Binocular rivalry without continuous perceptual report

Citation: Frässle, S., Sommer, J., Jansen, A., Naber, M. & Einhäuser, W. Binocular Rivalry: Frontal Activity Relates to Introspection and Action But Not to Perception. Journal of Neuroscience 34(5), 1738–1747, 29 January 2014. DOI: 10.1523/JNEUROSCI.4403-13.2014. Published, peer-reviewed article; complete PMC version, PMCID PMC6827584.

Reading and provenance

Read on 29 September 2026. Read the complete archived PMC article: all 1,278 extracted lines, including front matter, abstract, introduction, Methods, Results, Discussion, all three figure captions, both complete numerical tables and their captions, footnotes, declarations and bibliography. Visually inspected all three original figure images. Both tables were additionally rendered from the complete PMC HTML and visually inspected; these are faithful data renders, not images of the publisher's page layout. No standalone display equations occur; the transformations and contrast definitions in the Methods were read in full.

No written scientific supplement is cited or linked in the complete article or its PMC archive, and no article-specific supplement was identified in targeted discovery. The publisher website and PDF routes returned access errors, and the PMC PDF link returned a challenge page. These failures concern alternate formats of an article available in full as HTML, rather than a known unread scientific supplement. The independent publisher-page package check remains unavailable; do not turn “none identified” into an absolute claim that no additional file could exist. All scientific content available in the official full-text archive was read. Referenced papers, commentaries and search results were not opened as additional evidence.

Source HTML, extracted text, three figure images, two table data/render sets, access-failure responses and exact acquisition details are under papers/consciousness_connectomics/frassle2014_no_report*. The provenance manifest records URLs, sizes, SHA256 hashes and reading boundaries. No raw-data reanalysis was performed.

Question and experimental contrast

Does frontal BOLD activity associated with a spontaneous change in binocular-rivalry perception remain when observers no longer report every perceptual switch? The authors operationalize perception using two ocular proxies, allowing event-related fMRI analysis during passive viewing.

The crucial endpoint is switch-related activity for rivalry relative to physical replay, and how that difference changes with reporting. It is not all frontal activity, an overall conscious-versus-unconscious contrast, or a direct test of whether frontal neurons are necessary for experience.

Twenty healthy, right-handed participants aged 19–30 took part, including one author; the other nineteen were naive to the study's purpose. The within-person design crossed two stimulus/proxy types, rivalry versus replay, and active report versus passive viewing. There were 32 trials of 60.9 seconds: four repetitions of each of the eight conditions. Two sessions separated the stimulus types, whose order was balanced across participants.

• Motion rivalry / optokinetic nystagmus (OKN): large red and green gratings moved in opposite horizontal directions. The slow phase of the eye movement was used to infer the dominant direction. The stimulus subtended approximately 30 × 22.2 degrees and moved at 22.3 degrees/second.
• Luminance rivalry / pupil size: stationary, differently oriented red and green gratings also differed substantially in luminance. A smaller pupil indicated dominance of the brighter grating. The circular aperture was 14.7 degrees in diameter.
• Active report: participants continuously held one of two buttons with the right hand, switching buttons when the dominant percept changed. They had to choose the more dominant percept even during mixed, spatially piecemeal rivalry; mixed states were not separately reported.
• Passive viewing: the report requirement was removed. The ocular proxies supplied the event estimates. This instruction did not measure or guarantee the absence of attention, introspection or spontaneous self-monitoring.
• Replay: both eyes received the same, physically switching stimulus. Switch times reproduced button presses from the preceding active-report rivalry trial. Both replay conditions used this same source sequence. Replay did not reproduce gradual, piecemeal rivalry transitions.

Within each repetition the order was fixed: active rivalry, active replay, passive rivalry, passive replay. Consequently report condition was not independently randomized in time, and passive replay was matched to preceding active, rather than simultaneous passive, rivalry dynamics.

How the ocular proxies were constructed and checked

Right-eye video was recorded at 60 Hz. Pupil location supplied relative eye motion; the authors did not calibrate each person's gaze to display coordinates because absolute position was unnecessary. OKN fast phases and blinks were removed, gaps linearly interpolated, and slow-phase velocity smoothed over 500 ms. A zero-crossing counted as a transition only after sufficient preceding velocity, with adjacent events at least 400 ms apart.

Pupil area was standardized within each trial, blink-interpolated, smoothed over 167 ms and detrended with a 4.167-second sliding average. Transition detection required a preceding excursion of at least 0.3 standard deviations and the same 400-ms separation. These are processed physiological indicators, not unfiltered readouts of subjective experience.

During active-report rivalry, proxy traces changed around reports in the expected directions. The authors also optimized each individual's proxy-to-report latency over a 0–2-second range and measured time spent in agreement:

All four values exceeded the stated 50% chance level at p < .001. These are the paper's reported mean ± uncertainty values; the result is substantially stronger for OKN than pupil size during rivalry. Latency was selected using the same report data, without a described held-out validation set. Agreement with reports cannot independently validate every no-report percept, particularly because passive viewing changes the task context.

The optimal proxy transitions preceded reports by 0.52 ± 0.07 seconds for OKN and 0.42 ± 0.12 seconds for pupil size. In replay, the physical switch preceded OKN by 0.46 ± 0.03 seconds and the report by 0.60 ± 0.02 seconds; corresponding pupil values were 0.30 ± 0.11 and 0.61 ± 0.10 seconds. Thus a proxy leading a button press is not evidence that the proxy precedes awareness: the actual subjective switch time during rivalry is unknown.

Behavioral and imaging results

Reporting changes the measured process. During active-report trials, median dominance periods inferred from buttons were longer than those inferred from the ocular proxies: 1.78 versus 1.57 seconds for OKN trials, and 1.74 versus 1.48 seconds for pupil trials. The authors suggest brief episodes may affect ocular responses without crossing the threshold for manual report; proxy errors or differing sensitivity also matter when interpreting this discrepancy.

Comparing the same proxy across task conditions, passive viewing increased the median duration for stationary/pupil rivalry from 1.48 to 1.57 seconds (p = .004). Motion/OKN rivalry was essentially unchanged, 1.57 versus 1.58 seconds (p = .79). The finding therefore supports a reporting-related change for one stimulus set, not an established universal acceleration of rivalry by reporting. Attention, action and monitoring are possible contributors.

Active-report maps reproduce frontal involvement. Using button-derived switch times, rivalry relative to replay recruited a predominantly right-sided occipital, parietal and frontal pattern. Table 1 includes right middle frontal (MNI 38, 36, 32; z = 4.32), inferior frontal and superior frontal peaks. Maps used a voxel threshold p < .001 uncorrected and a cluster-extent threshold p < .01 familywise-error corrected.

Continuous proxies yield stronger statistical contrasts. Discrete ocular and button events produced broadly similar maps. Using continuous ocular traces increased statistical contrast strength in selected occipital, parietal and middle-frontal areas. To construct these regressors, the authors took absolute proxy values, inverted them so transitions became peaks, resampled to the 1.45-second scan interval, and convolved them with a canonical hemodynamic response. “Higher power” here describes stronger results in the fitted analyses; it is not a prospective power calculation or independent validation of the inferred perceptual trajectory.

The central report comparison is a difference of differences:

(active rivalry − active replay) − (passive rivalry − passive replay)

Both sides use the continuous ocular measures. This reduces an explanation based solely on adding a button press, because active rivalry and active replay both require reporting. It does not isolate introspection from attention, task difficulty, decision processes, action–perception interactions or transition-shape differences.

During passive viewing, the rivalry-versus-replay map retained occipital and parietal responses, and the Results explicitly report right inferior orbitofrontal activity. Middle-frontal activity was no longer significant. Accordingly, the abstract's broad statement about absent frontal activity should be narrowed to the implicated differential response, particularly middle frontal cortex; the paper does not show all frontal activity disappearing.

The direct active-versus-passive comparison identified bilateral middle-frontal differences, together with supplementary motor, insular, parietal and visual regions. Table 2 reports left middle frontal (−36, 32, 24; z = 4.24) and right middle frontal (40, 40, 34; z = 3.98). This comparison used a global-null conjunction across the two stimulus types, p < .001 uncorrected, with reported clusters exceeding 20 voxels. Its threshold differs from the corrected single-condition maps. Rejecting a global conjunction null is also not equivalent to separately establishing the effect for each component condition; the authors additionally report qualitative consistency between stimulus types.

What follows, and what remains uncertain

The strongest supported conclusion is that part of the conventional switch-related frontal BOLD response depends on the reporting context. It is unsafe to label every response found in an active-report rivalry task a neural mechanism that generates conscious perceptual changes.

The authors go further, arguing that frontal areas chiefly monitor rather than initiate switches. Their closing paragraph allows roles in other forms of rivalry. For this library, the causal interpretation needs further qualification:

1. No regional intervention: removing a task requirement does not selectively disable frontal circuitry. The study does not demonstrate that frontal cortex is causally dispensable or that posterior cortex alone suffices.
2. Difference versus absence: no significant rivalry-minus-replay effect can coexist with activity shared by rivalry and replay, sustained activity, weak or spatially mixed signals, or other information not captured by the GLM. No equivalence margin or Bayesian evidence for a frontal null is reported.
3. Monitoring is not isolated: passive viewing bundles reduced reporting demands with possible changes in attention, effort, expectation and introspection. The paper acknowledges attention as a possible explanation for altered rivalry dynamics. Fixed trial order adds a temporal confound.
4. The proxy bridge is conditional: active-report calibration supports its use, but passive perception is not independently verified at every inferred transition. Pupil/report agreement is modest, and preprocessing and continuous-regressor assumptions shape which events drive the fMRI estimate.
5. Mixed perception and replay: the forced binary report omits the extent of piecemeal rivalry. The authors argue this affects report conditions similarly, but do not directly measure that invariance. Abrupt replay and endogenous gradual transitions need not have equal task or attentional consequences.
6. Measurement and sample limits: twenty young adults, one author-participant, millimeter-scale BOLD with 6-mm spatial smoothing, two specific stimulus classes, and an uncorrected central interaction constrain generalization. This is an experiment about rivalry dynamics within already conscious participants, not the presence versus absence of consciousness as a whole.

Relation to Cogitate, connectomics and the project

Frässle removes continuous reports of the changing percept and estimates those changes from ocular physiology. Cogitate instead varies the relevance of clearly visible stimulus categories while participants remain engaged in a target-detection task; its task-irrelevant nontarget analyses do not create the same passive-viewing comparison. Frässle's report-calibration problem and Cogitate's task/visibility assumptions are related methodological issues, but the experiments do not measure an identical contrast. Neither makes task irrelevance synonymous with lack of attention.

For connectome-constrained modeling, this paper motivates treating task instructions, report generation and internal state as part of the experiment a model must reproduce. A structural network fitted only to a report-linked signal might reproduce monitoring or action-related computation rather than the intended perceptual process. That is a research implication, not a connectomic result: the paper reconstructs no synaptic map, tests no emulator, and provides no evidence about machine experience or personal continuity.