The rise in the number of households with a cell phone but no land line represents a problem that telephone interviewing has to face up to if it is to offer nationally representative samples. RDD samples have historically not included mobile phone numbers, and this has led to concern about coverage bias. In the UK cell-only households are currently about 8% of all households, but half of these are households where the oldest member is aged 18-24. There is thus a considerable risk of bias, both overall and within this group. This potentially affects all surveys, but concern has been expressed most publicly about possible bias in election surveys. This was certainly the case in the US in the run-up to the 2004 presidential election. Incorporating mobile numbers into RDD samples poses a number of problems, concerning both sampling and interviewing. It is fairly easy to develop sampling and weighting algorithms to cope with this, but a far bigger unknown is whether it is harder to interview a mobile phone sample, whether it is more expensive, and whether the results are biased in any way. To test the acceptability of interviewing via cell phones, GfK NOP Social Research conducted an experiment in the last UK Genera! Election, running a cell phone survey alongside a traditional telephone survey. The paper discusses the results of the experiment, in terms of the relative accuracy of the two surveys, and the differences in costs and response rates. The paper also examines some of the main differences - both demographic and attitudinal - between mobile-only households and those with landlines. The paper concludes with a discussion of the implications for future surveys.