SYNTH-103 · Guide

One session or two in a day

Two sessions in a day are the athlete's call: keep it whole if it fits, split it if it will not fit, costs too much or suits their day, count a second inside the week's budget and do not advise it for more signal: a close second bout raises some markers, enzymes are mixed, trained athletes show no clear performance gain; no trial compares double-threshold days with anything, no study measures daily time near VO2max across two sessions, no study links heart-rate drift to adaptation, and every gap and rule here is base practice

Main caveat: Few studies isolate the double. Most change glycogen, order, rest days or weekly work as well. This guide says which, study by study.
limited, mixed evidenceMostly from recreational to world-class athletes

The question it answers

An athlete has one session planned and a day that may not hold it, or has time for a second session and has heard that a second one gives a stronger signal. For threshold, for VO2max and HIIT, and for easy and long work: split, add, or leave alone, and what changes?

Short answer: most "twice a day" trials change something else as well (the fuel, the order of two different sessions, the rest days, the size of the week), so little here isolates the double. Treat a double as a scheduling tool with a known acute cost, not as an upgrade.

What this guide owns and what it leaves

Owns (decided here)Leaves to
Split, add or leave alone, by kind of workWho may run a double-threshold day and how → SYNTH-03; threshold formats → SYNTH-27
The stronger-signal case and how far it reachesHow many quality sessions a week → SYNTH-26; a session as a brief signal → PHYS-18
What a double is shown to cost, and what is unmeasuredEating in the gap → SYNTH-85, SYNTH-69; gaps to the next hard day → SYNTH-70; sleep → SYNTH-07; iron → SYNTH-97
Two easy sessions against one long one, as a choice for the dayWhether the long session is needed and how long → SYNTH-84; durability → SYNTH-32; lifting and riding on one day → SYNTH-31, SYNTH-70

Default — the decision, in order (what to prescribe first)

1

Keep the session whole when it fits the day and its cost is acceptable. Base practice; no study tests splitting against leaving it alone for adaptation.

2

A split is one session in two parts; a second session is a second session. Same planned work split in two takes one place in the week's count (SYNTH-03's wording, base practice). Added work takes a place in SYNTH-26's table (base practice).

3

Threshold: ask, then count. Two sessions in a day are the athlete's call at any weekly volume (the editor's decision; no study at this level): ask whether it suits them. Splitting one session takes one place in the count. Double threshold as a method still waits for SYNTH-03's four conditions (base practice), and threshold sessions a week follow SYNTH-27 (base practice).

4

VO2max and HIIT: keep it whole, shorten it or move it by default (base practice, set here). No study splits one hard interval session at matched work, for or against.

Key sessions stay ≥48 h apart by default, and a packed week is a planned block for well-trained athletes (SYNTH-70, base practice), with a third quality session only as a 2-4-week block (SYNTH-26, base practice). This is advice: an athlete who wants two interval sessions in a day decides (the editor's decision).

Do not refuse; say what is known and what it costs: quality in the second session, food in the gap, an easy day after.

5

Easy and long: the event decides. SYNTH-84's bands stand: optional up to ~2 h, spread in base for ~2-5 h with a regular long session in the last 8-10 weeks, a regular part of the specific phase beyond ~5 h (base practice).

6

Do not advise adding a session to chase signal (base practice). If they still want it, it is their call and it counts in the week. SYNTH-26's line against adding hard sessions to make up missing hours is the nearest rule (its count is base practice), and the signal case below has not been shown to reach performance in trained athletes, in trials too small to exclude it.

7

If two sessions fall on one day, put the key one first or on its own day (SYNTH-84 rule 2, base practice), set them as far apart as the day allows (base practice; a double-threshold day at least 5-6 h, SYNTH-03, base practice), eat in the gap (SYNTH-85; its gap bands are base practice), and let the next day be easy after a hard double (SYNTH-03, base practice).

1. Threshold

What exists. No trial compares a double-threshold day with anything, on any outcome, in any sport (SYNTH-03, B-). No split threshold session has been followed for adaptation, and none has been tried in cyclists. Below is the nearest evidence and what each study isolates. Both sides are thin.

For a split or a double.

Splitting cut the acute cost. It isolates the double at matched work: 6 × 10 min of moderate intervals, or 3 × 10 min twice with 6.5 h between, in 14 national-level men (11 skiers, 3 runners), running. sRPE load 743 vs 929, supine heart rate 60.7 vs 65.4 bpm, less next-morning fatigue and soreness (REF-159, B). Acute only.
The second part did not read worse. Heart rate fell 171 → 166 bpm and lactate 2.72 → 2.14 mmol/L from the first short session to the second, and perceived training quality showed no clear difference (REF-159, B). The authors suggest the lower cost could allow more time at this intensity; work was matched, so that is untested.
Some of the best train this way. World-class runners do three to four lactate-guided sessions a week, typically as morning and evening (REF-301, B+, narrative; its authors state no controlled study tests the model).What world-class skiers' 124 to 197 double days a year held is not in the abstract and the rise came with more easy sessions, so it is not a threshold result; it is read under volume below (REF-1559, B-, self-report, survivors only). Description, not cause.
Concentrating sub-threshold work raised power at 4 mmol/L in one week's block. Six sessions in 7 days, then 6 easy days: +4.0% vs −1.3% against regular training in 30 men, 15-min power no clear difference, +3.9% vs +0.2%, p = .14 (REF-137, B+). It isolates concentration; the entry describes no two sessions in a day.

Against.

The cheaper form is also a different session, and nobody knows what that does. In the single form heart rate rose 168 → 173 bpm, lactate 2.60 → 3.01 mmol/L and RPE 13.4 → 14.8; motivation before the second short session was 1.0 point lower, p = .041 (REF-159, B). No study links this drift to adaptation, so "splitting removes the stimulus" and "splitting loses nothing" are both unsupported.
More, shorter threshold sessions showed no clear extra gain. It isolates packaging across days, not within one: the same 136 min a week as two long or four short sessions.An 8-km time trial improved only within the two-session group (31.8 → 30.6 min, p = .04; four sessions p = .16), between-group effect trivial (ES 0.17), no between-group test, 9 and 6 analysed, 4 of 10 dropped out of the four-session arm against 1 of 10 (REF-1293, B-). It does not show that two beat four.
A closer gap did not clearly move the lactate curve further. It isolates the gap, with rest days also differing: a long ride plus sprints then intervals, 2 h or about 15 h apart, 3 weeks, 14 active men, 7 an arm.The curve shifted right in both: slope index 0.60 → 0.11 at 15 h and 0.56 → 0.10 mmol/L/min at 2 h, group × time P = .609, no equivalence test (REF-1547, B-; very likely REF-1535's men, where state 4 respiration and the fall in RPE favoured 2 h). No threshold session was done.
A second hard ride carries over. It isolates the double, which also doubles the work: a 75-min ride at 75% of VO2max 3 h after a first went with higher VO2, heart rate, temperature and stress hormones than the ride alone, in nine men (REF-1544, B-, one programme, one day). Calling that threshold work would be our label. It is an added ride, not a split.

What it means. The case for a split is an acute saving; the case against is only that nothing shows the saving is free or that the split builds as much, and nothing shows it builds less. SYNTH-03's default stands. For the athlete: a split is a fair way to get the session done; it is not shown to be a better session or a worse one.

2. VO2max and HIIT

What exists. No study measures daily time near VO2max across two sessions: the nearest reads each session on two double days (11, 10, 11 and 13 min in the top oxygen-uptake zone, pooled across morning and afternoon, REF-1543, B+), with no daily total and no one-session arm. No study splits one interval session at matched work. What is tested is moving several whole sessions into fewer days, which changes the double, the density and the calendar together.

For a double.

Compressing showed no clear loss in recreational men. Six sessions of 10 × 60 s in 5 days, 3 h apart with a meal between, or over 2 weeks: VO2max +7.7% and +6.1%, time to exhaustion +80.1% and +79.2%, no significant difference, 28 men, no equivalence test (REF-1542, B+). Four running sessions a week on 2 days or on 4: VO2max and Yo-Yo rose in all arms with no clear difference, 23 men, allocation not stated (REF-1538, B).
Interval output held across a packed week. Nine sessions in a week with three double days, 4-5 h and a meal between (the paper also prints at least 5 h): mean distance per 4-min interval 740 vs 736 m against three a week, 20 junior skiers (REF-1216, B+). A mean across the block, not first against second session.
Trained runners improved against usual training. Ten 5 × 4-min sessions in 7 days, with doubles at least 5 h apart: threshold velocity interaction P = .006; 5 km 2.7% and 2.3% faster, control 0.3% slower, P = .044; 33 athletes, 7 women (REF-1543, B+). It isolates the block. Nobody did these sessions on single days, and the block arms ran about two and three times the control's distance in the secondary sample (our arithmetic).
Oxygen uptake showed no clear change in two settings. Peak uptake 5.04 L/min in a first ~4-min effort and 5.00 L/min in a second 3 h later, no equivalence test, 12 men (REF-1546, B). Time at or above 90% of VO2max showed no clear difference (P = .753) with 3 × 3 min at minute 13, 36 or 69 of a 90-min ride, 16 cyclists (REF-1545, B). Neither is a test of output: the first effort was partly paced, and the second study fixed the power.

Against.

At matched work the packed week showed no gain, with a non-significant lean toward spreading. It isolates concentration, doubles included: the same nine sessions gave no evidence of a difference, with VO2max +0.9 spread and −0.8 mL/kg/min packed (p = .071, ES 0.67) and the 600-m trial −3 and −1 s (p = .280).Perceived recovery before hard sessions was median 6 and 4 in the loading week (p < .001); overall recovery-stress did not differ by the end (p = .510) (REF-1216, B+). The packed arm had no hard session in the week before retest, so the two causes cannot be separated.
Self-paced power fell when intervals followed a long ride. 297 vs 323 W, 1 h after a 90-min ride, 14 cyclists (REF-1097, B); high-intensity capacity was also compromised in REF-1096 (B, the authors' conclusion). The double, low glycogen, the order and the rest days all move together here. Three weeks of it brought about 10% on the time trial in both arms: no clear gain or loss from stacking, despite the lower interval power, 7 men an arm.
A real second effort slowed. World Championship pursuit finals ran about 1% slower than qualifying (4:18.7 and 4:21.1, P < .001, about 6.5 h apart) (REF-1546, B). Tactics and selection may explain part of it (our inference). No power was recorded.
VO2max showed no clear interaction in trained runners. P = .170 in 33; in a smaller analysis of the same trial (n = 29) P = .076 with +6.5% and +5.2% against +0.7%, while threshold (P = .071) and the 5 km (P = .053) also missed significance, so that analysis weakens the gains and this null alike.Time in the top heart-rate zone fell 47% → 28% and 45% → 16% through the block while time by speed held (REF-1543, B+); the authors read that as heart rate understating the work, so it is not a cost shown.The 10 × 60 s in REF-1542 (B+) was picked because it looked repeatable, so that trial says nothing on longer reps; 4-min reps as doubles appear only in the packed weeks above, where first and second sessions were not compared.

What it means. For a non-athlete who cannot spread sessions, a double day has shown no clear loss, in men. For an athlete it is a block, not a routine (base practice), and in the one trial against the same sessions spread it showed no evidence of a difference either way. Nothing here shows a double day adds to the sessions it contains. The FOR side is mostly recreational or uncontrolled for the double; the AGAINST side is small and confounded.

3. Volume: easy and long

What exists. In athletes, acute responses only. No training study compares one long session with the same time split for an event of 3 h or more (SYNTH-84, B-), and none in athletes reads adaptation. Both sides are thin, and each transfers from another sport or population.

For two sessions.

At low intensity the split was a little faster at no clear physiological cost. It isolates the double at matched distance: one ~3-h or two ~1.5-h roller-ski sessions 7 h apart, 13 national-level skiers. No significant heart-rate difference, split 1.9% faster, perceived muscular exertion 2.0 points lower straight after the split, training stress at 7 and 23 h with no clear difference (REF-1331, B-, abstract).
Short bouts added up in sedentary adults. Bouts of 10 min or more against one continuous bout: VO2max −0.50 mL/kg/min (95% CI −1.53 to 0.54), 8 trials (REF-1330, A-, mostly walking, every trial at high risk of bias). For a trained athlete this is a transfer, and a long one.
The molecular case for length is thin. PGC-1α gene expression was comparable at 60 and 90 min at 60% of VO2max, nine men (REF-1556, C+, abstract, no numbers). Over 2 h at low intensity, 12 national-level women changed little by what was reported: breathing 43.2 → 45.5 breaths/min in one gear, RPE up about 0.6 (REF-1555, B-, abstract, no heart rate).
Doubles are how the top skiers reach their hours. Double days 124 → 197 and sessions 441 → 519 a year, with more and longer easy sessions (REF-1559, B-). Description, in survivors, at 660 to 880 h a year by rough arithmetic (ours).

Against.

Only the long session rehearses the late hours. Pacing, fuel, contact points and the athlete's own fade (SYNTH-84 rule 6, base practice). The decline shows from about 80-140 min (SYNTH-84, B-): critical power held at 80 min and fell 9% by 2 h of heavy-intensity riding on placebo, and 60 g/h cancelled that fall (REF-145, B+); easy riding is a transfer. Two halves do not visit that state (inference).
Length goes with durability, and one gene appears only at length. Regular ≥90-min runs went with less economy loss over 90 min, +3.1% vs +6.0%, 26 men (REF-160, B, cross-sectional, not a trial). VEGFA expression rose only at 90 min (REF-1556, C+). Neither is adaptation shown by a longer session.
Fewer, longer sessions were not worse. Two sessions a week against four at matched weekly time: VO2max difference 0.7 mL/kg/min (95% CI −1.5 to 2.9, margin 3.5), 28 untrained adults, heavy and interval riding on back-to-back days, not a same-day split (REF-1329, B+). The one formal non-inferiority test here gives no reason to prefer the shorter form.
An added second session showed no clear gain in swimmers. It isolates extra volume, not a split: two 1.5-h swims a day against one for 6 weeks, no differences over the season, and sprint velocity fell in the doubled group while it rose in the other (REF-934, B, group sizes not given).Adherence in sedentary split-bout trials was mixed: raw drop-out 20% vs 16%, not tested, though split groups did more measured exercise, SMD 0.25, 6 trials (REF-1330, A-, every trial at high risk of bias).

What it means. Two easy sessions are a fair way to get easy time done, and easy work has little acute cost to save. What they do not give is the late hours, and for a long event that is what SYNTH-84 keeps the long session for. "Long sessions are unnecessary" is not a finding, and neither is "the split builds as much" or "the split builds less".

4. The stronger-signal case, laid out honestly

What rises (markers).

Close gap, low glycogen both ways, men who are not athletes, one project. At 2 h against 15 h, more nuclear TFEB and NFAT and more PGC-1α, PPARα and PPARβ/δ transcription (REF-1536, C+, no numbers or n in the abstract); MCT1 mRNA about 80% higher in 5 men (P = .038), and it rose in the control too (REF-1547, B-).
The second bout starts high. PGC-1α mRNA was about 8-fold up after the first of two 5 × 4-min sessions and stayed up through the second, with no difference by drink, 10 men (REF-1548, C+). The first session's pulse peaks within about 2 h (REF-1532, C+), so a reading after a close second bout may be two responses added, not one amplified (inference).
Low carbohydrate, pooled. 19 randomised studies: PDK4 (SMD 1.61), GLUT4 (1.38) and UCP3 (2.05) mRNA up; PGC-1α and TFAM no clear overall effect; PGC-1α up in a subgroup with a glycogen gap above 200 mmol/kg dw (SMD 2.08, I² = 75%) (REF-1537, A- for mRNA only). It isolates the fuel. A vote count before it had signalling up in 73% of 11 studies (REF-543, B+).
Stress hormones and IL-6. Larger after a second ride at 3 h than after one ride, with the work doubled (REF-1544, B-). A larger response is not harm, and could be stimulus: those papers cannot tell.

What does not follow. Papers from one project or programme count once below, and a review is not independent of any trial it includes.

Who, and what it isolatesMarkersEnzymes and contentPerformance
Untrained men, one leg each, the whole package (REF-1539, REF-1540, one programme, B-)PGC-1 family mRNA: no significant difference between legsCitrate synthase rose in both, more in the two-session legTime to exhaustion at 90% rose markedly more; gain in maximal workload reported as "identical", no equivalence test; no numbers
Active men, the gap at 2 h or 14-15 h (REF-1535, B-; REF-1536, C+; REF-1547, B-; one project)Gene signals higher at 2 h3 weeks: citrate synthase, mitochondrial area, PGC-1α protein no clear change in either arm; state 4 respiration favoured 2 hPeak VO2 rose in both; RPE fell more at 2 h; lactate curve no clear difference; no time trial
Recreational men, double with and without glucose against spread (REF-1538, B)Heat shock proteins rose in all, no clear differenceSuccinate dehydrogenase, vastus: +70% placebo double, +17% glucose double, +19% once a day; gastrocnemius +76%, +27%, +53%VO2max and Yo-Yo: no clear difference
Active people, the fuel, everyone doubled (REF-1548, C+; REF-1549, B)p38 higher after the second session on placebo; AMPK and PGC-1α no clear differenceCitrate synthase and COXIV rose, no clear difference250-kJ trial 211 → 244 W on 17 g, 203 → 219 W on 195 g between sessions, p = .02
Trained cyclists, ride then intervals with low glycogen (REF-1096, B; REF-1097, B)PGC-1α protein unchanged in both (REF-1096)Citrate synthase 45 → 54 and COX IV rose only in the stacked arm, within-arm (REF-1096); β-HAD protein up (REF-1097)About 10% on the time trial in both arms, both trials, 7 an arm
Elite men, the fuel, everyone doubled (REF-1541, B+)pACC +38 ± 19% overall, no clear group differenceCitrate synthase +11 ± 5% overall, no clear group differenceVO2max and 30-min power +5 ± 1% overall, no clear group difference; no once-a-day arm
Enzymes are mixed. Up in REF-1539, REF-1096 and the placebo arm of REF-1538, not in REF-1535, and not by fuel in REF-1541 or REF-1549. Twenty days of two a day went with citrate synthase and respiration about 40-50% up in 10 men, uncontrolled, in fixed order; respiration no longer differed 2 weeks later and citrate synthase stayed about 36% up, with no once-daily arm at that volume (REF-1557, B-): that is dose, not the split.
Below 300 mmol/kg dw, lower fuel cost work and added no detected signal. Starting at 88, 185 or 278 mmol/kg dw: capacity 18, 36 and 44 min, with AMPK and PGC-1α reported comparable at unequal work, 8 men, overnight rather than same-day (REF-1551, B-). Cutting carbohydrate after a double added no clear signal, nine men (REF-1550, C+).
Performance by status. Positive where people are untrained or only active (REF-1539, B-; REF-1549, B, which tests the fuel). No clear gain in trained cyclists (REF-1096, B; REF-1097, B) or, testing the fuel with everyone doubled, in elite men (REF-1541, B+).Pooled restriction in trained athletes: SMD 0.17 (−0.15 to 0.49), p = 0.29 (REF-218, A-); the entry does not list its 9 studies, so overlap with the trials above is likely (inference). Vote count: performance improved in 37% of 11 (REF-543, B+).
Markers can also read the other way. Respiration was about 20% lower and density markers 5-50% higher after 4 weeks of added intervals plus continuous work, with performance reported better days later, 27 athletes (REF-1534, B-, no control). Not a double-threshold week.

Both ways. FOR is real and narrow: some genes, in non-athletes, around one session, and a performance gain in an isolated leg and in active people restricted on fuel. AGAINST is also narrow: the trained nulls have 7 men an arm and 3 weeks, the elite null tests the fuel, and none rules out a small gain. So: a close second bout is not shown to be a stronger training stimulus for a trained athlete, and it is not shown not to be.

5. What it costs

Quality in the second session. Self-paced power fell after a long ride with low glycogen (REF-1097, B: 297 vs 323 W). It held, on means, with hours and a meal between (REF-1216, B+: 740 vs 736 m), and easy speed rose when split (REF-1331, B-: 1.9%). Speed held in a block but was partly prescribed (REF-1543, B+).At set power, perceived effort rose late in a ride with no clear oxygen change (REF-1545, B). On placebo, energy stores fell further in the second of two sessions (REF-1548, C+). No study compares free second-session output with first at the same hard work under a control.
Fuel. The fuel trials test the fuel, not the double. Restricting between sessions: no clear gain in elite men, 414 g vs 79.5 g (REF-1541, B+); a gain in active people, 17 g vs 195 g (REF-1549, B); higher bone-resorption marker on about 3 g/kg against about 12 g/kg, one day, a marker (REF-1550, C+).Across 4 weeks of hard doubles, 10 g/kg/day went with +65% glycogen and +10.7% trial power, against flat and +1.6% on 5 g/kg/day, 22 rowers (REF-1554, B+, in training, P not clearly between arms). SYNTH-85 stands: fuel the first, eat at once, cover the day.
The gap. Only two programmes vary it, and neither alone. 3 h against 6 h with one more meal: lower VO2, heart rate and most hormones at 6 h, IL-6 and IL-1ra no clear difference (REF-1544, B-); gap and food cannot be separated. 2 h against about 14-15 h with low glycogen: gene signals higher at 2 h, no clear change in content, rest days differing (one project, REF-1535, B-).Other gaps were fixed, not tested: 6.5 h (REF-159, B), 7 h (REF-1331, B-), 3 h (REF-1542, B+). No gap is tied to adaptation or performance in athletes; "as far as the day allows" is base practice, and a shorter gap is not shown to be a stronger stimulus.
Stress markers. Peak IL-6 8.8 pg/mL after a second ride at 3 h and 5.2 after one ride, work doubled, nine men (REF-1544, B-). In 16 fed elite rowers, IL-6 (about 7.5-fold) and hepcidin (about threefold) sat on a plateau through a second submaximal session 2.5 h later, p = 1.00, no single-session arm (REF-1552, B-). Different contrasts, so no contradiction. Neither reads harm.
Sleep. Moving a morning session from 07:00 to 09:00 added 1.0 ± 0.5 h of night sleep, 5.9 to 6.9 h, with no significant difference in time-trial change (100 m p = .629, 800 m p = .884), hand-timed, 27 adolescent swimmers on 8-day camps; both arms doubled (REF-1558, B+). It isolates the hour, not the double. No adult double-day study reads sleep with performance. One whole night awake is a different thing (REF-1533, B-, students, no training).
Injury: not measured for doubles. No controlled study exists. One observational link: among many factors in a living review of trail runners, injury went with two sessions a day (REF-783, B+, no size or dose here). A single run more than 10% longer than the past 30 days' longest went with more overuse injury in 5,205 runners (REF-1102, B+); that is length, not doubles, and whether two shorter runs avoid it is untested (inference).
Women: thin. Hepcidin stayed above a rested baseline to the next morning in 13 university runners on about 6 g/kg, uncontrolled, already up before the second session (REF-1553, B-): no cost of doubling shown. On a packed week women's recovery-stress state fell below men's, 11 vs 21 (p < .001), 10 women (REF-1216, B+).Women were 10 of 22 in REF-1554 (B+) and 7 of 33 in REF-1543 (B+) with no result by sex. Most signal and stress studies have no women; nothing reads menstrual phase.
Masters: almost nothing. No double-day study in the base. After one hard session, 9 masters cyclists showed no evidence of a difference from 8 young riders in sprint, strength and 30-min power over 48 h, and rated motivation, fatigue and soreness worse, not significantly in post hoc tests (REF-1296, B-); that is not a double. Widen the gaps on recovery signals, not on age (SYNTH-70, SYNTH-35, base practice).

Decision — who does what

Who-and-what columns are base practice throughout unless a source is named for that number; the right column says how sure.

WhoWhat to doSource and how sure
Anyone, session fits, cost acceptableKeep it wholeBase practice; no study tests the alternative for adaptation
Threshold, any weekly volume, session will not fit, costs too much or suits their dayTheir call: ask. Split morning and evening, one place in the count. Threshold sessions a week by SYNTH-27 (a second only above ~8 h/week, on separate days)Editor's decision; SYNTH-03, SYNTH-27, base practice. REF-159, B: acute, moderate intervals, 6.5 h gap, well-trained men, no adaptation. No study below that level, for or against
Two sessions in a day, any weekly volume, because it suits how their week runsTheir call: ask. Each real session counts in SYNTH-26's table; say which one is the key oneEditor's decision; the count is base practice. No study at this level
12 h+/week, Tier 3+, lactate meter, used to two a dayOne double-threshold day a week, held for a block, judged on a marker chosen in advanceSYNTH-03, base practice. No trial
Tier 4-5, elite volume, coach, lactate controlThe model as describedREF-301, B+, narrative, no controlled study
VO2max or HIIT, session will not fit, any levelShorten or move it by default. If they prefer to split, it is their call: say a split is untested and may not be the same sessionBase practice. No study of a split
Anyone wanting two interval sessions in a dayAdvise ≥48 h first (SYNTH-70, base practice); then their call. Say what was seen and what it costs: no clear loss in recreational men, with a meal between in the one trial that reports the gap; food in the gap; an easy day afterEditor's decision. REF-1542, B+; REF-1538, B. Not a test of equality; not athletes or women
Trained athlete planning a loading weekA block, if at all, not a routine; SYNTH-26, SYNTH-70 and SYNTH-90 set it (base practice). Expect a recovery dip, and do not judge it by heart rate aloneREF-1543, B+ against usual training; REF-1216, B+ no evidence of a difference against the same sessions spread
Anyone wanting more signalDo not advise it. If they still want it, it is their call and it counts in the weekSYNTH-26, base practice. Section 4: markers, not performance
Long ride then intervals as routineNo; if the intervals matter, first or on their own daySYNTH-84 rule 2, base practice; REF-1097, B, lower interval power; REF-1096, B; time trial about 10% either way, no clear difference
Easy work, event up to ~2 hTwo easy sessions are fineSYNTH-84, band is base practice; REF-1331, B-, acute, transfer
Event ~2-5 hSpread in base; regular long session in the last 8-10 weeksSYNTH-84, base practice
Event over ~5 h, or a first at that lengthKeep the long session in the specific phaseSYNTH-84, base practice
Moderate work before a key daySplit it for the lower cost if it suits them, or shorten or move it, and say what changesSYNTH-84 rule 3, base practice; REF-159, B
Hours to spare, adding an easy second sessionA volume step, not a signal trick; ramp by SYNTH-29 (base practice)REF-1559, B-, description; REF-934, B, an added swim showed no clear gain, group sizes not given
Any doubleEat in the gap by SYNTH-85 (gap bands base practice)REF-1541, B+, no clear gain from restricting; REF-1554, B+; REF-1550, C+; against them a gain on 17 g in active people, REF-1549, B. Fuel, not the double
Sleep is the concernMove the morning session later if the day allows (base practice)REF-1558, B+, adolescents, transfer of direction
Women, mastersSame rows; ask about food and iron (SYNTH-97); widen on recovery signals (base practice)Thin: REF-1553, B-; REF-1216, B+; REF-1296, B-

The four practical cases

A session that does not fit the day. Ask what it is for. Threshold or moderate: split if it suits them, at any weekly volume (their call), or shorten or move (base practice). VO2max or HIIT: shorten or move by default; a split is untested, and if they prefer it, it is their call (base practice).Easy: split freely, unless it is the long session the event needs (base practice, SYNTH-84). Say what was saved and what is unknown: nothing shows the split builds as much, or less.
A session that costs too much in one piece. At moderate intensity a split lowered the cost (REF-159, B). At low intensity there was little extra cost to remove (REF-1331, B-), so this reason is weak for easy work. For hard intervals no split is tested; cut reps instead (base practice). A lower cost is not a proven equal gain or a proven loss, since neither study measured adaptation.
Adding a second session for more signal. Not advised, and still their call. Some markers rise in non-athletes around one session (REF-1536, C+); in trained cyclists about 10% came either way (REF-1096, B; REF-1097, B). A second hard session also spends the week's budget (SYNTH-26, base practice). If a second session is wanted for hours, call it volume and ramp it (base practice).
Two easy sessions instead of one long one. Fine for events up to ~2 h and in base for ~2-5 h (base practice, SYNTH-84). In the last 8-10 weeks before a ~2-5 h event, and in the specific phase beyond ~5 h, keep the long one or move it (base practice). The split was skied a little faster with no clear next-day difference (REF-1331, B-). What it does not give is late-hours rehearsal (SYNTH-84 rule 6, base practice).

Conflicts and how they are resolved

"The drift is the stimulus" against no link at all. Base lines have said that splitting removes a stimulus. REF-159 measured a lower response, not a lost adaptation, and REF-1331 and REF-1555 limit it at low intensity. Resolved: keep a durability session whole for the late-hours rehearsal, as base practice, and do not cite drift as the reason.
[REF-1539] against [REF-1540]. Same laboratory and design: a gain in enzymes and exhaustion time, no significant difference in PGC-1 family mRNA. Resolved: one programme, and a performance result without the marker difference is a further reason not to steer by the marker.
[REF-1549] against [REF-1541]. A gain on low carbohydrate in active people, no clear gain in elite men, 2 weeks against 4. Both tested the fuel. Resolved: no restriction advice for athletes; SYNTH-85 stands.
[REF-1543] against [REF-1216]. A week of doubles beat usual training, and gave no evidence of a difference against the same sessions spread. Different comparators, and the first control ran about half to a third as far. Resolved: a block may help against running less; against the same work spread one trial shows no evidence of a difference, which does not exclude one either way.
[REF-1544] against [REF-1552]. Higher IL-6 against one bout; no extra rise against the first bout. Different contrasts, intensity and feeding. Resolved: quote each with its contrast.

Practical Application — how the assistant says it

"Split it, or add a second one for more signal?" Split if it will not fit or costs too much. We do not advise one for signal: markers rose, trained athletes showed no clear gain, and double-threshold days are untested.
"Is a double a stronger signal?" For some genes around one session, in men who are not athletes (REF-1536, C+). That is not shown to be more fitness: the same project's three-week arms, 15 men, showed no clear change in mitochondrial content, in a trial too small and short to exclude one, while state 4 respiration and the fall in RPE favoured 2 h (REF-1535, B-).
"Can I split my threshold session?" Yes, if it suits your day; that is your call at any weekly volume, and it takes one place in the week's count (base practice). One acute study shows it costs less (REF-159, B). Nobody has measured what it builds.
"Do I lose the stimulus if I split?" Nobody knows. Heart rate and lactate rose within the single form and fell from first to second session in the split form (REF-159, B), and no study links that rise to adaptation. Keep a race-specific long session whole for the rehearsal (base practice).
"I train under 8 hours a week. Can I train twice a day?" If it suits your week, yes; that is your call. Each real session counts, and no study tests this at your volume, for or against.
"Intervals twice a day?" Our default is ≥48 h between key sessions (base practice), and it is your call. If you do it, eat in the gap and keep the next day easy (base practice). Recreational men lost nothing clear by compressing (REF-1542, B+); skiers gave no evidence of a difference from spreading, and felt less recovered in the loading week (REF-1216, B+).
"How far apart?" As far as the day allows (base practice). No gap is tied to a training result in athletes. A short gap is not a stronger stimulus by any evidence that reaches performance.
Say what the base cannot judge: double-threshold days, a split VO2max or HIIT session, daily time near VO2max, drift as a stimulus, injury from doubles, masters, and menstrual phase.

What NOT to do

Say "two sessions give a bigger signal, so more adaptation", or use mRNA as a reason to add a session.
Say "splitting removes the stimulus" or "the drift is what you are after". Say "splitting loses nothing" either.
Count REF-1535, REF-1536 and REF-1547 as three tests, or REF-1539 and REF-1540 as two, or the four Ronsen papers in REF-1544 as four.
Use a both-arms-doubled trial (REF-1541, REF-1548, REF-1549, REF-1550, REF-1554) as evidence on whether to double.
Read REF-1293 as two sessions beating four, or as a test within one day.
Read REF-1097's 297 W as the cost of a second session alone; glycogen, order and rest days move with it.
Read REF-1543 as a test of doubles, or REF-1559 as doubles causing the top. Quote REF-1553 as a cost of doubling.
Prescribe restriction between sessions from REF-1549; it was active people for 2 weeks.

Evidence gaps

No study tests: double-threshold days against anything; a split VO2max or HIIT session at matched work; daily time near VO2max from one session or two; drift against adaptation; one long session against the same time split for adaptation, durability or a long event, in athletes; intervals done late in a long session over weeks; a single-session control at matched work for any second-bout marker study; the gap, alone, on any outcome in athletes; injury from doubles under control; overreaching from doubles; adult sleep on double days with a performance outcome; menstrual phase or a sex comparison on double days; masters on double days; whether adding doubles raises total hours outside the elite.

Limitations

Few studies isolate the double. Most change glycogen, order, rest days or weekly work as well. This guide says which, study by study; a reader should not add them up. "No study" means none in the base or in a search that was not systematic.
Not independent. One project (REF-1535, REF-1536, REF-1547); one programme each for REF-1539 with REF-1540, and for REF-1544's four papers. REF-543's tally includes REF-1541 and REF-1096; the overlap of REF-218 and REF-1537 with the trials here is not stated.
Small and short. Most arms 6 to 15, most trials 1 to 12 weeks, mostly men. A null here is not equality, and none has an equivalence or non-inferiority test except REF-1329, which is not a same-day comparison.
Practice, not evidence: every gap, count, hour band and order in the Default and the Decision table, and the rule that a split takes one place in the week.
How this entry was checked and changed
Why this level

(the default's direction, that a second session has shown no clear added gain in trained people and that splitting lowers the acute cost of moderate work, rests on B+ and B entries: REF-1216 (B+), REF-1097, REF-1096, REF-159 (B), with REF-1331 B- for easy work; REF-1542 (B+) and REF-1538 (B) speak only to compressing sessions in recreational men. B-, not B, because the main question is untested for threshold and for one split hard interval session, the signal evidence is mixed, and every gap, count and order in the default is base practice. The A- entries do not lift it: REF-218 and REF-1537 pool the fuel, not the double, and REF-1330 is sedentary walkers)

Changelog
2026-11-02: new framework on one session or two in a day, for threshold, VO2max and HIIT, and volume, built from REF-1535 to REF-1559 with REF-1532, REF-1533, REF-1534 and the existing entries on its Level line. Two sessions in a day made the athlete's call at any weekly volume, and ≥48 h between key sessions made a default rather than a bar (the editor's decision, 2026-11-02; SYNTH-03 and SYNTH-70 changed with it). SYNTH-26, SYNTH-84 and SYNTH-85 unchanged. Level B-.