R help - Apr 2004 - Spatial autocorrelation in linear regression models

The scope of my problem relates to bird count data.

Point counts (standing and listening at one point) were conducted along 
transects, every 500 m.

In literature, I've found reference to calculating an autocorrelate for 
each observation where if a species was found in an adjacent site, it is 
assigned a value of 1, otherwise it has a value of 0. This is simple. 
What I don't understand is how this can be included in a model without 
it having the opposite effect and becoming a predictor?

I'm using stepAIC for my model selection.

T
-- 
Trevor Wiens
twiens at interbaun.com