How long to rest between sets

Singer et al. (2024) pooled nine randomised trials and 19 measurements. Comparing rest longer than 60 seconds against 60 seconds or less, the standardised mean differences were 0.13 for the arms (95% CrI −0.27 to 0.51) and 0.17 for the thigh (95% CrI −0.13 to 0.43); whole-body measures slightly favoured shorter rest at −0.08 (95% CrI −0.45 to 0.29). The posterior probability that longer rest is better was 74% for arms and 88% for thigh. That is a lean, not a verdict.

The same paper reports no appreciable further benefit beyond 90 seconds. Treat that with some caution: in their own network analysis the largest thigh estimate belonged to the very-long category (180 seconds or more, SMD 0.32), above both 61–119 seconds (0.13) and 120–179 seconds (0.01). That is an internal tension in the best paper available, and 90 seconds should not be read as a ceiling.

The trained-lifter case is where this gets thin. Six of the nine included studies used untrained participants — roughly 63 trained participants in total — and the authors say plainly they had insufficient data to subanalyse trained lifters. The most-cited trial that did use them, Schoenfeld et al. (2016), found significantly greater 1RM squat and bench press and greater anterior-thigh muscle thickness after eight weeks of three-minute rest versus one minute, in 21 young trained men. A 2025 preprint restricted to trained males reports only trivial hypertrophy differences and argues longer rest may not confer the assumed advantage; it has not been peer-reviewed. The honest position is that this is contested in exactly the population most people reading this belong to.

The short answer — More than 60 seconds as a floor. Two to three minutes on heavy compounds if you are trained. Anywhere in that range is defensible, and no study supports a precise optimum.

Grgic et al. (2018) reviewed 23 studies and 491 participants and concluded that resistance-trained individuals need rest longer than two minutes to maximise strength gains, while 60–120 seconds is sufficient for untrained people. Iversen et al. (2021) turn that into a working prescription: 1–2 minutes untrained, two minutes or more trained.

Two qualifications. Robust strength gains still occur with rest under 60 seconds — Grgic's review says so explicitly — so the difference is degree, not kind, and a rushed session is not a wasted one. And that 491-participant base was 413 male, so the two-minute rule for trained strength work rests on a modest, male-dominated evidence base.

The most replicated finding in this literature is also the least glamorous: short rest costs you reps later in the session. Willardson & Burkett (2006) had 15 trained men perform five sets at a 15RM load; reps fell from set one to set five under every condition, and two minutes beat 30 seconds in the squat and beat both 30 seconds and one minute in the bench press. Senna et al. (2011) found the same shape across five sets at 10RM in 15 trained men.

Longo et al. (2022) is the study that links that acute effect to adaptation. Twenty-eight people trained unilateral knee extension at 80% 1RM for ten weeks. Quadriceps cross-sectional area rose 13.1% with 180-second rest and 12.9% with 60-second rest when volume load was equated by adding sets — against 6.8% and 6.6% in the arms where volume load was allowed to fall. The short-rest arm averaged 9.8 ± 2.9 reps across three sets versus 16.1 ± 5.2 for long rest. Singer et al. (2024) endorse this reading: the small benefit of resting beyond 60 seconds is perhaps mediated by reductions in volume load.

That is one isolation exercise in one population over ten weeks, not a general law. But it reframes the question usefully. Rest is not a growth input in its own right. It is a mechanism for protecting the hard sets you were going to do anyway — which is the currency the volume model on our methodology page already runs on.

The familiar prescription — 30–60 seconds for hypertrophy, three to five minutes for strength — is citable to de Salles et al. (2009), which recommends short rest at moderate intensity explicitly on growth-hormone grounds. That rationale has since been dismantled by the researchers who built it. Schoenfeld (2013) reviewed the hormone hypothesis and found the transient post-exercise spikes more plausibly reflect fuel mobilisation than tissue anabolism. Henselmans & Schoenfeld (2014) confirmed that rest under a minute does raise acute serum growth hormone and does lower the testosterone-to-cortisol ratio, but described the link from those changes to chronic hypertrophy as highly contentious and apparently weak. Across the longitudinal trials they reviewed, none found superior growth with shorter rest and one found the opposite.

So there is no hypertrophy-versus-strength trade-off in the rest data. Longer rest is at worst neutral for growth and is the better bet for strength. Singer et al. (2024) go further and say current 30–90 second hypertrophy guidelines warrant reconsideration.

The convention says long rest for compounds, short for curls. The guideline literature agrees: ACSM's 2009 position stand recommends 3–5 minutes between sets for heavy loading in the 1–6RM range, 1–2 minutes in the 6–12RM hypertrophy zone, and under 90 seconds for muscular-endurance work.

The best acute head-to-head does not support a clean split. In Senna et al. (2011), the machine chest fly and leg extension — single-joint work — improved at every step from one to three to five minutes, whereas the bench press plateaued between three and five. On that data the isolation exercises were, if anything, the more rest-sensitive ones.

A defensible practical position, which we will label as convention rather than finding: rest longer on heavy compounds because those are the lifts where a lost rep costs you the most, and because more muscle mass per set means the session is shorter anyway (Iversen et al., 2021). Not because curls recover faster. They may not.

Zhang et al. (2025) meta-analysed 19 studies and 313 participants: supersets improved training efficiency with an SMD of 1.74 (95% CI 0.46 to 3.01), roughly a 37% reduction in session time, with no detectable difference in total repetitions, volume load, maximal strength or hypertrophy.

Which exercises you pair is the part that matters. Agonist–antagonist pairs allowed more repetitions than traditional sets (SMD 0.68, 95% CI 0.20 to 1.17), while pairing two biomechanically similar exercises for the same muscle significantly reduced volume load (SMD −1.08, 95% CI −1.72 to −0.44). Rows with presses, curls with triceps extensions: fine. Two chest movements back to back: that is short rest with extra steps.

The measurable cost is subjective rather than adaptive — RPE was higher with supersets (SMD 0.77) and blood lactate higher both during and after, while creatine kinase and acute swelling did not differ. The sample was predominantly young trained males, and the chronic strength and hypertrophy nulls rest on fewer studies than the acute results do.

Probably not. Henselmans & Schoenfeld (2014) found no support for predetermined rest being preferable to auto-regulated rest. Alonso-Aubin et al. (2024) tested it directly in 13 trained lifters on the back squat: self-selected rest differed significantly in duration from a fixed two minutes, yet mean propulsive velocity, performance and muscle oxygenation did not differ. That is one small acute study and the chronic question is untested — but it argues for treating a rest timer as a default you are free to override, not a rule you obey.

Shojin's rest timer is built to match that ranking. Rest is per-exercise rather than one global number: whatever you set on an exercise in a template is what counts down, and otherwise it falls back to your profile default (90 seconds out of the box). It starts only when you complete a working set — warm-ups and drop sets never trigger it, and re-editing a set you already logged doesn't restart it. If you've linked two exercises into a superset, the timer stays quiet between the pair and starts when the round finishes, which is the pattern the antagonist-pairing evidence actually describes. You can add or take 15 seconds mid-countdown, or stop it entirely. Given how much of this literature suggests self-selected rest performs about as well as prescribed rest, it is a default, not an instruction.

What this article can't tell you is the exact number. The evidence supports a floor — more than 60 seconds, and two minutes or more for heavy work if you're trained — and it does not support a precise optimum. The strongest synthesis available reports effect sizes of 0.13 to 0.17 with credible intervals that span zero, and its own authors say they had too few trained participants to analyse separately. Most of what gets repeated confidently about rest comes from acute studies measuring reps inside a single session, which is a different question from what happens to muscle over ten weeks; Grgic et al. flag that inference gap explicitly. No study in the 2024 meta-analysis measured torso muscles at all, the interventions ran only five to ten weeks, and the trained-lifter samples across this whole literature are small and heavily male. We don't know whether the two-minute strength recommendation generalises to women or to older adults, because it largely hasn't been tested in them. And the 2026 ACSM position stand — an overview of 137 systematic reviews and more than 30,000 participants — makes recommendations on load, range of motion, sets, exercise order, frequency and volume, while its abstract's summary recommendations don't mention inter-set rest at all. Read that silence as a ranking of priorities, not as an endorsement of any duration.

Every number in this article traces to one of these. Where the evidence is contested or thin, the article says so rather than picking a side.