Previous evidence suggests that a wave with mixed mode data collection in an otherwise face-to-face panel survey will achieve a  lower response rate than other waves. But until now there has been no evidence as to whether the response rate can be expected to recover subsequently. In other words, is the hit on response rate temporary or permanent? In this paper we address that question by examining the relationship between response at wave 3 of the  Understanding Society Innovation Panel and the data collection protocol used at wave 2, when an experiment was carried 

out with mixed modes.