Perceptual learning and top-down influences in primary visual cortex

Visual perception is subject to the influences of stimulus context, behavioral goals and past experience. In order to understand the underlying neuronal mechanisms, Professor Wu Li and colleagues in Professor Charles Gilbert's neurobiology laboratory at Rockefeller University examined the effects of these influencing factors on visual processing in the primary visual cortex (V1). Their studies, which have been reported in Nature Neuroscience and Neuron, showed that the responses of V1 neurons to complex stimuli are modified by an ongoing process of perceptual learning, and that they are subject to higher order cognitive influences of attention, expectation and perceptual task.

Professor Wu Li

Institute of Cognitive Neuroscience & Learning, Beijing Normal University, China

Overview

In a shape discrimination task, V1 neurons showed striking changes in contextual influences, and the nature of these influences depended on the current perceptual task. This lead to very different V1 responses to identical visual stimuli when different visual discrimination tasks were performed. Even the process of contour integration, which is generally characterized as a hard-wired process driven by geometric attributes of visual stimuli, strongly depended on top-down influences and perceptual learning. The laboratory's findings suggest that V1 is continually involved in encoding learned information, and that task-specific top-down influences can dynamically and plastically modify the expression of local circuits in V1 for more efficient processing of recurring stimuli and tasks. This may serve as a general neuronal mechanism of perceptual learning.

Example stimuli and behavioural protocols

Visual stimuli were generated by the CRS VSG2/5 visual stimulus generator, and consisted of oriented anti-aliased 0.5 x 0.08° light bars presented on a grey background. The bar contrast ranged between 15% and 60%.

In the five-line stimulus array (Figure a), the central line was fixed in the center of the Receptive Field (RF) and the flanks were placed near the edge of the RF. In different trials, the stimulus flanks were laterally displaced in varying steps with step size of 0.1° or 0.15°. All lines were oriented at the preferred orientation of the recorded neuron.

In the stimulus presentation, all the stimuli in the same task were randomized within a block of trials, and different tasks were arranged in different blocks and were preceded by a couple of cuing trials with only task-relevant lines shown. Each stimulus was repeated 8-20 times in different trials. At the trial outset, a 0.08° fixation point (FP) was displayed in the center of the screen. Eye positions were sampled at 30 Hz by an infrared tracking system (Matsuda,K. et al., Soc. Neurosci. Abstr. 26, 744.2, 2000 - this works in a similar way to the CRS Video Eyetracker Toolbox) and single units were recorded by a spike sorting and acquisition system (Plexon Inc.). Within 600 ms after FP presentation, the subject was required to fixate within an invisible circular window of 0.4-0.5° in radius around the FP. After the subject maintained fixation for 191 ms (20 CRT frames at 105 Hz), the stimulus was presented for 476 ms. Another 191 ms later, the FP was extinguished and two 0.15° saccade targets were presented for 600 ms symmetrically around the screen center on an axis orthogonal to the orientation ofstimulus lines. The subject indicated its choice by making a saccade toward either target.

Figures

The figures are reproduced from Li W, Piech V, Gilbert CD. Perceptual learning and top-down influences in primary visual cortex. Nature Neuroscience. 2004 June 7(6):651-7.

(a) The stimulus conditions upon which two different discrimination tasks were performed.

(b,c) Typical psychometric function (green curve) measured for the bisection (b) and vernier (c) tasks, using the five-line stimuli. The black dashed curve was constructed by assuming that the subject responded to the offsets of the task-irrelevant flanks rather than the task-relevant flanks. Note that the subject’s behavioral responses were strongly associated with the cued task, but were not affected by the irrelevant stimulus flanks.

(d) Learning curves of a subject in a simplified bisection and vernier task, where either three side-by-side or three end-to-end lines were displayed horizontally at 1.4° eccentricity. Each data point was based on 300 responses. The bottom horizontal lines indicate thresholds measured 3 months after training.

Further Reading

Li W, Piech V, Gilbert CD.
Contour saliency in primary visual cortex.
Neuron. 2006 June 15; 50(6):951-62.

Li W, Piech V, Gilbert CD.
Perceptual learning and top-down influences in primary visual cortex.
Nature Neuroscience. 2004 June 7(6):651-7.

Li W, Gilbert CD.
Global contour saliency and local colinear interactions.
Journal of Neurophysiology. 2002 November; 88(5):2846-56.

Crist RE, Li W, Gilbert CD.
Learning to see: experience and attention in primary visual cortex.
Nature Neuroscience. 2001 May; 4(5):519-25.

Fahle M.
Perceptual learning: gain without pain?
Nature Neuroscience. 2002 October; 5(10):923-4.

Professor Wu Li

Professor Wu Li completed his B.S in Biophysics at the University of Science and Technology of China in 1989 then went on to complete his Ph.D. in Neurophysiology at the Shanghai Institute of Physiology, Chinese Academy of Sciences, Shanghai, China in 1994.

He then held positions as Postdoctoral Fellow in the Division of Neurobiology, University of California, Berkely, from 1994 then in 1996 Research Fellow at Max-Planck Institute for Biological Cybernetics, Germany. He then became Research Assistant Professor in the Laboratory of Neurobiology, The Rockefeller University, New York, USA.

He has now moved back to China, working as Professor at the Institute of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China.

Prof. Wu Li has written chapters in the following books:

Li, W. and Gilbert, CD.  Perceptual Learning.  In: The Senses: A Comprehensive References. Eds. Allan Basbaum et al, Elsevier Inc, Oxford, 2008, ISBN: 9780126394825

Li, W. and Gilbert, CD.  Perceptual learning: neural mechanisms.  In: The New Encyclopedia of Neuroscience.  Editor-in-chief: Larry Squire. Elsevier, In Press

Related Product Information

ViSaGe Visual Stimulus Generator

Our latest top-of-the-range system is based on 15 years of VSG development. Accurate timing guarantees no dropped or inserted frames; 14 bit DAC output. Now easier than ever to use, with VSG Toolbox for MATLAB.

High Speed Video Eye Tracker Toolbox

Low noise, high fidelity 250Hz Video Eyetracker. Far more accurate on diagonals that limbus tracker solutions. Ideal for oculomotor research and saccade contingent displays.

Copyright © Cambridge Research Systems Ltd. Click for details.

^ Back to top