Total alkalinity (TA) is an important parameter in determining the uptake capacity of anthropogenic CO2 by the ocean. So far, oceanic carbon cycle models do not accurately represent TA and its variations. A spectrophotometric method was developed to measure variations of TA during two JGOFS cruises to the Northeast Atlantic in the early summer of 1990 and 1991 and in Emiliania huxleyi batch cultures. Short-term precision averaged around ± 0.1 %. A discrepancy of < 0.5% with coulometric results was observed in Na2COa standards. In natural seawater photometric TA was lower than potentiometric and calculated (pCOl, TC02) TA by about 1 and 2%, respectively. Discrepancies varied with hydrographic and/ or biological regime. Possible reasons for methodological shortcomings were considered, but without certified TA standards for different sample types, it was not possible to make an absolute statement about the accuracy of the methods involved. Combining the cruise results, photometric TA ranged by 90 and 20 p.eq kgSW-l in the surface mixed layer (SML) and at sub-thermocline depths, respectively. Some horizontal variation in the SML was related to salinity, but most of it could be linked to coccolithophorid growth during a bloom in 1991. Associated small-scale changes in TA of up to 40 J.leq kgSW-t occurred over 40 km. Independent estimates of seasonal net production of PlC and its relation to that of particulate organic carbon (POC) were established. Based on preceding investigations, a seasonal and latitudinal sequence of changes in surface TA was proposed which was corroborated by the photometric results from this study. The culture experiments revealed reductions in photometric TA which were half of those expected from parallel changes in measured PlC and nitrate concentrations. Proposed explanations for this included methodological shortcomings of all three methods and increases in final TA due to algal sulphate uptake and/or organic acid release. As the main conclusion, further targeted intercomparisons of TA methods are needed to identify the causes for errors in various TA methods in samples covering realistic hydrographic and biological ranges.