# reference help needed on a fact about poles of L-functions

Suppose $\pi$ and $\rho$ are cuspidal automorphic representations on $GL(n)$ and $GL(m)$ respectively. Then the L-function $L(s,\pi \times \rho)$ has a pole iff and $m=n$ and $\pi$ is isomorphic to the contragradient of $\rho$ by some twist. Does anyone know some reference containing the proof of this fact?

I checked Rankin-Selberg convolution paper by Jacquet-P.S-Shalika. It mentioned this result and said the proof would appear somewhere.

Many thanks.

-
I'm also curious, but not so curious right now to pore through the literature. It seems that the J-PS-S paper was followed up by archimidean results by J (Contemp. Math.) and J-S (in the PS 60th bday volume). Perhaps these finish it off? Also, Cogdell's article "Analytic theory of L-functions for GL(n)" (on his webpage) has an exposition of this result. I don't know which of these or other articles is the definitive reference for this important result. –  Marty Oct 3 '10 at 18:41