I worked on chipmunks and used open-field tests, but never distinguished between different classes of freezing behavior. I am not aware of any such distinction, but maybe lab studies have been more precise (I am a behavioral ecologist and conduct my behavioral tests on wild animals in natural settings). My understanding is that you would distinguish between a freezing behavior associated with different components of a stress response and a freezing behavior associated with a lack of response to the stimuli?
I think that properly validated measures of stress reactivity (cortisol levels, defecation rates, heart rate and heart rate variability, body temperature) would help you disentangle different freezing behaviors. These would be associated with the stress reactivity measurements in different ways.
Thank you Pier-Oliver. I'm working in a research using Fear Conditioning, and the freezing variable is commonly used as a measure of conditioned learning. So, we want to divide the group in high/low freezing subjects: we want to realize genetic expression analysis in BDNF and Creb and compare the results between both groups. I've already divided them by 1 SD and quartiles, but I'm interested to know if somebody else has realized this kind of division before, and in what way.
Do you have a within-subject design, i.e. freezing data and gene expression of each individual animal? If yes, why then classifying the animals? Simply correlate the data by taking the individual data. That would be the strongest dataset! I guess there something like this from the Lessmann group in Magdeburg, Germany (correlation of individual BDNF levels from wildtype and BDNF +/- mice with their fear conditioning).
If classification, then try different classifications (3 third, 4 quarters, everything above mean + SED and below mean - SED,...) and simply check what looks best. In general classification criteria are somehow arbitrary.
Hi Mijail, here is a paper I can recommend: Pham, J.; Cabrera, S.M.; Sanchis-Segura, C.; Wood, M.A. (2009). Automated scoring of fear-related behavior using EthoVision software. Journal of Neuroscience Methods, 178, 323–326. It is based on video tracking using our EthoVision system, but the method should also work with other systems.