The present study explores phonetic reduction in the Buckeye Corpus [Pitt et al., Speech. Commun. 45, 89–95 (2005)] following up on the work of Johnson [Proceedings of the 1st Session 10th International Symposiam (2004), pp.], who counted the number of segments and syllables deleted from each word in the subset of the corpus that was available at the time. The first experiment presented here provides updated rates of segment deletion (over 25% of words) and syllable deletion (over 6% of words) as measures of reduction rates in the entire completed corpus. The second experiment investigates reduction in the duration of words as a function of several linguistic factors, including frequency, conditional probability, and rate of speech. The data are modeled using multiple mixed-effect linear regression to predict both the duration of each word and the reduction from the median duration of each word. Age, gender, and interviewer gender were not found to affect word duration. Citation length and current rate of speech were both strong predictors of word duration. Frequency and predictability from context also correlate strongly with word duration. Verbs were found to be significantly shorter in duration than other content words with the same citation length.