We simulate a nonlinearized Kalman , Kalman and a modified Kalman (linear) filter for suppressing a narrowband Gaussian interference in direct sequence spread spectrum receiver and examine the suitability of Signal-to-Noise Ratio (SNR) of the test statistic as a measure of performance of the receiver. We consider Gaussian autoregressive interference with a peaked spectrum and the three cases: small processing gain (PG) and short pseudonoise (PN) sequence, small PG and long PN sequence, and moderate PG and PN sequence. Based on the simulations, we conclude that for the two cases corresponding to small processing gain, if the thermal noise variance is small and the interference is strong, the Gaussian approximation to the test statistic does not yield the correct BER for any of the receivers. For small PG and short PN sequence, even though the SNR corresponding to nonlinear filter is significantly higher than the SNRs of the two linear filters, the BER of the non-linear is higher than that of the linear receivers. SNR is not a useful measure in these situations.