08/22/2026
Lifting to failure is usually treated as the honest way to train. In 158 men, stopping halfway to it built more strength than going all the way.
Bar speed falls inside a set as the muscle fatigues. Velocity loss is simply how far it has fallen when you rack the bar, and it works as a live readout of how close that set came to failure. A 25 percent loss means the last rep moved three quarters as fast as the first. A 50 percent loss means it moved half as fast, which is close to the point where nothing else could be lifted. Because the number is measured rather than guessed, it does the same job as reps in reserve without depending on anyone's honesty about how hard a set felt.
The physiology behind the split is fatigue itself. Extending a set forces recruitment of the largest, fastest motor units and drives the metabolic disturbance thought to trigger growth, which is why longer sets tend to build more tissue. The cost is that those final grinding reps are slow, and rehearsing slow reps trains the nervous system toward the wrong end of the force-velocity relationship. Stopping earlier keeps every repetition fast, preserves the fastest muscle fibre types, and appears to protect the coordination that expresses strength.
The trial randomised 158 resistance-trained men across twelve bench press programmes, crossing three loads with four stopping rules over eight weeks. Load mattered most. Training at 70 to 85 percent of maximum produced the largest gains in chest cross-sectional area, in one-repetition maximum, in bar speed against submaximal loads, and in repetitions to failure. Underneath that, the stopping rule split the outcomes: allowing a 50 percent velocity loss produced the greatest muscle growth and rep endurance, while stopping at 25 percent produced the greatest maximal strength.
The population is trained young men and the exercise is bench press alone, so nothing here transfers automatically to squats, to beginners, to women, or to older lifters. Twelve groups from 158 people leaves roughly thirteen per condition, which is thin for detecting anything but the largest differences and means the ranking within the grid should be held loosely. Eight weeks is short enough that early neural adaptation dominates and structural change is only beginning. The published summary reports directions rather than magnitudes, so how much better 25 percent was for strength, or 50 percent for size, is not stated. And the two variables were analysed as separate main effects, which means the correct reading is that heavy load and stopping rule each acted on their own, not that one particular cell of the grid was crowned.
This is not the first time the split has appeared, and the earlier version came from the same group. In 2017 they compared 20 against 40 percent velocity loss in the squat across eight weeks in 22 men. The lighter stopping rule matched the heavier one for squat strength and beat it for jump height, 9.5 percent against 3.5, while performing 40 percent fewer total repetitions. The heavier stopping rule produced more growth in two of the quadriceps muscles, but also shifted fibre composition away from the fastest type. Different exercise, seven times the sample, same pattern.
A broader analysis of proximity to failure across the wider literature found the dose-response models fit only modestly, and confidence intervals for strength spanned enough ground to leave the shape of the relationship open.
Load is the decision that matters most, and this trial says heavy served both goals. The stopping rule is the second decision, and it is a genuine trade rather than a matter of effort or discipline. Stopping while the bar is still moving well is not leaving results behind; it is choosing strength over size. Someone chasing a bigger chest should sit in the long, slow sets. Someone chasing a bigger bench should rack it earlier than instinct suggests, and take the repetitions saved into another set.
Pareja-Blanco et al., Med Sci Sports Exerc 2026;58(8):1773-1781 · PMID 41872733
Pareja-Blanco et al., Scand J Med Sci Sports 2017;27(7):724-735 · PMID 27038416
Schoenfeld et al., J Strength Cond Res 2017;31(12):3508-3523 · PMID 28834797
Robinson et al., Sports Med 2024;54(9):2209-2231 · PMID 38970765
Tundidor-Duque et al., Scand J Med Sci Sports 2026;36(3):e70265 · PMID 41855389