Precedence of the eye region in neural processing of faces

Individual differences in face salience and rapid face saccades

Humans saccade to faces in their periphery faster than to other types of objects. Previous research has highlighted the potential importance of the upper face region in this phenomenon, but it remains unclear whether this is driven by the eye region. Similarly, it remains unclear whether such rapid saccades are exclusive to faces or generalize to other semantically salient stimuli. Furthermore, it is unknown whether individuals differ in their face-specific saccadic reaction times and, if so, whether such differences could be linked to differences in face fixations during free viewing. To explore these open questions, we invited 77 participants to perform a saccadic choice task in which we contrasted faces as well as other salient objects, particularly isolated face features and text, with cars. Additionally, participants freely viewed 700 images of complex natural scenes in a separate session, which allowed us to determine the individual proportion of first fixations falling on faces. For the saccadic choice task, we found advantages for all categories of interest over cars. However, this effect was most pronounced for images of full faces. Full faces also elicited faster saccades compared with eyes, showing that isolated eye regions are not sufficient to elicit face-like responses. Additionally, we found consistent individual differences in saccadic reaction times toward faces that weakly correlated with face salience during free viewing. Our results suggest a link between semantic salience and rapid detection, but underscore the unique status of faces. Further research is needed to resolve the mechanisms underlying rapid face saccades.

A preference to look closer to the eyes is associated with a position-invariant face neural code

When looking at faces, humans invariably move their eyes to a consistent preferred first fixation location on the face. While most people have the preferred fixation location just below the eyes, a minority have it between the nose-tip and mouth. Not much is known about whether these long-term differences in the preferred fixation location are associated with distinct neural representations of faces. To study this, we used a gaze-contingent face adaptation aftereffect paradigm to test in two groups of observers, one with their mean preferred fixation location closer to the eyes (upper lookers) and the other closer to the mouth (lower lookers). In this task, participants were required to maintain their gaze at either their own group's mean preferred fixation location or that of the other group during adaptation and testing. The two possible fixation locations were 3.6° apart on the face. We measured the face adaptation aftereffects when the adaptation and testing happened while participants maintained fixation at either the same or different locations on the face. Both groups showed equally strong adaptation effects when the adaptation and testing happened at the same fixation location. Crucially, only the upper lookers showed a partial transfer of the FAE across the two fixation locations, when adaptation occurred at the eyes. Lower lookers showed no spatial transfer of the FAE irrespective of the adaptation position. Given the classic finding that neural tuning is increasingly position invariant as one moves higher in the visual hierarchy, this result suggests that differences in the preferred fixation location are associated with distinct neural representations of faces.

Emergence of brain-like mirror-symmetric viewpoint tuning in convolutional neural networks

Primates can recognize objects despite 3D geometric variations such as in-depth rotations. The computational mechanisms that give rise to such invariances are yet to be fully understood. A curious case of partial invariance occurs in the macaque face-patch AL and in fully connected layers of deep convolutional networks in which neurons respond similarly to mirror-symmetric views (e.g. left and right profiles). Why does this tuning develop? Here, we propose a simple learning-driven explanation for mirror-symmetric viewpoint tuning. We show that mirror-symmetric viewpoint tuning for faces emerges in the fully connected layers of convolutional deep neural networks trained on object recognition tasks, even when the training dataset does not include faces. First, using 3D objects rendered from multiple views as test stimuli, we demonstrate that mirror-symmetric viewpoint tuning in convolutional neural network models is not unique to faces: it emerges for multiple object categories with bilateral symmetry. Second, we show why this invariance emerges in the models. Learning to discriminate among bilaterally symmetric object categories induces reflection-equivariant intermediate representations. AL-like mirror-symmetric tuning is achieved when such equivariant responses are spatially pooled by downstream units with sufficiently large receptive fields. These results explain how mirror-symmetric viewpoint tuning can emerge in neural networks, providing a theory of how they might emerge in the primate brain. Our theory predicts that mirror-symmetric viewpoint tuning can emerge as a consequence of exposure to bilaterally symmetric objects beyond the category of faces, and that it can generalize beyond previously experienced object categories.

Inactivation of face-selective neurons alters eye movements when free viewing faces

During free viewing, faces attract gaze and induce specific fixation patterns corresponding to the facial features. This suggests that neurons encoding the facial features are in the causal chain that steers the eyes. However, there is no physiological evidence to support a mechanistic link between face-encoding neurons in high-level visual areas and the oculomotor system. In this study, we targeted the middle face patches of the inferior temporal (IT) cortex in two macaque monkeys using an functional magnetic resonance imaging (fMRI) localizer. We then utilized muscimol microinjection to unilaterally suppress IT neural activity inside and outside the face patches and recorded eye movements while the animals free viewing natural scenes. Inactivation of the face-selective neurons altered the pattern of eye movements on faces: The monkeys found faces in the scene but neglected the eye contralateral to the inactivation hemisphere. These findings reveal the causal contribution of the high-level visual cortex in eye movements.

Computational components of visual predictive coding circuitry

If a full visual percept can be said to be a 'hypothesis', so too can a neural 'prediction' - although the latter addresses one particular component of image content (such as 3-dimensional organisation, the interplay between lighting and surface colour, the future trajectory of moving objects, and so on). And, because processing is hierarchical, predictions generated at one level are conveyed in a backward direction to a lower level, seeking to predict, in fact, the neural activity at that prior stage of processing, and learning from errors signalled in the opposite direction. This is the essence of 'predictive coding', at once an algorithm for information processing and a theoretical basis for the nature of operations performed by the cerebral cortex. Neural models for the implementation of predictive coding invoke specific functional classes of neuron for generating, transmitting and receiving predictions, and for producing reciprocal error signals. Also a third general class, 'precision' neurons, tasked with regulating the magnitude of error signals contingent upon the confidence placed upon the prediction, i.e., the reliability and behavioural utility of the sensory data that it predicts. So, what is the ultimate source of a 'prediction'? The answer is multifactorial: knowledge of the current environmental context and the immediate past, allied to memory and lifetime experience of the way of the world, doubtless fine-tuned by evolutionary history too. There are, in consequence, numerous potential avenues for experimenters seeking to manipulate subjects' expectation, and examine the neural signals elicited by surprising, and less surprising visual stimuli. This review focuses upon the predictive physiology of mouse and monkey visual cortex, summarising and commenting on evidence to date, and placing it in the context of the broader field. It is concluded that predictive coding has a firm grounding in basic neuroscience and that, unsurprisingly, there remains much to learn.

Rapid, concerted switching of the neural code in inferotemporal cortex

A fundamental paradigm in neuroscience is the concept of neural coding through tuning functions 1 . According to this idea, neurons encode stimuli through fixed mappings of stimulus features to firing rates. Here, we report that the tuning of visual neurons can rapidly and coherently change across a population to attend to a whole and its parts. We set out to investigate a longstanding debate concerning whether inferotemporal (IT) cortex uses a specialized code for representing specific types of objects or whether it uses a general code that applies to any object. We found that face cells in macaque IT cortex initially adopted a general code optimized for face detection. But following a rapid, concerted population event lasting < 20 ms, the neural code transformed into a face-specific one with two striking properties: (i) response gradients to principal detection-related dimensions reversed direction, and (ii) new tuning developed to multiple higher feature space dimensions supporting fine face discrimination. These dynamics were face specific and did not occur in response to objects. Overall, these results show that, for faces, face cells shift from detection to discrimination by switching from an object-general code to a face-specific code. More broadly, our results suggest a novel mechanism for neural representation: concerted, stimulus-dependent switching of the neural code used by a cortical area.

When the whole is only the parts: non-holistic object parts predominate face-cell responses to illusory faces

Humans are inclined to perceive faces in everyday objects with a face-like configuration. This illusion, known as face pareidolia, is often attributed to a specialized network of 'face cells' in primates. We found that face cells in macaque inferotemporal cortex responded selectively to pareidolia images, but this selectivity did not require a holistic, face-like configuration, nor did it encode human faceness ratings. Instead, it was driven mostly by isolated object parts that are perceived as eyes only within a face-like context. These object parts lack usual characteristics of primate eyes, pointing to the role of lower-level features. Our results suggest that face-cell responses are dominated by local, generic features, unlike primate visual perception, which requires holistic information. These findings caution against interpreting neural activity through the lens of human perception. Doing so could impose human perceptual biases, like seeing faces where none exist, onto our understanding of neural activity.

Single neuron responses underlying face recognition in the human midfusiform face-selective cortex

Faces are critical for social interactions and their recognition constitutes one of the most important and challenging functions of the human brain. While neurons responding selectively to faces have been recorded for decades in the monkey brain, face-selective neural activations have been reported with neuroimaging primarily in the human midfusiform gyrus. Yet, the cellular mechanisms producing selective responses to faces in this hominoid neuroanatomical structure remain unknown. Here we report single neuron recordings performed in 5 human subjects (1 male, 4 females) implanted with intracerebral microelectrodes in the face-selective midfusiform gyrus, while they viewed pictures of familiar and unknown faces and places. We observed similar responses to faces and places at the single cell level, but a significantly higher number of neurons responding to faces, thus offering a mechanistic account for the face-selective activations observed in this region. Although individual neurons did not respond preferentially to familiar faces, a population level analysis could consistently determine whether or not the faces (but not the places) were familiar, only about 50 ms after the initial recognition of the stimuli as faces. These results provide insights into the neural mechanisms of face processing in the human brain.

Short-latency preference for faces in the primate superior colliculus

Face processing is fundamental to primates and has been extensively studied in higher-order visual cortex. Here we report that visual neurons in the midbrain superior colliculus (SC) display a preference for faces, that the preference emerges within 50ms of stimulus onset - well before "face patches" in visual cortex - and that this activity can distinguish faces from other visual objects with accuracies of ~80%. This short-latency preference in SC depends on signals routed through early visual cortex, because inactivating the lateral geniculate nucleus, the key relay from retina to cortex, virtually eliminates visual responses in SC, including face-related activity. These results reveal an unexpected circuit in the primate visual system for rapidly detecting faces in the periphery, complementing the higher-order areas needed for recognizing individual faces.

The neural code for "face cells" is not face-specific

Face cells are neurons that respond more to faces than to non-face objects. They are found in clusters in the inferotemporal cortex, thought to process faces specifically, and, hence, studied using faces almost exclusively. Analyzing neural responses in and around macaque face patches to hundreds of objects, we found graded response profiles for non-face objects that predicted the degree of face selectivity and provided information on face-cell tuning beyond that from actual faces. This relationship between non-face and face responses was not predicted by color and simple shape properties but by information encoded in deep neural networks trained on general objects rather than face classification. These findings contradict the long-standing assumption that face versus non-face selectivity emerges from face-specific features and challenge the practice of focusing on only the most effective stimulus. They provide evidence instead that category-selective neurons are best understood by their tuning directions in a domain-general object space.

Faces in scenes attract rapid saccades

During natural vision, the human visual system has to process upcoming eye movements in parallel to currently fixated stimuli. Saccades targeting isolated faces are known to have lower latency and higher velocity, but it is unclear how this generalizes to the natural cycle of saccades and fixations during free-viewing of complex scenes. To which degree can the visual system process high-level features of extrafoveal stimuli when they are embedded in visual clutter and compete with concurrent foveal input? Here, we investigated how free-viewing dynamics vary as a function of an upcoming fixation target while controlling for various low-level factors. We found strong evidence that face- versus inanimate object-directed saccades are preceded by shorter fixations and have higher peak velocity. Interestingly, the boundary conditions for these two effects are dissociated. The effect on fixation duration was limited to face saccades, which were small and followed the trajectory of the preceding one, early in a trial. This is reminiscent of a recently proposed model of perisaccadic retinotopic shifts of attention. The effect on saccadic velocity, however, extended to very large saccades and increased with trial duration. These findings suggest that multiple, independent mechanisms interact to process high-level features of extrafoveal targets and modulate the dynamics of natural vision.

Inactivation of face selective neurons alters eye movements when free viewing faces

During free viewing, faces attract gaze and induce specific fixation patterns corresponding to the facial features. This suggests that neurons encoding the facial features are in the causal chain that steers the eyes. However, there is no physiological evidence to support a mechanistic link between face encoding neurons in high-level visual areas and the oculomotor system. In this study, we targeted the middle face patches of inferior temporal (IT) cortex in two macaque monkeys using an fMRI localizer. We then utilized muscimol microinjection to unilaterally suppress IT neural activity inside and outside the face patches and recorded eye movements while the animals free viewing natural scenes. Inactivation of the face selective neurons altered the pattern of eye movements on faces: the monkeys found faces in the scene but neglected the eye contralateral to the inactivation hemisphere. These findings reveal the causal contribution of the high-level visual cortex in eye movements.

Significance

It has been shown, for more than half a century, that eye movements follow distinctive patterns when free viewing faces. This suggests causal involvement of the face-encoding visual neurons in the eye movements. However, the literature is scant of evidence for this possibility and has focused mostly on the link between low-level image saliency and eye movements. Here, for the first time, we bring causal evidence showing how face-selective neurons in inferior temporal cortex inform and steer eye movements when free viewing faces.

Emergence of brain-like mirror-symmetric viewpoint tuning in convolutional neural networks

Primates can recognize objects despite 3D geometric variations such as in-depth rotations. The computational mechanisms that give rise to such invariances are yet to be fully understood. A curious case of partial invariance occurs in the macaque face-patch AL and in fully connected layers of deep convolutional networks in which neurons respond similarly to mirror-symmetric views (e.g., left and right profiles). Why does this tuning develop? Here, we propose a simple learning-driven explanation for mirror-symmetric viewpoint tuning. We show that mirror-symmetric viewpoint tuning for faces emerges in the fully connected layers of convolutional deep neural networks trained on object recognition tasks, even when the training dataset does not include faces. First, using 3D objects rendered from multiple views as test stimuli, we demonstrate that mirror-symmetric viewpoint tuning in convolutional neural network models is not unique to faces: it emerges for multiple object categories with bilateral symmetry. Second, we show why this invariance emerges in the models. Learning to discriminate among bilaterally symmetric object categories induces reflection-equivariant intermediate representations. AL-like mirror-symmetric tuning is achieved when such equivariant responses are spatially pooled by downstream units with sufficiently large receptive fields. These results explain how mirror-symmetric viewpoint tuning can emerge in neural networks, providing a theory of how they might emerge in the primate brain. Our theory predicts that mirror-symmetric viewpoint tuning can emerge as a consequence of exposure to bilaterally symmetric objects beyond the category of faces, and that it can generalize beyond previously experienced object categories.

Representational geometry of incomplete faces in macaque face patches

The neural code of faces has been intensively studied in the macaque face patch system. Although the majority of previous studies used complete faces as stimuli, faces are often seen partially in daily life. Here, we investigated how face-selective cells represent two types of incomplete faces: face fragments and occluded faces, with the location of the fragment/occluder and the facial features systematically varied. Contrary to popular belief, we found that the preferred face regions identified with two stimulus types are dissociated in many face cells. This dissociation can be explained by the nonlinear integration of information from different face parts and is closely related to a curved representation of face completeness in the state space, which allows a clear discrimination between different stimulus types. Furthermore, identity-related facial features are represented in a subspace orthogonal to the nonlinear dimension of face completeness, supporting a condition-general code of facial identity.

Towards an optimization of functional localizers in non-human primate neuroimaging with (fMRI) frequency-tagging

Non-human primate (NHP) neuroimaging can provide essential insights into the neural basis of human cognitive functions. While functional magnetic resonance imaging (fMRI) localizers can play an essential role in reaching this objective (Russ et al., 2021), they often differ substantially across species in terms of paradigms, measured signals, and data analysis, biasing the comparisons. Here we introduce a functional frequency-tagging face localizer for NHP imaging, successfully developed in humans and outperforming standard face localizers (Gao et al., 2018). FMRI recordings were performed in two awake macaques. Within a rapid 6 Hz stream of natural non-face objects images, human or monkey face stimuli were presented in bursts every 9 s. We also included control conditions with phase-scrambled versions of all images. As in humans, face-selective activity was objectively identified and quantified at the peak of the face-stimulation frequency (0.111 Hz) and its second harmonic (0.222 Hz) in the Fourier domain. Focal activations with a high signal-to-noise ratio were observed in regions previously described as face-selective, mainly in the STS (clusters PL, ML, MF; also, AL, AF), both for human and monkey faces. Robust face-selective activations were also found in the prefrontal cortex of one monkey (PVL and PO clusters). Face-selective neural activity was highly reliable and excluded all contributions from low-level visual cues contained in the amplitude spectrum of the stimuli. These observations indicate that fMRI frequency-tagging provides a highly valuable approach to objectively compare human and monkey visual recognition systems within the same framework.

Progressive neuronal plasticity in primate visual cortex during stimulus familiarization

The primate brain is equipped to learn and remember newly encountered visual stimuli such as faces and objects. In the macaque inferior temporal (IT) cortex, neurons mark the familiarity of a visual stimulus through response modification, often involving a decrease in spiking rate. Here, we investigate the emergence of this neural plasticity by longitudinally tracking IT neurons during several weeks of familiarization with face images. We found that most neurons in the anterior medial (AM) face patch exhibited a gradual decline in their late-phase visual responses to multiple stimuli. Individual neurons varied from days to weeks in their rates of plasticity, with time constants determined by the number of days of exposure rather than the cumulative number of presentations. We postulate that the sequential recruitment of neurons with experience-modified responses may provide an internal and graded measure of familiarity strength, which is a key mnemonic component of visual recognition.

Deep Neural Networks and Visuo-Semantic Models Explain Complementary Components of Human Ventral-Stream Representational Dynamics

Deep neural networks (DNNs) are promising models of the cortical computations supporting human object recognition. However, despite their ability to explain a significant portion of variance in neural data, the agreement between models and brain representational dynamics is far from perfect. We address this issue by asking which representational features are currently unaccounted for in neural time series data, estimated for multiple areas of the ventral stream via source-reconstructed magnetoencephalography data acquired in human participants (nine females, six males) during object viewing. We focus on the ability of visuo-semantic models, consisting of human-generated labels of object features and categories, to explain variance beyond the explanatory power of DNNs alone. We report a gradual reversal in the relative importance of DNN versus visuo-semantic features as ventral-stream object representations unfold over space and time. Although lower-level visual areas are better explained by DNN features starting early in time (at 66 ms after stimulus onset), higher-level cortical dynamics are best accounted for by visuo-semantic features starting later in time (at 146 ms after stimulus onset). Among the visuo-semantic features, object parts and basic categories drive the advantage over DNNs. These results show that a significant component of the variance unexplained by DNNs in higher-level cortical dynamics is structured and can be explained by readily nameable aspects of the objects. We conclude that current DNNs fail to fully capture dynamic representations in higher-level human visual cortex and suggest a path toward more accurate models of ventral-stream computations.SIGNIFICANCE STATEMENT When we view objects such as faces and cars in our visual environment, their neural representations dynamically unfold over time at a millisecond scale. These dynamics reflect the cortical computations that support fast and robust object recognition. DNNs have emerged as a promising framework for modeling these computations but cannot yet fully account for the neural dynamics. Using magnetoencephalography data acquired in human observers during object viewing, we show that readily nameable aspects of objects, such as 'eye', 'wheel', and 'face', can account for variance in the neural dynamics over and above DNNs. These findings suggest that DNNs and humans may in part rely on different object features for visual recognition and provide guidelines for model improvement.

Intracerebral Electrophysiological Recordings to Understand the Neural Basis of Human Face Recognition

Understanding how the human brain recognizes faces is a primary scientific goal in cognitive neuroscience. Given the limitations of the monkey model of human face recognition, a key approach in this endeavor is the recording of electrophysiological activity with electrodes implanted inside the brain of human epileptic patients. However, this approach faces a number of challenges that must be overcome for meaningful scientific knowledge to emerge. Here we synthesize a 10 year research program combining the recording of intracerebral activity (StereoElectroEncephaloGraphy, SEEG) in the ventral occipito-temporal cortex (VOTC) of large samples of participants and fast periodic visual stimulation (FPVS), to objectively define, quantify, and characterize the neural basis of human face recognition. These large-scale studies reconcile the wide distribution of neural face recognition activity with its (right) hemispheric and regional specialization and extend face-selectivity to anterior regions of the VOTC, including the ventral anterior temporal lobe (VATL) typically affected by magnetic susceptibility artifacts in functional magnetic resonance imaging (fMRI). Clear spatial dissociations in category-selectivity between faces and other meaningful stimuli such as landmarks (houses, medial VOTC regions) or written words (left lateralized VOTC) are found, confirming and extending neuroimaging observations while supporting the validity of the clinical population tested to inform about normal brain function. The recognition of face identity - arguably the ultimate form of recognition for the human brain - beyond mere differences in physical features is essentially supported by selective populations of neurons in the right inferior occipital gyrus and the lateral portion of the middle and anterior fusiform gyrus. In addition, low-frequency and high-frequency broadband iEEG signals of face recognition appear to be largely concordant in the human association cortex. We conclude by outlining the challenges of this research program to understand the neural basis of human face recognition in the next 10 years.

Quick, eyes! Isolated upper face regions but not artificial features elicit rapid saccades

Human faces elicit faster saccades than objects or animals, resonating with the great importance of faces for our species. The underlying mechanisms are largely unclear. Here, we test two hypotheses based on previous findings. First, ultra-rapid saccades toward faces may not depend on the presence of the whole face, but the upper face region containing the eye region. Second, ultra-rapid saccades toward faces (and possibly face parts) may emerge from our extensive experience with this stimulus and thus extend to glasses and masks - artificial features frequently encountered as part of a face. To test these hypotheses, we asked 43 participants to complete a saccadic choice task, which contrasted images of whole, upper and lower faces, face masks, and glasses with car images. The resulting data confirmed ultra-rapid saccades for isolated upper face regions, but not for artificial facial features.

Local features drive identity responses in macaque anterior face patches

Humans and other primates recognize one another in part based on unique structural details of the face, including both local features and their spatial configuration within the head and body. Visual analysis of the face is supported by specialized regions of the primate cerebral cortex, which in macaques are commonly known as face patches. Here we ask whether the responses of neurons in anterior face patches, thought to encode face identity, are more strongly driven by local or holistic facial structure. We created stimuli consisting of recombinant photorealistic images of macaques, where we interchanged the eyes, mouth, head, and body between individuals. Unexpectedly, neurons in the anterior medial (AM) and anterior fundus (AF) face patches were predominantly tuned to local facial features, with minimal neural selectivity for feature combinations. These findings indicate that the high-level structural encoding of face identity rests upon populations of neurons specialized for local features.

Anterior face patches in the macaque have been assumed to represent face identity in a holistic manner. Here the authors show that the neural encoding of face identity in the anterior medial and anterior fundus face patches are instead driven principally by local features.

Brain-wide functional connectivity of face patch neurons during rest

In resting state functional magnetic resonance imaging (fMRI), areas showing coherent hemodynamic fluctuations across the brain are operationally defined to be functionally connected. However, it is unknown how the activity of single units residing within a voxel contributes to this network structure. Here we demonstrate a shared but restricted pattern of functional connectivity among neighboring neurons residing in functionally defined face patches. Unexpectedly, such neurons also exhibited a prominent inverse correlation with thalamic structures and brainstem neuromodulatory centers. Single unit maps differed from analogous maps obtained with local field potentials and seed-based fMRI. These findings suggest that during rest, individual cortical neurons have a restricted set of functional connections, which is governed in part by anatomical projections and in part by neuromodulation.

The brain is a highly organized, dynamic system whose network architecture is often assessed through resting functional magnetic resonance imaging (fMRI) functional connectivity. The functional interactions between brain areas, including those observed during rest, are assumed to stem from the collective influence of action potentials carried by long-range neural projections. However, the contribution of individual neurons to brain-wide functional connectivity has not been systematically assessed. Here we developed a method to concurrently measure and compare the spiking activity of local neurons with fMRI signals measured across the brain during rest. We recorded spontaneous activity from neural populations in cortical face patches in the macaque during fMRI scanning sessions. Individual cells exhibited prominent, bilateral coupling with fMRI fluctuations in a restricted set of cortical areas inside and outside the face patch network, partially matching the pattern of known anatomical projections. Within each face patch population, a subset of neurons was positively coupled with the face patch network and another was negatively coupled. The same cells showed inverse correlations with distinct subcortical structures, most notably the lateral geniculate nucleus and brainstem neuromodulatory centers. Corresponding connectivity maps derived from fMRI seeds and local field potentials differed from the single unit maps, particularly in subcortical areas. Together, the results demonstrate that the spiking fluctuations of neurons are selectively coupled with discrete brain regions, with the coupling governed in part by anatomical network connections and in part by indirect neuromodulatory pathways.

Clutter Substantially Reduces Selectivity for Peripheral Faces in the Macaque Brain

According to a prominent view in neuroscience, visual stimuli are coded by discrete cortical networks that respond preferentially to specific categories, such as faces or objects. However, it remains unclear how these category-selective networks respond when viewing conditions are cluttered, i.e., when there is more than one stimulus in the visual field. Here, we asked three questions: (1) Does clutter reduce the response and selectivity for faces as a function of retinal location? (2) Is the preferential response to faces uniform across the visual field? And (3) Does the ventral visual pathway encode information about the location of cluttered faces? We used fMRI to measure the response of the face-selective network in awake, fixating macaques (two female, five male). Across a series of four experiments, we manipulated the presence and absence of clutter, as well as the location of the faces relative to the fovea. We found that clutter reduces the response to peripheral faces. When presented in isolation, without clutter, the selectivity for faces is fairly uniform across the visual field, but, when clutter is present, there is a marked decrease in the selectivity for peripheral faces. We also found no evidence of a contralateral visual field bias when faces were presented in clutter. Nonetheless, multivariate analyses revealed that the location of cluttered faces could be decoded from the multivoxel response of the face-selective network. Collectively, these findings demonstrate that clutter blunts the selectivity of the face-selective network to peripheral faces, although information about their retinal location is retained.SIGNIFICANCE STATEMENT Numerous studies that have measured brain activity in macaques have found visual regions that respond preferentially to faces. Although these regions are thought to be essential for social behavior, their responses have typically been measured while faces were presented in isolation, a situation atypical of the real world. How do these regions respond when faces are presented with other stimuli? We report that, when clutter is present, the preferential response to foveated faces is spared but preferential response to peripheral faces is reduced. Our results indicate that the presence of clutter changes the response of the face-selective network.

Socially meaningful visual context either enhances or inhibits vocalisation processing in the macaque brain

Social interactions rely on the interpretation of semantic and emotional information, often from multiple sensory modalities. Nonhuman primates send and receive auditory and visual communicative signals. However, the neural mechanisms underlying the association of visual and auditory information based on their common social meaning are unknown. Using heart rate estimates and functional neuroimaging, we show that in the lateral and superior temporal sulcus of the macaque monkey, neural responses are enhanced in response to species-specific vocalisations paired with a matching visual context, or when vocalisations follow, in time, visual information, but inhibited when vocalisation are incongruent with the visual context. For example, responses to affiliative vocalisations are enhanced when paired with affiliative contexts but inhibited when paired with aggressive or escape contexts. Overall, we propose that the identified neural network represents social meaning irrespective of sensory modality.

Social interaction involves processing semantic and emotional information. Here the authors show that in the macaque monkey lateral and superior temporal sulcus, cortical activity is enhanced in response to species-specific vocalisations predicted by matching face or social visual stimuli but inhibited when vocalisations are incongruent with the predictive visual context.

Twenty years of investigation with the case of prosopagnosia PS to understand human face identity recognition. Part II: Neural basis

Patient PS sustained her dramatic brain injury in 1992, the same year as the first report of a neuroimaging study of human face recognition. The present paper complements the review on the functional nature of PS's prosopagnosia (part I), illustrating how her case study directly, i.e., through neuroimaging investigations of her brain structure and activity, but also indirectly, through neural studies performed on other clinical cases and neurotypical individuals, inspired and constrained neural models of human face recognition. In the dominant right hemisphere for face recognition in humans, PS's main lesion concerns (inputs to) the inferior occipital gyrus (IOG), in a region where face-selective activity is typically found in normal individuals ('Occipital Face Area', OFA). Her case study initially supported the criticality of this region for face identity recognition (FIR) and provided the impetus for transcranial magnetic stimulation (TMS), intracerebral electrical stimulation, and cortical surgery studies that have generally supported this view. Despite PS's right IOG lesion, typical face-selectivity is found anteriorly in the middle portion of the fusiform gyrus, a hominoid structure (termed the right 'Fusiform Face Area', FFA) that is widely considered to be the most important region for human face recognition. This finding led to the original proposal of direct anatomico-functional connections from early visual cortices to the FFA, bypassing the IOG/OFA (lulu), a hypothesis supported by further neuroimaging studies of PS, other neurological cases and neuro-typical individuals with original visual stimulation paradigms, data recordings and analyses. The proposal of a lack of sensitivity to face identity in PS's right FFA due to defective reentrant inputs from the IOG/FFA has also been supported by other cases, functional connectivity and cortical surgery studies. Overall, neural studies of, and based on, the case of prosopagnosia PS strongly question the hierarchical organization of the human neural face recognition system, supporting a more flexible and dynamic view of this key social brain function.

The cortical and subcortical correlates of face pareidolia in the macaque brain

Face detection is a foundational social skill for primates. This vital function is thought to be supported by specialized neural mechanisms; however, although several face-selective regions have been identified in both humans and nonhuman primates, there is no consensus about which region(s) are involved in face detection. Here, we used naturally occurring errors of face detection (i.e. objects with illusory facial features referred to as examples of 'face pareidolia') to identify regions of the macaque brain implicated in face detection. Using whole-brain functional magnetic resonance imaging to test awake rhesus macaques, we discovered that a subset of face-selective patches in the inferior temporal cortex, on the lower lateral edge of the superior temporal sulcus, and the amygdala respond more to objects with illusory facial features than matched non-face objects. Multivariate analyses of the data revealed differences in the representation of illusory faces across the functionally defined regions of interest. These differences suggest that the cortical and subcortical face-selective regions contribute uniquely to the detection of facial features. We conclude that face detection is supported by a multiplexed system in the primate brain.

Face neurons encode nonsemantic features

Significance Face neurons, which fire more strongly in response to images of faces than to other objects, are a paradigmatic example of object selectivity in the visual cortex. We asked whether such neurons represent the semantic concept of faces or, rather, visual features that are present in faces but do not necessarily count as a face. We created synthetic stimuli that strongly activated face neurons and showed that these stimuli were perceived as clearly distinct from real faces. At the same time, these synthetic stimuli were slightly more often associated with faces than other objects were. These results suggest that so-called face neurons do not represent a semantic category but, rather, represent visual features that correlate with faces.

Parallel functional subnetworks embedded in the macaque face patch system

During normal vision, our eyes provide the brain with a continuous stream of useful information about the world. How visually specialized areas of the cortex, such as face-selective patches, operate under natural modes of behavior is poorly understood. Here we report that, during the free viewing of movies, cohorts of face-selective neurons in the macaque cortex fractionate into distributed and parallel subnetworks that carry distinct information. We classified neurons into functional groups on the basis of their movie-driven coupling with functional magnetic resonance imaging time courses across the brain. Neurons from each group were distributed across multiple face patches but intermixed locally with other groups at each recording site. These findings challenge prevailing views about functional segregation in the cortex and underscore the importance of naturalistic paradigms for cognitive neuroscience.

Face Recognition by Humans and Machines: Three Fundamental Advances from Deep Learning

Deep learning models currently achieve human levels of performance on real-world face recognition tasks. We review scientific progress in understanding human face processing using computational approaches based on deep learning. This review is organized around three fundamental advances. First, deep networks trained for face identification generate a representation that retains structured information about the face (e.g., identity, demographics, appearance, social traits, expression) and the input image (e.g., viewpoint, illumination). This forces us to rethink the universe of possible solutions to the problem of inverse optics in vision. Second, deep learning models indicate that high-level visual representations of faces cannot be understood in terms of interpretable features. This has implications for understanding neural tuning and population coding in the high-level visual cortex. Third, learning in deep networks is a multistep process that forces theoretical consideration of diverse categories of learning that can overlap, accumulate over time, and interact. Diverse learning types are needed to model the development of human face processing skills, cross-race effects, and familiarity with individual faces.

Joint encoding of facial identity, orientation, gaze, and expression in the middle dorsal face area

The last two decades have established that a network of face-selective areas in the temporal lobe of macaque monkeys supports the visual processing of faces. Each area within the network contains a large fraction of face-selective cells. And each area encodes facial identity and head orientation differently. A recent brain-imaging study discovered an area outside of this network selective for naturalistic facial motion, the middle dorsal (MD) face area. This finding offers the opportunity to determine whether coding principles revealed inside the core network would generalize to face areas outside the core network. We investigated the encoding of static faces and objects, facial identity, and head orientation, dimensions which had been studied in multiple areas of the core face-processing network before, as well as facial expressions and gaze. We found that MD populations form a face-selective cluster with a degree of selectivity comparable to that of areas in the core face-processing network. MD encodes facial identity robustly across changes in head orientation and expression, it encodes head orientation robustly against changes in identity and expression, and it encodes expression robustly across changes in identity and head orientation. These three dimensions are encoded in a separable manner. Furthermore, MD also encodes the direction of gaze in addition to head orientation. Thus, MD encodes both structural properties (identity) and changeable ones (expression and gaze) and thus provides information about another animal's direction of attention (head orientation and gaze). MD contains a heterogeneous population of cells that establish a multidimensional code for faces.

Holistic face recognition is an emergent phenomenon of spatial processing in face-selective regions

Spatial processing by receptive fields is a core property of the visual system. However, it is unknown how spatial processing in high-level regions contributes to recognition behavior. As face inversion is thought to disrupt typical holistic processing of information in faces, we mapped population receptive fields (pRFs) with upright and inverted faces in the human visual system. Here we show that in face-selective regions, but not primary visual cortex, pRFs and overall visual field coverage are smaller and shifted downward in response to face inversion. From these measurements, we successfully predict the relative behavioral detriment of face inversion at different positions in the visual field. This correspondence between neural measurements and behavior demonstrates how spatial processing in face-selective regions may enable holistic perception. These results not only show that spatial processing in high-level visual regions is dynamically used towards recognition, but also suggest a powerful approach for bridging neural computations by receptive fields to behavior.

Closing the gap between single-unit and neural population codes: Insights from deep learning in face recognition

Single-unit responses and population codes differ in the "read-out" information they provide about high-level visual representations. Diverging local and global read-outs can be difficult to reconcile with in vivo methods. To bridge this gap, we studied the relationship between single-unit and ensemble codes for identity, gender, and viewpoint, using a deep convolutional neural network (DCNN) trained for face recognition. Analogous to the primate visual system, DCNNs develop representations that generalize over image variation, while retaining subject (e.g., gender) and image (e.g., viewpoint) information. At the unit level, we measured the number of single units needed to predict attributes (identity, gender, viewpoint) and the predictive value of individual units for each attribute. Identification was remarkably accurate using random samples of only 3% of the network's output units, and all units had substantial identity-predicting power. Cross-unit responses were minimally correlated, indicating that single units code non-redundant identity cues. Gender and viewpoint classification required large-scale pooling of units-individual units had weak predictive power. At the ensemble level, principal component analysis of face representations showed that identity, gender, and viewpoint separated into high-dimensional subspaces, ordered by explained variance. Unit-based directions in the representational space were compared with the directions associated with the attributes. Identity, gender, and viewpoint contributed to all individual unit responses, undercutting a neural tuning analogy. Instead, single-unit responses carry superimposed, distributed codes for face identity, gender, and viewpoint. This undermines confidence in the interpretation of neural representations from unit response profiles for both DCNNs and, by analogy, high-level vision.

Impact of interracial contact on inferring mental states from facial expressions

Although decades of research have shown that intergroup contact critically impacts person perception and evaluation, little is known about how contact shapes the ability to infer others' mental states from facial cues (commonly referred to as mentalizing). In a pair of studies, we demonstrated that interracial contact and motivation to attend to faces jointly influence White perceivers' ability to infer mental states based on facial expressions displaying secondary emotions from both White targets alone (study 1) and White and Black targets (study 2; pre-registered). Consistent with previous work on the effect of motivation and interracial contact on other-race face memory, we found that motivation and interracial contact interacted to shape perceivers' accuracy at inferring mental states from secondary emotions. When motivated to attend to the task, high-contact White perceivers were more accurate at inferring both Black and White targets' mental states; unexpectedly, the opposite was true for low-contact perceivers. Importantly, the target race did not interact with interracial contact, suggesting that contact is associated with general changes in mentalizing irrespective of target race. These findings expand the theoretical understanding and implications of contact for fundamental social cognition.

Inferior Occipital Gyrus Is Organized along Common Gradients of Spatial and Face-Part Selectivity

The ventral visual stream of the human brain is subdivided into patches with categorical stimulus preferences, like faces or scenes. However, the functional organization within these areas is less clear. Here, we used functional magnetic resonance imaging and vertex-wise tuning models to independently probe spatial and face-part preferences in the inferior occipital gyrus (IOG) of healthy adult males and females. The majority of responses were well explained by Gaussian population tuning curves for both retinotopic location and the preferred relative position within a face. Parameter maps revealed a common gradient of spatial and face-part selectivity, with the width of tuning curves drastically increasing from posterior to anterior IOG. Tuning peaks clustered more idiosyncratically but were also correlated across maps of visual and face space. Preferences for the upper visual field went along with significantly increased coverage of the upper half of the face, matching recently discovered biases in human perception. Our findings reveal a broad range of neural face-part selectivity in IOG, ranging from narrow to "holistic." IOG is functionally organized along this gradient, which in turn is correlated with retinotopy.SIGNIFICANCE STATEMENT Brain imaging has revealed a lot about the large-scale organization of the human brain and visual system. For example, occipital cortex contains map-like representations of the visual field, while neurons in ventral areas cluster into patches with categorical preferences, like faces or scenes. Much less is known about the functional organization within these areas. Here, we focused on a well established face-preferring area-the inferior occipital gyrus (IOG). A novel neuroimaging paradigm allowed us to map the retinotopic and face-part tuning of many recording sites in IOG independently. We found a steep posterior-anterior gradient of decreasing face-part selectivity, which correlated with retinotopy. This suggests the functional role of ventral areas is not uniform and may follow retinotopic "protomaps."

Moving a Shape behind a Slit: Partial Shape Representations in Inferior Temporal Cortex

Current models of object recognition are based on spatial representations build from object features that are simultaneously present in the retinal image. However, one can recognize an object when it moves behind a static occlude, and only a small fragment of its shape is visible through a slit at a given moment in time. Such anorthoscopic perception requires spatiotemporal integration of the successively presented shape parts during slit-viewing. Human fMRI studies suggested that ventral visual stream areas represent whole shapes formed through temporal integration during anorthoscopic perception. To examine the time course of shape-selective responses during slit-viewing, we recorded the responses of single inferior temporal (IT) neurons of rhesus monkeys to moving shapes that were only partially visible through a static narrow slit. The IT neurons signaled shape identity by their response when that was cumulated across the duration of the shape presentation. Their shape preference during slit-viewing equaled that for static, whole-shape presentations. However, when analyzing their responses at a finer time scale, we showed that the IT neurons responded to particular shape fragments that were revealed by the slit. We found no evidence for temporal integration of slit-views that result in a whole-shape representation, even when the monkey was matching slit-views of a shape to static whole-shape presentations. These data suggest that, although the temporally integrated response of macaque IT neurons can signal shape identity in slit-viewing conditions, the spatiotemporal integration needed for the formation of a whole-shape percept occurs in other areas, perhaps downstream to IT.SIGNIFICANCE STATEMENT One recognizes an object when it moves behind a static occluder and only a small fragment of its shape is visible through a static slit at a given moment in time. Such anorthoscopic perception requires spatiotemporal integration of the successively presented partial shape parts. Human fMRI studies suggested that ventral visual stream areas represent shapes formed through temporal integration. We recorded the responses of inferior temporal (IT) cortical neurons of macaques during slit-viewing conditions. Although the temporally summated response of macaque IT neurons could signal shape identity under slit-viewing conditions, we found no evidence for a whole-shape representation using analyses at a finer time scale. Thus, the spatiotemporal integration needed for anorthoscopic perception does not occur within IT.

The N170 event-related potential reflects delayed neural response to faces when visual attention is directed to the eyes in youths with ASD

Atypical neural response to faces is thought to contribute to social deficits in autism spectrum disorder (ASD). Compared to typically developing (TD) controls, individuals with ASD exhibit delayed brain responses to upright faces at a face-sensitive event-related potential (ERP), the N170. Given observed differences in patterns of visual attention to faces, it is not known whether slowed neural processing may simply reflect atypical looking to faces. The present study manipulated visual attention to facial features to examine whether directed attention to the eyes normalizes N170 latency in ASD. ERPs were recorded in 30 children and adolescents with ASD as well as 26 TD children and adolescents. Results replicated prior findings of shorter N170 latency to the eye region of the face in TD individuals. In contrast, those with ASD did not demonstrate modulation of N170 latency by point of regard to the face. Group differences in latency were most pronounced when attention was directed to the eyes. Results suggest that well-replicated findings of N170 delays in ASD do not simply reflect atypical patterns of visual engagement with experimental stimuli. These findings add to a body of evidence indicating that N170 delays are a promising marker of atypical neural response to social information in ASD. LAY SUMMARY: This study looks at how children's and adolescents' brains respond when looking at different parts of a face. Typically developing children and adolescents processed eyes faster than other parts of the face, whereas this pattern was not seen in ASD. Children and adolescents with ASD processed eyes more slowly than typically developing children. These findings suggest that observed inefficiencies in face processing in ASD are not simply reflective of failure to attend to the eyes.

Parallel Processing of Facial Expression and Head Orientation in the Macaque Brain

When we move the features of our face, or turn our head, we communicate changes in our internal state to the people around us. How this information is encoded and used by an observer's brain is poorly understood. We investigated this issue using a functional MRI adaptation paradigm in awake male macaques. Among face-selective patches of the superior temporal sulcus (STS), we found a double dissociation of areas processing facial expression and those processing head orientation. The face-selective patches in the STS fundus were most sensitive to facial expression, as was the amygdala, whereas those on the lower, lateral edge of the sulcus were most sensitive to head orientation. The results of this study reveal a new dimension of functional organization, with face-selective patches segregating within the STS. The findings thus force a rethinking of the role of the face-processing system in representing subject-directed actions and supporting social cognition.SIGNIFICANCE STATEMENT When we are interacting with another person, we make inferences about their emotional state based on visual signals. For example, when a person's facial expression changes, we are given information about their feelings. While primates are thought to have specialized cortical mechanisms for analyzing the identity of faces, less is known about how these mechanisms unpack transient signals, like expression, that can change from one moment to the next. Here, using an fMRI adaptation paradigm, we demonstrate that while the identity of a face is held constant, there are separate mechanisms in the macaque brain for processing transient changes in the face's expression and orientation. These findings shed new light on the function of the face-processing system during social exchanges.

The neurons that mistook a hat for a face

Despite evidence that context promotes the visual recognition of objects, decades of research have led to the pervasive notion that the object processing pathway in primate cortex consists of multiple areas that each process the intrinsic features of a few particular categories (e.g. faces, bodies, hands, objects, and scenes). Here we report that such category-selective neurons do not in fact code individual categories in isolation but are also sensitive to object relationships that reflect statistical regularities of the experienced environment. We show by direct neuronal recording that face-selective neurons respond not just to an image of a face, but also to parts of an image where contextual cues—for example a body—indicate a face ought to be, even if what is there is not a face.

Real-world structure facilitates the rapid emergence of scene category information in visual brain signals

In everyday life, our visual surroundings are not arranged randomly but structured in predictable ways. Although previous studies have shown that the visual system is sensitive to such structural regularities, it remains unclear whether the presence of an intact structure in a scene also facilitates the cortical analysis of the scene's categorical content. To address this question, we conducted an EEG experiment during which participants viewed natural scene images that were either "intact" (with their quadrants arranged in typical positions) or "jumbled" (with their quadrants arranged into atypical positions). We then used multivariate pattern analysis to decode the scenes' category from the EEG signals (e.g., whether the participant had seen a church or a supermarket). The category of intact scenes could be decoded rapidly within the first 100 ms of visual processing. Critically, within 200 ms of processing, category decoding was more pronounced for the intact scenes compared with the jumbled scenes, suggesting that the presence of real-world structure facilitates the extraction of scene category information. No such effect was found when the scenes were presented upside down, indicating that the facilitation of neural category information is indeed linked to a scene's adherence to typical real-world structure rather than to differences in visual features between intact and jumbled scenes. Our results demonstrate that early stages of categorical analysis in the visual system exhibit tuning to the structure of the world that may facilitate the rapid extraction of behaviorally relevant information from rich natural environments.NEW & NOTEWORTHY Natural scenes are structured, with different types of information appearing in predictable locations. Here, we use EEG decoding to show that the visual brain uses this structure to efficiently analyze scene content. During early visual processing, the category of a scene (e.g., a church vs. a supermarket) could be more accurately decoded from EEG signals when the scene adhered to its typical spatial structure compared with when it did not.

Researchers Keep Rejecting Grandmother Cells after Running the Wrong Experiments: The Issue Is How Familiar Stimuli Are Identified

There is widespread agreement in neuroscience and psychology that the visual system identifies objects and faces based on a pattern of activation over many neurons, each neuron being involved in representing many different categories. The hypothesis that the visual system includes finely tuned neurons for specific objects or faces for the sake of identification, so-called "grandmother cells", is widely rejected. Here it is argued that the rejection of grandmother cells is premature. Grandmother cells constitute a hypothesis of how familiar visual categories are identified, but the primary evidence against this hypothesis comes from studies that have failed to observe neurons that selectively respond to unfamiliar stimuli. These findings are reviewed and it is shown that they are irrelevant. Neuroscientists need to better understand existing models of face and object identification that include grandmother cells and then compare the selectivity of these units with single neurons responding to stimuli that can be identified.

Evolving Images for Visual Neurons Using a Deep Generative Network Reveals Coding Principles and Neuronal Preferences

What specific features should visual neurons encode, given the infinity of real-world images and the limited number of neurons available to represent them? We investigated neuronal selectivity in monkey inferotemporal cortex via the vast hypothesis space of a generative deep neural network, avoiding assumptions about features or semantic categories. A genetic algorithm searched this space for stimuli that maximized neuronal firing. This led to the evolution of rich synthetic images of objects with complex combinations of shapes, colors, and textures, sometimes resembling animals or familiar people, other times revealing novel patterns that did not map to any clear semantic category. These results expand our conception of the dictionary of features encoded in the cortex, and the approach can potentially reveal the internal representations of any system whose input can be captured by a generative model.

The bottom-up and top-down processing of faces in the human occipitotemporal cortex

Although face processing has been studied extensively, the dynamics of how face-selective cortical areas are engaged remains unclear. Here, we uncovered the timing of activation in core face-selective regions using functional Magnetic Resonance Imaging and Magnetoencephalography in humans. Processing of normal faces started in the posterior occipital areas and then proceeded to anterior regions. This bottom-up processing sequence was also observed even when internal facial features were misarranged. However, processing of two-tone Mooney faces lacking explicit prototypical facial features engaged top-down projection from the right posterior fusiform face area to right occipital face area. Further, face-specific responses elicited by contextual cues alone emerged simultaneously in the right ventral face-selective regions, suggesting parallel contextual facilitation. Together, our findings chronicle the precise timing of bottom-up, top-down, as well as context-facilitated processing sequences in the occipital-temporal face network, highlighting the importance of the top-down operations especially when faced with incomplete or ambiguous input.

Evidence for an attentional priority map in inferotemporal cortex

A fundamental dogma in the cognitive neurosciences is that attention is controlled by parietal and prefrontal areas. Here, we show that an area in the temporal lobe exhibits the properties of a priority map coding the focus of attention. We show this through whole-brain functional magnetic resonance imaging, electrophysiological single-unit recordings, and causal electrical stimulation. This discovery changes our understanding of the organization of visual pathways and the functions of attention networks.

From incoming sensory information, our brains make selections according to current behavioral goals. This process, selective attention, is controlled by parietal and frontal areas. Here, we show that another brain area, posterior inferotemporal cortex (PITd), also exhibits the defining properties of attentional control. We discovered this area with functional magnetic resonance imaging (fMRI) during an attentive motion discrimination task. Single-cell recordings from PITd revealed strong attentional modulation across 3 attention tasks yet no tuning to task-relevant stimulus features, like motion direction or color. Instead, PITd neurons closely tracked the subject’s attention state and predicted upcoming errors of attentional selection. Furthermore, artificial electrical PITd stimulation controlled the location of attentional selection without altering feature discrimination. These are the defining properties of a feature-blind priority map encoding the locus of attention. Together, these results suggest area PITd, located strategically to gather information about object properties, as an attentional priority map.

Universal Mechanisms and the Development of the Face Network: What You See Is What You Get

Our assignment was to review the development of the face-processing network, an assignment that carries the presupposition that a face-specific developmental program exists. We hope to cast some doubt on this assumption and instead argue that the development of face processing is guided by the same ubiquitous rules that guide the development of cortex in general.

From eye to face: The impact of face outline, feature number, and feature saliency on the early neural response to faces

The LIFTED model of early face perception postulates that the face-sensitive N170 event-related potential may reflect underlying neural inhibition mechanisms which serve to regulate holistic and featural processing. It remains unclear, however, what specific factors impact these neural inhibition processes. Here, N170 peak responses were recorded whilst adults maintained fixation on a single eye using a gaze-contingent paradigm, and the presence/absence of a face outline, as well as the number and type of parafoveal features within the outline, were manipulated. N170 amplitudes and latencies were reduced when a single eye was fixated within a face outline compared to fixation on the same eye in isolation, demonstrating that the simple presence of a face outline is sufficient to elicit a shift towards a more face-like neural response. A monotonic decrease in the N170 amplitude and latency was observed with increasing numbers of parafoveal features, and the type of feature(s) present in parafovea further modulated this early face response. These results support the idea of neural inhibition exerted by parafoveal features onto the foveated feature as a function of the number, and possibly the nature, of parafoveal features. Specifically, the results suggest the use of a feature saliency framework (eyes > mouth > nose) at the neural level, such that the parafoveal eye may play a role in down-regulating the response to the other eye (in fovea) more so than the nose or the mouth. These results confirm the importance of parafoveal features and the face outline in the neural inhibition mechanism, and provide further support for a feature saliency mechanism guiding early face perception.

Neural correlates of sparse coding and dimensionality reduction

Supported by recent computational studies, there is increasing evidence that a wide range of neuronal responses can be understood as an emergent property of nonnegative sparse coding (NSC), an efficient population coding scheme based on dimensionality reduction and sparsity constraints. We review evidence that NSC might be employed by sensory areas to efficiently encode external stimulus spaces, by some associative areas to conjunctively represent multiple behaviorally relevant variables, and possibly by the basal ganglia to coordinate movement. In addition, NSC might provide a useful theoretical framework under which to understand the often complex and nonintuitive response properties of neurons in other brain areas. Although NSC might not apply to all brain areas (for example, motor or executive function areas) the success of NSC-based models, especially in sensory areas, warrants further investigation for neural correlates in other regions.

Object Vision in a Structured World

In natural vision, objects appear at typical locations, both with respect to visual space (e.g., an airplane in the upper part of a scene) and other objects (e.g., a lamp above a table). Recent studies have shown that object vision is strongly adapted to such positional regularities. In this review we synthesize these developments, highlighting that adaptations to positional regularities facilitate object detection and recognition, and sharpen the representations of objects in visual cortex. These effects are pervasive across various types of high-level content. We posit that adaptations to real-world structure collectively support optimal usage of limited cortical processing resources. Taking positional regularities into account will thus be essential for understanding efficient object vision in the real world.