As long-term average speech spectrum (LTASS) and its loudness relationship across frequency have been used to derive hearing aid prescriptive formulae or to determine frequency response targets for hearing aids, it is important to understand how background noise affects LTASS. This study examined the effect of the type and level of background noise on LTASS. Thirty speakers read aloud an article in quiet; and in babble-noise, traffic-noise, and restaurant-noise presented at 50 dB SPL, 65 dB SPL and 80 dB SPL. Speech samples were collected and analyzed using the protocol proposed by Byrne et al. (1994). The LTASS in quiet was comparable to the "universal" LTASS (Byrne et al., 1994). In noise, speech level increased by 3 dB for every 10 dB increase in noise level, regardless of the type or spectral shape of the noise; and the spectral peak of the LTASS shifted from low- to mid-frequency. There were significant but small intensity differences between the male and female speech spectra but it is still feasible to combine them to form a composite spectrum. Implications of these findings for speech recognition and hearing aid fitting will be discussed.