Several general population surveys have shown that offering individuals a choice of responding to a survey via web or mail produces lower response rates than offering mail as the sole response mode. In this paper we present results of two experiments designed to assess the effect of offering a choice between mail and web response when only mail contacts are used, as is necessary in general public populations. The population for these experiments (undergraduate students at Washington State University) was chosen to overcome a limitation of previous studies, i.e. the fact that some members did not have web access. In support of prior theory and research, this research suggests that offering a choice produces a slightly lower response rate than that obtained when only offering a mail response option. At the very least, there is no evidence to support the common assumption that offering a choice will improve response. We also find that offering only a web response option produces a significantly lower response rate than either the mail-only or choice option. Both experiments also include additional treatments that reveal whether the negative effects of choice and offering web-only can be overcome through combining postal contacts, token financial incentives, and strategically timed emails to support web response. Results of these experiments have significant, but different, implications for surveys of general populations in which email contacts cannot be used (mail-only may be a better strategy) and situations in which postal and email contacts can both be used (a web-only response option is likely to be better).