A study demonstrated the utility of item analyses to investigate which items function well or poorly in a second language reading recall protocol instrument. Data were drawn from a larger study of 56 learners of German as a second language at various proficiency levels. Pausal units of scored recall protocols were analyzed using both classical item analysis and the Sato Caution Index. The assumptions of classical local independence and unidimensionality were found to be largely fulfilled in this analysis. Pausal units did not fulfill the noninvasiveness condition, but this was not found detrimental to the investigation. Results concerning item difficulty, point biserial correlations, the discrimination index, reliability, item weighting, and use of the Sato caution index are summarized. Results of the Sato caution analysis were found to corroborate those of classical analysis, indicating that item analyses of dichotomously scored recall protocol units can be applied and yield interpretable results. It is not suggested that the Sato Caution Index supplant classical item analyses, but that it be used whenever integrative language tests using summated item scores are used to make theoretical, instructional, or evaluative decisions. A 30-item bibliography is included. One text used and its item analysis results are appended. (MSE)