A study of the functions and structure of referential third- person pronominal chains in spontaneous English conversation as compared to those in written English is discussed. The study found the frequencies of these pronouns and of their chains to be generally much higher in the spoken corpus than in the written corpus. In the spoken corpus, "it" chains were the most common, "she" chains were the least common, and most were endophoric. In the written corpus, "they" chains were the most common and "she" chains the least common, and all were endophoric. Spoken chains were longer than written chains and contained more tokens, which can be explained at least in part by factors in the context of situation. It was also found that chain length was closely related to the number of pronouns per chain, and that the range for chain length was much larger in conversation. The study concludes that differences in the use of pronominal chains in conversation and writing are due to differences in referential strategy: while a speaker in conversation makes pragmatic word choices and has no real wish to vary his words, a writer makes stylistic choices and aims at variation in the written product, which acts as a constraint on the use of pronominal chains. (MSE)