本研究主要目的為編製適用於小學四、五、六年級學生的多複本閱讀理解測驗，提供一套可瞭解 閱讀理解能力發展情形的工具，也可用於閱讀低成就學生的補救教學成效評估，作為介入反應之 決定。題目型式為測量文章理解的題組。每套複本都有四個題組，每個題組都有一篇文章及10 個閱讀理解題目。四個題組的文章分別是短記敘文、短說明文、長記敘文、長說明文。40 道題目 中有兩題是不計分的，主要用以確認受試者是否認真作答。每個題組中的題目所測量的閱讀理解 成份為字彙觸接、字面理解、摘取大意、及推論理解。在信度考驗方面，四、五、六年級每個複 本的Cronbach's α 信度及折半信度多在 .8 以上。在效度方面，以「國民小學閱讀理解篩選測驗版 本A」為效標，所測得的效標關聯效度介於 .59~ .75 之間。本測驗除了提供百分等級常模外，也 提供能力值轉換常模，以便於繪出學生的閱讀理解能力成長曲線及計算成長係數。本研究也對於 教育應用上的建議與限制提供說明。

The purpose of this study was to develop a set of equivalent tests for reading comprehension progress monitoring for grades four to six. The reading comprehension equivalent tests can also be used as tools for evaluating instructional effects of remedial intervention. In this test, there were six alternate-forms for each grade. Four testlets were included in each alternate-form. Each testlet had one passage and ten test items. The four passages in one alternate-form included one short narrative, one short expository, one long narrative, and one long expository texts. There were 200-300 characters in short passages, and 400-500 characters in long passages. The test items were designed to measure lexical access, literal comprehension, summarization, and inferential comprehension. There were forty items in total in one alternate-form. Regarding reliability, the Cronbach's α coefficients and split-half coefficients of all alternate-forms were mostly above .80. Using “Reading Comprehension Screening Test for Elementary School Students-Version A” as criterion, the criterion validity coefficient were .67 (p < .001), .59 (p < .001), .75 (p < .001) for grades 4, 5, and 6 respectively. In addition, based on the data collected from six time-points in 2011 school year, the growth model analysis revealed that the reading comprehension abilities of the fourth, fifth, and sixth graders increased from the beginning to the end in one school year. In order for teachers to monitor students’ progress, this test provided not only the norm of percentile rank, but also the norm of ability value (theta score). Using the ability values between two time-points, a growth coefficient could be calculated. Several suggestions and one limitation for educational application were provided.