Visual and visuomotor processing of hands and tools as a case study of cross talk between the dorsal and ventral streams

Naturalistic Object Representations Depend on Distance and Size Cues

Egocentric distance and real-world size are important cues for object perception and action. Nevertheless, most studies of human vision rely on two-dimensional pictorial stimuli that convey ambiguous distance and size information. Here, we use fMRI to test whether pictures are represented differently in the human brain from real, tangible objects that convey unambiguous distance and size cues. Participants directly viewed stimuli in two display formats (real objects and matched printed pictures of those objects) presented at different egocentric distances (near and far). We measured the effects of format and distance on fMRI response amplitudes and response patterns. We found that fMRI response amplitudes in the lateral occipital and posterior parietal cortices were stronger overall for real objects than for pictures. In these areas and many others, including regions involved in action guidance, responses to real objects were stronger for near vs. far stimuli, whereas distance had little effect on responses to pictures-suggesting that distance determines relevance to action for real objects, but not for pictures. Although stimulus distance especially influenced response patterns in dorsal areas that operate in the service of visually guided action, distance also modulated representations in ventral cortex, where object responses are thought to remain invariant across contextual changes. We observed object size representations for both stimulus formats in ventral cortex but predominantly only for real objects in dorsal cortex. Together, these results demonstrate that whether brain responses reflect physical object characteristics depends on whether the experimental stimuli convey unambiguous information about those characteristics.

Significance Statement

Classic frameworks of vision attribute perception of inherent object characteristics, such as size, to the ventral visual pathway, and processing of spatial characteristics relevant to action, such as distance, to the dorsal visual pathway. However, these frameworks are based on studies that used projected images of objects whose actual size and distance from the observer were ambiguous. Here, we find that when object size and distance information in the stimulus is less ambiguous, these characteristics are widely represented in both visual pathways. Our results provide valuable new insights into the brain representations of objects and their various physical attributes in the context of naturalistic vision.

Temporal differences and commonalities between hand and tool neural processing

Object recognition is a complex cognitive process that relies on how the brain organizes object-related information. While spatial principles have been extensively studied, less studied temporal dynamics may also offer valuable insights into this process, particularly when neural processing overlaps for different categories, as it is the case of the categories of hands and tools. Here we focus on the differences and/or similarities between the time-courses of hand and tool processing under electroencephalography (EEG). Using multivariate pattern analysis, we compared, for different time points, classification accuracy for images of hands or tools when compared to images of animals. We show that for particular time intervals (~ 136-156 ms and ~ 252-328 ms), classification accuracy for hands and for tools differs. Furthermore, we show that classifiers trained to differentiate between tools and animals generalize their learning to classification of hand stimuli between ~ 260-320 ms and ~ 376-500 ms after stimulus onset. Classifiers trained to distinguish between hands and animals, on the other hand, were able to extend their learning to the classification of tools at ~ 150 ms. These findings suggest variations in semantic features and domain-specific differences between the two categories, with later-stage similarities potentially related to shared action processing for hands and tools.

Neural and behavioral signatures of the multidimensionality of manipulable object processing

Understanding how we recognize objects requires unravelling the variables that govern the way we think about objects and the neural organization of object representations. A tenable hypothesis is that the organization of object knowledge follows key object-related dimensions. Here, we explored, behaviorally and neurally, the multidimensionality of object processing. We focused on within-domain object information as a proxy for the decisions we typically engage in our daily lives - e.g., identifying a hammer in the context of other tools. We extracted object-related dimensions from subjective human judgments on a set of manipulable objects. We show that the extracted dimensions are cognitively interpretable and relevant - i.e., participants are able to consistently label them, and these dimensions can guide object categorization; and are important for the neural organization of knowledge - i.e., they predict neural signals elicited by manipulable objects. This shows that multidimensionality is a hallmark of the organization of manipulable object knowledge.

Semantic feature production norms for manipulable objects

Feature generation tasks and feature databases are important for understanding how knowledge is organized in semantic memory, as they reflect not only the kinds of information that individuals hold about objects but also how objects are conceptually represented. Traditionally, semantic norms focus on a variety of object categories and, as a result, have a small number of concepts per semantic category. Here, our main goal is to create a more fine-grained feature database exclusively for one category of objects-manipulable objects. This database contributes to the understanding of within-category, content-specific processing. To achieve this, we asked 130 participants to freely generate features for 80 manipulable objects and another group of 32 participants to generate action features for the same objects. We then compared our databases with other published semantic norms and found high similarity between them. In our databases, we calculated the similarity between objects in terms of visual, functional, encyclopaedic, and action feature types using Spearman correlation, Baker's gamma index, and cophenetic correlation. We discovered that objects were grouped in a distinctive and meaningful way according to feature type. Finally, we tested the validity of our databases by asking three groups of participants to perform a feature verification experiment while manipulating production frequency. Our results demonstrate that participants can recognize and associate the features of our databases with specific manipulable objects. Participants were faster to verify high-frequency features than low-frequency features. Overall, our data provide important insights into how we process manipulable objects and can be used to further inform cognitive and neural theories of object processing and identification.

Two distinct networks containing position-tolerant representations of actions in the human brain

Humans can recognize others' actions in the social environment. This action recognition ability is rarely hindered by the movement of people in the environment. The neural basis of this position tolerance for observed actions is not fully understood. Here, we aimed to identify brain regions capable of generalizing representations of actions across different positions and investigate the representational content of these regions. In a functional magnetic resonance imaging experiment, participants viewed point-light displays of different human actions. Stimuli were presented in either the upper or the lower visual field. Multivariate pattern analysis and a surface-based searchlight approach were employed to identify brain regions that contain position-tolerant action representation: Classifiers were trained with patterns in response to stimuli presented in one position and were tested with stimuli presented in another position. Results showed above-chance classification in the left and right lateral occipitotemporal cortices, right intraparietal sulcus, and right postcentral gyrus. Further analyses exploring the representational content of these regions showed that responses in the lateral occipitotemporal regions were more related to subjective judgments, while those in the parietal regions were more related to objective measures. These results provide evidence for two networks that contain abstract representations of human actions with distinct representational content.

Disentangling hand and tool processing: Distal effects of neuromodulation

Neural processing within a local brain region that responds to more than one object category (e.g., hands and tools) nonetheless have different functional connectivity patterns with other distal brain areas, which suggests that local processing can affect and/or be affected by processing in distal areas, in a category-specific way. Here we wanted to test whether administering either a hand- or tool-related training task in tandem with transcranial direct current stimulation (tDCS) to a region that responds both to hands and tools (posterior middle temporal gyrus; pMTG), modulated local and distal neural processing more for the trained than the untrained category in a subsequent fMRI task. After each combined tDCS/training session, participants viewed images of tools, hands, and animals, in an fMRI scanner. Using multivoxel pattern analysis, we found that tDCS stimulation to pMTG indeed improved the classification accuracy between tools vs. animals, but only when combined with a tool and not a hand training task. Surprisingly, tDCS stimulation to pMTG also improved classification accuracy between hands vs. animals when combined with a tool but not a hand training task. Our findings suggest that overlapping but functionally-specific networks may be engaged separately by using a category-specific training task together with tDCS - a strategy that can be applied more broadly to other cognitive domains using tDCS. By hypothesis, these effects on local processing are a direct result of within-domain connectivity constraints from domain-specific networks that are at play in the processing and organization of object representations.

Tool heads prime saccades

Tools are wielded by their handles, but a lot of information about their function comes from their heads (the action-ends). Here we investigated whether eye saccadic movements are primed by tool handles, or whether they are primed by tool heads. We measured human saccadic reaction times while subjects were performing an attentional task. We found that saccades were executed quicker when performed to the side congruent with the tool head, even though "toolness" was irrelevant for the task. Our results show that heads are automatically processed by the visual system to orient eye movements, indicating that eyes are attracted by functional parts of manipulable objects and by the characteristic information these parts convey.

Overlapping but distinct: Distal connectivity dissociates hand and tool processing networks

The processes and organizational principles of information involved in object recognition have been a subject of intense debate. These research efforts led to the understanding that local computations and feedforward/feedback connections are essential to our representations and their organization. Recent data, however, has demonstrated that distal computations also play a role in how information is locally processed. Here we focus on how long-range connectivity and local functional organization of information are related, by exploring regions that show overlapping category-preferences for two categories and testing whether their connections are related with distal representations in a category-specific way. We used an approach that relates functional connectivity with distal areas to local voxel-wise category-preferences. Specifically, we focused on two areas that show an overlap in category-preferences for tools and hands-the inferior parietal lobule/anterior intraparietal sulcus (IPL/aIPS) and the posterior middle temporal gyrus/lateral occipital temporal cortex (pMTG/LOTC) - and how connectivity from these two areas relate to voxel-wise category-preferences in two ventral temporal regions dedicated to the processing of tools and hands separately-the left medial fusiform gyrus and the fusiform body area respectively-as well as across the brain. We show that the functional connections of the two overlap areas correlate with categorical preferences for each category independently. These results show that regions that process both tools and hands maintain object topography in a category-specific way. This potentially allows for a category-specific flow of information that is pertinent to computing object representations.

Decoding stimuli (tool-hand) and viewpoint invariant grasp-type information

When we see a manipulable object (henceforth tool) or a hand performing a grasping movement, our brain is automatically tuned to how that tool can be grasped (i.e., its affordance) or what kind of grasp that hand is performing (e.g., a power or precision grasp). However, it remains unclear where visual information related to tools or hands are transformed into abstract grasp representations. We therefore investigated where different levels of abstractness in grasp information are processed: grasp information that is invariant to the kind of stimuli that elicits it (tool-hand invariance); and grasp information that is hand-specific but viewpoint-invariant (viewpoint invariance). We focused on brain areas activated when viewing both tools and hands, i.e., the posterior parietal cortices (PPC), ventral premotor cortices (PMv), and lateral occipitotemporal cortex/posterior middle temporal cortex (LOTC/pMTG). To test for invariant grasp representations, we presented participants with tool images and grasp videos (from first or third person perspective; 1pp or 3pp) inside an MRI scanner, and cross-decoded power versus precision grasps across (i) grasp perspectives (viewpoint invariance), (ii) tool images and grasp 1pp videos (tool-hand 1pp invariance), and (iii) tool images and grasp 3pp videos (tool-hand 3pp invariance). Tool-hand 1pp, but not tool-hand 3pp, invariant grasp information was found in left PPC, whereas viewpoint-invariant information was found bilaterally in PPC, left PMv, and left LOTC/pMTG. These findings suggest different levels of abstractness-where visual information is transformed into stimuli-invariant grasp representations/tool affordances in left PPC, and viewpoint invariant but hand-specific grasp representations in the hand network.

Domain-Specific Diaschisis: Lesions to Parietal Action Areas Modulate Neural Responses to Tools in the Ventral Stream

Neural responses to small manipulable objects ("tools") in high-level visual areas in ventral temporal cortex (VTC) provide an opportunity to test how anatomically remote regions modulate ventral stream processing in a domain-specific manner. Prior patient studies indicate that grasp-relevant information can be computed about objects by dorsal stream structures independently of processing in VTC. Prior functional neuroimaging studies indicate privileged functional connectivity between regions of VTC exhibiting tool preferences and regions of parietal cortex supporting object-directed action. Here we test whether lesions to parietal cortex modulate tool preferences within ventral and lateral temporal cortex. We found that lesions to the left anterior intraparietal sulcus, a region that supports hand-shaping during object grasping and manipulation, modulate tool preferences in left VTC and in the left posterior middle temporal gyrus. Control analyses demonstrated that neural responses to "place" stimuli in left VTC were unaffected by lesions to parietal cortex, indicating domain-specific consequences for ventral stream neural responses in the setting of parietal lesions. These findings provide causal evidence that neural specificity for "tools" in ventral and lateral temporal lobe areas may arise, in part, from online inputs to VTC from parietal areas that receive inputs via the dorsal visual pathway.

The dominant role of functional action representation in object recognition

Action representation of manipulable objects has been found to be involved in object recognition. Recently, studies have indicated the existence of two distinct action systems: functional action specifying how to use an object and structural action concerning how to grasp an object. Despite evidence revealing the systems' anatomical and functional differences, few preceding studies have dissociated their respective roles in object recognition. The present study aimed to tease apart their roles in the recognition of manipulable objects with a priming paradigm. Specifically, we used static stimuli (photos, Experiments 1 and 2) and dynamic stimuli (video clips, Experiments 3 and 4) depicting functional and structural action hand gestures as primes and measured the magnitude of functional and structural action priming effect in object recognition. We found that static and dynamic priming stimuli induced a robust action priming effect only for functional action prime-target pairs. Naming latencies of the target objects were shorter when functional action representations of the prime and target were congruent than when they were incongruent. Moreover, as compared to static priming photos, dynamic priming stimuli induced a larger functional action priming effect. By contrast, neither static nor dynamic priming stimuli elicited a structural action priming effect. Behavioral data from our four experiments provide consistent evidence of the dominant role of functional action representation in the recognition of manipulable objects, suggesting that action knowledge regarding how to use rather than grasp an object is more likely an intrinsic component of objects' conceptual representation.