| 摘要: | 本研究蒐集2003年至2012年，近10年國內「合作學習」相關文獻，共彙整121篇(127個比較篇數)，以後設分析法(meta-analysis)探究「合作學習」教學策略在整體、認知、情意態度、動作技能、社會行為等向度之學習成效。所探討之調節變項為：出版形式、教學對象、教學地區、教學科目、教學時間、教學情境、認知層次等七項，分別於各向度探討對學習成效之影響。
本研究以Hedges與Olkin(1985) 固定效果模式之統計技術，篩選「組間差異」為「合作學習」之比較組，以計算單純「合作學習」加權平均效果量，並搭配森林圖及效果量計算的共同語言(簡稱CL) 輔助分析學習成效；不重疊量數U3作為效果量的考驗力分析；同質性考驗為異質則以標準化殘差法刪除異質性研究；以漏斗圖檢視出版偏誤，繼之以Nfs.05值與容忍值校正出版偏誤的影響；最後，採用類別模式尋找影響實驗處理成效之調節變項。經統計分析後研究結論如下：
壹、 學習成效
整體成效及認知學習、情意態度、動作技能、社會行為等五個向度之加權平均效果量分別為：0.4522、0.4443、0.4639、0.3611、0.4457且均達顯著水準，表示實驗組(合作學習教學法)學習成效優於控制組(傳統教學法)，達Cohen中等程度之效果。
貳、 調節變項
除「出版形式」不是各向度之調節變項外，動作技能向度因同質性檢定為同質，故不探討調節變項。其餘各向度之調節變項判定結果如下：
一、「教學對象」是整體成效、認知學習向度之調節變項，且國小及大專生之調節效果優於國中及高中職。
二、「教學地區」是整體成效、認知學習、情意態度、社會行為向度之調節變項， 且中部、南部地區之調節效果優於北部地區。
三、「教學科目」是認知學習之調節變項，且非升學科目之調節效果優於升學科目。
四、「教學時間」是整體成效、認知學習、社會行為向度之調節變項。在整體與認知向度部分，以一至四週之教學時間最具調節效果；而社會行為向度則以八週以上最具調節效果。
五、「教學情境」是整體、認知學習成效之調節變項，且以面對面合作學習之調節效果優於資訊融入合作學習與虛擬教室合作學習。
六、「認知層次」是認知學習之調節變項，且以高層次認知能力之調節效果 優於基礎認知層次。
上述調節變項雖與整體研究文獻的效果量變化確實有關，但尚有其他潛在調節變項的影響。本研究根據上述研究結論，提供建議給教育政策推動及實務教學工作者與學生，作為相關規畫之依據。
This study collects nearly ten years of relevant literature on “cooperative learning” from 2003 to 2012 to compile 121 articles (127 comparison articles), and utilizes meta-analysis to explore the effectiveness of “cooperative learning” teaching strategy in holistic, cognitive, affective attitude, motor skill, social behavior dimensions. The seven moderators being explored are: publishing format, teaching audience, teaching area, teaching subjects, teaching time, teaching context, and cognitive levels, to explore the impact of each dimension on learning effectiveness.
This study uses Hedges and Olkin's Fixed-Effect Method (1985) to screen out “group difference” as the comparison group of “cooperative learning”, and calculates the simple weighted average of “cooperative learning”. They are then paired with forest plot and results of common language (CL) to assist the effectiveness of learning, and used the non-overlapping number U3 as the statistical power analysis of effectiveness. Standardized residuals are used to delete heterogeneous research if there is heterogeneity in homogeneity test. Funnel chart is used to inspect bias and then correct the effect of bias by Nfs.05 value and tolerance. Finally, the grouping mode is used to search the moderators influencing the effectiveness of experiment. After statistical analysis, the research conclusions are as follows:
1. Learning effectiveness
The weighted average values for the five dimensions of holistic, cognitive, affective attitude, motor skill, and social behavior are 0.4522, 0.4443, 0.4639, 0.3611, and 0.4457, respectively. They all reach the level of significance, showing that the experimental group (cooperative learning teaching) is better than the control group (traditional teaching), and reaching Cohen’s medium effect.
2. Moderator
Other than that “publishing format” is not moderators for each dimension, motor skill is determined as homogeneous in homogeneity test and its moderators will not be explored. The moderators for the remaining dimensions are shown as follows:
A. “Teaching audience” is the moderator of holistic effectiveness and cognitive learning, and the moderating effect on elementary school and college students are better than junior and senior high schools.
B. “Teaching area” is the moderator of holistic effectiveness, cognitive learning, affective attitude, and social behavior, and the moderating effects on the central and southern region are better than that of the northern region.
C. “Teaching subjects” is the moderator for cognitive learning, and the moderating effect on non-advancing courses is better than academic advancing courses.
D. “Teaching time” is the moderator for holistic effectiveness, cognitive learning, and social behavior. In the holistic and cognitive dimension, the teaching time from one to four weeks has the most moderating effect, while eight weeks and more have the most moderating effect on social behavior dimension.
E. “Teaching context” is the moderator for holistic effectiveness,cognitive learning and the moderating effect of face-to-face cooperative learning is better than integrating information into cooperative and virtual classroom learning.
F. “cognitive levels” is the moderator for cognitive effectiveness, and the moder- ating effect of complex-level cognitive is better than basic-level cognitive .
Although the aforementioned moderators are correlated to the effectiveness change of the overall research literature, the explaining power is not strong enough. There are still other effects of latent moderators. This study uses the above research conclusion to provide recommendation for education policy promoters, education practitioners and students to be used as the reference for relevant planning. |
