trend of using artificial neural networks for studying vision, and 2) the current knowledge of retinal image representations. As a first contribution, we show that a very popular type of convolutional neural networks (CNNs), called autoencoders, may develop human-like CSFs in the spatiotemporal and chromatic dimensions when trained to perform some basic low-level vision tasks (like retinal noise and optical blur removal), but not others (like chromatic) adaptation or pure reconstruction after simple bottlenecks). As an illustrative example, the best CNN (in the considered set of simple architectures for enhancement of the retinal signal) reproduces the CSFs with a root mean square error of 11% of the maximum sensitivity. As a second contribution, we provide experimental evidence of the fact that, for some functional goals (at low abstraction level), deeper CNNs that are better in reaching the quantitative goal are actually worse in replicating human-like phenomena (such as the CSFs). This low-level result (for the explored networks) is not necessarily in contradiction with other works that report advantages of deeper nets in modeling higher level vision goals. However, in line with a growing body of literature, our results suggests another word of caution about CNNs in vision science because the use of simplified units or unrealistic architectures in goal optimization may be a limitation for the modeling and understanding of human vision.

Color illusions also deceive CNNs for low-level vision tasks: Analysis and implications

The study of visual illusions has proven to be a very useful approach in vision science. In this work we start by showing that, while convolutional neural networks (CNNs) trained for low-level visual tasks in natural images may be deceived by brightness and color illusions, some network illusions can be inconsistent with the perception of humans. Next, we analyze where these similarities and differences may come from. On one hand, the proposed linear eigenanalysis explains the overall similarities: in simple CNNs trained for tasks like denoising or deblurring, the linear version of the network has center-surround receptive fields, and global transfer functions are very similar to the human achromatic and chromatic contrast sensitivity functions in human-like opponent color spaces. These similarities are consistent with the long-standing hypothesis that considers low-level visual illusions as a by-product of the optimization to natural environments. Specifically, here human-like features emerge from error minimization. On the other hand, the observed differences must be due to the behavior of the human visual system not explained by the linear approximation. However, our study also shows that more 'flexible' network architectures, with more layers and a higher degree of nonlinearity, may actually have a worse capability of reproducing visual illusions. This implies, in line with other works in the vision science literature, a word of caution on using CNNs to study human vision: on top of the intrinsic limitations of the L + NL formulation of artificial networks to model vision, the nonlinear behavior of flexible architectures may easily be markedly different from that of the visual system.

Decision-level adaptation in motion perception

Prolonged exposure to visual stimuli causes a bias in observers' responses to subsequent stimuli. Such adaptation-induced biases are usually explained in terms of changes in the relative activity of sensory neurons in the visual system which respond selectively to the properties of visual stimuli. However, the bias could also be due to a shift in the observer's criterion for selecting one response rather than the alternative; adaptation at the decision level of processing rather than the sensory level. We investigated whether adaptation to implied motion is best attributed to sensory-level or decision-level bias. Three experiments sought to isolate decision factors by changing the nature of the participants' task while keeping the sensory stimulus unchanged. Results showed that adaptation-induced bias in reported stimulus direction only occurred when the participants' task involved a directional judgement, and disappeared when adaptation was measured using a non-directional task (reporting where motion was present in the display, regardless of its direction). We conclude that adaptation to implied motion is due to decision-level bias, and that a propensity towards such biases may be widespread in sensory decision-making.

Visual aftereffects and sensory nonlinearities from a single statistical framework

When adapted to a particular scenery our senses may fool us: colors are misinterpreted, certain spatial patterns seem to fade out, and static objects appear to move in reverse. A mere empirical description of the mechanisms tuned to color, texture, and motion may tell us where these visual illusions come from. However, such empirical models of gain control do not explain why these mechanisms work in this apparently dysfunctional manner. Current normative explanations of aftereffects based on scene statistics derive gain changes by (1) invoking decorrelation and linear manifold matching/equalization, or (2) using nonlinear divisive normalization obtained from parametric scene models. These principled approaches have different drawbacks: the first is not compatible with the known saturation nonlinearities in the sensors and it cannot fully accomplish information maximization due to its linear nature. In the second, gain change is almost determined a priori by the assumed parametric image model linked to divisive normalization. In this study we show that both the response changes that lead to aftereffects and the nonlinear behavior can be simultaneously derived from a single statistical framework: the Sequential Principal Curves Analysis (SPCA). As opposed to mechanistic models, SPCA is not intended to describe how physiological sensors work, but it is focused on explaining why they behave as they do. Nonparametric SPCA has two key advantages as a normative model of adaptation: (i) it is better than linear techniques as it is a flexible equalization that can be tuned for more sensible criteria other than plain decorrelation (either full information maximization or error minimization); and (ii) it makes no a priori functional assumption regarding the nonlinearity, so the saturations emerge directly from the scene data and the goal (and not from the assumed function). It turns out that the optimal responses derived from these more sensible criteria and SPCA are consistent with dysfunctional behaviors such as aftereffects.

Sustained attention is not necessary for velocity adaptation

Prolonged adaptation to a stimulus, such as a drifting grating, lowers sensitivity for detecting similar stimuli, and changes their appearance, for example, making gratings of the same orientation appear of lower contrast and move more slowly. It has been suggested that adaptation is increased by sustained attention to the adapting stimulus and is decreased by distracting attention with a competing task. This paper describes a novel 2AFC (spatial) measure of adaptation in which adaptation and bias are carefully distinguished by the random interleaving of different test conditions. The experiment revealed significant adaptation of perceived velocity, but no effect of attentional distraction.

What is the primary cause of individual differences in contrast sensitivity?

One of the primary objectives of early visual processing is the detection of luminance variations, often termed image contrast. Normal observers can differ in this ability by at least a factor of 4, yet this variation is typically overlooked, and has never been convincingly explained. This study uses two techniques to investigate the main source of individual variations in contrast sensitivity. First, a noise masking experiment assessed whether differences were due to the observer's internal noise, or the efficiency with which they extracted information from the stimulus. Second, contrast discrimination functions from 18 previous studies were compared (pairwise, within studies) using a computational model to determine whether differences were due to internal noise or the low level gain properties of contrast transduction. Taken together, the evidence points to differences in contrast gain as being responsible for the majority of individual variation across the normal population. This result is compared with related findings in attention and amblyopia.

Motion adaptation does not depend on attention to the adaptor

Prolonged inspection of moving stimuli causes stationary stimuli to appear moving in the opposite direction to the adapting stimulus (the Waterfall effect). It has been claimed that distracting the viewer's attention from the adapting stimulus by a secondary task reduces the strength of adaptation. However, the method used to show the effect of distraction (the duration of the aftereffect) is potentially susceptible to bias. The experiments reported here show no effect in genuinely naïve subjects, or in experienced observers using a variety of cancellation procedures to measure the effect.

Wohlgemuth was right: distracting attention from the adapting stimulus does not decrease the motion after-effect

We determined whether distracting the observer's attention from an adapting stimulus could decrease the motion after-effect. Unlike previous studies we used a relatively bias-free 2AFC procedure to measure the strength of adaptation. The strength of motion adaptation was measured by the effects of a moving grating on the contrast discrimination (T vs. C) function for gratings moving in the same or opposite direction. As in previous reports, the effect of adaptation was to move the T vs. C function upwards and rightwards, consistent with an increase in the C50 (semi-saturation) response in the transduction function of the neural mechanism underlying the discrimination. On the other hand, manipulating the attentional load of a distracting task during adaptation had no consistent effect on contrast discrimination, including the absolute detection threshold. It is suggested that previous reported effects of attentional load on adaptation may have depended on response bias, rather than changes in sensitivity.

Evidence for a subtractive component in motion adaptation

Adaptation to a moving stimulus changes the perception of a stationary grating and also reduces contrast sensitivity to the adaptor. We determined whether the first effect could be predicted from the second. The contrast discrimination (T vs. C) function for a drifting 7.5 Hz grating test stimulus was determined when observers were adapted to a low contrast (0.075) grating of the same spatial and temporal frequency, moving in either the same or the opposite direction as the test. The effect of an adaptor moving in the same direction was to move the T vs. C function upwards and to the right, in a manner consistent with an increase in divisive inhibition. We also measured the effect of adaptation on the motion-null point for a counterphasing grating containing two components, one moving in the same direction as the adaptor and the other in the opposite direction. Adaptation increased the amount of contrast of the adapted component required to achieve the motion-null point. However, this shift could not be predicted from the effects of adaptation on contrast sensitivity. In particular, the balance point was shifted in gratings of high contrast where there was no effect of adaptation on contrast discrimination. We suggest that adaptation has a subtractive (recalibration) effect in addition to its effects on the contrast transduction function, and that this subtractive effect may explain the movement after-effect seen with stationary tests.

Adaptation and visual coding

Visual coding is a highly dynamic process and continuously adapting to the current viewing context. The perceptual changes that result from adaptation to recently viewed stimuli remain a powerful and popular tool for analyzing sensory mechanisms and plasticity. Over the last decade, the footprints of this adaptation have been tracked to both higher and lower levels of the visual pathway and over a wider range of timescales, revealing that visual processing is much more adaptable than previously thought. This work has also revealed that the pattern of aftereffects is similar across many stimulus dimensions, pointing to common coding principles in which adaptation plays a central role. However, why visual coding adapts has yet to be fully answered.

Spatially localized motion aftereffect disappears faster from awareness when selectively attended to according to its direction

In searching for the target-afterimage patch among spatially separate alternatives of color-afterimages the target fades from awareness before its competitors (Bachmann, T., & Murd, C. (2010). Covert spatial attention in search for the location of a color-afterimage patch speeds up its decay from awareness: Introducing a method useful for the study of neural correlates of visual awareness. Vision Research 50, 1048-1053). In an analogous study presented here we show that a similar effect is obtained when a target spatial location specified according to the direction of motion aftereffect within it is searched by covert top-down attention. The adverse effect of selective attention on the duration of awareness of sensory qualiae known earlier to be present for color and periodic spatial contrast is extended also to sensory channels carrying motion information.

The motion aftereffect reloaded

The motion aftereffect is a robust illusion of visual motion resulting from exposure to a moving pattern. There is a widely accepted explanation of it in terms of changes in the response of cortical direction-selective neurons. Research has distinguished several variants of the effect. Converging recent evidence from different experimental techniques (psychophysics, single-unit recording, brain imaging, transcranial magnetic stimulation, visual evoked potentials and magnetoencephalography) reveals that adaptation is not confined to one or even two cortical areas, but occurs at multiple levels of processing involved in visual motion analysis. A tentative motion-processing framework is described, based on motion aftereffect research. Recent ideas on the function of adaptation see it as a form of gain control that maximises the efficiency of information transmission at multiple levels of the visual pathway.

Space and time in visual context

No sensory stimulus is an island unto itself; rather, it can only properly be interpreted in light of the stimuli that surround it in space and time. This can result in entertaining illusions and puzzling results in psychological and neurophysiological experiments. We concentrate on perhaps the best studied test case, namely orientation or tilt, which gives rise to the notorious tilt illusion and the adaptation tilt after-effect. We review the empirical literature and discuss the computational and statistical ideas that are battling to explain these conundrums, and thereby gain favour as more general accounts of cortical processing.