240 
Fishery Bulletin 108(2) 
shell heights (Table 2). The nonlinear shell-height to 
meat-weight relationship showed exaggerated extremes 
of the distributions so that the ratio of maximum to 
mean meat weight was 158/27=5.9 for video data and 
138/29 = 4.8 for measuring boards (Table 3) compared to 
201/106=1.9 and 193/109=1.8 for shell heights (Table 2). 
Variance in meat-weight measurements increases as 
true meat-weight increases for video data and, to a 
lesser extent, for measuring boards (Fig. 5). 
The meat-weight composition data were more right 
skewed (gj=1.53) and flatter (g 2 = 6.22) than the meat- 
weight composition data from measuring boards 
(^ 1 =0.92 and ^2=2.61) or calipers (^=0.99 and g 2 =3.00). 
Errors in meat-weight data were left skewed and not as 
peaked for video (^ 1 =-0.80 and g 2 =2.48) than measur- 
ing board data (^ 1 =-1.06 andg\ 2 = 4.68). 
80 1 A 
Measurement (mm) 
i 1 1 1 1 1 
0 200 400 600 800 1000 
Distance from origin (mm) 
Figure 3 
(A) Video measurements for tiles in experiment 1. The verti- 
cal line shows the true value at 48.5 mm. ( B) Measurement 
errors (video measurement minus caliper measurement) 
for tiles in experiment 1 as a function of distance from the 
origin (DFO). The nonlinear LOESS regression line shows 
the overall trend in measurement errors as a function of 
DFO. 
Results from the assessment models 
Based on results from experiment 2 (Table 2) and 
assumptions listed above, video shell-height measure- 
ments for sea scallops with true sizes evenly distributed 
over 100-104.99 mm SH (i.e., the 100-mm bin with 
midpoint 102.5 mm) would fall into nine observed shell- 
height bins with midpoints from 77.5 to 117.5 mm (Table 
4). Measuring board shell-height measurements would 
fall into four observed shell-height bins with midpoints 
ranging from 92.5 to 107.5 mm (Table 4). 
Four model configurations were used. The “no mea- 
surement error” model configuration was fitted by as- 
suming no errors in shell-height data. The “bias only” 
model was fitted by assuming that shell-height data 
were biased (to the extent measured in experiment 2), 
but precise (with zero variance). The “imprecision 
only” model was fitted by assuming that shell-height 
measurements were imprecise (standard deviations 
from experiment 2), but not biased. The “impreci- 
sion and bias” model was fitted by assuming both 
types of shell-height measurement errors. 
Models which accommodated measurement errors 
fitted better, with substantially lower negative log 
likelihoods for both stocks, than models that ig- 
nored measurement errors. Differences in negative 
log likelihood were mostly for shell-height compo- 
sition data. Mean 2004-06 biomass and fishing 
mortality rates and coefficients of variation (CV) 
for biomass and fishing mortality estimates were 
similar for all model configurations (Table 5). 
Discussion 
The importance of body-size measurement errors 
and the need to accommodate them in modeling 
probably depends on the situation. Biological factors 
(growth rate, recruitment variability), assessment 
model type, quality and quantity of fishery and fish- 
ery-independent data may be important. Sea scallops 
may be an atypical case because they are a data-rich 
species. We suggest that the potential importance of 
body size measurement errors should be evaluated 
on a case by case basis, particularly if body-size data 
may be imprecise or biased. Simulation studies may 
be useful in determining the importance of experi- 
mentally derived body-size measurement errors on 
stock assessment results. 
In the sea scallop case, models that accommodat- 
ed measurement errors fitted substantially better, 
but there was little effect on point estimates and 
variances for recent biomass and fishing mortality. 
We hypothesize that effects on biomass and mortal- 
ity estimates would be larger in cases with positive 
biases in body-size measurements. For both video 
and measuring boards, the positive bias in meat 
weights due to the nonlinear relationship between 
body size and meat weight was mitigated to some 
extent by the negative bias in shell-height mea- 
