Levine, B. D., Friedman, B., & Stray-Gundersen, J. (1996). Confirmation of the "high-low" hypothesis: Living at altitude - training near sea level improves sea level performance. Medicine and Science in Exercise and Sports, 28(5), Supplement abstract 742.

Competitive runners (M = 27; F = 12; 18-29 yr) were exposed to a series of control and manipulated conditions. They were as follows:

  1. Comprehensive sea-level lab testing for aerobic parameters, maximal steady state VO2 (ventilatory threshold), running economy, and velocity at VO2max, as well as a 5000 m time trial.
  2. Supervised sea-level training for four weeks to establish control/baseline conditions.
  3. Repeated testing as done in testing #1.
  4. Four weeks of training camps. Ss were randomized to: (a) HL - live high (2500 m), train low (150 m) (N = 13); (b) HH - live high, train high (2500 m) (N = 13), and (c) LL - live low, train low (150 m) (N = 13).
  5. Laboratory and performance tests were repeated once again.


  1. Both altitude groups significantly increased VO2max (HL 3.5, HH 4%) due to an increase in red cell mass volume (HL 6.3%, HH 11%). Controls did not change in either of these factors.
  2. 5 km time improved in the three groups during initial sea-level training (22.3 sec). During the latter camps, it was further improved in HL group (13 sec), deteriorated in the HH group (3 sec) and the LL group (31 sec).
  3. Velocity at VO2max improved only in the HL group (0.4 mph) as did maximal steady state VO2 (6.3%).

It was concluded that 4 weeks of high-low training improves sea-level running performance more than an identical training camp at sea-level. The improvement is due to both a high altitude acclimatization effect (red cell mass, VO2max), and a low altitude training effect.

Implication. These results should be viewed cautiously. If true controls had been exerted then all groups should have experienced a sea-level training effect. Of particular importance should have been an explanation for the LL group's regression to worse than pre-baseline levels in performance. If good control had been exerted in the study, there would have been at least no detraining effect in the sea-level control group. The result that the HH group performed better than the LL group is contrary to other published works on trained groups, particularly when using Ss as their own controls.

The design of this experiment needs to conform more to standard intrasubject designs which controls for reliability and stability of conditions and parameters. Although HL is "confirmed" the unusual nature of the other group's results casts concerns about the fairness of the experiment for evaluating this phenomenon.

Return to Table of Contents for this issue.