The Sharpened Romberg Test (SRT) is a test of balance commonly used in Diving Medicine. Interpretation of an abnormal test can be confounded by several factors. This study was conducted to further evaluate the usefulness of the SRT. In the first part of the study, naval and civilian volunteers in a Naval Base were recruited as subjects. The SRT scores were recorded in two separate trials; once in the morning (4 attempts) and once in the evening (4 attempts) to evaluate the effect of practice on the SRT. In the second part of the study immediate pre- and post-dive scores in a group of divers were measured to evaluate: (1) the effect of decompression; (2) the effect of the normal post-dive fatigue; and (3) the vestibular effect of swaying after a boat ride. Comparisons were also made between the distributions of the SRTs of the normal subjects and those of a retrospective group of DCI patients treated at the Slark Hyperbaric Unit, Royal New Zealand Navy Hospital (RNZNH), Auckland. The SRT was found to have an early learning effect. Second attempts were significantly better than the first (p<0.001) within the same trial. However this learning effect plateaued by the third and fourth attempts. No difference was found between trials (morning and evening). There was a post-dive decline in the scores of the first attempts only (p<0.05). Subsequent second to fourth attempts were not affected by diving. The practice effect is only evident between the first and second attempts within the same trial but not between trials. The pre- and post-dive data showed that the SRT was not affected by decompression, post-dive fatigue or the vestibular sensation of swaying that is commonly experienced after a boat ride. Comparison of the distributions between controls and DCI patients showed a bimodal pattern. Fifty-four percent (54%) of the DCI patients had 'normal' scores (60 seconds), while 14% had scores between 16-35 seconds and 32% scored less than 15 seconds. In contrast, 95% of the control groups had 'normal' scores while 5% scored between 16-35 seconds. Therefore, accepting a score of less than 40 seconds as being "abnormal" will give the SRT a sensitivity of 46%, specificity of 95% and predictive value of 82%.