Erythroid cells express a unique form of beta spectrin I as a result of tissue-specific alternative pre-mRNA processing. Nonerythroid cells that express the beta spectrin I gene include four additional exons at the 3' end of the mature transcript, leading to elongation of the carboxyl terminus of the protein. The nonerythroid beta spectrin I isoform is not present in the red blood cell membrane skeleton; the erythroid isoform is not detected in other cell types. Therefore, developing erythroid cells acquire this tissue-specific pre-mRNA processing activity during differentiation. In the present study, we investigated the developmental timing of erythroid-specific pre-mRNA processing in human erythroid precursors. Partially purified human peripheral blood burst forming uniterythroid (BFU-E) cells were grown in culture for 5 to 12 days. beta Spectrin I mRNA transcripts were analyzed at different time points by S1 nuclease mapping. The processing of beta spectrin I transcripts was found to be exclusively erythroid from day 5 onward, indicating that erythroid-specific processing is not linked temporally to assembly of the mature erythroid membrane skeleton. Human erythroleukemia (HEL) cells had both erythroid and nonerythroid transcripts, indicating that both processing patterns can coexist. Induction of erythroid differentiation in HEL cells using hemin resulted in a partial switch toward the erythroid processing pattern of beta spectrin I transcripts. Using a genomic S1 probe that spans the erythroid polyadenylation signal, we found that a substantial portion of the transcripts detected by the erythroid cDNA S1 probe (in both cultured BFU-E and HEL cells) is incompletely processed pre-mRNA precursors. Poly(A) RNA selection before S1 analysis showed that the unprocessed transcripts are not polyadenylated. We conclude that (1) erythroid-specific pre-mRNA processing activity is present early in erythroid differentiation; (2) beta spectrin I transcripts that are unprocessed at the 3' end accumulate, awaiting either erythroid or nonerythroid processing pathways, from which observation we infer that the regulated alternative pathways are both inefficient; and (3) HEL cells offer a human cell culture model in which to study the balance between the two pre-mRNA processing pathways. We speculate that erythroid cells evolved this tissue-specific pre-mRNA processing machinery for other erythroid genes in addition to beta spectrin I.