Several factors influencing dream recall frequency (DRF) have been identified, but some remain poorly understood. One way to study DRF is to compare cognitive processes in low and high dream recallers (LR and HR). According to the arousal-retrieval model, long-term memory encoding of a dream requires wakefulness while its multisensory short-term memory is still alive. Previous studies showed contradictory results concerning short-term memory differences between LR and HR. It has also been found that extreme dream recall frequencies are associated with different electrophysiological traits related to attentional processes. However, to date, there is no evidence for attentional differences between LR and HR at the behavioural level. To further investigate attention and working memory in HR and LR, we used a newly-developed challenging paradigm, called MEMAT: it allows to study selective attention and working memory interaction, during memory encoding of non-verbal auditory stimuli. We manipulated the difficulties of the distractor to ignore and of the memory task. The performance of the two groups were not differentially impacted by working memory load. However, HR were slower and less accurate in presence of hard, rather than easy, toignore distractor, while LR were much less impacted by the distractor difficulty. Therefore, we show behavioural evidence towards less resistance to hard-to-ignore distractors in HR. Using a challenging task, we show for the first time attentional differences between HR and LR at the behavioural level. The impact of auditory attention and working memory on dream recall is discussed.