Academic journal article Perception and Psychophysics

A Face Recognition by Similarity (FRBS) Conjecture

Academic journal article Perception and Psychophysics

A Face Recognition by Similarity (FRBS) Conjecture

Article excerpt

Cohen (1963) investigated free recall in two lists of words. The first contained unrelated words. The second comprised words drawn from several semantic categories, where the number of categories was equal to the number of words in the first list. He found that recall of unrelated words was equal to the recall of categories. The face recognition by similarity (FRBS) conjecture proposes that this relation cannot be applied to face recognition. Following Cohen's design, two different experimental situations for generating two target faces were constructed. The findings showed that the number of correct recognitions of specific facial features belonging to the first target face (e.g., nose, chin) was greater than or equal to the number of categories of visually similar facial features belonging to the second target face (e.g., different long noses, round chins). In addition, theoretical underpinnings for the FRBS conjecture were suggested.

Between the 1950s and the early 1970s, a major effort was made in memory research to discover the relation between organizational processes in verbal free recall (e.g., Baddeley, 1976; Murdock, 1974; Tulving & Pearlstone, 1966; Wood, 1972). The general finding was that free recall was enhanced when the words in the list to be remembered were grouped under different categories.

One important question from that research program, which constituted a theoretical trigger for the present study, concerns the relationship between the recall of two different lists of words. One list, L, contains n words, which are divided into g categories (clusters, groups), with z words per category-that is, z = n/gthat are semantically related. The other list, 1, contains n' unrelated words, where n' = g. For example, L has n = 30 words, which are divided into g = 10 categories with z = 3 words per category. List 1 has n' = 10 unrelated words, n' = g. The question is whether the number of categories that will be correctly recalled from L [denoted by C(g, z), which depends on g and z] will be equal to the number of words that will be correctly recalled from 1 [denoted by W(n'), which depends on n']-that is, whether W(n') = C(g, z).

Cohen (1963) tested the hypothesis that W(n') = C(g, z) (called the chunk hypothesis) under two conditions: one that fulfilled and one that did not fulfill the total-time (T) requirement. Accordingly, the lists should be equal in their T: the multiplication of the number of words in a list (n) by the presentation time of a word (t)-that is, T = nt-since it was found that the number of words recalled in a single-trial free recall increased as a function of T (see Murdock, 1960, 1974).

On the basis of Miller's (1956) concept of a chunk (a group of words that "are related in some manner" [Cohen, 1963, p. 227]), Cohen phrased Miller's chunk hypothesis in the following mannen

A list comprising 20 categories of words and a list of 20 unrelated words are equivalent in the sense that each contains 20 units or chunks of information. . . . there should be no significant difference between the number of words recalled from a list of 20 unrelated words compared to the number of categories represented by the recalled words of a list of 20 categories, (p. 227)

He found that when T was held constant across lists, there was no significant difference between recall of unrelated words from List 1 and recall of categories from List L. This can be viewed as supporting the chunk hypothesis. However, when the T requirement was not fulfilled, the number of categories recalled was higher than the number of unrelated words recalled.

The major question of the present article is whether the chunk hypothesis [W(n') = C(g, z)] may apply to recognition of faces. Our attempt to answer this question led to development of the present article's conjecture.

The Face Recognition by Similarity (FRBS) Conjecture and Previous Preliminary Findings

To answer the question above, the following two crude analogies were drawn: (1) a list consisting of words and categories versus a face consisting of facial values and similarity groups per facial dimension, and (2) recall of unrelated words and categories versus recognition of facial values and similarity groups. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.