The study of English intensifiers has been of interest in sociolinguistic research. This paper analyzes the variation of common intensifiers very and really in a corpus of Academic English and the predicting language-internal and –external factors. All factors are related to a speaker’s evaluation, specifically to the semantic notion of positive, negative, or neutral prosody. Using a variationst approach, this paper furthers insight to the grammaticalization of intensifiers and how the effect of delexicalization can predict the semantic properties of a modified adjective. The significant factors predicting very and really were academic setting, semantic prosody, academic discipline, and gender. The study further develops a methodological framework for operationalizing semantic prosody and producing quantitative results. Finally, the paper discusses its support for the use of smaller corpora to examine and compare linguistic effects in more specific registers.