www.phidot.org

just posted, both on Gary's site, and on this mirror:

http://www.phidot.org/software/mark/download/

Major changes include:

An option was added under the Adjustments menu of the Results
Browser to specify the effective sample size for computing AICc and QAICc.
The reason for this adjustment is the different types of parameters in the
robust design capture-recapture and also robust design occupancy data types.
Parameters related to primary occasions can be considered to have different
sample sizes than parameters related to secondary occasions within primary
occasions. The effective sample size is now stored in the data file so that
all models in the results browser will have used the same effective sample
size to compute AICc or QAICc. The default value is still the value
computed by the MARK numerical routine.

Effective sample size calculations for the robust design occupancy
data types were changed to be the sum of the number of sites sampled in each
primary occasion, rather than just the number of sites sampled in the first
primary occasion. The effective sample size for the regular occupancy data
type is still the number of sites visited.

A bug with the specification of the Burnham Jolly-Seber data type
was fixed. However, this model is still not recommended for general use.
Rather, use the Pradel, Link-Barker, or POPAN data types. The Burnham
Jolly-Seber data type often does not converge to the maximum likelihood
estimates, whereas the other models usually do. Note, however, that the
Pradel lambda (λ) data type does not enforce the constraint that lambda(i)
> >= phi(i), so you can get nonsensical answers from this parameterization.
Hence, I generally recommend use of the Pradel recruitment (f) data type.