It is generally accepted that the target and the source in a verbal metaphors are not, in a given context, reversible. For instance, in a given situation one cannot simply reverse "my surgeon is a butcher" into "my butcher is a surgeon" (of course in a quite different context "my butcher is a surgeon" may be a viable metaphor). Even though visuals don't have a grammar (which helps in assessing target and source in many verbal metaphors), pictorial/visual metaphors also do NOT allow the two terms to be reversed. In this respect, though not in others, Carroll's (1994, 1996) proposals concerning the nature and identifiability of pictorial metaphors are wrong. Carroll also is wrong, in my view (see Forceville 1996) that prototypical visual/pictorial metaphors are of the "homospatially noncompossible" variety -- that is, what Forceville 1996 labels the MP2 (or: hybrid) subtype. This is just one type among others. Finally, the paper makes suggestions for how metaphor theory needs to be adapted to account for cinematic metaphors in which music and sound play a role.