Portfolio assessment has become the predominant best practice in writing assessment at the college level. Despite its clear superiority to previous methods of assessment, portfolio assessment has brought its own collection of challenges. Although the stated goal of much portfolio grading is to create an overall, or holistic, judgment of a student's ability based on a collection of writings, we know from Hamp-Lyons and Condon that creating a truly holistic assessment of a portfolio is "highly unlikely, if not impossible" because instructors "inevitably...consider one text in light of another,...weigh one against the other,...and make decision(s) that, while representing a judgment about the whole portfolio, (are) grounded in a weighing of the parts, rather than in a dominant impression of the whole" (180). This nonholistic variation in approach to portfolio assessment is further complicated when the portfolio contains impromptu (timed, exam-style essays) as well as revised writing. Many departments and institutions choose to have students include a sample of impromptu writing in portfolios along with drafts of revised essays that were written during the semester, creating what can be called "heterogeneous" portfolios. There are numerous motivations behind this decision: to have an example of what students can produce on their own (in contrast to what they can produce with feedback from instructors and peers), to have an example of a different style of writing to compare to the others, to try to eliminate/reveal plagiarism in the revised essays, and to provide practice for upcoming standardized writing tests, among other reasons. The author's home English department implements a large, cross-graded programmatic writing assessment system in various levels of developmental courses that uses heterogeneous portfolios for some of the reasons articulated above. Her own experience participating in this assessment system and close observations of others' methods of reading portfolios led her to suspect that, as Hamp-Lyons and Condon have found, teachers treat the impromptu and revised writings very differently. Not only do teachers treat these items differently than one another, but different teachers also treat these two types of writing quite differently than other teachers do. Hamp-Lyons and Condon hypothesized that this situation would have been ameliorated if the revised essays in the portfolios in their study had included drafts; presumably, the inclusion of drafts of these essays might have persuaded the teachers to pay more attention to the students' writing process rather than focusing on final products. The current study found that teachers have very different ways of approaching and assessing heterogeneous portfolios. Although the portfolio system studied here does include multiple drafts of all revised essays, all teachers don't necessarily read all of the drafts that are presented. On a logistical and administrative level, these differences are concerning--the differences among teachers could make the whole assessment program quite unreliable. This study illuminates a number of directions for future research to help teachers refine ways of using portfolios in the evaluation and teaching of writing. (Contains 1 note.)