An extensive survey of external pollen morphology in the monocot family Araceae was undertaken, using both light and scanning electron microscopy. Pollen of approximately 380 species was examined, representing 103 of the 111 currently recognized genera. The data collected were analyzed to determine polarities for the character-states of each major pollen character. The most primitive pollen in Araceae is regarded as monosulcate, boat-shaped, heteropolar, bilaterally symmetrical and small to medium-sized. It is shed in monads, has foveolate to reticulate exine sculpturing and is probably tectate-columellate. The evolution of pollen exine type in Araceae may be explained in part by pollinator type. The two most advanced exine types--psilate and spinose--are very closely associated with pollination by beetles and flies, respectively. Two embryological aspects of araceous pollen were investigated: nuclear number and starch content. Binucleate, starchless pollen clearly represents the primitive condition in Araceae. The presence of starch in aroid pollen is primarily related to pollen size--i.e., larger grains are much more likely to store starch. The phylogenetic relationships of pertinent infrafamilial aroid taxa were analyzed using pollen and all other available characters, resulting in an informal cladogram and a preliminary new classification for the family. Major deviations of this system from the standard classification of Engler include: the referral of Acorus to its own monotypic family, and of Gymnostachys to its own subfamily; the merger of the subfamilies Pothoideae and Monsteroideae; the dissolution of Calloideae; the radical emendation of the subfamily Lasioideae; the enlargement of the subfamily Philodendroideae to include several tribes from the Lasioideae and Aroideae; a drastic internal rearrangement of the Colocasioideae; the reduction in size of the Aroideae, plus the insertion of Pistia in this subfamily and the consequent abandonment of the subfamily Pistioideae. The relationship of the Araceae to other families was briefly considered. The Araceae are in general considered more closely related to the Alismatiflorae than to the Areciflorae, Typhales or Piperales. The Lemnaceae are not closely related to Pistia, and their relationship to the Araceae in general needs to be reassessed.