Some Formulae for Predicting Spelling Difficulty

Abstract

WIDE DISCREPANCIES exist in grade place ment of words in spelling lists (2, 3). Considerable discussion currently centers around the elimination of formal spelling lists in favor of teaching spelling words as the child's needs arise. This position would seem to increase further the difficulty vari ance. A number of stimulus variables relate to the difficulty of spelling words. The present paper at tempts to apply the most highly related of these to three measures of spelling difficulty, andthento de termine formulae for predicting spelling difficulty. In Hull's terms (8), a spelling word is a hetero geneous response chain with terminal reinforcement. The longer such a response chain becomes, the low er the probability that all responses will be correct. Accordingly, word length is considered tobe related to spelling difficulty. In support of this c o ne e pt, Ayers (1) found a rank difference correlation of . 88 between spelling difficulty and word length, and the writer (4) found a correlation of . 48 between number of correct spellings and word length for second grade children. Hildreth (6) felt that frequency ofoccurrenee related closely to the spelling difficulty of words. Most spelling lists are built upon a combination of expert judgment and frequency of occurrence. The relationship between frequency of occurrence and difficulty or amount learned is, however, not linear. Hull (8), for example, described the effects of a re petition as a loge function of repetitions. Howes and Solomon (7), working with recognition threshold, found that logjo of frequency more closely describes the relationship. In the present study, the log2 of frequency was used in lieu of frequency of occur rence. First, of three distributions (straight fre quency, logio and log2) log2 was the most normally distributed among the words in the sample, and sec ond, log2 can be grossly determined without special log tables.