918 
Fishery Bulletin 96(4), 1998 
Figure 1 
Map of the study area showing the location of sampling stations ( black dots). 
A = Thracian Sea; B = Chalkidiki gulfs; C = Thermaikos Gulf; D = Gulfs of 
central and southeastern Greece. 
or standard length to the nearest 0.1 mm). 
If more than 100 specimens of anchovy were 
captured in a tow, 100 randomly selected 
larvae were measured, and subsequent 
length frequencies were raised to the total 
number of larvae. Lengths were rounded to 
1-mm length groups (e.g. 2-2.99: 2.5 mm). 
Each station was assigned to day, night, 
or twilight hours according to the recorded 
time at the beginning of the tow. Twilight 
was designated as one hour before or after 
sunrise and sunset (Morse, 1989). 
The analysis of retention and catchability 
was made after pooling all available data 
and, subsequently, estimating mean catch 
per length group. For example, mean catch 
per m 2 in each length category was calcu- 
lated for 227 day, 146 night, and 84 twilight 
positive stations (Table 1). Pooling data over 
large temporal and spatial regimes inte- 
grates areal and temporal heterogeneity in 
distribution and abundance, and results of 
the analyses represent average conditions 
(Hewitt, 1981; Morse, 1989). 
Negative stations (where no anchovy lar- 
vae were caught by either of two nets) were 
not included in our analysis ( sensu Hewitt, 
1981); we assumed that they represent 
samples drawn from outside larval habitat. 
Given that the number of negative stations 
was very low in our data set ( 17 out of 474, of 
which 10 were from one cruise, 92ANC2, 
Table 1), inclusion or exclusion of such nega- 
tive stations was not expected to affect results. 
Catches were standardized to numbers 
per m 3 . This standardization is sufficient for 
comparing larval retention in paired differ- 
ent mesh-size nets but, in comparing day- 
time or twilight and nighttime catches, data 
had to be standardized to numbers per m 2 , 
by using maximum tow depth and volume 
of water filtered (Houde, 1977). Mean stan- 
dardized catches per length group were used 
to calculate the following ratios: day:night 
and twilight:night as well as 0.500-:0.250- 
mm mesh and 0.335-:0.250- mm mesh. Vari- 
ances of ratios were approximated as in Somerton 
and Kobayashi (1989). 
The calculation of mean catch and its variance fol- 
lowed the methods of Pennington ( 1983) for the delta- 
distribution of catch frequencies (see also Morse 
( 1989)). The estimators based on the lognormal model 
(delta distribution) are more efficient for marine data 
than the usual sample estimators (Lo et al., 1992; 
Pennington, 1996). In particular, they provide rea- 
sonable estimates for data sets that contain isolated 
large catches. 
Anchovies collected during 1995 were further sorted 
into yolksac, preflexion, flexion, and postflexion larvae. 
Eye pigmentation, functional mouth, and the forma- 
tion of intestine (Regner, 1985; Palomera et al., 1988; 
Clarke, 1989) were used to distinguish yolksac from 
preflexion larvae. The flexion stage begins at initial 
notochord flexion and ends (postflexion stage starts) 
