The effects of body-scale sample structure on the back-calculated lengths o
f roach, Rutilus rutilus (L.) were studied. Samples for growth determinatio
ns were collected from two separate lake basins using different fishing met
hods. As a result, the length ranges in the two samples were different (84-
248 mm in the Laitialanselka basin and 40-260 mm in the Enonselka basin). W
hen back-calculations were based on the datasets collected separately from
the two basins, the mean length at age 1 was higher in Laitialanselka than
in Enonselka in six out of eight cohorts. When the two body-scale samples w
ere pooled and used for both basins, the mean length at age 1 in Laitialans
elka never exceeded that in Enonselka. According to the separate body-scale
samples, the difference in the growth rate of roach between basins widened
during the 1980s, whereas with the pooled body-scale sample no such conclu
sion could be drawn. When roach <80 mm were omitted from the pooled body-sc
ale sample, the back-calculated length of the roach cohort 1989 at age 1 wa
s 5% higher than the length obtained with the whole dataset, whereas trunca
tion in the upper end of the dataset had no effect. With minimum sizes of 1
00 and 120 mm, the error was 13% and 25%, respectively. The effect of lengt
h truncation decreased clearly with increasing value of b in the power func
tion describing the body-scale relationship. The error caused by length tru
ncation was more than doubled when the value of b was changed from 0.8 to 0
.6. The results demonstrated that length truncation in the body-scale plot
should be taken into account in the back-calculation of fish lengths, espec
ially in fish populations having a clearly curvilinear body-scale relations
hip. (C) 2000 Elsevier Science B.V. All rights reserved.