Push-ups against the bench press
Can a bodyweight exercise build muscle like a barbell? A 2017 study by Kikuchi and Nakazato tested it on 18 young men. Half did the bench press at 40% of their one-rep max. The other half did push-ups, with each person’s position set so the load matched that 40%.
Both groups trained twice a week for eight weeks. Chest and triceps thickness went up in both groups, and so did the bench press max: from 60.0 to 65.0 kg with the bench press and from 61.1 to 64.2 kg with push-ups. The biceps grew only in the bench-press group.
- Chest muscle thickness grew in both groups
- Biceps grew only in the bench-press group
Make the push-up hard enough
In that study, every set went to failure. The researchers set each person’s push-up to the same load: a regular push-up, a push-up from the knees, or one from the knees with the hands on a raised support.
A 2015 study by Calatayud and colleagues went the other way with 30 students who already lifted. To make push-ups heavy enough for six-rep sets, they added an elastic band. After five weeks, band push-ups and the bench press raised strength by similar amounts, between 13.65% and 22.21% on the tests. The control group stayed the same.
In both studies the sets were hard for the person doing them: to failure in one, at a six-rep max in the other. If a full push-up is out of reach, push-ups for beginners starts from easier versions.
- Knees down, hands raised Kikuchi, 2017
- Knees down Kikuchi, 2017
- Regular Kikuchi, 2017
- Regular, elastic band Calatayud, 2015
What push-up count says about your heart
Yang and colleagues followed 1,104 male firefighters for ten years. At the start, each man did push-ups to a metronome at 80 beats a minute. The test ended at 80 push-ups, after three missed beats, or when he stopped. Over the next ten years the group had 37 cardiovascular events.
Men who managed more than 40 push-ups had 96% fewer events than men who managed 10 or fewer. Drag the slider to see each group’s rate. The numbers don’t go down evenly: 31 to 40 push-ups looks a little worse than 21 to 30, and the wide intervals show how few events each group had.
This is an association. The men averaged about 40 years old and had active jobs, so the result may not hold for women or for less active people. The authors say a push-up count could be a cheap fitness check, and that larger studies with more varied people are still needed. You can run a similar check at home with the push-up test.
- 0 to 10 1
- 11 to 20 0.36
- 21 to 30 0.16
- 31 to 40 0.25
- 41 or more 0.04
21 to 30: 0.16 times the event rate of the 0 to 10 group (95% CI 0.06 to 0.42).
1,104 male firefighters, 10 years, 37 events. A link, not proof of cause. Lines show the 95% interval.
Counting only the reps that do the work
In Pushgard the phone’s front camera counts your push-ups, and a rep counts only when two signals agree: your arms fold, and your shoulders grow in the frame as you come down. If only one of them happens, the rep doesn’t count.
Timing is checked too. A rep faster than 0.55 seconds or slower than 5 seconds, top to top, isn’t counted. How the counter refuses a half rep explains where those numbers came from, and what counts as a full push-up shows the depth from the side.
- Your arms fold
- Your shoulders grow in the frame
How often to do push-ups
The muscle study used two sessions a week for eight weeks: three sets to failure, with two minutes of rest between sets. That’s 16 sessions in total.
The WHO 2020 guidelines ask adults to do muscle-strengthening work for all major muscle groups on 2 or more days a week, at moderate effort or harder. Push-ups train your chest and triceps, so your legs need their own exercise. The guideline team found no evidence that doing more strength work brings more health benefit.
For form, see how to do a push-up. In the Pushgard push-up counter, every push-up the camera counts hits the monster in front of you.
16sessions in the push-up study: 3 sets to failure, 2 minutes of rest
2+days a week of strength work for all major muscle groups, by the WHO guideline
