AIMS: To analyse whether recommendations for the graduated frequency (GF) approach to measure alcohol consumption are justified in a multi-cultural comparative study. DESIGN: Representative surveys, conducted between 1995 and 2003, of 10 countries participating in the GENACIS project (Gender, Alcohol and Culture: an International Study). MEASUREMENTS: Usual quantity, usual frequency and mean consumption per day measured with three instruments: GF, generic quantity-frequency (QF) and beverage-specific quantity-frequency (QFBS). FINDINGS: The GF did not consistently yield higher volumes and quantities across all countries compared with the generic QF, while the QFBS resulted in higher quantities and higher volumes compared with the GF (in all but one country) and the QF. Frequencies were mostly higher on the GF compared with the QF and QFBS but there was also evidence of over-reporting of frequencies with the GF. Results for the GF suggested that it was implemented improperly in at least three of the 10 countries. CONCLUSION: The GF does not appear to be appropriate for cross-cultural research. It results in over-reporting of frequencies and appears to be too complex to be administered correctly in many countries. The best measure for these purposes appeared to be the QFBS particularly because it captures more effectively the variability of different alcoholic beverages with different ethanol contents and consumption with different vessel sizes.