It is well-documented that students' prior knowledge, cultural background, and language proficiency play a role in how they read, interpret, and respond to writing tasks (Barkaoui, 2007; Connor & Kramer, 1995; Hinkel, 2002). Essays written by students from different language backgrounds often differ in their linguistic, stylistic, and rhetorical characteristics and these features may affect the scores that students receive. This study explored the features of essays written on the SAT by students for whom English was not their best language (L2 students) and bilingual students who reported both English and another language as their best language, compared to students for whom English was solely their best language (L1 students). A sample of essay responses on 14 different prompts administered between October 2005 and January 2006 were coded on a variety of features including number of words, use of first-person voice, use of a five-paragraph structure, and types of examples offered (e.g., scholarly or personal experience). Multilevel logistic regression analyses indicated that L2 students had greater odds of using first-person and using personal experience in their essay responses; and had lesser odds of taking a mixed argument approach in responding to the essay prompt. There was substantial variability in the frequency of essay features by language group across prompt, suggesting that different prompts elicit responses with different features. The implications for test development and student instruction are discussed.