Due to several previous replications of increased internal consistency reliability associated with increases in the number of alternatives per choice point, alternate forms test-retest reliability of self-ratings on a noncognitive scale was studied. Increased stability reliability associated with increases in the number of alternatives per choice point is discussed in terms of past research, theory and practical concerns. Recently researchers have indicated renewed interest in effects upon psychometric scales due to varying the number of alternatives per choice point (3). Reviews addressing the effect of number of alternatives (options) per choice point (item) on scale reliability have suggested that the effect varies with stimulus conditions (I).* Replications of dependable increases in estimates of internal consistency have been associated with increases in the number of alternatives per choice point with Likert-type item formats under varied conditions (2).2 For a 12-item author-constructed Attitudes Toward Measurement Scale in the Likert-type format with 3, 5, 7, or 9 alternatives, alpha coefficients of .65 (n = 47), .72 (n = 44), .73 (n = 47), and .81 (n = 46), respectively, were ~roduced by the total of 184 graduate students of a measurement course.' For a 37-item author-constructed Attitudes Toward Students' Rights Scale with 3, 5, standard 5, 7, or 9 alternatives in the Likert-type format, alpha coefficients of .65 (n = 113), .72 ( n = 113), .74 ( n = 113), .78 (n= 113), and .69 (n = 113) were produced.2 For the "standard 5"format in this latter study all alternatives were verbally anchored; otherwise for both studies, only end-and mid-points were verbally anchored.In another study, the effect under consideration was studied with five forms of the Marlowe-Crowne Social Desirability scale having either 2, 4, 6, 8, or 10 alternatives per choice point. In that study, it was concluded that, for the even-numbered response format (with no midpoint) alpha coefficients increased with the number of alternatives, from 2 through 8, replicating the previous studies involving odd numbers of alternatives (4). Alternate forms with time interval estimates of reliability were studied in the present experiment to replicate Oaster's original finding that, up to some practical point, increases in the number of alternatives per choice point increase scale reliability.Reliability of self-ratings on 10 forms of the 16-item Texas Social Behavior Inventory (Forms A and B) with either 3, 5, standard 5, 7, or 9 alternatives per choice point was studied for 255 undergraduate students.' Participants were requested from classes for which instructors agreed to participate including 3 business, 3 education, 2 English, 4 psychology, 2 sociology classes, and 1 religion class in a private, church-related liberal arts college in the Midwest. The five forms were randomly distributed within classes, each with a 10-option answer sheet attached which was also unobtrusively coded for form number in column 60 with liquid paper. Those who complete...

investigations of serial learning have demonstrated superior mastery by subjects of initially presented and final material over centrally located material. Isolating an item in the sequence tends to increase the probability that item will be learned. Both effects were examined using as stimulus material extended, meaningful discourse with 108 graduate students. Two counterbalanced forms of a tape-recorded lecture resulted in significant serial section and isolation effects favoring initial over middle and final material, final over middle material, and isolated over nonisolated material. Item counterbalancing coupled with nonsignificant forms and interaction effects indicate results are not due to spurious inherent item difficulty differences.

No abstract

No abstract

The three studies presented are concerned with three kinds of multiple-choice reading comprehension questions that may be answered correctly at a higher-than-chance level when they are administered without the accompanying passage. An attempt was made to demonstrate that each of these kinds of high risk item does not necessarily lead to passage dependence invalidity. One study succeeded whereas two failed. The net results of these failures to show that the threats to validity are not necessarily damning was interpreted as evidence of the seriousness of the threats.EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT 1980, 40 READING comprehension test questions that can be answered correctly at a higher-than-chance level when administered without the passage(s) upon which they are based seem to be of dubious validity as measures of reading comprehension. However, the ability of examinees to answer items with the passage(s) out (PO) does not demonstrate that examinees, when taking the test with the passage(s) in (PI), actually do answer such items without reading and mentally processing the information on which the items are based. For this reason, the answerability of questions with their PO &dquo;is but a threat to valid measurement and not proof of invalidity&dquo; (Tuinman, 1973(Tuinman, -1974). Yet this threat has been taken by prominent writers (e.g., Pyrczak, 1975; Traxler, 1965, p. 792; Tuinman, 1970) as a very serious indictment of the invalidity of items. Pyrczak (1976, p. 919) listed three kinds of items that examinees may be able to answer with the PO with greater-than-chance success. These are: (1) items that tap general knowledge rather than new information contained in the passage, (2) interrelated questions that clue at University of Manitoba Libraries on June 18, 2015 epm.sagepub.com Downloaded from