Elementary Visual Features Research Articles

Strong perceptual consequences of low-level visual predictions: A new illusion

Predicting information is considered to be an efficient strategy to minimise processing costs by exploiting regularities in the environment, and to allow for adaptation in case of irregularities, i.e. prediction errors. How such errors impact conscious perception is unclear, especially when predictions concern elementary visual features. Here we present results from a novel experimental approach allowing us to investigate the perceptual consequences of violated low-level predictions about moving objects. Observers were presented with two squares moving towards each other with a constant speed, and reported whether they were in contact or not before they disappeared. A compelling illusion of a gap between the squares occurred when the leading edges of those squares contacted briefly. The apparent gap was larger than a physical and stable separation of 2.6 min of arc between the squares. The illusion disappeared only when the contact did not violate extrapolations of the contrast edge between the moving object and the background. The pattern of results is consistent with an early locus of the effect and cannot be explained by decisional biases, guesses, top-down, attentional or masking effects. We suggest that violations of the contrast edge extrapolation in the direction of motion have strong perceptual consequences.

Theory of neural coding predicts an upper bound on estimates of memory variability.

Observers reproducing elementary visual features from memory after a short delay produce errors consistent with the encoding-decoding properties of neural populations. While inspired by electrophysiological observations of sensory neurons in cortex, the population coding account of these errors is based on a mathematical idealization of neural response functions that abstracts away most of the heterogeneity and complexity of real neuronal populations. Here we examine a more physiologically grounded model based on the tuning of a large set of neurons recorded in macaque V1 and show that key predictions of the idealized model are preserved. Both models predict long-tailed distributions of error when memory resources are taxed, as observed empirically in behavioral experiments and commonly approximated with a mixture of normal and uniform error components. Specifically, for an idealized homogeneous neural population, the width of the fitted normal distribution cannot exceed the average tuning width of the component neurons, and this also holds to a good approximation for more biologically realistic populations. Examining eight published studies of orientation recall, we find a consistent pattern of results suggestive of a median tuning width of approximately 20°, which compares well with neurophysiological observations. The finding that estimates of variability obtained by the normal-plus-uniform mixture method are bounded from above leads us to reevaluate previous studies that interpreted a saturation in width of the normal component as evidence for fundamental limits on the precision of perception, working memory, and long-term memory.

Rapid Parallel Attentional Selection Can Be Controlled by Shape and Alphanumerical Category.

Previous research has shown that when two color-defined target objects appear in rapid succession at different locations, attention is deployed independently and in parallel to both targets. This study investigated whether this rapid simultaneous attentional target selection mechanism can also be employed in tasks where targets are defined by a different visual feature (shape) or when alphanumerical category is the target selection attribute. Two displays that both contained a target and a nontarget object on opposite sides were presented successively, and the SOA between the two displays was 100, 50, 20, or 10 msec in different blocks. N2pc components were recorded to both targets as a temporal marker of their attentional selection. When observers searched for shape-defined targets (Experiment 1), N2pc components to the two targets were equal in size and overlapped in time when the SOA between the two displays was short, reflecting two parallel shape-guided target selection processes with their own independent time course. Essentially the same temporal pattern of N2pc components was observed when alphanumerical category was the target-defining attribute (Experiment 2), demonstrating that the rapid parallel attentional selection of multiple target objects is not restricted to situations where the deployment of attention can be guided by elementary visual features but that these processes can even be employed in category-based attentional selection tasks. These findings have important implications for our understanding of the cognitive and neural basis of top-down attentional control.

Cross-sensory correspondences and symbolism in spoken and written language.

Lexical sound symbolism in language appears to exploit the feature associations embedded in cross-sensory correspondences. For example, words incorporating relatively high acoustic frequencies (i.e., front/close rather than back/open vowels) are deemed more appropriate as names for concepts associated with brightness, lightness in weight, sharpness, smallness, speed, and thinness, because higher pitched sounds appear to have these cross-sensory features. Correspondences also support prosodic sound symbolism. For example, speakers might raise the fundamental frequency of their voice to emphasize the smallness of the concept they are naming. The conceptual nature of correspondences and their functional bidirectionality indicate they should also support other types of symbolism, including a visual equivalent of prosodic sound symbolism. For example, the correspondence between auditory pitch and visual thinness predicts that a typeface with relatively thin letter strokes will reinforce a word's reference to a relatively high pitch sound (e.g., squeal). An initial rating study confirms that the thinness-thickness of a typeface's letter strokes accesses the same cross-sensory correspondences observed elsewhere. A series of speeded word classification experiments then confirms that the thinness-thickness of letter strokes can facilitate a reader's comprehension of the pitch of a sound named by a word (thinner letter strokes being appropriate for higher pitch sounds), as can the brightness of the text (e.g., white-on-gray text being appropriate for the names of higher pitch sounds). It is proposed that the elementary visual features of text are represented in the same conceptual system as word meaning, allowing cross-sensory correspondences to support visual symbolism in language. (PsycINFO Database Record

Representation of Naturalistic Image Structure in the Primate Visual Cortex.

The perception of complex visual patterns emerges from neuronal activity in a cascade of areas in the primate cerebral cortex. We have probed the early stages of this cascade with "naturalistic" texture stimuli designed to capture key statistical features of natural images. Humans can recognize and classify these synthetic images and are insensitive to distortions that do not alter the local values of these statistics. The responses of neurons in the primary visual cortex, V1, are relatively insensitive to the statistical information in these textures. However, in the area immediately downstream, V2, cells respond more vigorously to these stimuli than to matched control stimuli. Humans show blood-oxygen-level-dependent functional magnetic resonance imaging (BOLD fMRI responses in V1 and V2) that are consistent with the neuronal measurements in macaque. These fMRI measurements, as well as neurophysiological work by others, show that true natural scenes become a more prominent driving feature of cortex downstream from V2. These results suggest a framework for thinking about how information about elementary visual features is transformed into the specific representations of scenes and objects found in areas higher in the visual pathway.

Predictive remapping preserves elementary visual features across saccades

Predictive remapping preserves elementary visual features across saccades

Transformation-invariant visual representations in self-organizing spiking neural networks.

The ventral visual pathway achieves object and face recognition by building transformation-invariant representations from elementary visual features. In previous computer simulation studies with rate-coded neural networks, the development of transformation-invariant representations has been demonstrated using either of two biologically plausible learning mechanisms, Trace learning and Continuous Transformation (CT) learning. However, it has not previously been investigated how transformation-invariant representations may be learned in a more biologically accurate spiking neural network. A key issue is how the synaptic connection strengths in such a spiking network might self-organize through Spike-Time Dependent Plasticity (STDP) where the change in synaptic strength is dependent on the relative times of the spikes emitted by the presynaptic and postsynaptic neurons rather than simply correlated activity driving changes in synaptic efficacy. Here we present simulations with conductance-based integrate-and-fire (IF) neurons using a STDP learning rule to address these gaps in our understanding. It is demonstrated that with the appropriate selection of model parameters and training regime, the spiking network model can utilize either Trace-like or CT-like learning mechanisms to achieve transform-invariant representations.

Necker cube: Stimulus-related (low-level) and percept-related (high-level) EEG signatures early in occipital cortex

During observation of an ambiguous Necker cube, our percept changes spontaneously although the external stimulus does not. An EEG paradigm allowing time-resolved EEG measurement during endogenous perceptual reversals recently revealed a chain of ERP correlates beginning with an early occipital positivity at around 130 ms (Reversal Positivity, "RP"). In order to better understand the functional role of this RP, we investigated its relation to the P100, which is spatiotemporally close, typically occurring 100 ms after onset of a visual stimulus at occipital electrodes. We compared the relation of the ERP amplitudes to varying sizes of ambiguous Necker cubes. The main results are: (1) The P100 amplitude increases monotonically with stimulus size but is independent of the participants' percept. (2) The RP, in contrast, is percept-related and largely unaffected by stimulus size. (3) A similar pattern to RP was found for reaction times: They depend on the percept but not on stimulus size. We speculate that the P100 reflects processing of elementary visual features, while the RP is related to a processing conflict during 3D interpretation that precedes a reversal. The present results indicate that low-level visual processing (related to stimulus size) and (relative) high-level processing (related to perceptual reversal) occur in close spatial and temporal vicinity.

Emotion-induced trade-offs in spatiotemporal vision.

It is generally assumed that emotion facilitates human vision in order to promote adaptive responses to a potential threat in the environment. Surprisingly, we recently found that emotion in some cases impairs the perception of elementary visual features (Bocanegra & Zeelenberg, 2009b). Here, we demonstrate that emotion improves fast temporal vision at the expense of fine-grained spatial vision. We tested participants' threshold resolution with Landolt circles containing a small spatial or brief temporal discontinuity. The prior presentation of a fearful face cue, compared with a neutral face cue, impaired spatial resolution but improved temporal resolution. In addition, we show that these benefits and deficits were triggered selectively by the global configural properties of the faces, which were transmitted only through low spatial frequencies. Critically, the common locus of these opposite effects suggests a trade-off between magno- and parvocellular-type visual channels, which contradicts the common assumption that emotion invariably improves vision. We show that, rather than being a general "boost" for all visual features, affective neural circuits sacrifice the slower processing of small details for a coarser but faster visual signal.

Evidence for long-range feedback in target detection: Detection of semantic targets modulates activity in early visual areas

In a variety of attention and search tasks, single-cell recordings of the primate brain have frequently shown an enhancement of responses in early visual areas to selected target stimuli. This enhancement is observed only at longer latencies, suggesting the possibility that it reflects the action of feedback or return signals from upstream processing areas. However, in typical studies, targets are specified on the basis of elementary visual features; as these are coded at multiple levels of the visual system, it is impossible to determine where enhanced target processing begins. Using human functional magnetic resonance imaging (fMRI), we demonstrate enhancement of activity in early visual areas even when low-level visual information is insufficient for target detection to occur. We found enhanced activity in early visual areas to targets defined purely by semantic category, suggesting that feedback signals returning from at least as far forward as temporal lobe semantic processing can influence visual responses. These findings also suggest feedback signaling as a mechanism by which early and late brain systems coding for different properties of a target object can integrate their activity, allowing for the target object to dominate overall processing.

Age-related differences in auditory evoked responses during rapid perceptual learning

Objective Young and older adults can learn to rapidly discriminate between elementary visual and auditory features. While growing evidence supports the notion that such behavioral improvement is paralleled by neuroplastic changes in corresponding sensory areas during adulthood, studies have not examined practice-related improvement in older adults and the corresponding changes in neural activity. Methods We used event-related potentials (ERPs) to investigate the effects of age on rapid learning-related changes in listeners’ ability to identify two phonetically different vowels presented simultaneously. Results During the first hour of testing, young and older listeners showed comparable behavioral improvement in identifying both vowels. In young adults, learning was paralleled by enhanced amplitudes of early (130 ms) and late (320 ms) ERP waves over the right temporal lobe, as well as an increased negative wave over the midline parietal region, peaking at about 400 ms after sound onset. The practice-related changes over the right temporal lobe were not present in older adults whereas the learning effect observed over the parietal region was present in both young and older adults. In older adults, behavioral improvement was also associated with reduced N1 amplitude recorded at inferior and posterior temporal/occipital scalp sites while no such changes were observed in young adults. Conclusions Age-related differences in neural activity during learning suggest that neural networks supporting behavioral improvements in speech segregation and identification change during the course of aging. Significance This research highlights the role of practice on concurrent sound perception and may facilitate the development of training programs that may help older listeners to parse the auditory scene into component sound sources.

Synergy of features enables detection of texture defined figures

Traditional theories of early visual processing suggest that elementary visual features are handled in parallel by independent neural pathways. We studied the interaction of orientation and spatial frequency in the discrimination of Gabor random fields. Target textures differed from reference textures either in mean feature value, showing an edge-like transition between both textures (edge defined), or in the degree of feature homogeneity with smooth transitions (region defined). Irrespective of the kind of texture definition, we found strong cue summation for targets defined by both cues simultaneously, provided two conditions were fulfilled. First, they were barely discriminable when defined by one cue alone. Second, the target elements formed a closed 2D surface. Only marginal cue summation was observed when target elements were heterogeneously distributed in a predefined area, lacking a clear 2D shape. Our findings indicate that feature synergy enables figure-ground segregation when the information from independent feature-specific pathways is insufficient for solving this task.

Involvement of the human frontal eye field and multiple parietal areas in covert visual selection during conjunction search.

Searching for a target object in a cluttered visual scene requires active visual attention if the target differs from distractors not by elementary visual features but rather by a feature conjunction. We used functional magnetic resonance imaging (fMRI) in human subjects to investigate the functional neuroanatomy of attentional mechanisms employed during conjunction search. In the experimental condition, subjects searched for a target defined by a conjunction of colour and orientation. In the baseline condition, subjects searched for a uniquely coloured target, regardless of its orientation. Eye movement recordings outside the scanner verified subjects' ability to maintain fixation during search. Reaction times indicated that the experimental condition was attentionally more demanding than the baseline condition. Differential activations between conditions were therefore ascribed to top-down modulation of neural activity. The frontal eye field, the ventral precentral sulcus and the following posterior parietal regions were consistently activated: (i) the postcentral sulcus; (ii) the posterior; and (iii) the anterior part of the intraparietal sulcus; and (iv) the junction of the intraparietal with the transverse occipital sulcus. Parietal regions were spatially distinct and displayed differential amplitudes of signal increase with a maximal amplitude in the posterior intraparietal sulcus. Less consistent activation was found in the lateral fusiform gyrus. These results suggest an involvement of the human frontal eye field in covert visual selection of potential targets during search. These results also provide evidence for a subdivision of posterior parietal cortex in multiple areas participating in covert visual selection, with a major contribution of the posterior intraparietal sulcus.

Associative Visual Agnosia Resulting from a Disconnection Between Intact Visual Memory and Semantic Systems

We report the case of a patient (RC) who developed a severe visual agnosia, associated to alexia without agraphia, color anomia and amnesia, following an ischemic stroke in the territory supplied by the left posterior cerebral artery. Based on his proficient performance on tests evaluating analysis of elementary visual features, formation of viewer-centered and object-centered representations of visual stimuli and discrimination between drawings representing real and unreal objects, we concluded that the critical locus of deficit was a disconnection between the normally functioning visual memory store and the semantic system. RC's disturbance in visual processing of human faces paralleled his recognition disorder of other classes of objects. The possible contribution of neurobiological factors in determining RC's agnosic deficit is discussed.

The effects of unilateral visuospatial neglect on perception of Müller-Lyer illusory figures.

Left visuospatial neglect after right hemisphere damage is a lateralised disorder of spatial perception and cognition. A study is reported of the extent to which patients with left visuospatial neglect are susceptible to the illusory effects normally elicited by Müller-Lyer figures, in which inducing fins are located on one end or on both ends of a horizontal line. Seven patients with left visuospatial neglect and seven normal controls were tested on a task of horizontal-line bisection in which stimuli consisted of plain lines of three lengths (100, 150, and 200 mm), randomly interleaved with lines with unilateral or bilateral fins. As a group, normals made accurate bisection judgments in the baseline (no-fins) condition, and exhibited significant illusory effects in both the unilateral-fin and the bilateral-fin conditions. In contrast, patients made substantial rightward errors in the baseline condition, consistent with their neglect of the left end of the line. However, as a group, patients still exhibited significant illusory effects with left-sided outward-projecting fins on 100 mm lines and with left-sided inward-projecting fins on 150 mm lines. Moreover, at least one patient exhibited consistent illusory effects both for inward-projecting and for outward-projecting left-sided fins at all line lengths. Normal illusory effects in patients were also obtained with stimuli containing unilateral right-sided and bilateral fins. The existence of such effects with inducing elements on the contralesional extremity of horizontal line stimuli suggests preservation of low-level, perhaps preattentive, perceptual mechanisms responsible for coding elementary visual features. The results may be relevant for an understanding of the influence of attentional factors on illusory perception.