In two studies the effect of the distance between the tip of the rod and the frame sides (gap) in the rod-and-frame (RF) illusion was examined and the effect of a full-square condition was compared with that of two different frame amputations. In both studies, there were more rod-setting errors in the direction of the tilt of the inducing figure with a small gap than with a large one. These findings are consistent with the idea that in the case of small gap size local interactions contribute to determining the RF illusion. The actual length of the rod was varied in order to keep the gap constant across different frame tilts; therefore these findings cannot be due to the co-variation between gap size and frame tilt which is typical of standard apparatuses. The effect of frame amputations was compared to the full-square condition. According to Wenderoth and Beh, amputations that maintain the two orthogonal contours of the square produce the typical angular function of the RF illusion. This prediction was confirmed in both studies. However, results indicate that the full square has a stronger illusory effect in the case of a small degree of tilt of the inducing stimulus, irrespective of gap size. It is suggested that this 'square superiority' effect is related to global, not local, mechanisms. To pursue Wenderoth and Beh's observations, amputations close to the vertical meridian were used in one experiment and those close to the horizontal meridian in the second experiment. Contrary to predictions, these conditions produced overlapping results.