Reward learning and statistical learning independently influence attentional priority of salient distractors in visual search

Existing research demonstrates different ways in which attentional prioritization of salient nontarget stimuli is shaped by prior experience: Reward learning renders signals of high-value outcomes more likely to capture attention than signals of low-value outcomes, whereas statistical learning can produce attentional suppression of the location in which salient distractor items are likely to appear. The current study combined manipulations of the value and location associated with salient distractors in visual search to investigate whether these different effects of selection history operate independently or interact to determine overall attentional prioritization of salient distractors. In Experiment 1, high-value and low-value distractors most frequently appeared in the same location; in Experiment 2, high-value and low-value distractors typically appeared in distinct locations. In both experiments, effects of distractor value and location were additive, suggesting that attention-promoting effects of value and attention-suppressing effects of statistical location-learning independently modulate overall attentional priority. Our findings are consistent with a view that sees attention as mediated by a common priority map that receives and integrates separate signals relating to physical salience and value, with signal suppression based on statistical learning determined by physical salience, but not incentive salience.

On the influence of spatial and value attentional cues across individuals

The visual world provides a myriad of cues every instance that can be used to direct information processing. How does the brain integrate predictive information from disparate sources to modify visual priorities, and are combination strategies consistent across individuals? Previous evidence shows that sensory cues that are predictive of the value of a visually guided task (incentive value) and cues that signal where task-relevant stimuli may occur (spatial certainty) act independently to bias attention. Anticipatory accounts propose that both cues are comparably encoded into an attentional priority map, whereas the counterfactual account argues that incentive value cues instead induce a reactive encoding of losses based on the direction of attention. Here we adjudicate between these alternatives and further determine whether there are individual differences in how attentional cues are encoded. 149 participants viewed two coloured placeholders that specified the potential value of correctly identifying an imminent target if it appeared in that specific placeholder. Prior to the target’s presentation, an endogenous spatial cue indicated the target’s more likely location. The anticipatory and counterfactual accounts were used to motivate parametric regressors that were compared in their explanatory power of the observed data, at the group level and on data stratified by a clustering algorithm applied to identify individual differences. The algorithm revealed 2 subtypes in the population; whereas all individuals use spatial certainty cues a subset does not use incentive value cues. However, when used, the influence of incentive value cues reflects a counterfactual loss function. The data show that spatial certainty and incentive value act independently to influence visual priorities because they act at distinct points in information processing, and that theories of motivated attention must account for the non-uniform influence of incentive value on visual priorities.

Impaired selection of a previously ignored singleton: Evidence for salience map plastic changes

Spatial suppression of a salient colour distractor is achievable via statistical learning. Distractor suppression attenuates unwanted capture, but at the same time target selection at the most likely distractor location is impaired. This result corroborates the idea that the distractor salience is attenuated via inhibitory signals applied to the corresponding location in the priority map. What is less clear, however, is whether lingering impairment in target selection when the distractor is removed are due to the proactive strategic maintenance of the suppressive signal at the previous most likely distractor location or result from the fact that suppression has induced plastic changes in the priority map, probably changing input weights. Here, we provide evidence that supports the latter possibility, as we found that impairment in target selection persisted even when the singleton distractor in the training phase became the target of search in a subsequent test phase. This manipulation rules out the possibility that the observed impairments at the previous most likely distractor location were caused by a signal suppression maintained at this location. Rather, the results reveal that the inhibitory signals cause long-lasting changes in the priority map, which affect future computation of the target salience at the same location, and therefore the efficiency of attentional selection.

Across-trial spatial suppression in visual search

In order to focus on objects of interest, humans must be able to avoid distraction by salient stimuli that are not relevant to the task at hand. Many recent studies have shown that through statistical learning we are able to suppress the location that is most likely to contain a salient distractor. Here we demonstrate a remarkable flexibility in attentional suppression. Participants had to search for a shape singleton while a color distractor singleton was present. Unbeknown to the participant, the color distractor was presented according to a consistent pattern across trials. Our findings show that participants learn this distractor sequence as they proactively suppressed the anticipated location of the distractor on the next trial. Critically, none of the participants were aware of these hidden sequences. We conclude that the spatial priority map is highly flexible, operating at a subconscious level preparing the attentional system for what will happen next.

Concurrent statistical learning of target and distractor spatial probability collide on the same spatial priority map(s)

Statistical learning (SL) of both target and distractor spatial probability distributions adjusts the attentional priority of locations. In the presence of a single manipulation for each location, SL also induces indirect effects (e.g., changes in filtering efficiency due to an uneven distribution of targets), suggesting that SL-induced plastic changes are implemented within common spatial priority maps. Here we tested whether, when target- and distractor-related manipulations are concurrently applied to the very same locations, dedicated mechanisms might support the independent encoding of spatial priority in relation to the attentional operation involved. In three related experiments, human healthy participants discriminated the direction of a target arrow, while ignoring a salient distractor, if present; target and distractor spatial probability distributions were systematically manipulated in relation to each single location. Critically, the selection bias produced by the target-related SL was significantly reduced by an adverse distractor contingency. Conversely, the suppression bias generated by the distractor-related SL was erased, or even reversed, by an adverse target contingency. Our results suggest that independent and concomitant target- and distractor-related SL manipulations concur to the plastic adjustment of the same spatial priority map(s), with the resulting priority corresponding to some kind of weighted average of the SL processes.

Distracted by previous reward: integrating selection history, current task demands and saliency in a computational model

Attention can be biased by previous learning and experience. We present analgorithmic-level model of this bias in visual attention that predicts quantitatively howbottom-up, top-down and selection history compete to control attention. In the model,the output of saliency maps as bottom-up guidance interacts with a history map thatencodes learning effects and a top-down task control to prioritize visual features. Wetest the model on a reaction-time (RT) data set from the experiment presented in [1].The model accurately predicts parameters of reaction time distributions from anintegrated priority map that is comprised of an optimal, weighted combination ofseparate maps. Analysis of the weights confirms learning history effects on attentionguidance. The model is able to capture individual differences between participants.Moreover, we demonstrate that a model with a reduced set of maps performs worse,indicating that integrating history, saliency and task information are required for aquantitative description of human attention.

Proactive distractor suppression elicited by statistical regularities in visual search

Irrelevant salient objects may capture our attention and interfere with visual search. Recently, it was shown that distraction by a salient object is reduced when it is presented more frequently at one location than at other locations. The present study investigates whether this reduced distractor interference is the result of proactive spatial suppression, implemented prior to display onset, or reactive suppression, occurring after attention has been directed to that location. Participants were asked to search for a shape singleton in the presence of an irrelevant salient color singleton which was presented more often at one location (the high-probability location) than at all other locations (the low-probability locations). On some trials, instead of the search task, participants performed a probe task, in which they had to detect the offset of a probe dot. The results of the search task replicated previous findings showing reduced distractor interference in trials in which the salient distractor was presented at the high-probability location as compared with the low-probability locations. The probe task showed that reaction times were longer for probes presented at the high-probability location than at the low-probability locations. These results indicate that through statistical learning the location that is likely to contain a distractor is suppressed proactively (i.e., prior to display onset). It suggests that statistical learning modulates the first feed-forward sweep of information processing by deprioritizing locations that are likely to contain a distractor in the spatial priority map.

Seismic risk priorities of site and mid-rise RC buildings in Turkey

Especially, the large-scale loss of life and property caused by the significant earthquakes in recent years has brought the importance of research and measures to be taken on this issue. Determining and analysing the ever-increasing building stock of the cities and detecting and managing all information related to buildings are important in terms of spatial planning and urban transformation. This study aims to determine tectonic characteristics calculating a and b values of Gutenberg- Richter magnitude-frequency relation which forms the basis of earthquake statistics for all cities in Turkey and the reinforced-concrete buildings which are primarily risky in terms of urban transformation. For this purpose, a total of 1620, 5-storey buildings from all provinces of Turkey were assessed. Twenty reinforced concrete buildings from each province were taken into consideration which has 5-stories. The first stage evaluation method specified in the principles regarding the identification of risky buildings issued in 2013 by the Republic of Turkey, Ministry of Environment and Urbanization was used in this study. The performance scores for 1620 buildings were calculated by using this method. A risk priority map was created for the provinces, taking into account for these buildings. The study aims to determine risk priorities of site and mid-rise reinforced-concrete buildings among the cities. The results obtained were interpreted and recommendations were made.

Distractor suppression leads to reduced flanker interference

Recent studies using the additional singleton paradigm have shown that regularities in distractor locations can cause biases in the spatial priority map, such that attentional capture by salient singletons is reduced for locations that are likely to contain distractors. It has been suggested that this type of suppression is proactive (i.e., occurring before display onset). The current study replicated the original findings using an online version of the task. To further assess the suppression of high-probability locations, we employed a congruence manipulation similar to the traditional flanker effect, where distractors could be either congruent or incongruent with the response to the target. Experiment 1 shows that through statistical learning distractor suppression reduces the interference from incongruent distractors, as participants made less errors in high-probability versus low-probability conditions. In Experiment 2, participants were forced to search for a specific target feature (the so-called feature-search mode), which is assumed to allow participants to ignore distractors in a top-down manner. Yet even when this “top-down” search mode was employed, there was still a congruence effect when the distractor singleton was presented at the low-probability but not at the high-probability location. The absence, but not reversal, of a congruence effect at the high-probability location also further indicates that this distractor suppression mechanism is proactive. The results indicate that regardless of the search mode used, there is suppression of the high-probability location indicating that this location competes less for attention within the spatial priority map than all other locations.

History-driven modulations of population codes in early visual cortex during visual search

To find important objects, we must focus on our goals, ignore distractions, and take our changing environment into account. This is formalized in models of visual search whereby goal-driven, stimulus-driven and history-driven factors are integrated into a priority map that guides attention. History is invoked to explain behavioral effects that are neither wholly goal-driven nor stimulus-driven, but whether history likewise alters goal-driven and/or stimulus-driven signatures of neural priority is unknown. We measured fMRI responses in human visual cortex during a visual search task where trial history was manipulated (colors switched unpredictably or repeated). History had a near-constant impact on responses to singleton distractors, but not targets, from V1 through parietal cortex. In contrast, history-independent target enhancement was absent in V1 but increased across regions. Our data suggest that history does not alter goal-driven search templates, but rather modulates canonically stimulus-driven sensory responses to create a temporally-integrated representation of priority.