Bibliography

Ekman & Friesen Pictures of Facial Affect: the classic 110-photo set (happiness, sadness, anger, fear, surprise, disgust, neutral), Consulting Psychologists Press. NO DOI (slide set / book). Lower-confidence as a 'record' since not DOI-indexed; cite as Ekman & Friesen (1976), Palo Alto: Consulting Psychologists Press. KDEF original (Lundqvist, Flykt & Ohman, 1998) is similarly non-DOI; cite KDEF via Goeleven 2008 validation above.

The foundational visual-search demonstration that an angry face is detected faster and more in parallel ("pops out") among happy-face distractors than vice versa, framed as preattentive threat detection. This is the origin of the anger-superiority effect, and the very stimuli used here were later shown to carry a low-level contrast artifact (Purcell et al.), making this paper the anchor of the entire confound debate.

Early re-examination (co-authored with the Hansens) that scrutinizes whether the face-in-the-crowd advantage reflects genuine pop-out / parallel search, probing the search-slope signature of the original effect. An early flag that the "pop-out" interpretation of anger superiority needed harder evidence.

Verified abstract: shows the Hansen & Hansen angry faces contained extraneous dark areas (a contrast artifact) introduced when photographs were converted to black-on-white sketches. When artifact-free grey-scale angry/happy faces were used, no pop-out was found; all targets were located by serial, self-terminating search regardless of expression. The seminal demonstration that "anger superiority" can be a low-level image artifact rather than emotion per se.

Systematically manipulates individual schematic facial features (eyebrows, mouth, eyes) to identify which geometric components drive perceived threat and capture attention. Demonstrates that downward/inward eyebrows and mouth curvature are the primary "threat" carriers - precisely the low-level geometric cues that can confound emotion with shape in search tasks.

Using hybrid faces that combine one expression in low spatial frequencies with another in high spatial frequencies, shows that which expression observers perceive depends on the spatial-frequency band their task makes diagnostic. Demonstrates that emotional expression categorization is driven by selective use of spatial-frequency information.

Reinstates a threat-detection advantage using schematic (line-drawing) faces designed to equate many low-level properties, reporting faster/more efficient search for angry than friendly schematic faces among neutral distractors. A key pro-threat counter to the confound critique, but schematic stimuli themselves introduce their own feature confounds (later probed by Coelho, Horstmann, Becker).

Shows the search asymmetry can be reframed as a distractor-rejection effect: happy-face distractors are rejected more efficiently than angry ones, which can masquerade as an angry-target advantage. Highlights that "anger superiority" RT patterns are sensitive to whether the asymmetry lives in target detection or distractor processing.

Comprehensive review of the emotional-face visual-search literature, weighing evidence for genuine emotion-driven search guidance against low-level perceptual-confound and methodological explanations. The standard synthesis of the anger/happy superiority debate and the controls each side requires.

Validation paper for the KDEF set (original set: Lundqvist, Flykt & Ohman, 1998, Karolinska Institutet, no DOI). 70 actors x 7 expressions (neutral, happy, angry, fearful, disgusted, sad, surprised) x 5 angles. Canonical stimulus-control citation. DOI verified (HTTP 302 -> Taylor&Francis).

NimStim database: 43 actors, multiple expressions (happy, angry, fearful, sad, surprised, disgusted, neutral, calm), open/closed mouth versions. Widely used, freely available to researchers. DOI verified (HTTP 302 -> Elsevier).

FACES (Ebner/MPI): 171 models across 3 age groups, 6 expressions (neutrality, sadness, disgust, fear, anger, happiness), 2 sets each. Free for research. DOI verified (HTTP 302 -> Springer). An OSF expansion 'FACES-X' (node w8snv) was also seen in OSF search.

Radboud Faces Database (RaFD): 67 models (adults + children), 8 expressions (anger, disgust, fear, happiness, sadness, surprise, contempt, neutral) x 3 gaze x 5 camera angles. Free for academic use. DOI verified (HTTP 302 -> Taylor&Francis).

Reports an anger-superiority advantage using photographic real faces with multiple identities (rather than a single repeated identity), arguing the effect survives when identity is varied to reduce simple feature-matching. A pro-threat result that attempts to control the identity/low-level confounds raised by earlier critiques.

Foundational application of the Ratcliff diffusion model to threat processing: re-analyzed a non-competition single-string lexical decision task in high- vs low-anxious individuals. The diffusion model revealed a consistent processing advantage (higher drift rate) for threatening words in high-anxious participants that mean RT/accuracy analyses missed, challenging the claim that input competition is necessary for enhanced threat processing.

Using controlled photographic stimuli, finds that once low-level confounds are removed it is HAPPY faces, not angry faces, that are detected more efficiently in both single- and multiple-target search. Directly inverts the classic anger-superiority claim and argues prior threat advantages were driven by uncontrolled image features.

Demonstrates that the schematic angry-face search advantage can be reproduced with non-face control stimuli (e.g., crosses / abstract shapes sharing the angry face's geometry), showing the "threat" effect tracks low-level visual features rather than emotional meaning. A direct empirical attack on the emotion-per-se interpretation.

Shows that very fast saccadic/behavioral face detection can be substantially supported by low-level image statistics (e.g., amplitude spectrum / contrast structure) rather than high-level face processing. Establishes that the earliest, fastest detection responses are exactly where low-level confounds dominate.

Develops a signal-detection version of the face-in-the-crowd task grounded in a model of automatic, stimulus-driven threat detection, after three calibration studies equating angry/happy stimuli for perceptibility, homogeneity, and intensity. Sensitivity (d') showed the expected angry-detection advantage in neutral crowds, consistent with a stimulus-driven attentional shift toward angry targets.

Psychophysically measures the stimulus exposure time needed to categorize each of the six basic expressions, finding large differences across emotions (e.g., happiness categorized at very short exposures, others requiring much longer). Provides expression-specific information-accrual time courses.

A systematic head-to-head test of anger vs. happiness superiority across stimulus manipulations, co-authored by key figures in the debate (Savage, Lipp, Becker, Horstmann). Shows the direction of the search advantage depends on stimulus and procedural factors rather than reflecting a fixed threat-detection mechanism.

Uses eye-tracking to decompose the angry-face search advantage into where attention is first directed vs. how quickly a fixated face is recognized as the target. Helps localize whether the RT advantage arises at guidance/orienting or at the decision/ confirmation stage - directly relevant to which EA parameter the effect maps onto.

Meta-analysis distinguishing two often-conflated tasks: detecting that an emotional face is present vs. recognizing/categorizing which emotion it is. Finds the advantaged expression differs by task (a happy advantage in recognition, with detection patterns dissociating), showing "the emotional advantage" is task-dependent.

Shows that whether anger or happiness superiority is observed depends systematically on the composition of the stimulus set (which identities, expressions, and distractors are used). Reinforces that the search asymmetry is partly an artifact of stimulus-set selection rather than a stable property of emotional content.

Decomposes the happy-face recognition advantage into contributions from visual saliency and feature distinctiveness (the high-contrast smiling mouth) vs. affective meaning. Argues the smile's perceptual salience/distinctiveness largely drives its rapid, accurate categorization - a low-level account of the happy advantage.

TOP STAGING CANDIDATE for EZ-diffusion reproduction. Zenodo dataset (DOI 10.5281/zenodo.836829, CC-BY-4.0) accompanying Vesker et al. 2018 PLOS ONE. Facial-expression VALENCE-categorization task (positive vs negative). Two .xlsx files CONFIRMED by parsing sharedStrings: "Correct RT data Zenodo.xlsx" and "Accuracy data Zenodo.xlsx". Both share the long-format column scheme: subject_sex, photo_sex, photo_ID, subject_age, "subject ID (sex_age_subject-number)", "emotion (sad/angry/fear/happy/surprised-happy)", photo_valence (Negative/Positive), and the per-response value ("correct RT " for RT file; accuracy for the other). N = 96 distinct subject IDs across 4 age groups (6, 9, 12 yrs, Adults; ~12 male + 12 female per group). Provides RT + accuracy + emotional-expression condition per subject -> sufficient for EZ-diffusion (Pc, MRT, VRT per cell). Caveats: developmental sample (children + adults), binary task is positive/negative valence (no explicit "neutral"), and data are in wide-ish xlsx that need reshaping; emotion is the within-stimulus manipulation. Reproduction-suitability: HIGH (only verified open dataset here with trial-resolved RT+accuracy+expression).

Companion paper for the Zenodo 836829 dataset. PLOS ONE (open access). Describes the valence categorization task, stimuli (positive/negative expressions: happy, surprised-happy vs sad, angry, fear), age groups, and RT/accuracy analyses. Use for task design + condition definitions when reconstructing the EZ-diffusion pipeline.

Used a gating paradigm to track how emotion-expression recognition unfolds over exposure time across face, voice, and audiovisual signals. Discrimination was reached earlier with faces than voices, multisensory stimulation reduced the perceptual evidence needed (earlier isolation point), and emotions differed in accumulation speed, with 'fear' isolated fastest across senses.

Applied diffusion modeling to an emotion-identification task in 90 incarcerated men. Physical aggression was positively associated with drift rate (more efficient evidence accumulation) specifically for anger, and drift rate statistically mediated the link between aggression and heightened anger identification. Aggression was not associated with starting-point bias or threshold separation, implicating processing efficiency (v) rather than bias as the mechanism.

GitHub repo isacdaavid/emotional-faces-psychopy-task: a PsychoPy emotional-faces task. Repo resolves (last updated 2020-10-16) BUT has NO license file (404 on LICENSE, GitHub license=None) -> NOT safe for reuse/redistribution without contacting the author. Lower confidence / reference-only.

CHECKED per instructions. OSF node s46pr resolves ('Confidence database'); it has 0 child nodes and stores 360 files (data_*.csv / readme_*.txt, indexed by author name) plus Database_Information.xlsx. Filenames are author-keyed, not task-keyed; parsing the index xlsx sharedStrings surfaced NO emotion/facial-expression/angry/fearful task labels. Conclusion: no clear emotion-EXPRESSION categorization task found in the Confidence Database (it is dominated by perceptual/memory/value confidence tasks). Recorded as a negative result, not a candidate.

Combined a diffusion decision model with ERPs (P100, N170, P250) during facial emotion recognition, manipulating stimulus orientation (upright/inverted) and type (photo/sketch). Face inversion reduced drift rate and stimulus impoverishment increased the encoding component of non-decision time; N170 activity appeared to increase to counteract inversion's detrimental effect on drift rate, linking the N170 to evidence accumulation.

PsyArXiv/OSF preprint (DOI 10.31234/osf.io/a8yxf, resolves to osf.io/a8yxf). On-target methodologically: applies a drift-diffusion model to emotion judgments of facial expressions (masked vs unmasked). VERIFIED the preprint node exists but it has NO linked public OSF data component (relationships.node = none), so trial-level data could not be confirmed open. Lower confidence as a data source; useful as a DDM-on-expression methods reference.

Applied diffusion modeling to an emotion-identification task (anger/happiness/fear blends with manipulated contextual threat) in 92 incarcerated men. Psychopathy was associated with longer non-decision time (slower processing), especially for mostly-angry ambiguous faces, whereas externalizing was associated with shorter non-decision time and greater threshold separation, dissociating two disinhibitory phenotypes.

PsychoPy implementation of an emotional go/no-go task + preprocessing script. Zenodo DOI 10.5281/zenodo.6207865 (resolves HTTP 302), license 'other-open' (free for researchers; check terms in the record before redistribution). File: emotional_go_no_go-Psychopy.zip. Verified creators/files via Zenodo API. Closest verified, properly-licensed, citable open task code for emotional-face responses.

Links between-person differences in expression-recognition ability to how efficiently each observer samples the diagnostic facial regions (via reverse-correlation / bubbles methods). Better recognizers use diagnostic information more efficiently, grounding individual differences in perceptual sampling rather than just response bias.

Used diffusion-model analysis of a face-in-the-crowd visual search task in which participants detected angry and happy faces (and visually matched 'anti-expressions') among neutral faces. Drift rate was larger and non-decisional time shorter for normal emotional expressions than for anti-expressions, and subjective arousal ratings were positively associated with drift rate (and negatively with non-decision time). Threshold separation was also larger for normal expressions. Concludes that efficient detection of emotional faces is driven partly by faster evidence accumulation (higher v).

Combined DDM and N2pc ERP analysis of a dot-probe task (26 lonely vs 26 non-lonely). Found no evidence of social-threat hypervigilance in either DDM parameters or N2pc, but lonely individuals showed a decreased drift rate and increased between-trial drift variability, indicating less efficient perceptual decision-making not captured by standard behavioral measures.

Forty participants rated sad/neutral/happy faces presented for 8.3, 16.7, and 25 ms, with performance estimated via hierarchical drift diffusion models alongside subjective/objective awareness. Emotion-dependent responses appeared as decreased drift rate and accuracy during sad trials, and 16.7 ms was identified as optimal for subconscious emotional priming.

Methodological study of angry-vs-happy expression decisions showing that participant-sex and Face Sex x Expression effects depend on outlier-removal and distributional modeling choices. Recommends the Diffusion Model, ex-Gaussian, ex-Wald and related distributional models as replacements for mean-RT analyses of facial-expression decision times.

Two pre-registered studies (N=228; N=264) using DDM on judgments of six emotions (incl. anger) with upper/lower face masked or unmasked. Masking the lower face reduced accuracy and slowed responses, and DDM localized this to a decreased drift rate: participants accumulated emotional evidence more slowly when the expressive region was occluded, an effect that intensified later in the pandemic.

Applied a drift-diffusion model to an explicit happy-to-angry face-emotion labeling fMRI task in youth and adults (n=44), separating sensitivity (drift) from bias. Age was associated with perceptual sensitivity to facial affect, and anterior insula showed increasing response to emotion ambiguity with age; both valence- and ambiguity-encoding systems contributed to discriminating face emotions.

Two experiments (N=546) used a sequential priming weapon-identification task with angry/happy and Black/White male face primes, analyzed with the diffusion decision model. Racial bias was weaker when emotion-expression salience was heightened; modeling supported an initial-bias (starting-point) account, with the decision process starting closer to 'gun' after Black faces and this shift weakened when expression was salient.

Double-blind placebo-controlled study (n=120 men) giving 150 mg testosterone or placebo, then an anger/fear-vs-neutral morph judgment task analyzed with regression, signal detection, and drift diffusion modeling. Converging analyses showed testosterone lowered perceived sensitivity to angry facial expressions specifically (no effect for fear), suggesting testosterone attenuates threat sensitivity.

Participants (n=55) used prestimulus cues to judge whether targets were threatening or neutral, analyzed with SDT and hierarchical DDM. High-uncertainty threat cues both shifted the starting point toward the threat decision and increased the efficiency of evidence accumulation (drift rate) for threat and neutral decisions; anxiety was linked to a greater starting-point shift but not to drift-rate efficiency.

Developed and replicated a social approach-avoidance conflict paradigm using happiness/anger morphs to parametrically vary social reward, social threat, and reward-threat conflict. Conflict trials produced more intermediate approach-avoidance choices and slower RTs, and drift diffusion models showed social conflict was driven by noisier (less efficient) evidence accumulation.

Population-based fMRI study (n=134) using DDM to index sustained task-irrelevant attention to angry vs neutral faces (attentional negativity bias). DDM showed reduced drift rate for angry compared with neutral faces, and the size of this reduction tracked self-reported depression history; drift for angry faces correlated with insula/IFG and parietal activity.

Used hierarchical drift diffusion modeling of an emotional interference task (identify a facial emotion with a congruent/conflicting emotion word) in 86 adolescents/young adults with and without borderline personality symptoms. Emotion-related impulsivity predicted inefficient processing (lower drift rate) under conflicting negative emotional stimuli and a lower decision boundary, with age-related differences in facial emotion processing.

Comprehensive review of gaze cueing: how perceived gaze direction automatically shifts observer attention in the cued direction, manifesting as shorter RT for targets at gazed locations. Reviews mechanisms (reflexive vs voluntary), modulating factors (gaze eccentricity, face realism, counterpredictive cues), and individual differences. RT cueing effects are ~10–40 ms.

Demonstrates a Bayesian prior for direct (mutual) gaze: under uncertainty about another person's gaze direction, observers systematically assume they are being looked at. The prior operates across a range of gaze directions (not only left/right dichotomy), and its strength decreases with gaze eccentricity. Critically relevant to understanding the psychometric function for gaze direction discrimination.

Proposed and validated the dual-route model of gaze direction perception: perceived gaze direction is a linear combination of eye-region (iris/pupil position) and head orientation signals. Head orientation has both a repulsive effect (pushing perceived gaze toward the averted side) and an attractive effect (pulling toward the head direction). Explains why head+eye integration is necessary for veridical gaze perception.

First demonstration that observers rapidly extract a summary representation of a crowd's collective gaze direction, pooling information from multiple crowd members. Gaze was perceived as a global combination of head and pupil rotation cues, and pooling occurred at high-level stages of visual processing. This paper establishes that crowd-gaze direction is an ensemble percept analogous to global motion direction.

Used equivalent noise analysis to quantify how well observers average gaze directions from spatially distributed groups of faces. Derived two key metrics: internal noise (uncertainty per individual gaze element) and effective sample size (number of elements pooled). Found observers pool a limited number of gaze directions efficiently, with performance limited by internal noise rather than ideal summation.

Extended Florey et al. (2016) to compare spatial (simultaneous crowd) vs temporal (sequential single-face) gaze averaging using equivalent noise analysis. Found differences in effective sample size and internal noise between spatial and temporal presentation, suggesting distinct mechanisms. Provides a direct empirical framework for designing a crowd-gaze task.

Replicated and extended Sweeny & Whitney (2014) by presenting crowds with conflicting gaze direction cues (head vs eye pointing in opposite directions). Found that observers weight head orientation more heavily than eye rotation in forming crowd gaze ensemble percepts, and that the majority direction dominates when cues conflict.

Comprehensive review defining ensemble perception as the visual system's ability to extract summary statistics from groups of objects in a brief glance. Covers gaze direction, emotion, identity, and numerosity. Synthesizes both spatial (simultaneous) and sequential ensemble averaging, internal noise models, and the role in behavior. Cites ≥463 citations.

Found that gaze direction perception exhibits positive serial dependence — current gaze judgments are biased toward the most recently perceived gaze direction, equivalent to temporal smoothing. Effect is localized in visual space and operates on perceived (not retinal) gaze direction. Demonstrates that head and eye information are integrated before the serial dependence stage.

Reviews gaze direction adaptation aftereffects: prolonged exposure to averted gaze biases subsequent gaze judgments toward the opposite direction. Discusses measurement methods, neural locus, and theoretical implications of gaze direction norm-based coding. Covers both head and eye rotation contributions to perceived gaze direction.

Measured thresholds for gaze direction discrimination (head+eye integration) in schizophrenia vs controls, drawing an explicit comparison to motion coherence thresholds. Found reduced integration of head and eye cues in schizophrenia, analogous to previously reported impairments in motion coherence perception. Reports gaze direction discrimination thresholds (degrees) per participant.

Applied formal evidence accumulation models (diffusion model / LBA) to gaze cueing RT data from three datasets (N=171, 139,001 trials, pooled) using individual and hierarchical Bayesian fitting. The gaze cueing effect is best described as an attentional-orienting / non-decision-time effect: there was little evidence for a sustained reallocation of information-processing resources, with drift rate the least-included parameter (model-inclusion probability only 3–13%). Provides the first DDM decomposition of social gaze orienting.

Trained rhesus monkeys on a random dot direction discrimination task and measured coherence thresholds before and after MT lesions. The abstract states they used a dynamic random dot display allowing them to vary motion signal intensity amid masking noise and measured threshold intensity for successful direction discrimination. Each animal was individually calibrated to threshold coherence, establishing the per-subject coherence calibration paradigm.

Foundational RDM study: simultaneously measured a monkey's psychophysical threshold for motion-direction discrimination as a function of motion coherence and the sensitivity of single MT neurons, showing individual MT neurons can rival behavioral coherence thresholds. Origin of the % motion-coherence construct.

Studied motion-blind patient L.M. using random dot limited-lifetime stimuli to separate signal-dot lifetime effects from noise-dot masking. Different stimulus versions isolated limited-lifetime effects from additive noise as coherence was lowered. When signal dots had fixed lifetime and only noise-dot fraction was varied (the Newsome-Pare method), a different deficit pattern emerged compared to varying lifetime simultaneously.

Compared psychophysical and single-neuron MT sensitivity for direction discrimination in rhesus monkeys near psychophysical threshold using a stochastic RDM display. Both psychophysical and physiological data were obtained from the same animals on the same trials with the display matched to each neuron's preferences. Individual MT neurons matched psychophysical threshold and predicted the shape of the psychometric function, suggesting decisions can be based on signals from a small number of MT neurons.

Developed a numerical simulation model showing how MT neural signals support perceptual decisions about RDM direction. The model pools MT neurons and feeds downstream (LIP) decision circuitry, generating predictions about psychometric functions and the coupling between neural firing and behavioral choice. Explains the correlated firing observed in MT and its effect on sensitivity.

Recorded LIP neurons in rhesus monkeys performing fixed-duration direction-discrimination task with random-dot motion. LIP neural activity reflected motion-direction judgments, with activity correlating with the monkey's ultimate choice even before motion onset. The gradual build-up of activity reflected accumulation of sensory evidence and linked neural firing to choice accuracy at multiple coherence levels.

Recorded LIP activity in rhesus monkeys (n=2) during a free-response (RT) motion-direction discrimination task using random-dot stimuli at multiple coherence levels. Both accuracy and RT improved monotonically with coherence. LIP neurons showed ramp-like firing that predicted the monkey's choice, with steeper ramps for stronger motion and shorter RTs, suggesting LIP integrates time-varying motion signals to a threshold governing the speed-accuracy tradeoff.

Developed a diffusion-model theory unifying chronometric (RT vs coherence) and psychometric (accuracy vs coherence) functions for RDM psychophysics. The proportional-rate and power-rate diffusion models predict analytic expressions for both functions and subsume Pieron's Law. Human psychophysical experiments confirmed the coupling between RT and accuracy; provides the theoretical bridge between coherence (sensory strength) and DDM parameters (drift rate scales with coherence).

Comprehensive Annual Review article on how the brain implements deliberation and commitment in perceptual decision making. Focuses on studies using RDM paradigms in primates and humans, reviewing how MT encodes sensory evidence, LIP integrates that evidence over time, and a threshold mechanism terminates the decision. Connects neural mechanisms to the diffusion model framework.

Used the RDM reaction-time task with a 'sure-bet' option to show LIP neurons represent decision confidence within the evidence-accumulation framework; links coherence, RT, choice and confidence in one diffusion account.

RDK direction-discrimination (leftward vs. rightward, 200 ms fixed duration) with continuous confidence ratings; six motion coherence levels plus volatility condition. Trial-level CSV (data_Zylberberg_2016.csv) deposited in the Confidence Database (OSF s46pr, download: https://osf.io/download/6d2hn/) with columns: Subj_idx, Stimulus (0/1), Response (0/1), Confidence (0-1), RT_decConf (s), Coherence, Condition, Accuracy, Block, Date. N=3 subjects, 1536-2107 trials each; Shadlen lab, Columbia University 2015.

Published description of a canonical Random Dot Kinematogram plugin for jsPsych (web-based research). The plugin (@jspsych-contrib/plugin-rdk v2.0.0) is now maintained in the official jspsych-contrib monorepo under MIT license at https://github.com/jspsych/jspsych-contrib/tree/main/packages/plugin-rdk. Cited 38 times (Crossref). Plugin supports configurable coherence, direction, motion type, and records RT and keypress response. URL verified HTTP 200.

Four-experiment study on perception of opposite-direction motion in RDKs. Data CSVs (Data_Expt1-4.csv, ~800KB-1MB each) with per-trial columns: subject_Num, trial_Num, TargetAngle, AngularError, RespAngle, directBin, Duration, Segments, Gender, Coherence, RT, Block, Session, Tdiff. Two coherence levels (0.25, 0.50). Task is angle estimation/direction report in a circular-report paradigm, not a standard 2AFC left/right task. OSF fexmg resolves, CSV verified via curl.

Shows that choice history biases in perceptual decisions act on drift rate and drift-rate criterion, not only starting point, contradicting standard assumptions. Fit bounded-accumulation DDMs to human perceptual choice data across multiple task protocols and sensory modalities; history bias parameters depended on observers' previous choices and manifested primarily as drift-rate rather than starting-point shifts.

Published description of the Confidence Database (OSF s46pr), a repository of 145+ datasets from 8700+ participants and ~4 million trials spanning perceptual, memory, motor and cognitive confidence tasks. Datasets are structured in a common CSV format. The database includes at least two RDK/motion-coherence datasets (Zylberberg_2016 and Law_unpub) with trial-level RT + Coherence + Accuracy. Available at https://osf.io/s46pr/

OSF repository for a study on self-relevant decisions and RDM task perceptual processing. Contains multiple studies with raw data folders (Raw Data Study 1, 2, 3; RawData_Exp2) and LBA model analysis code (LBA_RDM_Exp1, LBA_RDM_Exp2 folders). Task description document also present. Public OSF project, last created 2020. LBA (Linear Ballistic Accumulator) analysis suggests the data has per-trial RT and accuracy. OSF gc7eh resolves (HTTP 200 verified).

International Brain Laboratory standardized mouse 2AFC visual decision-making task. 140 mice, 7 laboratories, training and behavioral data published. The task uses visual contrast gratings (not motion coherence), but provides open trial-level choice + RT data accessible via the ONE API (https://one.internationalbrainlab.org). Data portal is https://data.internationalbrainlab.org (HTTP 200 verified via redirect). Brain Wide Map data also at figshare.com/articles/preprint/Data_release_-_Brainwide_map_-_Q4_2022/21400815.

MIT-licensed PsychoPy implementation of a Random Dot Kinematogram experiment, based on the Burghoorn et al. 2019 article. Implements left/right direction discrimination with configurable coherence and dot parameters. Repository at https://github.com/AnonZebra/rdk-psychopy, last updated 2022-06-03. License: MIT. 0 stars (small but usable reference implementation).

OSF repository of MATLAB scripts implementing a Variable Coherence Random Dot Motion (VCRDM) active paradigm using Psychophysics Toolbox. Files include VCRDM_active_paradigm_v1p4.m (20KB main task), showLocalizerDots_random.m (5KB), createTRect.m (1KB). OSF jyzgm resolves (HTTP 200 verified). License not stated explicitly; public OSF repository.

RDK left/right direction-discrimination using PsychoPy2 with simultaneous 4-point confidence rating. Difficulty controlled by motion coherence via 6 independent ASA staircases targeting 3 accuracy levels (52%, 65%, 78% correct). Per-participant coherence thresholds estimated and applied. Trial-level CSV (data_Law_unpub.csv, OSF s46pr) has columns: Subj_idx, Stimulus, Response, Confidence, RT_decConf, blocki, triali, keypress, coh_level, coherence, isCalibTrial, stairID, accTarget. CSV verified download: https://osf.io/download/vgr27/

Applied Bayesian models to smooth pursuit eye movements elicited by random dot kinematograms at varying coherence and directional predictability. Found pursuit velocity is driven by both immediate sensory motion and anticipated direction, with reliability-based weighting. Bayesian models accounted for most but not all pursuit behaviour across noise and predictability conditions.

MIT-licensed PsychoPy random dot motion task implementation. Public GitHub repository at https://github.com/abcsds/RDM, last updated 2024-10-13. MIT license, no stars. Minimal but open PsychoPy RDM implementation.

Foundational review of binocular rivalry covering 160+ years of psychophysics. Reviews evidence that rivalry is implemented by neural mechanisms at multiple levels of the visual hierarchy rather than a single locus, with imaging and electrophysiology showing competitive interactions at early and high-level stages. Proposes a multi-level framework reconciling contradictions in the rivalry literature.

Introduces continuous flash suppression (CFS): distinct images flashed at ~10 Hz to one eye suppress a stimulus in the other eye for durations at least ten times longer than standard binocular rivalry. Demonstrates that strength of negative afterimage is reduced in proportion to degree of CFS suppression, implying that neurons contributing to afterimages are accessible to both-eye input but are not the direct correlates of perceptual awareness.

Proposes a computational model of binocular combination in which each eye exerts gain control on the other eye's signal in proportion to its own contrast energy, plus second-order gain control on the other eye's gain control. The model accurately predicts perceived phase of binocular combination stimuli and provides an equation linking per-eye contrast to perceived dominance.

Early demonstration that interocularly suppressed (invisible) images of erotic content can direct spatial attention, with the direction of cueing depending on the observer's gender and sexual orientation. Participants were unaware of the suppressed pictures yet attention was implicitly captured, establishing that content-specific processing occurs under interocular suppression before awareness.

Reviews neuroimaging and psychophysical evidence that rivalry entails competitive interactions at multiple neural sites including eye-selective sites. Rivalry greatly suppresses ventral pathway activity and attenuates visual adaptation, yet some information about suppressed stimuli reaches higher areas. Proposes a hybrid model where low-level inhibitory interactions interact with high-level excitatory influences to produce paradoxical rivalry properties relevant to visual awareness.

Introduces a dichoptic interocular suppression test (N=88) in which dynamic noise is shown to one eye and a target to the other; the participant reports when the target breaks through suppression. Provides a standardized psychophysical technique that quantifies interocular suppression strength per eye, yielding an index of sensory eye dominance that is more sensitive than binocular rivalry phase timing alone.

Critical evaluation of the b-CFS paradigm using upright vs inverted face detection across six experiments. Finds that upright faces are detected faster than inverted in both CFS and control (non-rivalrous) conditions, though the effect is larger under CFS. RT distributions are critical: when matched across conditions, effect sizes converge, suggesting CFS amplifies rather than creates the upright-face advantage. Challenges the assumption that b-CFS unambiguously measures unconscious processing.

Measures binocular rivalry using chromatic (red-green) and achromatic (luminance) gratings separately after 150 minutes of monocular patching in adult humans. Shows that deprivation effects manifest differentially for luminance-driven and chromatic-driven rivalry, indicating that colour channel and luminance channel are processed partially independently in binocular competition.

Manipulates coherence of random-dot motion stimuli (paralleling RDM coherence manipulations) in binocular rivalry, comparing to contrast manipulations. Finds that both contrast and coherence obey revised Levelt's propositions: increasing coherence/contrast in one eye increases that eye's dominance duration. At low coherence, dominance durations increase before decreasing at higher coherence — an effect absent in contrast experiments — reconciled via an adaptation mutual-inhibition model with a nonlinear input transfer function.

Uses a novel dichoptic suppression paradigm to titrate conscious vs. unconscious evidence in a decision task. Fits a diffusion model showing that removal of conscious awareness reduces (but does not eliminate) both the rate and variability of evidence accumulation. Unconscious evidence integrates with conscious evidence over time to boost or diminish accuracy. Critically: v is reduced but non-zero under suppression, and this reduced-v is not reflected in confidence.

Comprehensive review of Levelt's 1965 propositions on binocular rivalry at their 50th anniversary. Revisits original propositions, updates them against modern evidence, and demonstrates that modified propositions apply broadly to bistable perceptual phenomena beyond binocular rivalry. Argues that the propositions reflect canonical neural properties of competitive inhibition that generalize to visual perception broadly.

Fear-conditioned stimuli (devoid of intrinsic social content) gain faster access to awareness in a b-CFS paradigm than non-conditioned stimuli. Stimuli paired with electric shock broke suppression faster in the test phase (no shocks). Demonstrates that learned threat relevance — not just intrinsic biological saliency — modulates suppression duration via breaking CFS.

Re-analysis of existing b-CFS datasets plus a new dataset showing that condition effects (e.g., upright vs inverted faces) are highly correlated with participants' overall suppression duration and trial-to-trial RT variability. Advocates a latency-normalization method that removes between-subject suppression duration variability, outperforming z-score and log transformations. Discusses how individual differences in overall suppression duration may reflect prolonged shallow suppression phases that are more sensitive to content.

Compares eight sensory eye dominance (SED) tests on 30 normally-sighted participants. Tests #1–6 estimate a contrast balance-point (BP), defined as the interocular difference in contrast at which participants are equally likely to base their judgement on either eye. The paper evaluates reliability and agreement across methods, identifying which tests translate best to clinical and lab settings.

Shows that SED can arise from two distinct processes: (1) an imbalance in binocular combination/summation (integration pathway) and (2) an imbalance in binocular inhibitory suppression (competitive pathway). These can be dissociated using binocular integrative stimuli vs binocular rivalry stimuli, and their relative contributions vary across individuals.

Zenodo dataset (DOI 10.5281/zenodo.1006000) supplying EDF eyetracker files and extracted MATLAB .mat files from a binocular rivalry study examining how visual awareness modulates pupil size. Raw data is available in .edf format (Eyelink device output) and as processed .mat files containing eye position, pupil size, and keypress perceptual state variables. Also includes analysis code (analysis.zip, 183MB).

Documents substantial individual variation in both binocular rivalry (BR) dynamics and continuous flash suppression (CFS) interocular suppression, and shows these two measures are not highly correlated across individuals. Highlights that pooling across participants obscures large individual differences in how much each eye dominates, reinforcing the need for per-participant calibration.

Review by Ooi & He (the primary proponents of sensory eye dominance measurement) covering how SED arises from interocular imbalances in binocular combination and suppression, how it differs from motor eye dominance, and the methods used to quantify it. Reviews the relationship between SED, cortical organisation, and amblyopia treatment.

Compares the reliability of binocular rivalry, CFS, and binocular combination as tools to measure ocular dominance and deprivation-induced shifts. Finds that different tasks probe different aspects of the binocular system and differ substantially in reliability, with direct practical implications for choosing a calibration method.

OSF repository (https://osf.io/zqnbt) for Kim et al. (2021, Sci Reports) on CFS affective priming. Contains trial-level CSV data with columns person, RT, response, emot, condition, absRT, face, task for 20 participants, 4492 trials across two experiments (face: neutral/scramble/flip conditions). Also includes a full Bayesian hierarchical diffusion model in JAGS/MATLAB (Experiment1.txt) with drift rate v, boundary a, and non-decision time theta estimated per participant per condition.

Argues that two systems act in opposition across different timescales during monocular deprivation: a fast adaptation process shifting balance toward the deprived eye, and a slower homeostatic process restoring balance. This biphasic model reframes how interocular balance is defined and tracked across time, emphasising that the 'balance point' is dynamic and not a fixed parameter.

Applies drift-diffusion modelling to b-CFS breakthrough times for fearful vs. neutral faces (n=33) alongside 7T fMRI. DDM modelling shows fearful faces accumulate perceptual evidence (drift rate v) faster than neutral faces; faster breakthrough times correlate with greater amygdala and insula activation; higher v maps to frontoparietal and occipital cortex; lower decision boundary (a) maps to insula and PCC. Provides direct DDM parameter estimates for b-CFS face detection.

OSF repository (https://osf.io/4sjxz, created 2022-03-09) for a CFS priming study. Contains ScenePrimingCFS_Data.txt (3.6MB, confirmed trial-level), ScenePrimingCFS_stimuli.zip (stimuli), and ScenePrimingCFS_Analysis.R (R analysis). Column structure confirmed by CSV preview: participant; prime_category; prime_context; prime_filename; target_category; target_context; target_filename; congruency; mask_contrast; soa; iti; task; accuracy; rt. Mask contrast varies per trial (20% high vs subthreshold); SOA varies. Uses natural scenes not faces.

Evaluates the test-retest reliability of binocular rivalry dynamics (dominance phase durations, mixed percept proportions, and derived ocular dominance index). Finds that rivalry-based ocular dominance indices show moderate to good reliability, supporting their use as individual difference measures and as calibration reference points.

Zenodo dataset (DOI 10.5281/zenodo.8043798) for Prosper et al. (2023, Vision Research, DOI 10.1016/j.visres.2023.108278) on monocular deprivation effects on binocular rivalry. Per-participant CSV files organised as data_raw/participant_##/rivalry/deprivation_duration/file.csv for 25 participants with columns key, time (seconds), eye (-1=left, 1=right, 2=mixed). Also includes dominant_eye per participant. Data collected via passive keypress rivalry paradigm (Gabor gratings, mirror stereoscope). Includes Python analysis_notebook.ipynb and analysis_functions.py.

Six experiments using predefined exposure durations (rather than free detection RT) show that emotional expressions do NOT alter detection sensitivity (d') to faces as they break through CFS. Prior findings of faster RT for fearful faces in standard b-CFS are argued to reflect decision criterion, identification, and response production differences rather than genuine perceptual sensitivity differences. Forced-choice localization, presence/absence detection, and staircase thresholds all yield null emotion effects on sensitivity.

EEG study (N=84) measuring SSVEPs during binocular rivalry with orthogonal gratings. Shows that the SSVEP amplitude of the suppressed stimulus gradually increases while the dominant stimulus amplitude decreases throughout a percept, and that longer percepts have more gradual signal changes. Changes are rapid and similar near perceptual transitions. SSVEP time courses are well predicted by a noisy accumulation model where perceptual transitions occur when accumulated signal crosses a threshold.

Introduces CFS tracking (tCFS), a paradigm in which suppressed target contrast steadily increases to breakthrough then decreases back to suppression repeatedly, measuring both breakthrough thresholds and suppression thresholds. Key finding: breakthrough thresholds differ across target types (replicating bCFS results) but suppression depth (breakthrough minus suppression threshold) is strikingly uniform. Argues suppression occurs early in visual processing and is not modulated by target salience or complexity. Challenges interpretations that bCFS category differences reflect stronger evidence accumulation.

OSF repository (https://osf.io/h6wnb, created 2024-07-21) containing Psychtoolbox/MATLAB CFS RT task code (CFS_RT.tgz, 59MB). Tags: Conscious, Continuous, Flash, RT, Suppression, Unconscious. Description: 'Matlab Psychtoolbox Codes for CFS RT experiment.'

OSF repository (https://osf.io/jes4u, created 2024-06-23) for a study on face-likeness and breaking time in b-CFS using Mooney faces and face inversion effects. Contains Facelikeness.zip (14MB, stimuli or full data) and two large CSVs: kerr_results_withcontrol_nopixel.csv (4MB) and kerr_results_nocontrol_withpixel.csv (4MB). Tags include #bCFS, #face_likeness, #inversion_effect, #Mooney_face. The large CSV size suggests trial-level per-participant data, though column structure was not confirmed (CSV not accessible via direct download in this session due to OSF API returning only first file in listing).

Introduces reverse-b-CFS (rev-bCFS): measuring the time a visible (dominant) stimulus takes to be suppressed from awareness, complementary to standard b-CFS (measuring time for a suppressed stimulus to break into awareness). In two experiments with face stimuli, rev-bCFS captures conscious processing effects (e.g., face inversion slows suppression in rev-bCFS), while b-CFS shows larger inversion effects that include an additional unconscious component. Contrasting b-CFS with rev-bCFS isolates the specifically unconscious contribution to awareness access.

GitHub repository (https://github.com/hansvoll/binocularRivalry, GPL-3.0 license) combining PsychoPy rivalry code and Oculus Rift VR-based CFS. Updated 2025-12-21. Implements both binocular rivalry and CFS in PsychoPy/Python.

GitHub repository (https://github.com/Mudrik-Lab/CFSVM, MIT license, MATLAB) by Mudrik Lab. Description: 'Continuous Flash Suppression and Visual Masking experiments in MATLAB.' Updated 2025-12-18, 3 GitHub stars, 1 fork. Provides a MATLAB toolbox for running b-CFS experiments. MIT license permits unrestricted reuse.

OSF repository (https://osf.io/xpy7r, created 2025-10-20) for an amblyopia binocular rivalry study. Contains a data folder, analysis_codes folder, experimental_codes folder, and Preprint.pdf (1.8MB). Description states: 'We share preprint, experimental and analysis codes and data in this project repository.' Binocular rivalry dynamics were compared between amblyopic observers and controls, making per-eye dominance a central variable.

OSF repository (https://osf.io/f9gez) for Drew et al. (2026, European Journal of Neuroscience). Contains EEG and behavioural raw data (Behavioural folder in w5t2u), MATLAB/Fieldtrip analysis code (02_Analysis.zip in p3cta), and binocular rivalry paradigm code (BR_conflict_parametrization.zip in ngdxu). Paradigm was run with stereoscopic goggles, includes green-channel isoluminance calibration step. Uses MATLAB Fieldtrip toolbox for EEG preprocessing. Paradigm is passive rivalry (perceptual alternation recording), not b-CFS.

OSF repository (https://osf.io/5gzc8, last modified 2026-05-02) containing per-participant trial-level CFS data and R analysis script. Individual CSV files per participant with columns: date, participant, intSeq, intFlip, intGrating, intEye (which eye), contrastB, contrastR (per-eye contrast values per trial), timeB, timeR (per-eye timing per trial), typeFlip, intFlipped. The contrastB/contrastR columns indicate adaptive per-trial per-eye contrast adjustment. No authors identified from OSF API (admin contributor data not returned).

Critical review of the breaking continuous flash suppression (b-CFS) paradigm: its origins (Jiang, Costello & He 2007), the rationale for using breakthrough/suppression time as an index of unconscious processing, methodological pitfalls (between-subject variability, response criteria, low-level confounds), and an evaluation of what b-CFS studies can and cannot tell us about unconscious visual processing. Surveys the rapid growth of b-CFS publications since Tsuchiya & Koch (2005) and offers recommendations for design and interpretation going forward.

Defines the family of transformed up-down (staircase) adaptive procedures for sensory threshold estimation. Demonstrates that interleaving runs with specific reversal rules converges to precise percent-correct points on the psychometric function (70.7% for 1-up/2-down). The abstract highlights simplicity, high efficiency, robustness, and small-sample reliability as key advantages over fixed-level designs.

Introduced the continuous random walk (diffusion) model of decision making applied to memory recognition paradigms. Evidence is accumulated in parallel from probe-memory comparisons, modeled as a continuous random walk process terminating at a threshold. This paper established the mathematical framework (Wiener diffusion with drift, two absorbing boundaries) that became the DDM used for RDM paradigms.

Introduces QUEST (Quick Estimation by Sequential Testing), a Bayesian adaptive method that updates a prior over the psychometric threshold after each trial and places each new stimulus near the current threshold estimate. QUEST finds thresholds in fewer trials than staircases by optimally exploiting prior knowledge about psychometric function shape. (Abstract not available in OpenAlex; based on title, Perception and Psychophysics 1983, 2505 citations.)

Presents the Psi adaptive Bayesian method that jointly estimates threshold and slope of the psychometric function by selecting each stimulus to minimise expected entropy over the joint posterior. Unlike QUEST, Psi also characterises the psychometric function steepness, providing a complete characterisation of sensitivity. (Abstract not available in OpenAlex; based on title, Vision Research 1999, 699 citations.)

Reviews adaptive psychophysical procedures (simple up-down, transformed up-down, PEST, QUEST, maximum-likelihood methods) with practical guidance on choosing method, step size, and number of trials. Compares convergence properties and accuracy of threshold estimates across methods. (Abstract not in OpenAlex; based on title, Perception and Psychophysics 2001, 843 citations.)

Rigorous treatment of fitting psychometric functions via maximum likelihood, covering goodness-of-fit, the lapse-rate parameter, and sampling issues. Shows that ignoring lapse rates systematically biases threshold estimates, especially at high coherence. Recommends including a small upper-asymptote lapse parameter (lambda approximately 0.01-0.05) as standard practice. (Abstract not in OpenAlex; based on title, Perception and Psychophysics 2001, 2430 citations.)

Shows that six major 2AFC decision models all reduce to the DDM, which implements the statistically optimal SPRT. Proves there is always an optimal speed-accuracy tradeoff that maximises reward rate, and that this optimum is achieved by adjusting the decision boundary rather than the drift rate. Explicitly addresses optimal tradeoff between speed and accuracy maximising various reward functions.

Develops an attractor-based framework for bistable perceptual switches in which alternations are driven by noise rather than slow adaptation. Accounts for Levelt's Proposition IV (faster alternation rate with increasing stimulus strength) and predicts that increasing stimulus strength changes activity levels of dominant and suppressed populations in measurable ways. Implemented in both firing-rate mean-field and spiking networks.

Comprehensive review and empirical demonstration of the Diffusion Decision Model (DDM) for two-choice tasks. Shows how behavioral data — accuracy, mean RT, and RT distributions — decompose into drift rate (v, quality of information), boundary separation (a, response caution), and non-decision time (t0). Three experiments illustrate: stimulus difficulty affects v; speed-accuracy instructions affect a; stimulus proportions bias v and starting point z.

Identifies common dynamical properties across multiple neuronal competition models of binocular rivalry, regardless of architecture or gain function. Shows that all competition models predict a non-monotonic relationship between input strength and dominance duration: at low strength dominance increases with strength (unlike Levelt's Prop IV), while at higher strength it decreases. Concludes that Levelt's Prop IV needs refined experimental tests.

Presents the EZ-diffusion model: three closed-form equations that transform three observables (mean RT, variance of RT, response accuracy) into three DDM parameters (drift rate v — 'quality of information'; boundary separation a — 'response conservativeness'; non-decision time t0). Designed for data-sparse situations enabling individual-subject analysis without full distribution fitting. Monte Carlo simulations confirm parameter recovery and robustness.

Examines the relative roles of noise and adaptation in bistable switching by studying models where rhythmogenesis mechanism can be varied from adaptation-driven to noise-driven. Finds that to match observed dominance duration distributions (mean and coefficient of variation) and serial correlations, models must operate near the boundary between adaptation-driven and noise-driven regimes — both mechanisms are required.

Presents HDDM, a Python toolbox for hierarchical Bayesian estimation of DDM parameters. Simultaneously estimates individual subject parameters and group distributions, providing posterior uncertainty. Handles outliers, supports trial-by-trial covariates (e.g., fMRI), and requires fewer trials per participant than non-hierarchical methods. Parameter recovery studies show HDDM outperforms chi-square quantile and MLE methods.

Annual Review overview of sequential sampling / diffusion decision models in cognitive neuroscience. Explains how DDM parameters (drift rate v, boundary a, non-decision time t0) decompose task performance and are linked to neural mechanisms. Reviews applications including RDM paradigms in monkey and human studies where subthalamic nucleus and pre-SMA are linked to boundary parameter a.

Reviews current status and open issues in the DDM, covering parameter estimation methods, model comparison (DDM vs race models, LCA), neurophysiological grounding, and extensions. Discusses failure modes (contaminant RTs, starting-point variability, intertrial drift variability), benchmarks for parameter ranges across tasks, and the role of each parameter in generating RT distributions. No abstract in OpenAlex or Crossref; record based on title and citation context.

Extends psychometric function estimation from a binomial to a beta-binomial model to handle overdispersion resulting from non-stationarity in observer behaviour. The Bayesian implementation (psignifit 4) yields accurate credible intervals even when observers shift criterion or attention across a session. Explicitly addresses observer non-stationarity as a source of overdispersion that standard binomial models cannot handle.

Shows that instructional SAT cues and response deadlines produce different RT-accuracy tradeoff functions and different DDM parameter profiles. In three experiments, deadline manipulation shifts primarily the boundary, while instructional emphasis also affects non-decision time and sometimes drift rate. The distribution of choices and RTs depends on specific task instructions, not just the generic speed-accuracy instruction.

Simulation study comparing optimization criteria (chi-square multinomial, maximum likelihood, G-square, KS) and design choices for estimating diffusion-model (DDM) parameters as a function of the number of trials per condition. Examines how parameter recovery bias and variability change with trial counts ranging from very low to very high, providing practical guidance on how many trials are needed for adequately precise DDM parameter estimation and which optimization criterion performs best at low trial counts.

Proposes a hierarchical model of binocular rivalry in which independent elements accumulate visual evidence at one level while groups of coupled elements compete for dominance at another — with feedback inhibition once a group dominates. Shows rivalry alternation statistics satisfy normative constraints of continuous decision-making: integrating evidence over space and time, choosing categorically, and concurrently evaluating alternatives. Argues multistable perception may reflect a decision process in a volatile world.

Directly compares motion discrimination using random-dot stimuli under two conditions: (1) unambiguous (monocular RDM) and (2) binocular rivalry conditions using the same dots. Accuracy and speed are similar for monocular conditions, consistent with a race between two monocular diffusion processes, but reaction times increase under rivalry, revealing competitive interactions. Proposes a combined model incorporating both diffusion-model accumulation and rivalry-model competition elements to account for RTs under both ambiguous and unambiguous conditions.

Uses random dot motion (RDM) paradigm to show that subjective awareness of having reached a decision corresponds to the DDM accumulator crossing its bound. Participants adjusted a clock to indicate when they first felt they had made up their mind; subjective decision times (tSDs) were faster for stronger motion and fit a bounded DDM. DDM parameters from tSDs alone predicted choices (accuracy) for 4 of 5 participants, providing strong evidence that consciousness = threshold crossing. Proposes this mechanism may govern how non-conscious processes pierce consciousness more broadly.