Peripheral scope (exercise programming). The outcome the evidence actually measures here is joint range of motion (ROM), an intermediate outcome, not a patient-important one in its own right. It matters to a person only through function (reaching, dressing, gait) or through an injury, pain or soreness claim. The held evidence does not test ROM’s transmission to function; it shows no benefit for injury (insufficient evidence, restated 2026-11-04 from near-nulls) and near-nulls for stretching around a single session on soreness (with one small field effect for a 12-week routine); and low back pain is unproven. So this page ranks low in Layer 1: a dose is held for the ROM goal, and the commonly believed benefits are mostly unsupported. (ROM-as-intermediate is the wiki’s classification, per Surrogate Outcomes; the source does not grade ROM’s transmission to function.)
Evidence-tier note (confidence: low; the soreness drop-lever is the best-supported claim, still low; injury drop-lever, dose/timing/ROM low).
The two drop-lever claims (injury, soreness) rest on RCT meta-analyses, the soreness one on a Cochrane
review whose pooled before-or-after estimates are near zero with quite precise CIs (GRADE low; moderate
only for the single 12-week field trial; the smaller pre-exercise-only pools are wide); see their sections below.
(Restated 2026-11-04.) The injury drop-lever is graded low here, as on
Exercise Interventions and Sports Injury Prevention: its MA state is insufficient evidence, and the advice not to rely on stretching
stands because it costs nothing, not because a null is shown. The soreness claim does not lift the page
above low (re-judged 2026-11-12): it rests on one review line whose pools are GRADE low, Dupuy is not
independent of Herbert, and the 12-week field estimate is a single trial (see Synthesis).
The dose and ROM claims are weaker: they
rest on one source, the 2011 ACSM position stand, a high-tier guideline that grades each recommendation with the NHLBI A-D categories (A = RCTs, rich data; B = RCTs,
limited data; C = nonrandomized or observational; D = panel consensus). It gives grades but no pooled
effect sizes, no confidence intervals and no studied ranges for its dose numbers.
(Garber et al., 2011)
The prescription below is therefore a graded guideline position, not a meta-analytic estimate.
What stretching does: range of motion, and little else that has been shown
- ROM improves, transiently and then chronically. «Joint range of motion is improved transiently after flexibility exercise, chronically after approximately 3–4 wk of regular stretching at a frequency of at least two to three times a week», and «it may improve in as few as 10 sessions with an intensive program». Table 1 grades the statement that flexibility exercise improves and maintains joint range of movement at A (the printed row is garbled: «improve and maintain and joint range of movement»). (Garber et al., 2011)
- The gain reverses when stretching stops. «Improvements in joint range of motion reverse within 4–8 wk on cessation of stretching exercise, although there are variable responses among muscle–tendon groups (115,145,146,200,394).»; one report found that cutting from daily to 2-3 sessions a week maintained ROM. (Garber et al., 2011)
- Balance, possibly. «Flexibility exer- cises may enhance postural stability and balance (83), par- ticularly when combined with resistance exercise (38).» Two references, no grade attached to this sentence; ref 83 (Costa) is a study of a single acute stretching bout, not a training programme. (Garber et al., 2011)
- Injury, low back pain, soreness: no consistent link. The next sentence: «No consistent link has been shown between regular flexibility exercise and a reduction of musculotendinous injuries, pre- vention of low back pain, or DOMS (9,116,163,248,355,401).» (Garber et al., 2011)
An inconsistency inside the stand. Table 1 still lists flexibility exercise among the methods that «may reduce the risk of CVD events and musculoskeletal injury during exercise» (grade C, bundled with warm-up, cool-down and gradual progression), while the flexibility text says no consistent link has been shown. The risks section puts both halves in one sentence: «Commonly used methods to reduce musculoskeletal injury and complications, such as the warm-up and cool-down, stretching, and gradual progression of volume and intensity, seem to be helpful at least under some circumstances, but controlled studies substantiating the effectiveness of these methods are insufficient (9,123,124,164,248,355).» The Conclusions then list warming up, cooling down, gradual progression and technique as effective strategies, and leave stretching off that list. Read the stand as not asserting an injury benefit for stretching. (Garber et al., 2011) (the reconciliation is the wiki’s reading).
The prescription, if ROM is the goal (Table 2 grades)
| Parameter | Stand’s recommendation | Grade |
|---|---|---|
| Frequency | ≥2-3 d/wk; greatest gains with daily | B |
| Intensity | stretch to tightness or slight discomfort | C |
| Hold (static) | 10-30 s for most adults | C |
| Hold, older adults | 30-60 s may give greater benefit | C |
| PNF | 3-6 s contraction at 20-75% MVC, then 10-30 s assisted stretch | B |
| Total per exercise | 60 s (e.g. 2 x 30 s or 4 x 15 s) | B |
| Repetitions | 2-4 per exercise | B |
| Coverage | a series of stretches for each major muscle-tendon unit | B |
| Type | static (active or passive), dynamic, ballistic and PNF are all effective | B |
| Timing | most effective when the muscle is warm (light-to-moderate aerobic or resistance work, or moist heat / hot baths) | A (as printed; see note) |
| Progression | «Methods for optimal progression are unknown.» | none |
- The hold-time ceiling carries no range. «Holding a stretch for 10–30 s at the point of tightness or slight dis- comfort enhances joint range of motion, with little apparent benefit resulting from longer durations (21,95).» The stand states neither the longest hold studied nor an effect size, so «little apparent benefit» above 30 s is a graded-C reading of two references, not a located plateau. (Garber et al., 2011) (the no-studied-range reading applies the wiki’s threshold rule).
- Repetition and time course. «Repeating each flexibility exercise two to four times is effective, with enhancement of joint range of motion occurring during 3–12 wk (21,95,264).» (Garber et al., 2011)
- Time cost. «For most individuals, this routine can be completed within 10 min.» The routine is a series of stretches for the major muscle-tendon units. (Garber et al., 2011)
- Muscle temperature. «Flexibility exercise is most effective when the muscle temperature is elevated through light-to-moderate cardiorespiratory or muscular endurance exercise or passively through external methods such as moist heat packs or hot baths, although this benefit may vary across muscle–tendon units (29,249,326).» (Garber et al., 2011) The three references behind this grade-A row are a study of pre- versus post-exercise stretching (Beedle), a dynamic-versus-static warm-up study on power and agility (McMillian), and a critical review of stretching and injury (Shrier). That is not a body of RCTs on ROM, so treat the row as C-strength whatever its printed grade. (Garber et al., 2011) (the re-grading is the wiki’s reading of the reference list).
Stretching before strength or power work: an acute cost
- «Stretching exercises can have a negative effect on subse- quent muscle strength and power and sports performances, particularly when strength and power are important (248,396).» No grade is attached to this statement. The stand qualifies it: limited evidence exists on stretching programmes of different durations and types for different activities, particularly in people exercising to improve fitness, and «Further research is needed before making universal recommendations concerning the timing of stretching». (Garber et al., 2011)
- Its working position: «Nonetheless, based on the available evidence, whenever pos- sible, persons engaging in a general fitness program should perform flexibility exercise after cardiorespiratory or resis- tance exercise—or alternatively—as a stand-alone program.» (Garber et al., 2011)
- And the warm-up comparison: «For most persons, a dynamic, cardiorespiratory endurance exercise warm-up is superior to flexibility exercise for en- hancing cardiorespiratory or resistance exercise (especially with high duration and repetitions) performance (29,124, 248,249,326).» Performance is the outcome here, not injury. (Garber et al., 2011)
- Warm-up as an injury lever is a separate question, carried on Exercise Interventions and Sports Injury Prevention. The held warm-up meta-analysis does not test stretching. It compares structured injury-prevention programmes (strength, balance, landing work) against «warm-up as usual (e.g., only running and stretching)» in under-19s, and where a control arm is specified it includes stretching (static and dynamic in four of the five specified); the programmes cut injury rates by about a third (IRR 0.64). (Ding et al., 2022) So a stretching-based warm-up is the comparator there, and the benefit comes from adding other content, which fits the stretching injury result in the Synthesis below (no benefit shown); the trial design cannot show that stretching contributes nothing, since five of the programmes also contained stretches.
Stretching against muscle soreness: a measured near-zero effect
The Cochrane review pools randomised trials of stretching soon before or soon after exercise against no stretching, with delayed-onset muscle soreness (DOMS) rescaled to 0-100. Twelve trials: eleven small trials (10-30 people stretching), of which ten were laboratory trials using a standardised bout of unaccustomed, mostly eccentric exercise of one muscle group and one was a cross-over trial of post-match stretching in Australian rules footballers; plus one large pragmatic internet trial (Jamtvedt 2010, 2377 adults doing their own recreational exercise for 12 weeks). Almost all used static stretching. (Herbert et al., 2011)
| Timing of stretching | Soreness day 1, MD (95% CI) | Soreness day 2, MD (95% CI) | Trials / N (day 1) | GRADE |
|---|---|---|---|---|
| Before exercise | -0.52 (-11.30 to 10.26) | 0.72 (-11.20 to 12.64) | 3 / 70 | low |
| After exercise | -1.04 (-6.88 to 4.79) | 1.12 (-4.63 to 6.87) | 4 / 127 | low |
| Either, pooled | -0.93 (-6.05 to 4.20) | 1.04 (-4.14 to 6.22) | 7 / 197 | low |
| Before and after, 12 weeks (field trial) | worst weekly soreness -3.80 (-5.17 to -2.43) | — | 1 / 2377 | moderate |
Negative = less soreness with stretching, on a 100-point scale; control-group means were 29-49 on day 1 in the small trials and 25 for weekly worst soreness in the field trial; day 3 pools are also near zero (pooled -0.28, -6.79 to 6.22). Every pool had I2 = 0%. (Herbert et al., 2011)
Magnitude, in the authors’ words. «The estimates were quite precise so, in the primary analyses, it was possible to rule out the possibility that the average effect of pre-exercise or post-exercise stretching was to reduce soreness by more than six points on a 100-point scale.» On the field trial’s 4 points: «This effect, though statistically significant, is very small.» (Herbert et al., 2011) The six-point bound holds, roughly, for the combined before-or-after pools (lower CI limits -6.05, -4.14 and -6.79 on days 1-3). The pre-exercise-only pools are 2-3 small trials and cannot exclude a benefit of about 11-16 points (lower limits -11.30, -11.20, -15.82); post-exercise alone reaches -7.49 on day 3.
The verdict. «The evidence from randomised studies suggests that muscle stretching, whether
conducted before, afer, or before and afer exercise, does not produce clinically important reductions
in delayed-onset muscle soreness in healthy adults.» The review states no minimal important difference;
clinically important is the authors’ judgment, not a pre-set threshold.
(Herbert et al., 2011)
[searched: minimal / minimum / worthwhile / smallest / clinically important across chunks 01-02]
The same field effect looks larger as a share of people. In Jamtvedt 2010, «the risk of experiencing bothersome soreness in a week was 24.6% in the stretch group and 32.3% in the control group (OR 0.69; 95% CI 0.59 to 0.82; Analysis 4.2).» The protocol behind it: seven lower-limb and trunk muscle groups, each held at least 30 s, before and after vigorous exercise, for 12 weeks. The review lists this trial’s stretching as 900 s per session, the top of its 40-900 s range. (Herbert et al., 2011) That is about 8 percentage points fewer weeks with bothersome soreness (32.3% - 24.6%), bought with at least about 7 minutes of stretching per session by the protocol (7 x 30 s before and after; the review lists 900 s) for 12 weeks. On the continuous scale the authors call the same effect very small; as a share of people-weeks it reads larger. Whether the trade is worth it is a Layer-3 preference; the evidence does not decide it.
Laboratory and field agree. «However the estimates from 10 laboratory- based studies and two community-based study suggest there are similar (near-zero) effects in both settings.» The second field trial (post-game stretching in Australian rules footballers) pointed the other way: soreness 3 points higher at 15 h and 4 points higher at 48 h, both CIs crossing zero. (Herbert et al., 2011)
Why the authors think the low GRADE may not undercut the null, and what that argument leaves out. The laboratory pools were downgraded for three reasons: risk of bias, indirectness («Such studies provide an indirect evaluation of the effects of stretching on soreness caused by recreational exercise.»), and too few studies to assess publication bias. No trial was blinded (soreness is self-reported) and «Allocation concealment was certain in Jamtvedt 2010 alone.» The authors’ argument, which they introduce as «some reasons to believe that the findings may not be seriously biased»: first, consistency of estimates across studies and comparisons; and «Also, biases due to non-concealment of allocation, lack of blinding, loss to follow-up and failure to analyse by intention-to-treat typically produce exaggerated effects (Chalmers 1983; Colditz 1989; Moher 1998; Schulz 1995) yet this review found consistent evidence of near-zero effects.» (Herbert et al., 2011) The argument has two limits: it answers the risk-of-bias downgrade only, not indirectness, which is why soreness from everyday training rests on the field data; and it covers biases that inflate a benefit, not a bias toward the null such as poor adherence in an unsupervised field arm. In Jamtvedt, «outcome reporting was in- complete (72% of all required reports).» (Herbert et al., 2011)
One sensitivity analysis gives a larger number. Adding a small trial with remarkably small SDs (mean 3.8 on 100, about one-sixth of the other trials’; excluded from the primary analysis because it would have taken 50-96% of the weight) moves day-1 soreness to -6 points (95% CI -9 to -4). (Herbert et al., 2011)
Who was studied. «In all but one study participants were young healthy adults (age less than 40 years or mean age less than 26 years); only Jamtvedt 2010 had substantial representation of older adults (mean age 39.9 years, SD 12.6).» No trial targeted older adults. (Herbert et al., 2011)
The original rationale is gone; one mechanism leaves a gap. The practice may have arisen from the 1960s-70s belief that soreness was muscle spasm, a view since discredited. «Contemporary beliefs about the mechanisms of muscle soreness do not suggest any means by which stretching before or afer exercise could reduce the soreness that follows that exercise session.» The authors then qualify it: if soreness comes from over-lengthened sarcomeres, adding sarcomeres in series or making tendons more compliant could help, so «This suggests the possibility that stretching may be more effective for people who have very short muscles, or if the stretching is repeated for weeks or months or years.» «The longest duration of stretching investigated in any of the trials conducted to date (Jamtvedt 2010) is 12 weeks.» (Herbert et al., 2011)
Out of scope for this review. Stretching once soreness has set in: «Some evidence suggests that once
muscle soreness has developed, stretching may provide a transient relief of soreness (Reisman 2005).»
(one cited study, not pooled). «The current review does not provide any evidence of an effect or otherwise
of stretching on risk of injury, performance or well-being.» The review did not assess harms (soreness was
its only outcome).
(Herbert et al., 2011)
[searched: adverse / harm across chunks 01-02]
Stretching among the other post-exercise recovery techniques
A meta-analysis of 99 studies compared single post-exercise recovery sessions in healthy adults aged 18-65 (any training status), pooling Hedges’ g per technique for DOMS and perceived fatigue. Stretching is the only technique whose interval does not reach into a reduction: SMD 0.15 (95% CI 0.00 to 0.29), 5 groups, 67 participants (positive = more soreness); electrostimulation, -0.28 (-0.59 to 0.03), and hyperbaric therapy, 0.55 (-0.12 to 1.22), are also null. (Dupuy et al., 2018) The authors: «we did not find any significant influence of stretching or electrostimulation on DOMS and fatigue.» (Dupuy et al., 2018) That interval is implausibly narrow for 67 people compared between groups (roughly ±0.35 to ±0.5 SD would be expected, depending on whether 67 counts one arm or both), which points to within-person designs or repeated timepoints counted separately; read its precision with caution, and the point estimate as “no reduction”, not as a shown increase.
| Technique (single session) | DOMS SMD (95% CI) | n | Fatigue SMD (95% CI) | n |
|---|---|---|---|---|
| Massage | -2.26 (-3.05, -1.47) | 158 | -2.55 (-3.49, -1.62) | 64 |
| Compression garments | -0.92 (-1.34, -0.50) | 160 | -0.88 (-1.34, -0.50)* | 28 |
| Active recovery | -0.94 (-1.61, -0.28) | 90 | 0.64 (-0.43, 1.70) | 33 |
| Water immersion (DOMS: all temperatures; fatigue: all 8 groups, which Table 2 lists as ≤15°C) | -0.47 (-0.77, -0.18) | 379 | -1.16 (-1.94, -0.39) | 75 |
| Cryotherapy | -0.53 (-1.04, -0.03) | 72 | not pooled | — |
| Electrostimulation | -0.28 (-0.59, 0.03) | 94 | (1 group) | 11 |
| Stretching | 0.15 (0.00, 0.29) | 67 | -0.21 (-1.04, 0.62) | 30 |
(Dupuy et al., 2018) Contrast water therapy -0.40 (-0.73, -0.07) and hyperbaric therapy 0.55 (-0.12, 1.22) on DOMS complete the table. *The compression fatigue CI repeats the DOMS CI, probably a copy slip in the source table; the source page carries the full rows and the table’s other transcription slips.
- The authors’ reading of stretching. «The overall SMD in our study confirmed that stretching had no positive impact on DOMS. Moreover, our results at <6 h after exercise indicated that stretching might even produce DOMS, as initially reported by Smith et al. (1993).» Smith 1993 compared static with ballistic stretching; the Cochrane review above lists it among its excluded studies. (Dupuy et al., 2018) (Herbert et al., 2011)
- What the comparison cannot carry. The paper never states what each technique’s SMD is measured
against, admits repeated-measure (within-person) designs, scores study quality (PEDro) but rates no
certainty for the body of evidence, and says «the beneficial
effect arising from the placebo effect of recovery techniques could not be eliminated».
(Dupuy et al., 2018)
[searched: passive / control / placebo across chunk 01]The ranking is therefore a ranking of unblinded single-session effects, specific and nonspecific together. Stretching’s null is less exposed to placebo inflation specifically, since a nonspecific boost would push toward benefit, not away from it; that argument does not cover the unstated comparator or the mix of within-person designs, which can bias either way (the same limit noted for Herbert’s bias argument above). The large massage figure is the most exposed (see Massage Therapy for Pain and Function; the general case is The Comparator Problem).
Synthesis
The stand and an RCT meta-analysis agree on injury, with shared backing, so not independent confirmation. The held RCT meta-analysis on Exercise Interventions and Sports Injury Prevention pools three stretching trials to no benefit shown for injury: «no beneficial effect for stretching (RR 0.963 (0.846–1.095))»; its discussion reports the same estimate with slightly different figures, «RR=0.961 (0.836–1.106)». (Lauersen et al., 2013) The stand’s «no consistent link» points the same way. Parameter check before comparing:
| Parameter | Garber 2011 (stand) | Lauersen 2013 (RCT-MA) | Same quantity? |
|---|---|---|---|
| Exposure | regular flexibility exercise, any form | stretching programmes (before or after exercise) | broadly yes |
| Outcome | musculotendinous injury (also LBP, DOMS) | lower-limb (± trunk) injury in the three stretching trials | overlapping, not identical |
| Design | narrative statement over 6 refs (Amako 2003 trial, Thacker 2004 SR, Herbert & de Noronha Cochrane, 3 narrative reviews) | pooled RR, 3 RCTs | no — narrative vs pooled |
| Population | apparently healthy adults (general) | stretching subgroup: 2 army-recruit RCTs (Pope 1998, 2000) + 1 internet RCT in adults >18 (Jamtvedt), no athletes | partly |
(Lauersen et al., 2013) Lauersen flags the narrow base itself: «this analysis only included two studies on army recruits and one internet-based study on the general popu- lation.» (Lauersen et al., 2013)
The comparison is a direction agreement on an overlapping outcome, not a type-E convergence. The two share backing: Lauersen says its stretching result «corresponds to earlier reviews», and two of those reviews (Thacker 2004; Herbert & Gabriel 2002, Lauersen refs 17-18) are also cited by the stand (refs 355, 164); Herbert also co-authors Pope 2000 and Garber refs 163-164. (Lauersen et al., 2013) (Garber et al., 2011) Read it as agreement (restated 2026-11-04; the page earlier read it as type-F, the later RCT-MA turning the stand’s hedged «no consistent link» into a measured result, and that reading is withdrawn). Re-judged as insufficient evidence by stratum on the injury page, the MA reaches the stand’s own state; it adds stratum detail (recruits, general adults), and the stand adds the outcomes the MA did not cover (low back pain, DOMS), at guideline grade only.
In knee and hip osteoarthritis, a mixed flexibility/skill category beat usual care (low certainty; it does not isolate stretching). On Knee Osteoarthritis and Modifiable Levers, an exercise-type network meta-analysis finds a flexibility/skill category beating usual care on OA pain and function at 8 weeks (low certainty), but the category pools stretching with balance and coordination work, so it does not isolate stretching. The same page holds a guideline null on the ROM surrogate under a combined programme in OA. Neither conflicts with this page: the stand’s ROM grade is for healthy adults doing a specified stretching dose; the OA results are a different population and a mixed exposure. A distinction, not a tension.
Soreness: the Cochrane update puts a size on the stand’s «no consistent link». The stand’s DOMS claim cites, among others, the 2007 edition of the Herbert & de Noronha Cochrane review (ref 163). (Garber et al., 2011) The 2011 update reaches the same direction with numbers. Parameter check:
| Parameter | Garber 2011 (stand) | Herbert 2011 (Cochrane) | Same quantity? |
|---|---|---|---|
| Exposure | «regular flexibility exercise» | stretching «soon before or soon afer exercise» (around one exercise bout, repeated over the following 2-3 days in two trials; one 12-week trial) | partly — acute peri-exercise vs regular training |
| Outcome | DOMS (bundled with injury and low back pain) | DOMS, 0-100 scale, days 1-3; weekly worst soreness | yes for DOMS |
| Evidence form | narrative statement over 6 references | pooled MDs with CIs, GRADE | no — statement vs estimate |
| Backing | includes the 2007 edition of this review | the update of that review | shared — not independent |
(Garber et al., 2011) (Herbert et al., 2011) Type-F: the update turns the stand’s hedge into a bounded estimate (the combined before-or-after pools exclude benefits much beyond 6-7 points on 100; the field trial’s estimate is 4 points, 95% CI 2.4 to 5.2), but it is the same review line, so it adds precision, not independent confirmation. The exposure row is the seam: the stand speaks of a regular flexibility routine, the review mostly of stretching around a single session. The one 12-week routine showed a small effect; routines longer than 12 weeks are untested.
One trial carries both general-adult results. The field trial behind the soreness estimate (Jamtvedt 2010, internet-run, adults doing recreational exercise) is also one of the three trials in Lauersen’s pooled injury estimate, and the review’s first author co-authored it. So for general adults outside the laboratory, stretching around exercise prevents neither injury nor meaningful soreness rests largely on one 12-week trial reporting two outcomes, not on two independent bodies of evidence. (Restated 2026-11-04.) For injury the state is insufficient evidence (the injury page’s re-judge), so the reading is not shown to prevent injury, nor meaningful soreness; the no-effect form is withdrawn for injury. (Lauersen et al., 2013) (Herbert et al., 2011) The laboratory near-null on soreness after a standardised eccentric bout rests on the ten small lab trials, not on Jamtvedt; the field estimate for everyday training (a small benefit) and the injury result in general adults (no benefit shown) both rest on Jamtvedt.
A third MA reads the same trial differently (2026-11-04). Liddle’s adult-recreational meta-analysis pools Jamtvedt as HR 0.75 (0.59-0.96) (read off its Figure 3) and lists the outcome as «Overall Injuries & Soreness»; Lauersen’s Table 1 counts 339 of 1079 vs 348 of 1046 for «Lower limb + trunk injury», a crude RR of about 0.94. (Liddle et al., 2023) (Lauersen et al., 2013) (inferred from Lauersen et al., 2013; Liddle et al., 2023) Lauersen’s analysis covers each trial’s primary outcome, which its Table 1 names for Jamtvedt, and Liddle’s hazard ratios are those trials reported from survival analyses; which Jamtvedt outcome Liddle took is not stated, and Jamtvedt is not held, so the page does not choose. If Liddle’s number is an injury-only hazard ratio, the general-adult no-benefit-shown result rests on one trial whose own analysis may point the other way; if it includes soreness, it is the soreness benefit above showing up again. Liddle pools no stretching estimate (its HR set pools Jamtvedt with two warm-up routines and a foot-ankle programme), so the no-benefit-shown result is not contradicted, only less settled; the injury page re-judges it as insufficient evidence, no benefit shown (2026-11-04). Detail: Exercise Interventions and Sports Injury Prevention.
Soreness: the recovery-technique meta-analysis agrees on direction, in a unit that cannot be lined up with the Cochrane points. Parameter check:
| Parameter | Herbert 2011 (Cochrane) | Dupuy 2018 (recovery MA) | Same quantity? |
|---|---|---|---|
| Exposure | stretching «soon before or soon afer exercise», mostly one bout (two trials repeated it over 2-3 days; one 12-week trial) | one post-exercise stretching session («repeated sessions of recovery … were not included») | partly — post-exercise single session is a subset |
| Outcome measure | mean difference on a 0-100 soreness scale, day 1 (18-30 h), day 2, day 3 | Hedges’ g per group (one composite where a study used several measures); a <6 h statement | no — points vs SD units |
| Comparator | no stretching | not stated in the methods | unknown |
| Population | young healthy adults, lab eccentric bouts; two field trials | healthy adults 18-65, any training status | broadly |
| Trials | 12, listed | 5 groups / 67 people, not listed | overlap cannot be checked directly |
| Estimate | either-timing day 1 -0.93 (-6.05 to 4.20) | 0.15 (0.00 to 0.29) | no — see outcome row |
(Herbert et al., 2011) (Dupuy et al., 2018) Type-F at most, and only for direction. Both find no soreness reduction from stretching, but Dupuy cites the Herbert reviews as the reason stretching «has not been recommended after exercise» and does not list its trials, so it is not independent confirmation; an SMD cannot be turned into the Cochrane points without the outcome SDs. One part of Dupuy’s result is a timepoint Herbert did not pool (its earliest window is 18-30 h): the suggestion that stretching may add soreness in the first 6 hours. The text gives no number for it (only a time-course figure) and cites Smith 1993, which Herbert excluded, so it stays a weak signal. (Dupuy et al., 2018) A possible hint of overlap: Herbert’s post-exercise day-1 pool (four trials) totals 67 stretching participants, the same n as Dupuy’s stretching row (five groups; its n column is not defined as arm or total). It is consistent with shared trials but does not establish them. (Herbert et al., 2011)
Decision relevance
If the goal is joint range of motion (a stiff hip or shoulder that limits something you want to do): stretch each tight area 2-3+ days a week, 60 s total per stretch (e.g. 2 x 30 s or 4 x 15 s; 30-60 s holds for older adults), after a workout or on its own, about 10 minutes for the whole body. Expect gains over 3-4 weeks and expect to lose them within 4-8 weeks of stopping. This is a graded B/C guideline dose, not a measured optimum. (operationalizes the Table 2 rows above).
Do not count on stretching to prevent injury. The RCT meta-analysis showed no benefit in recruits and general adults (insufficient evidence, not a tight null, by the injury page’s 2026-11-04 re-judge). If injury prevention is the aim of a warm-up, the held evidence favours a structured strength/balance/jumping warm-up programme over the usual warm-up (the review’s example: running and stretching), with about a third fewer injuries (IRR 0.64) in sport participants aged 18 or under (the Ding section above); for adults see Exercise Interventions and Sports Injury Prevention. For back pain the guideline found no consistent link: unproven, not shown useless.
A single bout of stretching around one session: skip it if next-day soreness is the only reason. It changes soreness over the next three days by about 1 point on a 100-point scale in the combined pools (no primary pooled point estimate exceeds 2.5 points; the combined CIs reach about 6-7 points of benefit, the small pre-exercise-only pools 11-16). In young healthy adults after a laboratory eccentric bout this is a measured near-zero effect (indirect for everyday training), not missing evidence. (operationalizes the Herbert estimates above). (2026-11-05) That near-zero verdict holds for the combined pools, and rests on Herbert’s own judgment of clinical importance, since the review sets no minimal important difference. The pre-exercise-only pools, whose CIs still admit 11-16 points of benefit, are insufficient evidence under the rule the injury page applies to an imprecise interval. The action is the same either way: skipping a single bout costs nothing.
A standing before-and-after routine: a small effect, worth it only if you value the trade. Over 12 weeks it cut the share of weeks with bothersome soreness from about 32% to 25%, for at least about 7 minutes of stretching per session (the review lists 900 s). (operationalizes the Jamtvedt estimate above).
For a beginner, the lever on soreness is the dose of new work, not stretching. The review describes DOMS as «usually caused by unaccustomed exercise, particularly exercise that requires primarily eccentric muscle action». (Herbert et al., 2011) So the practical default is to start new exercises, especially lowering-heavy ones such as downhill walking or slow eccentric lifts, at low volume and build up. This is standard practice, not trial-tested against stretching; the repeated-bout idea behind it is a mechanism note on Resistance Training Prescription - Load Sets and Frequency.
Untested, not refuted: stretching for longer than 12 weeks, or for people with very short muscles, which the review names as a conceivable route to an effect. (Herbert et al., 2011)
For soreness after a session, do not use stretching; if you want something, pick from the techniques that showed a signal and expect a modest effect at most. In the recovery-technique meta-analysis, massage, compression garments, active recovery, cold water (≤15°C; warm immersion showed no reduction), contrast water and cryotherapy showed pooled single-session reductions in SD units; stretching, electrostimulation and hyperbaric therapy showed none. Light active recovery or a cold soak are the cheap options. The ordering is suggestive only: each technique was pooled from a different set of trials, not compared head to head, the DOMS funnel plot leaned toward positive effects («23 studies out of 57 were distributed toward a positive effect»), and the estimates are unblinded, against an unstated comparator, carry no certainty rating and are untested for regular use; none can be sized in points on a 100-point scale. Details: Massage Therapy for Pain and Function, Cold-Water Immersion (which also carries the open question of cold water after strength training). (operationalizes the Dupuy section above) (Dupuy et al., 2018)
Before heavy lifting, sprinting or jumping, warm up with movement, not long static holds. The stand states the acute strength/power cost without a grade and calls the evidence for general exercisers limited, so this is a low-cost default, not a firm finding.
Layer-1 rank: low. For a novice whose large lever is starting aerobic and resistance work at all, stretching is optional. Its strongest case is a person whose limited ROM blocks an activity they value.