Articulatory motion is a skilled sequential movement in which a phonemic string is converted into continuous movements in the articulatory coordinate system. In generating articulatory movements, the phonemic target or task is often defined as a compact region of either articulatory, vocal-tract geometric, or acoustic space. In this paper, statistical analysis of the phonemic targets obtained from the empirical articulatory and acoustic data is described. The target distribution shows different variants in each space for different classes of sounds. The variance of the target is examined for different conditions of syllable, stress, and utterance speed. Also examined is which articulatory, geometric, or acoustic space is appropriate for specifying the target in forming the articulatory movements using the trajectory formation model with a stochastic target representation [M. Honda and T. Kaburagi, ICSLP 1, 179--184 (1994)].