For the beverages (Table 3), repeatability and reproducibility values were related to 10 mg l1 spiking level. RSD% values for drinks are generally lower than those resulting from the solid matrices. This could be due to a poor sample handling. All the recoveries were higher than 75% except for azocarmine G and metanil yellow. However, their recoveries were higher than 60%. This could be explained considering the filtration step before HPLC analysis: perhaps the filter partly retained the dyes. The repeatability and reproducibility were quite low.