633 
Changes in the probability density 
function of larval fish body length 
following preservation 
Pierre Pepin 
Fisheries and Oceans 
RO. Box 5667 
St. John's, Newfoundland, Canada A! C 5X! 
E-mail address: pepin@athena.nwafc.nf.ca 
John F. Dower 
William C. Leggett 
Department of Biology, Queen's University 
Kingston, Ontario, Canada K7L3N6 
The influence of body size on physi- 
ological and developmental pro- 
cesses during the early life history 
of fishes has been clearly demon- 
strated (Miller et ah, 1988; Houde, 
1989; Pepin, 1991). Because of this 
and to provide a direct comparison 
of field collections with laboratory 
observations that often use mea- 
surements of fresh specimens, nu- 
merous studies have quantified the 
effects of preservation and han- 
dling on the length of larval fish 
(Table 1). Most of the research has 
used laboratory-reared animals for 
which changes in length are as- 
sessed using either comparisons 
among treatments (i.e. AN OVA) or 
departures from a one-to-one rela- 
tionship (i.e. regression) (Table 1). 
Shrinkage in length is the predomi- 
nant response to preservation 
(Table 1), and the amount of shrink- 
age is influenced by handling 
(Theilacker, 1980, 1 966: Theilacker 
and Porter, 1995; Fox, 1996). 
Changes in larval body length 
due to preservation are relatively 
small (3-15%) although variations 
of up to 1 mm are not uncommon 
(Table 1). The contrast among spe- 
cies led Jennings (1991) to suggest 
that specific correction factors 
would be required. An alternative 
was Hjorleifsson and Klein-Mac- 
phee’s (1992) simple model of the 
relative change in larval lengths, 
based on a review of previous stud- 
ies, which showed clear evidence of 
the effects of body length and pres- 
ervation time on shrinkage. Of 
course, as with any such review, the 
effects of unaccounted for or con- 
founding variables are unknown 
(Pepin and Miller, 1993). However, 
Hjorleifsson and Klein-Macphee 
( 1992 ) predicted a maximum mean 
shrinkage of -15% for the smallest 
larvae (-2 mm) followed by an ex- 
ponential decrease in relative 
shrinkage with increasing length. 
This raises an important point. Al- 
though differences in mean larval 
body length may appear significant 
from a statistical perspective, the 
importance of correcting for the ef- 
fects of preservation are most pro- 
nounced at the level of the indi- 
vidual because morphological mea- 
surements are used in the analysis 
of each larva’s physiological condi- 
tion (e.g. Theilacker and Porter, 
1995) to assess the relative state of 
a population. Even small changes 
in length resulting from shrinkage 
can create important biases at this 
level of analysis. In contrast, the 
small relative effects of preserva- 
tion on body length (Hjorleifsson and 
Klein-Macphee, 1992) is unlikely to 
substantially influence estimates of 
length-frequency distributions. 
Fundamental issues yet to be 
addressed in the assessment of lar- 
val shrinkage include 1) the effect 
of preservation on the distribution 
of larval length measurements 
within a given length interval 
rather than the mean, and 2) the 
contribution of investigator-in- 
duced error on that distribution. 
The former point is of particular 
importance because the application 
of correction factors to individual 
larvae must maintain the status of 
that animal in relation to others 
within the population. Investigator- 
induced error may be equally impor- 
tant. Few studies have attempted to 
quantify the variance in repeated 
estimates for a given operator (Jen- 
nings, 1991; Hjorleifsson and Klein- 
Macphee, 1992; Fox, 1996) and none 
have contrasted bias and variance 
among operators. Our failure to 
address these questions to date 
probably results from 1) the small 
sample sizes (i.e. generally <100) 
presented in most studies of the 
effects of preservation (Table 1), 
and 2) the fact that usually only a 
single investigator is involved in 
making the measurements of the 
larvae. Nonetheless, these issues 
are important because they provide 
the basis for narrowing the possible 
sources of error in interpretations 
of physiological processes that in- 
fluence ichthyoplankton population 
dynamics. 
In this study we report on an 
evaluation of the shrinkage effects 
of preservation on several species 
of larval fish collected as part of a 
field study. We consider changes in 
both the mean and variance of the 
distribution of individuals within 
narrow length intervals and assess 
the null hypothesis that there are no 
differences among species. We also 
Manuscript accepted 11 September 1997. 
Fishery Bulletin 96:633-640 (1998). 
