Individual Difference in Serial Position Effects with Connected Discourse

Abstract

The large body of evidence amassed about the serial-position effect has led to its description as one of the best esmblished and most pervasive phenomena in verbal (Slamecka, 1967, p. 111). In its most general form, the effect refers to a clear indication that with ordered items initial and final material appears to be more readily mastered than middle material. The basic curve consistently emerges even though its slope and symmetry have been experimentally altered (Marshall & Werder, 1972) and attributedifferentiated Ss exhibit curves of differing heights and asymmetry (Fagan, 1972) . The present study examined this effect with connected discourse and investigated ics generalizability to individual Ss. A 20-min. audio-taped lecture was presented to 25 graduate students. Evenly spaced throughout the lecture were 30 items which were tested using a multiple-choice examination with randomly ordered questions (KR-20 reliability = .77). At the conclusion of testing, items were placed in serial order and divided into initial, middle, and final sections (10 items per section) with section and total scores computed for each subject. Section scores resulted in a typical effect with initial, middle, and final means of 6.68, 5.16, and 5.72, respectively. A repeated-measures analysis of variance detected a significant serial effect (F:,u = 8.45, p < ,001) favoring initial and final over middle items. The inference drawn from serial-position studies is that Ss master material at a certain level in the initial section, do less well toward the middle, and improve again at the end. For the analysis of individual differences this pattern was expected. Three other basic patterns were examined, including: (1 ) Ss who never get worse and always get better ( A ) , (2 ) Ss who never get better and always get worse ( B ) , and ( 3 ) Ss who get better and then get worse ( C ) , the opposite of the expected pattern. The expected pattern was exhibited by 11 Ss (Ms = 7.0, 4.18, 6.27), while Pattern A typified 3 Ss (Ms = 7.0, 7.33, 8.67), Pattern B was exhibited by 8 Ss (Ms = 6.5, 4.88, 3.75), and Pattern C by 3 Ss (Ms = 5.67, 7.33, 6 .0) . In short, 14 of 25 Ss responded with atypical patterns. While the total test mean of the expected pattern group (17.45) was nearly identical to that of the total group (17.56), the panern deviation means suggest differences in ability may be related to serial effects. Pattern B Ss had by far the lowest test mean (15.13), Pattern C Ss had a mean test score (19.00) exceeding that of the total group, and Pattern A Ss had by far the highest test mean (23.00), even chough theoretically pattern and test score are independent. Although restrictions of sample size limit interpretation, that the majority of Ss responded in atypical ways suggests that applications and generalization of the effect must remain tenuous pending more extensive investigation of individual differences in learning connected discourse.