Loading fundamentals 06: Exercise selection

Short answer. Muscles grow unevenly, and the exercise you choose changes where within a muscle the growth lands. Use a small number of exercises per muscle, varied deliberately across positions and held long enough to progress on. The same finding means the single-site measurements most trials in this branch rely on are less precise than their decimal places suggest.

---

What it is

The default answer is that exercise selection is preference and equipment: a muscle either gets stimulated or it does not, and anything that loads it will do.

That answer assumes a muscle grows as a single unit. It does not, and the consequence runs further than program design. It reaches back into how you read every effect size in this branch.

The proposed mechanism

Muscles are not uniformly activated. Activation varies by region within a single muscle, and it varies with joint angle, with the direction of the resistance, and with which head of a multi-headed muscle has the mechanical advantage in a given position.

Recruitment is not purely by size. The size principle says motor units are recruited smallest first as force demand rises. Layered on top of that is neuromechanical matching: units are preferentially recruited when they are mechanically better placed to contribute to the specific movement being performed. Two exercises that both load the quadriceps do not necessarily call on the same units, because the units best positioned to help differ between a squat and a leg extension.

Local strain differs across a muscle. Pennation angle, tendon architecture and the distribution of force along the muscle belly mean that a given exercise does not stretch or load every region equally.

Each of these predicts the same thing: where growth happens should depend on what you did to cause it.

What has actually been tested

A controlled trial with volume and intensity equated. Participants were randomly assigned to perform either Smith machine squats or leg extensions, and growth was measured at proximal, central and distal regions of the rectus femoris and vastus lateralis. All three regions of the rectus femoris grew significantly in the leg extension group. In the squat group, only the central region of the vastus lateralis grew significantly (Zabaleta-Korta et al., 2021).

Same muscle group, same volume, same intensity, different distribution of growth.

Variation helps, up to a point. A systematic review of eight studies covering 241 young men found that varying exercise selection can influence hypertrophy and strength. Some systematic variation appeared to enhance regional adaptations and maximize dynamic strength, while excessive and random variation may compromise gains (Kassiano et al., 2022).

That is an unusually specific finding for this literature. It says the answer is neither "one exercise per muscle forever" nor "rotate constantly."

Muscle length has a regional effect, and a modest one. A meta-analysis examining whether training at different muscle lengths produces different regional growth reported a small overall effect: a standardized mean difference (SMD) of 0.10, a gap between groups of a tenth of a standard deviation. The probability of a meaningful positive effect rose across the muscle from 50.7% at proximal sites to 69.8% mid-belly to 80.8% distal (Varovic et al., 2024). That analysis was a preprint at the time we read it.

And there is a measurement problem that follows from all of the above. If adaptation is uneven across a muscle, a study that measures one site is measuring that site. The same research group has argued exactly this: measurements of muscle size and architecture taken at a single site can misrepresent what happened to the muscle as a whole (Nunes et al., 2024). Adaptations are frequently uneven across regions, and even between synergist muscles in the same group.

Our reading

Exercise selection determines distribution, not just amount. The controlled evidence is limited but consistent with the mechanism, and the mechanism is well specified enough to trust more than usual. If you want a whole muscle to develop, particularly a multi-headed one, no single exercise is likely to do it.

The practical rule is a small number of exercises per muscle, held long enough to progress on. Enough variation to cover the regions and heads, not so much that you never accumulate progression on any single movement. The systematic review's distinction between systematic and random variation is the useful part: rotating deliberately across positions is not the same as changing exercises because you are bored.

Now the part that matters more than the training advice.

This finding retroactively qualifies a large share of the evidence in this branch, including the evidence we have cited (Nunes et al., 2024). Almost every hypertrophy trial we have quoted measured muscle thickness or cross-sectional area at one or two sites. If growth is regional, those numbers describe the sites measured. An intervention that appeared to produce no difference may have moved growth from one region to another. An intervention that appeared to work may have worked at the site someone happened to scan.

We do not think this overturns the conclusions elsewhere in this branch. The effects we have described are mostly about how much work gets done rather than where, and those should be less site-sensitive. But it does mean the precision of every effect size in this branch is worse than the decimal places imply, and it is a reason to treat small differences between interventions with more suspicion than the confidence intervals alone would justify.

This is one of the reasons OMNID programs keep a stable core of exercises for each muscle through a block, with the variation chosen for the positions it loads: not a single fixed lift, and not a constantly rotating list. The point of the variation is coverage. The point of the stability is that you cannot progress on a movement you keep replacing. (→ Loading fundamentals sub-hub)

Where the goals diverge

Building a balanced physique. This is the goal the finding was made for. Multi-headed muscles need exercises that differ in the position they load, not just in name.

Building maximal strength in a lift. Specificity dominates. Variation is a cost here rather than a benefit, because progression on the tested movement is the objective. (→ Loading fundamentals 05: Progressive overload)

Limited time or equipment. Compound movements cover more muscles per unit of time, and the regional finding does not change that. The other economies available to you are covered separately. (→ Loading fundamentals 01: Rest between sets; Loading fundamentals 03: Training frequency) It means accepting less complete development within each muscle, which for most people is a reasonable trade.

Addressing a visibly lagging region. Adding volume to a lagging muscle runs into the same per-session limits as adding it anywhere else. (→ Loading fundamentals 03: Training frequency) This is where the evidence is most directly useful and least precise. It supports choosing an exercise that loads that region well; it does not tell you which exercise that is for your anatomy.

The case against

The controlled evidence is thin. One randomized trial comparing two exercises (Zabaleta-Korta et al., 2021), plus a systematic review of eight studies in 241 young men (Kassiano et al., 2022). That is a small base for a claim this widely repeated.

Three of four sources share an author network, including the measurement critique we lean on hardest. The controlled trial is the one independent source here.

Regional measurement has its own error. Ultrasound at defined sites is reproducible but not precise enough to make fine distinctions between adjacent regions, and studies differ in where they scan.

The muscle length regional analysis is small and was a preprint. An SMD of 0.10 with a 50.7% probability of meaningful benefit at proximal sites (Varovic et al., 2024) is close to nothing at that end of the muscle.

Our measurement caveat cuts both ways. If single-site measurement is unreliable, it is unreliable for the regional studies too, which are also using it.

What is not known

Which exercise best loads which region, for any given muscle, with enough confidence to prescribe. The principle is better established than any specific application of it.

How much variation is enough. The systematic review distinguishes systematic from random variation without quantifying either.

Whether regional differences accumulate into visible differences over years, or whether they wash out.