This study compared the results of a usability inspection conducted under two separate conditions: An explicit concurrent think-aloud that required explanations and silent working. 12 student analysts inspected two travel websites thinking-aloud and working in silence to produce a set of problem predictions. Overall, the silent working condition produced more initial predictions, but the think-aloud condition yielded a greater proportion of accurate predictions as revealed by falsification testing. The analysts used a range of problem discovery methods with system searching being favoured by the silent working condition and the more active, goal playing discovery method in the think-aloud condition. Thinking-aloud was also associated with a broader spread of knowledge resources.