The effective response is calculated in nonlinear composite wires and
strips which are modeled as two-dimensional random conductance network
s with lateral size L and width delta much less than L. We consider a
two-component nonlinear conductance network which consists of two type
s of conductors. The first component is assumed to be nonlinear and ob
eys a current-voltage (I-V) characteristic of the form I = sigma(1)V+c
hi(1)V(3), while the second component is linear with I=sigma(2)V, wher
e sigma(1) and sigma(2) are the linear conductances and chi(1) is the
nonlinear conductance. We invoke a renormalization-group (RG) analysis
to rescale the strip repeatedly by small-cell transformations to obta
in a chain of nonlinear conductors, for which exact formulas of the ef
fective linear response sigma(e) and nonlinear response chi(e) are ava
ilable. We calculate sigma(e) and chi(e) as a function of the volume f
raction for various conductance ratios and examine the dependence on d
elta. We observe large enhancements in the nonlinear response under ap
propriate conditions, as well as interesting crossover from one- to tw
o-dimensional behaviors as delta increases. Numerical simulations are
performed to verify the RG calculations. Scaling exponents are determi
ned in the RG and compared with available estimates; good agreements a
re found. Possible generalizations of the present investigation are di
scussed.