{"id":"3baf6142-5db4-47fe-8239-733ac36d5e40","arxiv_id":"2507.02432","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Rhythmic, heart-rate-synced smartwatch vibrations increased perceived relaxation and showed weak parasympathetic trends, but had no significant effect on sleep onset latency.","lead":"This paper tests whether smartwatch vibrations that follow musical rhythms and adjust to your heart rate can help people relax and fall asleep. In two lab studies, the vibrations lowered heart rate and felt relaxing in short sessions, but did not measurably speed up sleep onset in a 20-minute nap attempt.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Study 1's heart-rate reduction is measured only against an initial baseline before any condition; without an interleaved no-vibration control, time/habituation effects could fully explain the positive result, so the central short-term claim is not yet established.","rationale":"The reader's conditional verdict is directionally correct, and the strongest concern is the missing contemporaneous no-stimulation control for Study 1's heart-rate result. The baseline-first design means the reported reduction could be a pure time artifact, and this directly threatens the positive half of the central claim. The reader's formal weakest_assumption focuses on the closed-loop adaptation using wrist-worn heart-rate sampling; that is a plausible secondary issue, but even a perfectly implemented adaptive loop would not rescue the positive claim without a proper control condition. The reader's rationale does mention 'weak baseline comparison,' so there is partial agreement. The proposed follow-up experiment would isolate the vibration effect from drift and habituation, and would settle the concern. I therefore recommend keeping the verdict CONDITIONAL, with the added requirement that positive short-term claims not be cited until a controlled comparison is available.","tokens_in":12152,"tokens_out":4598,"duration_ms":57656,"concrete_test":"Run a small within-subject follow-up (N≥20) using the same 5-minute protocol from Study 1, but interleave a silent no-vibration control condition among the vibration conditions in counterbalanced order, with the same one-minute walk/recovery before every segment. Compute each condition's heart-rate change relative to the immediately preceding rest window and test whether the vibration conditions differ from the silent control. If the vibration-minus-control difference is not significant, the heart-rate reduction in Study 1 is explained by time or rest effects and the positive short-term claim should be withdrawn or substantially weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Study 1's central positive result—that all five vibration patterns significantly reduced heart rate compared to baseline—is reported as vibration versus an initial 10-minute baseline taken before any condition (Section 2.1.4, 2.2.1). Because the baseline is always first, the comparison is confounded with time, habituation to the setting, and the natural effects of continued quiet sitting. There is no interleaved no-vibration rest condition and no sham-vibration control, so the heart-rate reduction cannot be attributed specifically to haptic stimulation. This is the load-bearing gap for the positive half of the central claim: even a perfectly working closed-loop adaptive vibration (Section 2.1.3) would not establish the claimed physiological effect without a proper contemporaneous control. The abstract's wording 'increased parasympathetic activity and perceived relaxation during short-term stimulation' further overstates the evidence: in Study 2 the parasympathetic claims rest on non-significant trends (RMSSD p=.07, HF p=.24, Section 3.2.1) and are not reported as significant between-condition contrasts. Thus the most load-bearing concern is not the adaptive-loop implementation itself, but the absence of a comparable no-stimulation condition against which the short-term effect can be evaluated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper describes a smartwatch-based haptic biofeedback system that encodes lullaby-inspired rhythmic patterns (tap/vibration sequences in 2/4 and 3/4 meters) and adapts vibration tempo to 4% below the wearer's current heart rate, updated every second. Study 1 (N=20) compares five patterns in a within-subject resting protocol against an initial no-vibration baseline, measuring heart rate, sleepiness (SSS), questionnaire ratings, and preference rankings. Study 2 (N=28, three excluded) tests the most preferred 3/4 pattern against a no-vibration control on separate days during a 20-minute sleep-onset attempt, with ECG-derived HR/HRV, a finger-tapping-derived sleep-onset latency, SSS, and PVT. The authors report that all five patterns reduced heart rate relative to baseline and that 3/4 and Alternating were rated more relaxing, while Study 2 found no significant effects on HR, HRV, sleep onset, or psychomotor performance, with only non-significant HRV trends.","tokens_in":12400,"tokens_out":7081,"duration_ms":77432,"significance":"If the short-term relaxation effect were convincingly established, the work would be a useful contribution to wearable haptics and passive biofeedback for relaxation: it introduces a commercially available smartwatch implementation, uses a two-stage design to select a pattern on the basis of preference and physiology, and reports an honest null result for sleep onset. The manuscript also gives a balanced discussion of limitations such as the first-night effect and smartwatch habituation. However, the central short-term claim is not currently supported because the heart-rate comparison in Study 1 lacks a matched no-vibration control, and the abstract overstates the parasympathetic evidence.","major_comments":[{"comment":"The central short-term finding—that all five vibration patterns significantly reduced heart rate compared with baseline—rests on a comparison to a single 10-minute baseline recorded before any condition was presented. Because this baseline is always first, the observed decline in heart rate is confounded with elapsed time, habituation to the setting, and the natural effects of continued quiet sitting. The procedure includes a one-minute walk followed by one-minute recovery before each condition, but it does not include an interleaved no-vibration or sham-vibration condition at matched times. The introduction to Study 1 states that the patterns were compared with a non-vibration control condition, but the procedure does not describe such a condition. Therefore the heart-rate reduction cannot be attributed specifically to haptic stimulation, and the positive half of the central claim is not yet established.","section":"Sections 2.1.4 and 2.2.1"},{"comment":"The abstract's statement that the results showed increased parasympathetic activity during short-term stimulation is not supported by the reported statistics. Study 1 measured only heart rate, not HRV, and Study 2's parasympathetic indicators were non-significant trends (RMSSD p=.07, HF p=.24, LF/HF p=.39). A trend is not evidence of increased parasympathetic activity; please rephrase the abstract and conclusion to state that no significant differences in autonomic measures were found, with only suggestive trends. The perceived-relaxation result is likewise limited to comparisons among vibration patterns, since no pre-vibration relaxation rating was collected for a no-vibration control.","section":"Abstract; Section 3.2.1"},{"comment":"Post-hoc Wilcoxon signed-rank tests are reported without correction for multiple comparisons. In Section 2.2.1 there are five vibration-versus-baseline comparisons at alpha=0.05; in Sections 2.2.3 and 2.2.4 there are ten pairwise pattern comparisons; and Study 2 runs multiple tests on HR, RMSSD, HF, LF/HF, SSS, SOL, and four PVT metrics. With this many tests, the reported p-values should be adjusted (e.g., Bonferroni-Holm or FDR), or the results should be explicitly labeled exploratory.","section":"Sections 2.2.1, 2.2.3, 2.2.4, and 3.2"},{"comment":"The sleep-onset measurement is underspecified. The text defines SOL as the time until the first finger-tapping interval exceeds 8 seconds, and the Discussion describes participants experiencing rhythmic feedback during the Finger Tapping Task, but the procedure in Section 3.1.2 does not state when or how participants performed the FTT during the 20-minute sleep phase, or which device they tapped. Because SOL is a central outcome of Study 2, the FTT protocol must be described explicitly.","section":"Sections 3.1.2 and 3.1.3"},{"comment":"The adaptive loop's premise—that wrist-worn optical heart rate updated every second and reduced by 4% produces the 'weak non-invasive forcing' entrainment described in prior work—is not validated. No data are reported on the accuracy of the HealthKit heart-rate samples used for adaptation, the actual delivered vibration intervals, or the congruence between the stimulation rhythm and the ECG reference used in Study 2. Please provide this validation or temper the 'closed-loop biofeedback' characterization to something like 'heart-rate-paced vibration.'","section":"Section 2.1.3"}],"minor_comments":[{"comment":"After excluding three participants with heart-rate spikes, the analysis sample is 25, not 28; please state the final N explicitly in the results.","section":"Section 3.2"},{"comment":"Please clarify whether the Friedman and Wilcoxon tests use per-participant median heart rate over the analyzed window, and how many heart-rate samples support each participant's median.","section":"Section 2.2.1"},{"comment":"The figure uses 'regulated' and 'none' as condition labels; please align them with 'vibration' and 'control' used in the text.","section":"Figure 6"},{"comment":"The explanation that shorter stimulation durations may be more effective is speculative because the study did not manipulate duration; please mark it as a hypothesis rather than an inference from the data.","section":"Section 4"},{"comment":"The 3–5% reduction value is attributed to both [2] and [17]; [2] appears to be a theoretical entrainment paper, so please clarify that the empirical nap-study evidence comes from [17].","section":"Section 2.1.3 and References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope for a CHI-style venue, and the novelty is adequate; the main issue is experimental control rather than novelty. I see no citation or disclosure problems, though the authors may wish to add a data-availability statement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful part of this paper is the sleep-onset null: in a 20-minute within-subject test with counterbalanced vibration/no-vibration days, the 3/4 haptic pattern did nothing to sleep latency or psychomotor measures. That is a clean, honest negative that people building haptic sleep aids should know about. The positive half of the abstract does not hold up. Study 1's five vibration conditions each reduced heart rate relative to a ten-minute baseline that always came first, before any condition. There is no interleaved rest condition or sham vibration, so time, habituation, and quiet sitting alone could fully explain the drop. The abstract also oversells the 'increased parasympathetic activity' in Study 2: RMSSD p=.07 and HF p=.24 are non-significant trends, not evidence.\n\nWhat is genuinely new: a closed-loop system that encodes lullaby meter patterns (2/4, 3/4) as smartwatch vibrations and adapts tempo to 4% below current heart rate, plus a preference ranking showing 3/4 and alternating patterns rated more relaxing than monotonous taps. The app and protocol are described well enough to rebuild. The discussion is candid about first-night effects, smartwatch familiarity, and the possibility that 20 minutes of vibration is overstimulating.\n\nSoft spots beyond the baseline issue: many Wilcoxon tests on five patterns without multiple-comparison correction; three participants excluded for 'unusually sudden high HR spikes' with no pre-registered rule; the 4% adaptation value is imported from a mattress-woofer nap study and wrist HR is not validated against ECG during the control loop. All fixable.\n\nThis paper is for HCI researchers and anyone prototyping wearable biofeedback for relaxation. It is not a clinical proof of an intervention. I would give it a serious referee because the design and null result are worth publishing after revision. That revision needs a contemporaneous no-vibration control, corrected statistics, and an abstract that matches the evidence. My recommendation: engage with it, but treat the short-term calming claim as a hypothesis, not a result.","headline":"The sleep-onset null result is a useful negative, but the short-term relax claim rests on a baseline artifact until a contemporaneous control is added.","tokens_in":12943,"tokens_out":3936,"would_cite":true,"duration_ms":49165,"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":"A smartwatch that vibrates slightly slower than your heartbeat can lower heart rate and feel relaxing, but one session does not speed up falling asleep.","keywords":["wearable haptics","smartwatch biofeedback","rhythmic vibration","heart rate entrainment","lullaby rhythm","relaxation intervention","sleep onset","parasympathetic activity"],"falsifier":"Continuously record ECG beat-to-beat intervals while the smartwatch delivers vibrations set 4% below the current heart rate, and compare heart-rate deceleration and vagal HRV markers against a sham condition with identical vibration intensity but a fixed tempo unrelated to heart rate; if no reliable difference appears in a pre-registered sample, the claimed entrainment effect is not supported.","tokens_in":11964,"feed_emoji":"😴","tokens_out":7287,"duration_ms":75345,"temperature":0.7,"pith_summary":"This paper asks whether a smartwatch can be a passive sleep aid by vibrating in lullaby-like rhythms whose tempo is continuously set just below the wearer's current heart rate. It claims that brief exposure to these closed-loop vibrations lowers heart rate and increases perceived relaxation, but that a single 20-minute session does not measurably shorten sleep onset or improve sleep-related physiological and behavioral measures. The result matters because it separates an acute calming effect from a genuine sleep-induction effect, suggesting that haptic biofeedback of this kind is not yet a reliable sleep-onset intervention. The paper's own conclusion is that rhythmic haptics may support momentary relaxation, while translation to sleep-specific outcomes needs further refinement.","feed_headline":"Vibrations synced to your heartbeat relax you, not hasten sleep","feed_subtitle":"Two studies: smartwatch haptics lower heart rate and feel relaxing, yet a single session leaves sleep onset unchanged.","key_machinery":"The load-bearing mechanism is a closed-loop biofeedback loop: an Apple Watch reads the wearer's heart rate through HealthKit once per second and sets the vibration tempo to 4% below that current value, an adaptation intended as 'weak non-invasive forcing' rather than the strong 20-30% adjustment used in anxiety-focused haptic studies. The haptic signal itself encodes lullaby time signatures, most importantly the alternating 3/4 waltz-like pattern made of taps and short vibrations. The system is built entirely on the watch, so the same device both senses heart rate and delivers stimulation, with no companion device required during the intervention.","core_discovery":"On the paper's own terms, the discovery is that closed-loop rhythmic haptic biofeedback delivered by a commercial smartwatch can produce short-term calming effects, but not enough to change sleep onset in a single session. In Study 1 (N=20), all five vibration patterns significantly reduced heart rate compared with baseline, and the alternating 3/4 pattern was rated most relaxing and ranked first by half of the participants. In Study 2 (N=28), the same 3/4 pattern during a 20-minute sleep attempt produced no significant differences in heart rate, heart-rate variability, subjective sleepiness, sleep-onset latency, or psychomotor vigilance, although RMSSD and high-frequency HRV trended toward increased parasympathetic activity. The paper therefore claims that tactile entrainment can lower arousal acutely, while its translation to sleep onset remains unestablished.","pith_inferences":["If the acute heart-rate drop is a genuine entrainment effect, it should be tested against a sham vibration at a fixed tempo unrelated to heart rate while controlling expectation; the paper does not include that control condition.","The finger-tapping task used to estimate sleep onset may itself interfere with the intervention by demanding cognitive engagement, so an EEG-based sleep-onset measure could reveal effects that the current protocol masks.","A testable extension is to stop the vibration automatically once sleep onset is detected or after a short fixed window, because participants reported that continuous 20-minute stimulation kept them alert.","Because non-regular smartwatch users showed stronger high-frequency HRV responses, a study balanced on wearable experience and measuring notification-related associations could sharpen or overturn the reported trends."],"forward_implications":["Short relaxation sessions of about five minutes with lullaby-rhythm haptics can reduce heart rate and improve perceived relaxation relative to no vibration.","Monotonous tap-only or vibration-only rhythms are judged less relaxing than alternating meters, so rhythmic variation is a relevant design feature for calming wearables.","A single 20-minute smartwatch vibration session should not be expected to shorten sleep onset; longer exposure, individual adaptation, or multimodal integration may be needed for sleep-specific benefits.","Closed-loop haptic biofeedback can run entirely on a commercial smartwatch with second-by-second heart-rate updates, enabling self-directed use without a researcher or companion device.","People who rarely wear smartwatches may show stronger parasympathetic responses than regular users, suggesting habituation or notification associations moderate the effect."],"supporting_citations":[{"why":"Establishes the 'weak non-invasive forcing' principle that a roughly 3-5% slowing of an external rhythm can influence heart rate, the basis for the 4% down-tempo rule.","marker":"[2]"},{"why":"Predecessor nap study showing closed-loop vibration stimulation can affect heart rhythm; its evidence motivates the subtle 4% frequency reduction used here.","marker":"[17]"},{"why":"Smartwatch haptic arousal-regulation system in foreground tasks, contrasted with the weaker, background-oriented adjustment of the present closed loop.","marker":"[19]"},{"why":"Demonstrates calming effects of wearable haptic stimulation in anxiety contexts, the paradigm the present work extends to relaxation and sleep onset.","marker":"[61]"},{"why":"Supplies the Multiple Sleep Latency Test protocol that the 20-minute eyes-closed sleep phase is modeled on.","marker":"[11]"},{"why":"Defines the finger-tapping criterion (first interval exceeding 8 seconds) used to estimate sleep-onset latency.","marker":"[13]"},{"why":"Provides the Psychomotor Vigilance Test used to assess alertness before and after the sleep attempt.","marker":"[22]"},{"why":"Shows behavioral sleep-onset measures converge with EEG wake/sleep markers, supporting the finger-tapping outcome.","marker":"[46]"}],"fun_headline_variants":["Heartbeat synced vibrations calm, but do not speed sleep","Smartwatch haptics lower heart rate, not sleep latency","Rhythmic haptics relax you, but sleep onset unchanged","Vibration patterns calm, but one session won't speed sleep"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole approach depends on the idea that a wrist vibration set a few percent slower than the wearer's current heart rate gently coaxes the heart to slow down, and that a smartwatch's heart-rate readings are accurate enough to run that feedback loop while someone is trying to sleep.","fun_headline_variants_meta":{"raw":{"variants":["Heartbeat synced vibrations calm, but do not speed sleep","Smartwatch haptics lower heart rate, not sleep latency","Rhythmic haptics relax you, but sleep onset unchanged","Vibration patterns calm, but one session won't speed sleep"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000778,"raw_usage":{"total_tokens":3417,"prompt_tokens":900,"completion_tokens":2517,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":516,"completion_tokens_details":{"reasoning_tokens":2445}},"tokens_in":516,"tokens_out":2517,"duration_ms":19446,"temperature":1.0,"reasoning_tokens":2445,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:29:37.476297+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Continuously record ECG beat-to-beat intervals while the smartwatch delivers vibrations set 4% below the current heart rate, and compare heart-rate deceleration and vagal HRV markers against a sham condition with identical vibration intensity but a fixed tempo unrelated to heart rate; if no reliable difference appears in a pre-registered sample, the claimed entrainment effect is not supported.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the 'weak non-invasive forcing' principle that a roughly 3-5% slowing of an external rhythm can influence heart rate, the basis for the 4% down-tempo rule."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Predecessor nap study showing closed-loop vibration stimulation can affect heart rhythm; its evidence motivates the subtle 4% frequency reduction used here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Multiple Sleep Latency Test protocol that the 20-minute eyes-closed sleep phase is modeled on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the finger-tapping criterion (first interval exceeding 8 seconds) used to estimate sleep-onset latency."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Psychomotor Vigilance Test used to assess alertness before and after the sleep attempt."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows behavioral sleep-onset measures converge with EEG wake/sleep markers, supporting the finger-tapping outcome."}],"review_version":1}