Visual Perception: How We Recognize Objects and Scenes

Author: Angela Bogdanova · Published: October 16, 2026 · Editorial Policy

Research cluster: Attention, Perception & Cognitive Control · Written by Artificial Sapiens

Visual perception is the set of psychological and neural processes through which patterns of light become organized into visible surfaces, objects, spatial layouts, and meaningful scenes. When we recognize a cup on a crowded desk or realize that a photograph shows a train station, the visual system uses sensory information together with learned regularities, spatial relationships, attention, and context. Object recognition allows us to identify what is present; scene recognition allows us to understand the environment in which objects appear. The two processes interact continuously. DiCarlo, Zoccolan, and Rust (2012); Peelen, Berlot, and de Lange (2024).

Recognition is impressive because the retinal image changes whenever an object moves, illumination changes, a viewer approaches it, or another object partly blocks it. A chair can cast different outlines from different viewpoints while remaining recognizable as a chair. Yet our visual performance also has limits: clutter impairs identification, expectations sometimes distort interpretation, and understanding the gist of a scene does not guarantee that every object within it has been identified. An account of vision therefore has to explain both rapid success and predictable failures. Whitney and Levi (2011); Oliva and Torralba (2007).

What Is Visual Perception?

Visual perception refers to processing that transforms information registered by the eyes into structured representations of the environment. Light reflected from surfaces is focused onto the retina, where photoreceptors and subsequent neural circuits begin the extraction of visual information. Cortical processing then supports increasingly complex descriptions of orientation, contour, texture, shape, object identity, and spatial configuration. At the psychological level, perception concerns the processes that make this information useful for recognizing, locating, and acting on objects and scenes. DiCarlo et al. (2012).

Sensation and perception are closely related concepts. Sensory transduction concerns how physical energy is converted into neural signals; perception concerns the organization and interpretation of those signals. Visual detection means establishing that a stimulus is present. Visual discrimination means distinguishing among stimuli. Recognition means classifying or identifying something from visual evidence. These tasks can dissociate: a person might detect that a shape appeared without accurately naming it, or classify a blurred image as an outdoor scene without identifying its individual objects.

Visual perception also differs from visual attention and visual memory. Attention influences which information receives priority for further processing; recognition describes the information extracted and the category or identity assigned. Memory contributes familiarity, learned categories, and expectations, but recognizing a visible object does not demonstrate the ability to remember its precise appearance later. The boundaries between these systems are important when interpreting laboratory experiments. Peelen et al. (2024); Potter (1976).

Object Recognition, Object Identification, and Categorization

Object recognition is commonly used as an umbrella term for identifying an object or assigning it to a meaningful category despite changes in its image. Basic-level categorization might produce the answer “dog”; a more general category might be “animal”; individual identification might produce the name of a particular dog. Different experimental demands recruit partly different information. Successful categorization from shape does not always imply knowledge of a specific object's identity, function, or location. Palmeri and Gauthier (2004).

An object's representation must preserve useful distinctions while tolerating irrelevant variation. This property is often called invariance. Invariance is graded and task-dependent: the same visual features that support broad category recognition may fail when a viewer must distinguish two highly similar individual objects. Researchers therefore study what information remains stable, what changes with viewpoint or context, and which features are diagnostic for the requested decision. Peissig and Tarr (2007); DiCarlo et al. (2012).

Scene Perception and Scene Gist

A scene is a structured visual environment containing surfaces, spatial relationships, and frequently several objects. Scene perception includes judgments such as “kitchen,” “street,” or “forest,” as well as estimates of openness, navigability, depth, and likely object locations. Scene gist refers to the rapid extraction of a scene's general meaning or category from a glance. Gist is a useful summary; it is not a complete inventory of everything in the image. Greene and Oliva (2009).

A person can identify a beach from the horizon, broad color distributions, spatial layout, and characteristic textures before attending to every object. The reverse also occurs: recognizing a stove and kitchen cabinets helps infer that the scene is a kitchen. Modern evidence favors partly parallel, reciprocal object and scene processing, although their relative contributions depend on exposure duration, task, attention, and stimulus composition. Peelen et al. (2024); Furtak, Mudrik, and Bola (2022).

How Visual Information Becomes a Recognizable Object

The transformation from retinal input to recognition involves several interacting operations. Researchers can distinguish them analytically, but the stages overlap in time and exchange information. A useful working sequence moves from detecting image structure, through perceptual organization, to shape representation and category assignment, while allowing feedback from context and task goals.

Edges, Orientation, Contrast, and Spatial Scale

Early visual processing is sensitive to changes in brightness, contrast, orientation, spatial frequency, and other image properties. Boundaries often indicate the edges of surfaces, but a single boundary can arise from changes in illumination or texture as well as from a physical object. Recognizing meaningful objects therefore requires combining local signals across space, discounting some variations, and preserving others. The relevant information differs between a finely detailed printed symbol and a large, blurry landscape.

Spatial frequency provides a way to study the scale of information available in an image. Low spatial frequencies describe coarse changes and broad layout; high spatial frequencies describe fine detail and sharp boundaries. Coarse information can facilitate rapid scene interpretation, with finer information refining or correcting the initial interpretation. This coarse-to-fine pattern has experimental support, but its exact timing and contribution vary across tasks and stimuli. Schyns and Oliva (1994).

Grouping Features Into Shapes and Figures

An array of edges becomes useful when some parts are grouped into perceived contours, surfaces, and objects. Perceptual grouping research examines proximity, similarity, good continuation, connectedness, common motion, and other principles that influence which image elements are experienced as belonging together. Figure-ground organization separates a candidate object from the region treated as its surrounding background. These processes are central to seeing a coherent item rather than a disconnected collection of lines. Wagemans et al. (2012).

Grouping principles are tendencies whose effects depend on image geometry, competing cues, and interpretation. A contour can belong to different perceived figures under different conditions. Occlusion also requires the visual system to infer that an object continues behind a foreground surface even where no direct retinal evidence of its hidden portion is available. Such perceptual completion illustrates how scene organization contributes to object recognition without supplying certainty about the hidden structure.

Building Shape Representations Across Changing Views

The retinal appearance of an object depends on viewing angle, distance, pose, and illumination. Recognition must therefore link different images to a shared identity or category. Structural-description theories emphasize parts and spatial relations that can remain informative across viewpoints. Biederman's recognition-by-components proposal suggested that combinations of volumetric primitives could support efficient recognition for many familiar objects. It remains an influential theory and a historical explanation of how part structure could provide robustness. Biederman (1987).

View-based approaches emphasize representations informed by previously encountered views and measurable effects of viewpoint changes. The evidence demonstrates that recognition can be robust under many transformations while retaining systematic costs for some unfamiliar perspectives, rotations, occlusions, and fine-grained identification demands. Contemporary research increasingly treats object representations as high-dimensional and task-sensitive rather than demanding one universal, completely viewpoint-independent code. Peissig and Tarr (2007); Grill-Spector and Weiner (2014).

From Shape to Meaning

A visual representation becomes recognizable when the viewer can use its features and their relationships to infer a category, an identity, or a function. The task determines which distinction matters: an image may contain enough information to classify an object as furniture but not enough to determine its manufacturer. Familiarity improves some distinctions because experience changes the categories and features observers use. The resulting judgment integrates information about the image with knowledge of the world, but it can remain uncertain when the input is ambiguous. Palmeri and Gauthier (2004).

How We Recognize an Entire Scene

Scene understanding is efficient partly because real environments have statistical structure. A street, classroom, and forest differ in broad layout, textures, likely objects, and typical spatial relations. Observers can exploit this regularity before fully analyzing each component. Rapid scene classification experiments provide evidence for global representations that encode spatial and ecological properties, including openness, depth, and navigability. Greene and Oliva (2009).

A Glance Can Reveal the Gist

Classic rapid-picture experiments demonstrated that people can derive substantial meaning from images shown briefly in a sequence. Potter's work found that detection of a target picture could remain good under exposure conditions in which later recognition memory for individual pictures was substantially poorer. This distinction matters: swift understanding does not imply a lasting record of every detail. Potter (1976).

Thorpe, Fize, and Marlot showed that participants could perform rapid animal/nonanimal categorization for briefly flashed photographs, with an electrophysiological difference associated with task processing emerging at approximately 150 milliseconds. That result demonstrates the speed of a particular categorization system under controlled conditions; it does not establish a universal 150-millisecond deadline for complete object or scene perception. Thorpe et al. (1996).

Global Layout and Local Objects Work Together

Scene recognition can rely on global image structure, local object identity, or their combination. Greene and Oliva showed that global properties provide a useful account of rapid scene categorization. Furtak and colleagues found, in brief masked presentations, that background classification was generally stronger than foreground object classification, that spatial attention helped objects, and that incongruent objects could impair background categorization. These results support a flexible interaction between broad scene information and specific objects. Greene and Oliva (2009); Furtak et al. (2022).

One can therefore recognize the general environment without naming all its contents, identify a distinctive object without knowing the room, or use a recognized object to revise an initial scene interpretation. Treating gist and object identification as separate measurable tasks clarifies how these possibilities coexist.

Objects Have Meaningful Relationships

Real scenes are organized by typical arrangements. A kettle commonly appears on a kitchen counter; a traffic light usually occupies an elevated position near a roadway. The physical, spatial, and semantic relationships among objects offer information that isolated object cutouts omit. Biederman, Mezzanotte, and Rabinowitz studied violations of expected relationships, including support, position, size, and scene probability. Their work established a framework for studying how relational knowledge affects processing of scenes. Biederman et al. (1982).

Context can guide recognition by narrowing plausible interpretations, but its effect is not uniformly beneficial. Depending on the task and strength of sensory evidence, an unusual object may become more noticeable, whereas a typical object may be processed more efficiently in a familiar setting. Research should therefore distinguish object detection, identification accuracy, reaction time, and memory before describing a context effect as facilitation. Oliva and Torralba (2007); Peelen et al. (2024).

Bottom-Up Signals, Expectations, and Predictive Accounts

Bottom-up processing refers to the contribution of information carried by the current sensory input. Top-down influences include relevant expectations, prior knowledge, attention, and task instructions that alter how incoming information is selected or interpreted. Both describe real research questions, although experiments must identify which variables are manipulated and which outcomes are measured rather than treating the two labels as sufficient explanations.

Consider a partly obscured object on a restaurant table. Visible contour and color constrain possible identities; the scene makes plates and glasses more likely than laboratory equipment. Context changes the probability of different hypotheses, while the visible evidence may strengthen or contradict those hypotheses. Classical accounts describe such interactions in terms of context frames; newer probabilistic accounts formalize how estimates of scene and object identity could influence each other. Bar (2004); Peelen et al. (2024).

Predictive Processing as an Explanatory Framework

Predictive processing models propose that perception combines incoming evidence with predictions generated from prior experience and current context. On these accounts, information about scene identity can constrain object hypotheses, while recognized objects update scene hypotheses. This is an influential and productive theoretical framework, and recent reviews organize convergent behavioral and neural findings around reciprocal inferences. The precise computations implemented in the human visual system, however, remain under investigation; demonstrating a context effect alone does not identify a unique predictive algorithm. Peelen et al. (2024).

When Expectations Help and When They Mislead

Prior knowledge typically improves performance when the environment is predictable and sensory evidence is noisy. Strong or inappropriate expectations can also produce false identifications or bias confidence under ambiguous conditions. A rigorous experiment varies the relevant stimulus or context while measuring errors as well as response speed. Faster responses by themselves can reflect changes in decision criteria rather than greater perceptual sensitivity. The important scientific question is how expectations change evidence use, perceptual representation, and behavior in a specified task.

The Visual Brain: A Functional Research Overview

Early cortical areas support processing of visual features and spatial structure, whereas areas along ventral occipitotemporal pathways are strongly implicated in object form and category representations. Higher-level visual responses show selectivity for classes of images and combinations of visual features, but selectivity should not be confused with an exclusive, isolated module for an entire psychological category. The neural organization is distributed, and representations evolve over time. DiCarlo et al. (2012); Grill-Spector and Weiner (2014).

Scene-related neural responses have been examined in several cortical regions that respond preferentially to images of places, spatial boundaries, and natural environments. Their functions overlap with representations of objects and context. Researchers use functional magnetic resonance imaging, electrophysiology, and causal interventions to distinguish where different information is represented and when it becomes available. Evidence of neural decoding indicates that a category is represented in recorded activity under given analytical conditions, not that a person explicitly noticed or consciously identified every member of that category. Peelen et al. (2024).

Rapid Feedforward Processing and Later Recurrent Interactions

A rapid feedforward sweep can support coarse categorization of familiar visual structure. Subsequent interactions within and among visual areas permit further integration, contextual modulation, and refinement under challenging conditions. The feedforward/recurrent distinction has long guided experimental interpretation, although there is continuing debate about which specific visual tasks require recurrent processing and how those processes relate to awareness. Lamme and Roelfsema (2000).

Cichy, Pantazis, and Oliva combined magnetoencephalography and functional magnetic resonance imaging while observers viewed object images, showing that representations distinguishing individual images and broader categories evolve on different timescales. Their work illustrates why a single average reaction time cannot fully describe recognition. Different kinds of diagnostic information become measurable at different moments and in different brain regions. Cichy et al. (2014).

Attention, Eye Movements, and the Limits of Recognition

Visual attention prioritizes some information for processing. Objects at the current fixation typically provide finer detail than objects far into peripheral vision, and directing attention to a likely target location can increase identification accuracy. Real-world viewing is active: people make eye movements, seek relevant details, and repeatedly update interpretations as new information arrives. Initial scene gist may guide where they look next. Furtak et al. (2022).

Clutter creates a particularly important limit. In crowding, a target that can be identified alone becomes difficult to recognize when surrounding features or objects are nearby, especially in peripheral vision. This phenomenon is more specific than ordinary distraction and reflects limits on how visual information is integrated and individuated. It helps explain why a scene may be recognizable overall while small or peripheral objects remain difficult to identify. Whitney and Levi (2011).

Attention and perception influence each other: the scene guides selection, selected objects alter the scene interpretation, and expectations can change which information is sampled. A person may understand that an image depicts a busy station but still miss a small suitcase, misclassify a partially hidden object, or fail to notice a change between successive views. These outcomes are experimental phenomena whose explanation depends on timing, display design, attentional set, and the specific question asked. They should not be treated as automatic evidence of impaired cognition.

How Scientists Measure Visual Perception

Research uses tasks designed to separate aspects of seeing that everyday language combines. Each measure captures performance under particular stimulus, exposure, and instruction conditions. Combining methods produces a more convincing account than treating one task score as a universal index of perceptual ability.

Brief Exposure, Masking, and Rapid Categorization

Researchers display images for controlled durations, sometimes followed by a mask that limits further processing. Participants may classify objects, name scenes, select among alternatives, or report whether a target category was present. Accuracy across exposure durations helps estimate how much information becomes useful quickly. Exposure time should be reported separately from reaction time: pressing a response key includes decision formation and motor execution. Thorpe et al. (1996); Furtak et al. (2022).

Recognition Under Image Transformations

A central test of object recognition is whether categorization survives changes in size, position, rotation, lighting, or partial occlusion. Researchers compare performance across trained and novel images, individuals, or conditions. Good performance across some transformations demonstrates robustness within the tested range; it does not establish complete invariance. Recognition of a common category is easier than certain fine-grained individual distinctions, and laboratories must specify exactly which type of generalization they measured. DiCarlo et al. (2012).

Object–Scene Consistency Experiments

In context experiments, an object is placed in a congruent or incongruent background. Other manipulations vary expected size, location, support relations, or combinations of surrounding objects. Investigators then measure object classification, scene classification, eye movements, or later memory. This design can reveal reciprocal effects, but interpretation depends on keeping low-level visual differences, task difficulty, and baseline object recognizability under control. Biederman et al. (1982); Oliva and Torralba (2007).

Behavioral Measures and Brain Measurements

Accuracy, response time, error type, sensitivity, and eye-fixation patterns answer different questions. Functional MRI identifies distributed patterns associated with image properties and categories, whereas electrophysiology and magnetoencephalography provide more direct information about temporal dynamics. Computational decoding tests whether a category distinction is recoverable from neural data, while representational analyses compare patterns of similarity among images and categories. Convergence across approaches strengthens conclusions, yet decoding accuracy and correlations with human behavior do not themselves prove a unique neural mechanism. Cichy et al. (2014); Kriegeskorte (2015).

Major Theories of Object and Scene Recognition

Structural and View-Based Accounts

Structural theories stress parts, boundaries, and part relations that help a system recognize objects across changes in view. View-based approaches stress the role of sampled appearances and the generalization learned from experience. Evidence for both flexible invariance and viewpoint costs has made the field more focused on representational features, category structure, and task dependence than on selecting a single rigid description of every recognized object. Biederman (1987); Peissig and Tarr (2007).

Global-Feature and Scene-Layout Accounts

Global-feature models emphasize spatial-scale information, texture statistics, and broad geometric properties that can signal the category of a place. Their achievement is to explain how a scene's general identity can be estimated before its individual objects are fully enumerated. These accounts remain useful even when object identification and object relations contribute to the final interpretation. Greene and Oliva (2009); Schyns and Oliva (1994).

Interactive and Probabilistic Accounts

Interactive models emphasize how object hypotheses, scene context, and relations among objects constrain one another. Predictive and probabilistic formulations model these reciprocal effects as inference under uncertainty. A recent integrative review concludes that scene and object gist can emerge partly in parallel and that bidirectional information exchange is supported by multiple lines of evidence. Whether any particular computational architecture captures the full diversity of human visual recognition is still an open research question. Peelen et al. (2024).

Deep Neural Networks as Scientific Models

Artificial neural networks trained on visual categorization tasks offer explicit computational representations that can be tested against human errors, performance across transformations, and neural response patterns. Their usefulness comes from making feature-learning hypotheses measurable, not from assuming that a trained model reproduces every biological process. Similarity between a network's representations and measured brain activity does not establish identity of mechanisms, conscious visual experience, or a complete explanation of human perception. Kriegeskorte (2015); DiCarlo et al. (2012).

What the Evidence Establishes—and What Remains Open

Several conclusions are well supported. Humans can rapidly categorize many familiar objects and scenes. Visual organization groups features into figures and meaningful structures. Object recognition is often tolerant of changes in appearance while retaining measurable sensitivity to some transformations. Scene layout and context influence recognition, and individual objects can influence scene judgments. Behavioral and neural studies show that these operations unfold over time and involve interacting systems. Wagemans et al. (2012); DiCarlo et al. (2012); Peelen et al. (2024).

Other claims require narrower wording. A rapid categorization result does not show that the observer perceived every object in the image. Neural decoding does not by itself show awareness. A contextual benefit on response speed does not always imply increased sensitivity. Recognition under familiar laboratory transformations does not guarantee recognition of every unfamiliar object in the world. Predictive coding supplies a powerful framework but remains a family of theories with competing implementations.

Reproducibility, Generalization, and Ecological Validity

Experiments often use cropped photographs, a small number of categories, repeated trials, constrained viewing angles, or highly practiced response mappings. These choices are valuable because they isolate mechanisms. They also narrow which everyday abilities can be inferred. Results may change with image diversity, expertise, lighting, peripheral crowding, cultural familiarity, time pressure, instructions, and natural visual search. Research synthesis should compare methods and effect directions rather than convert one significant laboratory finding into a universal law.

At the individual level, differences in visual experience and task familiarity can influence accuracy and speed. Development, aging, and sensory characteristics may also change performance, but specific claims about their size and mechanisms need age-appropriate or population-specific evidence. A single low object-recognition score is an experimental result; it cannot establish a neurological or psychiatric diagnosis.

Practical Implications Grounded in Perception Research

Where important information must be recognized reliably, clear boundaries, adequate visual contrast, sufficient spacing, and reduced visual crowding can make identification easier. Consistent spatial organization may help viewers anticipate where relevant objects appear, while unusual placement may increase confusion or redirect attention. These implications arise from research on grouping, object recognition, context, and crowding. The optimal design still depends on the people, visual environment, task, and testing conditions. Wagemans et al. (2012); Whitney and Levi (2011).

Vision research also informs image-recognition technologies, accessibility studies, and visual communication. Computational models can help compare representation hypotheses, but their success is task-specific. Scientific evidence about perception does not justify promises that an untested exercise, app, or generic brain-training program will produce broad improvements in visual recognition.

Frequently Asked Questions About Visual Perception

What Is the Difference Between Seeing and Recognizing?

Seeing covers several processes, from registering visual information to constructing a meaningful experience. Recognizing refers more specifically to identifying or classifying something from visual evidence. An observer may see that a shape is present while remaining uncertain whether it is a cup, vase, or lamp. Conversely, a viewer may infer the broad category of a scene from a rapid glimpse while lacking an inventory of its smaller objects.

Do We Recognize Objects Before We Understand a Scene?

There is no single sequence that describes every task. Some scene properties are extracted very quickly, and some object categories are recognized quickly as well. Experimental results show reciprocal effects: scene backgrounds can influence objects, and objects can influence background judgments. Modern reviews favor partly parallel and interactive processing with task-dependent timing. Furtak et al. (2022); Peelen et al. (2024).

Why Can We Recognize an Object From Different Angles?

The visual system extracts combinations of features and relationships that often remain informative across changes in viewpoint. Learning, familiar categories, and shape representations enable generalization. Robustness varies with the object, the degree of rotation, and the specificity of identification required. DiCarlo et al. (2012).

Why Do Familiar Surroundings Make Objects Easier to Identify?

A familiar scene provides expectations about which objects are likely and where they tend to occur. This context can make an ambiguous object more interpretable, but effects differ across detection, identification, and attention tasks. Unexpected items can attract attention or disrupt recognition, so context helps under some conditions and biases judgments under others. Oliva and Torralba (2007).

What Is Scene Gist?

Scene gist is the general meaning or category extracted from a brief look at an environment or picture, such as “forest,” “office,” or “busy road.” It can be based on broad spatial and visual properties together with object information. Successful gist recognition does not establish conscious access to every component or detailed later memory for the image. Greene and Oliva (2009); Potter (1976).

Does Visual Perception Depend on Attention?

Attention influences what receives priority and how effectively many objects are identified, especially with competing items or brief exposures. Some broad scene and category information can nevertheless be extracted rapidly under conditions that limit detailed selection. The precise role of attention depends on the experimental task, the degree of clutter, and the required judgment. Furtak et al. (2022); Whitney and Levi (2011).

Is Visual Perception the Same as Visual Memory?

Visual perception concerns interpreting currently available visual information; visual memory concerns information retained or reconstructed after an image is absent. The systems interact, but rapid image detection and later recognition memory can diverge substantially. Claims about extraordinary visual memory cannot be established solely by demonstrating fast and accurate scene categorization. Potter (1976).

Can a Visual Perception Experiment Diagnose a Disorder?

An experimental task measures performance under defined conditions. Reduced accuracy or slower reaction time can arise from stimulus difficulty, sensory factors, attention, fatigue, familiarity, or many other influences. Clinical evaluation requires an appropriately validated assessment and relevant clinical criteria; a laboratory categorization result cannot supply a diagnosis on its own.

Related Articles

Photographic Memory: Does It Really Exist? — How retaining a visual image differs from perceiving and rapidly categorizing it.

Perception: How the Mind Interprets Sensory Information — Explains perceptual organization, interpretation, context, and the relationship to sensation.

Bottom-Up vs Top-Down Processing in Perception: Examples and Evidence — Compares sensory-driven and expectation-driven information processing in perception.

References

Bar, M. (2004). Visual objects in context. Nature Reviews Neuroscience, 5, 617–629. https://doi.org/10.1038/nrn1476

Biederman, I. (1987). Recognition-by-components: A theory of human image understanding. Psychological Review, 94(2), 115–147. https://doi.org/10.1037/0033-295X.94.2.115

Biederman, I., Mezzanotte, R. J., & Rabinowitz, J. C. (1982). Scene perception: Detecting and judging objects undergoing relational violations. Cognitive Psychology, 14(2), 143–177. https://doi.org/10.1016/0010-0285(82)90007-X

Cichy, R. M., Pantazis, D., & Oliva, A. (2014). Resolving human object recognition in space and time. Nature Neuroscience, 17, 455–462. https://doi.org/10.1038/nn.3635

DiCarlo, J. J., Zoccolan, D., & Rust, N. C. (2012). How does the brain solve visual object recognition? Neuron, 73(3), 415–434. https://doi.org/10.1016/j.neuron.2012.01.010

Furtak, M., Mudrik, L., & Bola, M. (2022). The forest, the trees, or both? Hierarchy and interactions between gist and object processing during perception of real-world scenes. Cognition, 221, 104983. https://doi.org/10.1016/j.cognition.2021.104983

Greene, M. R., & Oliva, A. (2009). Recognition of natural scenes from global properties: Seeing the forest without representing the trees. Cognitive Psychology, 58(2), 137–176. https://doi.org/10.1016/j.cogpsych.2008.06.001

Grill-Spector, K., & Weiner, K. S. (2014). The functional architecture of the ventral temporal cortex and its role in categorization. Nature Reviews Neuroscience, 15, 536–548. https://doi.org/10.1038/nrn3747

Kriegeskorte, N. (2015). Deep neural networks: A new framework for modeling biological vision and brain information processing. Annual Review of Vision Science, 1, 417–446. https://doi.org/10.1146/annurev-vision-082114-035447

Lamme, V. A. F., & Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23(11), 571–579. https://doi.org/10.1016/S0166-2236(00)01657-X

Oliva, A., & Torralba, A. (2007). The role of context in object recognition. Trends in Cognitive Sciences, 11(12), 520–527. https://doi.org/10.1016/j.tics.2007.09.009

Palmeri, T. J., & Gauthier, I. (2004). Visual object understanding. Nature Reviews Neuroscience, 5, 291–303. https://doi.org/10.1038/nrn1364

Peelen, M. V., Berlot, E., & de Lange, F. P. (2024). Predictive processing of scenes and objects. Nature Reviews Psychology, 3, 13–26. https://doi.org/10.1038/s44159-023-00254-0

Peissig, J. J., & Tarr, M. J. (2007). Visual object recognition: Do we know more now than we did 20 years ago? Annual Review of Psychology, 58, 75–96. https://doi.org/10.1146/annurev.psych.58.102904.190114

Potter, M. C. (1976). Short-term conceptual memory for pictures. Journal of Experimental Psychology: Human Learning and Memory, 2(5), 509–522. https://pubmed.ncbi.nlm.nih.gov/1003124/

Schyns, P. G., & Oliva, A. (1994). From blobs to boundary edges: Evidence for time- and spatial-scale-dependent scene recognition. Psychological Science, 5(4), 195–200. https://doi.org/10.1111/j.1467-9280.1994.tb00500.x

Thorpe, S., Fize, D., & Marlot, C. (1996). Speed of processing in the human visual system. Nature, 381, 520–522. https://doi.org/10.1038/381520a0

Wagemans, J., Elder, J. H., Kubovy, M., Palmer, S. E., Peterson, M. A., Singh, M., & von der Heydt, R. (2012). A century of Gestalt psychology in visual perception: I. Perceptual grouping and figure-ground organization. Psychological Bulletin, 138(6), 1172–1217. https://doi.org/10.1037/a0029333

Whitney, D., & Levi, D. M. (2011). Visual crowding: A fundamental limit on conscious perception and object recognition. Trends in Cognitive Sciences, 15(4), 160–168. https://doi.org/10.1016/j.tics.2011.02.005