I agree with the theory given by the RJ Captain; the Colgan flight probably experienced a tail surface stall. A tailplane stall is an icing phenomenon that rarely occurs yet requires not only counter-intuitive but immediate action by the crew. Worst yet, it tends to occur after selection of flaps or the final landing flap setting, which means the aircraft is at a low altitude. The accident aircraft was about 5 miles from the airport, on final approach and some 1500 feet above the ground during the last communication. That puts it exactly where the landing flap position would be selected by the crew.

A tail surface stall means lift is lost on the horizontal stabilizer. But unlike wing lift, tail lift is in a downward direction. With few exceptions, all airplanes are nose-heavy, meaning the center of gravity (CG) is forward of the center of wing lift. The center of wing lift is where all the lift can be considered to be concentrated at. Think of the center of lift as the fulcrum of a teeter-totter; it is where the teeter totter is supported. Since an airplane's CG is forward of the center of lift, there needs to be a downward force acting on the tail at all times for stable flight. This downward force is refined with elevator control by the pilot flying (or autopilot) to adjust pitch. If the tail stalls, downward lift suddenly ceases and the tail is then violently forced upward and the nose downward. Why? Again, because the CG is forward of the center of lift. It's like two people balanced on a teeter-totter and one person suddenly jumps off. 

Ice accretion is a bad thing on any lifting surface because it disrupts smooth airflow and results in a loss of lift. If enough ice has accumulated on the leading edge of the horizontal stabilizer, the stabilizer will loose downward lift and require additional up elevator or nose-up trim for stable flight. When flaps are lowered, depending on the aircraft, the center of lift will move further aft, further from the CG. This makes the plane even more nose heavy. In addition, the effect of lowering flaps can change the airflow across the tail in a worsening way. This added nose down force along with the change in airflow from the flaps can be enough to cause a horizontal stabilizer that has severe ice on the leading edge, to stall.

I have spent over 250 hours in flight simulators during all of my airline training and have never been shown or have been asked to demonstrate recovery from a tailplane stall. FAA requires steep turns, normal stalls and wind shear procedures as part of any training program, but not tailplane stalls. I really don't know why. I have a feeling this will change.