TOPEX/POSEIDON (T/P) collinear differences have been used to extract errors in the M2 tide model of Cartwright and Ray (1991). The fluctuation of the power of these differences clearly shows the need for corrections at a period of about 60 days, close to the aliased periods of the M2 and S2 tides. The level of the necessary changes at this period is about 4 cm (rms). Further detailed analysis of collinear differences in 4¿¿4¿ bins, allowing for determination of up to 6 luni-solar and ''ocean weather'' tides (annual and semiannual), shows that the M2 corrections are well discriminated from the others and have a global power of 3.5 cm (rms). Both in-phase and out-of-phase components of the corrections have similar broad scale characteristics dominated by a few meridionally oriented bands. The T/P-derived M2 model shows rms deviations of 3.1 cm with data at 80 selected ocean island and bottom gauges compared with 3.9 cm for the Cartwright and Ray (1991) model. A weighted combination of Geosat and T/P-derived models achieves residuals of 2.8 cm (rms) with this global tide gauge data.