Sampling was performed on board the research vessel Sarmiento de Gamboa during a cruise carried out in December 2009 (late autumn). The study area was located to the northwest and southwest of Mallorca (Balearic Islands, western Mediterranean Sea, Fig. 1) at the shelf-break (>200 m depth) and the median slope (700-900 m depth). A 280 m² midwater trawl (10 mm mesh size in the cod-end) and a 3 m² (3 mm mesh size) Isaacs- Kidd Midwater Trawl (IKMT) were employed to capture adult and juvenile fish. The depth where the fishing nets operated was controlled using a SCANMAR sensor. Fishing operations were performed both near the surface (40-80 m) and near the Deep Scattering Layer (DSL) at 400 m depth, where the echosounder recorded the highest acoustic signal at 18 and 38 kHz (Olivar et al. 2012). The vessel speed was maintained at 4 knots for the large midwater trawl and 3 knots for the IKMT. After on-board fish identification using specific literature (Hulley 1981, 1984), the specimens were frozen and stored at -20°C to prevent otolith damage. The water column temperature was obtained with the SBE 911plus CTD at each station.

The collection of fish larvae and transformation stages was carried out at the same stations with a 0.25 m² Hydro-Bios MultiNet with 0.3 mm mesh size. Oblique tows at five discrete depths from 200 m to the surface were performed during a 24-h cycle (8 in daytime and 8 at night). Ichthyoplankton samples were preserved in 5% buffered formalin and sorted once in the laboratory. The pH was maintained at ca. 8 to prevent otolith degradation.

Fish larvae were sorted and identified in the laboratory using relevant literature (Tåning 1918, Moser and Watson 2006). Larvae were ascribed to preflexion, flexion and postflexion stages (according to the urostyle bending), or transformation stages (according to photophore development) (Kendall et al. 1984), and were measured to 0.1 mm precision under a microscope employing an ocular micrometre.

The standard length (SL) was measured before otolith extraction for 208 individuals, of which 45 were larvae (<16 mm SL), 49 were individuals at the transformation stage (17-20 mm SL), and the remaining 114 were juveniles and adults (>20 mm SL). For each individual, the left and right sagittae otoliths were extracted from the ear cavity. In the larval and transformation stages, the otolith extraction was performed using a fine needle, with the help of polarized light of the microscope for a better location of the otoliths. After drying, a drop of Cristal Bond 590^TM thermolabile resin was poured over the otolith for further preservation. In the juveniles and adults, we first made a cross-sectional incision at the back of the fish head, near the rear of the operculum. Then, the two sagittae were extracted with forceps according to Secor et al. (1991). After the extraction, both otoliths were cleaned with a 5% solution of KOH to remove the attached organic tissue and then rinsed with water. The mounting and polishing were performed according to the methods described by Morales-Nin (1992) and Stevenson and Campana (1992). An otolith of each fish was mounted on a slide to polish its sagittae section, fixing the inner face with thermolabile resin.

The otolith length was defined from maximum diameter in larvae and from rostrum to posterior margin according to Tuset et al. (2008). In the smallest larvae (5-7 mm SL), the transparency of the otoliths allowed us to count growth increments without need for polishing. Otoliths from the transformation, juvenile and adult stages were manually polished with grinding lapping papers (0.1-3 μm). In otoliths from larger specimens, the polishing process was interspersed with micrographs of the increments located in the margin of the otolith (the narrowest ones) to ensure the identification and the count of these increments and prevent any loss in the case of over-polishing. To increase transparency in thicker otoliths, the polishing process was carried out on both sides, taking care not to affect the maximum diameter of the otolith. Finally, the otoliths were ultra-polished with an abrasive consisting of a MicroCloth PSA 10/ PK 8 disc previously impregnated with an abrasive solution with 0.05 μm of alumina. Daily increments were counted from images obtained with a digital camera mounted on an optical microscope (Zeiss Axioskop 2 Plus) and with the Image-Pro Plus v.5.0 image analysis software (magnification: 100×, 400×, 630× and 1000×). Double readings were performed in the postrostrum otolith section. If the first reading was not coincident with the second one, a third reading was performed to guarantee the correct interpretation of age.

To determine whether the increment width was outside the resolution threshold of the optical microscope, some otoliths from the transformation stage (n=1), and juvenile and adult stages (n=5) were also examined at high resolution. Polished otoliths were immersed in a 1% HCl solution for between 60 and 300 seconds, rinsed with water and allowed to air dry for 24 hours. Finally, the otoliths were mounted, covered with gold-palladium and then observed using a scanning electron microscope (SEM).

In the Mediterranean Sea, C. maderensis is not sexually mature until it reaches ≥40 mm SL (Hulley 1981, 1984). Specimens collected in this study were not sexed, so we cannot confirm whether there are size differences between sexes in individuals of C. maderensis from the western Mediterranean Sea. Therefore, in this study, specimens of both sexes were considered together. Linkowski et al. (1993) observed that individuals of C. maderensis from the northeastern Atlantic Ocean showed no significant size differences between males and females. The relationship between the fish length and the otolith diameter was determined by linear regression to ensure that the growth of the otolith is proportional to the somatic growth of the species. As each increment is assumed to represent one-day growth, the relationship between the number of day-increments and SL was fitted by a von Bertalanffy growth curve based on Equation 1:

where L _t is standard length in mm, L _∞ is the asymptotic size in mm, k in days⁻¹ is the Brody growth rate coefficient, which measures the rate at which the growth rate declines, and t ₀ corresponds to the age at which L is 0 (von Bertalanffy 1938).

The absolute growth rate (g _t ) is given by g _t =k(L _∞ -L _t ), and when t=0 (w), the absolute growth rate is w=k×L _∞ . Parameters k and L _∞ were fitted and estimated using R software.

Finally, a second von Bertalanffy growth curve considering adult specimens only (i.e. SL>19 mm) was fitted in order to assess for growth differences between our specimens from the western Mediterranean Sea and from the northeastern Atlantic Ocean measured by Linkowski et al. (1993) (individuals with SL<19 mm were not considered by these authors). The comparison of these two curves was performed without fixing parameters or fixing one, two or all three parameters using the fisheries stock-assessment method package (FSA; Ogle 2016) in R software. According to this method, when the best explanatory models are those in which most of the parameters have not been fixed, this indicates that there are differences between the curves. By contrast, when the best models are those in which most parameters have been fixed, this indicates similarities between the curves. The best models were selected using the Akaike information criterion (AIC, Burnham and Anderson 2002). Models with AIC differences ≤2 were considered equivalent (see Burnham and Anderson 2002, 2004).

The observation and the counting of otolith increments were feasible using both optical microscope and SEM. Three different zones were differenciated in adult specimens: larval (Fig. 3A, B), metamorphic (Fig. 3C, D), and juvenile-adult (Fig. 3E, F). Table 1 shows the SL ranges and the number of increments for each of these regions. The larval region was composed of a core (Fig. 3B) followed by a succession of an average of 27 increments (n=9, σ=2.98). The mean radius between the centre of the core and the outer part of the first increment was 2.95 μm (n=3, σ=0.84), while the average radius of the larval zone was 125.34 μm (n=19, σ=30.46). Within the larval region, the increment thickness increased progressively following the exponential function:

where y corresponds to the thickness of the increments in μm and x is the radius in μm (r²=0.91). The mean increment thickness in this region ranged from 1.4 to 4.2 μm (n=14, σ=0.88). The end of the larval region was well defined by a dark band of transition between the larval and metamorphic zones (Fig. 3A). This band was 5-10 μm wide and could also be detected using an optical microscope.

The metamorphic zone (Fig. 3C, D) extended from the external limit of the larval region (Fig. 3A) to the beginning of the juvenile-adult region (Fig. 3E, F). Increments in this region showed no clear pattern and were sometimes even unnoticeable or incomplete and irregular. This greatly hindered counting of the number of increments in this area, so estimates for this region should be taken with caution. Additionally, it made it difficult to obtain accurate measures of the mean increment thickness in this region, which was 142.46 μm (n=14, σ=42.23). Finally, the juvenile-adult zone (Fig. 3E, F) consisted of well-defined increments that showed the typical radial structure and rhythmic growth patterns, as in other species. The mean increment thickness was 4.94 μm (n=22) and showed a decreasing trend towards the otolith edge (to <1μm) (Fig. 3F). In addition to the three development regions mentioned above, from one to three translucent regions were also observed along the radial axis of the otolith of adult individuals (Fig. 4). The radius of each of these translucent regions relating to both the fish SL and the number of increments in the otolith is shown in Table 2.

The age interpretation of the otoliths under the optical microscope with coincident readings was successful in 59 individuals between 3.5 mm and 64 mm SL (Table 1). Assuming that the frequency of increment formation in the otoliths of C. maderensis is daily, these individuals would be between 7 and 332 days old. The relationship between the number of increments and the SL, considering the full-size range of individuals, renders an asymptotic size of 70.59 mm and growth rates of 0.33 (for the smallest larvae) to 0.03 (for the largest adult specimen), with a growth rate decline through the development of 0.01mm (Fig. 5, Tables 3 and 4).

Table S1 shows the estimates of the parameters of the von Bertalanffy growth curves that we obtained when considering adult specimens only (SL>19mm), in order to assess growth differences between our specimens and those measured by Linkowski et al. (1993) (Fig. 6) in the northeastern Atlantic Ocean. The comparison of these two curves revealed that the best explanatory model (Model 1, Table 5) was the one in which the three parameters of the growth curve (L _inf, k and t ₀) had been fixed. This indicates that growth differences between our specimens and those measured by Linkowski et al. (1993) in the northeastern Atlantic Ocean are not statistically significant. Models 1, 2, 3 and 4 (Table 5) can be considered equivalent as their ∆AIC values were <2. For body size, von Bertalanffy models predicted a SL of 72.82 mm for the Mediterranean individuals and 77.94 mm for the Atlantic ones (Table S1). However, considering the standard error obtained, body size differences between the Mediterranean and Atlantic specimens should be considered with caution.