Dialects vary in their allophonic patterns, which can affect listeners' phonological and lexical representations. I explore how different exposure to dialect-specific allophonic patterns for two vowels in American English, /ae ai/, affects listeners' lexical processing behaviors across three perception tasks: perceptual similarity, priming, and visual-world eye-tracking. I evaluate claims about phonologization of allophonic patterns, which have largely relied on production data in existing literature. I contrast listeners from northeastern Ohio, a region with less/non-phonologized /ae/ allophony and more/phonologized /ai/ allophony, with listeners from central/southwestern Ohio, a region with more/phonologized /ae/ allophony and less/non-phonologized /ai/ allophony. I found that listeners with more exposure to dialect-specific allophonic variation for a specific linguistic variable perceived mismatching allophones and matching allophones as less similar than listeners with less exposure. These results suggest that exposure affects the degree of phonological contrast present in listeners' cognitive representations of allophones. These results also provide preliminary support for late abstractness theories of phonologization, in which phonological abstractness gradually develops during a diachronic change. I also observed that the two linguistic variables differed in how they were treated during lexical processing. For /ae/, listeners treated words with mismatching allophones just like words with mismatching phonemes during lexical disambiguation. Listeners with more exposure also showed a lexical processing advantage via earlier differentiation of allophonic mismatch from match trials as compared to listeners with less exposure. These results indicate that both listener groups have phonologized /ae/ allophony, but more exposure to /ae/ allophony still confers an additional lexical processing advantage relative to less exposure. For /ai/, listeners treated words with mismatching allophones as more ambiguous than words with mismatching phonemes during lexical disambiguation. Listeners with more exposure also showed a slight lexical processing advantage in the form of earlier differentiation of allophonic mismatch from match trials as compared to listeners with less exposure. These results suggest that neither listener group has phonologized /ai/ allophony, but more exposure to /ai/ allophony was still associated with more accumulation of phonological abstractness via a stronger degree of phonological contrast in perceptual similarity. The results of this study add nuance to existing theories of how allophonic patterns are cognitively represented by examining perception data. Additionally, this study contributes to a stronger understanding of the role of exposure to dialect-specific allophonic variation in online lexical processing. Finally, this study showed that the phonological status of allophones is highly variable and sensitive to many factors, including exposure and the linguistic variable itself. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://bibliotheek.ehb.be:2222/en-US/products/dissertations/individuals.shtml.]