I've been round and round on this. The best solution I have found is to use this traceable glass lab thermometer as my bench mark:
http://www.carolina.com/product/equ...).do?keyword=745416&sortby=bestMatches&page=1
I have several digital thermometers including two type "k" thermocouple models. One of those is a traceable version. The problem is that if you switch thermocouples, it is not longer considered lab calibrated (traceable). You can, however, calibrate it yourself if you have a reliable benchmark to check it against.
Additionally, I've found that using boiling or ice water to check/calibrate a thermometer can be tricky. The ice water must be the consistency of a slurpee. IOW, a fine ice slush with little or no free water in it. You must keep the probe away from the sides and bottom of the container and you must allow plenty of time for the probe to equalize with the slush. This can be as much as five minutes and sometimes more to get it stable within a fraction of a degree. It's similar with boiling water. I've found that sometimes boiling water will read slightly above boiling at the surface, but not at any depth. I'm not sure why this happens, but it may have to do with the atmospheric pressure or something. I think the water may be boiling very near the bottom of the kettle where the heat is applied, but at that depth, the pressure is slightly greater than atmospheric. When the water rises to the surface, it may be slightly superheated and read a bit higher than 212*F. The worst situation is when a thermometer reads correctly at boiling, but not at freezing etc or when is is not consistent throughout the range.
I'm with Bobby on this. It seems you cannot trust any of them absolutely, but the calibrated lab thermometer I linked above is the best thing I have found so far. It checks out at both boiling and freezing and I don't see how it could get out of whack so long as you don't break it completely.
If you want to read something interesting on the subject, google how Mr. Fahrenheit developed his temperature scale and how much trouble he had establishing the freezing and boiling points of water. He was facing much the same dilemma with nothing truly reliable to use as a benchmark. I don't have a link to the info, but some searching should easily turn up the history. Wiki probably has something on the subject. I find that kind of thing fascinating. Fortunately for us brewers, if we can get within a degree or two of the true temperature it's generally good enough. There will likely be more variance throughout the grain bed or fermenter than that anyway, so we are often also chasing a moving target. The Easy-Read traceable is not very expensive and one of the best brewing accessories I have ever purchased. It is also truly easy to read as the name implies. The graduations are big and highlighted by the yellow background. You can easily read it to a tolerance of 1/2 degree F and maybe better than that if you use a magnifying glass. I'm good with being able to read it to +/- one degree without difficulty.
FWIW, the relatively cheap CDN Pro-Accurate digital has proven to be very reliable and accurate. I've been using one for about six years IIRC and it has always been quite accurate (as far as I can tell). I've also dropped it into near boiling water twice and it survived without any damage or loss of accuracy. I don't know how many times you could get away with that, but twice is pretty damn rugged IMO.
I have two clocks as well and you know how that goes!
