460 



Fishery Bulletin 104(3) 



Table 2 



Comparison between three likelihood-based models for mark and recapture of horseshoe crabs tLimulus polyphemus). Each 

 model allowed a differed pattern of variation in recapture probabilities. The models are listed in order from least to most complex. 

 In model 1, the recapture probabilities were set to be constant for both release cohorts (prespawning season period and prepeak- 

 spawning period) and for all three recapture occasions. Model 2 allowed recapture probabilities to be time-specific, but equal for 

 release cohorts. Model 3 allowed recapture to be time and cohort specific. Lower values for Akaikie information criteria (AIC) 

 and /lAIC indicate a better model fit; 4AIC = is the best fitting model. Likelihood ratio test (LRT) compares two nested models, 

 and a significant P-value indicates the more complex model is supported. 



Likelihood ratio test 



Model 



Variation in recapture 

 probabilities 



AIC 



4AIC 



Models compared 



P-value 



Constant 



Time-specific 



Time and cohort-specific 



-348512.08 

 -350225.72 

 -350220.83 



1713 

 

 5 



Models 1 and 2 

 Models 2 and 3 



<0.0001 



0.77 



fects the accuracy of abundance estimates. Therefore, it 

 was important to evaluate thoroughly whether tagged 

 animals had different catchability than untagged ani- 

 mals. We tested for a temporary delay in spawning by 

 comparing relative risk of recapture among the two 

 release periods and by fitting a likelihood model that 



incorporated separate recapture probabilities for each 

 release cohort (Table 2). Evidence did not support the 

 hypothesis that initial capture and tagging temporarily 

 affected spawning behavior (x2=2.03, df =2, P=0.36). 

 The difference between the two release periods in the 

 risk of being recaptured during peak spawning was 



