Recent developments in procedures for conducting pairwise multiple comparisons of means prompted an empirical investigation of several competing techniques. Monte Carlo results revealed that the newer multistage sequential procedures maintain their familywise Type I error probabilities while exhibiting power that is superior to the traditional competitors. Of all procedures examined, the modified Peritz (1970) procedure (Seaman, Levin, Serlin, & Franke, 1990) is generally the most powerful according to all definitionsof power. At the same time, when computational ease and convenience are taken into consideration, Hayter's (1986) procedure should be regarded as a viable alternative. Beyond pairwise comparisons of means, the versatile Holm (1979) procedure and its modifications (Shaffer, 1986) are very attractive insofar as they represent simple, yet powerful, data-analytic tools for behavioral researchers.Researchers who are investigating differences among three or more experimental groups are often interested in the pairwise differences between group means. The choice of a multiplecomparison procedure (MCP) with which to assess these differ-

The multiple-comparison procedure originally proposed by R. A. Fisher (1935) for the 1-way analysis of variance context has several desirable properties when K(ihe number of groups) is equal to 3. In this article, the logic of the procedure is described in conjunction with those properties. A discussion follows of how the Fisher procedure can be similarly applied in a number of other K = 3 (and, more generally, 2-degree-of-freedom) hypothesis-testing situations. Finally, the Fisher logic is combined with recent sequential applications of the Bonferroni inequality to illustrate the utility and versatility of that combination for the applied researcher.

Methods for determining whether 2 means are practically equivalent are discussed.Existing equivalency-testing procedures are reviewed and compared. An equivalence confidence interval is proposed and compared with the traditional confidence interval for a mean difference. The use of this new interval is described for both confirmatory and exploratory research and is shown to fit within the context of good-enough methods. Adaptations and extensions of the interval are proposed.For the past 20 years, biostatisticians and biometricians have been developing methods for assessing the equivalence of the effects of two drug formulations on various blood concentration profile variables (e.g.,

No abstract

This study investigated the effects of 3 different writing tasks (descriptive, narrative, and expository) and 3 different writing prompts (bare, vocabulary, and prose model) on 937 writing samples culled from 330 novice learners enrolled in 15 classes of Levels 1 and 2 high school French. In order to assess the quality, fluency, syntactic complexity, and accuracy of the writing samples, the researchers employed 4 evaluation methods: holistic scoring, length of product, mean length of T-units, and percentage of correct T-units. Results indicate that the descriptive task was the easiest and the expository task the most difficult. The prose model prompts produced the highest mean scores, and the bare prompts produced the lowest mean scores. Based on these findings, the researchers question whether the description of a novice writer in the ACTFL Proficiency Guidelines (1986) should be used as a blueprint for curriculum development and textbook construction for secondary novice foreign language learners.AT WHAT POINT IN FOREIGN LANGUAGE (FL) instruction should writing be introduced, and what can reasonably be expected of novice writers? The ACTFL Proficiency Guidelines (1986), which delineate expectations for four levels of FL writers, may be seen as a response to this question and, in this regard, have had a major influence on establishing curricular goals for writing since their publication in 1986. Nevertheless, a survey of the literature of FL writing reveals little of the empirical research necessary to establish a solid basis for decisions with respect to the FL writing curriculum.A number of articles on FL writing have appeared over the last 2 most have dealt either with theoretical considerations or with the application of theory to instruction. Few researchers have directly studied writing instruction in FL classes. 1 Consequently, practitioners have had to depend on research from writing in English as a first language (L1; e.g., Emig, 1971;Hunt, 1965) or English as a Second Language (ESL; e.g., Raimes, 1979;Zamel, 1976Zamel, , 1985Zamel, , 1987 when developing instructional procedures for teaching second language (L2) writing. In fact, Silva (1990) found that new approaches based upon L1 writing research are sometimes adopted in wholesale fashion and accepted uncritically in L2 writing instruction. Not only has writing been underrepresented in FL/L2 research, secondary school learners have been virtually ignored in such research. Krapels (1990) found that the few studies conducted in DENISE PAIGE WAY Dorchester School District Two 102 Greenwave Boulevard