Neural Correlates of Fixated Low- and High-level Scene Properties during Active Scene Viewing

Finding the meaning in meaning maps: Quantifying the roles of semantic and non-semantic scene information in guiding visual attention

In real-world vision, people prioritise the most informative scene regions via eye-movements. According to the cognitive guidance theory of visual attention, viewers allocate visual attention to those parts of the scene that are expected to be the most informative. The expected information of a scene region is coded in the semantic distribution of that scene. Meaning maps have been proposed to capture the spatial distribution of local scene semantics in order to test cognitive guidance theories of attention. Notwithstanding the success of meaning maps, the reason for their success has been contested. This has led to at least two possible explanations for the success of meaning maps in predicting visual attention. On the one hand, meaning maps might measure scene semantics. On the other hand, meaning maps might measure scene features, overlapping with, but distinct from, scene semantics. This study aims to disentangle these two sources of information by considering both conceptual information and non-semantic scene entropy simultaneously. We found that both semantic and non-semantic information is captured by meaning maps, but scene entropy accounted for more unique variance in the success of meaning maps than conceptual information. Additionally, some explained variance was unaccounted for by either source of information. Thus, although meaning maps may index some aspect of semantic information, their success seems to be better explained by non-semantic information. We conclude that meaning maps may not yet be a good tool to test cognitive guidance theories of attention in general, since they capture non-semantic aspects of local semantic density and only a small portion of conceptual information. Rather, we suggest that researchers should better define the exact aspect of cognitive guidance theories they wish to test and then use the tool that best captures that desired semantic information. As it stands, the semantic information contained in meaning maps seems too ambiguous to draw strong conclusions about how and when semantic information guides visual attention.

Neural Mechanisms Determining the Duration of Task-free, Self-paced Visual Perception

Humans spend hours each day spontaneously engaging with visual content, free from specific tasks and at their own pace. Currently, the brain mechanisms determining the duration of self-paced perceptual behavior remain largely unknown. Here, participants viewed naturalistic images under task-free settings and self-paced each image's viewing duration while undergoing EEG and pupillometry recordings. Across two independent data sets, we observed large inter- and intra-individual variability in viewing duration. However, beyond an image's presentation order and category, specific image content had no consistent effects on spontaneous viewing duration across participants. Overall, longer viewing durations were associated with sustained enhanced posterior positivity and anterior negativity in the ERPs. Individual-specific variations in the spontaneous viewing duration were consistently correlated with evoked EEG activity amplitudes and pupil size changes. By contrast, presentation order was selectively correlated with baseline alpha power and baseline pupil size. Critically, spontaneous viewing duration was strongly predicted by the temporal stability in neural activity patterns starting as early as 350 msec after image onset, suggesting that early neural stability is a key predictor for sustained perceptual engagement. Interestingly, neither bottom-up nor top-down predictions about image category influenced spontaneous viewing duration. Overall, these results suggest that individual-specific factors can influence perceptual processing at a surprisingly early time point and influence the multifaceted ebb and flow of spontaneous human perceptual behavior in naturalistic settings.

Aging changes the interactions between the oculomotor and memory systems

The use of multi-modal approaches, particularly in conjunction with multivariate analytic techniques, can enrich models of cognition, brain function, and how they change with age. Recently, multivariate approaches have been applied to the study of eye movements in a manner akin to that of neural activity (i.e., pattern similarity). Here, we review the literature regarding multi-modal and/or multivariate approaches, with specific reference to the use of eyetracking to characterize age-related changes in memory. By applying multi-modal and multivariate approaches to the study of aging, research has shown that aging is characterized by moment-to-moment alterations in the amount and pattern of visual exploration, and by extension, alterations in the activity and function of the hippocampus and broader medial temporal lobe (MTL). These methodological advances suggest that age-related declines in the integrity of the memory system has consequences for oculomotor behavior in the moment, in a reciprocal fashion. Age-related changes in hippocampal and MTL structure and function may lead to an increase in, and change in the patterns of, visual exploration in an effort to upregulate the encoding of information. However, such visual exploration patterns may be non-optimal and actually reduce the amount and/or type of incoming information that is bound into a lasting memory representation. This research indicates that age-related cognitive impairments are considerably broader in scope than previously realized.

Meaning maps detect the removal of local semantic scene content but deep saliency models do not

Meaning mapping uses human raters to estimate different semantic features in scenes, and has been a useful tool in demonstrating the important role semantics play in guiding attention. However, recent work has argued that meaning maps do not capture semantic content, but like deep learning models of scene attention, represent only semantically-neutral image features. In the present study, we directly tested this hypothesis using a diffeomorphic image transformation that is designed to remove the meaning of an image region while preserving its image features. Specifically, we tested whether meaning maps and three state-of-the-art deep learning models were sensitive to the loss of semantic content in this critical diffeomorphed scene region. The results were clear: meaning maps generated by human raters showed a large decrease in the diffeomorphed scene regions, while all three deep saliency models showed a moderate increase in the diffeomorphed scene regions. These results demonstrate that meaning maps reflect local semantic content in scenes while deep saliency models do something else. We conclude the meaning mapping approach is an effective tool for estimating semantic content in scenes.

Deep saliency models learn low-, mid-, and high-level features to predict scene attention

Deep saliency models represent the current state-of-the-art for predicting where humans look in real-world scenes. However, for deep saliency models to inform cognitive theories of attention, we need to know how deep saliency models prioritize different scene features to predict where people look. Here we open the black box of three prominent deep saliency models (MSI-Net, DeepGaze II, and SAM-ResNet) using an approach that models the association between attention, deep saliency model output, and low-, mid-, and high-level scene features. Specifically, we measured the association between each deep saliency model and low-level image saliency, mid-level contour symmetry and junctions, and high-level meaning by applying a mixed effects modeling approach to a large eye movement dataset. We found that all three deep saliency models were most strongly associated with high-level and low-level features, but exhibited qualitatively different feature weightings and interaction patterns. These findings suggest that prominent deep saliency models are primarily learning image features associated with high-level scene meaning and low-level image saliency and highlight the importance of moving beyond simply benchmarking performance.

Meaning maps capture the density of local semantic features in scenes: A reply to Pedziwiatr, Kümmerer, Wallis, Bethge & Teufel (2021)

Pedziwiatr, Kümmerer, Wallis, Bethge, & Teufel (2021) contend that Meaning Maps do not represent the spatial distribution of semantic features in scenes. We argue that Pesziwiatr et al. provide neither logical nor empirical support for that claim, and we conclude that Meaning Maps do what they were designed to do: represent the spatial distribution of meaning in scenes.

Spontaneous perception: a framework for task-free, self-paced perception

Flipping through social media feeds, viewing exhibitions in a museum, or walking through the botanical gardens, people consistently choose to engage with and disengage from visual content. Yet, in most laboratory settings, the visual stimuli, their presentation duration, and the task at hand are all controlled by the researcher. Such settings largely overlook the spontaneous nature of human visual experience, in which perception takes place independently from specific task constraints and its time course is determined by the observer as a self-governing agent. Currently, much remains unknown about how spontaneous perceptual experiences unfold in the brain. Are all perceptual categories extracted during spontaneous perception? Does spontaneous perception inherently involve volition? Is spontaneous perception segmented into discrete episodes? How do different neural networks interact over time during spontaneous perception? These questions are imperative to understand our conscious visual experience in daily life. In this article we propose a framework for spontaneous perception. We first define spontaneous perception as a task-free and self-paced experience. We propose that spontaneous perception is guided by four organizing principles that grant it temporal and spatial structures. These principles include coarse-to-fine processing, continuity and segmentation, agency and volition, and associative processing. We provide key suggestions illustrating how these principles may interact with one another in guiding the multifaceted experience of spontaneous perception. We point to testable predictions derived from this framework, including (but not limited to) the roles of the default-mode network and slow cortical potentials in underlying spontaneous perception. We conclude by suggesting several outstanding questions for future research, extending the relevance of this framework to consciousness and spontaneous brain activity. In conclusion, the spontaneous perception framework proposed herein integrates components in human perception and cognition, which have been traditionally studied in isolation, and opens the door to understand how visual perception unfolds in its most natural context.