Charlie99:
With the wide range of "normal" temperature, the more reasonable (and more historically accurate) explanation is that it was rounded off to 37C, and the apparent precision of 98.6F is just an artifact of the units conversion.
Actual average oral temp is closer to 98.2F / 36.8C. See Wikipedia "Thermoregulation".
My "duh" moment on this was about 15 years ago while living in Japan. I went out and got a little digital thermometer, reading in Celsius of course, to check the fever on my daughter and converted 98.6F to C. Only then did I realize that 98.6F is a nice round 37C.
How it really started:
History
The Fahrenheit temperature scale became popular through its use on the first reliable, commercially-available, mercury-in-glass thermometers. Daniel Gabriel Fahrenheit manufactured such thermometers in Amsterdam from about 1717 until his death in 1736. The scale we know as the Fahrenheit scale was the last of three scales he used.
As the zero point on his scale Fahrenheit chose the temperature of a bath of ice melting in a solution of common salt, a standard 18th century way of getting a low temperature in the laboratory (and in the kitchen, as in an old-fashioned ice cream churn). He set 32 degrees as the temperature of ice melting in water. For a consistent, reproducible high point he chose the temperature of the blood of a healthy person (his wife), which he measured in the armpit and called 96 degrees. (The number arises from beginning with a scale of 12 intervals, like a one-foot ruler, and then doubling the number of steps as instruments become more precise, making 24 intervals, then 48, and finally 96.)1
Fahrenheits successors used the boiling point of water to calibrate their thermometers, which they set at 212 degrees in order to retain the size of Fahrenheit's degree. Fahrenheit himself did make thermometers that could read as high as 600°F, but the largest market was for thermometers for weather observations, and on these the part of the thermometer from 130°F to 212°F would be wasted.
So, the 37C = 98.6 is more of an accident then planed.
Also, Wikipedia cannot used as a cited source in most upper level papers as the data is not confirmed and anyone can change it. However, Wikipedia can be used as a pointer to "good" sources which can be cited