本研究是以自我發問解釋作為閱讀的策略，兩班選修普通化學的師院非數理系學生，分為實驗組和控制組，實驗組學生先在上課時練習自我發問的閱讀方法。正式施測時，兩組學生都作有聲思考的閱讀，再比較實驗組和控制組學生的閱讀推論。並進一步探討全體學生的先前知識、閱讀推論與閱讀的學習成就之關係。除了上述的統計結果之外，並分析學生的心智模式與迷失概念；比較兩組學生的閱讀過程類型；以及學生對於整篇文章所作的口語資料之分析。
分析學生的閱讀推論，無論是實驗組或控制組，每一位學生都有與理解文意或精緻化有關的推論，其中利用相關知識而產生的推論，控制組的百分比多於實驗組，而其他類型的推論，則沒有顯著的差異。
對於推論以外的表達，實驗組22位學生中，有19位學生口語資料中有關於這一類的表達，佔76％。控制組20位學生中，有10位學生有關於這一類的表達，佔50％。其中提出疑問或討論的問題的百分比，實驗組多於控制組。
實驗組與控制組學生對於正式施測時學習成就的測驗成績，其t考驗的t值沒有達到alpha=.05的統計之顯著水準。但對於保留測驗而言，如果用單側考驗時(假設實驗組的成績優於控制組)，則 P=.03，達到統計之顯著水準，所以實驗組正式施測的學習成就測驗成績，與控制組沒有顯著的差異；但實驗組的保留測驗成續，優於控制組。
因為先前知識與閱讀推論，對閱讀的學習成就均有顯著的影響，所以進一步以先前知識與各類型的推論等變項，與閱讀的學習成就進行逐步迴歸分析，結果先前知識與統整推論兩個變項的F值，分別為8.184、8.626，均達到統計的顯著水準。其中學生的先前知識對學習成就有較大的預測力，可以解釋全部變異量的17％，而統整的推論可以解釋全部變異量的14％，合計為31％。
對於心智模式的分析，不論實驗組或控制組學生屬於比較完整型、完整型等兩種心智模式的比較多。但比較實驗組與控制組含有迷思概念型的人數，分別為18％與45％，控制組則較多，可見實驗組的心智模式比較少迷思概念。統計結果，沒有迷失概念的學生，學習成就的測驗成績比較好。而有無迷思概念與邏輯型推論有顯著的相關存在。
至於文章的性質與結構：科學性理論的說明，主要是邏輯型推論和統整型的推論；不是科學性理論的說明時，主要是相關知識的推論和邏輯型推論。而學生多半在讀完一個完整的句子時作推論。在文章開始和結束時，閱讀推論可能比其他段較多。但文章比較長時，後面幾段的閱讀推論，可能比其他段較少。
最後本研究對科學教材中環境教材的編寫、科學課程中融入的環境教材，以短文為宜，避免冗長。對環境教材的教學方法，建議參考實驗組的方式，讓學生練習自己提出問題，並訓練學生對科學性的課文，提出統整的推論與問題，有助於學生對課文的了解，與提高學習成就，並幫助學生解決關鍵的觀念問題。對未來研究的建議希望能增加對環境文章的興趣之測驗，探討對環境文章的興趣之變因，與閱讀推論的關係。

The study investigated the reading strategy of self-questioning and self-explaining used by teacher college students in reading environmental texts. The students were divided into two groups: experimental and control. The students of the experimental group practiced self-questioning in the class. The control group students were given no specific method to follow. While reading the environmental texts, both groups were required to think aloud. After this they were tested on the contents of the environmental text. Inferences of the two groups were compared using a Mann-Whitney U Test. In addition, the relationship of the students' background knowledge, inferences, and learning achievements was investigated by stepwise regression. Following the protocols of interviewing, students' mental models and misconceptions were analyzed, types of reading processes compared, and protocol data of students in reading the whole text summarized.
After analyzing the students' inferences, it was found that every student of both groups made inferences of understanding or elaboration of the text. Concerning inferences of background knowledge, the control group's frequency was greater than the experimental group. As for questions or problems, the experimental group raised more than the control group.
A t-test of the two groups in immediate examinations showed no differences. In delayed examinations, however, the t-tests showed the experimental group to be significantly better than the control group.
The results of stepwise regression indicated that background knowledge can explain 17% of the variance, while the integrate inferences can explain 14% of the variance. Total explanation of the variance was 31%.
On the analysis of mental modes, the misconceptions of students of experiment group was 18%, while the control group was 45%.
As for the types of reading processes, students of the experimental group preferred to think aloud when reading. The Control group students preferred to think aloud after reading either one paragraph or the whole text.
For scientific text, the main inferences that students use were logic and integrate inferences. For non-scientific text they used background and logic inferences. It was also found that reading inferences were more often used at the beginning and end of a text.
The study suggests that environmental texts should be as concise as possible. It demonstrates that self-questioning is a useful strategy for helping students to understand key problems and issues. The study will stimulate additional research on the relationship of interest and inference in students involved in reading environmental texts.