For Phase 1, we used an existing set of archival data from the Word Generation project (Lawrence, Francis, Paré-Blagoev, & Snow, 2017). Collected in 2009, 2010, and 2011, in the Bay Area in California, the data include spring and fall assessments of academic vocabulary from over ten thousand middle school students.

The test had three different forms, each with 50 items. There was some overlap across test forms, resulting in 110 unique items. For the current study, we used items from two of the forms: 89 unique items in total.

Each item is in a multiple choice format, with a single correct response option (KEY) and three distractors. The student is asked to choose the response option that best matches the meaning of the underlined word in the item stem (the TARGET).

Some of the students who took the test were monolingual English speakers, but the majority were learning English, and spoke another language at home. They were classified by the district as either limited English proficient (LEP), initially fluent English proficient (IFEP), or reclassified fluent English proficient (RFEP).