In this study, an investigation into the students' English Placement Test (EPT) performances at Iowa State University were compared with their self-assessments and instructors' judgments of students' writing. This investigation was framed within an argument-based approach to validity framework of Li (2015) and Chapelle et al. (2008), with a particular focus on the extrapolation inference. Surveys and interviews were used to investigate how 92 undergraduate ESL students enrolled in 101B (Academic Writing I) and 101C (Academic Writing II) ESL writing courses self-assessed their writing proficiency, and how six instructors judged these students' abilities. Survey data were used to conduct binary logistic regression analysis in which placement levels were predicted using self-assessment. Interview data were used to confirm the trends found in the survey data and these were qualitatively analyzed using APPRAISAL analysis. Results show that self-evaluations and judgments about students' writing process and grammar and lexis were found to be statistically significant in predicting placement level although with very small effects. Qualitatively, the 101C students and instructors were slightly more critical than the 101B students and instructors toward students' writing process, and conversely, more lenient toward students' grammar and lexis. The findings do not support the extrapolation assumptions but rather the conditions of rebuttal of the extrapolation inference. The findings demonstrate that incorporating self-assessment and instructors' judgments for placement test validation purposes may be challenging in part because the ability range between 101B and 101C is narrow. Locally, the findings from this research can help ESL writing educators understand and use the self-perceptions and attitudes that undergraduate non-native speakers of English may hold toward their own writing.