Slide 11 of 21
Notes:
Roitblat et al. (1990) modeled the dolphin's decision processes as a sequential sampling problem. Although Wright and Sands had to treat each of the pigeon's observations as a single look, Roitblat, et al., were able to treat each click as an independent observation. The dolphin tended to emit a train of clicks at each target, ended either by a switch to scan another target or by a decision (indicated by a touch of a target in front of the matching stimulus). Following sequential sampling theory, Roitblat, et al. assumed that: (1) Each click has some cost to the animal. (2) Each click provides a sample of evidence about the target that returned the echo. (3) Information from successive looks is combined to identify the stimulus. (4) The animal attempts to minimize the number of looks subject to meeting a confidence criterion for identifying the stimulus. (5) As a result, confidence in identifying the stimulus grows monotonically with increasing numbers of clicks, but at the expense of making those clicks. (6) As a result of averaging multiple samples, the marginal contribution of each additional echo to the animal's confidence in identifying the target diminishes with the number of echoes already directed at that target.
