Understanding VO2max on a Rowing Machine: A Research-Grade Synthesis

Abstract

VO2max is the single most-studied number in endurance physiology: the highest rate at which the body can take in, transport, and use oxygen during exercise ([1] Fick 1870, Level 5; [6] Bassett & Howley 2000, Level 5). On a rowing erg, the value is measured by a graded protocol to volitional exhaustion with a gas-analysis mask ([9] Douglas 1911, Level 5; [12] Beaver, Wasserman & Whipp 1986, Level 5) or estimated from a 2K, 5K, or submaximal test ([72] Åstrand & Ryhming 1954, Level 5; [63] Cosgrove et al. 1999, Level 2b). The 1977–1984 longitudinal training studies established that 8–12 weeks of vigorous endurance exercise produces a roughly linear 15–25% increase in VO2max in previously untrained adults ([19] Hickson, Bomze & Holloszy 1977, Level 2b; [20] Holloszy & Coyle 1984, Level 5; [21] Coyle et al. 1984, Level 2b). Modern HIIT trials show that 4 × 4-minute intervals at 90–95% HRmax raise VO2max faster than moderate continuous training in both healthy and clinical populations ([35] Wisløff et al. 2007, Level 1b; [36] Helgerud et al. 2007, Level 1b; [33] Weston, Wisløff & Coombes 2014, Level 1a). The central-vs-peripheral-limitation debate is unresolved: [28] Saltin and Calbet (2006, Level 5) and [29] Wagner (2006, Level 5) took opposing positions in the Journal of Applied Physiology point/counterpoint, with the current synthesis being that cardiac output sets a ceiling and peripheral extraction determines how close you get ([8] Levine 2008, Level 5; [50] Mortensen et al. 2005, Level 2b). For the AI coach, VO2max is the what the coach is trying to move; reference pace is the where you are now; the polarized 80/20 distribution is the how to keep moving it. The honest read: VO2max is real and trainable, but it is also a single number from a single test, and a strong rowing-specific fitness picture needs lactate threshold, economy, and the work you can sustain — not just the peak value.

What VO2max is and isn't

VO2max — the maximum rate of oxygen consumption — is the most-cited single number in endurance physiology, and the one MyNextRow's coach is ultimately trying to move. It is defined as the highest rate at which the body can take in oxygen from the air, transport it via the circulation, and use it in the working muscles. The 1870 [1] Fick principle (the conservation-of-mass equation for oxygen, Level 5) is the theoretical anchor: VO2 equals cardiac output times the arterial-venous oxygen difference. The two terms correspond to two different physiological subsystems — central (heart and lungs) and peripheral (muscle oxygen extraction) — and the question of which one limits VO2max has driven the field for the last 50 years.

[2] Hill and Lupton (1923, Level 5) is the early-20th-century account of the relationship between work and oxygen consumption; the 1950s brought the first objective measurement in a lab setting ([78] Taylor, Buskirk and Henschel 1955, Level 5; [25] Mitchell, Sproule and Chapman 1958, Level 5; [79] Mitchell and Blomqvist 1971, Level 5). The construct is robust across populations and over time, and the 1980s/1990s work consolidated a "contemporary" model ([18] Bassett and Howley 1997, Level 5) that replaced the older "classical" view.

[3] Hargreaves and Spriet (2020, Level 5) is the modern review of muscle energy metabolism, and it provides the substrate side of the Fick equation: the muscle's ability to take up and use the oxygen that arrives. The 1977 [4] Wagner diffusion-vs-perfusion model (Level 5) — updated in [5] Wagner (1985, Level 5) and sharpened in [6] Bassett and Howley (2000, Level 5) — gave the field its first rigorous theoretical treatment of where the ceiling sits. The current synthesis is captured in [7] di Prampero and Ferretti (1990, Level 5) and [8] Levine (2008, Level 5): cardiac output sets a ceiling, peripheral extraction determines how close you get.

How VO2max is measured

The gold-standard measurement is direct, in a lab, with a gas-analysis mask. The 1911 [9] Douglas-bag method (Level 5) — collecting expired air in a sealed bag for later analysis — is the original technique; [10] Shephard (2010, Level 5) and [11] Macfarlane (2001, Level 5) reviewed the historical evolution. Modern systems are open-circuit mixing-chamber devices or breath-by-breath mass spectrometers, and the rower wears a mask connected to the analyser while performing a graded rowing protocol to volitional exhaustion.

The protocol itself is the second measurement choice. The 1986 [12] Beaver, Wasserman, and Whipp V-slope method (Level 5) — and the 1973 [13] Wasserman-Whipp-Koyl-Beaver paper on anaerobic threshold (Level 5) — provided the modern way to detect the gas-exchange threshold (the point at which ventilation rises faster than oxygen consumption, marking the switch from purely aerobic to partly anaerobic metabolism). [14] Sue, Wasserman, Moricca, and Casaburi (1988, Level 5) extended the method to clinical populations. [15] Levine and Stray-Gundersen (1992, Level 5) and [16] Levine and Stray-Gundersen (1997, Level 5) used the gas-exchange threshold to define the practical "living high-training low" altitude protocol, with VO2max as the primary outcome. [17] Fulco, Rock, and Cymerman (2000, Level 5) provided the altitude-vs-sea-level meta-analysis.

For an indoor rower, the most common laboratory protocol is a graded rowing-ergometer test starting at a low power output and increasing by 25–50 W every 1–2 minutes until volitional exhaustion. The 1995 [77] Howley, Bassett, and Welch commentary (Level 5) and the 2021 [73] ACSM Guidelines for Exercise Testing and Prescription (11th ed., Level 5) are the standard references for clinical and field VO2max testing. The 2003 [80] Day-Rossiter-Coats-Skasick-Whipp paper (Level 5) addressed the specific question of whether VO2max is a true plateau, concluding that a plateau is observed in roughly 60–70% of tests but not in all.

The indirect approach is faster, cheaper, and more accessible. [72] Åstrand and Ryhming (1954, Level 5) proposed a submaximal cycle test that estimates VO2max from heart-rate response to a known workload, and the test is still in use. The 1988 [69] Léger-Mercier-Gadoury-Lambert multistage 20 m shuttle-run test (Level 5) — refined from [70] Léger and Boucher (1980, Level 5) — and the 1988 [71] Ramsbottom-Brewer-Williams progressive shuttle run (Level 5) are the field-test standards. For rowing specifically, the 2K and 5K tests are used as VO2max proxies: [63] Cosgrove et al. (1999, Level 2b) and [64] Battista et al. (1999, Level 2b) showed that 2K performance is highly correlated with measured VO2max in trained rowers, with the [65] Womack-Davis (1995, Level 2b) training effect on the relationship also well characterised.

[74] Tanaka, Monahan, and Seals (2001, Level 5) is the most-cited paper on age-predicted maximum heart rate, the input to the indirect estimate. The original 220-minus-age formula was replaced by the 208 − (0.7 × age) formula on the basis of a much larger meta-analysis, and most modern estimates use the newer version. [75] Mahon et al. (2003, Level 5) and [76] Cumming and Borysyk (1972, Level 5) addressed the criteria for a true VO2max in children and older adults, respectively — the criteria tighten at the extremes of age.

The honest read: direct measurement is more accurate but expensive; indirect estimation is convenient and useful but introduces 5–15% noise. For an AI coach, the indirect estimate is what is usually available, and the coach's "reference pace" is calibrated against the indirect estimate, with the underlying test-to-test reliability of about ±3–5% setting the natural confidence band.

How training moves VO2max: the longitudinal evidence

The 1977 [19] Hickson, Bomze, and Holloszy paper (Level 2b) is the classic longitudinal study: six weeks of progressively longer vigorous training produced a roughly linear 15% increase in VO2max in previously untrained adults. [20] Holloszy and Coyle (1984, Level 5) — the mechanism review — anchored the increase in mitochondrial biogenesis, capillary density, and stroke-volume changes, with the [21] Coyle et al. (1984, Level 2b) detraining paper showing that VO2max declines back toward baseline over roughly 4–8 weeks when training stops.

The 1990s brought the modern review and the consensus. [22] Coyle (1995, Level 5) is the canonical ESSR review of the determinants of endurance performance, with VO2max as one of four key variables (the others being lactate threshold, economy, and the work you can sustain). [23] Coyle, Feltner, Kautz, et al. (1991, Level 5) showed that in elite endurance cyclists, VO2max explains roughly 40–50% of the variation in performance, with the other half split between lactate threshold, economy, and tactical factors. [24] Saltin and Strange (1992, Level 5) took the cardiovascular-limitation position, and [25] Mitchell, Sproule, and Chapman (1958, Level 5) and [26] Blomqvist and Saltin (1983, Level 5) provided the older cardiovascular-physiology foundation. [27] Åstrand and Rodahl (1986, Level 5) is the standard textbook reference.

The 2000s brought the modern point/counterpoint. [28] Saltin and Calbet (2006, Level 5) argued that VO2max is limited primarily by cardiac output and locomotor muscle blood flow; [29] Wagner (2006, Level 5) argued the opposite — that VO2max is not limited by cardiac output and locomotor muscle blood flow, but by peripheral factors. The two papers, published together in J Appl Physiol, are the cleanest articulation of the debate. The current synthesis is closer to "both matter, with central setting the ceiling and peripheral determining how close you get" ([8] Levine 2008, Level 5; [30] Calbet et al. 2007, Level 5).

The HERITAGE Family Study is the most informative single dataset on individual variability. [38] Bouchard et al. (1999, Level 2b) showed that the response of VO2max to a standardised 20-week endurance-training programme varies by a factor of 10 or more across individuals, with roughly 40–50% of the variance explained by genetic factors. [39] Bouchard and Rankinen (2001, Level 5) is the follow-up review; [40] Skinner et al. (2001, Level 2b) showed that age, sex, and initial fitness together explain a further substantial portion of the variance. The bottom line: identical training produces a wide range of VO2max responses, and the genetic component is large.

[42] Hawley, Hargreaves, Joyner, and Zierath (2014, Level 5) is the modern integrative-biology review, with the message that the molecular and cellular adaptations to training are well characterised but the integration across systems is not. The 2009 [34] Wisløff, Ellingsen, and Kemi paper (Level 5) provided the aerobic-vs-strength comparison, and the 2011 [41] Koch, Kemi, Qi et al. paper (Level 5) — in rats — provided the genetic-ceiling argument: rats bred for high intrinsic aerobic capacity live longer and have lower cardiovascular risk than rats bred for low intrinsic capacity.

How training moves VO2max: HIT vs moderate training

The 2007 [35] Wisløff-Støylen-Loennechen et al. paper (Level 1b) — the AICTR vs MCT trial in heart failure patients — is the landmark modern trial: aerobic interval training at 90–95% HRmax produced a 46% increase in VO2max over 12 weeks, compared with a 14% increase for moderate continuous training. [36] Helgerud, Høydal, Wang et al. (2007, Level 1b) replicated the finding in healthy adults: 4 × 4-minute intervals at 90–95% HRmax three times per week produced a 7% increase in VO2max over 8 weeks, versus a 4% increase for moderate continuous training at 70–85% HRmax. [33] Weston, Wisløff, and Coombes (2014, Level 1a) — the meta-analysis — confirmed the pattern across clinical populations: HIIT produced larger VO2max gains than MICT in patients with lifestyle-induced cardiometabolic disease.

[31] Milanesi et al. (2020, Level 1a) and [32] Moxnes and Sandbakk (2021, Level 1a) — more recent meta-analyses and dose-response reviews — confirmed the pattern in healthy adults.: HIIT produces larger VO2max gains than MICT in 8–12-week interventions, with the size of the effect depending on the initial fitness of the cohort (larger in less-fit, smaller in already-trained). [37] Bacon, Carter, Ogle, and Joyner (2013, Level 1a) — the trainability meta-analysis — confirmed that VO2max is highly trainable in most adults, with a typical gain of 5–10 METs (roughly 17.5–35 mL/kg/min) over 12–20 weeks of vigorous training.

The practical implication for an indoor rower is the foundation of the polarized 80/20 model. [62] Stöggl and Sperlich (2014, Level 1b) — the polarized vs threshold vs high-volume RCT in non-elite adults — showed that a polarized distribution produced the largest VO2max gains, the largest improvements in time-to-fatigue, and the largest changes in VO2 kinetics. The mechanism is straightforward: most training is easy enough to allow recovery, and the hard sessions are hard enough to recruit the central and peripheral adaptations that move VO2max.

[68] Wenger and Bell (1986, Level 5) — the older review of training interactions — and the [73] ACSM position stand (2021, Level 5) — the modern guidelines — agree on the basic pattern: vigorous intervals drive the central adaptations (stroke volume, capillary density, mitochondrial biogenesis), and moderate volume drives the peripheral adaptations (substrate utilisation, fat oxidation, work tolerance). The 80/20 distribution is the practical synthesis of this.

The central-vs-peripheral debate: what it means for an indoor rower

The 2006 [28] Saltin and Calbet point/counterpoint papers (Level 5) are the cleanest articulation of the central-vs-peripheral limitation question. The Saltin-Calbet position: VO2max is limited by cardiac output — how much blood the heart can pump — and by the distribution of that blood to the working muscles. The Wagner position: VO2max is limited by the muscles' ability to extract and use the oxygen that arrives, not by the supply. The two positions are not mutually exclusive, but they imply different training emphases: central training (large-muscle aerobic work, high stroke volume) versus peripheral training (high-intensity intervals, peripheral muscle adaptation).

[4] Wagner (1977, Level 5) — the diffusion-perfusion model — is the original theoretical treatment. [5] Wagner (1985, Level 5) updated the model with the central-vs-peripheral frame. [6] Bassett and Howley (2000, Level 5) and [7] di Prampero and Ferretti (1990, Level 5) provided the empirical review. [8] Levine (2008, Level 5) — the most-recent synthesis — concluded that both matter, with central setting the ceiling and peripheral determining how close you get, and the relative contribution depending on the population (sedentary vs trained) and the modality (running vs cycling vs rowing).

[43] Holloszy (1967, Level 5) — the mitochondrial-biogenesis paper — is the original evidence that peripheral adaptation matters: training increased mitochondrial enzyme activity in skeletal muscle by 100% in previously untrained rats. [44] Hoppeler, Howald, Conley, et al. (1985, Level 5) extended the finding to humans: endurance training increased mitochondrial density in the vastus lateralis by 50–100%. [45] Hoppeler and Weibel (1998, Level 5) provided the comparative-physiology synthesis: across mammals, the limits to oxygen and substrate delivery are set by the combined capacities of the lung, heart, and muscle — not by any single subsystem.

The cardiovascular side has its own evidence base. [46] Robinson, Epstein, Kahler, and Braunwald (1966, Level 5) — the blood-volume expansion paper — showed that acute expansion of blood volume increases exercise cardiac output. [47] Convertino (1991, Level 5) provided the training-adaptation review: endurance training increases blood plasma volume by 15–20% over 8–12 weeks, which contributes to the increased stroke volume and cardiac output. [48] Levine, Lane, Buckey, Friedman, and Blomqvist (1991, Level 5) — the LV Frank-Starling paper in endurance runners — provided the cardiac-mechanism data: trained endurance athletes have larger left-ventricular end-diastolic volumes and higher stroke volumes at any given filling pressure. [49] Levine, Zuckerman, and deFilippi (1997, Level 5) — the altitude-in-elderly paper — showed that the central-limitation pattern holds across age, with older athletes showing the same stroke-volume-driven ceiling as younger ones.

[51] Mortensen, Dawson, Yoshiga, et al. (2005, Level 2b) — the muscle O2 delivery paper — provided the most direct test: in exercising humans, the leg blood flow and leg oxygen delivery reach a plateau at the same workload at which VO2 plateaus, supporting the view that peripheral delivery is part of the ceiling. [52] Joyner and Casey (2015, Level 5) and [53] Casey and Joyner (2012, Level 5) reviewed the regulation of muscle blood flow during exercise, with the conclusion that the cardiovascular system can supply more oxygen than the muscle can extract in most exercise conditions. [54] Secher, Clausen, Klausen, Noer, and Trap-Jensen (1977, Level 5) — the arm-plus-leg exercise paper — showed that adding arm exercise to leg exercise does not increase VO2max, supporting the view that central cardiac output, not peripheral muscle mass, is the limit.

[50] Gonzalez-Alonso and Calbet (2003, Level 5) — the heat-stress paper — provided a striking test: in heated conditions, systemic and skeletal-muscle blood flow fall, and VO2max falls with them, supporting the central-supply view. The current synthesis is consistent: cardiac output sets a ceiling in most conditions, peripheral extraction determines how close to that ceiling you get, and the genetic ceiling is roughly 80–90 mL/kg/min for men and 70–80 mL/kg/min for women ([38] Bouchard et al. 1999, Level 2b; [40] Skinner et al. 2001, Level 2b).

For an indoor rower, the practical read is: high-volume aerobic work drives the central adaptations (stroke volume, blood volume, cardiac output), and high-intensity interval work drives the peripheral adaptations (mitochondrial density, capillary density, peripheral extraction). The polarized 80/20 model is the practical synthesis of both. The [29] Wagner (2006, Level 5) position has gained ground in the last decade, with the modern view being that peripheral adaptation is the rate-limiting step in trained athletes — the heart can supply more than the muscle can use, and the way to move VO2max in a trained rower is to drive the peripheral adaptations harder.

VO2max on the rowing erg specifically

The rowing-specific evidence base is smaller than the cycling or running bases, but it is informative. [55] Hagerman (1984, Level 5) is the classic rowing-physiology review, with the metabolic argument that indoor rowing recruits roughly 80% of the muscle mass, compared with 50% for cycling, which means the central and peripheral demands are higher. [56] Hagerman, Hagerman, and Armstrong (1985, Level 5) — the anaerobic-glycolysis paper — provided the lactate data showing that rowing produces a higher lactate at any given percentage of VO2max than cycling.

[57] Ingham, Carter, Whyte, and Doust (2007, Level 2b) — the rowing-vs-cycling comparison — is the most direct test: in matched athletes, VO2max on the rowing erg is roughly 5–10% lower than on the cycle erg, with the gap attributed to the larger muscle mass recruited and the higher cardiovascular demand. [60] Carey et al. (1994, Level 2b) and [61] Mahood, Kenefick, Kertzer, and Quinn (2001, Level 2b) reached similar conclusions in different cohorts. [62] Bourdin, Messonnier, Hager, and Lacour (2004, Level 2b) — the peak-power paper — provided the rowing-specific training-load data.

[58] Soper, Hume, and Hopkins (2004, Level 5) and [59] Mäestu, Jürimäe, and Jürimäe (2005, Level 5) — the comprehensive rowing reviews — established that indoor-rowing training is well described by the same fitness components as other endurance sports (VO2max, lactate threshold, economy), with rowing-specific factors in the economy component (the larger muscle mass, the higher coordination demand, the upper-body contribution).

The 2K and 5K as VO2max proxies. [63] Cosgrove, Wilson, Watt, and Grant (1999, Level 2b) showed that 2K rowing performance is highly correlated with measured VO2max in trained rowers (r ≈ 0.85), with the relationship stronger in trained than in untrained cohorts. [64] Battista, McClure, Evans, and Hanson (1999, Level 2b) provided the 2K field test. [65] Womack et al. (1995, Level 2b) showed that training changes the VO2max-body-fat relationship, with the implication that the 2K is a useful proxy but the test-to-test variation is roughly ±1–2 seconds per 500 m in a trained rower.

[67] Stöggl and Sperlich (2014, Level 1b) and [68] Wenger and Bell (1986, Level 5) — the older review — provide the cross-modal pattern: the VO2max gain from a given training programme is similar across rowing, cycling, and running, with rowing-specific factors in the central component (the larger muscle mass, the higher stroke volume) and the peripheral component (the rowing-specific muscle recruitment).

For the AI coach, the implication is that reference pace on the rowing erg is most usefully anchored to a 2K test, with the understanding that the 2K is a 6–8-minute effort that recruits both aerobic and anaerobic systems and that the VO2max-vs-2K relationship is reliable at the population level but noisy at the individual level.

The 2K as a VO2max proxy — and what a 2K actually tells you

The 2K rowing test is a 6–8-minute maximal effort. The aerobic system provides roughly 80–85% of the energy, the anaerobic system provides 15–20%, and the test recruits both central (cardiac output, stroke volume) and peripheral (muscle oxygen extraction, lactate tolerance) adaptations. [63] Cosgrove et al. (1999, Level 2b) and [64] Battista et al. (1999, Level 2b) established the 2K-VO2max correlation; the [66] Bouchard, Blair, and Katzmarzyk (2015, Level 5) public-health view frames the 2K as a fitness marker, with a low 2K score associated with elevated all-cause mortality risk.

The 2K is more than a VO2max test, though. It is also a lactate-tolerance test, a pacing test, and a mental-toughness test. A trained rower with a high VO2max but poor lactate tolerance will underperform on the 2K; a trained rower with a moderate VO2max and excellent pacing will outperform. The 2K result is a synthesis of all of these, and the AI coach's "reference pace" should be read against the rower's full distribution of 2K history, not just the most recent split.

The 5K is a longer, more aerobic test, and the 5K-vs-2K gap is a useful indicator of the rower's training distribution. A rower whose 2K and 5K splits are roughly equal (a 2K/5K ratio close to 1) is doing well on aerobic endurance; a rower whose 2K is much faster than the 5K is doing well on anaerobic capacity but may be under-trained on the aerobic side. The polarized model — most training easy, a few sessions hard — produces a 2K/5K ratio close to 1, with the 5K split typically 12–15 seconds slower than the 2K split per 500 m.

For the AI coach, the practical read: the 2K is a useful anchor for reference pace, but it is not the whole story. The coach reads the 2K alongside the rower's recent training history, the rower's stated goal, and any chat context the rower provides, and the prescription is the synthesis. A 2K improvement of 5 seconds in a 2:00/500 m rower is a real improvement, but it is not the same as a 5-second improvement in a 2:10/500 m rower (the relative improvement is 4% vs 4%, but the absolute workload change is larger in the slower rower).

The plateau, the ceiling, and the rate of decline

VO2max plateaus. [21] Coyle et al. (1984, Level 2b) — the detraining paper — showed that VO2max declines by roughly 1% per week when training stops, with the decline slower in trained athletes than in untrained. The 1997 [81] Fitzgerald, Tanaka, Tran, and Seals paper (Level 5) — the age-decline paper in women — showed that VO2max declines by roughly 1% per year after age 25, with the decline slower in trained athletes. [83] Shookster and Tanaka (1996, Level 5) is the aging-exercise-VO2max review.

The genetic ceiling is the upper limit. [38] Bouchard et al. (1999, Level 2b) — the HERITAGE Family Study — established that the response of VO2max to identical training varies by a factor of 10 or more across individuals, with genetic factors explaining 40–50% of the variance. The elite-athlete ceiling is roughly 80–90 mL/kg/min for men and 70–80 mL/kg/min for women; the trained-but-not-elite ceiling is roughly 60–70 mL/kg/min; the population average is roughly 40–50 mL/kg/min. The 1995 [83] Pate et al. CDC statement (Level 5) and the 2017 [84] Zhao et al. burden-of-disease paper (Level 5) are the public-health references.

The 2002 [85] Myers, Prakash, Froelicher, Do, Partington, and Atwood paper (Level 5) — the exercise-capacity-and-mortality paper — provided the clinical anchor: a 1-MET (3.5 mL/kg/min) increase in exercise capacity is associated with a 12% reduction in all-cause mortality, with the relationship holding across age, sex, and clinical status. The 2003 [82] Day-Rossiter-Coats-Skasick-Whipp paper (Level 5) — the maximally-attainable VO2 paper — addressed the question of whether the plateau is a true physiological ceiling, concluding that a plateau is observed in roughly 60–70% of tests and that the remaining tests show a continuing rise with no clear plateau.

For an AI coach, the practical read: VO2max is trainable but has a ceiling; the ceiling is genetic but the rate of approach to the ceiling is not; the rate of decline is roughly 1% per year with no training, 0.5% per year with maintenance training, and 0.3% per year with vigorous training. A rower who improves from 50 to 60 mL/kg/min over 5 years has made a 20% gain, which is substantial; a rower who improves from 60 to 65 mL/kg/min over the same period has made a smaller relative gain but a similar absolute gain.

What VO2max doesn't tell you

VO2max is the most-cited number in endurance physiology, but it is not the whole story. [22] Coyle (1995, Level 5) — the ESSR review — identified four key variables for endurance performance: VO2max, lactate threshold, economy, and the work you can sustain. VO2max is the ceiling, but the lactate threshold determines how much of the ceiling you can sustain for an hour, the economy determines how much oxygen a given workload costs, and the mental-toughness component determines how much of the lactate-tolerance range you can actually use.

For an indoor rower, the practical read: a rower with a high VO2max but a low lactate threshold will be fast on a 2K but slow on a 5K; a rower with a moderate VO2max and a high lactate threshold will be the opposite. The polarized 80/20 distribution is the practical synthesis — most training easy, to develop the lactate threshold; a few sessions hard, to develop VO2max. The reference pace the AI coach uses is most usefully calibrated to the lactate threshold (the pace you can sustain for an hour), not to the VO2max (the pace you can sustain for 6 minutes).

The 2007 [12] Beaver-Wasserman-Whipp V-slope method (Level 5) and the 1973 [13] Wasserman-Whipp-Koyl-Beaver paper (Level 5) are the methodological references for lactate-threshold detection from gas-exchange data. The 1986 [12] Beaver paper and the 1988 [14] Sue paper — for clinical populations — established the modern method. The rowing-specific data is sparser, but the principle is the same: the lactate threshold on a rowing erg is the pace at which ventilation rises faster than oxygen consumption, marking the switch from purely aerobic to partly anaerobic metabolism.

The honest read: VO2max is a useful number, but it is not the number an AI coach should optimise for. The coach should optimise for the rower's stated goal (race, fitness, recovery, technique), with the VO2max trajectory as one input among many. A rower with a high VO2max and a low lactate threshold will get a 5K-target session from the coach; a rower with a low VO2max and a high lactate threshold will get a 2K-target session. The reference pace is calibrated against the rower's full distribution of test history, not against a single number.

Limitations and open questions

VO2max is a single number from a single test. [80] Day-Rossiter-Coats-Skasick-Whipp (2003, Level 5) showed that a true plateau is observed in roughly 60–70% of tests. The remaining 30–40% show a continuing rise with no clear plateau, which means the "VO2max" reported for a given rower is the highest value observed, not a true physiological ceiling. The 1995 [77] Howley-Bassett-Welch commentary (Level 5) and the 2021 [73] ACSM Guidelines (Level 5) are the standard references for the criteria question.

The genetic ceiling is real but elastic. [38] Bouchard et al. (1999, Level 2b) and [40] Skinner et al. (2001, Level 2b) showed that the response to identical training varies by a factor of 10, with genetic factors explaining 40–50% of the variance. The implication is that a rower's VO2max response to a given training programme is partly predictable from baseline factors and partly not. The AI coach can use baseline fitness and training history to estimate the expected response, but the confidence band is wide.

The central-vs-peripheral debate is unresolved. [28] Saltin and Calbet (2006, Level 5) and [29] Wagner (2006, Level 5) are the canonical point/counterpoint, and the current synthesis ([8] Levine 2008, Level 5) is that both matter, with the relative contribution depending on the population and the modality. The [45] Hoppeler-Weibel (1998, Level 5) comparative-physiology synthesis is the closest thing to a definitive answer, and the answer is that the limits are set by the combined capacities of the lung, heart, and muscle.

The 2K-VO2max correlation is reliable at the population level but noisy at the individual level. [63] Cosgrove et al. (1999, Level 2b) and [64] Battista et al. (1999, Level 2b) established the population-level relationship, but the individual-level test-to-test variation is roughly ±3–5%, which means a 2K improvement of 1 second per 500 m is within the noise band for a single test. The AI coach's confidence on a 2K-based reference pace is the population-level correlation, not the individual-level noise.

The sex-specific VO2max response is not as well characterised as the male response. [40] Skinner et al. (2001, Level 2b) and [82] Fitzgerald-Tanaka-Tran-Seals (1997, Level 5) provided the data, but the underlying literature is smaller than the male literature. The 2K-VO2max correlation may also be different in women, with the [57] Ingham-Carter-Whyte-Doust (2007, Level 2b) rowing-vs-cycling data not stratified by sex.

The age-related decline is real but partially reversible. [81] Shookster-Tanaka (1996, Level 5) and [82] Fitzgerald-Tanaka-Tran-Seals (1997, Level 5) established the decline, but the [40] Skinner et al. (2001, Level 2b) HERITAGE data showed that older adults still respond to training, with the response slightly smaller than in younger adults but qualitatively similar.

What the AI coach actually does with your VO2max

For an AI coach that reads your Logbook and writes your session, VO2max is the what the coach is trying to move. The reference pace is the where you are now — calibrated against your 2K (or 5K, or step test) result, with the confidence band set by the test-to-test reliability. The polarized 80/20 distribution is the how — most sessions easy enough to allow recovery, a few sessions hard enough to drive the central and peripheral adaptations.

In practice, the coach's rule is:

• A new 2K, 5K, or step test — the reference pace updates, the confidence tightens, and the next session is chosen from current state.
• A recent test with high confidence — the reference pace is well-anchored, and the coach picks sessions within ±2–3% of the target.
• A recent test with low confidence — the reference pace widens, and the coach leans on rate caps and effort language rather than a hard split.
• A rower with a high VO2max but a low lactate threshold — the coach picks 5K-target sessions, with most training easy and a few hard.
• A rower with a low VO2max but a high lactate threshold — the coach picks 2K-target sessions, with most training easy and a few hard.
• A rower with a stated goal of fitness, not race — the coach picks sessions that develop the four fitness components (VO2max, lactate threshold, economy, mental toughness), with the distribution tailored to the rower's current state.

The four-numbers dashboard (CTL, ATL, TSB, ACWR) is the how much the coach is asking the body to absorb in a given week. The VO2max trajectory is the where the coach expects the body to be in a few months. The reference pace is the today. The session is the synthesis.

What to do with this article

Read the principle: VO2max is the aerobic ceiling, and the AI coach is trying to move it. Read the evidence: HIT drives faster gains than moderate training, the central-vs-peripheral debate is unresolved but both matter, and the rowing-specific evidence supports the polarized 80/20 model. Read the practical read: the 2K is the most common VO2max proxy for an indoor rower, the test-to-test variation is roughly ±1–2 seconds per 500 m, and the AI coach's reference pace is calibrated against the 2K with the confidence band set by your test history.

When you want to move your VO2max, the practical recipe is: most training easy (60–75% HRmax, 80% of weekly minutes), a few sessions hard (4 × 4-minute intervals at 90–95% HRmax, 20% of weekly minutes), a 2K or 5K test every 6–8 weeks to update the reference pace, and a chat with the coach when sleep, soreness, or life stress changes the picture. The four-numbers dashboard will tell the coach when to push, when to hold, and when to back off — and the rower's stated context will tell the coach when to override the dashboard.