Fitness from cross-training such as rowing transfers only partly to swimming, cycling and running, and never exceeds the effect of training the sport itself
Interpretation
Tanaka's 1994 review concluded some VO2max transfer exists between modes but cross-training effects never exceed sport-specific training, with swimming transferring least. Millet, Vleck and Bentley's 2009 review describes VO2max as mode-specific, with more transfer from running to cycling than the reverse. Millet et al. 2002 (4 elite triathletes, 40 weeks) found bike and run training transferred to each other but swimming didn't. Flynn et al. 1998 (20 well-trained runners) found added cycle-ergometer intervals improved 5 km time as much as added running intervals over 6 weeks. The direction is consistent, but studies are small and old, and no trial has tested rowing ergometer training on triathlon or running performance. Wilson et al. 2014 link ergometer volume and previous back injury to low back pain in rowers.
Practical implication
Use the rowing machine for a reason (injury, travel, off-season variety) rather than in place of swim, bike or run time when those are possible, and build erg volume gradually with good technique. Rowing-specific transfer is inferred from cycling and running data.
Sources
5 sources, independently verified.
Some transfer of VO2max training effects exists from one mode to another, more noticeable when running is the cross-training mode; swim training may give minimal transfer; cross-training effects never exceed those of sport-specific training; cross-training may be a useful supplement during injury rehabilitation, overtraining or psychological fatigue. Abstract read on Springer; no numerical effect sizes in the abstract.
VO2max is specific to exercise modality: runners usually reach a higher VO2max on a treadmill, cyclists reach similar values cycling and running; in triathletes cycling and running VO2max are generally similar (evidence described as inconclusive). Muscles adapt specifically to a task, sometimes without VO2max change; more physiological transfer from running to cycling than vice versa. Abstract read on Springer.
Systems-model analysis of 40 weeks of swim, bike and run training: training-performance relationships were significant for running (r = 0.74) and swimming (r = 0.37); cycling and running training transferred to each other (r = 0.56); swimming showed no transfer to or from cycling or running; running training had a major effect on competition performance (r = 0.52). Very small sample, modelling study.
Both groups kept normal running and added 3 interval sessions a week for 6 weeks, either running or on a cycle ergometer. 5 km treadmill time trial improved over time (1096.6 s pre, 1076.1 s at 3 weeks, 1068.6 s at 6 weeks) with no difference between groups; running economy similar; no group differences in testosterone, cortisol or creatine kinase. Cycling, not rowing, was the cross-training mode.
Lower back is the most commonly reported injury site in rowers; among studies reporting 12-month data, low back pain incidence was 31.8% to 51%. Previous lumbar spine injury and ergometer training volume were the most significant risk factors. Pelvic rotation at either end of the stroke is preferred to lumbar flexion/extension, which novices often show; technique deteriorates with intensity and duration. Data largely retrospective.