The concept of a commutative convolution measure algebra (c.c.m.a.), due to Taylor [24], has proved to be very important in the theory of harmonic analysis on locally compact abelian groups. Taylor, in effect, defines a c.c.m.a, to be the predual of a commutative W*-algebra d endowed with a cocommutative comultiplication c; i.e., a W*-morphism c: ~'-->~' ® d which obeys the diagrams dual to those characterizing a commutative associative multiplication. C.c.m.a.'s include LI(G) and M(G) for locally compact abelian groups G, and they share many of the important features of both algebras. If ~' is a commutative Hopf W*-algebra, i.e., if .~ is a W*-algebra together with a cocommutative comultiplication c, we shall call the predual a commutative predual algebra. Significant examples of such preduals are the Fourier algebra A(G) and the Fourier-Stieltjes algebra B(G) of a locally compact group G. These were introduced by Eymard [8], and have also been studied by Walter [26], among others. B(G) is by definition the linear combinations of (continuous) positive definite functions on G, and A(G) is a certain closed ideal in B(G). If G is abelian with dual group (~, then BIG ) is isomorphic to M((~) (Bochner's theorem), and A(G) is isomorphic to LI(G) under the inverse Fourier transform. Thus, for G non-abelian, B(G) and A(G) may be regarded as generalizations of M((J) and LI(~J), respectively. Our work here is motivated by the question of symmetry for M(G) when G is abelian. It is well-known (cf. [22], Chapter 5) that, unless G is discrete, M(G) is asymmetric; i.e., there is a complex homomorphism h of M(G) and a measure/~ in M(G) with h(/~*)4: h(#), where * is the involution in M(G). Furthermore, if G is not discrete, then the Hewitt-Kakutani phenomenon holds; namely, G contains a compact independent perfect set Q such that every bounded linear functional of norm less than or equal to 1 on the continuous measures supported on Qw - Q extends to a complex homomorphism of M(G). By analogy one might expect that B(G) is asymmetric if G is a noncompact locally compact group, and, that in fact there are always involution closed subpredual spaces L~ B(G) such that every element in L* of norm less than or equal to one extends to a complex homomorphism of B(G). However, if G is the group of Euclidean motions, then B(G) is symmetric (see Theorem 1.3), so such is not the case. The paper is organized as follows. In Section l we state our main theorem (Theorem 1.3), which characterizes those groups G for which B(G) is symmetric among those compactly generated or Lie groups which are a compact extension of a normal abelian subgroup. Examples are given to show why the hypothesis