Reading the Mind in the Eyes Test - Revised (RMET-R)

Version: v1 (current)

An advanced theory of mind assessment measuring the ability to infer complex mental states from facial expressions in the eye region.

Overview

The Reading the Mind in the Eyes Test - Revised (RMET-R) is the updated version of the classic Mind in the Eyes task, with improved stimuli, more sophisticated mental state terms, and enhanced psychometric properties. Participants view photographs showing only the eye region of faces and select which complex mental state (from four options) best describes what the person is thinking or feeling.

Unlike basic emotion recognition tasks, the RMET-R requires nuanced social-cognitive inference about sophisticated mental states like "contemplative," "suspicious," "playful," "skeptical," or "thoughtful." The task taps into advanced theory of mind abilities essential for understanding others' intentions and navigating complex social situations.

The RMET-R is considered the gold-standard measure of adult-level mentalizing ability and is extensively used in autism research, social neuroscience, empathy studies, and clinical assessment of social cognition deficits.

Scientific Background

Classic Findings:

• Theory of Mind Sensitivity: Requires inferring internal mental states from minimal visual cues
• Autism Marker: Individuals with autism spectrum conditions score significantly lower (mean ~22/36 vs. ~26/36 in controls)
• Sex Differences: Women typically score 1-2 points higher than men on average
• Brain Correlates: Activates medial prefrontal cortex (mPFC), temporoparietal junction (TPJ), and superior temporal sulcus (STS)
• Empathy Relationship: Correlates moderately with self-reported empathy (r ~ 0.3-0.4)
• Stability: Test-retest reliability ~0.63, indicating moderate stability

Key Mechanisms:

• Mentalizing: Attributing mental states to others (beliefs, desires, intentions)
• Social Perception: Extracting social information from facial features
• Semantic Knowledge: Understanding nuanced emotion/cognition vocabulary
• Integrative Processing: Combining perceptual cues with social knowledge

Seminal Papers:

• Baron-Cohen, Wheelwright, Hill, Raste, & Plumb (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
• Baron-Cohen et al. (2015): Elevated fetal steroidogenic activity in autism
• Oakley, Brewer, Bird, & Catmur (2016): Theory of mind is not theory of emotion

Why Researchers Use This Task

1. Autism Research: Core social-cognitive assessment; identifies mentalizing deficits even in high-functioning individuals
2. Social Neuroscience: Map neural networks supporting theory of mind using fMRI/EEG
3. Empathy Studies: Correlate mentalizing ability with empathy, prosocial behavior, and relationship quality
4. Clinical Assessment: Evaluate social cognition in schizophrenia, personality disorders, social anxiety, depression
5. Developmental Research: Track development of advanced mentalizing across lifespan (childhood through aging)
6. Intervention Studies: Measure improvements in social cognition following therapy or training

Current Implementation Status

Fully Implemented:

• ✅ Built-in 36-item revised version (plus one practice item) with standardized eye stimuli
• ✅ Custom item sets with uploaded eye images
• ✅ Four-option forced-choice format
• ✅ Correct answer key for automated scoring
• ✅ Response time (latency) measurement per item
• ✅ Practice phase with correct/incorrect feedback
• ✅ Optional per-word translations of mental state terms
• ✅ Optional item-order and answer-order randomization
• ✅ Optional per-item or global response timeout

Not Yet Implemented:

• ❌ Child version (simplified terms and stimuli)
• ❌ Short-form version (10-12 items for screening)
• ❌ Adaptive administration based on performance

Configuration Parameters

Options

Parameter	Type	Default	Description
rmet_image_width_px	number	400	Display width of the eye image in pixels (form range 200-800; clamped to 100-1200 at runtime). A per-item image_width_px can override this.
rmet_response_timeout_ms	number	0	Max time per item in milliseconds (0 = unlimited). A per-item response_timeout_ms can override this.
rmet_randomize_order	boolean	false	Randomize item presentation order
rmet_randomize_answers	boolean	false	Randomize the order of the four answer options
rmet_practice_enabled	boolean	true	Run a practice phase (with feedback) before the main task

Instructions

Parameter	Type	Default	Description
main_instructions	html	''	Main task instructions (rich text)
hint_instructions	html	''	Hint/help text (rich text)
practice_instructions	html	''	Pre-practice instruction screen (shown only when practice is enabled)
trials_instructions	html	''	Pre-main instruction screen shown after practice (shown only when practice is enabled)

Items

Parameter	Type	Default	Description
rmet_items	array	built-in 36-item set	Main items. Each item: id, image_url, correct, options (4 words), optional block, optional response_timeout_ms. Falls back to the built-in RMET set when empty.
rmet_practice_items	array	single item P	Practice items (same shape; block defaults to 'practice').
rmet_translations	object	omitted	Optional map of English option word → translated label, emitted only when at least one non-blank translation exists.

The item editor exposes a spreadsheet with the columns: Image (dropdown of uploaded filenames), ID, Correct, Option 1, Option 2, Option 3, Option 4, Block, Timeout (ms).

Data Output

Markers

The task emits a trial_order marker at start (order = array of item ids). Per-trial onset markers are prefixed with the task slug (derived from the task name, or the kind rmet_revised by default): rmet_revised_trial_start for main trials and rmet_revised_practice_trial_start for practice trials. Additional markers include stimulus_shown, image_onset, stimulus_aoi, feedback_shown (practice only), and the centrally emitted response_recorded.

If a configured rmet_items or rmet_practice_items list is present but every row fails validation (missing id, image_url, or fewer than 4 options), the task falls back to the built-in item bank and emits rmet_revised_config_rejected (rejected_count, field, reason) so the fallback is visible in the data rather than silent.

{
  "type": "stimulus_shown",
  "ts": "2024-01-01T00:00:01.000Z",
  "hr": 1234.56,
  "data": {
    "trial_index": 1,
    "is_practice": false,
    "stimulus_id": "rmet_0_1",
    "image_url": "assets/RMET/image-001.jpg",
    "correct_answer": "playful",
    "options": ["playful", "comforting", "irritated", "bored"],
    "block": null
  }
}

Response Data

Emitted via recordResponse (trial grain):

{
  "trial_index": 1,
  "stimulus_id": "rmet_0_1",
  "source": "button",
  "image_url": "assets/RMET/image-001.jpg",
  "correct_answer": "playful",
  "options": ["playful", "comforting", "irritated", "bored"],
  "response_value": "playful",
  "choice_label": "playful",
  "response_correct": true,
  "responded": true,
  "is_practice": false,
  "block": null,
  "latency_ms": 4820
}

source is one of button, timeout, or moderator_advance. response_correct is null when the trial was not actually answered (timeout / moderator advance). choice_label is the translated label of the chosen word (falls back to the English word).

Summary Artifact

Emitted at task end as rmet_summary_<taskIndex>.json (only when at least one main trial was recorded and not in preview):

{
  "task_kind": "rmet_revised",
  "task_index": 0,
  "total_trials": 36,
  "overall": {
    "total": 36,
    "valid_responses": 36,
    "correct": 28,
    "score": 28,
    "total_possible": 36,
    "proportion_correct": 0.778,
    "mean_rt_ms": 4650,
    "median_rt_ms": 4120,
    "timeouts": 0
  },
  "breakdowns": {},
  "trials": [
    {
      "trial_index": 1,
      "stimulus_id": "rmet_0_1",
      "image_url": "assets/RMET/image-001.jpg",
      "correct_answer": "playful",
      "options": ["playful", "comforting", "irritated", "bored"],
      "response_value": "playful",
      "response_correct": true,
      "responded": true,
      "source": "button",
      "block": null,
      "latency_ms": 4820,
      "is_practice": false
    }
  ]
}

When practice is enabled and at least one practice trial was recorded, a parallel practice object (same overall/breakdowns/trials shape) is added.

Example Research Configurations

Standard Clinical/Research Administration

Items: Full 36-item revised version
Presentation: Randomized order
Timing: Self-paced (response timeout = 0)
Practice: Practice phase enabled (with feedback)
Scoring: Total correct out of 36
Analysis: Compare to normative data (mean ~26, SD ~3.5)

Screening Version (Time-Limited)

Items: 36 items
Timing: 10-second limit per item
Purpose: Reduce ceiling effects in high-functioning samples
Analysis: Speed-accuracy trade-off

Autism Research Protocol

Items: 36 items
Presentation: Randomized
Additional: Record eye-tracking to examine fixation patterns
Analysis: Compare ASD vs. controls on accuracy and viewing strategies

Intervention Study (Pre-Post)

Items: 36 items at baseline and follow-up
Order: Different randomization at each timepoint
Analysis: Change in total score after social skills training
Interpretation: Increase of 3+ points = meaningful improvement

Participant Experience

1. Instructions: The configured main instructions (rich text) describe the task, e.g. "You will see photographs of people's eyes. For each photo, choose which word best describes what the person is thinking or feeling."

2. Practice Phase (if rmet_practice_enabled):

   • Pre-practice instruction screen (practice_instructions)
   • For each practice item: view the eye photo, then choose among the four word buttons
   • Receive visual feedback after each choice (green check for correct, red X for incorrect; ~1 s)
   • The built-in default has a single practice item

3. Transition (if practice enabled): A pre-main instruction screen (trials_instructions) is shown before the scored trials begin.

4. Main Task:

   • For each item:
     • View the eye photograph
     • Read the four mental state options (a 2×2 grid of buttons)
     • Click/tap the best answer
     • No feedback during the main task
   • Optional per-item response timeout advances automatically when reached

5. Completion: A summary artifact is generated with per-trial results and accuracy.

Design Recommendations

General Guidelines

• Self-Paced: Recommended to allow time for careful consideration
• No Time Pressure: Timed administration reduces validity by adding speed demands
• Randomization: Counterbalances order effects across participants
• Glossary: Helpful for non-native speakers or individuals unfamiliar with complex emotion terms
• Practice: Ensures participants understand task format

Scoring and Interpretation

Normative Data (Baron-Cohen et al., 2001):

• Typical Adults: Mean ~26/36 (SD ~3.5), range 18-34
• Female Advantage: Women score ~1-2 points higher on average
• Autism Spectrum: Mean ~22/36 (SD ~4), range 10-32
• Cutoff: Scores ≤21 suggest mentalizing difficulties (not diagnostic alone)

Scoring Methods:

• Total Correct: Primary outcome (0-36)
• Response Time: Secondary measure; very slow RTs may indicate difficulty or overthinking
• Item Analysis: Some items are more difficult; can examine patterns

Population-Specific Adaptations

Adolescents (13+ years):

• Full 36-item version appropriate
• Consider glossary for less familiar terms
• Scores slightly lower than adults (developmental trajectory)

Older Adults (65+):

• Full version appropriate
• Larger font size (20-24px) recommended
• High-contrast mode for visual accessibility
• Scores generally stable with aging (no major decline in healthy aging)

Non-Native Speakers:

• Critical: Glossary should be mandatory
• Consider translating mental state terms (requires validation)
• Cultural differences in expression recognition possible

Autism Spectrum:

• Standard administration appropriate
• Expect lower scores but wide variability
• Some individuals develop compensatory strategies (explicit rule-learning)
• Scores don't correlate with intelligence in ASD

Clinical Populations:

• Schizophrenia: Moderate impairment (mean ~21-23)
• Borderline Personality Disorder: Variable performance
• Social Anxiety: Typically normal or slightly reduced scores
• Depression: May show slight reduction during acute episodes

Common Issues and Solutions

Issue	Solution
Ceiling effects (many perfect or near-perfect scores)	Add time pressure (10s/item) or use more difficult items
Participant unfamiliar with vocabulary	Enable glossary feature; provide definitions before task
Very slow responses (>10s per item)	Normal; allow self-pacing; only flag if consistently >15s
Low motivation/engagement	Emphasize no right/wrong, focus on "best fit"; use progress bar
Cultural differences in expressions	Acknowledge limitation; use caution comparing across cultures
Participants overthink responses	Instruct to go with "first impression" or "gut feeling"

Validity and Reliability

Psychometric Properties:

• Internal Consistency: Cronbach's α = 0.60-0.70 (modest due to item diversity)
• Test-Retest: r = 0.63 over 1 year (moderate stability)
• Convergent Validity: Correlates with other ToM measures (r ~ 0.3-0.5)
• Discriminant Validity: Low correlation with basic emotion recognition (r ~ 0.2-0.3)
• Sensitivity: Discriminates autism from typical development (Cohen's d ~ 0.8-1.2)

References

• Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (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.

• Baron-Cohen, S., & Wheelwright, S. (2004). The empathy quotient: An investigation of adults with Asperger syndrome or high-functioning autism, and normal sex differences. Journal of Autism and Developmental Disorders, 34(2), 163-175.

• Oakley, B. F., Brewer, R., Bird, G., & Catmur, C. (2016). Theory of mind is not theory of emotion: A cautionary note on the Reading the Mind in the Eyes Test. Journal of Abnormal Psychology, 125(6), 818-823.

• Prevost, M., Carrier, M. E., Chowne, G., Zelkowitz, P., Joseph, L., & Gold, I. (2014). The Reading the Mind in the Eyes test: Validation of a French version and exploration of cultural variations in a multi-ethnic city. Cognitive Neuropsychiatry, 19(3), 189-204.

• Vellante, M., Baron-Cohen, S., Melis, M., Marrone, M., Petretto, D. R., Masala, C., & Preti, A. (2013). The "Reading the Mind in the Eyes" test: Systematic review of psychometric properties and a validation study in Italy. Cognitive Neuropsychiatry, 18(4), 326-354.

See Also

• Facial Expression Recognition - Basic emotion recognition (different construct)