Two experiments examined Ward, Avons, and Melling's (2005) proposition that the serial position function is task, rather than modality, dependent. Specifically, they proposed that for backward testing the 2-alternative forced-choice (2AFC) recognition paradigm is characterised by single-item recency irrespective of the modality of the stimulus presentation. In Experiment 1 the same nonword sequences, presented both visually or auditorily, produced qualitatively equivalent serial position functions with 2AFC testing. Forward testing produced a flat serial position function, while backward testing produced two-item recency in the absence of primacy. In order to rule out the possibility that the serial position functions for visual stimuli were the product of sub-vocal rehearsal, Experiment 2 employed articulatory suppression during the presentation phase. Serial position function equivalence was again observed together with a modest impairment in overall recognition rates. Taken together, these data are consistent with the Ward et al. proposition and further support the existence of a visual memory that can facilitate storage of visual-verbal material (e.g. Logie, Della Sella, Wynn, & Baddeley, 2000). However, the observation of two-item recency contradicts the original duplex account of single-item recency traditionally observed for backwards recognition testing of visual stimuli (Phillips & Christie, 1977).