Neither, on current evidence: in a 2023 meta-analysis of 13 studies and 1,016 adults, no difference in muscle growth was detected between machines and free weights3. Where they differ is strength, and only in a predictable way: people got strongest at the kind of lift they trained.
That makes the gym argument less about which tool builds muscle and more about which lifts you want to be good at. Here is what was compared, what was found, and where the evidence is still thin.
What was compared
First, the words. Free weights are dumbbells and barbells: you move them along a path you control and balance yourself. Machines, in this research, are devices with a pin-loaded plate stack that guide the load along a fixed path3. Cable stations without a fixed path were left out of the 2023 review3.
A meta-analysis is a study of studies: it pools the results of many trials into one estimate. The 2023 review by Haugen and colleagues only included trials that put one group on free weights and another on machines, for at least 6 weeks, in adults aged 18 to 60 without chronic disease3.
| What | Number |
|---|---|
| Studies | 13 |
| Adults | 1,016 (789 men, 219 women) |
| Training status | 6 studies of trained lifters, 7 of untrained |
| Length | 6 to 12 weeks, 2 to 4 sessions a week |
| Studies that measured muscle size | 5 |
| Competitive athletes | none |
Typical match-ups were the barbell back squat against the leg press or a Smith machine squat, and the barbell bench press against a machine chest press3.
Results are reported as an effect size (a standardised mean difference): how big a gap between groups is, compared with how much people normally differ. Statistically significant means the gap is too large to be likely down to chance.
In plain terms: this is the most complete head-to-head summary available, built from short trials in ordinary adults.
What the studies found
No clear difference in muscle size. Across the studies that measured it, the difference between groups was an effect size of −0.06 (95% confidence interval −0.40 to 0.29, p = 0.75)3. A confidence interval is the range the true value most likely sits in; this one sits across zero. Both groups grew3.
Strength was specific to the tool. Tested on a free-weight lift, the free-weight groups gained more (effect size 0.21, 95% CI 0.03 to 0.39, p = 0.023)3. Tested on a machine, the machine groups tended to gain more (0.29, CI −0.02 to 0.60, p = 0.064), a trend that could still be chance3. Sports scientists call this specificity: you get strongest at the exact movement you practise.
Each on its own lift, the gains matched. When each group was tested on the lift it trained, free weights on free weights and machines on machines, there was no difference (0.08, CI −0.11 to 0.27, p = 0.39)3. A neutral test of strength (an isometric, or static, push) also showed no difference (−0.08, p = 0.66)3.
Machine vs free-weight training: pooled effect sizes · SMD
An earlier meta-analysis by Heidel and colleagues found the same shape2. Free-weight training gained more strength on free-weight tests (effect size 0.66), machine training gained more on machine tests (0.78), a neutral device showed similar gains (0.13), and muscle growth did not differ (−0.01)2.
In plain terms: both tools grew muscle about equally in these trials, and each made people better at its own version of the lift.
Two trials worth a closer look
The pooled numbers hide individual trials that are easy to picture.
In an 8-week trial, 46 young adults (26 of them women) trained each muscle 2 to 3 times a week with either free weights or machines1. Biceps and quadriceps muscle thickness increased in both groups, with no difference between them1. Machine bench press strength rose more in the machine group (13.9% against 8.6%), while free-weight bench press, squat and Smith machine squat rose 11% to 19% in both groups alike1.
In another 8-week trial, 38 resistance-trained men did the same program with either barbells or machines: full squat, bench press, prone bench pull and shoulder press4. Training variables were matched, so the only difference was the tool. Each group gained most on the modality it trained, but across all eight strength tests the changes were similar (differences of 1.8% or less)4. Ultrasound-measured muscle area rose in every muscle checked, again with no difference between groups (2.0% or less)4.
What this does and does not mean
It does not mean the two are identical. A barbell squat asks you to balance and control the load; a leg press does not. The 2023 authors point out that the squat moves more like a jump than a leg press does, and they tie the strength results to the SAID principle ("specific adaptations to imposed demands")3.
It also does not prove free weights build more "total" muscle through stabilising work. The authors raise that idea but label it speculation: extra muscle activity in a single session has an uncertain link to long-term growth3. They also speculate that a mix of both might work best, which no included trial tested3.
One sub-result cuts the other way. For upper-body strength, the machine groups gained more (effect size 0.25, p = 0.003); for lower-body strength there was no difference3. That rests on a handful of studies, so read it as a hint, not a rule.
In plain terms: the muscle result is "no detectable difference", the strength result is "you get good at what you practise", and the rest is open.
What people who trained with each actually did
These trials describe, rather than prescribe. The groups trained 2 to 4 times a week for 6 to 12 weeks, usually with matched sets and reps, and swapped only the tool3. The authors' own conclusion is that the choice comes down to "individual preferences and goals"3. The earlier review adds that people after general strength and muscle "may choose whichever activity they prefer and are more likely to adhere to"2.
Two practical readings follow from that, as description:
- If a goal is a specific lift, the trials that measured it found the biggest gains came from training that lift. Someone who wants a heavier squat is tested on a squat.
- If the goal is muscle, the trials found no edge either way, so the deciding factors become the ones they held constant: sets, effort and consistency. See how many sets per muscle per week and what progressive overload means.
The studies at a glance
| Study | Who was studied | Key result | Year |
|---|---|---|---|
| Schwanbeck et al., *J Strength Cond Res*1 | 46 young adults (26 women), 8 weeks | Biceps and quadriceps thickness rose with no group difference; machine bench press rose more with machines (13.9% vs 8.6%) | 2020 |
| Heidel et al., *J Sports Med Phys Fitness*2 | Meta-analysis of trials to September 2020 | Strength favoured the trained modality on its own test; muscle growth no different (effect size −0.01) | 2022 |
| Haugen et al., *BMC Sports Sci Med Rehabil*3 | Meta-analysis, 13 studies, 1,016 adults aged 18 to 60 | Muscle growth no different (−0.06, p = 0.75); strength specific to the tool; head to head no different | 2023 |
| Hernández-Belmonte et al., *Med Sci Sports Exerc*4 | 38 trained men, 8 weeks, matched programs | Strength changes similar across 8 tests (≤1.8% apart); muscle area rose similarly (≤2.0% apart) | 2023 |
| Amanuma et al., *J Bodyw Mov Ther*5 | 8 untrained women, one leg each way, 9 weeks | Thigh muscle thickness rose with both lunges and leg press; no group-by-time difference | 2025 |
Limits worth knowing
- Few studies measured muscle. Only 5 of the 13 in the 2023 review assessed muscle size3.
- Rough measures. Size was measured by ultrasound, circumferences, skinfolds or air displacement, methods the authors say "vary greatly" in sensitivity; they ask readers to be "careful with the interpretation"3.
- Short trials. Six to twelve weeks. Nothing here covers years of training.
- Not everything matched. The review could not adjust for differences in range of motion, intensity, volume or frequency between studies3.
- No competitive athletes. None of the included studies examined them3.
- One result leans on one trial. Removing a single large study made the free-weight-test advantage non-significant (p = 0.153)3.
- Abstract only for the 2022 review, the 2020, 2023 and 2025 trials; we cite only numbers stated in those abstracts. The 2023 review was read in full text.
Latest research
The newest trial we found is a 2025 study in 8 untrained young women, in which each woman trained one leg with lunges and the other on an inclined leg press, 3 days a week for 9 weeks5. Ultrasound thickness of two thigh muscles, at two points along each, rose on both sides: for example, +13.3% with free weights and +12.1% with the machine at mid-thigh on the outer quadriceps5. No region showed a difference between the legs5.
Thigh muscle thickness change after 9 weeks, by region · %
Free weights (lunge) Machine (leg press)
A within-subject design (each person is her own control) removes differences between people, which helps a small study. But eight people is still eight people, and the percentages are group means. It fits the pooled picture; it does not settle it.
The short version
- In pooled trials, machines and free weights built about the same amount of muscle.
- Strength was specific: each group got best at the kind of lift it trained.
- The muscle evidence is thin: few studies, short durations, rough measures.
- The authors put the choice down to goals and preference, and the tool you keep using is the one that gets trained.
Educational overview only — not medical advice.
