Rated ease of learning as a common scale of learning-speed of various language materials

Abstract

The measures of meaningfulness (M) and association value can predict learningspeed only when experimental materials are of the same class. If different classes of test materials are used, the two measures are known to reduce their predictability. In a previous study where two differing classes of materials (two-letter syllables and digits with the same level of M) were used, more Ss were found to recall digits than two-letter syllables. (Imae & Umemoto, 1966)In order to find a general scale which can prove valid when used across different classes of test material, the “rated ease of learning” measure (REL) was tested in the present study. A 7-point scale was constructed, ranging from “very hard to memorize” to “very easy to memorize” and a total of 104 items were selected from among 4 different classes of material (26 items from each class). The 4 classes of stimuli used in the present study were Japanese two-letter syllables (corresponding to English CVC trigrams), digits, alphabetical monograms and digrams. Mvalues of the alphabetical monograms were not computed in Japan, but those of the items in the three remaining classes were all known.A total of 69 college students rated all the 104 items, arranged at random, during atest session. The scale average for each item was calculated by assigning values ranging from 7 to 1 to each corresponding point on the scale-7 to “very easy to memorize” and 1 to “very hard to memorize”. The RELs obtained for the alphabetical monograms are shown in Table 1. The Spearman's γ was computed between M's obtained from American Ss by Anderson and those corresponding RELs given in Table 1. The correlation was only. 10.The relationship between REL and M is shown in Fig. 1, and correlations between the two measures appear high within each class. As shown in Fig. 1, however, even at the same level of M, different materials have different RELs.Five minutes after the rating task, Ss were told to recall as many items as possible. The median percentage of Ss who incidentally recalled the items was then calculated. The relationship between the percentage of Ss who recalled the items and the M values is shown in Fig. 2. The M measure predicted the percentage of Ss who recalled the items within, one particular class of material. When stimuli were of different classes, however, the percentage of Ss who recalled the items was significantly different at the same level of M. Digits were recalled most often, alphabetical digrams next, followed by twoletter syllables.In Fig. 3, the percentage of Ss who recalled each item is shown as a joint function of the class of material and the REL. A group of dots enclosed by the dotted circle are not significantly different in the percentage of Ss who recalled the items. It is possible to combine all the items belonging to the different materials at each of the scale points, and, in Fig. 4, the median percentage of Ss who recalled the combined items are plotted as a function of the REL.In Fig. 4, it is shown that the percentage of Ss who recalled the combined items increases monotonously as a function of the REL, and all the 6 dots in Fig. 4 are significantly different from one another in terms of the percentage of Ss who recalled the items. It is concluded that the REL is very robust in predicting learningspeed across different classes of material, and sensitive enough to predict differences in learning-speed when the stimuli judged have a difference of only one scale unit on the REL scale.