An improved quasi-Lagrangian primitive equation model is used to investigate existing cumulus parameterization schemes. Comparative forty-eight hour forecasts were made utilizing seven different cumulus heating functions, one being zero cumulus latent heat release. Initial data were that of a nondeveloping easterly wave in the Carribbean. Of the six forecasts which included latent heat release, two predictions involving a time-independent vertical heat release function proved to be the most errant. The parameterized heating of Krishnamurti and Moxim, Yamasaki with time-dependent heat release function, convective adjustment, and cumulus scale mass conservation, produced comparable results. Time mean maximum ascending motions of the four similar forecasts ranged from a low of −2 times 10−4 mb sec−1 with Krishnamurti & Moxim heating to a high of near −6 times 10−4 mb sec−1 for Yamasaki heating. Convective adjustment and cumulus scale mass conservation techniques tended to produce poorer low-level forecasts because of a strong dependency upon the initial moisture distribution. Concurring results among forecasts were consistent with the previous tropical scale analysis of Charney. For the easterly wave of moderate intensity, computed vertical motions indicated essentially no coupling of upper and lower tropospheric motion outside convective areas and only minimal coupling in regions of cumulus latent heat release. Seventy to eighty percent of the ascending motion and eddy kinetic energy produced within the wave was estimated to be the direct result of parameterized cumulus latent heat release.