Numerous studies have found better memory for multimodal than unimodal stimuli. In these studies, however, multimodal stimuli consist not only of multiple modalities, but also of more varied information than unimodal. Therefore, in the present study, we investigated encoding variability as an explanation for the multisensory benefit. Written words were studied together with two different orienting questions that promoted processing of same modality (both visual or both auditory) or different modality information (one visual and one auditory). In Experiment 1, recognition memory did not differ between constant modality and varied modality conditions. In Experiment 2, we replicated this effect with items that were repeated at a lag and we found an advantage of any type of encoding variability (within and between modality) compared to a condition in which the same orienting question was repeated. In Experiments 3 and 4, these findings were replicated when there was a day delay between study and test. We conclude that we did find an encoding variability benefit on memory, but no multisensory benefit. This conclusion challenges the sensory integration explanation and provides support for encoding variability as an explanation of the multisensory benefit.