Abstract
Orientation identification tasks often show improvement with practice, but training with zero external noise may be more efficient: multi-session training in zero external noise showed essentially full transfer to high external noise, while training with high external noise showed limited transfer to low external noise (Dosher & Lu, 2005). Here, we examined how training in zero or in high external noise transferred to multiple external noise levels as assessed in each session. In four sessions, observers were trained with feedback in a peripheral (5.4 deg) two-alternative orientation task (-55+/-10 deg) in either zero or high external noise (in two separate groups of n=7), and were also assessed at four external-noise levels (0, 0.8, 0.17, 0.33) without feedback at the beginning (pretest) and end (posttest) of each session. Contrast thresholds tracked a low accuracy level (65%) to minimize learning without feedback (Liu et al, 2012). Transfer to different retinal locations was tested in the fifth session. Results: 1) The contrast threshold improved in both training groups under their respective training conditions, with faster learning observed in the zero external noise group (log-log threshold vs block slope: -0.28 vs -0.07); 2) training improved performance in all external noise levels in both groups; 3) within-session improvement from pretest to posttest was larger in the first few sessions when most learning occurred and overnight consolidation (from previous-day posttest to current day pretest) was also most apparent early in training; 4) there was substantial location transfer. The integrated reweighting theory (IRT, Dosher et al, 2013) successfully captured these results with the same learning rate, because external noise in the stimulus adds noise to and perturbs learned weights from stimulus representations (encoding) to decision (decoding)(Lu, et al, 2010). Empirically, assessing multiple external noise levels throughout training revealed transfer across external noise conditions throughout the training process.