Training monotony is the week's mean daily load divided by the standard deviation of those daily loads and training strain is the week's load multiplied by monotony
And in the 25 athletes the constructs were built on, illness clustered where each athlete crossed a threshold identified for him individually, mostly on strain, not where anyone crossed a shared number
Protocol
Athletes logged every session with the session-RPE method — a rating of the session as a whole multiplied by its duration in minutes — producing one load number per session and, summed, per day.
Banal illness was recorded and related to three indices computed from that series: load as a rolling 6-week average, monotony as the daily mean divided by the standard deviation, and strain as load × monotony.
The abstract describes the thresholds that separated ill from healthy periods as "individually identifiable" — read as: fitted per athlete against that athlete's own series, not applied as one shared cut-point across the group.
Key Findings — what the primary source actually states
n=25 experienced athletes. They logged training by combining the RPE of each exercise session with how long the session lasted — i.e. session-RPE × minutes, the same instrument as REF-500.
Three indices were correlated with banal illness, a marker Foster names as frequently cited for overtraining syndrome: load (rolling 6-week average), monotony (daily mean ÷ standard deviation), and strain (load × monotony).
**The result: a high percentage of the illnesses could be accounted for by athletes going over training thresholds that were *individually identifiable*, mostly tied to training strain.** The threshold is a per-athlete quantity in the source itself.
The stated rationale for looking at monotony at all: animal studies had pointed to unremitting heavy — monotonous — training as making overtraining syndrome more likely. The paper's conclusion is modest: simple monitoring of training may let athletes reach their training goals while keeping unwanted outcomes to a minimum.
What the abstract does NOT contain, and what therefore does not appear in this entry: any numeric cut-point for monotony or strain, the percentage of illness explained, the sports of the 25 athletes, or the observation period.
The formulas as the applied literature operationalises them
Stated explicitly and attributed to Foster 1998 in peer-reviewed use (Clemente et al. 2019, PLOS ONE 14(12):e0225522, DOI 10.1371/journal.pone.0225522, which cites this paper for both):
| index | formula | unit |
|---|---|---|
| monotony (weekly) | mean of the seven daily loads ÷ standard deviation of those seven daily loads | a ratio — unit-free, so it survives a change of load scale |
| strain (weekly) | sum of the week's session loads × monotony | carries the load unit — comparable only within one load scale |
What the formula does mechanically — arithmetic, not a finding
Monotony is exactly the reciprocal of the coefficient of variation of daily load, so it is dominated by how many zero days the week contains. Worked from the formula itself (equal-sized sessions, population SD):
| week shape | monotony |
|---|---|
| 7 identical days, no rest | undefined (SD = 0) |
| 6 identical days + 1 rest day | 2.45 |
| 5 identical days + 2 rest days | 1.58 |
| 3 hard + 2 easy + 2 rest, loads 150/50/0 | 1.21 |
| 4 moderate + 1 big day + 2 rest, loads 100/245/0 | 1.21 |
Two consequences worth saying out loud: a week that already contains a rest day can still score "over 2", and two weeks with completely different session content — one big day versus five even ones — can land on the same monotony.
Afonso et al. 2021 (Front Sports Act Living 3:661200, DOI 10.3389/fspor.2021.661200) make the same point about the construct: the index is dominated by session duration, and "if weekly training sessions have similar duration and perceived exertion levels, that week will be considered monotonic, even if the specific contents and training stimuli diverged widely".
How this differs from ACWR — one line
ACWR is a between-week question (is this week bigger than the recent norm?); monotony is a within-week one (was every day the same shape?). Two weeks can carry an identical total, an identical ACWR and completely different monotony. Both are ratios built from the same daily-load series, and both inherit the ratio problems REF-365 records for ACWR.
Field values, so a number can be read against something
Professional soccer, 27 players over 10 weeks: weekly monotony peaked at 3.8 A.U., weekly strain ranged 885-1751 A.U. (Clemente 2019, load in session-RPE units).
A track-and-field preparatory camp: middle/long-distance runners averaged ~2.2 A.U., throwers ~1.6 A.U. (Cristina-Souza et al. 2019, J Phys Educ Sport 19(Suppl 1) Art 26:173-177, DOI 10.7752/jpes.2019.s1026).
Values above 2 are ordinary in real training weeks and are not by themselves an event.
Practical Application
Compute monotony as the week's mean daily load divided by the standard deviation of the seven daily loads, and strain as the week's total load times that monotony — then read both against the athlete's own history, never against a shared cut-point. Foster's own result is that the thresholds that mattered were identified per athlete; a population number is not what the source found.
To move the index, change the shape of the days, not the total. One genuine rest day, or one deliberately easy day, does more to monotony than any reshuffling of the hard work — because the standard deviation of daily load is what the ratio is built on.
Read a high monotony as a question about the week's structure, not as a diagnosis. It says the days looked alike. It does not say the athlete is overreached, and it says nothing about what the sessions contained.
If the load unit is TSS rather than session-RPE × minutes, monotony still transfers and strain does not. Monotony is a unit-free ratio. Strain carries the load unit, so the published strain magnitudes above cannot be compared with a TSS-derived strain — build a within-athlete baseline before a strain number means anything.
Do not quote 2.0 as a limit the way it is usually quoted. See Limitations: it is a practitioners' convention with a soft citation trail, not a number this base can trace to the primary source.
Handle the degenerate week explicitly: seven training days of near-identical load drive the standard deviation toward zero and the index toward infinity. Report "no rest day this week" rather than a spectacular number.
Illness is a clinician's question once it persists. These indices describe training structure; recurrent infection, unexplained fatigue or a performance decline that outlasts a recovery week belong with a physician, and with the overtraining criteria in REF-206, not with a ratio.
Limitations
What was read for this entry. The primary full text was not accessible — the paper is closed access (Unpaywall reports
is_oa: false; the publisher's full-text page returns HTTP 402).Everything attributed to Foster above comes from the PubMed abstract read verbatim (PMID 9662690, NCBI E-utilities), and nothing from the paper's Results section beyond what its own abstract states.The formulas, the field values and the critique come from named secondary sources with their own DOIs: Clemente et al. 2019 (PLOS ONE, full text read, cites Foster 1998 for both formulas), Cristina-Souza et al. 2019 (J Phys Educ Sport, full text read), Afonso et al. 2021 (Front Sports Act Living, open-access full text read), and Haddad et al. 2017 (Front Neurosci 11:612, DOI 10.3389/fnins.2017.00612), a review of the session-RPE method that restates both indices and attributes them to Foster 1998.The famous thresholds are not verified against the primary source. The applied literature states that "training monotony should not exceed 2.0 a.u. in adult athletes to avoid maladaptive responses to training" and attributes it to Foster 1998 together with a speed-skating book chapter (Foster, de Koning, Gemser & van Ingen Schenau 1999) — but the 1998 abstract carries no such number and points the other way, calling the thresholds individually identifiable. Treat 2.0 as a coaching convention with a soft citation trail, not as a finding of this paper.
Three commonly quoted bands have no source this base could find and are deliberately absent here: an "ideal 1.3-1.5" monotony range, a "below 1.0 = chaotic" floor, and strain bands (low / moderate / high at fixed absolute values). They are not in the primary abstract and no cited source states them; do not reproduce them.
The construct measures shape, not content. Identical durations and identical RPEs read as monotonous even when the sessions were physiologically different, and the index is dominated by session duration.
The evidence itself is thin for the weight the construct carries. n=25, observational, 1998, one cohort, per-athlete thresholds fitted to each athlete's own data — which is close to fitting the answer — and no control for the ordinary determinants of upper-respiratory illness (travel, sleep, season, exposure). A construct being universally used is not evidence that it predicts.
Whether these indices predict anything in endurance sport specifically is unsettled. The replication literature quoted here is team-sport and track-and-field, describes seasonal variation rather than outcomes, and this base holds no endurance-specific test of monotony or strain against illness, injury or performance.
Scale transfer is unproven. Every number in this entry originates from session-RPE × duration. No source here has tested monotony or strain computed from a power-derived stress score.
How this entry was checked and changed
Content correction 2026-09-27
Worked-example table: the "4 moderate + 1 big day" row had no stated loads, so its 1.20 could not be recomputed; loads are now given (100/245/0) and the value recomputed from the formula (1.21, population SD) so the "same monotony from different weeks" point is exact arithmetic. Other rows re-verified (2.45, 1.58, 1.21).
Everything attributed to Foster re-checked against the abstract — unchanged; field values and formulas remain attributed to the named secondary sources.