SYNTH-39 · Guide
When session RPE and TSS disagree
Read it as a list of ordinary causes, not as a test of central fatigue
The decision this framework makes
What to conclude — and what not to conclude — when a session's internal load (sRPE × duration) and its external load (a power- or pace-derived training-stress score) point in different directions.
Key Findings — what each number actually is
sRPE is a validated instrument with a defined protocol. Foster's method: about 30 minutes after the session the athlete rates the whole effort on the Borg CR-10 scale; sRPE × session duration in minutes gives a load in arbitrary units.Foster 2001's own abstract reports a "highly consistent" relationship between the session-RPE score and an HR-based objective method across steady-state cycling, interval cycling and basketball, with the sRPE score running higher in absolute terms, and gives no correlation coefficients; the r 0.75-0.90 range and the lactate / %VO2max criteria once quoted for sRPE are not in Foster's abstract and have been deleted from REF-500.
It is one number for a whole session. Its resolution is the session. It cannot distinguish a hard first half from a hard second half, and it is coloured by how the session ended.
A power- or pace-based stress score is a function of the stored threshold. It is computed against an anchor that was measured at some point in the past. When the anchor is wrong, the score is wrong by the same proportion for every session, in the same direction, silently.
Perceived effort can move with no peripheral change whatsoever. After 90 min of a demanding cognitive task, 16 subjects cycling to exhaustion at 80% peak power lasted 640 ± 316 s against 754 ± 339 s after watching neutral documentaries (about 15% shorter, p=0.003), and rated perceived effort significantly higher throughout — while cardiorespiratory and muscle-energetic responses were largely unaffected (REF-370, randomised crossover).The RPE rose at the same rate in both conditions, so the mentally fatigued riders simply reached their maximal RPE and stopped earlier.This is the decisive fact for this card: an elevated internal load with unchanged external load is documented after a 90-min laboratory cognitive task (its extension to an ordinary bad mental day is inference), not a marker of a specific fatigue state.
The environment moves internal cost at a fixed external load. With fluid fully replaced, between minutes 15 and 45 of a 45-min ride at 59.2% VO2max heart rate rose 11% and VO2max fell 15% at 35 °C, versus +2% and a non-significant 5% fall at 22 °C (REF-3496, 10 trained men). The same watts genuinely cost more in the heat.
Subjective measures track training response more sensitively than most objective ones (REF-368) — which is a reason to take a high sRPE seriously, not a reason to treat it as diagnostic.
1. The rule
A divergence licenses a question, not a diagnosis. It says one or more ordinary causes are present. The honest response is to name the candidates and check the cheap ones.
2. The candidate list, in the order worth checking
sRPE high relative to the external score (the candidates and their order are the base's own practical judgement, not a ranked evidence base; a refId marks the items that have a source):
1
Heat, humidity, altitude, headwind, or a heavy bike/pack — the environment raised the internal cost of the same work (heat: REF-3496, 35 vs 22 °C; humidity, altitude, wind and weight are added from practice, not from that entry).
2
The stored threshold is too high — every session then scores low, consistently, in the same direction. This is the cause worth catching, because it is silent and systematic.
3
Sleep debt, life stress, a cognitively brutal day — perceived effort rises with the physiology largely unaffected (tested for a 90-min cognitive task in REF-370; sleep debt and life stress are inference, not tested there).
4
Illness, or the days before it becomes obvious.
5
Under-fuelling during or before the session.
6
Session shape — surging, punchy terrain, group riding: repeated hard transitions cost more than the average suggests.
7
Accumulated fatigue from the preceding days, which is the one interpretation people jump to first and which belongs sixth or seventh on the list.
sRPE low relative to the external score:
1
The stored threshold is too low — the athlete has improved and the score has not been told.
2
Favourable conditions: cool, tailwind, downhill-net, fast group, fresh legs.
3
A session that scored high on time-in-zone arithmetic but was psychologically easy — long steady work often reads harder on paper than in the legs.
4
Under-reporting: athletes systematically understate fatigue, particularly under a coach's eye (REF-500's social-desirability caveat).
3. What the divergence is NOT
It is not a test of central fatigue. No study in this base measures central fatigue from the gap between a perceived-effort score and a power-derived one, and REF-370 shows the internal side can move on its own for entirely mundane reasons.
It is not a readiness score. It describes a session that already happened.
Agreement is not reassurance. Two summaries can agree while the athlete is under-fuelled, ill or losing form; "internal matched external, therefore nothing is wrong" is a false inference and should never be stated.
A single session's divergence is not a trend. The instruments are read within the athlete against their own baseline (the base's own practice as carried in REF-500, in the direction the Saw 2017 guidance argues for); acting on a rolling shift rather than one day is the base's own practice, carried in REF-500 as a methodological convention — REF-368's abstract does not state a trigger rule.
Practical Application
When the two numbers disagree, ask about the conditions before you ask about the athlete. Heat, wind, terrain and load are the cheapest things to check, and heat alone raised heart rate 11% at fixed watts (REF-3496); that they explain more divergences than fatigue does is practical judgement, not a measured frequency.
Check the stored threshold first when a gap persists in one direction (the base's own reasoning, not a tested rule). If most sessions in a block score lower than they felt, the stored threshold is probably too high; if most feel easier than they score, it is probably too low. That makes a threshold test the first check — after illness, under-fuelling and accumulated fatigue have been asked about, since a block that keeps feeling harder than it scores fits those too.
Say "this is consistent with several things" and name them. Never convert a gap into a single cause, and never call it central fatigue — nothing in this base licenses that reading.
Act on a rolling shift across several sessions, not on one (the base's own practice; no study here sets the window). One divergent session is an ordinary day; a week of them is a question worth taking to the athlete.
Ask for the sRPE about half an hour after the session, on the same 0-10 scale every time (Foster's protocol as carried in REF-500; the 30-min window is the protocol's convention, not a tested finding). Ratings taken mid-session, or on a scale that keeps changing, are not comparable to the athlete's own history.
Never overrule the athlete's high sRPE with a low computed score. Subjective well-being measures track the training response more sensitively than physiological markers such as resting blood values and heart rate (REF-368 — a review of wellness self-report, not of sRPE against a power score; applying it here is a transfer); the computed number is a function of a stored anchor, and the athlete is the measurement.
What NOT to do
Do not diagnose overreaching from a divergence. That diagnosis has its own criteria and its own entries (REF-206 and the entries in Section 29).
Do not adjust the plan on one session's gap.
Do not average sRPE and the computed score into a single "true" load — they measure different things and the disagreement is the information.
Do not read agreement between them as evidence that the athlete is fine.
Do not compare one athlete's sRPE to another's; reading the scale within-person is the base's own practice as carried in REF-500.
Limitations
The divergence rule itself is reasoning, not a measured finding. No study in this base quantifies how large a gap has to be before it means anything, and this card deliberately proposes no threshold.
sRPE was validated against an HR-based objective score (Foster 2001), not against a power-derived stress score; the two have never been calibrated against each other here. The r 0.75-0.90 range once quoted for sRPE is in no abstract and has been deleted from REF-500.
Power- and pace-derived stress scores are proprietary constructs with no primary entry in this base; their internal weighting is not something this card can inspect.
REF-370 is 16 participants in a laboratory randomised crossover (B+) — a clean demonstration of the mechanism, not a field prevalence.
The candidate ordering in section 2 is practical judgement about what is most common and cheapest to check, not a ranked evidence base.
Nothing here applies to strength sessions, which this card does not cover.
How this entry was checked and changed
Why this level
(synthesis; the systematic review REF-368 and the consensus REF-206 are A-level, but the instrument entry REF-500 is C+ and REF-370, REF-3496 and REF-195 are B+; the divergence rule and the candidate list are reasoning over them, not measured findings)
Content correction 2026-09-27
"REF-370: time to exhaustion fell 12%, RPE rose 0.5-1.0 Borg points, HR/lactate/VO2 identical" -> abstract: 640 ± 316 s vs 754 ± 339 s (≈15% shorter, p=0.003), RPE significantly higher (p=0.007), physiological responses "largely unaffected"; randomised crossover, 16 subjects, 80% peak power.
"REF-500: sRPE validated against TRIMP, lactate, %VO2max, %HRmax at r 0.75-0.90" -> Foster 2001's abstract reports a highly consistent relationship with an HR-based method in cycling and basketball and gives no r-values; the r range is labelled an untraced composite figure.
"REF-368: decisions come from rolling shifts" -> REF-368's corrected entry labels the rolling-shift / trigger rules as the base's own practice; attributed accordingly.
REF-3496 drift restated over the measured window (minutes 15-45 at 59.2% VO2max, 10 men) with the non-significant 5% cool-condition fall.
Level parenthetical: REF-370 described as one crossover of 16 rather than "A-level"; synthesis level B+ unchanged.
Level parenthetical: the mental-fatigue mechanism was called A-level; its evidence in the base is B+ or lower.
Resync 2026-09-28: "the r 0.75-0.90 range and lactate / %VO2max criteria quoted in REF-500" (Key Findings and Limitations) -> the corrected REF-500 has deleted that range; Foster 2001's abstract gives no coefficients. Both lines now say the figure is not in any abstract and no longer in the entry.
Resync 2026-09-28: "REF-370 (B+ by the current rubric, though its entry still reads A); the entry's '12% shorter' and '0.5-1.0 Borg points' are not what the abstract reports" -> REF-370 is now B+ and carries the abstract's numbers (640 vs 754 s, ~15%; RPE significantly higher); stale caveat removed.Cross-reference "RPE rises with heart rate, lactate and VO2 unchanged" -> "physiological responses largely unaffected", the entry's wording (the HR/lactate/VO2 detail was removed from it).
2026-10-01: cross-reference repointed from a merged framework to the one that replaced it (SYNTH-01→31, 05→34, 09→70, 12/15→32, 19→26).
Verification 2026-10-01: 19 ref-bearing sentences read against REF-368, REF-366, REF-370, REF-500, REF-3496, REF-3495, REF-195 in the KB.md snapshot; 9 OK. All numbers quoted from REF-370 (640 ± 316 vs 754 ± 339 s, p=0.003, n=16) and REF-3496 (+11% / −15% vs +2% / n.s. 5%, 59.2% VO2max, 10 men, minutes 15-45) match their entries; no NUMBER or WRONG finding.
Verification 2026-10-01: rows counted = 11 candidate causes (section 2) + 6 Practical bullets. None is a measured finding about an sRPE-vs-TSS gap; 6 are 'mixed' (a real source for a neighbouring fact — heat, mental fatigue, Foster protocol, under-reporting — carried over by reasoning).
Verification 2026-10-01: the card is honest that the rule is reasoning (section 1, Limitations) and proposes no threshold.The real problems are (a) three refIds stretched past what the entry tested (REF-3496 for humidity/altitude/wind/weight and for a frequency claim; REF-370 for sleep and life stress; REF-368 for sRPE vs power), (b) the unlabelled 'threshold problem until proven otherwise … not a recovery decision' rule, which contradicts the card's own 'never convert a gap into a single cause', (c) level B+ with a C+ instrument entry.
Verification 2026-10-01: 'Borg CR-10' and 'about 30 minutes' are 'full text; not checked against the abstract' in REF-500; the card labels the 30-min window as a convention, the CR-10 scale is not labelled (not flagged as a finding).
Verification 2026-10-01: minor wording — candidate 7 says accumulated fatigue 'belongs sixth or seventh on the list'; it is seventh.