I was initially going to make this piece one of those ones about the danger of cherry-picking evidence. It’s quite easy to find research supporting any given view in health/fitness. That’s not actually a problem with science, per-se. It’s expected that just by chance you’ll have the odd result emerge that runs contrary to the expected, and in the case of health sciences especially where differences in protocols, sample size and measurement techniques can influence results meaningfully, it’s a near certainty. Well-meaning individuals without a grasp of the totality of the evidence can be easily misled by believing the first piece of evidence that they come across, and less well-meaning individuals can easily cherry-pick evidence to support their positions and sell you stuff.
Instead I decided to just make it a comparison of two papers that I came across which, per the titles and abstracts, appear diametrically opposed, but upon reading aren’t too far apart. They deal with training frequency and its effects on hypertrophy, and given that that’s a sort of hot topic I figured it’d be worth looking at. I suppose it is an illustration that, even where at surface level, papers seem to entirely contradict each other, it’s possible to reconcile differences by reading them in depth as opposed to rejecting either/both.
That brings me to the two papers I’m going to talk about today. They were both published in the Journal of Strength and Conditioning Research (JSCR), and I found them because they were side by side in my suggestions.
The first is called “High-Frequency Resistance Training Is Not More Effective Than Low-Frequency Resistance Training in Increasing Muscle Mass and Strength in Well-Trained Men” by Gomes et al (2018).
The second is called “High Resistance-Training Frequency Enhances Muscle Thickness in Resistance-Trained Men” by Zaroni et al (2019 – available ahead of print 2018).
Other than the hilariously opposite titles, they share enormous similarities in their methodology, which I’ve summarised below.
1 – Participants – 23 males aged 26.2yrs +/-4.2 (Gomes) and 18 males aged 26.4yrs +/-4.6 (Zaroni).
2 – Resistance training experience – 6.9yrs +/- 3.1 (Gomes) and 6.4yrs +/- 2.4 (Zaroni).
3 – Training protocol comparison – comparing split workouts of 10-15 sets/bodypart performed once a week with a volume-matched equivalent full-body routine performed 5x/wk. In both instances sets were performed to failure.
In both instances squat and bench press 1RM were measured to assess strength (the Zaroni paper also measured row 1RM) and both took measures of muscle mass. The Gomes study used DEXA to report total and segmental fat-free mass, whilst the Zaroni study used ultrasound.
In both instances diet was measured across the study, with no differences between groups.
Here’s a look at the respective training protocols
Whilst in concept things are a little different, they are reasonably close replications. A point that I’ll return to is that the Zaroni paper actually compares 2x/wk frequency for upper body to 5x/wk frequency, as unlike in the Gomes paper where biceps and triceps training was paired with upper body pushing/pulling, there were separate arms days in this protocol. I’ll return to that point, because it either does or doesn’t do much to explain the results.
As for the results, the titles give it away. In the Gomes paper, no statistically significant differences were found for muscle and strength increases across the board. In the case of the Zaroni paper, of the three sites where muscle thickness was measured, two (elbow flexors and vastus lateralus) experienced significantly greater growth in the high frequency condition. No significant differences were found for strength gains between either condition.
There were also measures of training volume/load and DOMS, which I will cover later in the article.
Heres a look at their respective results tables
If you’ve had a look at them, you probably noticed that in the majority of measures the changes in the absolute values favoured the higher frequency condition. However, the vast majority didn’t reach statistical significance. This doesn’t surprise me enormously, given the small number of subjects, high level of interindividual variability, and comparatively low degree of change in each measure. These factors make statistically establishing differences between the groups very difficult.
However, I do think that there is value in observing the trends within the data, provided that any conclusions drawn from it are sufficiently tentative, and so I’ve collated the results of the most comparable measures in one table and included my interpretation and notes for each. I used total FFM from the Gomes paper and didn’t include all of the segmental data as it is less noisy – it should capture any changes in size across the body without introducing error by mismeasurement of the segments themselves, etc.
The first thing to note is that whilst only the muscle thickness data from the Zaroni study strongly favours higher frequency training, none of the measures favour the lower frequency condition. It seems to me that there is either no difference, or high frequency is marginally better. In the case of strength measures, the absolute changes in the squat outcomes and bench in the Gomes study are enough to make me tentatively think that high frequency might be better, but in the absence of statistically significant support I don’t hold that opinion strongly.
In my “notes” I mentioned which of the two groups was “weaker” to begin with based off of the averages supplied. There were not actually statistically significant differences in strength at baseline, so I’d take that with a grain of salt. However, the inability to statistically differentiate the groups with certainty is not the same as there being no difference (hence my making of inferences on the data in the first place), and if there were a difference in the strength of the groups at baseline that might also provide an alternative explanation for differences in gains – you’d generally expect smaller/weaker trainees in an otherwise homogenous group to gain faster, but if the differences in baseline characteristics are trivial, so would the expected effect of said differences be on their rate of progress. Still, in all 3 cases where the strength gains appeared greater in higher frequency training, the higher frequency group may also have been weaker to begin with, and so it’s worth noting. You’ll also note that I shaded them light blue – I said I’d tentatively conclude that they may have benefited from the higher frequency condition, meaning with a considerable degree less certainty than the thicknesses in the Zaroni study. Part of why I’m inclined to think there might be a difference is that a number of strength measures appeared to benefit, and that the two that didn’t may be partly attributable to the training protocol itself (I discuss that later).
Likewise, I mention that the confidence intervals overlap but that the range is higher for the three measures that I said tentatively favour the high frequency condition – statistically that’s not really appropriate. That the confidence intervals overlap means that it’s within the accepted bounds of probability that the true value for the two groups is the same (I explain confidence intervals below), so again, consider that with a grain of salt.
Looking at the studies side by side, I’d be inclined to infer that the higher frequency condition did impart some advantage to size and strength, although I wouldn’t hold that position very strongly.
Illustrating that nicely is the graph below from the Zaroni study (Fig 7), which shows standardised differences between the Total and Split groups. Dots to the left of the vertical dashed line support the higher frequency condition, and dots to the right support the lower frequency condition. The lines either side of the dots (error bars) show a 90% confidence interval. Because the dot is an “estimation” of effect, the error bars show the range of estimated effects that the data suggests, with 90% certainty, the “true” value lies.
To make that clearer, imagine measuring the height of 100 men aged 18-22 from a given city. Your average height within your sample won’t necessarily be the same as the average height of the entire city’s population, but might cause you to say “the average height of the city at large is likely to sit in this range (confidence interval) and our best estimate of it is x (dot),” – that’s more or less what this shows.
The grey zone around the dashed line indicates the smallest worthwhile change, and so error bars that lie in that zone would indicate a difference that is practically unimportant.
As you can see, all of the dots favour higher frequency, however a number of confidence intervals reach into the smallest worthwhile change, or cross the dashed line, meaning that it is unclear that there is a difference between groups, or even which condition is favoured.
So to wrap up the results – how different are they between the two studies? In my view, they aren’t as different as the titles would suggest. The majority of effects favour higher frequency in both studies, but the degree to which they do so is in many cases limited.
Hypotheses and interpretations
My analysis of the results above (beyond just reporting the findings in absolute terms) is already hypothetical. I’ve said that I think overall both studies favour the higher frequency condition, and that in some respects the large amount of individual difference and small-ish magnitude of gains (and differences in gains) makes establishing that statistically difficult.
That question can be answered somewhat by the most recent meta analysis of the role of frequency in hypertrophy, which found that when volume was equated frequency did not appear to effect gains. Where volume was not equated, higher frequency does improve gains, suggesting that increasing frequency is most useful as a means for increasing effective training volume.
A hypothesis that appeals to me is that of a maximum productive volume per session. I’ve written about it before here, and Menno Henselmans more recently made a similar point, suggesting 9-13ish sets near failure per session (I believe James Krieger has suggested 8-10). If returns on per-session volume increase to that point, then either fail to increase or diminish, it may also explain the lack of pronounced observed advantage to higher frequencies in the current studies, as both conditions fall some way short of the peak of the inverted U of optimal per-session volume. The Gomes study mentions prior research suggesting that as few as 3 sets per session can maximise the MPS response in their introduction, but other research shows dose-dependant increases in MPS to 6+ sets.
That said, were benefits to higher frequencies only particularly notable in higher volume conditions, then per my logic above the triceps would have appeared to benefit more from increased frequency in the Zaroni study, and this didn’t seem to be the case.
In a prior hypertrophy article I mentioned how neuromuscular fatigue can limit per-session productive work and cited evidence, including the Wernbohm review on training variables and hypertrophy, suggesting that moderate per-session volumes likely yield the best returns. I suspect that moving from low frequencies to moderate frequencies, where per-session volume is not diminished too much, yields more benefit than moving to very high frequencies. Consider this paper in which doing 16 sets in 1 session produced less gains than doing 8 sets twice/wk. I also mentioned how the fatigue from doing moderate volumes (approx 8 sets) can persist for 3-4 days, making performing additional work too frequently difficult.
My overall impressions are this – the main benefits to increasing frequency are the ability to accrue more effective volume across the week. Where volumes are sufficiently high, distributing work over 2 or more sessions is likely helpful, but the recovery demands of doing volumes sufficient to maximise per-session returns probably limit the number of maximally productive sessions.
Further research (including meta analyses) in which the effects of manipulating frequency are stratified depending on per-session and per-week volumes could shed some light on this.
Some curious notes – speaking of the Zaroni study particularly, if frequency did impart any particular benefits I’d expect there to be differential effects between the muscles. The legs are trained once/wk in the split routine and 5x/wk in the total routine, and so you might expect greater benefits to be seen in the squat/vastus lateralus measurements. To a degree, this was played out – the squat had the most convincing differences in strength between the two conditions (although I still only tentatively considered it “better”) and the VL measurements were one of the two muscles to see statistically significant improvements in gains between conditions.
By the same token I’d expect triceps/bench press to benefit the least from increasing frequency – they were already hit twice a week (once in the “chest” day and once in the “triceps” day). Whilst the same can be said for biceps (or elbow flexors per the muscle thickness results), anecdotally back training is less biceps intensive than pressing training is triceps intensive. Again, that did play out, with there being no differences in bench press strength increases between conditions and the triceps being the only muscle to not statistically significantly benefit from increased frequency, although the effect size estimates do seem to favour total.
The same re differences in frequency can’t be said for the Gomes study, which consolidated all bicep/tricep training to the same days as chest/back in the low frequency condition.
When looking at my collated data from the studies, the two strength measures that didn’t appear at all different between conditions were also the bench press and row in the Zaroni study. In the 3 strength measures that I tentatively said may benefit from the higher frequency condition, exposure to relevant training increased from 1-5x. What’s more, in the Gomes study, exposure to the exact movement increased from 1-5x.
In the case of the Zaroni paper, the increase in frequency of relevant training was lesser, and because of the greater variety of movements used (“chest” training consisted of a variety of presses, rather than just bench) exposure to the exact movements were also not increased as much. This may explain why there appeared to be no difference between the high and low frequency protocols in those two measures.
What can we certainly conclude
The most conservative interpretation I can make would be that distributing a moderate weekly volume across a number of smaller sessions likely has no downsides, even when per-session volumes are significantly reduced, when sets are taken to failure. It is possible that with training that is sufficiently submaximal, higher per-session volumes would be necessary to maximise stimulus (as the effectiveness of per-set work is reduced – see this article for a bit more on that).
More broadly, a number of viable options for distributing volume exist and provided that the same total volumes can be achieved, large differences in training effect likely won’t be seen for different frequencies.
From the present studies there are compelling practical reasons to perform higher frequencies than 1/wk.
In the Gomes paper DOMS was measured using a visual-numeric scale. The low frequency condition reported significantly greater DOMS at the beginning, middle and end of the study. That this difference was present at all time points suggests that adaptation to the higher per-session volumes was not sufficient to mitigate the extra DOMS that they induced. An alternative hypothesis provided in the paper is that by exposing the trainees to more bouts of exercise the repeated bout effect was increased in the high frequency condition, but as experienced trainees I’d presume that they were already reasonably acclimated to training. Reducing DOMS without reducing gains in size/strength (and possibly promoting them) is a plus.
On a similar note, the Zaroni paper measured internal training load (ITL). The internal training load is calculated by multiplying the time under tension for a whole session by the session RPE (ie “how long were you exposed to the stimulus, and how hard was it”). There was no significant difference in ITL between the groups – this is to be expected as both groups performed sets to muscular failure (ie, the same difficulty per set).
However, in both papers the total volume performed was greater in the high frequency condition, likely as reduced neuromuscular fatigue on a set by set basis meant that per-set outpit was better. This means that a greater amount of volume was performed across the week in the Zaroni paper without the measure of training stress (ITL) being higher.
On total volumes, the HFRT condition in the Gomes paper performed 13.9% more volume load across the study. The accumulated total load lifted was 22.3% higher for the Total group in the Zaroni paper.
The degree to which the additional volume was “effective” might be up for debate, given that there were limited statistically significant differences in outcome measures between the groups. However, the returns on one additional set/wk are estimated to be small – 0.37% per Schoenfield et al’s systematic review.
What the information about DOMS, total volume performed and ITL between the two papers suggest, however, is that increasing frequency is a viable means to improve training output without increasing perception of effort or post-session soreness for a given volume of work.
Whether that necessarily makes training multiple body parts per-session easier than using a split is another question. Anecdotally, many of my clients have found working multiple bodyparts (especially when using large compound movements like the powerlifts) more taxing physically and psychologically than bodybuilder-esque training. It takes more warming up and often more time per session. However, a middle ground exists (such as upper/lower splits) in which it’s possible to work each muscle group multiple times across a week without having to train in a full-body fashion.
Finally, there appeared to be significant heterogeneity in results between the individuals. It’s probable that different individuals benefit more/less from increasing training frequency and volume overall. Consider this study by Hammarstrom et al, in which, in a within-subject design, some individuals massively increased gains by tripling training volume, whilst others yielded no detectable benefit. Research suggests differences in genetic susceptibility to muscle damage, and given that high levels of muscle damage preclude further effective work within a session, interfere with training adaptations, and lengthen recovery time, it is also possible that individuals who are more susceptible to muscle damage will be less resilient to higher per-session and weekly volumes. “Optimal” training configurations likely thus differ from person to person based off of factors such as this, and so there probably isn’t a catch-all method of distributing training volume.
What I might speculative/tentatively conclude
Having acknowledged that the current evidence suggests largely that manipulating frequency is a useful tool for maximising effective volume performed and managing the fatigue and burden of performing it, there are a couple of more speculative interpretations that I might make, although I wouldn’t hold any of them particularly strongly at this point.
Firstly – there may be independent advantages to higher frequencies, especially where weekly volumes are high. The fatigue and performance decrement associated with high per-session volumes and the capped anabolic response might make distributing high volumes into multiple moderate volume sessions preferable. I suspect that moving from low (1) to moderate (2-3) frequencies will help. Moving from moderate to high frequencies may help less, as total weekly volumes rarely warrant it, and returns on per-session volumes don’t diminish that rapidly.
In the present studies there are trends in the results that favour the higher frequency conditions despite them being below what I would consider optimal per-session volumes. Given the absence of any indication that the higher frequency condition was detrimental, if I had to bet which was better I’d say that.
After all that, what would change in my programming for hypertrophy as pertains to frequency? In reality, not that much.
– Neither very high or very low frequencies appear to confer great benefits when volume is equated
– Where manipulating frequency can allow a higher volume of productive work, it may be beneficial to do so
As such, I’ll be continuing to prescribe mostly moderate per-session volumes at frequencies of 2-3x/wk per bodypart for the majority of people. When intra-session fatigue or time availability inhibits people performing more quality work, but recovery between sessions is not an issue, I’d consider increasing frequency provided the big rocks of training and progression remain in check.
Despite the opposite titles, there’s a lot of similarity between the results of these papers. They fall within a body of literature that doesn’t suggest magical benefits to higher frequency, but also doesn’t preclude their being benefits to manipulating it. Increasing training frequency appears to be a viable means to increase the effectiveness of work performed and reduce recovery costs. Where, and whether, there is an optimal distribution of training is yet to be determined, but focusing on distribution over quality and quantity of work is probably missing the forest for the trees.
If you LIKED this article, join my mailing list – it’s FREE and I can send you stuff directly to your inbox. You can also follow me on Instagram, where I frequently post training and diet analysis and advice on my stories.
I also have a podcast, Weakly Weights, available on iTunes and Podbean, where I discuss training for powerlifting in-depth. You can join our mailing list, too, for free sample programs when we discuss them.