Individual variation 01: Are you a non-responder?

Short answer. True non-response to resistance training appears to be rare. When a study separated real individual differences from measurement noise, very few people failed to respond, although real differences in how much people respond remained. A disappointing twelve weeks is weak evidence about you: extend the observation window and judge trends over months.

---

What it is

The observation that people given the same training program finish with wildly different results, and the question that follows: is some of that you, and can you find out?

Every other entry in this publication reports group averages and then warns you that a group average is not a person. This entry is about what that warning means, and it is the one place where we can partly cash it in.

The proposed mechanism

Genetic and biological variation. People differ in satellite cell content, myonuclear addition capacity, hormone profiles, fiber type distribution and gene expression responses. Any of these could plausibly produce different adaptation from identical training.

Everything else. Sleep, nutrition, stress, prior training history, recovery capacity, technique, and how faithfully the sets were performed.

And a third category that is not a mechanism at all. Measurement error. Muscle size measured twice on the same unchanged person will not give the same number twice, and neither will a one-repetition maximum. Some of the difference between two people's results is the instrument rather than the people.

What has actually been tested

The famous ranges are real. Five hundred and eighty-five participants, 342 women and 243 men across eight study centers, trained the non-dominant elbow flexors for twelve weeks with cross-sectional area measured by magnetic resonance imaging (MRI). Changes in muscle size and in one-repetition maximum spanned enormous ranges, with strength changes reported from 0% to 250% (Hubal et al., 2005). Similar spreads appear across the literature.

Those ranges do not measure what they appear to. Any observed change in an individual is the sum of three things: the true effect of the intervention, within-person biological variability, and measurement error. A range of individual change scores contains all three, and separating them requires a design built to do it (Atkinson and Batterham, 2015).

The standard method needs a control group. The standard deviation of individual responses is calculated by taking the variance of change scores in the training group and subtracting the variance in a non-exercising control group. A positive result indicates real differences in trainability. A result of zero or below indicates the observed variability is attributable to measurement error and biological noise rather than to true individual differences.

Almost nobody has done this. In resistance exercise specifically, one recent analysis noted that only four studies had statistically evaluated response variability in this way, and of those, one concerned back pain and two used light-load training in clinical populations (Soendenbroe et al., 2026).

When it was done properly, true non-response was rare. In fifty-eight healthy men averaging 72 years, across sixteen weeks of heavy resistance training against a sedentary control arm, individual changes were classified against measurement error (Soendenbroe et al., 2026). 82% were classified as robust or excellent responders and only 5% as poor responders. Substantial real inter-individual variability was present, so this is not a finding that everyone responds identically. It is a finding that almost nobody fails to respond.

And the obvious explanation did not work. In that same analysis, training compliance and one-repetition maximum progression did not explain the variability. Lower baseline levels were associated with greater improvement without fully accounting for the differences.

A better design exists and is only now being used. Repeating the whole training period in the same people, and asking whether they respond the same way the second time, tests reproducibility of individual response directly (Räntilä et al., 2025). A within-subject design training one limb and using the other as an internal control has also been proposed specifically to isolate true individual response from biological variation (Chaves et al., 2025).

Our reading

True non-response to resistance training appears to be rare, and you are far more likely to be measuring badly than to be one.

The 5% figure (Soendenbroe et al., 2026) comes from one study in one population, so we would not carry the number itself very far. What we would carry is the shape: when researchers built a design capable of separating real individual differences from noise, the population of true non-responders was small.

Real individual differences do exist. They are just much smaller than the raw ranges imply. The spread reported by Hubal and colleagues is not evidence that some people cannot build strength. It is evidence that individual change scores are a noisy measurement, and the noise is large relative to the signal over twelve weeks.

The compliance finding is the one that should change your thinking. The easy explanation for someone progressing slowly is that they are not training hard enough, and in the study that looked, compliance did not explain the variation. That cuts both ways: it removes the comfortable dismissal, and it removes the comfortable self-blame.

So the practical position is that a disappointing twelve weeks is weak evidence about you. It is more likely to reflect measurement, timing, and the ordinary variation in a short window than a fixed property of your physiology. The response is to extend the observation window, not to conclude something about yourself, and to trust a trend across months above a snapshot at week twelve. (→ Loading fundamentals 05: Progressive overload; Adherence 01: The first month)

And this is where we cash in the warning attached to every other entry here. When we say an effect size of 0.2 is a group average and not a prediction about you, the fair extension is that your own result is also not a clean reading of your physiology. Both numbers are noisier than they look. (→ Loading fundamentals sub-hub)

Where the goals diverge

Early in training and progressing slowly. The measurement argument applies most strongly here, because twelve weeks is short and early changes include fluid shifts that are not growth. (→ Inflammation 02: Delayed-onset muscle soreness)

Years in and stalled. Different problem. Rate of adaptation slows with training age, and a stall after five years is expected rather than diagnostic.

Considering genetic testing. Nothing in this literature supports commercially available tests as predictive of training response. The studies that isolated true individual response were not able to explain it with the variables they measured.

Deciding whether to change programs. The measurement argument is also an argument against frequent switching. If twelve weeks cannot reliably tell you whether you responded, it cannot reliably tell you whether a program change worked either. (→ Loading fundamentals 06: Exercise selection)

The case against

Our headline figure comes from older men. Sixteen weeks, one population, one laboratory. Younger trained lifters may distribute differently, and we have extended a shape rather than a number.

Substantial real variability was still present. We have emphasized the rarity of non-response and should equally note that the same analysis reported strong evidence of real inter-individual variation. Both are true and the second is easier to lose.

Four studies is a thin base for a claim this structural. Most of the responder literature has not used designs capable of supporting its own conclusions, which is a criticism of the field and also a limit on what we can say from it.

The repeated-training and within-limb designs are recent. Their results may revise this picture, and if true individual response turns out larger than current estimates, our framing will need changing rather than qualifying.

And absence of an explanation is not absence of a cause. Compliance and progression did not explain the variability. Something did.

What is not known

Whether the rarity of true non-response holds in young trained lifters.

What drives the real individual variation that remains after noise is removed. It has been isolated and not explained.

How long an observation window needs to be before an individual result becomes informative about that individual.

Whether individual response is reproducible at all, which is the question the repeated-training design was built to answer and which does not yet have a settled answer.