Four MA translation students and five expert translators translated two texts from Danish into English and two from English into Danish. In each language direction, one task was performed while thinking aloud and one without thinking aloud. All tasks were performed on a computer, and all keystrokes were logged with Translog. The aim was to determine what influence, if any, the think-aloud condition might have on translation speed, on the amount of revision undertaken, and on pause-defined segmentation. It was expected that thinking aloud might slow down the process of translation, but that it would have no effect on revision or on segmentation. In both groups, significant effects were found on speed. Thinking aloud delayed translation by about 25%. No significant effects on revision were found. However, contrary to expectation, significant effects on segmentation were discovered. The think-aloud condition forced both groups of translators to process text in smaller segments.