Experiences with wireless carriers in the United States seem heavily dependent on location and personal travel routes, which is why anecdotal evidence about coverage is generally not very useful. Comparing their coverage maps, it looks clear that AT&T technically has more network coverage:

AT&T: http://www.wireless.att.com/coverageviewer/#?type=voice

T-Mobile: http://www.t-mobile.com/coverage/pcc.aspx

If I lived in an urban area (where T-Mobile is likely to have already deployed both standard 3G and HSDPA+ coverage, and AT&T's network has a higher than average proportion of heavy iPhone users), my experiences with either carrier are likely to be different than if I lived and stayed in the suburbs or rural areas (where T-Mobile is more likely to either have EDGE coverage or nothing at all).

I travel around ND,SD, and MN - T-Mobile is the worst here. It has no coverage once you leave a city and sometimes not even in them. AT&T and Verizon were amazing compared to T-Mobile. Sprint does ok in parts, but has some big dead spots.

So what? I'm on AT&T and have a 3G connection everywhere but inside concrete bunkers in my city. But that says nothing about coverage. You can say coverage is about HAVING "bars"; better coverage refers to having more area with bars. Coverage of an entire carrier can't be determined by a single case.

Additionally, measuring "bars" is useless to the measurement of coverage (1) because the bars on one phone could mean something totally different than the bars on another phone (see iPhone reception miscalculation) and (2) because how many bars you have, or signal strength or reception, is determined in part by how close you are to the signal source. So what if you have perfect reception if you live right next to a cell tower?