Students' sensory preferences (SP), that is, the ways or channels they prefer to receive the information, have been proposed as personal factors influencing science learning. Sorne studies relate improvement in learning to the adequacy of instructional resources to individual SP. A widely used instrument to determine SP is the VARK questionnaire. However, the individual SP has to be reliable to be considered useful to improve instruction. There are not enough studies on the reliability of this questionnaire. This study aimed to assess the reliability (temporal stability) of SP, as measured by the VARK instrument. Also, the reliability of the student's subjective perception of his/her personal SP was evaluated. A mixed methodology was used, combining techniques for obtaining both quantitative and qualitative information. The sample was made of 582 male and female students, from 7th to 11th grades. The questionnaire was administered twice to a subsample of students, with a few months' delay. Besides, a small group of students was interviewed. In the analysis of the results, using chi-square tests (X2) and a repeated measure MANCOVA, no massive permanent changes in the group distribution of SP were found along with secondary education. This would be useful for teachers to adapt learning materials to their students' preferences. The individual SP was stable enough under a permissive criterion, but not when a stricter standard was used. Besides, the reliability of students' self-perception of SP was low.