During a 5-day period in late winter 1984, temperature were measured at 9 depths beneath a drifting spar buoy to a relative accuracy of 2 mK. The depths ranged from 15 to 145 m in a mixing layer that was nearly homogeneous to over 250 m. A diurnal temperature signal was observed, whose maximum amplitude decreased from 40 mK at 15 m to 20 mK at 145 m. Surface heat fluxes, in particular the diurnal variation in solar radiation, are too small to account for the diurnal variations in mixed layer heat content. Instrumental errors are ruled out. The Lagrangian experiment minimized horizontal advection, which had in any case been quantified using a towed CTD-SeaSoar. The SeaSoar was used to survey the upper 380 m on horizontal scales from 1–40 km around the drifting spar, but revealed no advection that could account for the diurnal signal. It is concluded that the spar was trapped in a convergence zone, where vertical velocities possibly as large as 0.04 m s−1 carried diurnally heated surface water down to 145 m within hours.