The purpose of this study was to investigate the effects of cognitive load during practice on university wind students' learning. Cognitive load was manipulated through instrument family (woodwind or brass) and the amount of repetition used in practice (highly repetitive or random). University woodwind and valved-brass students (N = 46) completed two practice sessions and two retention testing sessions. Participants practiced three seven-pitch tasks in either a blocked, repetitive order or in a random order. Performance trials were scored for accuracy, speed and evenness. At 24-hour retention, woodwind players who had practiced in a random order were able to play significantly faster, F(4,80) = 4.448, p = .003, h2 = .15, and more evenly, F(4,80) = 4.464, p = .003, h2 = .16, than woodwind players who had practiced in a blocked order. However, for brass players, blocked practice supported better accuracy, F(4,64) = 3.508, p = .012, h2 = .15, and speed, F(4,64) = 4.489, p = .003, h2 = .18, than random practice. A secondary research question examined participants' judgment of learning. At the end of the second practice session, participants predicted the tempo they expected to be able to play the tasks at during the 24-hour retention session. A significant correlation between predicted and performed tempo was found for brass players using blocked practice (r = .360, p = .04).