Dynamic information processing states revealed through neurocognitive models of object semantics

Representational formats of human memory traces

Neural representations are internal brain states that constitute the brain's model of the external world or some of its features. In the presence of sensory input, a representation may reflect various properties of this input. When perceptual information is no longer available, the brain can still activate representations of previously experienced episodes due to the formation of memory traces. In this review, we aim at characterizing the nature of neural memory representations and how they can be assessed with cognitive neuroscience methods, mainly focusing on neuroimaging. We discuss how multivariate analysis techniques such as representational similarity analysis (RSA) and deep neural networks (DNNs) can be leveraged to gain insights into the structure of neural representations and their different representational formats. We provide several examples of recent studies which demonstrate that we are able to not only measure memory representations using RSA but are also able to investigate their multiple formats using DNNs. We demonstrate that in addition to slow generalization during consolidation, memory representations are subject to semantization already during short-term memory, by revealing a shift from visual to semantic format. In addition to perceptual and conceptual formats, we describe the impact of affective evaluations as an additional dimension of episodic memories. Overall, these studies illustrate how the analysis of neural representations may help us gain a deeper understanding of the nature of human memory.

No need to integrate action information during coarse semantic processing of man-made tools

Action representation of man-made tools consists of two subtypes: structural action representation concerning how to grasp an object, and functional action representation concerning the skilled use of an object. Compared to structural action representation, functional action representation plays the dominant role in fine-grained (i.e., basic level) object recognition. However, it remains unclear whether the two types of action representation are involved differently in the coarse semantic processing in which the object is recognized at a superordinate level (i.e., living/non-living). Here we conducted three experiments using the priming paradigm, in which video clips displaying structural and functional action hand gestures were used as prime stimuli and grayscale photos of man-made tools were used as target stimuli. Participants recognized the target objects at the basic level in Experiment 1 (i.e., naming task) and at the superordinate level in Experiments 2 and 3 (i.e., categorization task). We observed a significant priming effect for functional action prime-target pairs only in the naming task. In contrast, no priming effect was found in either the naming or the categorization task for the structural action prime-target pairs (Experiment 2), even when the categorization task was preceded by a preliminary action imitation of the prime gestures (Experiment 3). Our results suggest that only functional action information is retrieved during fine-grained object processing. In contrast, coarse semantic processing does not require the integration of either structural or functional action information.

Semantic cognition in healthy ageing: Neural signatures of representation and control mechanisms in naming typical and atypical objects

Effective use of conceptual knowledge engages semantic representation and control processes to access information in a goal-driven manner. Neuropsychological findings of patients presenting either degraded knowledge (e.g., semantic dementia) or disrupted control (e.g., semantic aphasia) converge with neuroimaging evidence from young adults, and delineate the neural segregation of representation and control mechanisms. However, there is still scarce research on the neurofunctional underpinnings of such mechanisms in healthy ageing. To address this, we conducted an fMRI study, wherein young and older adults performed a covert naming task of typical and atypical objects. Three main age-related differences were found. As shown by age group and typicality interactions, older adults exhibited overactivation during naming of atypical (e.g., avocado) relative to typical concepts in brain regions associated to semantic representation, including anterior and medial portions of left temporal lobe (respectively, ATL and MTG). This provides evidence for the reorganization of neural activity in these brain regions contingent to the enrichment of semantic repositories in older ages. The medial orbitofrontal gyrus was also overactivated, indicating that the processing of atypical concepts (relative to typical items) taxes additional control resources in the elderly. Increased activation in the inferior frontal gyrus (IFG) was observed in naming typical items (relative to atypical ones), but only for young adults. This suggests that naming typical items (e.g., strawberry) taxes more on control processes in younger ages, presumably due to the semantic competition set by other items that share multiple features with the target (e.g., raspberry, blackberry, cherry). Together, these results reveal the dynamic nature of semantic control interplaying with conceptual representations as people grow older, by indicating that distinct neural bases uphold semantic performance from young to older ages. These findings may be explained by neural compensation mechanisms coming into play to support neurocognitive changes in healthy ageing.

Unique contributions of perceptual and conceptual humanness to object representations in the human brain

The human brain is able to quickly and accurately identify objects in a dynamic visual world. Objects evoke different patterns of neural activity in the visual system, which reflect object category memberships. However, the underlying dimensions of object representations in the brain remain unclear. Recent research suggests that objects similarity to humans is one of the main dimensions used by the brain to organise objects, but the nature of the human-similarity features driving this organisation are still unknown. Here, we investigate the relative contributions of perceptual and conceptual features of humanness to the representational organisation of objects in the human visual system. We collected behavioural judgements of human-similarity of various objects, which were compared with time-resolved neuroimaging responses to the same objects. The behavioural judgement tasks targeted either perceptual or conceptual humanness features to determine their respective contribution to perceived human-similarity. Behavioural and neuroimaging data revealed significant and unique contributions of both perceptual and conceptual features of humanness, each explaining unique variance in neuroimaging data. Furthermore, our results showed distinct spatio-temporal dynamics in the processing of conceptual and perceptual humanness features, with later and more lateralised brain responses to conceptual features. This study highlights the critical importance of social requirements in information processing and organisation in the human brain.

Beyond core object recognition: Recurrent processes account for object recognition under occlusion

Core object recognition, the ability to rapidly recognize objects despite variations in their appearance, is largely solved through the feedforward processing of visual information. Deep neural networks are shown to achieve human-level performance in these tasks, and explain the primate brain representation. On the other hand, object recognition under more challenging conditions (i.e. beyond the core recognition problem) is less characterized. One such example is object recognition under occlusion. It is unclear to what extent feedforward and recurrent processes contribute in object recognition under occlusion. Furthermore, we do not know whether the conventional deep neural networks, such as AlexNet, which were shown to be successful in solving core object recognition, can perform similarly well in problems that go beyond the core recognition. Here, we characterize neural dynamics of object recognition under occlusion, using magnetoencephalography (MEG), while participants were presented with images of objects with various levels of occlusion. We provide evidence from multivariate analysis of MEG data, behavioral data, and computational modelling, demonstrating an essential role for recurrent processes in object recognition under occlusion. Furthermore, the computational model with local recurrent connections, used here, suggests a mechanistic explanation of how the human brain might be solving this problem.

In recent years, deep-learning-based computer vision algorithms have been able to achieve human-level performance in several object recognition tasks. This has also contributed in our understanding of how our brain may be solving these recognition tasks. However, object recognition under more challenging conditions, such as occlusion, is less characterized. Temporal dynamics of object recognition under occlusion is largely unknown in the human brain. Furthermore, we do not know if the previously successful deep-learning algorithms can similarly achieve human-level performance in these more challenging object recognition tasks. By linking brain data with behavior, and computational modeling, we characterized temporal dynamics of object recognition under occlusion, and proposed a computational mechanism that explains both behavioral and the neural data in humans. This provides a plausible mechanistic explanation for how our brain might be solving object recognition under more challenging conditions.

Oscillatory Dynamics of Perceptual to Conceptual Transformations in the Ventral Visual Pathway

Object recognition requires dynamic transformations of low-level visual inputs to complex semantic representations. Although this process depends on the ventral visual pathway, we lack an incremental account from low-level inputs to semantic representations and the mechanistic details of these dynamics. Here we combine computational models of vision with semantics and test the output of the incremental model against patterns of neural oscillations recorded with magnetoencephalography in humans. Representational similarity analysis showed visual information was represented in low-frequency activity throughout the ventral visual pathway, and semantic information was represented in theta activity. Furthermore, directed connectivity showed visual information travels through feedforward connections, whereas visual information is transformed into semantic representations through feedforward and feedback activity, centered on the anterior temporal lobe. Our research highlights that the complex transformations between visual and semantic information is driven by feedforward and recurrent dynamics resulting in object-specific semantics.

Integrated deep visual and semantic attractor neural networks predict fMRI pattern-information along the ventral object processing pathway

Recognising an object involves rapid visual processing and activation of semantic knowledge about the object, but how visual processing activates and interacts with semantic representations remains unclear. Cognitive neuroscience research has shown that while visual processing involves posterior regions along the ventral stream, object meaning involves more anterior regions, especially perirhinal cortex. Here we investigate visuo-semantic processing by combining a deep neural network model of vision with an attractor network model of semantics, such that visual information maps onto object meanings represented as activation patterns across features. In the combined model, concept activation is driven by visual input and co-occurrence of semantic features, consistent with neurocognitive accounts. We tested the model’s ability to explain fMRI data where participants named objects. Visual layers explained activation patterns in early visual cortex, whereas pattern-information in perirhinal cortex was best explained by later stages of the attractor network, when detailed semantic representations are activated. Posterior ventral temporal cortex was best explained by intermediate stages corresponding to initial semantic processing, when visual information has the greatest influence on the emerging semantic representation. These results provide proof of principle of how a mechanistic model of combined visuo-semantic processing can account for pattern-information in the ventral stream.

Convolutional neural network-based encoding and decoding of visual object recognition in space and time

Representations learned by deep convolutional neural networks (CNNs) for object recognition are a widely investigated model of the processing hierarchy in the human visual system. Using functional magnetic resonance imaging, CNN representations of visual stimuli have previously been shown to correspond to processing stages in the ventral and dorsal streams of the visual system. Whether this correspondence between models and brain signals also holds for activity acquired at high temporal resolution has been explored less exhaustively. Here, we addressed this question by combining CNN-based encoding models with magnetoencephalography (MEG). Human participants passively viewed 1,000 images of objects while MEG signals were acquired. We modelled their high temporal resolution source-reconstructed cortical activity with CNNs, and observed a feed-forward sweep across the visual hierarchy between 75 and 200 ms after stimulus onset. This spatiotemporal cascade was captured by the network layer representations, where the increasingly abstract stimulus representation in the hierarchical network model was reflected in different parts of the visual cortex, following the visual ventral stream. We further validated the accuracy of our encoding model by decoding stimulus identity in a left-out validation set of viewed objects, achieving state-of-the-art decoding accuracy.

Am I looking at a cat or a dog? Gaze in the semantic variant of primary progressive aphasia is subject to excessive taxonomic capture

Object naming impairments or anomias are the most frequent symptom in aphasia, and can be caused by a variety of underlying neurocognitive mechanisms. Anomia in neurodegenerative or primary progressive aphasias (PPA) often appears to be based on taxonomic blurring of word meaning: words such as "dog" and "cat" are still recognized generically as referring to animals, but are no longer conceptually differentiated from each other, leading to coordinate errors in word-object matching. This blurring is the hallmark symptom of the "semantic variant" of PPA, who invariably show focal atrophy in the left anterior temporal lobe. In this study we used eye tracking to characterize information processing online (in real time) as non-aphasic controls, semantic and non-semantic PPA participants completed a word-to-object matching task. All participants (including controls) showed taxonomic capture of gaze, spending more time viewing foils that were from the same category as the target compared to unrelated foils, but capture was more extreme in the semantic PPA group. The semantic group showed heightened capture even on trials where they ultimately pointed to the correct target, demonstrating the superiority of eye movements over traditional testing methods in detecting subtle processing impairments. Heightened capture was primarily driven by a tendency to direct gaze back and forth, repeatedly, between a set of related foils on each trial, a behavior almost never shown by controls or non-semantic participants. This suggests semantic PPA participants were accumulating and weighing evidence for a probabilistic rather than definitive mapping between the noun and several candidate objects. Neurodegeneration in PPA thus appears to distort lexical concepts prior to extinguishing them altogether, causing uncertainty in recognition and word-object matching.

Neural representation for object recognition in inferotemporal cortex

We suggest that population representation of objects in inferotemporal cortex lie on a continuum between a purely structural, parts-based description and a purely holistic description. The intrinsic dimensionality of object representation is estimated to be around 100, perhaps with lower dimensionalities for object representations more toward the holistic end of the spectrum. Cognitive knowledge in the form of semantic information and task information feed back to inferotemporal cortex from perirhinal and prefrontal cortex respectively, providing high-level multimodal-based expectations that assist in the interpretation of object stimuli. Integration of object information across eye movements may also contribute to object recognition through a process of active vision.

Understanding What We See: How We Derive Meaning From Vision

Recognising objects goes beyond vision, and requires models that incorporate different aspects of meaning. Most models focus on superordinate categories (e.g., animals, tools) which do not capture the richness of conceptual knowledge. We argue that object recognition must be seen as a dynamic process of transformation from low-level visual input through categorical organisation to specific conceptual representations. Cognitive models based on large normative datasets are well-suited to capture statistical regularities within and between concepts, providing both category structure and basic-level individuation. We highlight recent research showing how such models capture important properties of the ventral visual pathway. This research demonstrates that significant advances in understanding conceptual representations can be made by shifting the focus from studying superordinate categories to basic-level concepts.