Many English language proficiency (ELP) tests used for university admissions and placement now include integrated writing tasks that require examinees to use external sources when writing. Integrated writing tasks improve test authenticity and impact, but they raise several validity questions, such as what academic language skills they engage and whether performance on these tasks varies with examinee ELP level. This study addresses these questions with reference to an integrated writing task from the Canadian Academic English Language (CAEL) Test that involves reading, listening, and writing in academic contexts. Responses by 59 students to one of the CAEL integrated writing tasks are analyzed in terms of various grammatical, discourse, sociolinguistic, strategic, content, and source use aspects and compared across ELP levels (high and low) and score levels. The findings indicate that both ELP level and score level had significant effects on most writing features examined in the study, except for syntactic complexity. Additionally, except for syntactic complexity, all writing dimensions examined in the study were significantly associated with writing scores. The findings and their implications for the validity argument of source-based writing tasks are discussed. De nombreux tests de comp & eacute;tence linguistique en anglais utilis & eacute;s pour les admissions universitaires et l'attribution de stages comprennent d & eacute;sormais des & eacute;preuves & eacute;crites int & eacute;gr & eacute;es. Celles-ci exigent que les candidat<middle dot>e<middle dot>s fassent appel & agrave; des sources externes au moment de la r & eacute;daction. Si ce type d'& eacute;preuve am & eacute;liore la repr & eacute;sentativit & eacute; et l'incidence d'un test, il soul & egrave;ve toutefois des questions en mati & egrave;re de validit & eacute;; quel est le niveau de langage universitaire mobilis & eacute; et dans quelle mesure la r & eacute;ussite de cette & eacute;preuve d & eacute;pend-elle des aptitudes linguistiques des candidat<middle dot>e<middle dot>s? Cet article aborde ces questions en se r & eacute;f & eacute;rant & agrave; l'& eacute;preuve & eacute;crite int & eacute;gr & eacute;e du test CAEL (Canadian Academic English Language) qui comprend des sections de compr & eacute;hension & eacute;crite, de compr & eacute;hension orale et d'expression & eacute;crite en contexte universitaire. Les r & eacute;ponses des & eacute;tudiant<middle dot>e<middle dot>s & agrave; l'une des & eacute;preuves & eacute;crites int & eacute;gr & eacute;es du CAEL font l'objet d'une analyse selon diff & eacute;rents crit & egrave;res d'ordre grammatical, discursif, sociolinguistique, et strat & eacute;gique, ainsi que divers aspects li & eacute;s au contenu et & agrave; l'utilisation des sources. Ces donn & eacute;es sont ensuite compar & eacute;es aux & eacute;chelons de ma & icirc;trise de l'anglais (bons et faibles) du test et aux r & eacute;sultats des candidat<middle dot>e<middle dot>s. Les conclusions montrent que le niveau de ma & icirc;trise de la langue et le classement final ont eu une incidence significative sur la plupart des crit & egrave;res r & eacute;dactionnels examin & eacute;s dans l'& eacute;tude, & agrave; l'exception de la complexit & eacute; syntaxique. Hormis ce dernier & eacute;l & eacute;ment, tous les aspects d'& eacute;criture analys & eacute;s dans le cadre de cette recherche ont & eacute;t & eacute; fortement li & eacute;s aux r & eacute;sultats obtenus en expression & eacute;crite. Les conclusions et leur incidence sur l'argument de la validit & eacute; des & eacute;preuves d'expression & eacute;crite & agrave; partir de sources sont examin & eacute;es.
