224 
Fishery Bulletin 107(2) 
the available current directions. In addition, the sig- 
nificance of current speed and direction in determining 
whether the northern fish was off bottom was tested 
using Generalized Additive Modeling (GAM; Venables 
and Ripley, 1994). This was done by modeling off-bot- 
tom status (coded 0 for on bottom and 1 for off bottom) 
as a binomial response to a smooth function of current 
speed and direction, with significance based on analy- 
sis of deviance. The test was conducted independently 
for each month that the fish was at liberty, excluding 
daytime hours when fish remained on the bottom (see 
Results section). 
Selective tidal transport model 
To determine if the selective use of tidal currents by 
northern rock sole was an important component of their 
seasonal horizontal migrations, we developed a selec- 
tive tidal stream transport model similar in its basic 
design to that discussed in Arnold and Holford (1995). 
Assuming that all horizontal movement occurred during 
off-bottom periods, we constructed a migration path 
as follows. Starting from the release location, the fish 
was assumed to drift at the current speed predicted for 
each off-bottom time and in the mean current direction 
during each vertical excursion. In addition, the fish was 
allowed to swim at a specified speed, also in the mean 
current direction during each vertical excursion. The 
specified fish swimming speed was held constant over 
the entire path, but because the true swimming speed 
was unknown, this speed was iteratively varied until 
the distance between the final path location and the 
actual capture location was minimized. Thus, for each 
0.5-h or 1-h recording time starting with the release 
location, latitude and longitude positions were advanced 
by using the combined drift and swimming speed and 
mean current direction for each excursion. Tag locations 
along the selective tidal stream transport path were 
converted to latitude and longitude positions using great 
circle formulae. 
To verify the accuracy of the predicted migration 
path, the predicted depth at each location along the 
path was plotted against the bottom depth (maximum 
24-h depth) measured by the tag. Path depths were pre- 
dicted with inverse distance-weighted surface interpola- 
tion (ArcMap 9.2 with Spatial Analyst Extension, ESRI, 
Redlands, CA) using National Imagery and Mapping 
Agency (NIMA) depth sounding data for the eastern 
Bering Sea continental shelf. 
Results 
Depth data and vertical excursions 
The archival tag depth data contained relatively high 
frequency variation from three sources. First, verti- 
cal excursions away from the bottom were identifiable 
by sharp decreases in tag depth (Fig. 2). Second, tidal 
height fluctuations were evident in the tag depth record 
(Fig. 2 A inset), an indication that fish were settled on 
the bottom. Third, rapid changes in depth resulted from 
horizontal movements along steep bottom gradients. 
The identification of vertical excursions was clearer 
for the northern fish than for the southern fish. The 
northern fish inhabited areas of relatively flat bottom 
where estimated tag bottom depths ranged from 48 to 
74 m; therefore rapid depth changes clearly reflected 
vertical excursions away from the bottom. The south- 
ern fish, by comparison, inhabited an area of complex 
bathymetric contours where bottom depths collected 
from tags ranged from 9 to 161 m. Movements along 
a steep bottom gradient were evident for the southern 
fish, particularly during April 2004 when fish depth 
increased and decreased more than 100 m within a 
24-h period (Fig. 2B). Because of this complexity, there 
were occasions when it was unclear whether changes in 
tag depth resulted from a vertical excursion or a quick 
movement along a bottom gradient. For this reason, 
not all vertical excursions were identifiable for this 
fish. More gradual movements across bottom gradients 
were observed for both fish, and most often coincided 
with periods of vertical excursions, indicating that the 
excursions were related to the horizontal movement of 
the fish. In a few cases with the southern fish, however, 
movements across bottom gradients occurred in the 
absence of vertical excursions (Fig. 2B inset). 
The frequency, duration, and distance of vertical ex- 
cursions away from the bottom were similar for the two 
fish. A total of 78 distinct excursions away from the 
bottom were identified for the northern fish during a 
period of 314 days, and 154 excursions were identified 
for the southern fish during a period of 620 days (Table 
1). These excursions were relatively rare, accounting 
from 2.0% (southern fish) to 2.6% (northern fish) of 
the time at liberty. Average excursion durations were 
2.6 hours (northern fish) and 2.1 hours (southern fish); 
average excursion extent (i.e., maximum distance off 
bottom) was 14 m with a maximum of 64 m (Table 1). 
The frequency of vertical excursions varied seasonally; 
most excursions occurred from winter to spring for the 
northern fish and during fall and spring for the south- 
ern fish. Excursions were infrequent during summer 
months (Fig. 2). 
The timing of vertical excursions was related to both 
diel and tidal factors. For the northern fish, 90% of the 
excursions occurred at night, whereas for the southern 
fish, 85% occurred at night (although not all vertical 
excursions could be identified with certainty). Both fish 
underwent vertical excursions that sometimes occurred 
over a series of consecutive nights (Fig. 3). In addition 
to being limited to nighttime, vertical excursions oc- 
curred during particular stages of the tide cycle. For 
example, during the beginning of September 2003, the 
southern fish made consecutive nightly excursions, but 
only before the dominant low tide (Fig. 3B). Examina- 
tion of tidal current directions (northern fish only), 
revealed the fish did not indiscriminately choose night- 
time periods, but made nightly vertical excursions only 
when tidal currents were in certain directions. For ex- 
