To what extent do we process the nondominant object in a morphed figure? Evidence from a picture–word interference task