Twelve casual, five elite, one floor
The approved design targeted twenty-four participants in two groups of twelve. Seventeen completed sessions: twelve casual rowers and five elite, grouped at recruitment by training and recent competition, then corroborated by the background questionnaire. Mean logging spanned 681 s with 630 s of rowing, 237 strokes, 22.8 spm, and 1,713 m at 61.6 W — and mean power still ran from 8.2 W to 165.7 W under the same instruction.
That twentyfold range is what an open task produces when rowers of very different backgrounds are told to row naturally and left alone. Elite median rate sat at or below the casual median while work per stroke was 4.3× higher: length and force, not frequency.
The single-condition structure is unchanged from the protocol: one instrumented RowSim bout after orientation, then REQ, the adapted UES-SF, and a semi-structured interview. The tradeoff remains: no multi-condition causal isolation. What changed is that the loop closed.
N=17 arrived (12 casual / 5 elite). Protocol allowed 24.
Who arrived
Twelve casual. Five elite.
Expertise is stratified; the projected floor session is shared. Protocol allowed twenty-four.
01 · ArrivedSeventeen completed sessions: twelve casual and five elite. The protocol allowed twenty-four.
Twelve casual and five elite rowers share one instrumented floor-MR session. Protocol allowed 24; 17 arrived.
Open task, about ten minutes
Participants were given no performance target and no instruction to row continuously. The script reads: there is no right or wrong way; row naturally; pause, rest, or stop at any point. Thirteen of seventeen sessions contain no pause at all. Four people stopped: P8 seventeen times for 175 s — and reported mild motion sickness if they stared too long — P5 six times for 64 s, P15 six times for 62 s, and P10 twice for 39 s. Casual pauses were brief and scattered; P10’s two breaks were self-imposed pieces. P12 had disclosed vestibular dysfunction and reported no difficulty. Two cases. Not a safety claim.
The display was one configuration for everyone: fluid magenta against white, palettes off, numeric HUD off, Wizard-of-Oz off. Watts scaled ink, peak force scaled impulse, boat speed set the carrier; displacement then ×0.42 / ×0.55 / radius 95. Twelve of seventeen raised that magenta unprompted. Eight measured it against water; P11 liked that it was not the sea. Palettes exist in the software — holding one was a study choice.
Consent went out at least forty-eight hours ahead and was signed in person. Identifiers here are P1–P17. The PDF is the archival copy of the same anonymous results.
Session shape
Open task, one floor.
One experience arc — not a battery of disconnected tasks. About ten minutes, no target.
01 · ConsentInformed consent signed in person after the forty-eight-hour review window.
Session shape: consent → orientation → open RowSim bout (~10 min) → REQ/UES → interview → debrief.
Logs, adapted UES-SF, REQ, interview
UES-SF was administered as twelve items: nine published, three replaced (focused attention has two items, perceived usability four). Subscales are means of administered items, not the published ÷3. Overall UES 4.37 (SD 0.43); REQ composite 4.54 (SD 0.32). No score sat below the midpoint on any subscale. Rhythm disruption sat at the floor (1.06); “I would recommend this experience” at 4.82; preference for the projection over a standard display at 4.59.
Sensor and PM5 logs feed behavioural metrics; REQ and UES feed subjective constructs; interviews feed six themes. Synthesis is done. Where the strands converge, confidence rises modestly. Where they diverge — pausing, absorption, novelty decay — the divergence is the finding.
Interpretive load sat on the floor: sixteen of seventeen gave the minimum for difficulty keeping rhythm. Focused attention is the one subscale off the ceiling because its two items split. “I was absorbed in the activity” averaged 4.41 (SD 0.62); “I lost myself in this experience” averaged 3.24 (SD 1.44) — the widest spread in either instrument. While filling the form, participants asked aloud whether “lost” was meant positively or negatively. The item is clear about its construct and ambiguous about its valence.
Instruments
Three strands. One synthesis.
Logs, questionnaires, and interviews are synthesised. Divergence is a finding, not a gap.
01 · StreamsLogs, REQ, adapted UES-SF, and interviews were collected from the same seventeen sessions.
Sensor/PM5 logs, REQ, adapted UES-SF, and interviews converge. Synthesis is done.
Output separates. Engagement does not.
All seven physical-output measures separate the groups; none of the six engagement measures does. The identical Mann–Whitney test, on the identical seventeen people, detected p ≤ .0006 on five force measures and p = .027 on drive-to-cycle ratio. It found nothing on how the display was experienced. Closest is aesthetic appeal at p = .195.
A participant producing 8.2 W and a participant producing 165.7 W returned overall engagement of 4.00 and 4.33. Twenty-fold physical range, 1.58 scale points of engagement. Elite rowers found the display easier, not harder — and were more conservative about replacing the PM5 they train with.
Non-detection is not equivalence. Five elite participants cannot prove a small effect is absent. They were enough to detect large differences in how the rowing was done.
Work per stroke 4.3× at or below the casual median rate.
Group comparison
Output separates. Engagement does not.
Each bar is −log₁₀ p. The spine is p = .05. Ranked as in the thesis.
01 · ForceSeven physical-output measures cross the line. Ranked as in the thesis, not regrouped.
Ranked Mann–Whitney p-values. Physical output sits left of p = .05; engagement does not.
Trained rowers occupy a different region
Plot stroke rate against work per stroke and the groups do not slide along one line. Elite sessions sit in a high-work band at or below the casual median rate. That is the textbook signature of trained rowing, and it is independent confirmation that recruitment-time labels were real.
Engagement does not track how hard people rowed (UES × power ρ = −0.07). Among casual participants only, paused time associates with engagement at ρ = −0.57 (p = .054) — diluted, not strengthened, when elite pausers are added, because P10’s pauses were pieces.
Rate and work
Trained rowers occupy a different region.
Seventeen sessions. Stroke rate against work per stroke. No target was given.
01 · CasualTwelve recreational sessions rise first. Rate spreads; work stays low.
Seventeen sessions: stroke rate against work per stroke. Elite band at or below the casual median rate.
What the interviews do that the tables cannot
Reflexive thematic analysis of the seventeen interviews produced six themes. Three support the design position, two qualify it, and one challenges it. All seventeen described coupling. Ten described deliberate exploration. Five, unprompted, predicted novelty decay within a single ten-minute exposure.
P9 named why leaving the numbers matters: optimising a metric versus having fun without one. P10 asked for hold-water by its technical name. P5 did not want the same pattern after five or six minutes. P16, who did not vary his rowing, returned the lowest engagement in the study. P17, the counter-case, said the display pulled attention toward the erg rather than away from it.
01 / 26Coupling drove engagement
Interview themes
Six themes. Not a transcript wall.
01 · Coupling drove engagementSupportsReal-time mapping
Six interview themes as tour stages. Three support the design, two qualify it, one challenges it.