One of the most perplexing issues being faced by survey researchers in some countries is the inability to use email addresses for surveying random samples of the general public. In this paper recent results from a program of research aimed at developing a mixed-mode approach to solving this problem will be discussed. Results from four experiments conducted by the authors will be reported in which postal address-based samples of households, arguably the best household sample frame available in the U.S. are contacted by mail and asked to respond over the web, with mail questionnaires being sent to non-respondents. Experimental results from these studies will be summarized. They suggest that about half of the households contacted will respond to these surveys, with two-thirds of those responses being received over the Internet. Our purpose in this paper will be to provide an overview of implementation techniques that are effective, and not effective, in achieving these results and outline additional research that is needed in the U.S., European countries, and elsewhere, to learn the extent to which these mixed-modes can provide a promising paradigm for designing and implementing mostly-web surveys of households, as a potential replacement for random digit dialing telephone surveys.