Back Squat Tempo: What the Evidence Actually Supports

By Szymon Spławski, a lifter and the developer of Repko, a four-phase tempo timer for iPhone. That is a commercial interest in you caring about tempo, so it is stated up here rather than buried at the bottom. Every prescription below carries a confidence rating, and the honest-gap section lists what the evidence does not support.

Two controlled trials have manipulated eccentric duration on the back squat itself. They reached opposite conclusions. That disagreement is the most useful thing in this literature, because the likely reason for it tells you more about how to prescribe tempo than either study does alone. This page works through what has actually been measured on this lift, separates it from what has been extrapolated, and marks the difference on every row.

• The two direct trials conflict, and the concentric is the probable reason. Kojic (2025) found a 4-second eccentric produced larger strength and hypertrophy gains than a 1-second eccentric, in untrained lifters whose concentric was also controlled at 1 second. Hao (2026) found a 4-second eccentric produced the smallest strength gain and significantly reduced jump and sprint performance, in trainees whose concentric was explosive.
• Velocity loss is far better evidenced than tempo. Three squat-specific trials and two meta-analyses converge on a usable rule: stop the set at 10 to 20 percent velocity loss for jump and sprint qualities, allow 20 to 40 percent when hypertrophy is the goal. The evidence for prescribing a specific eccentric duration is nowhere near that quality.
• Depth matters more than tempo, and half squats are the risk. In a 10-week trial, the half-squat group gained nothing and was the only group whose pain, stiffness and disability scores rose. There is no comparable safety signal for any tempo prescription, in either direction.

The two direct trials, and why they disagree

Almost every tempo recommendation you will read for the back squat is extrapolated from other exercises. Two studies actually manipulated eccentric duration on the squat over a training block, and they are worth reading against each other rather than picking whichever one supports the prescription you already liked.

Kojic, Mandic and Duric (2025) randomised 18 untrained participants to a fast eccentric (1-0-1-0) or a slow eccentric (4-0-1-0) for seven weeks at 60 to 70 percent of repetition maximum, with sets, rest and load matched. The slow group gained more one-repetition maximum strength (effect size 1.60 against 0.99) and more vastus lateralis cross-sectional area (1.74 against 1.37). On its own this reads as a clean win for slow eccentrics.

The design detail that matters is buried in the notation: the fast group's concentric was also one second. Both groups lifted the bar at a controlled speed. Nobody in that study was trying to move the bar fast.

Hao and colleagues (2026) ran eight weeks with three groups, all of whom pushed the concentric explosively: 4-0-X-0, 2-0-X-0, and a self-selected tempo. On squat one-repetition maximum there was no significant difference between groups. The significant findings were on the other outcomes, and they went the other way: the four-second eccentric group's countermovement jump fell significantly (effect size -0.90), squat jump fell (-0.83), and 30-metre sprint declined significantly, while the two-second group improved on all three.

Read together, the plausible reconciliation is that a long eccentric is not free. It buys time under tension and it costs the ability to express force quickly. When the concentric is deliberately slow anyway, as in Kojic, that cost is invisible because the study never measured anything requiring speed. When the concentric is explosive and jump and sprint are measured, as in Hao, the cost shows up clearly.

A third trial complicates even this. Gao and Li (2026) put 30 collegiate soccer players through six weeks of complex training with 5-0-X-0, 1-5-X-0 and 2-0-X-0, matched for load and volume, and the slow-eccentric group produced the largest one-repetition maximum increase at 17.6 percent, while its squat jump height fell 6.1 percent. That is the strength result Hao did not find and the jump penalty Hao did find, in the same group. The strength question is genuinely open.

The best synthesis available is Amdi and King's 2025 meta-analysis, and it is thin by its own account: nine studies, 166 participants in total, 46 percent trained, 77 percent male, and not squat-specific. Its estimates are worth quoting precisely. Shorter eccentrics were better for countermovement jump at moderate certainty (g = -0.73). Maximal strength overall was uncertain (g = 0.25, 90 percent confidence interval -0.17 to 0.67). Hypertrophy was essentially null (g = 0.05, interval -0.22 to 0.33). The one subgroup that favoured longer eccentrics was trained participants (g = 0.33, interval 0.07 to 0.60) and volume-load-matched trials (g = 0.25, interval 0.04 to 0.45).

So the honest headline is not "slow eccentrics build more muscle". On the current evidence, eccentric duration is a weak lever for hypertrophy, an uncertain one for strength, and a reliable one for making your jump worse.

Back squat tempo prescriptions

Every row below names what supports it. Rows marked as convention have no supporting trial and are labelled as such rather than dressed up.

Pause squats against rebound squats

One trial has compared these over a training block. Martínez-Cava and colleagues (2021) randomised 26 men to a roughly two-second pause or a rebound, for ten weeks of velocity-based squat training that differed in nothing else. Both groups improved on most neuromuscular tests. The pause group's effect sizes ran 0.76 to 1.12, the rebound group's 0.45 to 0.92. Sprint and Wingate results were not significant but leaned to the pause group; countermovement jump and the 10 to 20 metre sprint split leaned to rebound.

That direction is coherent, and it is what a coach would predict: removing the stretch-shortening contribution biases the training toward concentric force production, keeping it biases toward elastic qualities. But this is 13 lifters per group and the differentiating results are effect-size trends rather than significant interactions. Treat it as one good study pointing a plausible way, not as a settled finding.

On hypertrophy, there is nothing. A search for paused-squat hypertrophy trials returns no such study. If someone tells you paused squats build more muscle, they are telling you what they believe, not what has been measured.

Time under tension is not a free-standing variable

Schoenfeld, Ogborn and Krieger's 2015 meta-analysis is the citation everyone reaches for, and it is worth quoting for what it actually concluded: hypertrophic outcomes are similar across repetition durations from 0.5 to 8 seconds. The authors also flagged that durations beyond 10 seconds per repetition appear inferior, while noting themselves that a lack of controlled studies makes that hard to state definitively.

The more useful finding is Martins-Costa and colleagues (2022), who equated time under tension at 36 seconds per set across a 3-second and a 6-second repetition protocol for ten weeks. There were no differences in cross-sectional area and none in one-repetition maximum. Their conclusion is the sentence to keep: training volume cannot be considered separately from time under tension. If you double the repetition duration and hold the set structure, you have not added a stimulus, you have traded repetitions for seconds.

Morton and colleagues (2019) closed the mechanistic door on the usual explanation. Taking sets to failure with different loads and different repetition durations produced significantly different total time under load, repetition counts, volume and electromyographic amplitude, and yet glycogen depletion in both fibre types and phosphorylation of anabolic signalling proteins showed no difference at all. Fibre activation at failure appears to be independent of how long the repetition took.

Acute markers do move. Wilk and colleagues (2021) had national-level powerlifters perform five sets to failure at 80 percent, comparing 5-0-3-0 against 2-0-2-0: the slow condition produced higher lactate and higher creatine kinase, with no difference in testosterone, growth hormone, insulin-like growth factor 1 or cortisol. Gepfert and colleagues (2021), in ice hockey players, found the opposite hormonal pattern, with the fast tempo producing higher growth hormone, insulin-like growth factor 1 and cortisol. Two acute studies, contradictory hormonal results, neither with a hypertrophy outcome. This is not a foundation to prescribe from.

Velocity loss, the lever that is actually well evidenced

If you want a variable with real evidence behind it on this lift, it is not eccentric duration. It is where you stop the set.

Pareja-Blanco and colleagues (2017) trained two groups for eight weeks, terminating sets at 20 percent or 40 percent velocity loss. Both gained similar one-repetition maximum strength, but the 20 percent group did so performing 40 percent fewer repetitions, improved countermovement jump by 9.5 percent against 3.5 percent, and preserved the myosin heavy chain IIx fraction that the 40 percent group reduced. Rodríguez-Rosell and colleagues (2021) extended this to 10, 30 and 45 percent thresholds and found the same pattern, with the 10 percent group posting the best jump improvement at 11.9 percent.

The largest trial to date, Sánchez-Valdepeñas and colleagues (2026), randomised 51 trained men across four thresholds. There was a significant group effect for one-repetition maximum, with 20 percent velocity loss largest. Hypertrophy trended toward the 40 percent group but the interaction did not reach significance, and the authors say so. Jukic and colleagues' 2023 meta-analysis reports the same shape across the wider literature: velocity loss threshold did not influence strength gains, higher thresholds slightly favoured hypertrophy, and lower thresholds preserved jump and sprint.

One caution that gets left out of most velocity-based training articles. Banyard and colleagues (2017) tested whether you can predict a back squat one-repetition maximum from the load-velocity profile. The best method reached a correlation of 0.93, which sounds usable, but every prediction differed significantly from the measured one-repetition maximum, and the velocity at one-repetition maximum was unreliable between sessions with a coefficient of variation of 22.5 percent. Velocity loss as a set-termination rule is well supported. Velocity as a way to guess today's one-repetition maximum is not.

Depth beats tempo, and half squats are the actual risk

Pallarés and colleagues (2020) trained 53 resistance-trained men for ten weeks at full, parallel or half depth, with individual range of motion standardised. The full-squat group was the only one to improve strength at all three depths tested. The parallel group came second. The half-squat group produced no meaningful neuromuscular or functional improvement at all, and was the only group whose pain, stiffness and physical functional disability scores increased significantly.

Bloomquist and colleagues (2013) found the same asymmetry over 12 weeks: deep-squat training transferred to shallow strength, shallow training barely transferred to deep, and the deep group gained more front-thigh cross-sectional area and a 15 percent better squat jump.

This matters for tempo prescription because the two are often confused as safety measures. Slowing the eccentric to "protect the knees" while cutting depth is doing the opposite of what the evidence supports. Hartmann and colleagues' review places peak retropatellar compressive stress at around 90 degrees of knee flexion, with stress falling as you descend past it, and concludes that half and quarter squats with the supramaximal loads they permit are the long-term degenerative risk. Zavala and colleagues (2021) measured patellofemoral joint stress directly and found it greatest below parallel within a given load, which does not fully reconcile with Hartmann. Both are reported here because they disagree and you should know that.

High bar against low bar, weaker than you have been told

The standard claim is that the low-bar squat shifts load to the hip and the high-bar to the knee. The best-controlled biomechanical test of that claim did not find it.

Larsen and colleagues (2024) used motion capture, force plates and statistical parametric mapping across the whole concentric phase at three-repetition-maximum loads, with stance width standardised. They found no significant differences in net joint moments, muscle forces or muscle-specific moments at hip, knee or ankle between the two bar positions, and concluded that placement should be chosen on comfort. The same group's 2021 study, where stance width was allowed to vary, did find the expected hip and knee differences. Read together, stance width appears to be doing much of the work that bar height gets credited with.

Lee and colleagues (2026) reached a similar place from electromyography, reporting that no combination of bar position and stance width produced consistent activation advantages. The one finding that ties bar position specifically to the eccentric phase is Murawa and colleagues (2020), who found higher low-bar activation during the eccentric at 60 and 65 percent of one-repetition maximum in 12 powerlifters. That is a small study at submaximal loads, and it is the whole of the evidence for tempo interacting with bar position.

What a slow eccentric does to the tendon

A common justification for slow eccentrics is tendon adaptation. The direct measurement points the other way.

Earp and colleagues (2016) combined ultrasonography, motion capture and force plates on parallel back squats at 60 percent of one-repetition maximum, comparing a slow fixed tempo, a volitional tempo and a maximum-speed jump squat. Peak quadriceps tendon length, patellar tendon force and rate of force development were all greater in the jump condition than in the slow tempo. The tendon did stretch more during the early eccentric of the slow condition, behaving viscously, but peak strain across the movement was higher when the bar moved fast.

If tendon stiffness is the adaptation you want, the evidence says load it quickly, not slowly. That is the opposite of what most tempo programming assumes.

Common errors

Treating time under tension as a stimulus you can add. You cannot hold set structure and load constant and increase time under tension. Martins-Costa 2022 equated it and found nothing.

Prescribing a slow eccentric to an athlete in season. Hao 2026 is the clearest result in the whole literature: a four-second eccentric significantly reduced countermovement jump, squat jump and 30-metre sprint over eight weeks.

Cutting depth to spare the knees. The half-squat group in Pallarés 2020 was the only group that got worse on pain and disability.

Setting today's load from a velocity-predicted one-repetition maximum. Banyard 2017 found every prediction method differed significantly from the real value, with a 22.5 percent coefficient of variation on velocity at one-repetition maximum.

Assuming form breaks down visibly under fatigue. Brice and colleagues (2020) took skilled lifters to failure at 80 percent and found joint ranges of motion unchanged. What changed was the distribution: knee moment fell while hip and lumbo-pelvic moments rose. The technique looks the same and the loading is not.

Frequently asked questions

What tempo should I use for the back squat?

For general strength and hypertrophy, anything with a controlled eccentric between two and four seconds and an explosive concentric intent will do, because the evidence does not distinguish within that band. Use 2-0-X-0 as a default. Choose a longer eccentric only when you are in a block with matched volume load and no jump or sprint requirement, and a shorter one when speed qualities matter.

Do slow eccentrics build more muscle on the squat?

There is no good evidence that they do. The only meta-analysis on eccentric duration puts the hypertrophy effect essentially at zero (g = 0.05), and Schoenfeld's repetition-duration meta-analysis found outcomes similar anywhere from 0.5 to 8 seconds. The one squat trial showing a hypertrophy advantage for a four-second eccentric used 18 untrained participants over seven weeks.

Is the low-bar squat more hip-dominant?

Less clearly than commonly claimed. When stance width was standardised and loads were heavy, no significant differences in joint moments or muscle forces were found between bar positions. Stance width appears to explain much of the difference that bar height usually gets credited with.

Should I squat slowly to protect my knees?

No evidence supports that, and the depth literature points the other way: half squats produced no gains and were the only condition associated with increased pain, stiffness and disability in a ten-week trial. There is no injury or pain outcome for any tempo prescription on this lift, in either direction.

How long should the pause be in a paused squat?

Unknown. The only training trial used a single pause of roughly two seconds. No study has compared pause durations, so any specific number is convention.

The honest gap: what we don't know

No randomised trial has tested back squat eccentric tempo in trained lifters with a hypertrophy outcome. The nearest thing is 18 untrained participants over seven weeks. No trial has isolated pause duration. No trial has compared paused against rebound squats for hypertrophy at all. No trial has crossed tempo with depth, bar position or range of motion; every one of those variables has been studied alone. No adequately powered study has looked at whether any of this differs by sex. No study has measured injury, pain or joint symptoms as an outcome of a tempo prescription, which means the safety arguments made in both directions are equally unsupported. The longest squat tempo intervention found is eight weeks.

The numbers in this page's tables sit inside ranges the evidence supports, but their precise selection is coaching judgement rather than a result anyone has measured.

The same standard applies to the tool. Repko has no camera, no sensor, and no way of knowing whether your fourth second was actually a fourth second. It marks each phase and trusts you to follow it, which means it removes the counting, not the discipline. A device that verifies your tempo does not exist, not in this app and not in any other. What exists is a timer that holds the number you already chose, at rep eight, when you are tired and counting badly. If you were hoping for more than that, nobody can sell it to you yet.

Closing

The back squat has better tempo evidence than most lifts and it still does not support the confidence with which tempo is usually prescribed. Two direct trials disagree. The meta-analytic hypertrophy effect is zero. The one thing that replicates cleanly is that long eccentrics cost you speed. Meanwhile the levers with real evidence behind them, depth and where you stop the set, get a fraction of the attention.

Pick a tempo you can actually hold, keep the concentric fast unless you have a reason not to, squat deep, and put your programming effort into the variables that have been measured.