> Whether a function whose natural domain consists of lists should coerce > a scalar into a one-element list is purely a matter of convenience. > > That the built-in GeometricMean does do that whereas the documentation's > example geom does not is perhaps just a sloppy oversight in constructing > that example.
GeometricMean[x + y]
During evaluation of In:= GeometricMean::vecmat1: Argument x+y is neither a non-empty vector nor a non-empty matrix. >>