ECVP Symposium


Space, colour, natural vision, and conscious robots: A symposium to honour Tom Troscianko

Sunday 2nd September 2012, at ECVP 2012, Alghero, Sardinia

Chair: Petroc Sumner


An afternoon of talks at Tom's favourite conference, the European Conference on Visual Perception, discussing some of the many and varied research areas that Tom was involved and interested in.


Mark W. Greenlee and Lothar Spillmann

Seeing through Tom's eyes: Perception of isoluminant chromatic contours


Certain aspects of vision are altered for isoluminant chromatic contours: spatial phase discrimination, Vernier offset acuity, apparent motion, velocity discrimination of moving contours and the Ouchi illusion. Tom Troscianko studied all of these phenomena and reported most of his results in a set of papers published in 1987-1988 (Troscianko, T. (1987). Vision Research, 27(4), 547–554.; Troscianko, T., & Fahle, M. (1988). JOSA A, 5(6), 871–880; Troscianko, T., & Harris, J. (1988). Vision Research, 28(9), 1041–1049). Tom suggested that the breakdown in visual performance (or the reduced illusory jitter in the Ouchi figure) was related to greater positional uncertainty of isoluminant chromatic contours. Tom attributed the uncertainty to a lack of inhibitory surrounds in postreceptoral mechanisms that encode isoluminant edges. We review Tom's work on perception at isoluminance and discuss to what extent research over last 25 years provides support for his early and insightful observations.


Patrick Cavanagh and Mark Wexler

The motion of pure colour: It's all in the jitter


Among Tom Trościanko’s many interests, he had an extended affair with pure colour stimuli, running many studies on their fascinating effects like slowed motion, and their jazzy, unstable appearance. Tom, always an iconoclast, did not buy the fashionable arguments that motion was colour-blind and argued instead that isoluminant stimuli acted as ordinary luminance stimuli but with positional jitter. He attributed this spatial scrambling to the larger positional uncertainty of early color-selective units. He then showed that spatial scrambling can produce all reported effects of isoluminant stimuli: apparent slowing, loss of global shape in kinematograms, slowed reaction times, whereas spatial jitter preserved properties not lost at isoluminance, such as symmetry perception. It followed, according to Tom, that isoluminant visual displays do not isolate real, higher-level chromatic mechanisms, as performance was first of all contaminated by this low-level jitter. As an indication that the motion contribution from color was not a property of high-level chromatic pathways, Tom, together with several others, showed that motion responses to isoluminant stimuli survived in brain-damaged patients that saw no color at all. He then developed techniques to isolate pure-color responses by adding uncorrelated dynamic luminance noise to the isoluminant display, extending the rationale of the Ishihara plates. He was able to show that a weak motion response survives even this superimposed noise field. Together with Mark Wexler in Paris, we are now extending Tom’s ideas about position jitter to new motion phenomena for luminance and colour-defined stimuli.


Manfred Fahle and Tom Troscianko

From stereo vision over isoluminance to perceptual learning


I will outline three topics Tom and I worked on together –stereo and colour vision, both in healthy subjects and in patients – and relate them to my recent work on perceptual learning. (1) On stereo vision, a topic Tom had touched hardly at all before and never again after our article, we identified retinal image quality, spatial frequency, luminance, contrast, temporal factors, motion, size and retinal location as factors constraining stereo vision. (2) On motion perception at isoluminance, we were able to explain the slow-down experienced subjectively at isoluminance of moving colour stimuli by decreased positional accuracy, rather than by a lack of temporal accuracy. We suggested that isoluminant stimuli behave like low-contrast non-isoluminant stimuli and successfully modeled the results accordingly. (3a)The study of two patients suffering from achromatopsia questioned the traditional view that colour information is carried exclusively by colour-opponent parvocellular channels, concluding that chromatic discrimination can be subserved by a non-parvocellular channel. (3b) Testing patients after peripheral retinal detachment, we found isoluminant flicker fusion frequencies being severely decreased even though anomaloscope results were normal, demonstrating the potential of isoluminant stimuli for clinical tests. The improvement, over time, in these patient’s performance, started my interest in perceptual learning.


David Tolhurst, Rikesh Rajani, Tom Troscianko, and Isabel Gibson

The colour opponency assumptions in a V1-based model for predicting perceived differences in natural scenes


We are developing a V1 based computer model to explain observers’ suprathreshold judgements of the perceived magnitude of differences between naturalistic images (Lovell et al, 2006, ACM TAP, 3, 155-178; To et al, 2010, JoV, 10(4):12, 1-22.). We now question some of the fundamental assumptions within the model, particularly the formulations of the Red-Green and Blue-Yellow colour opponent processes. Presently, we use a Macleod-Boynton transform which assumes (i) that R/G opponency is only between L and M cones and (ii) that colour opponent changes are coded totally independently of luminance. Then, isoluminant L/M and S-modulated sinewave gratings (Mullen, 1985, J.Physiol. 359, 381-400) map directly and separately into the R/G and B/Y mechanisms. However, there is much evidence that the R/G system also includes input from S-cones, while the S-cone in the B/Y system might be synergistic with L or M cones. The assumption of isoluminance is questionable. We have conducted rating magnitude experiments to estimate the perceived differences between isoluminant 3.8 degrees square patches differing in hue in L*c*h space by 10, 20 or 30 degrees. The perceived difference depends highly on the patch luminance, falsifying our basic assumption that the colour opponent planes in our model are isoluminant.


Petroc Sumner, Aline Bompas, and Grace Kendall

What task drove the evolution of human colour vision?


The red-green dimension of human colour vision appears to be optimized for finding fruit in leaves at about arms reach (Parraga, Troscianko & Tolhurst, 2002), but is this ‘picking fruit’ task the one where trichromacy provides the largest advantage over red-green colour blindness? Other authors had assumed that spotting a fruiting tree at distance (between trees) was key. We tested this directly in a naturalistic setting by asking trichromats and dichromats to spot fruit pieces in bushes at different distances. We found that performance diverged with distance from 4m to 12m – i.e. the advantage of trichromacy grows with distance. Interestingly however, for the shortest distance (1 m) the advantage of normal colour vision also appears greater than at 4 m. Thus both theories (arms-length and between-tree) may be right.


Daniel Osorio and Sarah Zylinski

Cuttlefish coloration - tricks of camouflage and show


Cuttlefish draw together a number of Tom's interests on natural images, camouflage, illusions and lighting. We will present an overview of this fascinating creature and new findings that illuminate the cuttlefish’s art.



Iain D. Gilchrist, Christina J. Howard, and Tom Troscianko

When the eyes predict judgments about real moving scenes


Our visual environment changes continuously and so in turn do our judgments. In a series of studies we developed a set of methods to study this dynamic relationship and to investigate if fixation behavior could give an insight into the link between the changing visual world and our changing judgments. In one experiment participants watched a video of a football match and indicated the likelihood of an imminent goal with a joystick. We found that the variability of fixation position across participants was related to judgments of imminent goal likelihood. Participants tended to be fixating the same part of the video as one another a few seconds before they increased their reported likelihood of a goal. We also found that experts got their eyes to the relevant parts of the scene earlier. In subsequent work we investigated participants making a continuous suspiciousness judgment while viewing a set of four CCTV videos. We found that the eyes were directed to the video with the highest level of reported relative suspiciousness. The methods we developed open up the possibility of studying a wide range of tasks in which the visual stimuli are complex and dynamic and the judgment is continuous.


Matthew Stainer, Ben Tatler, and Kenneth Scott-Brown

Examining attention allocation in multiplex viewing: How many scenes are seen?


Multiplex displays are a popular visualisation tool in entertainment and professional use. Most research examines how attention is allocated in single scene viewing, but multiplex displays present the visual system with a number of additional challenges. Perhaps nowhere are the demands of viewing such displays more evident than in the CCTV Control Room where operators can be required to simultaneously monitor up to 100 scenes. In a series of experiments we examine attention allocation across the multiplex and tease apart several potential causes of processing difficulty. Using a modified version of the flicker paradigm with multiple scenes containing a single changed item, we use change detection performance as an index of attention allocation. Unsurprisingly, change detection performance decreases as scene number increases. There are many potential reasons for this difficulty with multiplex arrays. Across a set of experiments we show that performance is influenced by the information content of the multiplex rather than semantic similarity between scenes or the physical continuity of content across scenes. The underlying factors governing attention allocation in multiplex displays appear surprisingly similar to those for single scene viewing, raising questions about whether a mutliplex of scenes is treated perceptually as a single scene.


Ronald Rensink

The perception of correlation in datasets


Humans are remarkably good at getting the gist of a scene from a quick glance. Can this ability be used in the visualization of complex datasets? It will be shown that the perception of correlation in scatterplots is rapid, being largely complete within 150 ms of presentation. This process can be characterized by two simple laws: a linear Fechner-like law for precision and a logarithmic Weber-like law for accuracy. Results show a surprising degree of invariance for scatterplot symbol: different sizes, colours, and shapes have little effect on precision or accuracy. Other forms of visualization exhibit similar patterns. These results suggest that correlation perception is a sophisticated process, likely playing an important role in rapid scene perception. At a more general level, they also suggest that information visualization can be a useful domain in which to study visual cognition.

Copyright © All Rights Reserved