Monitoring the attention of others is fundamental to social cognition. Most of the literature on the topic assumes that our social cognitive machinery is tuned specifically to the gaze direction of others as a proxy for attention. This standard assumption reduces attention to an externally visible parameter. Here we show that this assumption is wrong and a deeper, more meaningful representation is involved. We presented subjects with two cues about the attentional state of a face: direction of gaze and emotional expression. We tested whether people relied predominantly on one cue, the other, or both. If the traditional view is correct, then the gaze cue should dominate. Instead, people employed a variety of strategies, some relying on gaze, some on expression, and some on an integration of cues. We also assessed people’s social cognitive ability using two, independent, standard tests. If the traditional view is correct, then social cognitive ability, as assessed by the independent tests, should correlate with the degree to which people successfully use the gaze cue to judge the attention state of the face. Instead, social cognitive ability correlated best with the degree to which people successfully integrated the cues together, instead of with the use of any one specific cue. The results suggest a rethink of a fundamental component of social cognition: monitoring the attention of others involves constructing a deep model that is informed by a combination of cues. Attention is a rich process and monitoring the attention of others involves a similarly rich representation.
In the attention schema theory, the brain constructs a model of attention, the attention schema, to aid in the endogenous control of attention. Growing behavioral evidence appears to support this proposal. However, a central question remains: does a controller of attention actually benefit by having access to an attention schema? We constructed an artificial, deep Q-learning, neural network agent that was trained to control a simple form of visuospatial attention, tracking a stimulus with its attention spotlight in order to solve a catch task. The agent was tested with and without access to an attention schema. In both conditions, the agent received sufficient information such that it should, theoretically, be able to learn the task. We found that with an attention schema present, the agent learned to control its attention spotlight and learned the catch task to a high degree of performance. Once the agent learned, if the attention schema was disabled, the agent could no longer perform effectively. If the attention schema was removed before learning began, the agent was drastically impaired at learning. The results show how the presence of even a simple attention schema provides a profound benefit to a controller of attention. We interpret these results as supporting the central argument of AST: the brain evolved an attention schema because of its practical benefit in the endogenous control of attention.
Wilterson, Andrew; Nastase, Samuel; Bio, Branden; Guterstam, Arvid; Graziano, Michael
Abstract:
The attention schema theory (AST) posits a specific relationship between subjective awareness and attention, in which awareness is the control model that the brain uses to aid in the endogenous control of attention. We proposed that the right temporoparietal junction (TPJ) is involved in that interaction between awareness and attention. In previous experiments, we developed a behavioral paradigm in human subjects to manipulate awareness and attention. The paradigm involved a visual cue that could be used to guide a shift of attention to a target stimulus. In task 1, subjects were aware of the visual cue, and their endogenous control mechanism was able to use the cue to help control attention. In task 2, subjects were unaware of the visual cue, and their endogenous control mechanism was no longer able to use it to control attention, even though the cue still had a measurable effect on other aspects of behavior. Here we tested the two tasks while scanning brain activity in human volunteers. We predicted that the right TPJ would be active in relation to the cue in task 1, but not in task 2. This prediction was confirmed. The right TPJ was active in relation to the cue in task 1; it was not measurably active in task 2; the difference was significant. In our interpretation, the right TPJ is involved in a complex interaction in which awareness aids in the control of attention.
Pacheco, Diego A; Thiberge, Stephan; Pnevmatikakis, Eftychios; Murthy, Mala
Abstract:
Sensory pathways are typically studied starting at receptor neurons and following postsynaptic neurons into the brain. However, this leads to a bias in analysis of activity towards the earliest layers of processing. Here, we present new methods for volumetric neural imaging with precise across-brain registration, to characterize auditory activity throughout the entire central brain of Drosophila and make comparisons across trials, individuals, and sexes. We discover that auditory activity is present in most central brain regions and in neurons responsive to other modalities. Auditory responses are temporally diverse, but the majority of activity is tuned to courtship song features. Auditory responses are stereotyped across trials and animals in early mechanosensory regions, becoming more variable at higher layers of the putative pathway, and this variability is largely independent of spontaneous movements. This study highlights the power of using an unbiased, brain-wide approach for mapping the functional organization of sensory activity.
Extrapolation -- the ability to make inferences that go beyond the scope of one's experiences -- is a hallmark of human intelligence. By contrast, the generalization exhibited by contemporary neural network algorithms is largely limited to interpolation between data points in their training corpora. In this paper, we consider the challenge of learning representations that support extrapolation. We introduce a novel visual analogy benchmark that allows the graded evaluation of extrapolation as a function of distance from the convex domain defined by the training data. We also introduce a simple technique, context normalization, that encourages representations that emphasize the relations between objects. We find that this technique enables a significant improvement in the ability to extrapolate, considerably outperforming a number of competitive techniques.
Antony, James W.; Cheng, Larry Y.; Brooks, Paula P.; Paller, Ken A.; Norman, Kenneth A.
Abstract:
Competition between memories can cause weakening of those memories. Here we investigated memory competition during sleep in human participants by presenting auditory cues that had been linked to two distinct picture-location pairs during wake. We manipulated competition during learning by requiring participants to rehearse picture-location pairs associated with the same sound either competitively (choosing to rehearse one over the other, leading to greater competition) or separately; we hypothesized that greater competition during learning would lead to greater competition when memories were cued during sleep. With separate-pair learning, we found that cueing benefited spatial retention. With competitive-pair learning, no benefit of cueing was observed on retention, but cueing impaired retention of well-learned pairs (where we expected strong competition). During sleep, post-cue beta power (16–30 Hz) indexed competition and predicted forgetting, whereas sigma power (11–16 Hz) predicted subsequent retention. Taken together, these findings show that competition between memories during learning can modulate how they are consolidated during sleep.
Bejjanki, Vikranth R.; da Silveira, Rava Azeredo; Cohen, Jonathan D.; Turk-Browne, Nicholas B.
Abstract:
Multivariate decoding methods, such as multivoxel pattern analysis (MVPA), are highly effective at extracting information from brain imaging data. Yet, the precise nature of the information that MVPA draws upon remains controversial. Most current theories emphasize the enhanced sensitivity imparted by aggregating across voxels that have mixed and weak selectivity. However, beyond the selectivity of individual voxels, neural variability is correlated across voxels, and such noise correlations may contribute importantly to accurate decoding. Indeed, a recent computational theory proposed that noise correlations enhance multivariate decoding from heterogeneous neural populations. Here we extend this theory from the scale of neurons to functional magnetic resonance imaging (fMRI) and show that noise correlations between heterogeneous populations of voxels (i.e., voxels selective for different stimulus variables) contribute to the success of MVPA. Specifically, decoding performance is enhanced when voxels with high vs. low noise correlations (measured during rest or in the background of the task) are selected during classifier training. Conversely, voxels that are strongly selective for one class in a GLM or that receive high classification weights in MVPA tend to exhibit high noise correlations with voxels selective for the other class being discriminated against. Furthermore, we use simulations to show that this is a general property of fMRI data and that selectivity and noise correlations can have distinguishable influences on decoding. Taken together, our findings demonstrate that if there is signal in the data, the resulting above-chance classification accuracy is modulated by the magnitude of noise correlations.
Recent advances in experimental techniques have allowed the simultaneous recordings of
populations of hundreds of neurons, fostering a debate about the nature of the collective
structure of population neural activity. Much of this debate has focused on the
empirical findings of a phase transition in the parameter space of maximum entropy
models describing the measured neural probability distributions, interpreting this phase
transition to indicate a critical tuning of the neural code. Here, we instead focus on the
possibility that this is a first-order phase transition which provides evidence that the
real neural population is in a `structured', collective state. We show that this collective
state is robust to changes in stimulus ensemble and adaptive state. We find that the
pattern of pairwise correlations between neurons has a strength that is well within the
strongly correlated regime and does not require fine tuning, suggesting that this state is
generic for populations of 100+ neurons. We find a clear correspondence between the
emergence of a phase transition, and the emergence of attractor-like structure in the
inferred energy landscape. A collective state in the neural population, in which neural
activity patterns naturally form clusters, provides a consistent interpretation for our
results.
Cara L. Buck; Jonathan D. Cohen; Field, Brent; Daniel Kahneman; Samuel M. McClure; Leigh E. Nystrom
Abstract:
Studies of subjective well-being have conventionally relied upon self-report, which directs subjects’ attention to their emotional experiences. This method presumes that attention itself does not influence emotional processes, which could bias sampling. We tested whether attention influences experienced utility (the moment-by-moment experience of pleasure) by using functional magnetic resonance imaging (fMRI) to measure the activity of brain systems thought to represent hedonic value while manipulating attentional load. Subjects received appetitive or aversive solutions orally while alternatively executing a low or high attentional load task. Brain regions associated with hedonic processing, including the ventral striatum, showed a response to both juice and quinine. This response decreased during the high-load task relative to the low-load task. Thus, attentional allocation may influence experienced utility by modulating (either directly or indirectly) the activity of brain mechanisms thought to represent hedonic value.