The conceptualisation of apraxia has often been linked to the clinicalassessment of the disorder (Rothi et al., 1997). Thus Ideomotor Apraxia(hereafter IMA) has been described as a failure to imitate actions. One of thetasks more widely employed by clinicians is that devised by De Renzi et al.,(1980). The 24 gestures of this test vary according to a 2x2x2 design, the threefactors being (i) gestures composed of a single posture (‘positions’) vs. asequence of postures (‘sequences’); (ii) involving hand and/or arm vs. fingers;(iii) symbolic vs. non-symbolic (see De Renzi and Faglioni, 1999, for detailsabout the items). Specific predictions involving all these gestures were madebased on a model that comprises a Long-Term Memory (LTM), a ‘MotorProgramming’ (MP) component, and a Working Memory (WM) (Baddeley,1986) including a Short-Term Memory component specific for gestures (STM;Smyth and Pendleton, 1988; Rumiati and Tessari, in press) and the SupervisoryAttentional System (SAS, Norman and Shallice, 1986) (Fig. 1).The gesture to be imitated is temporarily held in the STM (while the SASselects, whenever necessary, alternative strategies for the gesture encoding), andthen transferred to the MP system. When a gesture is symbolic, the WM loadcan be reduced by accessing its LTM representation. IMA per sewouldcorrespond to a breakdown at MP level, while damage to the WM componentwould produce an inability to maintain the gesture (STM) or to selectappropriate strategies (SAS) beforethe MP stage.The WM and LTM components have an overall probability wof maintainingand transferring to MP an accurate copy of the input gesture. When a non-