Subjects made smooth pursuit eye movements with a target moving horizontally at 15 deg/sec. At a specified location the target either: (1) suddenly vanished; or (2) jumped to the fovea with target retinal velocity and feedback becoming 0 (target stabilized at the fovea). In each type of trial, the subjects either: "looked" at the target, "pushed" the target, or "passively" gazed. When the target vanished, eye velocity decreased exponentially with a short time-constant (tau approximately 0.10 sec), regardless of whether the subjects were "looking," "pushing" or "passively" gazing. However, some subjects while "pushing" (using an imaginary target) did generate low velocity smooth movement (1-2.5 deg/sec) late in the offset. When the target was stabilized at the fovea, eye velocity also decreased, but with a relatively long time-constant (tau = 0.4-0.8 sec). The time-constant was the same with both "looking," and "pushing", but was shorter for some subjects with "passive" gazing (tau = 0.1-0.5 sec). These findings show that smooth pursuit offset is influenced by the presence of a target, but is relatively independent of attentional mode. All of the pursuit offset responses can be simulated using a model of the pursuit system with target velocity and position inputs, and an internal positive feedback loop enabled by target presence.