{"id":"40167a5a-5603-4946-9ebe-3fd2d1561828","arxiv_id":"2501.05378","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A two-smartphone setup records the readiness potential before basketball free throws, but the signal does not distinguish successful from missed shots.","lead":"Researchers tested whether two off-the-shelf smartphones could record brain waves and body movements at the same time while people shot basketball free throws. The setup did capture the brain's preparation signal before each shot, but that signal did not predict whether a shot went in.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Movement-onset validity is the load-bearing risk: the paper's own PLD validation shows left-hand/hip motion, not right-wrist motion, at the accelerometer-derived t=0, so the RP and pose results may be aligned to the wrong event. An independent video or motion-capture onset marker should settle this.","rationale":"I read the paper in good faith. The strongest claim is feasibility of a two-smartphone, head-mounted-EEG setup for capturing brain dynamics during natural movement, and the main supporting evidence is a statistically significant readiness potential at Cz and other fronto-central channels. The paper deserves credit for transparent reporting: FDR-corrected tests, a null performance-RP result, a movement-onset validation attempt, and public MATLAB code and raw data (except video). Those supports do not, however, remove the need for a valid time zero. The onset detection is the hinge: the accelerometer threshold defines the epoch window, and the validation section shows that the most consistent PLD motion at that time zero is in the left hand and hips, not the right wrist that generated the threshold. That is an internal inconsistency, not merely a departure from consensus. If the onset marker is a different event, the RP could still be a genuine readiness potential preceding that event, but the paper's specific claim about readiness potentials for basketball free-throw shooting would be misaligned, and all performance-aligned pose statistics would inherit the same offset. My proposed check, re-epoching to a video-derived onset and measuring sensor latency directly, would settle whether the concern lands. Because this is exactly the condition the reader identified and because the requested validation is feasible with existing data, I do not move the verdict; the paper should remain conditional pending that validation.","tokens_in":24283,"tokens_out":8830,"duration_ms":91579,"concrete_test":"Using the already-recorded synchronized video (or a new motion-capture session), define an independent onset per trial as the first video frame in which the ball or the right wrist begins its shooting-relevant displacement (or the frame of ball release). Re-run the RP analysis with the same preprocessing and statistics but with epochs aligned to this video-derived onset instead of the accelerometer threshold, and quantify IMU-to-EEG latency with an impulse test (e.g., tap the wrist sensor while EEG and PLD are recording). If the significant Cz negativity from -400 to 0 ms persists with unchanged timing and morphology, the accelerometer onset is validated; if it shifts, smears, or disappears, the reported RP is not specifically locked to the free-throw movement.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central empirical support is the readiness potential (RP) at fronto-central channels before 'movement onset' (Results, Presence of the RP; Fig. 4C). Every EEG epoch, pose window, set-point latency, and hit/miss correlation is time-locked to t=0 determined from the right-wrist accelerometer threshold (baseline mean + 1 SD; Materials and Methods, Onset detection), so the validity of that t=0 is load-bearing. The paper's own validation (Results, Movement onset validation) does not confirm it: at the accelerometer-defined onset, the body parts with significant motion across participants were the left index finger, left pinky, left thumb, left wrist, and both hips, not the right throwing hand/wrist. Since the right wrist is the very sensor whose threshold defines t=0, the IMU event could be a preparatory/postural adjustment, a ball-handover movement, or a cross-device synchronization delay rather than shot initiation. The reported RP morphology could survive such a shift if the same non-throwing movement precedes each shot by a roughly constant interval, so the presence of an RP-like waveform is not itself evidence that the onset marker is correct. The authors list 'minimal delays in synchronization' as a limitation but do not quantify them, and no independent ground-truth onset is reported. This makes the central feasibility claim, capturing the RP for basketball free-throw shooting, conditional rather than established; the portable-setup conclusion may still be true, but the shot-locked interpretation is not yet secured.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a portable, low-cost mobile brain/body imaging setup consisting of two Android smartphones, a wireless 32-channel EEG amplifier, and a wrist-worn IMU, used to record brain activity and full-body pose while 26 basketball players performed 120 free throws each. The authors analyze (i) the readiness potential (RP) preceding the throwing movement, (ii) whether RP amplitude differs between successful and unsuccessful shots, and (iii) whether pose landmarks differ between hits and misses. They report a significant negative deflection at Cz and adjacent fronto-central channels from -400 to 0 ms before their accelerometer-derived movement onset, no significant relationship between RP amplitude and shooting success, and exploratory pose differences in 10 of 26 participants with small explained variance. The paper concludes that the setup is sufficient to monitor complex human movement and associated brain dynamics outside the laboratory.","tokens_in":24549,"tokens_out":3699,"duration_ms":37193,"significance":"If the movement-onset alignment is valid, this is a useful methodological contribution to mobile EEG and MoBI research. The setup is genuinely portable, low-cost, and lightweight; the authors provide open code and data availability; and the statistical reporting is largely transparent, including FDR corrections, effect sizes, and an honest presentation of null results for the RP-performance relationship. The pose analysis is explicitly exploratory, and the small effect sizes are not overstated. The central risk is the validity of the accelerometer-derived movement onset, which the paper's own validation does not confirm. Because every EEG epoch, pose window, set-point latency, and performance correlation is time-locked to that event, this issue is load-bearing for the main feasibility claim. The paper would be strengthened substantially by independent validation of the onset marker, for example via video annotation, EMG, or a second motion sensor on the ball.","major_comments":[{"comment":"The validation of the movement-onset marker does not confirm that the right-wrist accelerometer threshold identifies the onset of the throwing movement. The binomial tests across participants show significant motion at the accelerometer-defined t=0 in the left index finger, left pinky, left thumb, left wrist, and both hips, but not in the right throwing wrist. Since the right-wrist IMU defines t=0 for all subsequent analyses, this result suggests that the threshold may be capturing a preparatory postural adjustment, a ball-handover movement, or a cross-device synchronization artifact rather than the initiation of the free-throw. The RP-like morphology could still arise if the same non-throwing movement consistently precedes each shot by a roughly constant interval, but in that case the paper's central claim of capturing the readiness potential for the basketball free-throw is not established. The authors should validate the onset against an independent ground-truth marker (e.g., manual video annotation, EMG, or a second sensor on the ball) or, at minimum, quantify and discuss the temporal offset between the detected event and the actual throwing onset.","section":"Results, 'Movement onset validation'; Figure 3"},{"comment":"The threshold-based onset detection procedure is described only qualitatively. No trial-to-trial or participant-level variability of the detected onsets is reported, and the set-point latency is given only as a group mean (544 ms after onset) without a range or standard deviation. In addition, the Limitations section acknowledges 'minimal delays in synchronization' (citing Iwama et al., 2024) but does not quantify the jitter or latency between the EEG, PLD, and IMU streams, despite the fact that all streams were aligned via linear interpolation on EEG timestamps. Because the entire analytic pipeline depends on the temporal alignment of these streams, the authors should provide quantitative synchronization error estimates and assess the stability of the onset detection across trials and participants.","section":"Materials and Methods, 'Onset detection'; Limitations"}],"minor_comments":[{"comment":"The Methods state that the presence of the RP was evaluated using t-tests, but the Results report Wilcoxon signed-rank tests; please reconcile this discrepancy.","section":"Materials and Methods, 'Presence of the RP' vs. Results, 'Presence of the RP'"},{"comment":"The abstract reports effect sizes as r = 0.50 to 0.77, but the Results do not explain how these are derived from the reported z-values; please specify the computation (e.g., r = z / sqrt(N)).","section":"Abstract and Results, 'Presence of the RP'"},{"comment":"The description of the binomial tests is incomplete: it is not stated whether the test is performed on the proportion of participants showing a significant Wilcoxon effect at each body part, and the correction for multiple comparisons across body parts is not explained in enough detail to be reproduced.","section":"Results, 'Movement onset validation'"},{"comment":"The sentence 'Inspection of the PLD signals confirmed that hand/wrist motion was among the first body parts involved in initiating a basketball free-throw motion sequence' appears to contradict the statistical validation reported in the Results, which found the most consistent movement at the left hand and hips; please clarify or qualify this statement.","section":"Discussion, 'Movement onset validation'"},{"comment":"The reference list contains an entry for 'Hayes, M.H.S. and Patterson, D.G. (1921)' that is not cited anywhere in the text; please remove it or cite it appropriately.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the movement-onset validation. The paper's own validation undermines the assumption that t=0 is the onset of the throwing movement, and this is load-bearing for the main feasibility claim. The concern is not insurmountable: an independent video-based or EMG ground-truth onset test, or a clear demonstration that the detected event has a constant temporal relationship to the true onset, would resolve it. The rest of the paper is transparently reported and the exploratory nature of the pose analysis is appropriately indicated. I would recommend major revision rather than rejection because the setup and data are potentially valuable to the MoBI community, and the flaw is addressable within the paper's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nShort version: this is a useful feasibility study. The authors show that two off-the-shelf smartphones—one streaming wireless EEG, one running MediaPipe pose tracking—plus a wrist-worn IMU can record a recognizable readiness potential (RP) in a natural basketball free-throw setting. The RP result is credible: FDR-corrected negative deflection at Cz and adjacent channels from -400 to 0 ms, with reasonable effect sizes. The null result on RP amplitude vs. shot outcome is reported honestly, with effect sizes rather than just p-values. The pose-performance analysis is explicitly exploratory, and the effects are small (max mean R² ≈ 0.045), which the authors acknowledge.\n\nWhat's genuinely new: the specific two-smartphone setup, and the first attempt to compare RP between successful and unsuccessful free throws. That combination isn't in the prior literature.\n\nThe soft spot is the movement-onset definition. Every EEG epoch and pose correlation is time-locked to a t=0 defined by the right-wrist accelerometer threshold (baseline mean + 1 SD). But the paper's own validation shows that at this t=0, the significant body-part movements are in the left fingers, left wrist, and hips—not the right throwing hand. That's a red flag, because the sensor is on the right wrist. It may be that t=0 captures an early postural adjustment or support-hand movement rather than shot initiation. The RP morphology doesn't settle it: a slow negative drift would survive if the same non-throwing movement precedes every shot by roughly the same interval. The authors list synchronization delays as a limitation but never quantify them. This makes the shot-locked interpretation conditional, not established.\n\nThe fix is straightforward: validate t=0 against an independent ground-truth marker—a video frame of ball release, a motion-capture system, or a second IMU on the throwing hand. The authors promise data and code on GitHub, so a referee can ask for that.\n\nOverall, worth a serious referee. The central feasibility claim is likely true, but the onset-validation gap needs addressing before the results can be taken at face value.","headline":"A credible mobile-EEG feasibility study with a load-bearing but fixable movement-onset validation gap.","tokens_in":25119,"tokens_out":3569,"would_cite":true,"duration_ms":35375,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Two off-the-shelf smartphones plus a wireless EEG cap can record the readiness potential during real basketball free throws.","keywords":["mobile EEG","readiness potential","basketball free throw","human pose estimation","MediaPipe","mobile brain/body imaging","inertial measurement unit","ecological validity"],"falsifier":"Take the same two-phone recordings and redefine movement onset using a high-speed camera or the first detectable motion of the right throwing hand rather than the wrist-accelerometer threshold; if the significant negative deflection at Cz disappears or shifts outside the -400 to 0 ms window, the claim that the portable setup captures the readiness potential is not established.","tokens_in":24061,"feed_emoji":"🏀","tokens_out":5263,"duration_ms":48222,"temperature":0.7,"pith_summary":"Twenty-six experienced basketball players shot 120 free throws each while wearing a 32-channel wireless EEG cap and a wrist-worn motion sensor, with two smartphones recording brain signals and full-body pose. The paper's central claim is that this pocketable, low-cost rig is enough to capture the neural preparation for a skilled whole-body movement in a real court setting. The main supporting result is a clear readiness potential, a slow negative voltage shift over fronto-central scalp, starting around 400 ms before the detected movement onset, with the strongest effect at electrode Cz. The paper also reports that readiness-potential amplitude did not distinguish successful from unsuccessful shots, while pose analysis showed participant-specific movement differences between hits and misses in 10 of 26 players. If correct, this establishes that ecologically valid, out-of-the-lab studies of brain-body coupling are feasible with consumer hardware.","feed_headline":"Phone rig records the brain's readiness signal before free throws","feed_subtitle":"A two-smartphone setup captured the readiness potential in 26 players, but the signal did not predict whether shots went in.","key_machinery":"The load-bearing tool is the synchronized recording chain. One smartphone receives 32-channel EEG at 250 Hz from a head-mounted wireless amplifier along with the amplifier's inertial sensors; a second smartphone runs MediaPipe Pose Landmark Detection at 15 Hz to track 33 body landmarks; and a wrist-worn IMU streams acceleration at 60 Hz. Lab Streaming Layer binds the streams into a single file. To time-lock brain activity, the analysis first detects a set-point when the right wrist crosses eye level on the pose stream, then walks backward through the wrist acceleration signal to find movement onset as the last sample below a threshold of baseline mean plus one standard deviation. EEG epochs from -2.5 s to movement onset are cleaned with ICA-based artifact removal and averaged to produce the readiness potential. The synchronization of EEG, pose, and IMU timestamps to that movement onset is what makes a comparison between neural preparation and subsequent movement outcome possible.","core_discovery":"The discovery this paper argues for is that a minimally invasive, low-cost two-smartphone system can simultaneously capture human pose and EEG well enough to observe a canonical brain signature of action preparation in a natural skilled task. Averaged across participants, the EEG shows a readiness potential over fronto-central channels, with a significant negative deflection at Cz from -400 to 0 ms before movement onset (mean -6.54 to -13.52 µV; z = -2.53 to -3.92; FDR-corrected p = 0.049 to 0.003; r = 0.50 to 0.77). However, the amplitude of the RP was not related to shooting outcome at the group or single-trial level (largest mean R² = 0.047, i.e., 4.7% explained variance), and only a minority of participants showed pose differences between successful and unsuccessful attempts. The authors present this as evidence that the setup is sufficient for monitoring movement and brain dynamics outside the laboratory, not that the RP forecasts performance.","pith_inferences":["The null RP-performance relationship may be a timing artifact of the onset marker: if the wrist IMU threshold captures a preparatory crouch or ball hand-over rather than throw initiation, single-trial RP estimates would be systematically jittered, biasing point-biserial correlations toward zero; re-epoching around a video-verified release frame would test this.","The pose differences, lower wrist before the shot, higher wrist at set-point, and stable head, suggest a compact marker set for automated coaching feedback, but the small explained variance and individual specificity mean such feedback would need to be personalized rather than generic.","Because the setup is cheap and untethered, a natural extension would be to run the same rig with elite versus novice shooters; expertise differences in RP amplitude found in earlier work might reappear in between-subject designs even though within-subject trial outcome did not correlate."],"forward_implications":["Readiness potentials can be elicited and recorded during whole-body, goal-directed motor skills outside the laboratory, extending the RP beyond finger presses and isolated movements.","RP amplitude before a free throw does not predict make versus miss in this population; performance is better linked, for some individuals, to body posture during execution.","A two-smartphone rig with a wireless EEG cap is enough to align neural, inertial, and pose data streams in a natural setting, reducing cost and mobility barriers.","Pose-based set-point detection plus reverse thresholding provides a practical event marker for natural movements that lack a discrete button-press onset.","This supports further mobile brain-body imaging studies and possibly portable neurofeedback applications in real environments."],"supporting_citations":[{"why":"Defines the readiness potential, the brain signal the portable setup is designed to capture.","marker":"Kornhuber & Deecke, 1965"},{"why":"Demonstrated low-cost wireless EEG during walking, the foundational feasibility result for mobile EEG.","marker":"Debener et al., 2012"},{"why":"Supplies the Pose Landmark Detection model used for real-time full-body motion tracking.","marker":"Bazarevsky et al., 2020"},{"why":"Supplies the MediaPipe framework that runs the pose landmark model on the smartphone.","marker":"Lugaresi et al., 2019"},{"why":"Provides the Lab Streaming Layer Android apps that synchronize EEG, pose, and IMU streams into one recording.","marker":"Blum et al., 2021"},{"why":"Provides the reverse-computation threshold method used to estimate movement onset from the wrist accelerometer.","marker":"Verbaarschot et al., 2015"},{"why":"Prior report of the RP outside the laboratory during a whole-body movement, the closest benchmark for this study.","marker":"Nann et al., 2019"},{"why":"Discusses RP onset and morphology and the limits of trial-averaged estimates, invoked in interpreting the observed signal.","marker":"Schurger et al., 2021"},{"why":"Connects RP amplitude to expertise and quiet-eye mechanisms, the background for asking whether RP predicts performance.","marker":"Mann et al., 2011"}],"fun_headline_variants":["Two phones and a headset capture brain's pre-shot signal","Pocketable EEG rig sees brain's readiness potential in free throws","Low-cost phone setup records brain activity during real basketball shots","Brain signal before free throws captured by smartphone EEG rig"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the wrist-worn accelerometer's threshold, baseline mean plus one standard deviation, marks the true onset of the shot, because EEG epochs, pose epochs, and every performance correlation are aligned to that time zero; the paper's own validation found the most consistent motion at that instant in the left hand and hips, not the right throwing hand.","fun_headline_variants_meta":{"raw":{"variants":["Two phones and a headset capture brain's pre-shot signal","Pocketable EEG rig sees brain's readiness potential in free throws","Low-cost phone setup records brain activity during real basketball shots","Brain signal before free throws captured by smartphone EEG rig"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000328,"raw_usage":{"total_tokens":1922,"prompt_tokens":1125,"completion_tokens":797,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":741,"completion_tokens_details":{"reasoning_tokens":728}},"tokens_in":741,"tokens_out":797,"duration_ms":7614,"temperature":1.0,"reasoning_tokens":728,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:13:35.267712+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the same two-phone recordings and redefine movement onset using a high-speed camera or the first detectable motion of the right throwing hand rather than the wrist-accelerometer threshold; if the significant negative deflection at Cz disappears or shifts outside the -400 to 0 ms window, the claim that the portable setup captures the readiness potential is not established.","supporting_citations":[],"review_version":1}