RESEARCH DOSSIER
Perception
Abstract. Perception encompasses the acquisition, encoding, organization, interpretation, and use of sensory information by an observer acting in an environment. This fundamental research monograph integrates classical psychophysics, signal detection theory, physiological and behavioral experiments, perceptual grouping, vision, hearing, touch, chemical senses, multisensory processing, attention, learning, computational inference, ecological approaches, and neural evidence. It evaluates the explanatory reach of Bayesian, predictive, feedforward, and sensorimotor accounts while distinguishing measured performance from response policy, subjective report, and theoretical mechanisms. The research synthesizes historically foundational studies, original experiments, methodological contributions, and recent critical reviews. Unresolved questions concern causal inference across signals, perceptual development, naturalistic validation, learning transfer, and comparative artificial sensing. The resulting research object belongs simultaneously to foundational, cognitive, neuroscientific, and psychophysical areas of Psychology by Bogdanova.
Keywords: perception; psychophysics; sensory coding; signal detection theory; visual perception; auditory perception; somatosensation; multisensory integration; perceptual organization; perceptual learning; Bayesian perception; predictive processing; active perception; Artificial Era
Perception as a Scientific Object
Perception is the set of processes by which an organism obtains, transforms, organizes, and uses information about its body and environment through sensory activity. A stimulus is a physically describable event; sensory transduction converts part of its energy into neural activity; perceptual organization relates signals across space, time, and modality; an observer's report or action provides measurable evidence about the resulting estimate. These operations can be studied jointly while retaining distinctions among sensation, perceptual representation, attention, judgment, and motor response. A scientific definition therefore specifies an organism or system, a sampling environment, a sensory task, and the criteria by which its performance is evaluated. The same retinal pattern can admit different interpretations because illumination, occlusion, past experience, and intended action alter the information relevant to the observer. Kersten et al. (2004) provides a computational account of ambiguity that helps formulate these relations explicitly.
The central empirical problem is underdetermination: sensory measurements generally admit more than one possible state of the world. Receptor responses fluctuate, signals overlap, and physical causes may be hidden by occlusion or environmental noise. Perception nevertheless supports stable behavior because information is distributed over features, prior observations, other modalities, and active sampling. Stability does not imply exact reconstruction. A perception can be reliable for a behavioral question while inaccurate relative to a laboratory-defined physical quantity. The investigator must distinguish the environmental property being estimated from the observable used as a proxy for that property, and from the observer's decision policy. The statistical decision framework of Green & Swets (1966) makes this distinction especially concrete in uncertain detection tasks. Its logic extends to the interpretation of reports, discrimination responses, and confidence.
The present research dossier synthesizes original psychophysical experiments, sensory physiology, formal models, methodological research, and comparative behavioral findings. Its scope is the general organization of perception across modalities, levels of explanation, and active contexts. Specialized mechanisms such as speech perception, motion processing, and perceptual learning appear where they illuminate a common principle; each also has a distinct experimental literature deserving focused treatment elsewhere. A successful integrative account must connect calibrated physical stimulation to behavior through identifiable intermediate operations, state what evidence distinguishes alternatives, and acknowledge the circumstances in which a proposed mechanism ceases to predict performance. Marr (1982/2010) supplies the foundational distinction among a computational problem, the representations and algorithms used to solve it, and their physical implementation.
The Historical Establishment of Perceptual Science
Perception became a quantitatively tractable psychological subject through the development of controlled stimulation and repeatable reports. Earlier philosophical investigations of appearance and knowledge supplied questions that experimental psychologists reformulated as measurable relations. Weber's work on discrimination and Fechner's subsequent synthesis made thresholds and just-noticeable differences central instruments of inquiry. Fechner's two-volume Elemente der Psychophysik, published in 1860, explicitly organized the relationship between physical magnitude and experienced intensity as an experimental research program Fechner (1860). Later psychophysics broadened these relations into multiple tasks rather than a single universal stimulus-to-sensation function. The foundational achievement remains methodological: an observer's performance can be studied systematically by manipulating the physical causes of sensory evidence and estimating the form of a response function.
Twentieth-century research clarified that perception depends on the structure of the observer as well as stimulation. Work on visual thresholds examined the limitations imposed by photons, optics, and neural variability; the experiments of Hecht et al. (1942) helped establish how very weak light can be detected under controlled conditions. Studies of perceptual constancy and Gestalt grouping investigated why distributed sensations are experienced as organized objects and events. Gibson (1979) subsequently emphasized the information available to moving observers in ordinary environments, including optic flow and the opportunities for action afforded by surfaces. These different traditions drew attention to controlled signal measurement, structural organization, and ecological sampling. They remain historically significant because each exposes a different evidentiary problem.
Neuroscience introduced direct characterization of receptive fields and hierarchical processing. The original electrophysiological experiments of Hubel & Wiesel (1962) documented properties of neurons in cat visual cortex, providing evidence for selective responses to oriented features and binocular inputs. These data influenced later models of visual coding, although neuronal selectivity does not by itself prove a particular conscious percept. Formal computational analysis developed alongside work on attention, probabilistic inference, and multisensory integration. Contemporary research returns to the original psychophysical demand that mechanisms predict measured behavior, while expanding measurement to gaze, neural activity, naturalistic scenes, and developmental change. Historical theories are interpreted here as research programs with distinctive predictions rather than as successive replacements in a simple chronology.
Stimulus, Receptor, and Sensory Transduction
A perceptual experiment begins with a physically specified stimulus that can be reproduced and calibrated. In vision, relevant variables include spectral distribution, luminance, contrast, retinal size, orientation, and temporal modulation. In audition, pressure variation, frequency content, duration, and spatial configuration matter. Touch involves contact forces, vibration, temperature, and body position; chemosensation involves volatile or dissolved substances and receptor interactions. These properties determine the information potentially available to the system, while receptor dynamics determine which differences are encoded at a particular moment. Experiments on light detection Hecht et al. (1942) show why stimulus specification and adaptation state are indispensable: a nominally identical physical input can generate different sensitivity across testing conditions.
Transduction is a series of physical and biochemical transformations that convert external or bodily events into activity usable by neural circuits. Receptors vary in dynamic range, adaptation, baseline activity, temporal precision, and the geometry of their sampling fields. Encoding is consequently selective. Receptive-field research such as Hubel & Wiesel (1962) demonstrates that selectivity can be investigated experimentally through changes in firing associated with particular visual patterns. Orientation and spatial-frequency tuning in human perception were also investigated psychophysically through the adaptation and sensitivity experiments of Blakemore & Campbell (1969). The convergence of behavioral and physiological evidence supports specialized encoding operations while leaving questions about the precise mapping between individual neural responses and experienced structure.
Sensory coding must be understood as context-dependent measurement. Optical focus, eye movements, background illumination, preceding stimulation, and nearby patterns all affect the effective input. A measurement made at the display cannot substitute for an estimate of what reaches the receptors. Researchers therefore use calibrated displays, controlled sound delivery, standardized viewing distance, and records of adaptation when those variables matter. Stimulus-response functions must be interpreted within the tested conditions rather than extrapolated to unmeasured environments. The inferential gap between calibrated input and observer report is a central reason for combining psychophysics with physiology, modeling, and repeated tasks instead of treating a single threshold as a complete description of perception.
Thresholds and the Classical Psychophysical Program
An absolute threshold summarizes a criterion level of detection under a specified task, whereas a difference threshold describes the magnitude of a change that observers discriminate at a given performance level. The familiar Weber fraction expresses a proportional relationship between a just-noticeable increment and a baseline stimulus in ranges where that approximation holds. Fechner's formulation associated systematic changes in stimulus intensity with changes in subjective magnitude Fechner (1860). The empirical success of threshold methods lies in their ability to produce repeatable functions and comparisons. Their scientific interpretation requires specification of response alternatives, attention, adaptation, uncertainty, and the fitting criterion. Thresholds are estimates conditional on an experimental procedure, not context-free properties attached permanently to a sensory organ.
Magnitude estimation offered a different measurement route: observers assign numbers to perceived intensity rather than only detecting or discriminating change. Stevens (1957) presented the power law as a broad account of relations between physical stimulus magnitude and subjective estimates across sensory continua. This framework made variation between modalities and tasks experimentally accessible. Yet numerical judgments depend on instructions, anchors, available response scales, and the range of presented stimuli. Evaluating a psychophysical law therefore requires testing whether a functional form predicts new observations under explicit conditions. Competing equations can fit a restricted range similarly; comparison becomes more informative when models make divergent predictions at the boundaries or across tasks.
Contemporary psychophysics uses threshold and scaling experiments as complementary operations. A forced-choice discrimination task emphasizes comparative information, whereas a magnitude-rating task involves numerical mapping and response conventions. The result of either can be informative when its parameters and assumptions are transparent. The studies by Wichmann & Hill (2001a) and Wichmann & Hill (2001b) show that fitting, sampling, and uncertainty estimation can change what is legitimately concluded from psychometric data. Trials near the steep region of a psychometric curve often provide high information about its location, but exclusive concentration near one point may conceal lapses and asymptotic errors. Scientific synthesis therefore treats functional estimates as fitted evidence requiring validation.
Signal Detection and Perceptual Decision
Signal detection theory separates the discriminability of sensory evidence from a decision criterion controlling whether an observer reports that a signal is present. In a yes-no task, hits, misses, false alarms, and correct rejections are distinct outcomes. Under specific distributional assumptions, sensitivity may be expressed by d-prime, while criterion measures summarize a response tendency Green & Swets (1966). This separation addresses an old weakness of classical threshold concepts: changes in willingness to respond can mimic changes in sensitivity. Experimental manipulations of signal probability, payoff, and instruction help test whether the observer shifts criterion, sensitivity, or both. The choice of model matters because unequal variances and dependent observations complicate the simplest Gaussian summary.
Receiver-operating-characteristic curves describe discrimination across different decision criteria, offering more information than a single accuracy figure. Forced-choice paradigms reduce some forms of criterion dependence but introduce their own assumptions about comparison, memory, interval order, and independence. A response-time distribution may reveal changes in speed-accuracy strategy, and confidence ratings provide another imperfect window onto internal evidence. Distinct measures should not be combined as if they represented one directly observable magnitude. The framework of Green & Swets (1966) encourages researchers to distinguish latent evidence strength, policy, and the behavioral function used to recover them. That methodological distinction is relevant to visual, auditory, tactile, and artificial classification systems alike.
A complete detection study reports the signal and noise distributions or their operational proxies, number of trials, balancing procedure, task instructions, and uncertainty in fitted estimates. It also asks whether observers adjusted their policy over time and whether practice altered the sensory representation itself. Studies of learning and attention are especially vulnerable to conflating criterion and sensitivity. External validation through alternative tasks can reveal whether a reported perceptual improvement transfers beyond one response format. The central conclusion is methodological: detection is an inference from measurements under a decision rule, and the rule becomes a substantive part of the explanation.
Psychometric Functions and Adaptive Experimentation
Psychometric functions relate a continuously varied stimulus property to the probability of a defined behavioral response. Their slopes, asymptotes, and locations are interpretable only in relation to the task and model. In a two-alternative forced-choice task, the chance-performance level differs from that in a yes-no task; lapses and occasional guessing can affect the upper asymptote as well. Wichmann & Hill (2001a) studied parameter fitting, sampling, and goodness of fit, while Wichmann & Hill (2001b) examined bootstrap-based confidence intervals. These studies establish that the uncertainty attached to an estimated threshold can be substantial when trial counts are small or stimulus levels are poorly distributed. A credible report communicates uncertainty and the limits of model identifiability.
Adaptive staircases select subsequent stimulus levels using responses observed so far. Watson & Pelli (1983) developed QUEST, a Bayesian adaptive psychometric method that illustrates the efficiency of maintaining an updating probability distribution over a threshold parameter. Adaptive designs can reduce wasted trials relative to fixed stimulus grids, but they do not remove the need to check prior assumptions, lapse rates, model mismatch, and stopping rules. An informative design samples more than the transition region if slope and asymptotes must also be estimated. Reliability can be strengthened through repeated sessions and independent starting points, especially when subjects learn the task during the measurement itself.
Experimental software makes complex stimulus control reproducible when researchers preserve the implementation details. Brainard (1997) described the Psychophysics Toolbox, and Peirce et al. (2019) documented PsychoPy2 as infrastructure for behavioral experimentation. Such tools cannot guarantee photometric calibration or timing precision on arbitrary hardware. Researchers must validate their actual displays, audio chains, response devices, and event logging. Transparent code, materials, and trial-level data support inspection of stimuli and analyses. Sensitivity studies should report exclusion rules, counterbalancing, observer characteristics, and whether the tested stimulus range contained sufficient evidence to estimate the intended psychometric parameters.
Visual Coding of Contrast and Form
Visual perception uses patterns of contrast and spatial variation to extract contours, textures, shapes, and other informative features from the retinal image. The optics of the eye and the sampling of photoreceptors restrict the input; neural filters with different spatial and temporal properties provide selective descriptions of that input. Blakemore & Campbell (1969) used psychophysical adaptation to investigate sensitivity to orientation and spatial scale in human vision. Their results became part of a convergence between behavioral estimates of tuned channels and physiological evidence of selective responses. Hubel & Wiesel (1962) documented receptive-field organization in cat visual cortex, illustrating both the power and the species and method limits of a physiological model. Perceptual form requires relations among many local measurements.
Contrast sensitivity is defined by detection performance across spatial patterns, frequencies, luminance levels, and temporal conditions. Apparent visibility is affected by eccentricity, crowding, external noise, optical blur, and neighboring contours. A stimulus highly recognizable near fixation may become confusable when placed in peripheral vision with surrounding objects. Rosenholtz (2016) reviewed the abilities and limits of peripheral vision, making the distinction between isolated acuity and object recognition in clutter particularly important. Researchers therefore test conditions representative of the target behavior rather than assuming foveal laboratory performance describes ordinary scene perception. Apparent failures can reflect sampling limits, grouping competition, or task uncertainty.
Form perception involves segmentation: an observer must assign visible boundaries to objects, surfaces, or background regions. Similarity, proximity, continuity, common region, and closure can all structure a visual display, but their relative contributions change with context. A mechanism may operate through learned regularities, neural connectivity, depth cues, or object-based attention. The scientific task is to predict when a particular contour belongs to one object and when competing organizations become likely. Behavioral reports, eye movements, adaptation, and masking provide complementary evidence. No single image of an illusion establishes a mechanism; controlled changes in geometry and task permit competing explanatory accounts to be distinguished.
Perceptual Grouping and Object Segmentation
Perceptual organization concerns how elements become perceptually associated into coherent units. Neighboring edges, synchronized changes, shared motion, and compatible surface structure often contribute to grouping. Gestalt research historically emphasized relations within the field rather than reducing the whole percept to independent point sensations. A contemporary experimental analysis identifies grouping cues separately and tests how they interact in ambiguous displays. Gestalt principles describe systematic patterns, while proposed mechanisms must make predictions about timing, neural population responses, attention, and behavior. Feature-based search experiments Treisman & Gelade (1980) helped demonstrate that identifying separate attributes and selecting conjunctions can produce distinct patterns of performance.
The same physical fragment can support several object interpretations depending on boundaries, surrounding texture, and likely occlusion. Partial contours may be completed perceptually, and figure-ground assignment can reverse when cues are balanced. A rigorous interpretation distinguishes the geometry of the stimulus, the structure reported by observers, and the computational operation posited to relate them. Kersten et al. (2004) framed object interpretation as inference under uncertain sensory evidence, offering one way to formalize competing scene interpretations. Bayesian models can explain some grouping effects by incorporating regularities of natural environments, but the correctness of a fitted probability model depends on how its assumptions are tested across displays.
Grouping also interacts with task goals and temporal dynamics. An observer searching for a face may favor configurations and diagnostic features differently from a person comparing local edge orientations. Theories of attention propose selective mechanisms for resolving competition, though attention can be influenced by perceptual organization as well as influence it. Treisman & Gelade (1980) remains a landmark experimental theory of feature conjunctions, but subsequent research across display types and response requirements complicates universal claims about serial versus parallel processing. The strongest conclusions are those anchored in measured search functions, independent manipulations of grouping cues, and comparisons with alternative explanations.
Motion, Temporal Change, and Optic Flow
Motion perception uses changes in sensory structure to estimate displacement, direction, speed, and object or self-motion. A succession of static images can produce movement experience under suitable temporal and spatial conditions; local receptive fields cannot always distinguish motion direction from changes in orientation. Adelson & Bergen (1985) presented spatiotemporal energy models, in which filters operating over space and time provide an explicit computational account of motion-sensitive responses. This approach connects physical transformations of image sequences to measurable tuning properties. It also clarifies why apparent motion and motion aftereffects offer mechanistic evidence rather than serving merely as visual curiosities. Conditions involving contrast, speed, and ambiguous patterns can discriminate among models.
Moving observers produce structured optic flow even when objects remain stationary in the world. Gibson (1979) examined how this information supports locomotion and the perception of surfaces and layout. Perceptual research must therefore distinguish image motion caused by objects, the eyes, the head, or the observer's displacement. Vestibular and proprioceptive signals contribute to this discrimination. Experiments that fixate an observer in front of a display may isolate a visual computation yet omit important information normally present during movement. Ecological and laboratory approaches complement one another when a model predicts performance across controlled and active-sampling conditions.
The brain must integrate noisy motion samples across space and time. Motion integration can yield coherent direction from local signals, although opposite directions, transparent motion layers, and moving backgrounds create ambiguities. Auditory signals may alter visual interpretation: Sekuler et al. (1997) reported that sound changes perception in a visual motion display. Such demonstrations provide a route to studying temporal binding across modalities, but precise claims depend on stimulus timing, congruency, and alternatives. A scientific account includes how motion evidence is encoded, how competing sources are combined, and what a specific behavioral response reveals about the inferred movement.
Depth, Distance, and Spatial Constancy
Spatial perception estimates three-dimensional layout from two-dimensional retinal projections together with multiple additional signals. Binocular disparity, ocular convergence, motion parallax, perspective, texture gradients, shading, occlusion, and familiar size contribute under different viewing conditions. No single cue provides reliable metric distance in every environment. Observer motion can turn spatial ambiguity into a more informative sequence by revealing parallax and changing occlusion boundaries. Gibson (1979) emphasized the significance of environmental structure and self-produced movement, while Kersten et al. (2004) analyzed how uncertain cues can be interpreted through probabilistic assumptions about scenes.
Perceptual constancy describes relative stability in apparent object properties despite changes in the sensory input. An object viewed at several distances produces different retinal sizes; color appearance changes with illumination; perceived shape may remain stable across viewpoint. Constancy requires an account of how the visual system uses contextual evidence about surfaces, lighting, and geometry. It is best tested by measuring judgments across manipulated conditions, because a stable label, a matching response, and the underlying perceptual estimate need not coincide. Differences between geometric accuracy and functional effectiveness are especially relevant in complex natural scenes.
Reliability-based cue integration offers a specific explanation for how information sources can be combined. When errors are approximately independent and distributions suitably characterized, estimates with lower variance receive greater weight. The original visual–haptic study by Ernst & Banks (2002) demonstrated behavior compatible with near-optimal integration under its tested conditions. This result does not imply that every spatial percept always follows statistically optimal combination: cue conflict, causal uncertainty, attention, developmental change, and inaccurate estimates of reliability all matter. Spatial perception is a productive domain for contrasting parameterized normative models with alternative heuristics through held-out conditions.
Auditory Perception and Scene Organization
Auditory perception begins with pressure fluctuations transformed into neural responses that preserve information about spectral composition and temporal structure. The auditory system must extract sources, locations, rhythms, and meaningful patterns even when signals overlap. A sound's amplitude is a physical quantity, whereas perceived loudness depends on frequency, duration, context, and listener state. Masking and discrimination tasks can distinguish frequency selectivity from response strategy when stimuli and competing explanations are specified. Bregman (1990) assembled a foundational account of auditory scene analysis: the perceptual organization of sound into streams and sources, including grouping across time. Its core question concerns how a listener identifies persistent environmental causes within complex mixtures.
Temporal coherence, spectral proximity, harmonic relations, onset timing, and spatial differences may all support auditory grouping. A sequence of alternating tones can be perceived as one stream or multiple streams depending on rate and frequency separation. The measurable outcomes include identification, detection, subjective streaming reports, and performance while attention is directed elsewhere. These outcomes can dissociate, so evidence about a listener's conscious report should be interpreted alongside objective performance. The same acoustic mixture can be organized differently under distinct task demands. Auditory grouping illustrates that perceptual organization must integrate incoming evidence with temporal expectations rather than evaluating each moment in isolation.
Auditory and visual information also constrain each other. McGurk & MacDonald (1976) described the influence of incongruent lip movements on heard speech, showing that audible content can be perceptually altered by visual articulatory information. The magnitude and direction of such effects depend on the speaker, listener, stimulus quality, and experimental design; the finding supports crossmodal influence without proving that all language perception is obligatorily fused. A complete account of auditory perception therefore connects peripheral coding, streaming, selective attention, learning, and multisensory context. Speech-specific phonological mechanisms are an important neighboring research object rather than a replacement for the wider auditory science.
Somatosensation, Haptics, and Bodily Localization
Somatosensory perception comprises processes that estimate contact, pressure, vibration, temperature, pain-related signals, and bodily configuration. Haptic perception actively combines cutaneous feedback with proprioception and movement. A finger exploring an edge samples a sequence of contacts rather than receiving a single static image of an object. Performance depends on receptor sensitivity, the spatial pattern of contact, movement velocity, and expectations about the material or shape being explored. Investigators use controlled tactile patterns, vibrotactile discrimination, force sensors, and tasks requiring object identification. The neural events associated with contact are only part of the information used in a behavioral judgment: posture and exploratory action determine which sensory evidence is available.
Touch and vision provide a particularly clear experimental test of multisensory estimation. In the study by Ernst & Banks (2002), judgments combining visual and haptic cues to object size were consistent with weighting based on relative reliability. This finding illustrates how a precise computational model can predict shifts in perceptual estimates when one cue becomes noisier. The interpretation assumes relevant conditions about cue dependence, uncertainty, and the possibility that observations concern a common object. Touch can dominate some judgments, and vision others, depending on the available information. Scientists must distinguish evidence for a weighted estimate from stronger claims that a particular neural computation or awareness mechanism has been uniquely identified.
Bodily localization additionally raises issues of reference frames. A touch occurs at a location on the skin but must often be coordinated with the changing position of the limb, gaze, and the environment. Posture, movement, and context can alter localization accuracy and confidence. A rigorous test separates sensory discrimination from coordinate transformation and decision demands, and can use perturbations of proprioceptive or visual cues to identify contributions. These distinctions are crucial for rehabilitation, tactile interfaces, and comparative models of embodied sensing. The same design principles applied to visual and auditory detection—calibration, task clarity, uncertainty, and replication—remain indispensable here.
Olfaction, Taste, and Chemical Information
Olfaction and gustation provide information about molecules through receptor systems whose natural inputs vary in concentration, mixture composition, airflow, temperature, and contact conditions. An odor is a complex physical mixture rather than a single subjective label. Taste qualities, olfactory identity, oral somatosensation, and retronasal smell jointly contribute to flavor judgments. Perceptual experiments use controlled concentrations, discrimination tasks, threshold estimates, identification tests, and mixture comparisons. A scientific report must document delivery method, adaptation, the order of trials, and the vocabulary supplied to the observer. Chemical perception illustrates how stimulus sampling and subjective categorization are intertwined: category names can vary across languages and experience even when discrimination performance is similar.
Mixtures can produce suppression, enhancement, masking, or qualitative changes in perceived identity depending on their components and testing conditions. Learned associations with foods, hazards, and environments may influence recognition and pleasantness. These effects require separate assessment of detection, intensity judgment, familiarity, and hedonic response. A participant's dislike of an odor does not establish lower sensory sensitivity, while poor naming can coexist with successful recognition. Classical psychophysical distinctions remain useful: Stevens (1957) provides a historical framework for magnitude judgments, and Wichmann & Hill (2001a) supplies modern reminders about fitted response functions. Researchers should avoid inferring a single perceptual deficit from a composite identification score without examining the component operations.
Chemical senses also reveal strong interactions between perception and action. Sniffing modifies odor delivery; chewing and swallowing change concentrations reaching receptors; approach or avoidance influences further exposure. Ecological interpretations emphasize the sampling process and its consequences for the environment Gibson (1979). Physiological accounts explain receptor and central processing, while behavioral studies show how learned relevance and motivation alter task performance. Comparative research requires careful translation of the task across species and avoids assuming that human verbal intensity ratings are universal measures of perceptual organization. The overall lesson is that perception is an experimentally accessible transaction between a sensing system and the structure it samples.
Selective Attention and Perceptual Access
Attention alters the allocation of processing across competing stimuli, locations, features, and tasks. It can improve sensitivity to relevant information, change the spatial resolution of processing, and influence which objects reach explicit report. Treisman & Gelade (1980) proposed feature integration theory and tested its predictions through search, texture, localization, and identification paradigms. The studies established a productive experimental distinction between detecting a simple feature and selecting conjunctions under competition. Contemporary interpretations distinguish attentional selection, encoding, decision, and response; a change in search slope alone cannot uniquely specify the underlying algorithm. The relevant outcome may depend on display size, similarity, eye movements, and whether the observer has time to sample sequentially.
Attention can have mixed effects on perceptual performance. Yeshurun & Carrasco (1998) reported experimental conditions in which attention improved or impaired performance in a way interpreted through enhanced spatial resolution. Increased resolution may help when a task requires fine differentiation but interfere when a coarser representation is more informative. This finding illustrates the inadequacy of equating attention with a uniform improvement in every perceptual dimension. Valid conclusions require attention cues that alter selection while keeping sensory stimulation comparable, along with independent measures of task accuracy and response policy. Investigators also need to examine whether observed effects reflect exogenous orienting, expectations, or strategic adjustments rather than assuming an undifferentiated attentional resource.
Perceptual access further involves the relation between attention and awareness. Simons & Chabris (1999) studied sustained inattentional blindness during dynamic events and documented failures to report unexpected information when attention was engaged elsewhere. Such failures establish limits on report and recognition under task conditions; they do not provide a direct census of every representation that might have been processed unconsciously. Experimental controls for expectation, memory, response format, and awareness assessment are essential. Attention, perceptual organization, and consciousness remain related research objects whose boundaries become clear only when operational definitions and methods are specified independently.
Multisensory Binding and Crossmodal Influence
Natural environments provide many simultaneous cues about the same event. A moving face supplies visual articulatory information while speech produces sound; a grasped object supplies shape information through both touch and vision. The research problem is to determine when signals are combined, which is given greater weight, and when the system treats them as separate causes. Ernst & Banks (2002) provides original evidence consistent with reliability-dependent integration of vision and haptics. Alais & Burr (2004) investigated the ventriloquist effect and reported behavior compatible with near-optimal audiovisual integration in its tested setting. These studies support formal comparison between the variability of individual cues and the outcome under combined stimulation.
Crossmodal effects are not restricted to judgments with explicit sensory weights. McGurk & MacDonald (1976) demonstrated that visible articulation can change an auditory speech judgment; Sekuler et al. (1997) showed that sound can alter a visual motion percept. The coexistence of these findings establishes that perceptual systems exchange information across modalities, but it does not entail a single fusion algorithm for speech, motion, and object size. Spatial and temporal coincidence, learned correspondence, reliability, and task relevance may each affect grouping. Researchers must test whether two signals are perceived as originating from one event before predicting a fully integrated estimate.
Laboratory integration studies gain explanatory force through controlled cue conflict and uncertainty manipulations. If vision is made unreliable while haptic precision is held constant, a reliability-weighting model predicts a directional shift in estimates under its assumptions. A causal-inference model predicts additional dependence on whether the observer regards the cues as belonging to one cause. Distinguishing such accounts requires independent reliability estimates, model comparisons, confidence intervals, and out-of-sample predictions Knill & Pouget (2004). Multisensory perception thus offers a bridge from sensory coding to general theories of uncertainty, learning, and contextual interpretation.
Bayesian Models of Perceptual Inference
Bayesian approaches describe how uncertain sensory evidence can be combined with prior probabilities to estimate the causes of stimulation. The posterior distribution represents uncertainty after considering available observations, while the likelihood describes the relation between possible causes and measurements. Kersten et al. (2004) examined object perception as Bayesian inference, and Knill & Pouget (2004) discussed the representation of uncertainty in neural coding. A Bayesian model becomes empirically informative when its likelihood, prior, loss function, and response rule have distinguishable consequences. It is possible to fit many perceptual choices with flexible priors; identifying an explicit prediction in advance and testing it on independent displays provides stronger support than retrospective resemblance.
Normative optimality depends on its stated assumptions. Reliability weighting is optimal for a particular objective when sensory estimates have appropriate error distributions, their statistical dependence is handled correctly, and the decision cost is specified. Real observers may use approximate calculations or heuristics, possess inaccurate uncertainty estimates, or optimize a task different from the one specified by investigators. Ernst & Banks (2002) demonstrates how well-chosen controlled manipulations can test a concrete statistical prediction in combined visual and tactile perception. Its success in an experimental domain gives conditional support to a model, rather than establishing that every perceptual system implements a universal Bayesian computer.
Inference models make powerful comparative tools when contrasted with mechanistic alternatives. A feedforward filter model may explain a rapid perceptual effect through stimulus statistics without an explicit representation of prior beliefs. Conversely, contextual effects may require interactions that a restricted feedforward model cannot reproduce. A fair evaluation fits each model with comparable data, checks predictive performance on new conditions, and reports parameter identifiability. Models may share predictions while implying different neural mechanisms. The scientific synthesis therefore distinguishes the normative question of how an estimate could be improved from the psychological question of what operations actually produce the observed judgment.
Predictive Processing and Competing Neural Accounts
Predictive-processing theories propose that expectations influence sensory interpretation and, in some implementations, that neural populations encode differences between predictions and inputs. Their scope ranges from task-level descriptions to specific hypotheses about cortical circuits. Peelen et al. (2024) reviewed evidence that context and object–scene relations influence rapid visual recognition, including interactions between scene gist and object information. Such observations establish contextual modulation, but require additional experiments to identify whether a predictive-coding architecture, recurrent processing, attention, or feature-based feedforward mechanisms best explains the pattern. The distinction matters because several neural computations can create similar behavioral facilitation.
The evidence for a computational theory must be evaluated at the level at which that theory makes predictions. Hodson et al. (2024) reviewed direct tests of predictive coding and active inference, identifying modest support for predictive coding and highlighting explanations available from alternative feedforward models. Their analysis demonstrates why a broad explanatory vocabulary cannot substitute for discriminating tests. An experiment manipulating expectation can measure response changes without proving that the specific neuronal signal is a prediction error. Researchers can improve tests through multivariate neural measurements, controlled violations of expectations, explicit generative and alternative models, and independent data capable of rejecting at least one implementation.
Predictive accounts have potential relevance to attention, adaptation, hallucination-like reports in experimental paradigms, and multisensory causal interpretation, but the evidence differs by domain. Neurophysiological recordings, behavior, and computational simulations describe different observables. Felleman & Van Essen (1991) documented distributed hierarchical connections in primate cortex, an anatomical foundation for considering recurrent interactions; connectivity alone cannot establish the message passed by each pathway. Productive research thus integrates architectural feasibility with predictions about temporal order, neural representations, psychometric functions, and behavior in natural settings. Strong theoretical synthesis retains the explanatory reach and uncertainty of each specific hypothesis.
Ecological Perception and Active Sampling
Ecological approaches examine information that remains available to an organism while it moves through an environment. Gibson (1979) treated visual perception as the pickup of structured information, including optic flow and stable relations associated with surfaces, events, and affordances. An affordance concerns opportunities for action relative to the organism's capabilities and environmental layout. This perspective changes the experimental unit from a fixed retinal image to a temporally extended observer–environment interaction. It is especially relevant when head movement, locomotion, manipulation, and exploration change the evidence available for estimating an object's properties. The approach can be tested by measuring the consequences of active versus constrained viewing.
Action creates information. Moving around an occluded object can reveal hidden surfaces; rotating a grasped object supplies contact changes; orienting the head can improve acoustic localization. Experiments can vary an observer's control of sampling while holding aspects of the stimulus distribution comparable. A stronger active-perception explanation predicts how particular movements alter uncertainty and future performance rather than merely noticing that the observer moves. Laboratory trials with immobilized heads or fixed fixation isolate mechanisms with precision, but generalization to natural behavior depends on verifying the same relationships in active contexts. Ecological validity is thus an empirically testable property of a research design.
An ecological description and a computational analysis can be compared at different explanatory levels. The structured flow available to an observer defines useful information; a model may specify neural filters, probabilistic estimation, or learned policies that exploit it. Adelson & Bergen (1985) formalized motion-sensitive image computations, while Gibson (1979) emphasized the behaviorally available structure of motion during locomotion. The scientific question becomes which combinations of environmental variables, sensory operations, and actions predict performance across settings. Explanatory progress follows from explicit, independently tested connections between these levels, particularly where a laboratory model and a naturalistic prediction would otherwise diverge.
Adaptation, Learning, and Perceptual Plasticity
Adaptation changes perception as a consequence of recent stimulation. Exposure to a pattern can alter later sensitivity, apparent orientation, motion, or other judgments, reflecting changes in multiple possible stages of processing. Blakemore & Campbell (1969) used adaptation to investigate orientation and spatial-scale selectivity. A scientifically careful account distinguishes relatively short-term adaptation from durable learning and from shifts in response policy. To attribute an effect to sensory coding, researchers require controls for fatigue, decision criterion, practice, stimulus-specific exposure, and retention. The temporal course of the effect and its transfer to unfamiliar tasks help separate competing mechanisms.
Perceptual learning refers to relatively lasting changes in performance induced by perceptual experience. Watanabe et al. (2001) reported improvement in sensitivity to an exposed motion direction even when that direction was task-irrelevant and initially below awareness under the experimental conditions. This challenged simple accounts requiring explicit attention to every learned feature. Watanabe & Sasaki (2015) reviewed task-relevant and task-irrelevant visual perceptual learning and discussed models separating feature-based from task-based plasticity. The accumulated research suggests that improvements can involve changes in sensory representation, readout, decision rules, or combinations of these, with the balance depending on paradigm and training history.
A rigorous learning study needs a pretraining baseline, appropriate control conditions, an assessment of retest effects, and a retention or transfer test. Learning that is restricted to a trained orientation or location is informative about specificity but does not uniquely identify a cortical locus. Transfer to untrained stimuli can reflect generalized skills or learned policies as well as altered sensory coding. The methodological value of Watanabe & Sasaki (2015) lies in integrating apparently discordant findings through explicit distinctions among stages of plasticity. This enables predictions that can be examined through task changes, altered reward relevance, controlled stimulus exposure, and converging behavioral and neural measures.
Development, Individual Variation, and Cultural Context
Perceptual abilities develop through interactions among sensory maturation, movement, experience, learning, and social environments. The reliability of a sensory channel can change with age, environmental opportunity, and the demands of a task. Developmental studies must account for changing response abilities: an infant's looking preference and an adult's forced-choice judgment are not interchangeable indicators of perceptual experience. Comparing age groups demands appropriate psychophysical tasks and explicit measurement invariance. Sensory fusion, category learning, perceptual attention, and active exploration may follow partly different trajectories. A convincing developmental explanation specifies what changed—input quality, integration, experience, response strategy, or a combination—and includes longitudinal or carefully controlled comparative evidence.
Observers differ in acuity, adaptation, attention, learned categories, and the strategies used to sample information. Variability can reflect visual or auditory function, exposure, language, task familiarity, fatigue, and contextual expectations. Between-person variation is informative when measurement reliability and task equivalence have been established. Wichmann & Hill (2001b) illustrates the importance of interval estimates for psychometric parameters, a lesson amplified when comparing groups with different response consistency. An observed mean difference may coexist with broad overlap and may reverse when a task is reformulated. Researchers should report distributions and evaluate alternative explanations before translating group results into claims about general perceptual capacities.
Cultural conditions may shape learned recognition, salience, categorical descriptions, and opportunities for perceptual expertise. A cross-cultural experiment requires matched stimulus presentation, comprehensible instructions, equivalent response formats, and attention to familiarity with testing environments. Some perceptual constraints originate in shared sensory physiology, but a universal assertion must be supported by sufficiently diverse observations. Naturalistic studies add ecological breadth while making causal isolation more demanding. A robust synthesis therefore treats culture and development as testable influences on specific perceptual operations and avoids treating any single population as a default model for the whole of human perception.
Neural Systems and the Limits of Localization
Perceptual behavior depends on many interacting neural populations from peripheral receptors through subcortical and cortical circuits. A physiological response to an edge, tone, or movement reveals selectivity under the tested stimulation but does not identify the full process by which an observer recognizes an object. Hubel & Wiesel (1962) demonstrated receptive-field properties in the cat visual cortex, while Felleman & Van Essen (1991) examined distributed hierarchical connectivity in primate cerebral cortex. These works remain foundational for analyzing specialized responses and interactions across processing stages. Their interpretation should preserve the differences between species, experimental preparations, and the behavioral tasks to which particular measurements can reasonably be generalized.
Neural correlates can be constrained by lesion studies, stimulation, adaptation, functional imaging, and electrophysiology, but each method permits different causal inferences. A region activated during a discrimination task may participate in sensory coding, attention, working memory, response preparation, or feedback. Strong designs compare tasks with matched nonperceptual demands and manipulate the candidate mechanism where intervention is feasible. Neural decoding performance should not automatically be equated with conscious awareness or the format of a perceptual representation. Marr (1982/2010) clarifies why implementation-level evidence becomes more informative when aligned with an independently specified algorithmic account and a behavioral problem.
A multilayered explanation relates physical input, nervous-system activity, perceptual decisions, and task performance without collapsing their observables. Temporal information may distinguish early sensory coding from later selection or report. Anatomical connectivity provides constraints on plausible communication pathways, while model comparison asks which computations best predict trial-level outcomes. The resulting account remains open to revision as stimulus sets and methods diversify. Perceptual neuroscience gains its strongest generality through joint behavioral and physiological predictions tested on independently collected observations rather than through the isolated identification of a neural location.
Perceptual Illusions and Their Scientific Interpretation
Perceptual illusions are systematic differences between a measured physical property and an observer's report or behavior under specified conditions. They are useful because a controlled discrepancy can reveal how a perceptual process depends on context, organization, or prior experience. An illusion is not itself evidence of an inferior sensory system; the interpretation may be appropriate for a class of natural scenes while giving an unusual response to an artificial display. A scientific experiment specifies the relevant physical standard, response task, effect size, and competing accounts. Kersten et al. (2004) illustrates how ambiguous retinal images can generate different object interpretations under alternative scene assumptions.
Motion displays, audiovisual speech, and spatial cue conflicts illustrate several different families of perceptual discrepancy. Adelson & Bergen (1985) provided a computational basis for studying motion-energy responses; McGurk & MacDonald (1976) demonstrated visual alteration of speech judgments; Alais & Burr (2004) assessed audiovisual spatial integration under controlled uncertainty. These findings differ in their stimulus structure and measured outcomes. They should therefore be compared at the level of explanatory mechanisms, not grouped under a single label as proof of a universal perceptual shortcut. The critical question is which manipulation changes the predicted response and whether the same model continues to work under altered temporal or sensory conditions.
Apparent illusions may also result partly from task interpretation, memory demands, or response conventions. Trials with open-ended descriptions, forced choices, matching methods, and objective action outcomes can yield different measurements. Replication across methods provides evidence about the robustness of an effect, while individual variation may reveal strategy or experience dependence. Quantitative models become informative when they predict the direction and magnitude of a new discrepancy in advance. Illusions are best treated as controlled perturbations that expose the organization of perceptual processing and the conditions under which a given organization supports successful behavior.
Perception, Memory, and Conscious Report
Perception unfolds over time and interacts with prior learning, immediate retention, and the expectations carried into a scene. A judgment about what is present may require maintaining and comparing samples, especially in sequential discrimination. Memory limitations can therefore lower observed accuracy even when early sensory encoding remains effective. Experimental tasks should distinguish simultaneous comparison from delayed recognition and report the relevant interval durations. Sensory adaptation, priming, and perceptual learning have different timescales and evidentiary signatures. Watanabe & Sasaki (2015) illustrates how training-dependent performance changes can involve feature-level and task-level plasticity rather than a single homogeneous memory process.
Conscious perception is often operationalized through report, identification, confidence, or behavioral access, and these measures need not agree. The original inattentional-blindness experiments of Simons & Chabris (1999) showed failures to notice unexpected events under focused task demands. Their significance is a measurable dissociation between information physically present and information subsequently reported. Interpretive caution is required when claiming that an unreported event was wholly unprocessed: awareness tests have sensitivity limits, and indirect influences may remain. Designs combining objective discrimination, subjective confidence, and variations in attention are better suited to identifying which operations depend on reportability.
The same object may be recognized successfully while details of color, position, or temporal order are inaccurately remembered. Researchers can test these distinctions with short delays, unexpected recognition probes, and response criteria chosen to limit guessing. The perceptual experience inferred from verbal description is a scientific construct linked to evidence, not a direct record accessible independently of the observer's behavior. General accounts of perception should therefore retain separate operational measures for sensory precision, decision, working memory, and awareness. This distinction becomes especially important when comparing humans with systems that produce fluent descriptions without necessarily using the same sensory architecture.
Artificial Perception and Comparative Systems
Artificial systems can estimate object identity, scene structure, sound classes, and movement from camera, microphone, and other sensor data. Their observable success creates a comparative research opportunity: algorithms and biological observers may solve related tasks with different representations, training histories, and sampling policies. Marr (1982/2010) supplies a useful language for separating the computational objective from a particular implementation, while Knill & Pouget (2004) offers principles for specifying uncertainty in estimation. Comparative work begins by defining common tasks, physical input conditions, error measures, and generalization tests. Matched outcomes alone cannot establish that two systems employ the same mechanisms or have the same forms of conscious experience.
A controlled comparison might expose human observers and an artificial classifier to matched changes in blur, illumination, occlusion, or conflicting modalities. Accuracy, calibration, sensitivity to distribution shift, and active information gathering can then be evaluated separately. Artificial models may exploit correlations that differ from those used by biological observers, while human performance can reflect ecological experience absent from artificial training data. The comparative framework must document training exposure, test-set independence, preprocessing, and the extent to which both systems receive the same relevant evidence. Theoretical language about agency or perception should be anchored to operational characteristics of the investigated systems.
Within the Artificial Era, perceptual science contributes methods for studying nonbiological sensing, inference, and action without importing unmeasured psychological attributes. Sensor fusion, adaptive sampling, uncertainty monitoring, and continual learning are empirically accessible functional properties. Human–artificial relations add another dimension when a person relies on an artificial system to interpret images, audio, or an environment: the joint performance depends on trust calibration, error communication, and task allocation. These are research questions connecting perception to broader cognitive partnership. They should be studied through controlled interactions and transparent measures rather than treated as conclusions entailed by the presence of an advanced model.
Critical Evidence Appraisal and Open Problems
The strength of an empirical claim depends on the study design that supports it. Repeated psychophysical measurements are valuable when the stimulus calibration, observer population, trial structure, exclusions, analysis code, and uncertainty are documented. Wichmann & Hill (2001a) and Wichmann & Hill (2001b) show how model fitting and interval estimation shape the meaning of thresholds. Pelli (1997) and Brainard (1997) describe experimental software that helped standardize visual stimulus delivery, but the actual physical performance of displays still requires validation. A theoretically appealing effect can appear more stable than it is when sampling is narrow, conditions are selected after inspection, or the model has enough flexibility to accommodate many outcomes.
Distinct studies provide different grades of evidence. A randomized manipulation of contrast can establish a causal effect on discrimination performance within the experiment; a correlational association between cortical activity and perception has other interpretations. An optimal-integration fit can demonstrate behavioral compatibility with a statistical principle, while leaving several implementational mechanisms possible. Hodson et al. (2024) provides a recent example of critical scrutiny of predictive-coding claims in relation to competing accounts. Across laboratories, investigators need independently collected samples, preregistered primary contrasts when appropriate, shareable stimulus descriptions, and effect estimates that incorporate variation across observers and testing settings.
Central unresolved problems include how perception binds information across modalities and time; how naturalistic scenes alter task-level models; when uncertainty is represented explicitly; how attention changes perceptual encoding versus response; what drives learning and its transfer; and which neural measurements discriminate rival computational accounts. The relative contribution of culture, development, and active sampling requires experimental equivalence across populations. Comparative artificial systems introduce additional questions about differences between calibrated classification and biologically situated perception. Addressing these problems calls for discriminating predictions, shared benchmarks that resist overfitting, and evidence that spans sensory, behavioral, computational, and ecological scales.
Conclusion
Perception is an organized, adaptive process that links structured stimulation, selective sensory encoding, contextual estimation, and behavior. Its foundational experimental accomplishments include calibrated discrimination tasks, psychophysical functions, signal-detection methods, and physiological characterization of receptive fields. Historical and modern research demonstrates that the same sensory data may support multiple interpretations, and that the observer's action, experience, and task determine which information is available and relevant. Fechner (1860), Green & Swets (1966), and Hubel & Wiesel (1962) exemplify distinct stages in this scientific history: quantitative measurement, decision-theoretic analysis, and the investigation of sensory mechanisms. Their explanatory contributions remain useful when their methodological assumptions are made explicit.
A scientific synthesis of perception must coordinate evidence from vision, hearing, touch, chemical senses, multisensory tasks, learning, and neural systems. Ernst & Banks (2002) supports reliability-based sensory integration under its conditions; Watanabe et al. (2001) demonstrates a form of task-irrelevant perceptual learning; Simons & Chabris (1999) reveals the limits of report under attentional competition. These studies identify real experimental phenomena while leaving room for debate about the operations that generate them. Competing ecological, feedforward, recurrent, Bayesian, and predictive accounts can be made more informative by testing distinct consequences on independent data. The most durable conclusions are those supported by convergent measurement, carefully specified uncertainty, and clear separation of behavioral observation from explanatory hypothesis.
The future of perceptual research lies in linking rigorously controlled experiments to environments in which observers explore, move, learn, communicate, and cooperate with technological systems. This approach preserves the strengths of classical psychophysics while expanding the range of testable tasks and subjects. For Psychology by Bogdanova, the present dossier establishes Perception as one coherent research object shared across several Science areas. Specialized dossiers may develop particular sensory mechanisms and experimental traditions without duplicating its canonical identity. The accumulated research provides a foundation for understanding biological perception and for constructing comparable, explicitly operational studies of artificial sensing in the Artificial Era.
References
- Adelson, E. H., & Bergen, J. R. (1985). Spatiotemporal energy models for the perception of motion. Journal of the Optical Society of America A, 2(2), 284–299. https://doi.org/10.1364/JOSAA.2.000284
- Alais, D., & Burr, D. (2004). The ventriloquist effect results from near-optimal bimodal integration. Current Biology, 14(3), 257–262. https://doi.org/10.1016/j.cub.2004.01.029
- Blakemore, C., & Campbell, F. W. (1969). On the existence of neurones in the human visual system selectively sensitive to the orientation and size of retinal images. The Journal of Physiology, 203(1), 237–260. https://doi.org/10.1113/jphysiol.1969.sp008862
- Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10(4), 433–436. https://pubmed.ncbi.nlm.nih.gov/9176952/
- Bregman, A. S. (1990). Auditory scene analysis: The perceptual organization of sound. MIT Press. https://doi.org/10.7551/mitpress/1486.001.0001
- Ernst, M. O., & Banks, M. S. (2002). Humans integrate visual and haptic information in a statistically optimal fashion. Nature, 415(6870), 429–433. https://doi.org/10.1038/415429a
- Fechner, G. T. (1860). Elemente der Psychophysik (2 vols.). Breitkopf und Härtel. https://wellcomecollection.org/works/jewtj34d
- Felleman, D. J., & Van Essen, D. C. (1991). Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex, 1(1), 1–47. https://doi.org/10.1093/cercor/1.1.1-a
- Gibson, J. J. (1979). The ecological approach to visual perception. Houghton Mifflin. https://search.worldcat.org/title/4929611
- Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics. Wiley. https://search.worldcat.org/title/890266
- Hecht, S., Shlaer, S., & Pirenne, M. H. (1942). Energy, quanta, and vision. The Journal of General Physiology, 25(6), 819–840. https://doi.org/10.1085/jgp.25.6.819
- Hodson, R., Mehta, M., & Smith, R. (2024). The empirical status of predictive coding and active inference. Neuroscience & Biobehavioral Reviews, 157, Article 105473. https://doi.org/10.1016/j.neubiorev.2023.105473
- Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. The Journal of Physiology, 160(1), 106–154. https://doi.org/10.1113/jphysiol.1962.sp006837
- Kersten, D., Mamassian, P., & Yuille, A. (2004). Object perception as Bayesian inference. Annual Review of Psychology, 55, 271–304. https://doi.org/10.1146/annurev.psych.55.090902.142005
- Knill, D. C., & Pouget, A. (2004). The Bayesian brain: The role of uncertainty in neural coding and computation. Trends in Neurosciences, 27(12), 712–719. https://doi.org/10.1016/j.tins.2004.10.007
- Marr, D. (2010). Vision: A computational investigation into the human representation and processing of visual information (reissue of the 1982 edition). MIT Press. https://doi.org/10.7551/mitpress/9780262514620.001.0001
- McGurk, H., & MacDonald, J. (1976). Hearing lips and seeing voices. Nature, 264(5588), 746–748. https://doi.org/10.1038/264746a0
- 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
- Peirce, J., Gray, J. R., Simpson, S., MacAskill, M., Höchenberger, R., Sogo, H., Kastman, E., & Lindeløv, J. K. (2019). PsychoPy2: Experiments in behavior made easy. Behavior Research Methods, 51, 195–203. https://doi.org/10.3758/s13428-018-01193-y
- Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10(4), 437–442. https://pubmed.ncbi.nlm.nih.gov/9176953/
- Rosenholtz, R. (2016). Capabilities and limitations of peripheral vision. Annual Review of Vision Science, 2, 437–457. https://doi.org/10.1146/annurev-vision-082114-035733
- Sekuler, R., Sekuler, A. B., & Lau, R. (1997). Sound alters visual motion perception. Nature, 385, 308. https://doi.org/10.1038/385308a0
- Simons, D. J., & Chabris, C. F. (1999). Gorillas in our midst: Sustained inattentional blindness for dynamic events. Perception, 28(9), 1059–1074. https://doi.org/10.1068/p281059
- Stevens, S. S. (1957). On the psychophysical law. Psychological Review, 64(3), 153–181. https://doi.org/10.1037/h0046162
- Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1), 97–136. https://doi.org/10.1016/0010-0285(80)90005-5
- Watanabe, T., Náñez, J. E., & Sasaki, Y. (2001). Perceptual learning without perception. Nature, 413(6858), 844–848. https://doi.org/10.1038/35101601
- Watanabe, T., & Sasaki, Y. (2015). Perceptual learning: Toward a comprehensive theory. Annual Review of Psychology, 66, 197–221. https://doi.org/10.1146/annurev-psych-010814-015214
- Watson, A. B., & Pelli, D. G. (1983). QUEST: A Bayesian adaptive psychometric method. Perception & Psychophysics, 33(2), 113–120. https://doi.org/10.3758/BF03202828
- Wichmann, F. A., & Hill, N. J. (2001a). The psychometric function: I. Fitting, sampling, and goodness of fit. Perception & Psychophysics, 63(8), 1293–1313. https://doi.org/10.3758/BF03194544
- Wichmann, F. A., & Hill, N. J. (2001b). The psychometric function: II. Bootstrap-based confidence intervals and sampling. Perception & Psychophysics, 63(8), 1314–1329. https://doi.org/10.3758/BF03194545
- Yeshurun, Y., & Carrasco, M. (1998). Attention improves or impairs visual performance by enhancing spatial resolution. Nature, 396, 72–75. https://doi.org/10.1038/23936