Endoscopic examinations of the larynx were recorded on 49 horses at rest and while exercising on a 5% inclined high-speed treadmill for 8 minutes at a maximum speed of 8.5 m/sec. Subjective laryngeal function scores at rest and while exercising were based on the degree and synchrony of arytenoid abduction. Arytenoid abduction was expressed as a left:right ratio of rima glottidis measurements. Horses with arytenoid cartilage asynchrony at rest (grade 2) could not be distinguished from normal horses (grade 1) when exercising because full abduction was maintained throughout the exercise period. Five horses with incomplete left arytenoid abduction at rest (grade 3) maintained full abduction during exercise; one grade 3 horse had dynamic collapse of the left side of the larynx. All horses with laryngeal hemiplegia at rest (grade 4) had dynamic collapse of the left side of the larynx during exercise. Forty-two horses with a resting left:right arytenoid abduction ratio ≥.71 consistently had complete arytenoid abduction at exercise. Seven horses with a leffcright ratio <.71 consistently showed dynamic collapse at exercise. There was no significant difference in the exercising left:right ratio between normal horses (grade 1) and grade 2 or grade 3 horses. These results suggest that horses with arytenoid asynchrony at rest do not suffer progressive collapse of the rima glottidis during exercise, and that incomplete arytenoid abduction at rest is an unreliable predictor of such collapse. Surgical treatment of all grade 2 horses and some grade 3 horses may be inappropriate.