NOTE Pepin et al .: Changes in the distribution of larval fish body length following preservation 
635 
evaluate the magnitude ofintra- and interoperator dif- 
ferences in performance for repeated measurements 
of the same specimens. 
Materials and methods 
The study was conducted on Conception Bay, Canada 
(47°45'N, 53°00'W), during the period of 12 July to 4 
August 1995. Sampling was performed daily from 
CSS Shamook (23-m boat length) at a single site near 
the head of the bay. 
Larvae from a number of species were obtained 
from vertical plankton hauls made with a square 
trawl 4 m long with a 4-m 2 mouth fitted with 333-pm 
mesh nitex and an oversize codend 20 cm in diam- 
eter and 30 cm long. The net was lowered to a depth 
of 20-30 m and retrieved at a rate of 1 m/s. Net de- 
sign and deployment protocol were chosen to mini- 
mize trauma to larvae. On deck, the net was washed 
and the codend contents poured into a 20-L plastic 
bucket. Live and freshly dead ichthyoplankton were 
immediately sorted with flexible forceps and trans- 
ferred to petri dishes filled with chilled seawater. 
Moribund larvae, indicative of death or extremely 
poor condition prior to capture (O’Connell, 1981; Otto 
and Boggs, 1983; Takizawa et al., 1994), were ex- 
cluded from our samples to avoid possible bias. Each 
larva was assigned an indentification number, ten- 
tatively identified to the lowest taxonomic level pos- 
sible, and recorded on videotape with a camera 
mounted on a Wild M3C dissecting microscope (S- 
type mount, 0.5x objective). Individual larvae were 
immediately transferred into 1.5 mL microcentrifuge 
tubes filled with 2% buffered formaldehyde. Identi- 
fications were confirmed in the laboratory and stan- 
dard lengths were determined to the nearest 0.1 mm 
using an Optimas® image analysis system. Measure- 
ments of the video-recorded fresh standard lengths 
were performed at the Northwest Atlantic Fisheries 
Centre by an operator with more than 10 years of 
experience in the study of larval fish. All laboratory 
analyses performed on preserved larvae were con- 
ducted approximately five months after the sampling 
cruise at Queen’s University by a novice operator 
with less than 1 year of experience in the study of 
larval fish. Each preserved larva was extracted from 
its microcentrifuge tube, videotaped in a manner 
identical to that employed for recently captured lar- 
vae, and measured (standard length) with an 
Optimas® image analysis system. All measurements 
were performed by clicking on a series of points through 
the centre of the head and along the notochord. 
To contrast the precision and accuracy of length 
measurements, the experienced operator repeated a 
set of measurements on a sample of randomly se- 
lected larvae from the initial videotape records. In 
addition, a random subsample of 72 preserved cape- 
lin larvae ( Mallotus villosus) were selected from the 
collection for replicate measurements by the experi- 
enced and novice operators. In each instance, both 
operators independently videotaped each larva for 
both the initial and repeated measurements. In ad- 
dition, the experienced operator performed duplicate 
measurements on approximately one half of the 
samples measured by both operators. All measure- 
ments were performed with only knowledge of speci- 
men number and magnification level. 
We evaluated individual differences in body length 
following preservation (Al = l preserved - l fresh )- For 
analysis, data were binned into 1-mm length inter- 
vals for fresh specimens (e.g. 1 < x < 2). The hypoth- 
esis that there was no significant difference in mean 
body length ( H 0 :Al =0) was evaluated by using a 
two-sided /-test. Comparison among species was ac- 
complished by using anANOVA( H 0 : A l x = for 
species 1 to n). A species was included in the analy- 
sis only if there were at least three specimens within 
a length interval. Homogeneity of the variances in 
fresh (f) and preserved (p) larval lengths within each 
length interval was assessed to/ using a one-sided 
F-test of the null (H 0 \Sf>=s p ) and alternative 
(H A :sf<si) hypotheses. Changes in individual rank, 
within a length interval, were assessed by using 
Kendall’s rank correlation coefficient (x). Operator 
precision and accuracy were evaluated by using gen- 
eral linear models. 
Results 
Operator accuracy 
The experienced operator exhibited consistent per- 
formance in repeated measurements of both fresh 
and preserved capelin larvae (Fig. 1; Table 2). Re- 
gression analysis of the independent measurements 
by the experienced operator showed no significant 
departure from an intercept of 0 and a slope of 1 for 
both fresh and preserved specimens (Table 2). Re- 
sidual variances for the two treatments were not sig- 
nificantly different (F 40 3g =l.l, P>0.05). 
Comparison of measurements by experienced and 
novice operators revealed no significant bias on the 
part of either operator although the results did ap- 
proach significance (0.1>P>0.05) and the slope was 
not significantly different from 1 (Table 2). However, 
the residual variance about the novice-experienced 
operator regression was significantly greater than 
the residual variance of repeated measurements 
