<p>In the analysis of operant behavior, there is little empirical research on the relationship between observing responses and primary stimulus generalization. This work aimed to investigate eye fixations when S+ and S- dimensions were varied on generalization tests. Ten university students participated. Their training consisted of a MULT VI 1&#xa0;s EXT schedule followed by MULT VI 2&#xa0;s EXT schedule. Discriminative stimuli were three Gabor line tilts. S+ and S- had 45º and 135º slopes, respectively. After participants achieved discrimination indices of 75%, generalization tests in extinction began. There were two different conditions: (1)&#xa0;S+ was replaced by stimuli with angles of 15^ο, 30^ο, 45^ο, 60^ο, and 75^ο (five participants). (2) S- was replaced by 105^ο, 120^ο, 135^ο,^, 150º, and 165º (five participants). In both training and tests, eye tracking equipment recorded observing responses defined as visual fixations. S+ variations yielded sharp observing response gradients. However, S- variations yielded flattened, bell-shaped, and U-shaped observing response gradients. These data contribute to the limited information on human observing during tests of primary stimulus generalization. The study provides a methodology for accomplishing a more complete characterization of behavioral processes that may be operative when normally capable adults are exposed to variations in S+ and S- on generalization tasks.</p>