We examined the relationship between C-test and criterion-test scores to better understand the C-test construct. We meta-analyzed Pearson's r coefficients and computed summary effects for subgroups, where criterion construct was the subgrouping variable. We summarized the evidence from 239 C-test studies, published in English. Any correlational study about foreign-language education was eligible. If a C-test was at least one passage in length, its information was recorded. Studies that did not include statistical information needed for meta-analysis were excluded, but other information was used to summarize aspects of the C-test domain.

Summary effects indicate that C-test scores correlate most strongly with general language proficiency scores. However, there were too few studies in some criterion subgroups for analysis. Confidence intervals for subgroup summary effects overlapped across meta-analyses. Additional work is needed to understand the C-test construct. We recommend creating a C-test meta-analysis bank and adding information from study reports not published in English to create a cumulative C-test domain (Cumming, 2014). Our study involves the most comprehensive study-retrieval process to date; there was little evidence of publication bias in results.

我们检查了C测验和标准测验分数之间的关系，以更好地理解C测验的结构。我们对Pearson的r系数进行了荟萃分析，并计算了子组的汇总效果，其中标准构造是子组变量。我们总结了以英文发表的239个C检验研究的证据。任何有关外语教育的相关研究均符合资格。如果C测验的长度至少为一个传代，则记录其信息。不包括荟萃分析所需的统计信息的研究被排除在外，但其他信息被用来概括C检验领域的各个方面。

摘要效果表明，C测验分数与一般语言能力分数之间的相关性最强。但是，在某些标准子组中进行分析的研究太少了。亚组汇总效应的置信区间在荟萃分析中重叠。需要更多的工作来理解C检验的构造。我们建议创建一个C检验荟萃分析库，并添加未以英文发布的研究报告中的信息，以创建一个累积的C检验域（Cumming，2014年）。我们的研究涉及迄今为止最全面的研究检索过程。几乎没有证据表明出版偏向于结果。