The experiment analyzed serial position curves in recall, global recognition (comparing probes to whole lists) and local recognition (comparing probes to specific items in a list). Input order, output order, and the spatial order to be memorized were deconfounded by presenting and probing items in random order in different spatial positions. Primacy and recency effects over input position were observed for all three tasks. Only primacy emerged over output position. Spatial position affected only recall and local recognition but not global recognition accuracy. Latency data provided additional information, sometimes deviating from the patterns of accuracy. The results support an attentional gradient as one source of the primacy effect. They are compatible with input and output interference or decay, but are difficult to explain by a temporal distinctiveness account. No support was found for response suppression and edge effects as mechanisms to explain recency.