Starting from the generate-recognise model of free recall, we will be addressing two points. First, we will discuss whether the enactment effect, i.e. better memory of self-performed actions (SPTs) compared to watched experimenter-performed actions (EPT) and to verbal tasks (VT) is due to enhanced relational and/or item information of SPTs. Second, we will propose that at least two different types of relational information must be distinguished: categorical and episodic (order) information. We will demonstrate (a) that categorical relational information is used equally in SPTs, EPTs and VTs, (b) that episodic relational information is more efficient in pure EPT than in pure SPT conditions, (c) that this advantage is lost when the two encoding conditions are mixed, and (d) that episodic relational information can be effectively used with short but not with long lists. This variability of episodic relational information determines whether SPT shows higher memory performances than EPT. Finally, we conclude that the SPT effect is based on item information because the SPT advantage is observed even though relational information is not enhanced. Key words: Enactment effect, item and relational information, generaterecognise models of free recall.