Heater host tee, radiator hose tee, or drain plug fitting. If the drain plug is threaded 1/4"NPT or 3/8"NPT, that's one of the better places.
Where possible, the coolant temperature should be measured somewhere near the back of the block, since that's about the hottest part of the engine. The back of the cylinder head is best, the back of the block is second best. I'm sure it's possible to drill and tap the cylinder head to accept a sensor, but I haven't been able to experiment with that as yet.
I'm considering installing a drain valve in that location, but I'm working with RENIX blocks that have the sensor at the rear of the head - and I can always add the late-model thermostat housing if I need another sensor or a tap.
When installing a temperature sensor of any sort, it is important that the sensor tip is inserted into a near-constant a flow of the "fluid" as possible - whether that fluid is gaseous or liquid. While installing a "T" fitting will work find for a pressure sensor, it will eventually result in "falsing" for a temperature sensor - unless you can make it a flow loop with an inlet and an outlet. This is true for any case where you are trying to read temperature - for liquid or gas.
It is also possible to braze a bung into the radiator tank - if you do this, use the "hot" tank - where the upper hose goes into the radiator. Place the bung as close to the hose fitting as possible, since this will give a more accurate reading. Since the radiator is copper, you can easily braze or hard solder a copper, brass, or aluminum bung into the tank - once you have stripped the paint.
There is also the small fact that the temperature gage for the 1997 4.0 may not respond in the same manner to temperature as the 1995 and eariler units - it may not be a true "thermistor." Until this can be verified, I would also advise against tapping the existing signal to feed a gage. However, if you can verify that the later temperature sensor responds to temperature variation along the same curve, it is possible to split the signal - but I'd probably also advise the use of diodes to separate the signals.
Oh - it's possible to get reducing T fitting that "reduce on run" (one arms small than the other) and "reduce on branch" (base of T smaller than arms) - and you can nearly always one up/down one size, with going up/down two sizes possible with a little looking. I'd suggest doing this before trying to bush a branch downwards, since you want to make a plumbing system with as few joins as possible (reducing thereby the potential for leaks.) As long as the run of the T has a constant flow, and the sensor tip can reach into that flow, you should be able to get a true reading.
In fact - a likely place to put a "T" would be in the heater supply line from the water pump, since I have been summarily informed (subject to confirmation, of course! :laugh3: I'm a natural sceptic) that the heater core is subject to constant supply flow. The heater outlet on the water pump housing is threaded 3/8"NPT, and I commonly stick a longer pipe nipple into there and an elbow - since having a rubber hose running right behind a spinning pully makes me nervous... See my posts on this subject elsewhere.
5-90