Let X-n = Sigma(i=1)(infinity) alpha(i) epsilon(n-i), where the epsilo
n(i) are i.i.d. with mean 0 and finite second moment and the ai are ei
ther summable or regularly varying with index epsilon (- 1, - 1/2). Th
e sequence {X-n} has short memory in the former case and long memory i
n the latter. For a large class of functions K, a new approach is prop
osed to develop both central (root n rate) and noncentral (non-root n
rate) limit theorems for S-N = Sigma(n=1)(N) [K(X-n) - EK(X-n)]. Speci
fically, we show that in the short-memory case the central limit theor
em holds for SN and in the long-memory case, S-N can be decomposed int
o two asymptotically uncorrelated parts that follow a central limit an
d a noncentral limit theorem, respectively. Further we write the nonce
ntral part as an expansion of uncorrelated components that follow nonc
entral limit theorems. Connections with the usual Hermite expansion in
the Gaussian setting are also explored.