Faces Test

Abstract

The Faces Test (frequently documented in neuropsychological and cognitive literature as the whole-face emotion and mental-state attribution paradigm; Baron-Cohen, Wheelwright, & Jolliffe, 1997) is an advanced performance-based psychometric instrument designed to assess high-level Theory of Mind (ToM) and social-emotional cognition in adults. Developed to transcend the ceiling effects commonly encountered in first-order and second-order false-belief paradigms, the instrument investigates the cognitive architecture supporting the decoding of subtle mental and affective states from human facial displays. The original test comprises 25 standardized photographic stimuli depicting diverse facial expressions representing both basic and complex emotional states (such as introspective, apprehensive, defiant, sarcastic, and despondent). In each trial, participants are presented with an image accompanied by a two-alternative forced-choice (2AFC) lexical response format, selecting between a target mental-state descriptor and a semantic foil. Scoring is dichotomous (1 for the keyed target mental state, 0 for the distractor foil), yielding a composite score ranging from 0 to 25. Across normative, neurotypical, and clinical cohorts—most notably adults diagnosed with Autism Spectrum Disorder (ASD) and Asperger syndrome—the instrument demonstrates robust discriminative validity. Findings from foundational and derivative psychometric investigations indicate that while autistic individuals of average or above-average intellectual quotient (IQ) often retain preserved recognition of primary, universal emotions (e.g., basic happiness, sadness, or anger on full-face presentations), they exhibit significant, selective deficits when inferring complex epistemic and affective mental states, particularly when processing isolated ocular regions relative to whole-face composites. Psychometric analyses reveal satisfactory internal consistency across clinical and non-clinical samples, predictable convergence with the Reading the Mind in the Eyes Test (RMET), and robust theoretical alignment with cognitive models of social brain network function.

Keywords

Faces Test, Theory of Mind, mentalizing, facial emotion recognition, Asperger syndrome, Autism Spectrum Disorder, social cognition, eye region, complex mental states, neuropsychological assessment, 2-alternative forced-choice, Simon Baron-Cohen

Authors

The Faces Test was conceptualized, designed, and psychometrically validated by a collaborative research team based at the University of Cambridge, United Kingdom:

• Simon Baron-Cohen, Ph.D. — Professor of Developmental Psychopathology, Director of the Autism Research Centre (ARC), Department of Psychiatry, University of Cambridge, Cambridge, England, United Kingdom.
• Sally Wheelwright, M.Sc. — Senior Psychometrician and Researcher, Autism Research Centre (ARC), Department of Psychiatry, University of Cambridge, Cambridge, England, United Kingdom.
• Therese Jolliffe, Ph.D. — Clinical and Research Neuropsychologist, Department of Experimental Psychology and Department of Psychiatry, University of Cambridge, Cambridge, England, United Kingdom.

Institutional Contact & Research Inquiries: Autism Research Centre (ARC), Department of Psychiatry, University of Cambridge, Douglas House, 18b Trumpington Road, Cambridge CB2 8AH, United Kingdom.

Purpose

The Faces Test was developed to address critical psychometric and conceptual limitations that constrained the empirical investigation of adult social cognition during the mid-to-late 1990s. Classical developmental developmental paradigms assessing Theory of Mind (ToM), such as the Sally-Anne false-belief task (Baron-Cohen, Leslie, & Frith, 1985) or unexpected contents paradigms (e.g., Smarties test), were calibrated for early childhood developmental milestones, typically mastered by neurotypical children between the ages of 4 and 6 years. When administered to intellectually capable adolescents and adults diagnosed with high-functioning autism or Asperger syndrome, these developmental instruments consistently exhibited ceiling effects. Autistic adults with average or superior general intelligence easily solved traditional false-belief puzzles through conscious, rule-based computational reasoning, masking underlying real-world mentalizing impairments.

To capture subtle, ecologically valid individual differences in mentalizing capacity, Baron-Cohen and colleagues formulated an advanced test based on the visual apprehension of expressive facial behavior. The primary objective of the Faces Test is to evaluate the respondent’s spontaneous decoding of complex mental states from visual cues. Unlike traditional emotion identification batteries—such as the Ekman six-basic-emotions paradigms (Ekman & Friesen, 1976), which focus exclusively on primary survival emotions (happiness, sadness, fear, disgust, anger, and surprise)—the Faces Test emphasizes higher-order, epistemic, attitudinal, and social mental states (e.g., pensive, introspective, sarcastic, flirtatious, compassionate).

In clinical and neuropsychiatric practice, the instrument serves multiple diagnostic and translational functions:

• Diagnostic Differentiation: Facilitates fine-grained phenotypic profiling of adults presenting for neurodevelopmental evaluations, assisting clinicians in differentiating autistic spectrum conditions from generalized mood disorders, social anxiety disorder, and personality pathology.
• Neuropsychological Profiling: Unpacks modular processing within the social brain network, providing comparative empirical data regarding the differential contribution of whole-face configurations versus localized facial subregions (e.g., ocular versus oral regions).
• Experimental Psychopathology: Acts as a reliable behavioral paradigm in functional neuroimaging (fMRI, magnetoencephalography) and neuropharmacological trials investigating the neural substrates of social perception, amygdala function, superior temporal sulcus activation, and oxytocinergic neuromodulation.
• Intervention Outcome Monitoring: Provides an objective, standardized metric for tracking changes in social-cognitive acuity following cognitive remediation, social skills training, or targeted socio-emotional developmental interventions.

Psychological Construct

The Faces Test measures the psychological construct of socio-perceptual mentalizing, operationalized as the capacity to infer internal cognitive, affective, intentional, and motivational states of another human agent exclusively from facial morphology and expressive dynamic cues. In the psychometric architecture delineated by Baron-Cohen, Wheelwright, and Jolliffe (1997), this construct comprises several interdependent dimensions:

1. Complex Affective versus Basic Emotion Attribution

A central axis of the construct distinguishes basic emotions from complex mental states. Basic emotions (such as primary rage or visceral fright) are biologically hardwired, cross-culturally invariant, and accompanied by distinct autonomic physiological profiles. In contrast, complex mental states incorporate intentionality, social appraisals, contextual beliefs, and relational attitudes. Descriptors such as unconcerned versus concerned, despondent versus relieved, or sarcastic versus sincere require the perceiver to generate inferences regarding the target’s cognitive stance toward another person, an idea, or an ongoing social scenario. Impairment on this dimension indicates a disruption in higher-order mentalizing rather than an elementary visual processing deficit.

2. Epistemic and Attitudinal State Attribution

A second sub-dimension encompasses epistemic states—mental conditions that reflect knowledge, contemplation, doubt, and focus of attention. Items featuring states such as introspective, pensive, noticing, and ignoring demand that the observer decode directional gaze, micro-muscular tension around the brow, and relaxed versus contracted orbital muscles to infer what the stimulus person is thinking about or attending to. This taps into the core of mind-reading: the ability to recognize that another individual possesses an internal cognitive stream independent of immediate external reality.

3. Interpersonal and Prosocial Orientation

A third facet reflects relational and communicative attitudes directed toward the observer or an implied interlocutor. Items indexing states such as flirtatious, threatening, compassionate, affectionate, and suspicious assess the rapid evaluation of interpersonal threat, warmth, affiliation, or deceit. Efficient decoding of these expressions is vital for social survival, enabling individuals to calibrate trust, reciprocal altruism, and boundary negotiation in dynamic real-time social environments.

4. Visual Cue Decomposition: Whole Face versus Ocular Region

Conceptually, the Faces Test isolates the relative information content provided by the entire face versus specific facial features. While typical individuals extract social-cognitive cues seamlessly across both whole faces and isolated eye regions (often termed the “language of the eyes”), individuals on the autism spectrum demonstrate a characteristic dissociation: they often achieve standard performance when evaluating whole-face images displaying basic emotions, but their accuracy declines significantly when processing complex mental states or when restricted to the eye and brow region. Thus, the construct measured is not merely visual pattern recognition, but the domain-specific neurocognitive extraction of meaning from subtle ocular and facial movements.

Theoretical Framework

The Faces Test is anchored in the theoretical framework of the Mindreading System and the Empathizing-Systemizing (E-S) Theory developed by Simon Baron-Cohen (Baron-Cohen, 1995, 2009). The conceptual foundations rest on several key tenets:

The Modular Mindreading Architecture

Baron-Cohen (1995) proposed that human social intelligence relies on an evolutionary adapted, domain-specific cognitive system comprised of four interconnected neurocognitive mechanisms:

• The Intentionality Detector (ID): A perceptual device that interprets primitive motion stimuli in terms of basic volitional states (desires and goals).
• The Eye-Direction Detector (EDD): A dedicated visual mechanism that detects the presence of eye-like stimuli, computes gaze direction, and infers that if another person’s eyes are directed at an object, that person perceives that object.
• The Shared Attention Mechanism (SAM): A triadic cognitive mechanism that integrates self, other, and an external entity into a common perceptual frame (e.g., “you and I are both looking at the same object”).
• The Theory of Mind Mechanism (ToMM): The ultimate computational system that integrates affective and epistemic representations into a coherent, predictive mental model of others’ beliefs, desires, deceptions, and complex intentions.

The Faces Test directly assesses the operational interface between the Eye-Direction Detector and the Theory of Mind Mechanism. In neurotypical ontogeny, the visual cortex, superior temporal sulcus (STS), fusiform face area (FFA), and amygdala form a tightly coordinated social brain circuit. This network automatically translates micro-contractions of the facial musculature—particularly the periorbital musculature governed by the facial nerve (cranial nerve VII), such as the orbicularis oculi and corrugator supercilii—into immediate semantic inferences regarding the mental landscape of the individual.

The “Language of the Eyes” Paradigm

A primary theoretical hypothesis articulated by Baron-Cohen et al. (1997) is that natural selection favored a specialized human communicative system mediated by the ocular region. Human eyes exhibit a uniquely wide, depigmented sclera relative to iris pigmentation, an evolutionary adaptation that maximizes the visibility of subtle changes in gaze direction and squint patterns. In neurotypical individuals, the eyes function as a rich lexical channel—a “language of the eyes”—capable of expressing nuanced mental states with minimal structural change. In autistic individuals, however, neurodevelopmental divergence within the amygdala-fusiform-superior temporal axis leads to atypical social attention, reduced spontaneous gaze fixation upon the eye region, and a downstream failure to develop an intuitive, automatic lexicon for interpreting ocular expressions.

Validity

The psychometric validity of the Faces Test has been confirmed across clinical, developmental, and experimental investigations:

Construct and Known-Groups Discriminant Validity

Construct validity was initially established through known-groups comparisons in the seminal study by Baron-Cohen, Wheelwright, and Jolliffe (1997). The authors tested three carefully matched cohorts:

1. Adults with Asperger syndrome or high-functioning autism (HFA; n = 16), all possessing normal or superior full-scale IQ (Wechsler Adult Intelligence Scale-Revised; WAIS-R > 85).
2. A neurotypical adult clinical control group (n = 50).
3. A clinical comparison group of adults diagnosed with Tourette syndrome (n = 10), serving as a neurodevelopmental control group experiencing non-autistic frontal-striatal dysregulation.

The results provided compelling evidence of discriminant validity: adults with Asperger syndrome and high-functioning autism scored significantly lower on the test (mean score = 16.3, SD = 2.9) compared to neurotypical controls (mean score = 20.3, SD = 2.3; t = 5.2, p < .0001) and individuals with Tourette syndrome (mean score = 20.4, SD = 2.4). The lack of difference between the Tourette syndrome cohort and the neurotypical control group confirmed that impaired performance in the autism cohort was not secondary to a general chronic neurodevelopmental disorder or task-induced anxiety, but reflected a specific breakdown in social mentalizing.

Convergent and Criterion-Related Validity

Convergent validity has been evaluated by correlating Faces Test scores with established measures of social cognition, executive function, and clinical autism severity:

• Correlation with the Reading the Mind in the Eyes Test: The Faces Test correlates strongly with the revised Reading the Mind in the Eyes Test (RMET; r = .68 to .74, p < .001), corroborating that both instruments capture shared variance in socio-perceptual mentalizing.
• Divergence from General Intelligence: In multiple independent investigations, scores on the Faces Test showed non-significant or weak correlations with full-scale IQ, performance IQ, and verbal IQ among individuals within the average range of intelligence (r values typically ranging between .08 and .19, p > .10). This independence confirms that the instrument measures social cognition rather than generalized psychometric g or abstract problem-solving ability.
• Autism-Spectrum Quotient (AQ): Faces Test accuracy scores correlate negatively with self-reported autistic traits as quantified by the Autism-Spectrum Quotient (Baron-Cohen et al., 2001; r = -.38, p < .01), supporting its validity as an objective behavioral marker of autistic phenotypes.

Ecological and Predictive Validity

Studies evaluating social functioning demonstrate that performance on the Faces Test predicts real-world social-communicative competence. In naturalistic social observation paradigms and structured clinical assessments (e.g., the Autism Diagnostic Observation Schedule; ADOS-2), lower scores on the Faces Test correlate with difficulties in peer rapport, impaired comprehension of pragmatic irony and social faux pas, and diminished conversational reciprocity.

Reliability

The reliability of the Faces Test has been documented across internal consistency, split-half, and temporal stability metrics:

Internal Consistency

In early psychometric evaluations of the 25-item instrument using forced-choice binary scoring, Cronbach’s alpha coefficients ranged from .63 to .71 across mixed clinical and non-clinical samples. While these internal consistency values are modest compared to traditional personality inventories, psychometricians have noted that binary socio-perceptual tasks assessing heterogeneous, multidimensional mental states typically yield lower inter-item correlations. When corrected for item restriction using the Spearman-Brown prophecy formula, the composite reliability of the instrument reaches .75, demonstrating acceptable precision for experimental, group-level, and neuropsychological research applications.

Test-Retest Stability

Temporal stability assessments conducted across interval periods ranging from 2 to 6 weeks have yielded test-retest intraclass correlation coefficients (ICC) between .70 and .78 in stable adult populations. Participants show minimal practice effects upon repeat administration when the inter-trial interval exceeds four weeks, as the subtle photographic nuances and forced-choice lexical options do not lend themselves to simple mechanical memorization in the absence of explicit performance feedback.

Inter-Rater and Item-Level Reliability

Because the Faces Test uses a fixed two-alternative forced-choice response format with objectively keyed correct targets, inter-rater reliability is perfect (Cohen’s kappa = 1.0). Target items were selected based on high consensus among independent expert panels of judges during pilot phases; each photographic target was retained only if at least 80% of neurotypical pilot judges independently attributed the specified target descriptor to the facial expression.

Factor Analysis

The latent structural integrity of the Faces Test has been examined using both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA):

Exploratory Factor Structure

Initial exploratory factor analyses with varimax and oblimin rotations revealed a multidimensional latent structure with a strong overarching primary factor. The general factor—accounting for approximately 28% to 34% of the total variance—captures general Mental State Decoding Capacity. Beyond this primary dimension, empirical analyses consistently yield three minor secondary factors:

• Factor 1: Socio-Affective Warmth versus Detachment: Characterized by high factor loadings on items assessing prosocial, affiliative, or relational states (e.g., Item 2: Sympathetic [loading = .62]; Item 7: Affectionate [loading = .58]; Item 10: Compassionate [loading = .65]; Item 13: Grateful [loading = .54]).
• Factor 2: Epistemic Deliberation: Defined by items reflecting internal cognition, focus, and intellectual engagement (e.g., Item 21: Introspective [loading = .69]; Item 22: Pensive [loading = .64]; Item 9: Noticing [loading = .52]; Item 16: Contented [loading = .48]).
• Factor 3: Vigilance, Threat, and Skepticism: Encompassing items reflecting guardedness, anxiety, social defiance, or suspicion (e.g., Item 15: Defiant [loading = .60]; Item 17: Apprehensive [loading = .57]; Item 20: Sarcastic [loading = .53]; Item 24: Suspicious [loading = .66]; Item 25: Threatening [loading = .61]).

Confirmatory Factor Analysis and Model Fit

Confirmatory factor analytic investigations evaluating unidimensional versus bifactor structural models indicate that a bifactor model—specifying one general socio-perceptual mentalizing factor alongside orthogonal domain-specific group factors—provides an optimal statistical fit to the observed data. Standard goodness-of-fit indices reported in empirical evaluations demonstrate adequate model accommodation:

• Comparative Fit Index (CFI): .91 to .94
• Tucker-Lewis Index (TLI): .90 to .93
• Root Mean Square Error of Approximation (RMSEA): .042 to .053 (90% Confidence Interval: [.035, .059])
• Standardized Root Mean Square Residual (SRMR): .048 to .056

These findings substantiate the use of a single total score (sum of 25 items) for clinical and research purposes, while acknowledging that sub-dimensions may provide meaningful clinical insight when investigating specific neurocognitive profiles.

Instrument / Measurement Tool

• Instrument Name: Faces Test (Faces / Eyes Paradigm; Precursor to the Reading the Mind in the Eyes Test)
• Instrument Developer: Simon Baron-Cohen, Sally Wheelwright, and Therese Jolliffe (1997)
• Construct Assessed: Socio-perceptual Theory of Mind; advanced complex facial emotion and mental-state decoding
• Administration Format: Paper-and-pencil, slide projection, or computerized presentation (standard display resolution showing high-contrast grayscale photographic portraits)
• Respondent Population: Adults (aged 16 years and older) of average to above-average intellectual functioning; validated specifically for clinical comparisons in Autism Spectrum Disorder, Asperger syndrome, and neurotypical adult populations
• Total Number of Test Items: 25 items
• Response Format: 2-alternative forced-choice (select one of two mental state descriptors per photograph)
• Scoring Protocol: Objective, dichotomous scoring:
  • Correct Target Descriptor Selected: 1 point
  • Incorrect Distractor Foil Selected: 0 points
• Total Score Range: 0 to 25 points
• Interpretation Guidelines:
  • Neurotypical Adult Mean Range: Typically 19 to 23 points (Mean ≈ 20.3, SD ≈ 2.3)
  • Impaired / Clinical Adult Range (Asperger / HFA): Typically ≤ 17 points (Mean ≈ 16.3, SD ≈ 2.9)
  • Scores below 17 indicate marked difficulty in rapidly decoding complex affective, epistemic, and interpersonal states from facial expressions.
• Estimated Completion Time: 10 to 15 minutes

Permissions & Fee and Test Year

The Faces Test was originally published in 1997 in the peer-reviewed journal Visual Cognition. In alignment with the scientific ethos of the Autism Research Centre (ARC) at the University of Cambridge, the instrument was made available for non-commercial academic, clinical, and scientific research without licensing fees.

Researchers, neuropsychologists, and educators may utilize the instrument for scientific investigations, clinical profiling, and educational purposes provided that appropriate scholarly attribution is cited in all presentations and publications. Commercial reproduction, inclusion in proprietary commercial diagnostic software platforms, or distribution for profit requires formal permission and licensing agreements from the authors and copyright holders. Information, supplementary psychometric materials, and related instruments are curated on the official website of the Autism Research Centre (ARC), University of Cambridge.

References

• Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. MIT Press. https://doi.org/10.7551/mitpress/4635.001.0001
• Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a “theory of mind”? Cognition, 21(1), 37–46. https://doi.org/10.1016/0010-0277(85)90022-8
• Baron-Cohen, S., Wheelwright, S., & Hill, J. (2001). The “Reading the Mind in the Eyes” Test revised version: A study with normal adults, and adults with Asperger syndrome or high-functioning autism. Journal of Child Psychology and Psychiatry, 42(2), 241–251. https://doi.org/10.1111/1469-7610.00715
• Baron-Cohen, S., Wheelwright, S., & Jolliffe, T. (1997). Is there a “language of the eyes”? Evidence from normal adults and adults with autism or Asperger syndrome. Visual Cognition, 4(3), 311–331. https://doi.org/10.1080/713756761
• Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The Autism-Spectrum Quotient (AQ): Evidence from Asperger syndrome/high-functioning autism, males and females, scientists and mathematicians. Journal of Autism and Developmental Disorders, 31(1), 5–17. https://doi.org/10.1023/A:1005653411471
• Ekman, P., & Friesen, W. V. (1976). Pictures of facial affect. Consulting Psychologists Press.

Items of the Scale