The simplest method to determine temperature is from the rise of resistance of the winding. You will need an accurate method to measure the resistance hot and cold.

Copper has a coefficient that means a rise of .4% per degC and is near enough over the working range. ( 100degC rise means 40% increase in resistance).

This gives average temperature rise, the centre will be hotter than the outside.

You could wind a few turns of thin copper wire into the area that you want to measure and make your own resistance thermometer but you will have to be crafty to make it work on load with ac volts induced in it.

those little thermal devices you see tied into the windings of the motors you describe are actually thermostats. you cannot read temperature from them. most are designed to open and break the circuit at 80 degree C.  there are some that are rated at 105 degree C.

most are rated at 10 amps (will take up to 10 amp) but open at rated temperature, and close once they cool.

for very accurate temperature sencing, motor manufacturers will use "RTD'S". the common ones are 100ohm platinum, and the resistance rises with temperature. you cannot however run the load through them, they are for temperature measurement only. they can be imbedded directly into the coil or right into the stator coil slot. sometimes they are also imbedded into the bearing housings for bearing temp monitoring.

lastly, they are very expensive.  typically i used to pay $100 for a set of three which i used in large motor repair if the client required them.

those hand held lazer temperature measuring units are quite accurate and should give you the basic picture of what is going on within a few percentage points.

if you really want to go the RTD route, you should be able to order them through an industrial electric supply house or through a motor shop.

An ordinary cheap silicon diode isn't too bad as a temperature sensor.  Feed a constant current through it and the forward voltage drop is fairly close to linear with temperature.  I built a digital thermometer once with one, put the diode into a probe so it was immersible and calibrated it in icewater and boiling water.  It was linear to within 2% or so over that temperature range.  Thermocouples are also good, and cheap if you buy a spool of the wire and make your own.  I think that's what most digital multimeters with temperature ranges use.  See http://www.omega.com for more variety than anyone could possibly use.