The Processing of Features and Dimensions

特徵、向度之知覺過程

張學揚(S. Y. Chang)

《Journal of Medical Sciences》 3卷1期 (1979/01) Pp. 791-802

摘要

Garner認為刺激的任何單一部份（unique event）的知覺必須考慮到刺激的整體，他對刺激屬性（attribute）的定義，包含了特徵（feature）及向度（dimension）二方面的意義，特徵的意義為刺激中是否存在某種屬性，例如圖形中有無角度的屬性存在，至於向度的意義為刺激中存在某種屬性，而此屬性包含了某種程度值的概念，例如角度的大小向度為何。由於特徵，向度的意義不同，Garner假設其二者之知覺過程亦有差異，本實驗意圖以附合刺激屬性定義之簡易圖形為刺激羣（stimuli set）來探索求證Garner的假設。其實驗過程中讓被試在不能獲得充份刺激信息的限制情況（state limitation）下，經由瞬間顯示器（tachistoscope）作辨認測驗（identification task），從辨認測驗結果中的錯誤分佈（error distribution）情形來研究特徵屬性不同的刺激在測驗中的錯誤分佈是否與向度屬性刺激的錯誤分佈相同。實驗I-A，所有的刺激羣為變化特徵屬性產生而成，其辨認之錯誤分佈是呈不對稱的（asymmetric），即被試者在限制情況下知覺到的刺激，往往會減少或省略刺激所具之特徵屬性。實驗I-B，所有的刺激羣是為變化向度屬性產生而成，基於刺激與刺激之間之差異僅為值之多寡不同，所以相似性高。在相似的知覺下，犯錯的機率應是相同，因此被試者之辨認測驗結果顯示錯誤分佈是呈對稱的（symmetric）。I-A與I-B的實驗結果，以錯誤分佈為指標，顯示刺激的特徵與向度二種屬性是具有不同的知覺機能，均能支持Garner的假說。在實驗II中刺激羣之變化同時包括特徵、向度二者屬性而成，此實驗之總數據可能由於複雜的刺激圖形，加強了個別差異的影響，並不能清晰地附合Garner的假說，但是若能個別審視被試者的測驗結果，不同的錯誤分佈型態仍然存在於部分數據當中。

並列摘要

Garner proposed that the perception of any unique event must be considered in terms of the total set of stimuli. It is necessary for us to make descriptions about properties of stimuli. Features and dimensions are two major subclasses of an attribute, or a component property. It is very interesting to know whether the two subclasses of an attribute are processed differently or not. In order to approach this problem, identification task, state limitation and error distribution were used in this study. In experiment I-A, the stimuli set generated from feature, the error distribution from identification task was asymmetric, subjects always 10st feature in the identification task under the state limitation. In experiment I-B, the stimuli set generated from dimension, the error distribution was symmetric, subjects made those errors on the basis if similarity. The results of experiment I-A and I-B fit Garner's hypothesis confirmly. In the experiment II, the stimuli set is generated from both feature and dimensions. The results of overall data were not clear, but when you check the data individually, there still indicated some different function between dimension and feature under the state limination.