Stoner et al.: An assessment of discard mortality for two Alaskan crab species, based on reflex impairment 
341 
condition. Models were fitted by the method of maxi- 
mum likelihood for binary data (i.e., dead or alive) with 
the regression module of Systat 12 (SYSTAT Software, 
Inc., San Jose, CA) (Peduzzi et al., 1980). A backward 
stepwise approach was used to determine the most 
parsimonious model for mortality, with an alpha value 
of 0.15 to remove a variable from the full model. This 
model for mortality was described by 
Log e (p /(I - p)) - a + p'x, 
Where p - proportion of y - 1; 
y = 1 if crab was determined to be dead, and 0 
if alive; 
a = intercept; 
j8' = model coefficients; and 
x = the model matrix of explanatory variables. 
The maximum likelihood estimates of mortality (p) were 
calculated as 
p = e <a+ P' x) / 1 + g(a+/3'*>_ 
Initially, the data for each species were split randomly 
into equal halves, one representing a learning set and 
the other a test set. The most parsimonious logistic 
model was developed with the learning set and validated 
with the test set. After cross-validation, a final model 
was fitted to the entire data set. Finally, the logistic 
model for each species was used to develop a surface plot 
showing the probability of mortality based upon fixed 
values for the key observations of crab condition. 
Results 
Laboratory results 
The six reflex actions identified for testing with C. bairdi 
(Table 1) were highly reliable and consistent among 
individuals. Strong responses in leg flare, leg retrac- 
tion, eye retraction, and mouth close were observed with 
every individual every time they were tested. Weak chela 
closure was observed just once, and the kick response to 
lifting the abdominal flap was weak in eight instances 
(seven females) and missing entirely in one test (also 
female). The weak responses generally occurred in the 
same individuals in duplicate trials. 
Field evaluation of reflex impairment 
Crabs collected in the recapture nets demonstrated a 
wide range of size, shell condition, reflex impairment, 
and injury level (Table 2), and reflex actions were lost at 
different frequencies. When just one reflex was absent, 
kick and leg retraction were the reflexes lost most fre- 
quently (Table 3). Among the crabs where one to five 
reflexes were lost, leg retraction, kick, and leg flare were 
most often lost, and eye retraction and mouth closure 
were absent least frequently. 
Table 3 
Percentages of reflex actions lost in Chionoecetes bairdi 
(Tanner crab) and C. opilio (snow crab). When just one 
reflex was absent in a crab, it was considered the 1 st reflex 
lost. The right column represents the percentage of spe- 
cific reflexes that were lost among all of the crabs where 
between one and five reflexes were lost. 
Reflex 
. st reflex lost % of total losses 
C. bairdi 
(72 = 18) 
(72 = 147) 
Kick 
44.4 
26.5 
Leg retraction 
44.4 
32.0 
Leg flare 
5.6 
19.0 
Chelae closure 
5.6 
14.3 
Eye retraction 
0 
4.1 
Mouth closure 
0 
4.1 
C. opilio 
(72 = 35) 
(72=210) 
Kick 
51.4 
23.3 
Leg retraction 
31.4 
30.0 
Leg flare 
14.3 
24.8 
Chelae closure 
0 
13.8 
Eye retraction 
2.9 
4.8 
Mouth closure 
0 
3.3 
Mortality increased with increasing reflex impair- 
ment, regardless of the specific index used (Fig. 1). 
However, mortality at a given impairment value was 
consistently higher with index A than with index B, 
and the differences were greatest between index values 
of 1 and 5. Hence, we judged index A, from the count 
of completely lost reflexes (scores=0), to provide a more 
sensitive predictor of mortality than the index based 
on reflexes scored as lost or weak. A lost reflex was 
more definitive and less ambiguous than a weak reflex; 
therefore, reflex impairment index A was used for all 
subsequent analyses. 
Appropriateness of experimental holding 
Adequacy of the field experiment depended upon two 
important experimental requirements. First, hold- 
ing needed to be sufficiently long to allow detection of 
delayed mortality in crabs. Second, the holding methods 
themselves did not induce mortality. 
Most of the mortality in the two species occurred 
within the first few days of collection (Fig. 2). For both 
species, all the crabs with reflex impairment equal to 
6 died within the first hour after reflex assessment. 
Among the balance of the individuals with less impair- 
ment, 80% of the mortality occurred within 3 days for 
C. bairdi and 2 days for C. opilio. For both species, 
>95% died by day 7. Not surprisingly, time to mortality 
decreased with increasing reflex impairment (Fig. 3). 
Some crabs with low reflex impairment died in hold- 
ing, but these crabs tended to have substantial physical 
injuries. Time to mortality averaged ~3 days for these 
individuals, and a high variability in time to mortal- 
