How to build a bigger back

Why a back plan needs two kinds of pull

The latissimus dorsi is the largest contributor to the vertical pull. The mid-back muscles between the shoulder blades contribute more heavily to the horizontal pull, and the spinal erectors work in both. , 2022).

That comes to roughly 8 to 11 hard sets, which is a reasonable place to start and can grow towards the top of the reviewed range as you get used to the work. , 2026).

How heavy should back work be?

, 2021). , 2022).

How to make back work actually progress

Back movements are the easiest place in the gym to hide a lack of progress, because so much of the work can be shifted into a swing of the torso or a shrug of the shoulders. The fix is unglamorous: fix the movement, record it, and beat the record.

Splitting back work across two sessions a week is the practical way to fit the sets in. A frequency meta-analysis found greater hypertrophy at twice a week than once at matched volume, and a later review found little frequency effect once weekly sets were equal (Schoenfeld, Ogborn and Krieger, 2016; Schoenfeld, Grgic and Krieger, 2019).

Back training claims the reviews do not support

• Wide grip for width, close grip for thickness. No review establishes that grip width changes which part of the back grows.
• Underhand pull-downs build the lower lats. This comes from activation readings, not growth measurements.
• You must deadlift to build a back. Nothing in the reviewed evidence makes any single exercise necessary.
• Training in the fully stretched position grows the lats more. A Bayesian meta-analysis of 12 studies found only trivial regional differences between longer and shorter working lengths (Varovic et al., 2025).

What does make a difference is much less interesting: enough hard sets, enough effort, load going up, and enough protein and food behind it. , 2018).

The shoulders do a fair amount of work in every row and pull-down, which makes them the natural next thing to get right.

What the reviews behind this page actually found

Every practical instruction above rests on reviews that pool many training studies rather than on any single trial. Here is each finding on its own, so you can see exactly what is and is not established.

• Muscle size increased as weekly hard sets rose, across a meta-analysis of 15 studies (Schoenfeld, Ogborn and Krieger, 2017).
• Dose-response modelling of weekly volume and frequency describes the same curve flattening as volume climbs (Pelland et al., 2026).
• Counting hard sets is a workable way of quantifying training volume, though no optimal number is established (Baz-Valle, Fontes-Villalba and Santos-Concejero, 2021).
• Low and high loads produced similar muscle growth, with heavier loads better for maximal strength (Schoenfeld et al., 2017).
• Growth occurred across a wide repetition range when sets were taken close to failure (Schoenfeld et al., 2021).
• Stopping a repetition or three short of failure produced similar growth to training to failure (Refalo et al., 2023).
• Training to failure was not required for muscle growth in a review of failure versus non-failure training (Grgic et al., 2022).
• Training a muscle twice a week compared favourably with once a week in an earlier frequency meta-analysis (Schoenfeld, Ogborn and Krieger, 2016).
• Once weekly sets were matched, frequency itself made little difference (Schoenfeld, Grgic and Krieger, 2019).
• Machines and free weights produced comparable muscle growth across 13 studies (Haugen et al., 2023).
• A second, independent meta-analysis also found no hypertrophy difference between machines and free weights, with strength gains specific to the test used (Heidel, Novak and Dankel, 2022).
• Systematic variation in exercise selection may help adaptations, while excessive or random rotation may hinder them (Kassiano et al., 2022).
• Training at longer versus shorter mean muscle lengths produced trivial differences in regional growth (Varovic et al., 2025).
• Protein above roughly 1.6 g per kg of bodyweight a day added no further benefit in a meta-analysis of 49 studies (Morton et al., 2018).