This paper tests two hypotheses about how well five different methods—the Question Understanding AID (QUAID), the Survey Quality Predictor (SQP), expert review, the Questionnaire Appraisal System (QAS), and cognitive interviews—predict problems (as measured by missing data, behavior codes, and response latency) that occur in typical survey conditions. We find partial support for both the complementary methods hypothesis (using the evaluations methods together will yield the best prediction of problems) and the test environment hypothesis (the more directly the evaluation method observes the response process, the better it will predict problems). In addition, we find evidence that the methods perform somewhat differently for items measuring subjective as opposed to objective characteristics.