Experiments on the spelling-to-sound regularity effect in word recognition

DAVID W. BAUER 0 1 2 0 The research was partially supported by the National Science Foundation Grant BNS76-82815. Requests for reprints should be sent to Keith E. Stanovich, Department of Psychology, Oakland University , Rochester, Michigan48063 1 University of California , San Diego, California 92037 2 KEITH E. STANOVICH Oakland University , Rochester, Michigan 48063 Subjects responded faster to words with regular spelling-to-sound correspondences than to words with irregular correspondences in both a naming task and a lexical decision task. The locus of the small, but significant, regularity effect was investigated in further experiments. When subjects were forced to respond faster than usual, via a response-deadline technique, no regularity effect was apparent in either their reaction times or error rates. In another experiment, no Regularity by Stimulus Quality interaction was obtained. It was concluded that the phonologicalreceding implied by the existence of the regularity effect takes place subsequent to lexical access. - the most compelling demonstrations of the use of phonological recoding in word recognition comes from one of the simplest of paradigms. Baron and Strawson (1976) had subjects read aloud lists of 10 words as fast as possible. The lists varied in the degree to which their component words followed spelling to-sound correspondence rules. lists composed of regular words (words following spelling-to-sound correspondence rules) were read faster than lists composed of irregular words (words not following such rules) when factors such as word frequency and word length were controlled. The data reported by Baron and Strawson (1976) suggest that the effect is sizable, approximately 165 msec/word. However, there are several problems with their procedure of having subjects read through lists of words. The problems all revolve around the fact that the time to read a given list is composed of the times for several component mental operations in addition to recognition time. For example, articulation latency, the time to execute the motor commands to name a word, is a component of the total time. It may be that irregular words simply take more time to articulate than regular words. However, no control for articulation latency is provided in the Baron and Strawson (1976) study. Additionally, the production latencies (the time to retrieve and initiate the motor commands for vocalizing a word) of irregular and regular words may differ. In short, the rather impressive regular-irregular difference displayed in the Baron and Strawson (1976) study may not be due to processes involved in recognition at all, but instead might simply be due to differences in the articulatory characteristics of the words. Gough and Cosky (1977) provide evidence that articulation latency may have been a factor in the large regular-irregular difference Strawson (1976). Gough and Cosky (1977) had subjects name single words and measured the time to initiate the response. Thus, differential articulation latencies were not implicated in the response times. Using regular and irregular words that were fairly closely matched on number of letters, number of syllables, form class, initial letter, initial phoneme, and frequency, Gough and Cosky (1977) found that reaction time to irregular words exceeded that to regular words by 27 msec. While this difference was statistically significant, it suggests a regularity effect considerably smaller than the 165 msec of Baron and Strawson (1976). Furthermore, Gough and Cosky (1977) reported no comparison of the production latencies of their sets of words. It is possible that their regular and irregular words differed in mean production latency and that response initiation processes subsequent to recognition for the 27-msec difference. Experiment 1 investigates this possibility. Subjects named regular and irregular words in an experimental situation like that of Gough and Cosky (1977). In addition, production latencies were obtained for the two sets of words. Finally, two simple measures of reading ability were administered in order to investigate whether performance in this laboratory task is related to actual reading proficiency. Subjects. The subjects were 16 undergraduate college students who were recruited through an introductory psychology subject pool for participation in the study. Stimuli and Apparatus. A PDP-l computer controlled the presentation of stimuli while recording responses and response latencies. A voice key was attached to the computer so that the reaction time could be measured. The stimuli were presented on a Hewlett Packard l311A display device and subjects sat approximately 70 cm from the display. The stimuli were 50 words with regular spelling-to-sound correspondences and 50 words with irregular spelling-to-sound correspondences. The irregular words were all either exceptions or minor correspondences as defined by Venezky (1970). The majority of both types of words were taken from Table 1 of the study by Baron and Strawson (1976). The two sets of words were equated on word frequency (mean regular frequency = 64.5, SD = 106; mean irregular frequency = 64.6, SD = 84; according to the Kucera & Francis, 1967, count) and word length (each set containing 1 three-letter word, 20 four letter words, 20 five-letter words, 8 six-letter words, and I seven-letter word). Whenever possible, a regular word was chosen so that there was a close graphemic correspondence between it and an irregular word, thus insuring that the two types did not differ in the sequential constraints of their component letters. The words were composed of uppercase letters. Four-letter words subtended a horizontal visual angle of approximately 3.27 deg. Procedure. Subjects were seated in front of the display and told that they would be seeing a series of words on the screen, one at a time. They were instructed to name each word as quickly and as accurately as possible. Following the instructions, the 100 words were presented in a sequence generated by a pseudorandom computer algorithm, subject only to the constraint that each stimulus appear exactly once. Subsequent to this block of trials, the production latencies were determined. Subjects were told that they would be seeing the same 100 words, but that they were not to respond until a plus sign appeared on the screen. One hundred trials were then completed in which a word was presented for 1.5 sec, followed by a blank interval that varied randomly between 1.4 and 3 sec. A plus sign then appeared, which was the signal for the subject to respond. Following the 100 production-latency trials, subjects were asked to read silently a 600-word passage taken from a popular magazine. They were instructed to read the passage for comprehension. The experimenter measured the reading time to the nearest .1 sec using a hand-held stopwatch. After reading the passage, the subject was asked several questions of content, to insure (...truncated)