To the good advice given elsewhere in the thread, I would add, "check your speedometer" (and thus your odometer). These things have some intrinsic error, which might either stack up atop or the other potential source of error: hanges you have made by tire/wheel size.
Knowing how much and in which direction your speedometer is fibbing can save you from tickets, too!
A few runs past one of those radar "sleds" at a steady and easily readable speed should suffice, although finding a stopwatch, or a wristwatch with a sweep second hand, and timing yourself through a measured mile or a "speedometer test section" on the freeway is the classic way of doing it.
You can easily see how recordkeeping can combine with the aspects of fillup discipline mentioned by others to let you partial out highway from city from all-around mileage.
If all you want is a good horseback guess, the fillup practices are probably the most important aspect, unless your speedometer is way off.
Also, the study should be conducted over a span of time longer than a couple of tankfuls but shorter than a season (differences in driving habits as the weather changes, and in some areas differences in gasoline formulation, can affect mileage). Cheers,
--Joe