Visual extinction, a parietal syndrome in which patients exhibit perceptual impairments when two objects are simultaneously presented in the visual field, is reduced when objects are correctly positioned for action, indicating that action helps patients' visual attention. Similarly, healthy individuals make faster action decisions on object pairs that appear in left/right standard co-location for actions in comparison to object pairs that appear in a mirror location, a phenomenon called the paired-object affordance effect. However, the neural locus of such effect remains debated and may be related to the activity of ventral or dorsal brain regions. The present fMRI study aims at determining the neural substrates of the paired-object affordance effect. Fourteen right-handed participants made decisions about semantically related (i.e. thematically related and co-manipulated) and unrelated object pairs. Pairs were either positioned in a standard location for a right-handed action (with the active object – lid – in the right visual hemifield, and the passive object – pan – in the left visual hemifield), or in the reverse location. Behavioral results showed a suppression of the observed cost of correctly positioning related pairs for action when performing action decisions (deciding if the two objects are usually used together), but not when performing contextual decisions (deciding if the two objects are typically found in the kitchen). Anterior regions of the dorsal stream (e.g. supplementary motor area) responded to inadequate object co-positioning for action, but only when the perceptual task required action decisions. In the ventral cortex, the left lateral occipital complex showed increased activation for objects correctly positioned for action in all conditions except when neither task demands nor object relatedness was relevant for action. Thus, fMRI results demonstrated a joint contribution of ventral and dorsal cortical streams to the paired-affordance effect. They further suggest that this contribution may depend on contextual situations and task demands, in line with flexible views of affordance evocation.