Warning: This is really nerdy!
One of the most confusing aspects of swimming to athletes that are not used to the water is breathing. In most sports, breathing is something we take for granted. Air is always around for us and there for the taking. Not so when your face is planted in a swimming pool or lake! In the water, you must actively work to breathe, and this work will take energy and slow you down. World class 50 meter swimmers are well aware of this fact. Many of them take 2 breaths or less for their entire race!
It is true that triathlon is a different sport than 50 meter sprint swimming. Taking too few breaths in the open water will reduce your oxygen supply, which can cause you to tire rapidly. Of course, taking too many breaths causes you to slow down. How do you know if you are breathing too little or too much? This experiment seeks to broaden understanding of how much time is lost with every breath. With this information, you can decide an appropriate breathing pattern for yourself. As with all of my experiments, the results are only as valid as the assumptions and experimental setup, and those are listed below. If you just want to see the conclusions and recommendations, skip to the bottom of this article.
Assumptions in the Experiment to determine time loss while breathing:
 The metric in this experiment is time lost per breath taken over a swimming distance of 25 yards. The 25 yard distance was selected because it is short enough to do a large volume of tests within the same study. A series of lengths were swum where the number of strokes per breath was controlled. Two, four and eight strokes per breath were investigated.
 Freestyle was the stroke utilized throughout this experiment. The athlete was free to determine their own pulling and flutter kicking technique. Technique was held consistent throughout the experiment.
 The response variable in this experiment is raw sprint time. Time was selected because the expected loss per breath is expected to be small (less than 0.10 seconds per breath). Such a small differential requires a precise measurement system, which can be obtained easier by a stopwatch compared to counting heart beats per minute.
 Each 25 yard data point was swum as fast as possible with the prescribed number of strokes per breath. An assumption made is that the time lost to breathe at high speeds (sprinting) is equivalent to time lost at slower speeds (in a lake). This assumption is reasonable, as the act of breathing is independent of arm or leg exertion. In other words, you work just has hard to turn your head to breathe if you are swimming fast or slow. However, this assumption will not be validated in this experiment.
 For each number of strokes per breath, a data set was collected when breathing just to the left side and a separate data set was collected when breathing just to the right side.
 The differences between left and right side breathing were determined by statistical methods. If it could be concluded that the left – right difference is not relevant to the experiment, the two data sets for each breathing requirement (left & right) would be merged and analyzed together.
 The various strokes per breath (2, 4, and 8) were evaluated completely on both sides of breathing (left/right). Eight repetitions were taken for each of these situations. Thus, the total number of experimental runs was 3 x 2 x 8 = 48.
 The tests were completed in a randomized order. Prior to the start of each test, the recorder announced the breathing pattern (breathe every 2, 4 or 8) and which side to breathe on. The athlete did not receive any further advance notice of what to do.
 Times were obtained by the recorder on the pool deck with a stopwatch to a resolution of 0.00 seconds. The recorder started the watch when the athlete assumed a still inwater starting position and stopped the watch when the athlete touched the opposite wall.
 Once started, the athlete took quantity4 underwater dolphin kicks before surfacing. This equated to approximately 5 yards of underwater swimming. This distance was sufficient to maintain speed off the start, but no so long as to interfere with the breathing results.
 The number of strokes taken over each length was collected. This was to ensure that the athlete was swimming in a consistent manner regardless of breathing pattern.
 All experimental tests were performed in the same lane, same pool and same direction of length. This was done to ensure equivalent environmental conditions for each test.
 Data was collected over a two day period where the days were consecutive. Exactly half of the experiment was completed on each day. It would have been ideal to collect all data in the same day and session, but it was too difficult for the athlete to perform the required number of sprints all at once. Since data was collected over consecutive days at the same facility, it is not likely to adversely affect results.
 A single athlete (Duane Dobko) was used as the test subject for the study. The assumption made in this test is that body type and swimming ability are independent of time lost per breath. Since different swimmers were not evaluated, this assumption could not be confirmed. The benefit to using an experienced swimmer is improved consistency, resulting in a more statistically valid result.
 The athlete wore the same equipment (brief style swimsuit) throughout the test. Wetsuits or fullbody suits were not worn at any time.
 Between repeats, adequate time was allowed for complete rest and recovery. Such time was at the discretion of the athlete.
Overall Results
Figure 1: Results – 25 yard sprint times with various breath holding patterns
Figure 1 demonstrates that the time lost per breath is smaller than expected (0.0441 seconds per breath versus an expected 0.1 seconds). This time loss can be used to estimate the effects of various breathing patterns in open water swimming. Results are shown in Figure 2.
Figure 2:
The values in Figure 2 were calculated assuming that the average swimmer takes 15 strokes for every 20 yards of swimming in open water, compounded over the distance in yards of each open water swim (for example, a 2.4 mile swim is 4,224 yards), and using the amount of time lost per breath (0.0441 seconds).
Figures 1 and 2 illustrate that the time lost by taking a few extra breaths is very small. For example, if you switched from 2 strokes per breath to 8 strokes per breath for an entire 2.4 mile Iron distance swim, your projected time would only improve by 52 seconds. For most of us, this gain would be more than offset by the increased fatigue due to lack of oxygen! The result suggests that breathing as much as possible is the way to go, even breathing every stroke. However, a more detailed analysis suggests otherwise. The tradeoff to breathing a lot lies in the corresponding increase in variation.
Analysis of Variation
The overall consistency of swim times was evaluated with different breathing rates. The intent is to determine if different breathing patterns create more or less variable results. The variation in strokes per breath is summarized in the Box/Whisker plot in Figure 3.
Figure 3 – Box / Whisker plots (definition of Box / Whisker on right)
Note: Click on above chart to enlarge.
Figure 3 illustrates how the variation in sprinting speed increased when the test subject increased breathing rate. The standard deviation of each data set was evaluated and shown in Figure 4.
Figure 4 – Scatter Plot, Standard Deviation of Data
As shown in Figure 4, the standard deviation (a measure of variation) increased a whopping 80% when the test subject increased the number of breaths per length from 2 to 9. The increase when the athlete doubled the number of breaths from 2 per length to 4 per length was 32%. Important to note is that the test athlete had significant swimming experience. It is likely (though not verified in this experiment) that the average triathlete would have an even greater increase in variation.
The large differences in variation suggest that a higher breathing frequency increases swimming speed variation. This makes sense, as head movement will tend to disrupt body positioning. This is why most swimmers find it easier to swim evenly with a snorkel.
Any increase in variation of swimming speed carries negative consequences for the triathlete in open water. As the athlete speeds up and slows down, their body position in the water must also rise and fall, forcing more water movement and greater energy expenditure to attain the same speed compared to a swimmer who swims more evenly and stays higher on average in the water. In other words, when you breathe more, you must work harder and use more energy to go the same speed. This hypothesis makes logical sense. However, it could not be confirmed in this experiment.
Number of strokes taken per length
The total number of stokes taken in each experiment was counted and documented. It is important that stroke count is reasonably consistent in order to accurately compare different breathing patterns. In other words, if the stroke count is varying in excess of 4 strokes per length, it would create a difference in the total number of breaths taken.
The maximum number of breaths taken in any 25 yard swim was 21, and minimum was 19. The average was 20 and the standard deviation was 0.6. These numbers make it sufficient to assume that number of breaths per length is consistent. The following relation of “strokes per breath” and “breaths per length” was established. This relationship is used throughout the experiment.
Stroke per Breath

Equivalent Breaths per length

2

9

4

4

8

2

Left Breathing versus Right Breathing
A more detailed Box / Whisker plot (Figure 5) was established in which the various breathing patterns were analyzed when the test subject breathed to the left side and to the right side.
Figure 5: Analysis of Left vs. Right breathing
Figure 5 illustrates that there is no difference in times or variation when the test subject breathed to the right side versus the left side. This result is to be expected for experienced swimmers. However, the result may differ for athletes with less experience. The data analysis is shown in Figure 6.
Figure 6
Note: Click on above chart to enlarge.
In figure 6, the pvalues and confidence intervals were calculated assuming that the left data and right data are independent variable sets (no paired comparison).
As shown in Figure 6, none of the pvalues showed any significant differences (to 95% confidence). The confidence intervals ranged from plus/minus 0.15 seconds at 2 breaths per length versus plus/minus 0.25 seconds at 9 breaths per length. These confidence intervals are sufficiently small to conclude that there is no relevant difference in time when the athlete breathes on the left side versus the right side.
From this data, it can be concluded that it is valid to merge all left and right data together into a single data set.
Statistical Analysis of Data
The data was analyzed in order to determine if the calculated time lost per breath taken in swimming freestyle (0.0441 seconds) was estimated from statistically valid data. The summary information is shown in Figure 7. The results show that the data gathered for this experiment is sufficient to make conclusions regarding time lost per breath.
The test for normality (ShapiroWilk method used) shows the pvalues all in excess of 0.05. With this result, it can be concluded that the data is insufficient to prove nonnormality. A much larger sample size would be required to prove normality. Thus, the conclusion that the data is normal is not definitive but is sufficient for this analysis.
The Kurtosis value in each data set is also reasonable. In a normal distribution, the excess Kurtosis is zero. Since all excess Kurtosis values are negative and small in this experiment, it can be concluded that the data is slightly flat, meaning that the characteristic peak associated with the middle of the bell curve is not quite as high as it should be. However, the Kurtosis values are considered small enough that the normal distribution is assumed for all data sets in this experiment.
An analysis of means may now be performed with the standard formulas associated with normality. The combined (left/right) data sets of 2 breaths per length were compared to 4 and 9 breaths per length, respectively. All of these comparisons were performed assuming independent data sets (no paired comparison).
The results of analysis of means (See Figure 7) showed a statistically significant difference between all data comparisons (to 95% confidence). Thus, a difference between 2 breaths per length versus 4 breaths per length, and 4 breaths per length versus 9 breaths per length can be considered conclusive. The confidence intervals range from approximately plus/minus 0.2 seconds up to plus/minus 0.35 seconds. The range of confidence intervals occurred due to the increase in variability when more breaths were taken in a swimming length. Based on this analysis, it can be concluded that an average loss of 0.0441 seconds per breath is a valid assessment.
Figure 7 – Statistical analysis summary
Overall Testing Conclusions
The experiment revealed two key findings as described in the following paragraphs.
The experiment established that there is a time loss associated with every breath you take when swimming freestyle. This loss in time is quite small, measuring only 0.0441 seconds per breath. When considered alone, this small value of time loss suggests that triathletes should not consider the number of breaths they take in a race, and may breathe as often as possible in order to maintain adequate oxygen supply.
However, the details of the experimental results cast doubt on this narrow focused conclusion. The variation in the test subject’s maximum swimming ability increased by 80% when breathing every two strokes versus breathing every 8 strokes. This suggests a loss of efficiency as breathing frequency increases. It can be projected that the inefficiency will add up over the long distances of open water swims in triathlon, increasing fatigue and decreasing total speed. As this experiment was narrow focused in evaluating maximum speed over 25 yards, it was not possible to confirm this statement.
So, what can a triathlete take away from this experiment? In my coaching clinics, I always recommend to my clients to take as many breaths as they need, but not to take any more. If you can get away with breathing every four strokes, then that is what you should do. If you start to get oxygen depleted, it is okay to breathe every 2 strokes. You will not lose a lot of time with this decision. However, be aware that your entire stroke will be less efficient when you breathe more. You must determine for yourself if this decrease in efficiency is going to be made up by the increase in oxygen intake. If you feel like you are drowning or suffocating, chances are that breathing more often is going to work well for you. If there is enough air in the tank, then charge ahead and don’t look up.