台灣留學生出席國際會議補助

2007年7月2日 星期一

Transfer (vs. specificity) following different amounts of perceptual learning in tasks differing in stimulus orientation and position

論文發表人:劉效樺 (加州大學爾灣校認知科學系博士班)

 

http://www.visionsciences.org/meeting.html

 

「知覺學習」指經由學習或是練習之後,在知覺作業表現的增進。這類知覺學習對作業的刺激屬性具有部分或全部的特定性,如:視覺刺激的旋轉或是視覺位置的改變。本研究中,我們探討不同刺激方向與位置所產生之遷移效果(反之,特定性)與初始練習量的函數關係。從先前的研究(Jeter et al., VSS 2004) ,我們已知中度精確的方向辨別作業,呈現部分遷移效果;因此,操弄不同的初始學習量,所產生的遷移效果得以被衡量。受試者在初始的學習階段,被要求對於在視野周邊的Garbor圖形(也就是 -35°±5°,或 55°±5°),如NWSE位置,做方向判斷作業(相對於基底角度是順時針或是反時針);遷移作業則是將視覺刺激轉換至相對的視野周邊NESW位置。方向判斷作業是在雜訊交錯或是無雜訊的條件下進行訓練與測驗。使用Staircase方法來測量不同練習量之後的對比閾限(contrast threshold)值。如預期,結果顯示接受4個區段(2天)與8個區段(4天)的初始訓練,比起未接受初始練習(0區段) ,所產生的遷移效果量大。初始訓練與遷移之後的訓練擁有約略連續的power函數。在初始訓練階段關係,多接受的訓練並未產生更大的遷移效果。此結果與預期訓練能產生部分遷移的簡單模式不一致;可能是對於額外練習之飽足效應(saturating benefits) 。這些結果將與先前的知覺學習文獻中的例子一併考慮。

 

Perceptual learning refers to the improvement in performance in perceptual tasks through learning or practice. Often, this learning is partially or wholly specific to stimulus features of the task, such as rotation of the stimulus or change of visual position. In this study, we investigated transfer (or conversely specificity) as a function of the extent of initial practice for tasks that differed in stimulus orientation and position. We chose to test a moderate-precision orientation discrimination task known to exhibit partial transfer (Jeter et al., VSS, 2004) so that the extent of transfer after different amounts of initial learning could be assessed. Subjects discriminated the orientation (clockwise or counter-clockwise of a base angle) of a peripheral Gabor in initial training (i.e., -35°±5° or 55°±5°) with locations such as the NW and SE positions and in a 90° rotated transfer task in the NE and SW positions. Discrimination was trained and tested in the presence and absence of white external noise. Staircase methods measured contrast thresholds as a function of practice. As expected, the results show improvements in initial performance for the transfer task following 4 blocks (2 days) and 8 blocks (4 days) of initial training compared to 0 (no) initial training. Initial training and training after transfer had approximately equal power function rates. Additional blocks of initial training yielded little improvement in the transfer task. These results were not consistent with a simple model in which each initial training block yielded a (fractional) benefit at transfer (p < 0.001), but may suggest saturating benefits with additional practice. These results are considered in the context of previous examples of perceptual learning in the literature.