The purpose of this paper is to investigate the variation of alveolar plosives between vowels and glides across word boundaries in American English using data from the Buckeye Corpus of Conversational Speech (Pitt et al. 2007) and a Maximum Entropy grammar (Hayes and Wilson 2008) framework. Through corpus analysis, we found that {t,d}-deletion, glottalization, flapping, and palatalization were the main variants in the sequences. A total of 5 types of variation were set by adding the case of no variation, and the distribution of variation according to voice of alveolar plosives and the type of glides (/w/ or /j/) that followed was examined. As a result, we observed that /t/ followed by /w/ predominantly showed glottalization, while with /d/ most remained unchanged. The /t/ and /j/ sequence revealed a rich set of variations, whereas /d/ followed by /j/ mostly remained unchanged or palatalized. Utilizing the MaxEnt grammar, we used the observed frequencies, constraint violations, and constraint weights to predict variant probability distributions. The probabilities predicted by MaxEnt grammar largely mirror the distributions observed in the Buckeye corpus, though some minor discrepancies were noted.