Search

As this summary suggests, some learning outcomes may be negatively affected by large-class instruction. Findings from more recent studies support this conclusion. In a study of large- and small-class instruction in educational psychology, Feldhusen (1963) found no significant differences in academic achievement, interest in teaching, or evaluations of the course and the teacher. However, attitudes toward teaching (as measured by the Minnesota Teacher Attitude Inventory) were significantly different (.05 level); also 40 per cent of the students in the large class indicated a preference for a small class. This important question of the effects of large-class instruction on learning outcomes other than academic achievement has been the subject of relatively little research. The present study was designed to compare some of the attitudes and perceptions, as well as the achievement, of students in large and small classes. Two hypotheses were tested: (1) Students taught general psychology in an instructional pattern of three large lecture sessions (approximately 200 students each) and one small-group discussion (15 students) per week would perform as well on objective measures of achievement as students taught in the traditional pattern of four lecturediscussion classes (30 to 50 students) per week; (2) Students' attitudes toward the subject matter, the instructor and instruction, other students, and future work in psychology would not be negatively affected by this change in the pattern of instruction.

THERE IS general agreement among statisti cians and research workers that sample bias re sulting from nonresponse is a common and serious problem in mailed questionnaire studies. This alleged sample bias is attributed to differences between respondent and nonrespondent groups. A 60 percent response to a mailed questionnaire study is considered to be good, but not sufficient to eliminate the main effects of this bias. An 80 percent response is considered to be necessary for the nonresident group to be sufficiently hetero genous and small to eliminate the main effects of nonresponse bias (12). The percentage of re sponse necessary for the elimination of the main effects of nonresponse bias is dependent on the amount of homogeneity of the population and on the nonrespondent group to be sufficiently hetero The nature of the bias would be a function of such situational factors as the nature of the problem, the nature of the population, etc. Rummel (9) discusses practices concerning the use of incomplete returns in questionnaire studies and cites a 1948 study by Shannon (10) which indicates that returns of 70 to 80 percent are characteristic of reputable educational research. Some writers have stressed the need for 100 per cent return of questionnaires to eliminate nonre sponse bias in certain kinds of investigations such as follow-up studies of high-school graduates (8). A number of writers have discussed the problem of sample bias from nonresponse and have pre sented economical methods of eliminating the main effects of this bias (2). Jackson and Roth ney (5) have compared the efficiency and cost of mailed questionnaires versus interviews. As one might suspect, more complete information and a larger proportion of response may be obtained by interviewing. However, the interview is many times as expensive. Norman reviewed the research on sample bias in mailed questionnaire studies in 1948 (7). The majority of the investigators in his review found significant differences between respondent and nonrespondent groups. He concluded that the re search amply demonstrated the existence of sam ple bias from nonresponse. His review indicated that responders tended to have higher intelligence test scores, higher scholastic averages, more loy alty or ties to the questionnaire sponsor, a higher educational level, more interest in the topic under investigation, a more favorable report to make, a more successful present status and a rural back ground. Rothney and Mooren (8) reported simi lar results. In a similar study, Mann (6) reported a lack of significant differences between respond rs and nonresponders. Mann's study compared responders and nonresponders on some data from standardized personality instruments and on a number of normative factors such as those dis cussed above. In considering the present status of research on non-response questionnaire bias, it seems to the present writers that the problem is deserving of study. It is still of practical importance, as studies are apparently being done which base conclusions upon partial questionnaire returns. In addition, personality differences on a standard base be tween respondent and nonrespondent groups have been the subject of relatively little study. The present study is designed to make compari sons between groups of responders and nonre sponders on a number of personality characteris tics as measured by the California Psychological Inventory (3).