AVA '99 Annual Meeting, College of Optometrists, London, 17 March 1999 page

Neurons in sensory systems encode and transmit information about attributes of the environment. Much of the information transmitted by spiking neurons appears to be encoded in the rate at which they fire. This rate necessarily has a positive value. We propose that the use of positive-valued signals to transmit information is a fundamental constraint on neural computation. We examine the implications of this constraint for models of motion detection. The detection of image motion is represented mathematically as a linear programming problem in which variables are restricted to having non-negative values. We propose an adaptive model that can detect a coherent motion signal from the responses of velocity tuned linear filters. Additional constraints are introduced as regularisers to stabilise the model. The regularisers are based upon the idea that image motion orthogonal to a one dimensional signal is assumed to be zero. It is shown how inhibition, motion opponency and contrast normalisation consistent with the properties of V1 complex and MT motion cells are motivated by the method of solution. The model exploits both divisive and subtractive velocity gain control to explain several properties of motion perception, namely, perceived biases in speed and direction and a subset of dynamic and static motion aftereffects including direction repulsion.

Search amongst new objects is more effective than search amongst old objects but the response is not necessarily faster

Gellatly, Angus and Cole, Geoff
Department of Psychology, University of Keele

This paper describes a series of visual search experiments in which the search display was presented to the observers very briefly. In the first two experiments the target was an illusory square containing stationary horizontal lines on a background of downward moving lines; non-targets were illusory squares containing upward moving lines. The comparison of interest is between new and old objects. New objects are those which appeared immediately following an auditory signal to search for a target. Old objects had been present on the screen for (usually) 500 ms before an auditory signal to indicate that one of them may become a target. The search display was always terminated within 500 ms. In Experiment 1, old and new objects always occurred in separate trials. In Experiment 2, each trial contained a mixture of old and new objects. In both experiments, new object targets were detected more frequently than old object targets. Experiments 3 and 4 replicated and extended this finding for the case in which targets, non-targets and background were all different shades of grey, showing the effect is not confined to a particular class of stimulus. Experiment 5 reverted to the use of motion defined stimuli but used reaction time as the dependent measure, it included a set size manipulation and search displays which were terminated by the participant’s response. Error data are consistent with the previous detection data but the RT data are not. The results will be discussed in terms of enhanced pre-attentive processing and of inhibitory response processes.

Can V1 mechanisms give higher sensitivities to medial axes of figure shapes?

Z. Li
Gatsby Computational Neuroscience Unit, University College London, 17 Queen Square, London WC1N 3AR, UK

Human observers are more sensitive to targets (e.g. bars) when they are placed in the centre of figure shapes such as circles or squares. Evidence of higher neural responses to figure centres has been observed in the primary visual cortex (V1). However, it is not clear whether the effects arise from V1 mechanisms or from higher centre feedbacks. We use a model of the primary visual cortex to show that some of these effects can arise from the contextual influences induced by the intracortical interactions within V1. The neural elements in the model consist of the pyramidal cells and inhibitory interneurons from layer 2-3 of the cortex, the local reciprocal connections between the two cell types, and the longer range horizontal connections from the pyramidal cells to other pyramidal cells and interneurons. The visual inputs are filtered by the classical receptive fields before being received by the pyramidal cells. The intracortical interactions make the pyramidal responses dependent on the stimulus outside their receptive fields, thus changing their sensitivities according to the contextual surround. We examine in particular how sensitivity enhancements to targets at the centre of figures depend on the shape, size, and other attributes (e.g. the texture features) of the figures, and whether the enhancement is also present at other locations inside the figures or even outside the figures. Our results are consistent with previous simulations from this model and suggest that V1 mechanisms contribute significantly to some aspects of pre-attentive segmentation.

The Effect of Subthreshold Sensory Stimulation on Learning in Visual Search

Stewart L. and Walsh V.
Institute of Cognitive Neuroscience, Alexandra House, 17 Queen Square, London, WC1N 3AR.

Performance on visual search tasks improves with learning. In previous experiments we have argued that improvements occur as a function of low level sensory changes rather than higher level cognitive ones. We tested this hypothesis by attempting to disrupt or improve the learning of visual search using transcranial magnetic stimulation (TMS). We established the threshold at which stimulation over area V5 could elicit moving phosphenes so that stimulation intensity could be set at a given percentage of this (80%). Subjects then trained on a form/motion conjunction search task for 1300 trials and received either no TMS, TMS at 3Hz TMS, or TMS at 10Hz or sham TMS on every trial. Stimulation did not disrupt visual search accuracy or reaction times. Only 3Hz stimulation had an effect on learning, consistent with other studies which have shown an effect of TMS on visually related learning. We suggest that this provides additional evidence that learning in visual search is due to sensory changes rather than cognitive strategies.

Multiple roles for memory in visual search

Shore, D and Klein, R.M.
Department of Psychology, Dalhousie University, Halifax, Nova Scotia, B3H4J1, Canada

It is a near universal assumption that memory, in several forms, contributes to visual search performance. This traditional view has been denied in a recent paper (Horowitz & Wolfe, 1998, Nat., 394, 575-577). After outlining fundamental problems with the interpretation of the data from this paper, we focus on recent findings which support the traditional view. Klein & MacInnes (in press, Psyc. Sci.) demonstrated a strong tendency against the refixation of items in a dense array and further showed that saccadic latencies are slower when a target is presented at a previously fixated location. This converges with the observation that simple detection is slowed following target presentation at a location occupied by a distractor in a serial search display (Takeda & Yagi, 1998, Proc. Int. Workshop on Adv. in Res. on Vis. Cog.: Select. and Integr., 211-216), as if previously attended items are inhibited from re-inspection (cf Klein, 1988, Nat., 334, 430-431). In a possibly related phenomenon, Watson & Humphreys (1997, Psych. Rev., 104, 90-122) demonstrated that previewing a subset of the distractors allows them to be visually marked, and thus discounted. These effects refer to the ongoing processing of a visual search display and thus implicate either perceptual or working memory systems. With respect to long-term memory, items which serve as distractors in one phase of an experiment are harder to find in a subsequent phase when they serve as targets (Lubow & Kaplan, 1997, J. Expt. Psych.: Hum. Perc. Perform., 23, 14-24). These three observations clearly show that information about distractor items influences performance on both the current and subsequent search trials.

An examination of the spatiotemporal integration of spatial frequencies

McSorley, E. and Findlay John M.
Centre for Vision and Visual Cognition, Department of Psychology, University of Durham, South Road, Durham, DH1 3LE

The experiments to be presented were designed to examine the importance of spatial frequency processing delays in the visual system on the efficacy of their subsequent integration. In the first experiment subjects were asked to detect a full bandwidth square-wave stimulus in various sequential displays, in some of which the first three harmonics were presented separately and sequentially. The effect of the presentation order of the harmonics, presented at short durations of 42 msec each running from either low to high or high to low in spatial frequency content, on square-wave detection was assessed. It was found that subjects incorrectly reported the presence of a square-wave significantly more often when the harmonics were presented in a low to high order. In experiment 2 subjects were asked to judge square-wave similarity with the same stimulus displays. Contrary to expectations, it was found that the high to low harmonic triplet was chosen significantly more often than the low to high stimulus to resemble a square-wave. The results suggest that the effective integration of spatial frequencies does not depend upon the low to high temporal structure imposed by initial spatial frequency processing delays. It is proposed that the integration of spatial frequencies is flexible and dependent upon task demands.

Attentional Asymmetries between Visual Fields in Visual Search Tasks.

Ellison, Amanda and Walsh, Vincent.
Department of Experimental Psychology, South Parks Road, Oxford University, Oxford, OX1 3UD

It has been suggested that attentional resolution is greater in the lower than in the upper visual field. As there is no corresponding asymmetry between the areas in the primary visual cortex where the input from upper and lower visual fields is processed, an "attentional filter" has been proposed to act in one or more higher visual cortical areas in order to constrict the availability of visual information to the level of awareness. To investigate this, a visual search array was presented to the entire visual field and reaction times from upper and lower visual fields compared. In a second experiment, subjects trained in one visual field before being transferred to the other. There was no significant difference between reaction times for targets presented in either upper or lower visual fields when the array was presented to the entire visual field. However, when the array was restricted to either the upper or lower visual fields, reaction times were significantly slower for detection in the upper visual field. This agrees with the proposal of a greater attention resolution in the lower visual field. The occurrence of significant transfer of suggests that the substrates for this attentional resolution must be at a higher level than low level visual processing areas, where receptive fields are small with a more orderly retinotopic map prevents transfer of information between fields.

Visual search for motion in depth: how do we find z-motion?

Julie M. Harris and Jane H. Sumnall
Neural Systems Group, Department of Psychology, University of Newcastle, Newcastle-upon-Tyne, NE1 7RU

We have measured how well observers can detect binocular 3-D motion when presented amidst stationary distractor dots, positioned randomly in depth. The target can be made to move laterally (x-motion), towards or away from the observer (z-motion) or along an intermediate trajectory that is a combination of the two. Z-motion, amidst 3-D distractors, is much harder to detect than x-motion. We suggested that this may be because motion mechanisms average signals from the left and right eye (Harris et al, 1998, Nature Neuroscience, 1, 165-168). This predicts that when presented with a combination of x- and z-motion, the visual system will respond as if only the x-motion is present. However, this prediction is not always true. As the amount of z-motion is increased, eventually its presence can improve performance over x-motion alone (Sumnall & Harris, 1998, Perception, 27, 51). Why is this? Observers often report seeing slight x-motion for a purely z-motion stimulus. It is possible that the ability to detect z-motion is based on imperfect averaging of left and right eye motion, resulting in a leak-through‚ of the x-motion signal. If so, then adding a small amount of x-motion jitter to a z-motion stimulus should result in reduced performance. Performance fell for z-motion as jitter was added, for 2 observers, but it did not fall to chance. For a third observer, performance fell much less quickly. There appears to be some effect of leak-through‚ of x-motion, but this cannot completely explain the ability to detect z-motion.

Effects of Dot Density in the Artificial Scotoma Paradigm

Welchman, A.E.
Neural Systems Group, Psychology Department, University of Newcastle, Newcastle-upon-Tyne, NE1 7RU

Ramachandran & Gregory (1991, Nature) report that a peripherally presented grey square (target patch) embedded in a surround of white noise will perceptually fade following 7-10 seconds of steady fixation. The area occupied by the patch is then perceived as filled with the "twinkling" noise texture of the surround. Here we investigate two types of target patch. In one case, we employ the original mean luminance patch of Ramachandran & Gregory; in the other we define the patch by a difference from the surround in motion coherence. We investigate the effects of dot density on exposure time required for perceptual fading.

Subjects viewed a central fixation cross and reported when the target patch, located 10 deg from fixation, had faded. The time taken for perceptual fading was measured as a function of dots in the stimulus. For the grey patch: Stimulus exposure time needed for perceptual fading increases as the density of dots surrounding it decreases. For the patch containing moving dots: Time to fade decreases as dot density decreases. These data point to the importance of the interaction between the properties of the target patch and the surrounding texture in determining the time course of perceptual fading in the artificial scotoma paradigm. Divergent effects are observed for patches defined by texture difference or motion coherence difference. This suggests that different mechanisms may be involved in the perceptual filling-in of each case.

Visual Search and motion - is it all relative?

Rushton, S.K. and Bradshaw, M.F.
Dept of Psychology, University of Surrey, Guildford, Surrey GU2 5XH

It has previously been reported that searching for oriented targets embedded within both moving and stationary oriented distractors is faster when the target is moving than when it is stationary (Driver and MacLeod, 1992; MacLeod, Driver and Crisp, 1988). The characteristics of cells in cortical area MT/V5 — namely its blindness to static elements (and poor resolution for orientation) has been used to account for this motion-advantage (and in some cases motion-disadvantage). Here we investigate whether this effect holds for both absolute and relative motion. Absolute motion is defined by the change in position of an element on the retina and must be recovered by taking eye position into account whereas relative motion is defined by the relative change of position of different elements in a scene and can be recovered without recourse to information about eye position. The sensitivity of the visual system is usually found to be much greater for relative than for absolute motion tasks (e.g. Snowden, 1992) but any relative influence on tasks, such as visual search, is unknown.

In the present experiments we dissociate absolute and relative motion of the stimulus elements by manipulating the presence and motion of a reference border in an otherwise featureless display. The results clearly show that both absolute and relative motion affect visual search tasks. The implications for the underlying physiology will be discussed.

Searching for change with and without awareness.

Ian M. Thornton(1) & Diego Fernandez-Duque(2)
Cambridge Basic Research, Cambridge, MA(1) Psychology, University of Oregon, Eugene(2)

Rensink (Visual Cognition, in press) has recently described a useful new method for exploring attentional processing, a method that marries standard visual search arrays with the change blindness flicker paradigm (Rensink et al., 1997, Psych. Sci; see Simons & Levin, 1997, TICS, for a review). In the current work, we use a variant of this methodology to examine the representation of change in the absence of awareness. We presented observers with a search array consisting of a ring of 8 or 12 rectangles (half horizontal, half vertical). The array was shown for 250 ms and then during an intervening blank period (250 ms) the orientation of one rectangle was switched and the whole array redisplayed for a further 250 ms. In one series of experiments, the redisplayed array was followed by a 2AFC task in which observers had to select the rectangle that changed. In a second series of experiments, the redisplayed array was followed by a color singleton, either at the location of change, or at a diametrically opposite location, and observers made a speeded red/green discrimination. In all experiments, observers also indicated whether they had been aware of any change in the initial displays. Our results show that a) observers can identify the changing object at above-chance levels even when they report having no conscious awareness of change and b) this ability does not rely on the redeployment of attention.

Safety Analysis of a Scanning Laser Display

Jones, R.J.M. and Liggens, E.
Protection and Performance Department, Centre for Human Sciences, DERA, Farnborough, UK

Scanning laser retinal displays (SLRDs) are a novel form of display which scan a modulated laser beam across the retina of the eye so as to form an image. Although safety assessments of such a system have been made, none have used the British and European laser safety standard, BS EN 60825:1 (1994). This paper describes such an assessment. The maximum safe exposures are given as corneal irradiances, and for both normal and scan failure situations. For scan failure the maximum permissible exposure is 2.55 mWcm-2 for an exposure of 0.25 seconds, which is the maximum time until the blink reflex terminates the exposure. For normal use the maximum permissible exposure is dependent on the exposure duration, screen size and the wavelength of the light, but for a virtual 15" monitor at 0.5m and an 8 hour exposure the limit is approximately 2mWcm-2 below 550nm, rising for longer wavelengths, to 35mWcm-2 at 630nm and 100mWcm-2 at 670nm. The paper also gives these limits in equivalent monitor luminosities. Finally, the laser safety standard is applied to conventional displays, and the limits compared with those found for SLRDs.

Selective attention and the ideal observer applied to binocular summation

Rose, D. and Pardhan, S.
Psychology Department, University of Surrey, Guildford, GU2 5XH.

Binocular contrast summation to threshold is a consequence of both higher sampling efficiency and lower equivalent internal noise for binocular targets (Pardhan and Rose, Perception 26, 1334, 1997; IOVS 39, S617, 1998). The contributions of each mechanism were estimated by adding varying amounts of external noise and extrapolating to zero, and differed depending whether the masking noise was correlated or uncorrelated between the eyes. A two-stage model is presented. First is a matrix of neurophysiological channels that are selective for all possible pairings of ocular dominance and stereoscopic disparity. At the second stage, attention is ‘ideally’ directed to (i.e. information is read out or used from) only the channels that respond to the signal. In our experiments, all targets were presented on the horopter but binocular, left eye and right eye targets were randomly intermixed. Binocularly correlated external noise impaired detection by injecting noise into the ideally attended channels (zero disparity and all ocular dominances) whereas uncorrelated external noise broadened attention: because the noise energy covered many disparities, its onset automatically drew attention away from the horopter.

Pre-attentive visual search and visual display technology.

Ward, P.A. and Carr, C.
Protection and Performance Department, Centre for Human Sciences, DERA, Farnborough, UK.

The traditional approach to assessing the impact of display technology on visual performance involves using either a rating/ranking or a reaction time paradigm. Both these approaches have fundamental drawbacks: inherent variability in the data, large numbers of subjects and long experimental sessions. This study adopted a radically different approach by using a pre-attentive visual task and signal detection analysis. Most display technologies vary most in physical and temporal characteristics at the individual pixel level. Cathode ray tube (CRT) and a liquid crystal displays (LCD) are good examples of displays with very different pixel characteristics, notably the diffuse shape of the CRT pixel and the highly defined shape of the LCD pixel. The stimulus luminance and contrast were matched for both displays, and temporal characteristics were matched to within 20ms. Each LCD pixel subtended 1.2 min of arc. Because the simplest stimulus which will demonstrate pixel effects is a line which is not aligned with the display pixel structure the task chosen was detection of a sloping line within a field of similar vertical distractors (Ward and Foster 1988). Target detection was assessed as a function of eccentricity. Six subjects were used and data were converted to the bias free estimate of performance d’. The results indicated no significant difference in pre-attentive visual search as function of display technology, indicating that the pixel shape did not significantly affect performance.

Ward P.A. and Foster D.H. (1988) Anisotropic thresholds fororientation discrimination in brief multi-element displays. Perception 17, 392

Long term stability of instrument myopia and its relation to tonic accommodation.

Ward P.A. and Viveash, J.P.
Protection and Performance Department, Centre for Human Sciences, DERA, Farnborough, UK

Tonic accommodation (TA) measurements were made before, and for 24 minutes after, 90 minutes use of a monocular optical device. Measurements of accommodation were made at 15 minute intervals during the 90 minute period. There was a significant change in TA from the pre-task to the post-task level. The shift in TA recovered over the post-viewing period. All the subjects demonstrated an inappropriate accommodation response when using the optical device which made them effectively myopic (instrument myopia (IM)). In all but one of the subjects the pre-task TA was significantly different from the first IM level. The level of IM varied greatly from subject to subject over a range from approximately 0.5 to 2.3D. In order to examine the long term stability of IM the subject data were pooled after first being normalised to remove the effects of inter-subject variability. This analysis indicated that the mean IM only varied within approximately 0.3D from its initial level. The analysis further indicated that the shift in TA is towards the mean IM level. Thus IM appears to be unlike other anomalous myopias, such as night myopia, in that it is not a simple return of the accommodation system to its resting state. Also IM itself influences TA, by effectively pulling the TA towards the IM level. The change in TA is similar to Ebonholtz (1983) hysteresis effects and requires approximately 30 minutes dark adaptation to recover.

Ebonholtz S.M. (1983) Accommodative hysteresis: a precursor for induced myopia? Invest. Ophthalmol. & Vis. Sci. 24,513-515.

Does attention have access to monocular input?

Solomon, J
The Institute of Ophthalmology, 11 - 43 Bath Street, London, EC1V 9EL

We used two types of paired display to determine whether or not visual search mechanisms have access to monocular input. In Type A both halves of each pair contained a Gabor target amongst perpendicular distractors. One target was horizontal; the other was vertical. Two spatial frequencies were used, one for one half, the other for the other half. The distribution of target and distractors within each half was identical so that when combined, the two targets formed a plaid. All of the distractors also formed plaids. When combined halves were viewed monocularly, the time required to locate the target did not vary with the number of distractors. We surmise that our target plaids are located by a pre-attentive mechanism. When the two halves were presented dichoptically, target location was even faster. This indicates that pre-attentive search mechanisms (possibly better described as texture-segregation mechanisms) have access to monocular input. Pairs of Type B differed from pairs of Type A in two ways: only one half of each pair contained a target and no Gabors formed plaids when the two halves were combined. With pairs of Type B, location time was proportional to the number of distractors. Thus, location of Type-B targets appears to require attention. For three of four observers, dichoptic and monocular displays produced similar location times. This indicates that attentive ("serial") search mechanisms do not have access to monocular input.

Visual search in Alzheimer’s disease

Andrea Tales, Tom Troscianko, Stuart Butler, Gordon Wilcock, Roy Jones and Jane Fossey.
School of Psychology, Cardiff University, PO Box 901, Cardiff CF1 3YG

We asked whether the presence of Alzheimer’s disease compared to normal ageing affected the performance of different kinds of visual search task. Of particular interest was whether Alzheimer's disease resulted in a particular impairment in a Conjunction search task in which the difference between target and distractor elements was defined by a conjunction of two features (orientation and lightness). Two control conditions were employed: a purely preattentive task and a single-dimension partly attentive task in which the target element was slightly larger than the distractors- a size task. The results suggested that individuals with Alzheimer’s disease have a particular impairment in the conjunction and not in the size task. This implies that Alzheimer’s disease particularly affects those mechanisms which compare across more than one feature type and spares the preattentive system. This study provides a double dissociation with previous work on Parkinson’s disease which points to an impairment at the preattentive and not the attentive levels in Parkinson’s disease. The two diseases therefore seem to affect different parts of the visual pathway.

webdesign by ablen