The first attempt to establish a relation between the Loop Current extension and deep flows in Yucatan Channel was made by Maul et al. <1985>; it was unsuccessful, probably because of the low spatial resolution of their observations. From September 8, 1999, to June 17, 2000, eight moorings with acoustic Doppler current profilers, current meters, and thermometers were deployed across the Yucatan Channel. The data from these arrays were used to compute time series of the transports below the level of the deepest isotherm observed in the Florida Straits, as required by a simple box model that restricts deep exchanges with the Gulf of Mexico to the Yucatan Channel. The surface extension of the Loop Current was inferred from 3 day advanced very high resolution radiometer imagery from October to May, when temperature gradients were sufficient to map the warm water unambiguously. The deep transports appear at first unrelated to the rate of change of the Loop Current extension, but filtering the series with a 20 day running mean increases the correlation between the low-pass series to 0.62, and up to 0.83 with a lag of 8.5 days, with Loop Current changes leading the deep flows. The cumulative deep transport, a quantity that favors lower frequencies, is very well related (correlations >0.9) to the surface extension of the Loop Current, also with a lag of about a week. These lags are not statistically significant but suggest a timescale for internal adjustment processes in the Gulf of Mexico. The empirical orthogonal function of the current best related to the area extension of the Loop Current represents a unidirectional flow across the entire deep section, flowing either toward or from the Gulf of Mexico, and includes a strong expression of the Yucatan Undercurrent.