Infrared (11 mm) radiances from GOES-8 and local radiosonde profiles, collected during the Cirrus Regional Study of Tropical Anvils and Cirrus Layers-Florida Area Cirrus Experiment (CRYSTAL-FACE) in July 2002, are used to assess the vertical distribution of Florida-area deep convective cloud-top height and test predictions as to its variation based on parcel theory. The highest infrared tops (Z11) reached approximately to the cold point at 15.4 km, though these are uncertain by about 1 km due to unknown cloud-environment temperature differences. Since lidar shows that visible ‘‘tops’’ are generally 1 km or more above Z11, visible cloud tops frequently penetrated the lapse-rate tropopause (~15 km). Further, since tropospheric ice concentrations were typically present up to ~1 km above the visible tops, lofting of moisture through the mean cold point was probably common. Morning clouds, and those near Key West, rarely penetrated the tropopause. As in previous studies, nonentraining parcel theory fails to explain either of these results, though it does show promise in explaining day-to-day variations over the peninsula. Moisture variations above the boundary layer account for much of the day-to-day Z11 variability, especially over the oceans. In all locations a 20% increase in mean mixing ratio between 750 and 500 hPa was associated with about 1 km deeper maximum cloud penetration, other things being equal. This sensitivity is too large to explain by simple dilution of parcel buoyancy through mixing, implying microphysical or dynamical feedbacks on cloud development. The evident influence of midtropospheric humidity on the depth of the tropical troposphere suggests an interesting climate feedback possibility for stabilizing midtropospheric relative humidity.