Research into survey methodology has revealed that respondents answer differently depending on the way in which questions are presented. To date there have been few empirical attempts to relate these differences to respondents’ cognition. To investigate this, we used eye tracking in three different survey modes: a computer-based mode, a pen-and-paper questionnaire and a face-to-face interview with show-cards. This novel method of obtaining cognitive measures across modes means that we are able to investigate attention and cognitive processing by coding which questions and answers are looked at the most in the different settings. The results confirm the attentional biases linked to some responses, such as a bias to the first and last response options in a list, as well as revealing differences in how response options were inspected across modes. These results are discussed and it is proposed that cognitive science can play a significant role in survey methodology.