SHERIDAN ET AL.: RECAPTURE PATTERNS OF TAGGED PENAEID SHRIMP 



tude, hereafter called "grids", along the Texas 

 coast. 



Port agents in Tamaulipas interviewed the cap- 

 tains of all vessels returning to the primary port of 

 Tampico. Unknown, but assumed relatively small, 

 amounts of catch and effort were potentially re- 

 ported in more southerly ports. Interviewers col- 

 lected catch and effort data by depth range and 10 

 minute lines of latitude between 22°N and 26°N. 

 These data were then recordable either within 9 m 

 depth zones or within grids as was done off Texas. 



Interviews recorded effort by specific 9 m depth 

 zones (Texas) or by actual depth ranges (Tamauli- 

 pas) per trip. Tamaulipas effort was assumed to fall 

 equally into adjacent 9 m depth zones if more than 

 one zone was covered by the stated depth range. The 

 average fishing depth per trip was then calculated 

 by weighting the hours expended in each 9 m depth 

 zone by the middepth of that zone (e.g., the 10-18 

 m zone had a middepth of 14 m), summing over all 

 depth zones, then dividing by the total effort ex- 

 pended on that trip. Average fishing depth for each 

 fleet was then compared by a t-test corrected for 

 unequal variances (Sokal and Rohlf 1969) using the 

 average depth for each of 2,008 Texas trips and 505 

 Tamaulipas trips as observations. 



Data Analysis 



Three-factor, model I analysis of variance ( ANOVA) 

 with unequal cell sizes was employed to test hypoth- 

 eses concerning the equivalence of treatment means 

 for several types of observations on recaptured 

 shrimp. The treatment factors were species (brown 

 or pink), sex, and release state (Texas or Tamau- 

 lipas). State was chosen as a treatment because the 

 level of fishing effort off Texas is much greater than 

 that off Tamaulipas (approximately 200 vessels use 

 the port of Tampico, whereas there are nearly 2,000 

 vessels registered in Texas alone). Four attributes 

 of shrimp movement were examined by ANOVA: 

 1) distance travelled before recapture, assumed to 

 be a straight line, 2) days at large, 3) apparent speed 

 of movement, and 4) recapture depth. All four 

 variables exhibited nonnormal (skewed) error distri- 

 butions, as indicated by the Shapiro-Wilk test 

 statistic (Shapiro and Wilk 1965), but the effects of 

 nonnormality are thought to be minimal with large 

 sample sizes (Underwood 1981). Variances of all 

 variables were found to be heterogeneous (F-max 

 test for unequal cell sizes; Sokal and Rohlf 1969). 

 Data were log(x -i- l)-transformed prior to ANOVA 

 (Underwood 1981), and F-max tests on transformed 

 data indicated homogeneity of variances. Multiple 



comparison of treatment means of transformed data 

 employed Fisher's LSD (least significant difference) 

 because of unequal cell sizes (Milliken and Johnson 

 1984). 



Circular scale data such as compass directions are 

 a special tj'pe of interval scale data (Zar 1984) that 

 cannot be examined by ANOVA because there is no 

 physical reason for any zero point and high or low 

 values are arbitrary (e.g., 45° is not a "larger" direc- 

 tion than 30°. and the mean of the 45° and 315° 

 is not 180° but 0°). Examination of the raw data 

 indicated that the assumption of unimodal distribu- 

 tions of recapture directions needed for hypothesis 

 testing with the recommended parametric test 

 (Watson- Williams statistic) would be violated. We 

 conducted multisample testing of grouped direc- 

 tional data using contingency tables (Zar 1984). 

 Before analysis, compass direction from release site 

 to recapture site was adjusted downward by 20° off 

 Texas and upward by 20° off Tamaulipas because 

 northerly movement parallel to shore (hereafter 

 termed "north") is 20° west of magnetic north 

 (340°) off southern Texas and 20° east of magnetic 

 north (020°) off northern Tamaulipas (Fig. 1). We 

 grouped the adjusted directional data into eight ar- 

 bitrary 45° divisions (0-44°, 45-89°, etc.) that ful- 

 filled the requirement of having no expected cell fre- 

 quency less than 4 (Zar 1984), with one exception. 

 Only one brown shrimp and one pink shrimp re- 

 leased off Tamaulipas were recaptured between 

 270° and 359°; thus the contingency tables compar- 

 ing these two data sets employed six 60° divisions 

 (15-74°, 74-134°, etc.) to avoid low cell frequencies. 



Differences in shrimp movement away from re- 

 lease sites were also tested by examining patterns 

 in recaptures per unit fishing effort (R/f). R/f ad- 

 justs for temporal and spatial variations in fishing 

 effort around each release site and integrates the 

 effects of distance and direction travelled (Gitschlag 

 1986). For each release, recaptures per lO'' hour of 

 effort were calculated north, within, and south of 

 the release grid from the release date through the 

 end of the study period. "North" was defined as all 

 grids lying between the northern latitude of the 

 release grid and the northern latitude of the grid 

 containing the northernmost recapture after each 

 release. "South" was defined as all grids lying be- 

 tween the southern latitude of the release grid and 

 the southern latitude of the grid containing the 

 southernmost recapture after each release. "With- 

 in" was defined as the release grid and all grids 

 directly east and west of it (recaptures in these grids 

 did not show longshore movement). Two-factor, 

 mixed model ANOVA with balanced cell sizes was 



299 



