Previous research has shown that prism adaptation (prism adaptation) can ameliorate several symptoms of spatial neglect after right-hemisphere damage. But the mechanisms behind this remain unclear. Recently we reported that prisms may increase leftward awareness for neglect in a task using chimeric visual objects, despite apparently not affecting awareness in a task using chimeric emotional faces (Sarri et al., 2006). Here we explored potential reasons for this apparent discrepancy in outcome, by testing further whether the lack of a prism effect on the chimeric face task task could be explained by: i) the specific category of stimuli used (faces as opposed to objects); ii) the affective nature of the stimuli; and/or iii) the particular task implemented, with the chimeric face task requiring forced-choice judgements of lateral 'preference' between pairs of identical, but left/right mirror-reversed chimeric face tasks (as opposed to identification for the chimeric object task). We replicated our previous pattern of no impact of prisms on the emotional chimeric face task here in a new series of patients, while also similarly finding no beneficial impact on another lateral 'preference' measure that used non-face non-emotional stimuli, namely greyscale gradients. By contrast, we found the usual beneficial impact of prism adaptation (prism adaptation) on some conventional measures of neglect, and improvements for at least some patients in a different face task, requiring explicit discrimination of the chimeric or non-chimeric nature of face stimuli. The new findings indicate that prism therapy does not alter spatial biases in neglect as revealed by 'lateral preference tasks' that have no right or wrong answer (requiring forced-choice judgements on left/right mirror-reversed stimuli), regardless of whether these employ face or non-face stimuli. But our data also show that prism therapy can beneficially modulate some aspects of visual awareness in spatial neglect not only for objects, but also for face stimuli, in some cases.