Recently, the mass levels of mesons have been successfully treated with the Schrodinger equations on the basis of the composite modeJ.l>,z> The validity of the nonrelativistic approximation was investigated by Nagasaki 3 > using the Dirac equation. In this note, the same problem is studied from a slightly different stand point. The model of Dirac particles interacting thr6ugh potential is assumed here; too. It is concluded that if there is the mass· difference between the constituents, the mass of a composite particle cannot be smaller than M=(mr+m2)/2, where the m 1 and m2 are the masses of the constituents, except for the 1 So state.The energy of a composite· particle is given by (Ho+ L:Oa V,_,(r)), Or= Ict,m ct,C2l, O!J'= -p~l)p~2)a(l>acz>+p~l>p~2)aO>aC2), o..~.= -pillp~2>+acr>aC2l and Qp= -p~1 !p~2>. The wave function 1/r(r) decomposed into the eigenfunction of the total angular momentum J, and parity P. Each eigenfunction has eight components.From Eq.(1), we can obtain two sets of simultaneous differential equations of first order for the eight amplitudes which are functions of the relative distance r only, corresponding parity states A) P= (-)J and B) P= (-)J+r. In those sets, the potentials are contained in terms of Ua(r) = Vs(r) -6V!l'(r) + Vp(r) ±4(Vv-(r)-V..;.(r)), U2±(r) = Vs(r)-Vp(r) ±2(Vv-(r) + V..~.(r)), Ua (r) = Vs(r) +2Vp(r) + Vp(r).Of the eight amplitudes, only the four amplitudes, written as GeJ(r) (j=O, 1, 2, 3), appear with the differential operators and the rests, written as aat(r) (j=O, 1, 2, 3), do not. Therefore, only the a,(r)'s are required to be continuous for the whole range of positive r. 4 J The aa(r)'s are represented in terms of a.(r)'s with (E± U(r)) in the denominator. For example, in the case of A), one of the amplitudes of spin' singlet part, aa0, is given in terms of the other Cleo as follows:( 3) where A= (mr-m2) /2.In order that the 1/r(r) 1s normalizable, the amplitudes may not have singular points. Therefore, for example, the potential Uz+(r) must satisfy -E< Uz+(r) for the whole range of r. Applying the similar discussions to all the aa(r)'s, we obtain -EJ< U2+(r) ro or r(ikr) ·for r>ro and by jJ(pr) for r