Abstract
Previously (VSS 2004) we reported experiments in which we investigated the bias and sensitivity in the apparent equidistance settings of two laterally separated stereoscopic probes. When each probe was presented in front of a square frontal platform, the variability of the adjustments was substantially reduced, compared with when the probes were presented in empty space, thus demonstrating that the perceived relative depth of the probes was mediated by the discontinuous platforms behind them. This result was found even when the two platforms were at different distances from the observer, but this situation also produced a substantial bias in the equidistance settings, with the probe in front of the farther surface being set farther away. This constant error could have arisen because of an incomplete integration of the local information specifying the depth of each probe relative to its platform with the more global information specifying the relative depth of the two platforms; or the error could have arisen from an underestimation of the depth separation of the two platforms. Here we report a new experiment that replicated the previous experiments but also contained conditions in which the probe in front of the more distant platform was adjusted to equal the perceived distance of the other platform, thus measuring the perceived relative depth between the two platforms. We again found that the probes appeared equidistant when the probe in front of the farther platform was farther away. When the probe was adjusted to match the distance of the other platform, however, the bias was much smaller and in the opposite direction. These results suggest that the bias in the equidistance settings of the two probes is primarily due to incomplete integration of local and global stereoscopic information. This result is compared with an analogous result previously found using monocular pictorial information (Meng & Sedgwick, 2002).
Supported by NSF Award # 0001809