Is a last session faster than race pace better?
Unclear: in 10 sub-elite runners a race-pace session 3 days out went with a 10 s faster 1500 m, a 115% one with 4 s, mixed. No direct test; a track trial ranked them the other way.
Spilsbury and colleagues asked whether the last interval session of a taper should be run at race pace or faster.
Ten trained male middle-distance runners each completed two conditions in random order, at least 3 weeks apart, each made of a 7-day week of their usual training and a 7-day taper; a 1500-m treadmill time trial was run before and after the usual week and after the taper.
The taper was prescribed from the group's prediction equations built on elite British runners' practice (REF-818): total running cut 30%, interval sessions kept, interval running cut 45%.
The two conditions differed only in the last interval session, three days before the final trial: 300-m repetitions with 90 s recovery on a treadmill at race pace (RP) or at 115% of race pace (HI).
The RP taper improved the 1500 m by 10.1 ± 1.6 s (90% confidence limits; 3.4%), rated "most likely beneficial"; the HI taper by 4.2 ± 12.0 s (1.4%), "unclear", with three runners slower afterwards.
Strength and rate of force development improved with tapering, with no evidence of a difference between the two conditions (MVC interaction p = 0.40; rate of force development p = 0.06). The authors recommend the race-pace taper for young sub-elite runners and caution that the faster final session is highly variable.
Our reading: in these runners the model-based, race-pace taper gave a consistent gain; the faster session, run at 2.7 km — more than the race distance — inside a modest volume cut, helped some and hurt others. The follow-up on a track with a 60% cut and a shorter session (REF-1201) found the opposite ranking, without a test between conditions in either paper.
Protocol
Participants: n = 10 sub-elite trained male middle-distance runners (800 and 1500 m); age 21.7 ± 3.0 y, height 182.9 ± 7.0 cm, body mass 73.4 ± 6.8 kg; personal best 4:17.5 ± 0:26.9 min (range 3:51.3-5:16.7); at least two years of training, two uninterrupted months before the study.
Design: counterbalanced crossover, order randomised; each condition = 7 days of usual training (control) + a 7-day taper; conditions at least 3 weeks apart; time trials on days 1, 9 and 17 of each condition (six in all), after familiarisation. Days 1 and 9 did not differ (RP 296 ± 20 vs 300 ± 20 s; HI 295 ± 22 vs 298 ± 16 s), so their mean is the control value. Tested in the indoor and pre-outdoor seasons (January-April).
Taper: prescribed from the control week by the elite-practice prediction equations; structure and interval sessions kept as far as possible.
Training achieved (Table 3): RP continuous 45 ± 13 → 33 ± 9 km (−27%), interval 10 ± 4 → 6 ± 2 km (−45%), total 55 ± 14 → 38 ± 9 km (−30%); HI 41 ± 15 → 30 ± 9, 9 ± 4 → 5 ± 2 and 50 ± 15 → 35 ± 9 km (−27%, −45%, −30%). Runs a week −10% (continuous −14 to −15%, interval sessions unchanged at 2). Continuous running slowed from 61-62% to 57-58% of race speed; interval running rose from 89% to 95% (RP) and 94% to 99% (HI).
Final laboratory interval session: day 14, three days before the last trial; fasted, 07:00-09:00, easy running only the day before; warm-up 10 min at 60% of race speed and two 10-s bursts at 90%; 300-m repetitions with 90 s recovery on a motorised treadmill, the number set by the prediction equations (2.7 ± 0.7 km in total); at season's-best race speed (RP) or 115% of it (HI).
Performance tests: fasted, 07:00-09:00, rest day before; isometric knee-extension maximal force (MVC) and rate of force development; 30 min later a 1500-m time trial on a self-paced ("on-response") treadmill at 1% gradient, speed and time hidden, distance visible, no encouragement. Test-retest coefficient of variation 0.9%.
Also measured: weighed food intake, accelerometer physical activity, peak blood lactate.
Analysis: two-way repeated-measures ANOVA with Bonferroni post hoc; magnitude-based inference against a smallest meaningful change of 1% (3.0 s), with 90% confidence limits and chances of benefit, triviality and harm.
Funding and interests: no funding from outside Loughborough University; no conflicts of interest.
Key Results
| Race-pace final session (RP) | Final session at 115% (HI) | |
|---|---|---|
| Final session speed | 5.8 ± 0.5 m/s | 6.6 ± 0.6 m/s (p = 0.002) |
| Final session volume | 2.7 ± 0.7 km | 2.7 ± 0.7 km |
| Final session heart rate | 169.0 ± 9.0 beats/min | 178.0 ± 7.0 (p = 0.001) |
| Final session RPE | 14.0 ± 1.0 | 17.0 ± 1.0 (p = 0.001) |
| Final session peak lactate | 3.8 ± 1.6 mmol/L | 9.9 ± 3.4 (p = 0.001) |
| 1500 m, control → after the taper | 298.8 ± 19.3 → 288.7 ± 18.5 s | 296.3 ± 18.4 → 292.1 ± 19.1 s |
| Improvement (± 90% confidence limits) | 10.1 ± 1.6 s (8.5 to 11.7 s); 3.4% | 4.2 ± 12.0 s (−7.8 to 16.2 s); 1.4% |
| Chances beneficial / trivial / harmful | 100% / 0% / 0% ("most likely beneficial") | 57% / 29% / 14% ("unclear") |
| Multiples of the smallest meaningful change (3.0 s) | 3.4 | 1.4 |
Individuals after HI: 6 of 10 improved beyond the smallest meaningful change; the 1st, 6th and 8th fastest runners got slower; one runner had his largest improvement of the study (5%) after HI.
Peak lactate after the trial rose with tapering (RP 7.1 ± 3.1 → 10.1 ± 2.6 mmol/L; HI 7.7 ± 2.4 → 9.3 ± 2.5; main effect p = 0.001); peak heart rate did not change.
Strength: MVC force rose with tapering (main effect p = 0.002; RP 663.4 → 722.3 N, +9%; HI 682.3 → 721.7 N, +6%; no interaction, p = 0.40); rate of force development rose (main effect p = 0.02; condition × time interaction p = 0.06).
Food and activity: energy intake and body mass unchanged; carbohydrate per km run higher in the taper; daily activity counts (p = 0.04) and moderate-to-vigorous activity (−1.6%, p = 0.03) lower in the taper; no difference between conditions in the three days after the final session.
Notes on the reporting:
No test of RP against HI for performance is reported; the comparison is two magnitude-based inferences, each against its own control (our note: magnitude-based inference has been criticised in the statistics literature as overstating evidence of benefit).
The final session's volume (2.7 km) was set by the prediction equations from uncontrolled usual training and exceeded the race distance; the worked example in Table 2 shows 6 × 300 m.
Trial times (about 289-299 s) were far slower than the runners' personal bests (mean 257.5 s), which the authors attribute to the treadmill, no opponents, no feedback and the early season.
The abstract and Conclusion call the HI result "possibly beneficial"; the Results classify it as "unclear" (57% beneficial, 14% harmful), and Table 5 marks it "possible".
Mechanisms
Measured: the HI final session was much harder (heart rate +9 beats/min, RPE 17 vs 14, lactate 9.9 vs 3.8 mmol/L) at the same volume; strength and rate of force development improved after both tapers; peak lactate after the trial rose.
Authors' interpretation: the ~30% volume cut from the elite-practice model may not leave enough recovery to benefit from a 115% session; a session longer than the race at that speed, so close to the trial, may have added fatigue in some runners; prior training load was uncontrolled and only one week of it informed the taper; runners did not rest more after the harder session; extra carbohydrate after it might have helped.The rise in strength may reflect recovery of function suppressed by heavy training rather than a gain.
Inference: with a single trial after each taper, the study cannot tell whether HI lowered the peak or delayed it (the authors note the same).
Practical Application
Is a last session faster than race pace better? Unclear: in 10 sub-elite runners a race-pace session 3 days out went with a 10 s faster 1500 m, a 115% one with 4 s, mixed. No direct test; a track trial ranked them the other way.
A modest, model-based taper worked consistently for sub-elite middle-distance runners: total running −30%, interval sessions kept, last session at race pace three days out — 3.4% faster on the treadmill (B, one trial; the prescription is the group's own model, REF-818).
If the last session is faster than race pace, keep it short and cut more volume around it: here 2.7 km at 115% inside a 30% cut was "highly variable"; the track follow-up with a 60% continuous-volume cut and 1,500 m at about 110% improved 7 of 8 runners (REF-1201). The two were not compared head to head (inference linking the two trials).
Strength recovered in the taper: knee-extensor strength (+6-9%) and rate of force development rose after a 7-day taper, whichever final session was used.
Limitations
Small and heterogeneous. Ten men whose personal bests ranged from 3:51 to 5:17; the authors note performance level did not explain who responded to HI.
No comparison between the conditions is reported; magnitude-based inference gives each condition a probability label, not a test of their difference.
Treadmill time trial, early season. Self-paced treadmill, fasted morning trials, no opponents, about 30-40 s slower than personal bests.
The final session was not standardised in volume (set from each runner's usual training; 2.7 ± 0.7 km), and only one week of usual training informed the prescription.
Unblinded. Runners could not be blinded to the change in load.
Accepted manuscript. Numbers were read from the author version in the university repository; the typeset article was not checked.
Funding: none external; no conflicts declared.
How this entry was checked and changed
Why this level
crossover with n = 10 (under 15), results reported as magnitude-based inferences for each condition rather than a test between them; read from the full text of the accepted manuscript (Nottingham Trent University repository), not the typeset version
Full text (checked 2026-10-26)
Performance was most likely improved after RP (mean ± 90% confidence limits, 10.1 ± 1.6 s), and possibly beneficial after HI (4.2 ± 12.0 s) (full text)
1500 m times: RP 298.8 ± 19.3 s before and 288.7 ± 18.5 s after the taper; HI 296.3 ± 18.4 s and 292.1 ± 19.1 s (full text)
The HI taper was unclear (SMC = 3.0 s, CI = -7.8 to 16.2 s, chances of a beneficial/trivial/harmful effect 57%, 29% and 14%) (full text)
Final session: speed 5.8 ± 0.5 vs 6.6 ± 0.6 m/s, heart rate 169.0 ± 9.0 vs 178.0 ± 7.0, RPE 14.0 ± 1.0 vs 17.0 ± 1.0, peak lactate 3.8 ± 1.6 vs 9.9 ± 3.4 (RP vs HI) (full text)
The volume of the final interval session was greater than 1,500 m race distance (2.7 ± 0.7 km) (full text)
Total running fell from 55 ± 14 to 38 ± 9 km (-30%) in RP and from 50 ± 15 to 35 ± 9 km (-30%) in HI (full text)
The mean coefficient of variation for time trial completion was 0.9% (full text)
The authors declare no conflicts of interest and the study did not receive funding from sources external to Loughborough University (full text)