Stimulus-response Mapping Research Articles

The attention-gated reinforcement learning model can explain experimentally observed changes in tuning curves that follow category learning

A problem with many traditional neural network learning algorithms is that they lack a biologically plausible method of assigning credit to the nodes in the earlier levels of the network that played a decisive role in the stimulus response mapping. In the Attention-Gated Reinforcement Learning Model (AGREL) [1] the credit assignment problem is overcome by the focal influence of attention through the feedback connections in the network. In each trial, a global reward value (conveyed by neuromodulators) is calculated and serves to increase the likelihood that successful behavior be repeated. The activity in the output level of the network (which is limited so that only a single node at a time may be active/attended) is back propagated through the network via feedback connections. The feedback connections, which are reciprocal in strength to the feedforward connections, allow the network to selectively target the lower level nodes that drove the current output, which in combination with the global reward factor, allows the network to enhance connections with nodes that caused successful behaviors, see figure 1. This attentional gating of the reward signal offers a biologically realistic solution to the credit assignment problem and in so doing provides a coherent and unifying framework for learning, with attentional selection at its core. AGREL has previously been shown to closely replicate the changes in tuning curves observed in two physiological categorization tasks [1]. In the current study, we first replicate the findings of two further physiological tasks using AGREL, and second, examine the effect of removing the feedback connections, on our models performance in these and the previous tasks.

Formal concepts expressed by compositional hierarchical Hebbian cell assemblies

Hebbian cell assemblies can be formalised as sets of tightly connected cells in autoand hetero-associative memories. Direct evidence for such has recently been obtained in multiple-unit recordings from rat hippocampal neurons [1,2]. These experiments suggest a hierarchical organisation where cliques are embedded in each other such that larger cliques represent less specific stimulus conditions. We here suggest an extended interpretation stating that the firing patterns may not just reflect nested categories but a lattice of concepts about stimulus-response mappings in the sense of formal concept analysis, an applied branch of set theory. We present an implementation of formal concept lattices in bidirectional associative memories that, in contrast to previous work by Belohlavek [3], satisfies Dale's principle and uses balanced excitation and inhibition [4]. Inhibitory cells have fixed, non-plastic synapses even if the model learns new concepts. As an extreme case a single global inhibitory cell is enough that controls the total level of activation. The excitatory cells can further learn incrementally using a Hebbian coincidence learning rule. A spiking neuron variant of the model is further presented and implications of the model for retrieval in auto-associative memories are outlined. Overall the model is well suited for representing hierarchical compositional relationships between entities in the form of correlated memory traces in the brain. Acknowledgements The author acknowledges support by the Engineering and Physical Sciences Research Council of the United Kingdom (EP/C010841/1, COLAMN – A Novel Computing Architecture for Cognitive Systems based on the Laminar Microcircuitry of the Neocortex) and the European Community (IST2001-35282, FACETS – Fast Analog Computing with Emergent Transient States).

Stimulus–Response compatibility based on affective arousal

Participants responded by producing weak or strong forces to pictures that varied with respect to affective arousal and horizontal position. The task relevance of these stimulus dimensions was manipulated between different groups of participants, with Stimulus–Response mapping varying between blocks of trials. The combination of the respectively task-irrelevant stimulus features with responses varied unpredictably on a trial-to-trial basis. In addition, the valence of the pictures and the combination of valence with response force were varied. When affective arousal was task relevant, there was an advantage of the pairing arousing→strong, not arousing→weak over the reverse mapping. No such affectively-based compatibility effect emerged when arousal was task irrelevant. In addition, effects of arousal were modulated by valence, but only when arousal was the relevant stimulus dimension.

Determinants of the benefit for consistent stimulus-response mappings in dual-task performance of four-choice tasks

Reaction times are shorter when the stimulus-response mappings for pairs of three-choice tasks are consistent (both corresponding or both mirrored) than when they are inconsistent. The benefit for consistent mirrored mappings is evident at the side positions for each task, for which the responses are crossed, but not at the middle position, for which the response is corresponding. In the present study, we report experiments in which we tested implications of an emergent mapping-choice account of the consistency benefit using pairs of four-choice tasks. This procedure allows crossed responses for all positions when the mapping is mirrored and use of mixed mappings for which one pair of stimuli and responses within a task has a corresponding assignment and the other a crossed assignment. Results showed that when a pure corresponding or pure mirrored mapping was used, there was a consistency benefit for both the middle and side positions. However, when the mixed mapping was introduced, the consistency benefit for that mapping depended on the overall complexity of the set of individual stimulus-response pairings for the combined tasks. A mapping choice between Tasks 1 and 2 is only one of several emergent processes that contribute to response-selection efficiency in dual-task contexts.

Transfer of orthogonal stimulus–response mappings to an orthogonal Simon task

When up-down stimulus locations are mapped to left-right keypresses, an overall advantage for the up-right/down-left mapping is often obtained that varies as a function of response eccentricity. This orthogonal stimulus-response compatibility (SRC) effect also occurs when stimulus location is irrelevant, a phenomenon called the orthogonal Simon effect, and has been attributed to correspondence of stimulus and response code polarities. The Simon effect for horizontal stimulus-response (S-R) arrangements has been shown to be affected by short-term S-R associations established through the mapping used for a prior SRC task in which stimulus location was relevant. We examined whether such associations also transfer between orthogonal SRC and Simon tasks and whether correspondence of code polarities continues to contribute to performance in the Simon task. In Experiment 1, the orthogonal Simon effect was larger after practising with an up-right/down-left mapping of visual stimuli to responses than with the alternative mapping, for which the orthogonal Simon effect tended to reverse. Experiment 2 showed similar results when practice was with high (up) and low (down) pitch tones, though the influence of practice mapping was not as large as that in Experiment 1, implying that the short-term S-R associations acquired in practice are at least in part not modality specific. In Experiment 3, response eccentricity and practice mapping were shown to have separate influences on the orthogonal Simon effect, as expected if both code polarity and acquired S-R associations contribute to performance.

A Domain-Independent Source of Cognitive Control for Task Sets: Shifting Spatial Attention and Switching Categorization Rules

To optimize task performance as circumstances unfold, cognitive control mechanisms configure the brain to prepare for upcoming events through voluntary shifts in task set. A foundational unanswered question concerns whether different domains of cognitive control (e.g., spatial attention shifts, shifts between categorization rules, or shifts between stimulus-response mapping rules) are associated with separate, domain-specific control mechanisms, or whether a common, domain-independent source of control initiates shifts in all domains. Previous studies have tested different domains of cognitive control in separate groups of subjects using different paradigms, yielding equivocal conclusions. Here, using rapid event-related MRI, we report evidence from a single paradigm in which subjects were cued to perform both shifts of spatial attention and switches between categorization rules. A conjunction analysis revealed a common transient signal evoked by switch cues in medial superior parietal lobule for both domains of control, revealing a single domain-independent control mechanism.

Effects of stimulus–response compatibility on inhibitory processes in Parkinson’s disease

Parkinson's disease (PD) is a neurodegenerative basal ganglia disorder accompanied by deficits in cognitive functions. One important executive function is the inhibition of responses. Due to basal ganglia damage, processes related to the selection of response are also dysfunctional. However, the relevance of deficits in response selection to processes related to response inhibition in PD is not clear. In this study we examined these processes by means of event-related potentials (ERPs) in two Go/Nogo tasks. In one task the stimulus-response mapping was compatible and in the other task it was incompatible with the meaning of the stimuli. The behavioural results show that PD patients were unaffected in the compatible response inhibition task but encountered problems in the incompatible one. In the ERPs the N2, generally reflecting response selection, was delayed for the PD compared to the control group. This suggests that response selection is delayed in PD. Moreover, the N2 was specifically enhanced in Nogo trials. This indicates that premotor inhibition, which is probably reflected by the Nogo-N2, is intensified in PD. The P3 was specifically attenuated and delayed after Nogo stimuli in the incompatible condition for PDs. Assuming that the Nogo-P3 reflects the evaluation of successful motor inhibition, our data show that this process is attenuated and delayed in PD but mainly in the incompatible task. The results suggest that inhibitory deficits in PD are only evident in complex (incompatible) stimulus-response mappings. These effects are probably due to an overstrain of striatal processes.

On the Natural Hierarchical Composition of Cliques in Cell Assemblies

Hebbian cell assemblies can be formalised as sets of tightly connected cells in auto- and hetero-associative memories. Direct evidence for such “cliques” has recently been obtained in multiple-unit recordings from rat hippocampal neurons. These experiments suggest a hierarchical organisation where cliques are embedded in each other such that larger cliques represent less specific stimulus conditions. We here suggest an interpretation stating that the firing patterns may not just reflect nested categories but a lattice of concepts about stimulus–response mappings in the sense of formal concept analysis, an applied branch of set theory. We present an implementation of formal concept lattices in bidirectional associative memories that in contrast to previous work satisfies Dale’s principle and uses balanced excitation and inhibition. Inhibitory cells have fixed, non-plastic synapses even if the model learns new concepts. As an extreme case a single global inhibitory cell is enough that controls the total level of activation. The excitatory cells can further learn incrementally using a Hebbian coincidence learning rule. Implications of the model for retrieval in auto-associative memories are further outlined. Overall the model is well suited for representing hierarchical compositional relationships between entities in the form of correlated patterns in technical cognitive systems and potentially the brain.

Evidence for Age-related Changes to Temporal Attention and Memory from the Choice Time Production Task

ABSTRACT The effect of aging on interval timing was examined using a choice time production task, which required participants to choose a key response based on the location of the stimulus, but to delay responding until after a learned time interval. Experiment 1 varied attentional demands of the response choice portion of the task by varying difficulty of stimulus–response mapping. Choice difficulty affected temporal accuracy equally in both age groups, but older participants' response latencies were more variable under more difficult response choice conditions. Experiment 2 tested the contribution of long-term memory to differences in choice time production between age groups over 3 days of testing. Direction of errors in time production between the two age groups diverged over the 3 sessions, but variability did not differ. Results from each experiment separately show age-related changes to attention and memory in temporal processing using different measures and manipulations in the same task.

The Brief Implicit Association Test

The Brief Implicit Association Test (BIAT) consists of two blocks of trials with the same four categories and stimulus-response mappings as the standard IAT, but with 1/3 the number of trials. Unlike the standard IAT, the BIAT focuses the subject on just two of each block's four categories. Experiments 1 and 2 demonstrated that attitude BIATs had satisfactory validity when good (but not bad) was a focal category, and that identity IATs had satisfactory validity when self (but not other) was a focal category. Experiment 2 also showed that a good-focal attitude BIAT and a self-focal identity BIAT were psychometrically similar to standard IAT measures of the same constructs. Experiment 3 presented each of six BIATs twice, showing that procedural variables had no more than minor influences on the resulting implicit measures. Experiment 4 further demonstrated successful use of the BIAT to measure implicit stereotypes.

Dual-task crosstalk between saccades and manual responses.

Between-task crosstalk has been discussed as an important source for dual-task costs. In this study, the authors examine concurrently performed saccades and manual responses as a means of studying the role of response-code conflict between 2 tasks. In Experiment 1, participants responded to an imperative auditory stimulus with a left or a right key press (manual task), a left or a right saccade (saccade task), or both. In Experiments 2 and 3, participants crossed their hands, and a modest (Experiment 2) or substantial (Experiment 3) degree of between-task response-code conflict through specific instructions was introduced. In Experiment 4, response codes across tasks were compatible, and stimulus-response mappings in both tasks were incompatible. Overall, the results indicate that performance not only in manual responses but also in saccades suffers from dual-task conditions, even though saccades were typically performed first and are usually assumed to be controlled quite independently. Moreover, the systematic introduction of response-code conflict between tasks modulated the pattern of dual-task performance. The authors propose confusability of response codes as an underlying mechanism of the observed effects of between-task crosstalk.

How task representations guide attention: Further evidence for the shielding function of task sets.

To pursue goal directed behavior, the cognitive system must be shielded against interference from irrelevant information. Aside from the online adjustment of cognitive control widely discussed in the literature, an additional mechanism of preventive goal shielding is suggested that circumvents irrelevant information from being processed in the first place. Participants had to react to 8 different words depicting clothing items that were presented in front of line drawings that could be either semantically related (clothes) or unrelated (animals with spatial orientation) to the target words. Participants either learned the stimulus-response (S-R) mappings by heart or used 1 task set (TS). In the S-R group, semantically related and unrelated distractors interfered with performance, whereas in the TS group, only semantically related distractors interfered, and unrelated distractors had no effect. It follows that task representations based on a general TS help to focus attention on relevant information, thereby preventing the processing of irrelevant information.

Mechanisms of masked priming: A meta-analysis.

The extent to which unconscious information can influence behavior has been a topic of considerable debate throughout the history of psychology. A frequently used method for studying subliminal processing is the masked priming paradigm. The authors focused on studies in which this paradigm was used. Their aim was twofold: first, to assess the magnitude of subliminal priming across the literature and to determine whether subliminal primes are processed semantically, and second, to examine potential moderators of priming effects. The authors found significant priming in their analyses, indicating that unconsciously presented information can influence behavior. Furthermore, priming was observed under circumstances in which a nonsemantic interpretation could not fully explain the effects, suggesting that subliminally presented information can be processed semantically. Nonetheless, the nonsemantic processing of primes is enhanced and priming effects are boosted when the experimental context allows the formation of automatic stimulus-response mappings. This quantitative review also revealed several moderators that influence the strength of priming. (PsycINFO Database Record (c) 2009 APA, all rights reserved).

On the neural control of social emotional behavior

It is known that the orbitofrontal cortex (OFC) is crucially involved in emotion regulation. However, the specific role of the OFC in controlling the behavior evoked by these emotions, such as approach-avoidance (AA) responses, remains largely unexplored. We measured behavioral and neural responses (using fMRI) during the performance of a social task, a reaction time (RT) task where subjects approached or avoided visually presented emotional faces by pulling or pushing a joystick, respectively. RTs were longer for affect-incongruent responses (approach angry faces and avoid happy faces) as compared to affect-congruent responses (approach-happy; avoid-angry). Moreover, affect-incongruent responses recruited increased activity in the left lateral OFC. These behavioral and neural effects emerged only when the subjects responded explicitly to the emotional value of the faces (AA-task) and largely disappeared when subjects responded to an affectively irrelevant feature of the faces during a control (gender evaluation: GE) task. Most crucially, the size of the OFC-effect correlated positively with the size of the behavioral costs of approaching angry faces. These findings qualify the role of the lateral OFC in the voluntary control of social-motivational behavior, emphasizing the relevance of this region for selecting rule-driven stimulus-response associations, while overriding automatic (affect-congruent) stimulus-response mappings.

Does training under consistent mapping conditions lead to automatic attention attraction to targets in search tasks?

Schneider and Shiffrin (1977) proposed that training under consistent stimulus-response mapping (CM) leads to automatic target detection in search tasks. Other theories, such as Treisman and Gelade's (1980) feature integration theory, consider target-distractor discriminability as the main determinant of search performance. The first two experiments pit these two principles against each other. The results show that CM training is neither necessary nor sufficient to achieve optimal search performance. Two other experiments examine whether CM trained targets, presented as distractors in unattended display locations, attract attention away from current targets. The results are again found to vary with target-distractor similarity. Overall, the present study strongly suggests that CM training does not invariably lead to automatic attention attraction in search tasks.

How the brain resolves high conflict situations: Double conflict involvement of dorsolateral prefrontal cortex

Executive control is a human ability that allows to overcome automatic stimulus-response mappings and to act appropriate in the context of a task where the selection of relevant stimuli and the suppression of interfering information are crucial. In order to address the question which brain areas are involved in the detection and processing of two simultaneously operating sources of interference derived from a spatial incompatibility task, we used functional MRI to contrast neural activity related to a double conflict situation to single incompatibility conditions. Results show signal increase of left dorsolateral prefrontal cortex when monitoring simultaneously presented conflict. There was no additional activity in the medial prefrontal cortex or anterior cingulate cortex although these regions are expected to play an important role in all types of conflict monitoring. Further analyses of conflict resolution and post-error adaptation pointed to different underlying functional mechanisms. While the resolution of high conflict was associated with rostral ACC activation, the post-error adaptation reflecting activity during post-error trials suggests a specific medial and lateral prefrontal network which was functionally distinct from conflict-related activity. Our results also suggest a major role for the basal ganglia during error detection and resolution.

An action sequence held in memory can interfere with response selection of a target stimulus, but does not interfere with response activation of noise stimuli

Withholding an action plan in memory for later execution can delay execution of another action if the actions share a similar (compatible) action feature (e.g., response hand). We investigated whether this phenomenon, termed compatibility interference (CI), occurs for responses associated with a target as well as responses associated with distractors in a visual selection task. Participants planned and withheld a sequence of keypress responses (with their right or left hand), according to the identity of a stimulus (A), and then immediately executed a keypress response (with their right or left hand) to a second stimulus (B), according to the identity of a target letter appearing alone or among distractor letters. Distractor letters were either response compatible or incompatible with the target and appeared either simultaneously with the target (Experiments 1A and 2) or 100 msec before the target (Experiment 1B). Also, stimulus-response mapping was either 1:1 (Experiment 1) or 2:1 (Experiment 2). Results showed that the response to the Stimulus B target was delayed when it required the same response hand as Stimulus A, as opposed to a different hand. Also, the target reaction time for Stimulus B was greater when the target was flanked by incompatible distractors than when it was flanked by compatible distractors. Moreover, the degree of CI was consistent across the compatible-, incompatible-, and no-distractor conditions, indicating that CI generalizes to responses associated with a target, but not to those associated with distractors. Thus, CI occurs at a response selection, not at a response activation stage. Implications for the code occupation account for CI (e.g., Stoet & Hommel, 1999, 2002) and an alternative account for CI are discussed.

RT distribution analysis of category congruence effects with masked primes

In the number magnitude decision task ("Is the number bigger/smaller than 5?"), response to a target (e.g., 3) is faster following a masked prime congruent with the target (e.g., 1) than it is following an incongruent prime (e.g., 9). This category congruence effect has been reported to be "interference-dominant" relative to a neutral prime (e.g., the # sign, the number 5) on the basis of the analysis of mean response time (RT). Using RT distribution analysis as well as mean RTs, we identified two bases for this pattern. One relates to the choice of neutral baseline: The # prime, unlike the digit prime, does not factor in the cost of perceptual transition between the prime and target, and therefore underestimates facilitation and overestimates the interference effect. The second basis of the interference-dominant pattern is a disproportionate slowdown of congruent trials in the slow RT bins. Furthermore, this slowdown is greater for primes that had been used as targets than it is with "novel" primes that have not been responded to as targets. We interpret the results as suggesting that the category congruence effect has two components with different time courses-one based on stimulus-response mapping, and the other on semantic categorization.

Task Complexity Enhances Response Inhibition Deficits in Childhood and Adolescent Attention-Deficit/Hyperactivity Disorder: A Meta-Regression Analysis

The ability to inhibit motor responses, as assessed by the stop-signal reaction time (SSRT), is impaired in children and adolescents with attention-deficit/hyperactivity disorder (ADHD). However, the between-study variation in effect sizes is large. The aim of this study was to investigate whether this variability can be explained by between-study variation in Go task complexity. Forty-one studies comparing children or adolescents diagnosed with ADHD to normal control subjects were incorporated in a random-effects meta-regression analysis. The independent variables were a global index of Go task complexity (i.e., mean reaction time in control subjects [RTc]) and a more specific index (i.e., spatial compatibility of the stimulus-response mapping). The dependent variable was the SSRT difference between ADHD and control subjects. The SSRT difference increased significantly with increasing RTc. Moreover, the SSRT difference was significantly increased in studies that employed a noncompatible, that is, arbitrary, mapping compared with studies that incorporated a spatially compatible stimulus-response mapping. These results indicate that inhibitory dysfunction in children and adolescents with ADHD varies with task complexity: inhibitory dysfunction in ADHD is most pronounced for spatially noncompatible responses. Explanations in terms of inhibition and working memory deficits and a tentative neurobiological explanation are briefly discussed.

Age differences in response selection for pure and mixed stimulus–response mappings and tasks

Two experiments examined effects of mixed stimulus–response mappings and tasks for older and younger adults. In Experiment 1, participants performed two-choice spatial reaction tasks with blocks of pure and mixed compatible and incompatible mappings. In Experiment 2, a compatible or incompatible mapping was mixed with a Simon task for which the mapping of stimulus color to location was relevant and stimulus location was irrelevant. In both experiments, older adults showed larger mixing costs than younger adults and larger compatibility effects, with the differences particularly pronounced in Experiment 1 when location mappings were mixed. In mixed conditions, when stimulus location was relevant, older adults benefited more than younger adults from complete repetition of the task and stimulus from the preceding trial. When stimulus location was irrelevant, the benefit of complete repetition did not differ reliably between age groups. The results suggest that the age-related deficit associated with mixing mappings and tasks is primarily due to older adults having more difficulty separating task sets that activate conflicting response codes.