有理式克漏字測驗與篇章結構測驗關係研究

Abstract

克漏字解題過程相關研究指出，透過事先嚴謹辨析篇章的關連性，克漏字測驗為一種可以檢測特定語言能力如銜接性(cohesion)的測驗工具(Bachman, 1982, 1985; Fotos, 1991; Stubbs& Tucker, 1974)。這種特別發展出來的克漏字測驗通常稱為有理式克漏字(the rational cloze)，有別於傳統上以特定每幾個字為單位來刪去字詞所產生的克漏字測驗。自民國七十一年起，有理式克漏字測驗包含在大學聯考英文科測驗裡(現在稱為大學指考英文科考試)。應用有理式測驗方法，這樣的評量可以產生以句構或是篇章為層面的試題。然而，自民國九十一年起，大學指定科目考試英文科測驗開始採用篇章結構此ㄧ新測驗形式。篇章結構測驗類似有理式克漏字測驗一樣帶有填補空缺訊息的性質。這種測驗從一篇短文當中擷取五個完整的句子作為需要填補闕漏訊息的空格。受試者必須從這五個選項當中做選擇來恢復文章的原貌。根據Halliday和Hasan (1976)的語言分析，銜接性提供聽者或是讀者填補文本中遺漏訊息所需的連續性。因為文章中的空白處會產生訊息的不連續性，我們可以假設銜接性(尤其是句子間的銜接性)對於篇章結構測驗的表現非常關鍵。然而，單就篇章結構測驗的分數也許只能表示受試者答題的良莠程度。整個答題過程中牽涉了哪些銜接性例如指涉性(reference)，連結性(conjunction)和詞彙銜接性(lexical cohesion)等也許無法直接反映出來。以Halliday和Hasan (1976)的理論為基礎，發展一個涵括三種銜接性次類別(如指涉性，連結性和詞彙銜接性)的有理式克漏字測驗來詮釋篇章結構測驗分數之意義是有需要的。 在理論上假設句子間的銜接性對上述兩種測驗評量的表現都很重要的同時，這兩種測驗具有什麼程度的相同性以及有理式克漏字的次類測驗如何預測篇章結構測驗的表現，就是有待研究的問題了。本研究旨在探討上述議題以及研究這兩個具有填補空缺訊息性質的測驗之間的關係。一共有三百五十四位來自台灣中部某高中的學生參與研究。在重複試驗的研究設計下，先在約五十分鐘的時間進行篇章結構測驗。之後在另外一個時段由同樣的受試者進行有理式克漏字測驗。兩測驗的間隔大約是兩星期。三個階段的資料分析分別著眼在(1)兩測驗的信度和有理式克漏字測驗的效度，(2)兩測驗的相同性檢驗，以及(3)篇章結構測驗的回歸分析。研究結果指出，兩測驗在統計上來說是不相等的測驗形式，而詞彙銜接性在篇章結構測驗的表現上最具有預測力。根據實驗結果，可以對英語教學與語言測驗做出建議。Considerable research on cloze procedures has shown that through meticulous, a priori identification of textual relationships, a cloze test can serve as an adequate measure yielding items that elicit specific language skills such as knowledge of cohesion (Bachman, 1982, 1985; Fotos, 1991; Stubbs& Tucker, 1974). This specifically constructed test is the product of the rational cloze procedure, as distinguished from the traditional, fixed-ratio deletion of every nth word. Since 1982, the rational cloze test format has been incorporated in the Joint College Entrance Examination (a term currently replaced by the Department Required English Test, DRET). With the rational approach, such a test can sample a variety of items at the syntactic and discoursal levels. However, the Discourse Structure Test (DST), a new test format with a gap-filling, cloze-like nature similar to that of the rational cloze test, has been adopted as an individual component in the DRET since 2002. The DST yields five blanks by extracting five complete sentences from a short passage. The test-takers have to choose among these alternatives to restore the text. Based on Halliday and Hasan’s (1976) linguistic analysis, cohesion provides the “continuity” that enables a listener or reader to restore the “missing pieces” in a text (p. 299). It can be hypothesized that knowledge of cohesion (especially intersentential cohesion) is critical for testees’ performance on the DST, in which each blank generates discontinuity in the text. Nonetheless, scores on the DST alone may only reveal how satisfactorily the test-takers tackle the test. Which aspects of cohesion (e.g., reference, conjunction or lexical cohesion) involved in the overall problem-solving process may not be projected straightforwardly. To interpret the DST scores on the theoretical ground of Halliday and Hasan’s (1976) paradigm, a rational cloze test (RCT) comprising three subtests assessing different dimensions of cohesion (i.e., reference, conjunction and lexical cohesion) would be desirable. With intersentential cohesion argued to be critical for successful closure of the DST and the cohesion-based RCT, questions arose as to the extent to which the two tests were equivalent and how the RCT subtests predicted performance on the DST. The present study aimed to address these issues and investigate the relationship between the two gap-filling tests. A total of 354 students at a senior high school in central Taiwan participated. In a repeated-measure design, the DST was administered first in a session lasting about fifty minutes. The RCT was distributed to the same participants in another session. The interval between the two sessions was approximately two weeks. For data analyses, three phases were performed on 1) the reliability of both tests and validity of the RCT; 2) test equating of both tests; and 3) regression of the DST on the RCT and its subtests. The results showed that the two test formats were statistically inequivalent and that lexical cohesion functioned as the most influential predictor on students’ performance on the DST. Based on research findings, pedagogical implications can be drawn for English instruction and language testing.