Adaptation I

Size adaptation effects are independent of spatial frequency aftereffects

Daniel H. Baker and Tim S. Meese

Repulsive adaptation effects for spatial frequency, motion and orientation (the tilt aftereffect) are well established, and support the notion of population coding in each of these domains. We have recently proposed (Meese & Baker, 2011, Journal of Vision, 11(1):23, 1-23) that the spatial extent of an object or texture is represented in a similar way. If so, adaptation effects should exist that are sensitive to object size (e.g. diameter) rather than the scale of a texture (spatial frequency). Using a matching task, we measured perceived size of 4c/deg gratings before and after exposure to an adaptor that jittered in space to cover the area of the largest target. All eight of our na?ve observers experienced a clear shift in perceived size - large targets looked larger and small targets looked smaller. The effect is similar in magnitude (10-20%) to spatial frequency repulsion effects (which we also replicate) but does not induce them: increasing perceived area does not increase perceived bar width. Size adaptation is robust to the relative orientation of adaptor and target, and even occurs for disparate objects such as gratings and faces. This implies adaptation of a broadly-tuned process which estimates the envelope of a stimulus.

The effect of surface material on colour constancy

Annette Werner and Lara Zebrowski

Unlike matt paper, most materials reflect light differently depending on the angle of the incoming light. We investigated how this may influence human colour constancy. Samples (circular patches, 90 mm) of different materials (paper, fur, tile, cloth, leather), were mounted on a sample holder (60o slope) and presented in the middle of a black viewing box (1.0 x 1.0 x 0.8 m). Illumination of the scene was provided by a computer controlled, calibrated LCD-projector (Panasonic PT AE 1000E). The observers (n=4) viewed the samples frontally (viewing distance 0.90 m), with their heads fixed. All samples appeared achromatic after 2 min adaptation to standard daylight (D65). Colour constancy was quantified by measuring the shift of the achromatic colour locus after adaptation to a new illuminant, picked from one of the cardinal axes in an equiluminant plane in colour space. Measuring the time course (0.2 – 15 s) of colour constancy for the different materials, we found significant differences at short adaptation times (0.2 s and 0.5 s). In general there was a trend for materials with a higher degree of mesostructure (e.g. fur) to be better compensated than those with a smoother surface texture (e.g. tile).

Statistical context alters perceived similarity: Gabor patches appear much more similar when large patch differences prevail

Alexander N Sokolov, Joachim Frederic Eicher and Paul Enck

Judgements of stimulus similarity are commonly used to infer underlying sensory representational spaces. Typically, however, stimulus pairs with small rather than large differences occur more frequently, ie possess higher overall base rates. Overall base rates and the stimuli encountered on the initial trials constitute statistical contexts that can alter sensory magnitudes: For example, the perceived speed is lower when fast rather than slow speeds are generally more frequent and/or dominate the series outset (frequency and primacy effects) [Sokolov et al, 2000, Perception & Psychophysics, 62(5), 998-1007]. Here, we ask if frequency and primacy effects occur with similarity judgements. Four separate groups of healthy adults used a 3-point scale (hardly, moderately, and very similar) to judge pair-wise similarity of six Gabor patches (spatial frequency, 1-10 cycle deg-1 in 0.2 log-steps). Patch pairs with either large or small differences possessed higher overall base rates and/or dominated the initial trials. The results indicate that independently of the content of initial trials, higher base rates of large, compared to small, patch differences yield much higher apparent similarity of the patches (ie the frequency effect). We show for the first time that statistical context modulates similarity judgements, suggesting important implications for computational models of sensory representations.

Adaptation in visual cortex equalizes population responses

Andrea Benucci and Matteo Carandini

Adaptation of neuronal responses in primary visual cortex (V1) has been proposed to discount changes in the statistics of the inputs. To test this hypothesis we recorded neuronal responses in cat V1 using 10x10 multielectrode arrays. Stimuli were sequences of gratings flashed for 32 ms. In control experiments all orientations had the same probability of occurrence (4-9%). In adaptation experiments one orientation (the adaptor) occurred with a higher probability (30-50%). Following adaptation, neurons selective for the adaptor gave similar average firing rates as the others despite being stimulated 3-7 times more often. Adaptation, therefore, maintains a flat probability of firing across the population. We described responses using a linear-nonlinear model. The linear filters are defined in stimulus orientation (tuning curves) and time. The model captured the data during both control and adaptation experiments with the same nonlinearity. Hence, the effects of adaptation are entirely captured by the linear filters. We conclude that adaptation is a homeostatic mechanism that maintains a flat distribution of responses in the face of biases in the ensemble of stimuli. It changes responsiveness by exactly the right amount, and exactly in the right neurons, to discount changes in the average firing of subpopulations. Supported by European Research Council and Medical Research Council.

Extreme motion aftereffects

Mark Wexler, Andrew Glennerster and Patrick Cavanagh

Motion aftereffects are either slow as compared to the inducing motion, or fast when their speed is imposed by the spatiotemporal features of a counterphase flicker. Here we show that when motion adaptation effects are induced by brief transients, they can be extremely fast, resulting in perceived motion that can be many tens of times faster that the inducer (and without the speed being imposed by counterphase flicker). As with regular motion adaptation, these effects are in the same direction as the inducing motion for brief adaptation, and in the opposite direction for longer adaptation. The jumps perceived are extreme, in that their amplitude, which varies from observer to observer, is at or just beyond each observer's maximal veridically perceived motion limit. These results are compatible with a process in which motions up to dmax are excluded, and with the minimal motion principle placing the perceived motion just beyond this excluded range.

Inferences of material properties based on frequency-band analyses

Qasim Zaidi and Martin Giesel

Successful functioning in the world often requires rapid and accurate assessment of material properties. We show that real materials exhibit characteristic configurations of low-level features, and the perception of properties like roughness, thickness, volume, can be altered by increasing or decreasing the relative energy in specific spatial-frequency bands of the image amplitude spectrum. These results suggest that simple neural “detectors” for material properties could just combine the outputs of sets of frequency-selective V1 neurons. Can such detectors be revealed through selective adaptation? We modified images of fabrics by increasing or decreasing relative energy in the specific frequency-bands identified by our image analyses (constant total energy). Baseline psychometric functions for each property were measured by side-by-side comparison of the original image with its manipulated versions. During the adaptation stage, dynamic white noise stimuli in specific frequency-bands were presented on the location of the original image, and the complementary notch-filtered dynamic noise presented on the location of the comparison image. Post-adaptation psychometric curves revealed that depending on adaptation frequency-band, observers judged fabrics as having systematically less surface volume, softer texture, or thinner weave/knit, as predicted. Consequently the visibility of specific frequency bands is directly linked to the perception of corresponding material properties.