The op asked for a self-contained device that must be inexpensive. 
Measuring G to about 5% would do that. It could be done with 
accelerometers or with an inexpensive pendulum (can be liquid - f.ex. 
mercury switch or liquid level sensor, or a free-swiging pendulum with 
2dof).

http://hyperphysics.phy-astr.gsu.edu/hbase/pendl.html

shows the formula for the period of a large amplitude pendulum. A 2cm 
pendulum will have a period of about 0.3 seconds and at 15 degrees swing 
deviates by less than 0.5% from the small amplitude pendulum period.

With 0.1G acceleration (~1m/sec^2) the period changes by more than 2%. 
The pendulum measures absolute acceleration since it is free-swinging in 
2dof. So it cannot differentiate between acceleration and deceleration 
(but and-ing its output with the brake signal would solve that - for 
example if the power comes from the brake light wire). It could be 
sensed with a hall or optical sensor. It would 'bob' all the time due to 
the car moving and people in it moving. It could provide the basic blink 
rate (~3Hz) and this could be multiplied up as needed.

A more expensive solution is to use mems accelerometers and measure 
sqrt(Gvert^2 + Gx^2 + Gy^2). The sqrt can be lost for math 
simplification. Then detecting 0.01G should be possible using 
accelerometers that are stable to 0.5% or so (S/N 46dB ?) after 
filtering. No ?

Peter
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist