86 
Fishery Bulletin 116(1) 
Table 1 
Procrustes distances for the mean body shape of blue 
jack mackerel (Trachurus picturatus) from the 3 areas 
sampled in this study in 2015, off the Madeira archi¬ 
pelago, off Peniche on mainland Portugal, and off the 
Canary Islands. Corresponding P-values (in parenthe¬ 
ses) obtained from a permutation test (10,000 permu¬ 
tation runs) are also given. The results indicate that 
all of the mean body shapes are statistically different 
between the 3 areas. 
Mainland Canary 
Region Madeira Portugal Islands 
Madeira - 0.03 0.03 
(cO.0001) (<0.0001) 
Mainland Portugal - - 0.03 
(cO.0001) 
Table 2 
Classification matrix from the discriminant analysis 
performed by using a jackknife procedure for cross-val¬ 
idation of the body shape of individual blue jack mack¬ 
erel (Trachurus picturatus) sampled from 3 areas in 
the northeast Atlantic Ocean in 2015: off the Madeira 
archipelago, off mainland Portugal, and off the Canary 
Islands. Values are percentages of individuals sampled 
in the areas (given in rows) that were classified into the 
area given in columns (values for correct classification 
are presented in bold). Overall classification success 
was 78.0%. 
Region 
Madeira 
Mainland 
Portugal 
Canary 
Islands 
Madeira 
67.0 
10.0 
23.0 
Mainland Portugal 
13.0 
83.0 
4.0 
Canaries 
11.0 
5.0 
84.0 
Figure 5 
Two-dimensional ordination plot of the samples of blue jack 
mackerel (Trachurus picturatus ), collected in 2015 from off the 
Madeira archipelago (black crosses), off mainland Portugal (open 
circles), and off the Canary Islands (black triangles), based on 
canonical variate analysis. Stars indicate class centroids per 
sampled area (each individual was allocated to the group with 
the nearest centroid). 
Canary Islands and Peniche (5.0%) and Ma¬ 
deira (11.0%), as well as between Peniche 
and the Canary Islands (4.0%) and Madeira 
(13.0%). However, the percentage of individu¬ 
als incorrectly classified between the referred 
locations was low. The proportion of correctly 
classified samples off Peniche (83.0%) and the 
Canary Islands (84.0%) to their original group 
was highest, showing a clear separation from 
samples off Madeira (67.0%). 
The 2-dimensional ordination plot (Fig. 5) 
showed a differentiation of stocks for the three 
areas studied, but with some overlapping of 
stocks mainly between Madeira and the two 
other areas; the Canary Islands and Peniche 
seem to be more distinct from each other. The 
first two canonical variate analysis accounted 
for 74.5% and 25.5% of the between-group 
variability. 
Otolith shape analysis 
The mean and standard deviations of the cu¬ 
mulative variance of the harmonics are shown 
in Figure 6. Only the first 15 harmonics were 
used for multivariate analysis because these 
were responsible for over 99% of the otolith 
shape variation. 
Figure 7 represents the outlines of the 
mean NEFDs, by region, plotted as an overlay 
picture to enhance differences in average oto¬ 
lith shapes. Overall, the highest shape varia¬ 
tion among areas occurred in the excisura os- 
tii and antirostrum. Variations also occurred 
in the dorsal and ventral edges of the otolith. 
The results of the MANOVA applied to compare the 
NEFDs by area and sex are shown in Table 3. Because 
neither differences in otolith shape between males and 
females within the three areas, nor interaction between 
area and sex, were found, all subsequent analyses were 
performed for sexes combined. 
The variables used in the stepwise discriminant 
analysis are represented in Table 4. For the analysis 
