Shared attention for action selection and action monitoring in goal-directed reaching.
Ontology highlight
ABSTRACT:
SUBMITTER: Mahon A
PROVIDER: S-EPMC7040085 | biostudies-literature | 2020 Mar
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Everyday behavior frequently involves encounters with multiple objects that compete for selection. For example, driving a car requires constant shifts of attention between oncoming traffic, rearview mirrors, and traffic signs and signals, among other objects. Behavioral goals often drive this selection process [1, 2]; however, they are not the sole determinant of selection. Physically salient objects, such as flashing, brightly colored hazard signs, or objects that are salient by virtue of learned associations with reward, such as pictures of food on a billboard, often capture attention regardless of the individual's goals [3-6]. It is typically thought that strongly salient distractor objects capture more attention and are more disruptive than weakly salient distractors [7, 8]. Counterintuitively, though, we found that this is true for perception, but not for goal-directed action. In a visually guided reaching task [9-11], we required participants to reach to a shape-defined target while trying to ignore salient distractors. We observed that strongly salient distractors produced less disruption in goal-directed action than weakly salient distractors. Thus, a strongly salient distractor triggers suppression during goal-directed action, resulting in enhanced efficiency and accuracy of target selection relative to when weakly salient distractors are present. In contrast, in a task requiring no goal-directed action, we found greater attentional interference from strongly salient distractors. Thus, while highly salient stimuli interfere strongly with perceptual processing, increased physical salience or associated value attenuates action-related interference.
| S-EPMC4526432 | biostudies-literature
Project description:Human gaze behavior prediction is important for behavioral vision and for computer vision applications. Most models mainly focus on predicting free-viewing behavior using saliency maps, but do not generalize to goal-directed behavior, such as when a person searches for a visual target object. We propose the first inverse reinforcement learning (IRL) model to learn the internal reward function and policy used by humans during visual search. We modeled the viewer's internal belief states as dynamic contextual belief maps of object locations. These maps were learned and then used to predict behavioral scanpaths for multiple target categories. To train and evaluate our IRL model we created COCO-Search18, which is now the largest dataset of high-quality search fixations in existence. COCO-Search18 has 10 participants searching for each of 18 target-object categories in 6202 images, making about 300,000 goal-directed fixations. When trained and evaluated on COCO-Search18, the IRL model outperformed baseline models in predicting search fixation scanpaths, both in terms of similarity to human search behavior and search efficiency. Finally, reward maps recovered by the IRL model reveal distinctive target-dependent patterns of object prioritization, which we interpret as a learned object context.
| S-EPMC8218821 | biostudies-literature
Project description:Understanding how goals control behavior is a question ripe for interrogation by new methods from machine learning. These methods require large and labeled datasets to train models. To annotate a large-scale image dataset with observed search fixations, we collected 16,184 fixations from people searching for either microwaves or clocks in a dataset of 4,366 images (MS-COCO). We then used this behaviorally-annotated dataset and the machine learning method of inverse-reinforcement learning (IRL) to learn target-specific reward functions and policies for these two target goals. Finally, we used these learned policies to predict the fixations of 60 new behavioral searchers (clock = 30, microwave = 30) in a disjoint test dataset of kitchen scenes depicting both a microwave and a clock (thus controlling for differences in low-level image contrast). We found that the IRL model predicted behavioral search efficiency and fixation-density maps using multiple metrics. Moreover, reward maps from the IRL model revealed target-specific patterns that suggest, not just attention guidance by target features, but also guidance by scene context (e.g., fixations along walls in the search of clocks). Using machine learning and the psychologically meaningful principle of reward, it is possible to learn the visual features used in goal-directed attention control.
| S-EPMC8218820 | biostudies-literature
Project description:In humans, sensory afferences are combined and integrated by the central nervous system (Ernst MO, Bülthoff HH (2004) Trends Cogn. Sci. 8: 162-169) and appear to provide a holistic representation of the environment. Empirical studies have repeatedly shown that vision dominates the other senses, especially for tasks with spatial demands. In contrast, it has also been observed that sound can strongly alter the perception of visual events. For example, when presented with 2 flashes and 1 beep in a very brief period of time, humans often report seeing 1 flash (i.e. fusion illusion, Andersen TS, Tiippana K, Sams M (2004) Brain Res. Cogn. Brain Res. 21: 301-308). However, it is not known how an unfolding movement modulates the contribution of vision to perception. Here, we used the audio-visual illusion to demonstrate that goal-directed movements can alter visual information processing in real-time. Specifically, the fusion illusion was linearly reduced as a function of limb velocity. These results suggest that cue combination and integration can be modulated in real-time by goal-directed behaviors; perhaps through sensory gating (Chapman CE, Beauchamp E (2006) J. Neurophysiol. 96: 1664-1675) and/or altered sensory noise (Ernst MO, Bülthoff HH (2004) Trends Cogn. Sci. 8: 162-169) during limb movements.
| S-EPMC2813281 | biostudies-literature
Project description:Attention control is a basic behavioral process that has been studied for decades. The currently best models of attention control are deep networks trained on free-viewing behavior to predict bottom-up attention control - saliency. We introduce COCO-Search18, the first dataset of laboratory-quality goal-directed behavior large enough to train deep-network models. We collected eye-movement behavior from 10 people searching for each of 18 target-object categories in 6202 natural-scene images, yielding [Formula: see text] 300,000 search fixations. We thoroughly characterize COCO-Search18, and benchmark it using three machine-learning methods: a ResNet50 object detector, a ResNet50 trained on fixation-density maps, and an inverse-reinforcement-learning model trained on behavioral search scanpaths. Models were also trained/tested on images transformed to approximate a foveated retina, a fundamental biological constraint. These models, each having a different reliance on behavioral training, collectively comprise the new state-of-the-art in predicting goal-directed search fixations. Our expectation is that future work using COCO-Search18 will far surpass these initial efforts, finding applications in domains ranging from human-computer interactive systems that can anticipate a person's intent and render assistance to the potentially early identification of attention-related clinical disorders (ADHD, PTSD, phobia) based on deviation from neurotypical fixation behavior.
| S-EPMC8062491 | biostudies-literature
Project description:BackgroundDistinguishing between actions that are more likely or less likely to be rewarded is a critical aspect of goal-directed decision making. However, neuroanatomic and molecular mechanisms are not fully understood.MethodsWe used anterograde tracing, viral-mediated gene silencing, functional disconnection strategies, pharmacologic rescue, and designer receptors exclusively activated by designer drugs (DREADDs) to determine the anatomic and functional connectivity between the orbitofrontal cortex (OFC) and the amygdala in mice. In particular, we knocked down brain-derived neurotrophic factor (Bdnf) bilaterally in the OFC or generated an OFC-amygdala "disconnection" by pairing unilateral OFC Bdnf knockdown with lesions of the contralateral amygdala. We characterized decision-making strategies using a task in which mice selected actions based on the likelihood that they would be reinforced. Additionally, we assessed the effects of DREADD-mediated OFC inhibition on the consolidation of action-outcome conditioning.ResultsAs in other species, the OFC projects to the basolateral amygdala and dorsal striatum in mice. Bilateral Bdnf knockdown within the ventrolateral OFC and unilateral Bdnf knockdown accompanied by lesions of the contralateral amygdala impede goal-directed response selection, implicating BDNF-expressing OFC projection neurons in selecting actions based on their consequences. The tyrosine receptor kinase B agonist 7,8-dihydroxyflavone rescues action selection and increases dendritic spine density on excitatory neurons in the OFC. Rho-kinase inhibition also rescues goal-directed response strategies, linking neural remodeling with outcome-based decision making. Finally, DREADD-mediated OFC inhibition weakens new action-outcome memory.ConclusionsActivity-dependent and BDNF-dependent neuroplasticity within the OFC coordinate outcome-based decision making through interactions with the amygdala. These interactions break reward-seeking habits, a putative factor in multiple psychopathologies.
| S-EPMC4871791 | biostudies-literature
Project description:Adaptive behavior relies on the selection of relevant sensory information from both the external environment and internal memory representations. In understanding external selection, a classic distinction is made between voluntary (goal-directed) and involuntary (stimulus-driven) guidance of attention. We have developed a task-the anti-retrocue task-to separate and examine voluntary and involuntary guidance of attention to internal representations in visual working memory. We show that both voluntary and involuntary factors influence memory performance but do so in distinct ways. Moreover, by tracking gaze biases linked to attentional focusing in memory, we provide direct evidence for an involuntary "retro-capture" effect whereby external stimuli involuntarily trigger the selection of feature-matching internal representations. We show that stimulus-driven and goal-directed influences compete for selection in memory, and that the balance of this competition-as reflected in oculomotor signatures of internal attention-predicts the quality of ensuing memory-guided behavior. Thus, goal-directed and stimulus-driven factors together determine the fate not only of perception, but also of internal representations in working memory.
| S-EPMC7533705 | biostudies-literature
Project description:Prompt execution of planned motor action is essential for survival. The interactions between frontal cortical circuits and the basal ganglia are central to goal-oriented action selection and initiation.1-4 In rodents, the ventromedial thalamic nucleus (VM) is one of the critical nodes that conveys the output of the basal ganglia to the frontal cortical areas including the anterior lateral motor cortex (ALM).5-9 Recent studies showed the critical role of ALM and its interplay with the motor thalamus in preparing sensory-cued rewarded movements, specifically licking.10-12 Work in primates suggests that the basal ganglia output to the motor thalamus transmits an urgency or vigor signal,13-15 which leads to shortened reaction times and faster movement initiation. As yet, little is known about what signals are transmitted from the motor thalamus to the cortex during cued movements and how these signals contribute to movement initiation. In the present study, we employed a tactile-cued licking task in mice while monitoring reaction times of the initial lick. We found that inactivation of ALM delayed the initiation of cued licking. Two-photon Ca2+ imaging of VM axons revealed that the majority of the axon terminals in ALM were transiently active during licking. Their activity was predictive of the time of the first lick. Chemogenetic and optogenetic manipulation of VM axons in ALM indicated that VM inputs facilitate the initiation of cue-triggered and impulsive licking in trained mice. Our results suggest that VM thalamocortical inputs increase the probability and vigor of initiating planned motor responses.
| S-EPMC8478854 | biostudies-literature
Project description:In daily life, humans interact with multiple objects in complex environments. A large body of literature demonstrates that target selection is biased toward recently attended features, such that reaches are faster and trajectory curvature is reduced when target features (i.e., color) are repeated (priming of pop-out). In the real world, however, objects are comprised of several features-some of which may be more suitable for action than others. When fetching a mug from the cupboard, for example, attention not only has to be allocated to the object, but also the handle. To date, no study has investigated the impact of hierarchical feature organization on target selection for action. Here, we employed a color-oddity search task in which targets were Pac-men (i.e., a circle with a triangle cut out) oriented to be either consistent or inconsistent with the percept of a global Kanizsa triangle. We found that reaches were initiated faster when a task-irrelevant illusory figure was present independent of color repetition. Additionally, consistent with priming of pop-out, both reach planning and execution were facilitated when local target colors were repeated, regardless of whether a global figure was present. We also demonstrated that figures defined by illusory, but not real contours, afforded an early target selection benefit. In sum, these findings suggest that when local targets are perceptually grouped to form an illusory surface, attention quickly spreads across the global figure and facilitates the early stage of reach planning, but not execution. In contrast, local color priming is evident throughout goal-directed reaching.
| S-EPMC4979628 | biostudies-literature
Project description:Human reaching movements require complex muscle activations to produce the forces necessary to move the limb in a controlled manner. How gravity and the complex kinetic properties of the limb contribute to the generation of the muscle activation pattern by the central nervous system (CNS) is a long-standing and controversial question in neuroscience. To tackle this issue, muscle activity is often subdivided into static and phasic components. The former corresponds to posture maintenance and transitions between postures. The latter corresponds to active movement production and the compensation for the kinetic properties of the limb. In the present study, we improved the methodology for this subdivision of muscle activity into static and phasic components by relating them to joint torques. Ten healthy subjects pointed in virtual reality to visual targets arranged to create a standard center-out reaching task in three dimensions. Muscle activity and motion capture data were synchronously collected during the movements. The motion capture data were used to calculate postural and dynamic components of active muscle torques using a dynamic model of the arm with 5 degrees of freedom. Principal Component Analysis (PCA) was then applied to muscle activity and the torque components, separately, to reduce the dimensionality of the data. Muscle activity was also reconstructed from gravitational and dynamic torque components. Results show that the postural and dynamic components of muscle torque represent a significant amount of variance in muscle activity. This method could be used to define static and phasic components of muscle activity using muscle torques.
| S-EPMC5623018 | biostudies-other