An evaluation of the t-unit as the basic unit of analysis of second language development in written form argues that despite the t-unit's advantages, the sentence-based analysis is superior for examining syntactic complexity in at least three ways: (1) it better characterizes learner knowledge; (2) it facilitates comparison across learners and across stages of second language development; and (3) it encourages dialogue between classroom teachers and researchers by providing a common unit of analysis. The t-unit completely eliminates full-clause coordination from any quantitative description of syntactic development, while sentence analysis allows such knowledge to be represented. In addition, the coordination index in sentence analysis allows comparison of the shift from the use of coordination by beginning learners to the use of embedding by more advanced learners, and finally, the use of the sentence, a learner-centered unit, gives both teachers and researchers access to data in the same form. It is suggested that whether used to measure average sentence length, average length of error-free sentences, or clauses per sentence, sentence-based analyses are useful in the analysis of written second language discourse. (MSE)