自2006年開始，台灣已陸續參加三屆國際閱讀素養調查研究（Progress in International Reading Literacy Stud，PIRLS），分別是PIRLS 2006、2011及2016。本研究以PIRLS教師問卷中，針對教師課室內閱讀教學的共同題為資料，探討台灣教師課室內閱讀教學活動頻率的變化情形，並以線性混合模型分析不同教學頻率對學生不同閱讀表現的主要效果，以及不同教學頻率與調查年度的交互效果。三次調查共有482位台灣教師完成教師問卷，及13,208位台灣四年級學生完成閱讀理解測驗。每位學生都有五種閱讀成績，分別是總成績、故事體成績、說明文成績、直接理解成績、與詮釋理解成績。三次調查中教師課室內閱讀教學的問卷共同題共有五大題，分別是：（1）教師以何種方式組織學生、（2）教師使用哪些閱讀材料、（3）教師進行哪些閱讀教學、（4）教師教導哪些閱讀策略、（5）學生閱讀後，教師請學生做哪些活動。結果發現：（1）相較於2006與2011的調查年度，2016年時有更多教師經常使用小組方式進行閱讀教學。但是不論教師偏好使用全班教學、小組教學或個別化教學，都不影響學生五種閱讀成績的表現；（2）相較於2006年，2011與2016這兩個調查年度，都有更多教師開始使用長篇小說進行閱讀教學，此外，2016年也有更多的教師更常使用「非小說」類的材料進行閱讀教學。而且，當教師愈常使用長篇小說進行閱讀教學，學生五種閱讀成績的表現就會愈好，且與調查年度無交互作用；（3）每天（或幾乎每天）會讓學生自行默讀的教師比例，隨著調查年度呈遞增趨勢。而且，當教師愈常讓學生自行默讀，學生的故事體成績與說明文成績就愈好，且與調查年度無交互作用。此外，會經常教學生略讀策略的教師比例，也隨調查年度而呈現遞增的現象。教師經常使用其它教學活動的比例，則出現2016與2006的高於2011年的現象。但教師使用這些教學活動的頻率不影響學生閱讀成績；（4）經常教學生就閱讀過的材料提出理由支持自己的看法，或就閱讀的材料進行預測，以及描述閱讀材料的風格與結構的教師比例，隨著調查年度呈遞增趨勢。教師經常進行其它閱讀策略教學的比例，在三次調查年度間無明顯變化趨勢。且教師進行閱讀策略教學的頻率，並不影響學生的任何一種閱讀表現。值得注意的是，2016年的資料發現「歸納與統整」的閱讀策略教學，對學生說明文成績有單純主要效果，且是教師愈常進行「歸納與統整」的閱讀策略教學，學生說明文成績就愈好；（5）經常要求學生就閱讀過的材料口頭回答或口頭概述的教師比例，也隨調查年度遞移逐漸增加。當教師愈常讓學生進行口頭概述或口頭回答問題，學生五種閱讀成績的表現就愈好，且與調查年度無交互作用。2016年時經常要求學生就閱讀過的材料和同學討論的教師比例，明顯高於2006與2011年的教師比例。經常在閱讀結束後以紙筆形式，要求學生就讀過的材料寫一些回應或進行小考的教師比例，在三次調查年度間無明顯變化趨勢。而且這些教學活動的頻率都不影響學生閱讀成績。

The reading performance of fourth-grade students in Taiwan has shown a trend of improvement since PIRLS 2006 to PIRLS 2016. Theoretically, because of the teachers' day-to-day reading instruction in classroom, teachers have inevitably taken a key role in affecting the reading performance of the students. So far, little is known on relationship between the reading instruction and student reading performance in Taiwan. Therefore, the current study focused on the impact of activities of teacher reading instruction in Taiwan on the reading outcomes of students. By analyzing the ten-year PIRLS data, the present study can not only understand which aspects of teachers' factors play key role on students reading performance, but also serve as an important empirical evidence for formulating future reading education policies. The linear mixed model was used to analyze the main effects of different teaching frequencies on students' different reading performance. And the interaction between different teaching frequencies and the survey year. In three surveys, a total of 482 Taiwanese teachers completed the teacher questionnaire and 13,208 fourth-grade students in Taiwan completed the reading comprehension test. Each student has five reading grades, which are total score, story score, explanatory text score, direct understanding of grades, and interpretation of comprehension scores. There are five major questions in the questionnaires for teachers' indoor reading teaching in three surveys, namely: (1) how often teachers use different types of grouping for reading instruction, (2) how frequently teacher asked their students to read various types of literary and informational texts, (3) how often frequently teacher teaching language skills, (4) how often frequently teacher asked their students to practice reading strategies in teachers' reading instruction (5) how often frequently teacher asked their students to do various types of responses after students reading a text. It was found that teachers preferred the ＂small grouping type＂ than ＂whole-class＂ in PIRLS 2016 survey, but there were no effects of grouping type on students' reading performance. However, when teacher more frequently asked students: to read longer fiction books with chapters, silent reading, and oral response on reading materials, the better performance of the reading scores of the students. It is worth noting that the frequency of reading skills or strategies teaching in teachers' reading instruction does not affect students' reading performance. Only in PIRLS 2016, when teachers more frequently asked students to make generalization and draw inferences, the better performance of the expository reading scores of the students. These findings are discussed in relation to reading education policy and the literature on instructional contexts.