1 Overview and key concepts
1.1 Definition of ambiguity in figures
Ambiguous figures are visual or conceptual representations for which the observer can form more than one meaningful interpretation. The ambiguity is not merely a matter of ignorance or missing knowledge; rather, the stimulus and the observer’s visual system together support multiple plausible percepts. In research contexts, ambiguous figures are valued because they make perceptual uncertainty observable in behavior, timing, and reported confidence.
1.2 Types of interpretive uncertainty
Interpretive uncertainty can emerge from several sources. Some stimuli provide incomplete information, leaving important features underdetermined. Others offer conflicting cues, such as contours that can be assigned to different structures. A third class arises from stimulus organization that supports switching between rival interpretations even when the physical input is constant. In all cases, the observer’s experience changes without a corresponding change in the stimulus.
1.3 Perceptual vs conceptual ambiguity
Perceptual ambiguity concerns how the visual system segments, groups, or identifies elements in the image. Conceptual ambiguity involves higher-level meanings—such as whether the same pattern is taken to represent one object rather than another—often influenced by learned categories and semantics. While these distinctions are analytically useful, real stimuli frequently combine both forms, with perception shaping what conceptual interpretation becomes accessible.
1.4 Related terms and common examples
Several terms frequently appear alongside ambiguous figures. “Bistable perception” describes situations where perception alternates between two stable interpretations. “Figure–ground ambiguity” refers to uncertainty about what counts as the figure (foreground) versus the background. Common examples include drawings that can be seen either one way or another, images with hidden outlines, and configurations where edges can be interpreted as boundaries of different objects.
2 Mechanisms of interpretation
2.1 Bottom-up visual cues
Bottom-up processing refers to the contribution of stimulus-driven signals, such as edge detection, contrast gradients, local motion (when present), and spatial grouping cues. In ambiguous figures, these cues may be compatible with multiple global organizations. When the evidence supporting different interpretations is approximately balanced, the observer’s percept becomes sensitive to subtle variations in signal quality and interpretation.
2.2 Top-down influences
Top-down influences include effects of task goals, prior experience, expectations, and semantic context. For example, if an observer is prompted to look for a particular object type, the visual system can weight features consistent with that category more strongly. This does not require that the stimulus change; the shift occurs in how the same input is interpreted and prioritized.
2.3 Attention and selection effects
Attention can determine which aspects of the stimulus are sampled and weighted. In many ambiguous figure tasks, the interpretation that becomes dominant often aligns with where attention is directed or what the observer is actively trying to verify. Selection effects are also reflected in how participants report or stabilize one percept when instructed to focus on specific details, such as one region or a particular contour.
2.4 Learning, expectations, and priors
Perceptual organization is shaped by statistical learning and the observer’s “priors,” meaning expectations about what shapes and structures typically co-occur. In ambiguous figures, priors can bias which interpretation feels more likely, especially when sensory evidence is weak or ambiguous. Over repeated exposure, participants may show changes in how readily certain percepts occur, suggesting that adaptation and learning alter interpretive preferences.
2.5 Competition and resolution dynamics
A common mechanistic view is that alternative interpretations compete for representation. Each candidate percept receives support from different feature groupings and contextual evidence. Resolution occurs when one interpretation sufficiently dominates the perceptual system, at which point the observer reports it as the current percept. Switching can happen when the balance of support changes due to internal noise, adaptation, or fluctuations in attention.
3 Classic categories of ambiguous figures
3.1 Reversible (bistable) figures
Reversible figures are stimuli that can flip between two stable interpretations while the input remains constant. The hallmark is alternation: the observer’s report changes over time between distinct percepts that are each locally plausible. Such figures are used to study the timing, persistence, and switching behavior associated with perceptual rivalry.
3.2 Multistable figures
Multistable figures support more than two distinct interpretations. Rather than only alternating between two states, the observer may cycle among several percepts, sometimes with different dominance durations. These stimuli are useful for examining how cognitive and perceptual systems handle a larger set of candidate hypotheses and whether some interpretations become preferentially stabilized.
3.3 Ambiguous figure–ground relationships
In figure–ground ambiguity, the question of “what is the figure” changes the perceived structure. Edges that could be taken as boundaries of the foreground may instead be interpreted as delineating background elements. The ambiguity often depends on cues such as closure, symmetry, and higher-level object expectations, which can tip the balance toward one assignment.
3.4 Illusory contours and hidden structure
Illusory contours appear when the visual system infers edges that are not explicitly drawn. Similarly, hidden structure can be revealed by completing contours or surfaces implied by sparse or fragmented information. These examples illustrate how perception constructs a coherent scene by integrating local cues into global interpretations, sometimes producing shapes that are not directly present in the physical stimulus.
3.5 Depiction ambiguity (e.g., shifts in object identity)
Depiction ambiguity involves uncertainty about what an image represents, such as whether a pattern depicts one object versus another, even if the low-level features remain constant. The observer’s identity-level interpretation can change as the visual system reorganizes features into different category-consistent structures. This category links ambiguous perception with meaning-making processes.
4 Perceptual psychology and experimental paradigms
4.1 Stimulus construction and control
Researchers typically design ambiguous figures to hold physical properties constant while manipulating interpretive variables. Control may include ensuring consistent contrast, spatial resolution, and presentation duration. Stimulus construction often requires careful attention to which cues are present (e.g., edge completeness, symmetry, and local contrast) because small changes can alter the balance between alternative interpretations.
4.2 Measuring interpretations over time
Because ambiguous perception often evolves within a trial, experiments commonly measure the trajectory of interpretation rather than only final reports. Participants may indicate the current percept continuously or at discrete time intervals, allowing researchers to quantify alternations and stability. This temporal emphasis distinguishes ambiguous-figure paradigms from standard classification tasks where decisions are typically instantaneous.
4.2.1 Reaction time and response logging
Reaction time measures can capture how quickly a participant arrives at an interpretation after stimulus onset or after a prompt to report. Response logging also includes event-based measures such as when participants press a key to indicate a switch. Interpreting reaction times requires care because reporting behavior may reflect both perceptual dynamics and motor or decision processes.
4.2.2 Dominance duration and alternation rate
Dominance duration refers to how long one interpretation remains primarily experienced. Alternation rate characterizes how frequently switches occur. These metrics can be analyzed across conditions to test how cue strength, instructions, or training influence perceptual persistence. Statistical models often account for individual variability and for distributions that may be skewed rather than normal.
4.3 Context manipulation and cue weighting
Context manipulations test how different sources of information contribute to interpretation. Researchers may change surrounding cues, provide semantic hints, or alter the task goal while keeping the ambiguous figure itself constant. Comparing outcomes across contexts reveals how the perceptual system reweights available evidence.
4.3.1 Instructions and task framing
Instructions can steer observers toward particular interpretations or encourage slower, more deliberative reporting. Task framing may also change what counts as “correct” or what evidence participants are asked to prioritize. Such manipulations are important because ambiguous figures are sensitive to goal-directed attention and to strategies that participants adopt.
4.3.2 Priming and adaptation
Priming refers to exposing participants to related stimuli before an ambiguous-figure trial, shifting subsequent interpretation tendencies. Adaptation involves repeated exposure that reduces sensitivity to certain features, which can make one percept more likely than another. These manipulations help identify how sensory history and learned associations affect perceptual organization.
4.4 Individual differences in susceptibility
Not all observers experience ambiguous figures with the same frequency or variability. Individual differences can reflect variation in attention control, perceptual learning, prior experience, and metacognitive reporting styles. Some studies find stable traits related to susceptibility to rivalry, while others suggest that training or recent exposure can shift performance within individuals.
5 Neural and cognitive accounts
5.1 Cortical processing and competitive pathways
Neural accounts often describe competitive interactions between candidate interpretations in visual cortical areas and their downstream partners. Signals corresponding to alternative percepts can exhibit differential activation, with dominance reflecting stronger or more sustained activity patterns. Evidence accumulation at the neural level can be influenced by adaptation and inhibitory interactions, providing a mechanistic basis for switching.
5.2 Predictive processing perspectives
Predictive processing frameworks treat perception as inference under uncertainty: the brain predicts sensory input and compares predictions with actual signals. Ambiguous figures then arise because multiple predictions can explain the input similarly well. Perceptual switching can occur when prediction errors and updating processes favor different hypotheses over time, guided by both sensory evidence and priors.
5.3 Decision and evidence accumulation models
Evidence accumulation models formalize how information is integrated over time until a decision threshold is reached. In ambiguous perception, alternative percepts can correspond to competing accumulators that race or interact. Dominance duration and switching frequency can be connected to parameters such as drift rate, threshold height, and noise, linking behavioral time courses to computational descriptions.
5.4 Role of consciousness and awareness
Conscious report typically tracks the interpretation that becomes dominant, but awareness may not perfectly match underlying neural representation. Some approaches distinguish between neural activity that changes due to processing and the moment when a participant becomes aware enough to report a switch. Investigating consciousness-related aspects of ambiguous figures can involve metacognitive measures, subjective confidence, and delayed-report paradigms.
6 Computational and engineering perspectives
6.1 Modeling ambiguous perception
Computational models aim to reproduce alternation and dominance patterns using principles of inference, competition, or constraint satisfaction. Models may represent multiple candidate interpretations and update their weights based on incoming evidence. The goal is not only to match final percepts but also to capture temporal dynamics such as switching probabilities and dominance distributions.
6.2 Stability, switching, and inference algorithms
Stability and switching can be modeled using dynamical systems that have multiple attractors—each attractor corresponds to a perceptual state. Switching can result from noise, adaptation, or changes in evidence. Inference algorithms may incorporate Bayesian updating, where posterior probabilities for interpretations fluctuate as evidence and prior beliefs change, leading to state transitions.
6.3 Computer vision challenges and opportunities
Ambiguous figures highlight challenges faced by computer vision systems that must infer 3D structure or object identity from 2D cues that do not uniquely determine a solution. Algorithms used for segmentation, depth estimation, or recognition may produce multiple plausible outputs in ambiguous regions. Studying ambiguous stimuli can motivate architectures that quantify uncertainty and represent multiple hypotheses rather than committing prematurely.
6.4 Generating ambiguous stimuli with constraints
Engineering approaches can generate or modify stimuli to achieve desired levels of ambiguity. Constraints can target cue balance, ensuring that competing interpretations are supported to similar degrees. Generation methods may rely on optimization over shape parameters, contour placement, or textures while enforcing physical and perceptual constraints so that the ambiguity is robust across viewers and display conditions.
7 Applications and uses in research
7.1 Probing perceptual inference
Ambiguous figures provide a controlled way to study how perception resolves underdetermination. Because multiple interpretations are available from the same input, researchers can test how sensory evidence is integrated with context and expectations to produce a single conscious experience at a given time.
7.2 Testing theories of attention and memory
These paradigms can be used to examine how attention modulates perceptual outcomes and how memory affects interpretation. For instance, manipulating cueing or instructional focus tests attentional weighting, while manipulating prior exposure or task-relevant knowledge assesses the contribution of memory traces to current perception.
7.3 Studying metacognition and confidence
When participants report not only which interpretation they see but also their confidence, ambiguous figures become tools for metacognitive research. Changes in confidence across dominance phases can reveal how subjective certainty relates to perceptual stability, and how people calibrate confidence under uncertainty.
7.4 Educational and outreach demonstrations
Because ambiguous figures are accessible and visually engaging, they are widely used in classrooms and public outreach. They provide an intuitive demonstration that perception is constructed rather than a direct copy of sensory input. When paired with discussion about experimental method and interpretation, they can also introduce core ideas about uncertainty and inference.
8 Methodological considerations and pitfalls
8.1 Stimulus reproducibility and calibration
Reproducibility depends strongly on display parameters and stimulus quality. Contrast differences, viewing distance, scaling, and compression artifacts can alter cue strength and shift dominance patterns. Calibration for luminance and careful control of image scaling help ensure that the intended ambiguity is preserved across sessions and equipment.
8.2 Demand characteristics and response bias
Participants may infer the study’s aims and adjust reporting accordingly. Instructions that emphasize one percept can create demand effects, increasing the likelihood of selecting that interpretation regardless of perceptual dynamics. Response bias can also arise from individual preferences for certain reporting actions, requiring careful experimental design and balanced procedures.
8.3 Visual factors: contrast, lighting, and viewing distance
Environmental and hardware factors shape perceived clarity. Lower contrast or unfavorable lighting can weaken sensory evidence and increase switching or uncertainty. Viewing distance affects effective spatial frequency content, potentially changing which edges or contours are salient. Accounting for these factors is crucial when comparing results across laboratories or devices.
8.4 Data interpretation and statistical approaches
Analyses must respect the temporal and distributional characteristics of dominance and switching data. Simple averages can obscure meaningful structure, such as heavy-tailed dominance durations or individual differences. Robust approaches often incorporate within-subject variability and appropriate distributional assumptions, and they separate perceptual effects from reporting or response-mechanism effects.
9 Practical notes for creating and testing ambiguous figures
9.1 Design principles for controlled ambiguity
Controlled ambiguity typically requires balancing competing cues so that multiple interpretations are plausible. Designers often use iterative pilot testing to find parameter ranges where neither interpretation is overwhelmingly favored. Consistency in line thickness, spacing, and noise levels helps maintain stable ambiguity across participants.
9.2 Safety and ethics in participant testing
Research with human participants should follow appropriate ethics review processes, including informed consent and the right to withdraw. While ambiguous figure tasks are generally low risk, researchers should consider accessibility (e.g., for color-vision differences or visual impairments) and provide accommodations when needed.
9.3 Reporting standards for experiments
Transparent reporting supports replication. Key details include stimulus specifications (resolution, contrast, size), presentation timing, response method, participant screening criteria, and analysis choices. Reporting should also describe instructions and any manipulations of context, priming, or task framing, since these strongly influence interpretation behavior.
9.4 Open datasets, repositories, and sharing formats
Sharing stimuli and data can accelerate progress and enable method comparison. Common practices include publishing stimulus files in standard image formats, documenting parameters needed for reconstruction, and providing scripts or metadata for experimental presentation. Open repositories also benefit cross-lab validation by allowing researchers to test robustness under different viewing conditions and participant groups.