Is Training To Failure Necessary?

Recent research casts light on an old debate

When Arthur Jones unleashed his training ideas upon the bodybuilding / strength training scene back in the 1970s, he started a firestorm of controversy.  His high intensity training prescription really stirred things up and they haven’t settled down since.

One of the key components of Arthur’s high intensity training was his insistence that in order to make maximum gains the trainee must “train to failure”.  What Arthur meant was that you had to perform an exercise to the point where you could no longer lift the weight in proper form.  For example, if you were performing an exercise with a weight that you could lift a maximum of 10 times, Arthur believed that you would have to lift that weight for all 10 repetitions in order to maximally stimulate growth.  If you ended the set before reaching 10 repetitions you would not grow as much as you would if you had lifted the weight for as many reps as you were capable of.  Strength coach and high intensity advocate Matt Brzycki summed up the argument for training to failure this way in his book A Practical Approach To Strength Training, “Simply, a submaximal effort will yield submaximal results.”(1)

Is training to failure as important as Arthur believed?  Does a submaximal effort really yield submaximal results?  Today, nearly 40 years after Arthur first began preaching training to failure, the debate continues.  In 2006 an international group of strength and conditioning researchers teamed up in an effort to definitively answer this question and their work casts some new light on an old debate.  Let’s have a look at that research.

The Research

The researchers believed that one of the reasons the question of training to failure was still unanswered was that all the research that had been previously done had failed to consider and control all the important training variables.  They  hypothesized that if all the training variables except training intensity were controlled that it would definitively show whether training to failure was better than training not to failure.

They recruited 42 competitive Spanish Basque ball players (Basque ball is the name for a variety of court sports similar to handball, raquetball, & jai lai).  These athletes had been training and playing Basque ball competitively for an average of 12.5 years and had been strength training twice weekly for at least 5 months prior to the start of the study.

Subjects were divided into three groups – one group that trained to failure, one group that trained to non-failure, and one control group – and began the 16 week training program.

Both training groups strength trained twice weekly and performed the same exercises, used the same weight (% of RM), and number of reps.  The training to failure group performed their exercises to failure, using 3 sets of 10 reps the first 6 weeks, 3 sets of 6 reps to failure the next 5 weeks, and  3 sets of 2-4 reps to failure during the final 5 weeks.  To keep total intensity and volume the same, the researchers had the non-failure training group perform 6 sets of 5 reps with the 10RM during the first 6 weeks, 6 sets of 3 reps during the next 5 weeks, and 3 sets of 2-4 reps during the final 5 weeks.  Table 1 sums up the sets, reps, and intensity for both training groups.

Table 1:  Training variables for each training group

Group Phase 1, 6 weeks Phase 2, 5 weeks Phase 3, 5 weeks
Training to Failure 3 x 10 reps,10RM weight 3 x 5 reps,5RM weight 3 x 2-4 reps85-90% 1RM
Non-failure training 6 x 5 reps,10RM weight 6 x 3 reps5RM weight 3 x 2-4 reps85-90% 1RM

RM = repetition maximum, i.e. the heaviest weight that the subject can lift for a specific number of repetitions.

The subjects were tested multiple times throughout the study on a variety of measurements, including maximum strength and power, endurance, body composition, and basal hormonal balance.

The Results

What were the results?  Did training to failure produce greater increases in strength than not training to failure as is preached by high intensity advocates?  No, it did not.

There was no significant different at the end of the study in the maximum strength of the two groups.  “In conclusion, both training to failure and training not to failure resulted in similar gains in 1 RM strength, muscle power output of the arm and leg extensor muscles, and maximum number of repetitions performed during the parallel squat.”

There were two significant difference between the groups at the end of the study.  First, the training to failure group improved local muscular endurance in the bench press (maximum number of repetitions they could perform with a fixed weight) significantly more (85% vs 69%) than the non-failure training group.  Second, the non-failure group significantly increased muscle power during the final phase of training while the training to failure group did not.


Despite the claims of high intensity proponents, this study demonstrates that training to failure is not more effective for increasing strength than training hard, but not to failure.  When other training variables (volume, frequency, training weight, total reps, etc.) are held constant, training to failure is only as effective as non-failure training.

Training Application

The training lesson to be taken from this study is training to failure is not necessary in order to maximize strength.  Training hard but terminating a set prior to failure is as effective for increasing strength as is training to failure.


1.  Brzycki, M, A Practical Approach To Strength Training, Masters Press, 1995, pg 38

2. Izquierdo M, Ibanez J, Gonzalez-Badillo J, Hakkinen K, Ratamess N, Kraemer W, French d, eslava J, Altadill A, Asiain X, Gorostiaga E., Differential effects of strength training leading to failure versus not to failure on hormonal responses, strength, and muscle power gains, J Appl Physiol, 2006, 100:1647-1656


Is Training To Failure Necessary? — 28 Comments

  1. Pingback: Is Training To Failure Necessary? | Training Science

  2. This is great, keep it up! It never quite made sense that training to failure was neccessay. Even HIT advocates admit training to failure is only a rough measure of work done, and is not needed for improvements. For HIT, it seems to be more of a way to more accurately measure progress than anything else.

  3. Hi Rich,

    I thoroughly enjoyed your debunking of HIT training, and reading about muscle factor training made me come up with the resolution to deviate from my 8-12 rep monoculture which I’ve been following for the last year or so. Keep up the great work!

    Still, I would like to get your opinion on a different piece of writing in favor of HIT which was posted on Drew Baye’s website (a HIT personal trainer):

    To quote from the linked article: “Rodney et al. [30] reported significantly greater gains (41.2% to 19.7%) in dynamic strength when training to muscular failure compared to submaximal sets of exercise. Similarly, Schott et al. [31] reported significantly greater gains in isometric strength when training to failure compared to stopping the exercise short of failure (24.9kg to 14.3kg) and Drinkwater et al. [32] reported significantly greater dynamic strength gains (9.5% to 5%), and also peak power for a bench press throw exercise when training to muscular failure compared to not training to failure (40.8W/10.6% to 25W/6.8%).” (p. 149)

    Basically, all the shortcomings of HIT exposed in your articles are contradicted, and both sides base their argument on piles of scientific studies. So, for a layperson (such as myself) there is really no way of finding out who’s right, and the main criterion for whom to believe becomes eloquence and persuasiveness rather than content matter.

    You have already pointed out some weaknesses in the evidence put forward by Carpinelli and the like – does this article basically provide a rehash of the same sources, or is there anything new which might make you reconsider any of your former conclusions?

    Thanks for providing this website, and greetings from Germany,

  4. I just read Body by Science, so I’m coming from that angle. McGuff and Little advocate high intensity (i.e. to concentric failure) but low dosage (i.e. 1 set to failure, not three) and low frequency (i.e. once per week, not three times per week) training. They would say that training too frequently at a high intensity does not provide adequate recovery time for fast muscle fibers, thereby limiting both potential increases in strength and also apparent performance on strength tests. They advocate HIT training not more than once per week – subjects in this study trained much more frequently than that. They actually cite research that often shows either similar results of more frequent training at lower intensity, but then ask, why would you bother when you’re just spending more time in the gym and increasing risk of injury?

    Also, high intensity training triggers a different kind of energy metabolism, due primarily to the release of epinephrine and norepinephrine and the associated wholesale release of emergency energy stores: glycogen from muscle cells and fat from fat cells – this is valuable in situations when the trainee is trying to lose fat or combat metabolic syndrome.

    I should mention that I’m really trying wrap my head around these debates, and Rich, you’re really engaging the debate at a very high level of scientific rigor, which is extremely uncommon. I really appreciate it.

    • makelemonade,

      I would point out to McGuff and Little that the preponderance of research does not support their theory. There is a large body of strength training research. If you wanted to know what, overall, we could learn by reviewing all of that research you would conduct something called a meta-analysis. The advantage of this type of analysis is it would give you the big picture. Instead of reviewing one or a few studies, you would include the results from every relevant study done. In this way you would get a clear picture of what that body of research had to tell us.

      In fact, several of these meta-analysis’ have been done and the results are quite clear – HIT as traditionally preached, has been shown to produce lesser gains than more traditional multi-set training (single set vs multiple sets – new research).

      Candidly, I was a HIT advocate for many years, making it difficult for me to accept the results of those meta-analysis. However, the research says what it says, my personal beliefs be damned. I was forced to abandon my belief in the superiority of HIT because the research simply didn’t support that belief.

      In my opinion, McGuff and Little will need to provide solid research that nullifies those meta-analysis before they can reasonably claim that research shows HIT to be as effective as multi-set training.

      • Hello Rich,

        There is something that I don’t understand about the “meta-analisys”. The article that Florian shared in the previous post, is it not a meta-analisys? Regarding meta-analisys, what guidelines define the relevance of a study? how are the claims of such studies challenged?
        I don’t understand why the analisys done by Kreiger is more important that the Fisher’s study, or the analisys done by McGuff and Little.

        Thanks for the help!

        • Hi, Jose.

          The study that Florian shared was not a meta-analysis. It was a review of some of the available research but did not include, nor attempt to include, a meta-analysis of the available research. Essentially what Fisher et al did in the study Florian cited was say, “study A says this, study B says this, study C says this” and so on. In contrast, a meta-analysis would takes the results of studies A, B, and C and combine them into one big study to see what the combined results produce. A meta-analysis is a statistical analysis of a group of studies and is a very different procedure than a review of a group of studies. Meta-analysis’ are more rigorous and have a lower chance of being influenced by researcher bias. When a team of researchers decide to conduct a meta-analysis they define the parameters of the study so as to include only relevant studies.

          Due to the fact that meta-analysis is designed to examine all the relevant research in order to reach a conclusion, their findings tend to be considerably more reliable and accurate than any single study or a review paper.

      • We should all be ready willing and able to perform 10 sets of 20 all to vein popping failure if studies confirmed this is true for optimal results. Unfortunately for people like this (me) this new information is a hard pill to swallow but facts are facts. Now if I know my 10 rep max, I go 8 with 2 in the tank plus 8 with 1 in the tank and then pretty darn close to 10 just below failure now. The body can be notorious for misleading the mind as to what is best which can swing from either far too easy to far too damaging for the muscles and or the CNS. I’ve found in this instance, the body is screaming THANK YOU and oozes with feelings of strength and health with a great night sleep. Thanks for submitting that study Rich.

  5. It would be interesting to know the difference in muscle power output in the final phase, or is it a basically no increase for the failure group compared to something normal in the non-failure group?
    Possilbe differences between both groups regarding muscle size are neglected as well which seems weird since that usually concerns bodybuilders like arthur jones

  6. Too bad the study used essentially a strawman of what true HIT training protocol would be, which is ONE set to failure, not THREE sets to failure per exercise. That invalidates the whole claim of this article and referenced study.

    Oy vey.

  7. Definitely an interesting take on the subject. Some of the evidence here conflicts with other studies I have seen done around this topic, which did note some positive results from training to failure.

    There has been more research into this recently with regards to the potential negative effects of training to failure – how the biomechanics of exercises are affected and the impact this has on form and subsequent risk of injury, for example. This was recently examined in this article:

    Just wondered people’s views on this? It seems premature to completely reject training to failure on the basis of a single study.

  8. Can someone explain to me how the group that trained to failure had an increase in muscular endurance while the non failure group didn’t. I thought if you increase your muscular endurance at above 60 % of your 1 rep mad you’ll also gain a little strength. And how did the non failure group have more muscular power at the end of the study can someone please explain to me, I would really appreciate it and thank you!

    • I suspect the HIT group had more endurance gains because they used higher rep ranges, especially in the first 6 weeks (10 reps for 3 sets vs. 5 reps for 6 sets).

      Per the SAID principle, if you train in a higher rep range, you will get better at doing things in a higher rep range than someone who trains at a lower rep range, and vice versa.

  9. My wife has been doing HIT for 2 1/2 years one year ago I started. twice a week 15 min. only. I’m 59 years young and played basketball three times a week until 10 years ago. playing weight 218. One year ago after 10 years of almost no intense physical activity I’m almost 270. Now Im around 245 gained approx 15 lbs muscle and look feel like a 35 year old. Sorry if this isn’t what you want to debate but HIT works in very little time. Think about it 52 weeks times 1/2 hour = 36 total hours. If your have little or no time HIT will build back muscle , strength and flip your metabolism. Dave Spokane WA.

  10. These studies are certainly interesting, but ultimately it is still difficult to determine the optimum way to train for many reasons.

    As a previous commenter has mentioned, the HIT protocol here uses multiple sets, whereas most HIT workouts use just 1 set per exercise.

    Most popular HIT protocols also use slower rep pace, either a 2/1/4 or 3/1/3 or even 5/1/10 (seconds for concentric, hold, and eccentric).

    Most HIT protocols advocate for different rep ranges than this study indicates too, favoring about 60-90 seconds for one set. So for a 3/1/3 pace, that would be in the 6-10 rep range, or for “superslow” it might be only 4-6 reps.

    And HIT is usually not performed twice a week, although that can vary due to individual recovery times according to HIT advocates. The goal is to make progress each session, and if progress is not made then to increase the rest time (train less frequently). That makes it difficult to standardize to test in a controlled study.

    So this study sadly does not indicate whether the commonly advocated HIT protocols are better or worse than other ways of training, like a typical 8-12 rep range, 3 sets, 1/0/1 pace, done 2 or 3 times a week.

    Even if HIT were merely equivalent, that might mean it was advantageous, as it can be much more time-efficient than multi-set, multiple days per week protocols.

    On the other hand, HIT might be at a disadvantage if the intensity level is so difficult to maintain – especially with continual progression – that trainees dread their workouts and skip them. Or HIT might lead to greater injuries on some exercises requiring more skill, such as back squats.

    So unfortunately, the debate rages on unabated….

  11. For efficiency, HIT wins hands down! Coming up to 4 decades, it continues to deliver. Where trainees go wrong is using it, any version of HIT, like a cookie cutter approach instead of making personal adjustments as needed. People need to learn to listen to their bodies and think for themselves.

  12. LOL- “3 sets of 10 at 10RM” just proves that these researchers (just like most of all the others) don’t recognize what failure is. These studies can be thrown out. If you failed at 10 reps you’d be lucky to get 8 reps on the next set (unless you had an unusually high percentage of slow-twitch fibers). Then you might get 6 reps on the last set. Someone with a lot of fast-twitch fibers might be struggling to get even 3 reps on the last set if he had really trained to failure on the previous 2 sets.

    • Very good point, probable why results are so surprising !
      It was disturbing me but thanks to your comment I now guess the results would have shown a benefit to true failure for the same volume…

  13. “Researchers found large changes in testosterone following a moderate-intensity protocol [70% 1RM, multiple subfailure sets], and no significant increases were found after numerous sets performed at 100 percent intensity [1RM]. This suggests that bodybuilders may benefit from lifting in a moderate repetition range of eight to 12. It appears the greater rise in testosterone may be the result of greater metabolic stress, such as increases in lactic acid following moderate-intensity, rather than maximal-intensity, training. Moderate intensity, high-volume exercise—eight to 12 reps and more than four sets—leads to greater increases in testosterone than low-volume, maximal-intensity exercise.” (Med Sci Sports Exerc. 36(9):1499-1506. 2004.; J Appl Physiol. 74(2):882-887. 1993.)

  14. Hi Rich,
    You commented on studies that refer to a number of repetitions at less than 1 or 2RM loads. This is more of a fatigue failure, yes?
    What studies have looked at short duration ( eg. 4sec isometric ‘hold’) at near 1RM, load ‘failure’ or what could be coined tensional failure?
    Might this approach elicit additive strength gains eg. if done in a cluster set pattern of 3 sets of 3×1 (30 sec between reps and 5 min between sets)

  15. To maximize strength gains I thought athletes should lift heavy weight to failure using 100% 8 rep max until failure because each additional set adds little to the final increase in strength since for each extra set the percent of 8 rep max decreases, being that failure is achieved only on the final set. Isn’t the above the reason why strength gains achieved using one set equal to that of multiple set? I wonder

    • Jelani,

      While it remains a contentious topic, meta-analysis actually shows multi-rep sets to be slightly more effective than a single set training. It’s interesting to note that when Arthur Jones first introduced the idea of “high intensity training” using 1 set to failure he stated in no uncertain terms that 1 set to failure was superior to multi-set training. He didn’t say it was as good as multi-rep training he claimed it was better than. Today, 1 set to failure proponents mostly claim 1 set to failure is equal to multi-set training. The “better than” has been replaced with “equal to”. Unfortunately for 1 set advocates, the research doesn’t support their claim.

      The reason multi-set is slightly better is because it trains a few more fibers than single set training.

      However, even multi-set training, as most commonly employed by most trainees, undertrains, or fails to train, many, many fibers in every whole muscle due to the very limited rep range used in most multi-set programs.


  16. I train to failure on every set. Been training for three years lifting 80 percent of 1 rep max. I have gained 35 pounds of muscle. Train twice a week upper lower split free weights i am 57-year-old

  17. I train with the purpose of triggering growth hormone release and elevating it when I run since that’s the hormone that stimulates muscle growth. I know I’m there when I’ve pushed through the sense of a pressure in my head and running becomes a little easier. At the end of my run (for example: 68% MHR for 90 minutes, typically with a 65 minute relief point), I’m sure I can run a lot longer, so it’s never “to fail”. My legs have developed the strength they need to prevent repetition injury, as my knee issues seem to be long behind me. I imagine I could continue to run to exhaustion for *more* HGH release, but doing that all the time would probably kill my motivation for running–then there wouldn’t be any muscle development.

  18. How can anyone say, what’s right an what isn’t? Not ONE body on this planet is the exact same, 100% the same. Recently, I had picked up a Super Slow tempo to failure, and the results have been tremendous. I am a seasoned lifter, 6 years, and training to this form has given the best results I’ve ever felt, especially contraction. If you’re looking at studies, look at Dr. Doug McGuff and PhD James Steele on their information on training to failure w/SS tempo.

  19. “How can anyone say, what’s right an what isn’t? Not ONE body on this planet is the exact same, 100% the same. ”

    This kind of argument uses fundamental fact to refute an argument, yet it’s often applied to refute an argument that has never been made (Strawman fallacy). The argument itself can be used to dismiss the field of physiology entirely, after all “How can *anyone* (specifically in this case, the physiologists who conducted the study) say what is right and what isn’t?” (Reductio ad absurdum exposes the fallacy).

    I think most institutionally trained physiologists understand that no one is 100% the same. That’s why their studies consist of evenly distributed groups and filled with statistical language (mean and standard deviation), to wrap it around that fact. If such an investigator presented this argument in a scientific journal to dismiss the study, it would seem rather silly to everyone involved.

    But your contribution after that cringe-worthy argument is completely valid and worth investigation under scientific controls and peer review.

    However, if the super slow tempo to failure technique wasn’t considered in the study (I don’t think it was), the study do not and can’t be used to contradict your personal results (it’s not making that argument), but neither can your results be used to contradict a study that never considered it. It only draws attention to a relatively *new* technique that enhances training to failure, possibly making it better in general than the not training to failure technique, as it was applied in the study.

Leave a Reply

Your email address will not be published. Required fields are marked *