A hypothesis is advanced to explain the peak in film drop production as a function of bubble size observed by Blanchard and Syzdek <1988>. Evidence is presented that for bubble diameters less than about 3 mm, bubble films do not disintegrate but roll up, forming a toroidal ring that gains mass as it advances toward the film edge. It is suggested that film drops are created when this rolled-up mass hits the edge of the film and hypothesized that the peak in production arises because of a competition between an increase in the mass of the rolled-up toroid and a decrease in its speed as bubble size increases. In this scenario the location of the break in the film is an important parameter. Measurements of this break point are provided. The theory provides an estimate of the bubble size dependence of the number and size of the film drops. The predicted number of droplets per bubble is less and the droplet sizes are larger than the values reported by Blanchard and Syzdek. The theory uses one adjustable parameter that fits the onset of film drop production observed by Blanchard and Syzdek and, in so doing, glosses over the complexities of the toroid impact with the film edge. The possibility that some film drops are perhaps, in reality, jet drops generated by bubbles created by film drop impacts with the water's surface is advanced.¿ 1997 American Geophysical Union