This paper examines the use of portfolio evaluation in accrediting higher education teachers in the United Kingdom. It explains portfolios as an approach to authentic assessment and discusses the issue of reliability in portfolio assessment when teacher accreditation depends on it. Researchers investigated one course, collecting data on 53 assessments from college records. The portfolios of 20 of those were regraded by trained assessors. The original 53 assessments were subjected to various statistical tests. The percentages of exact and close agreement were computed for six groups of items from the 74 comprising portfolio assessment. The experimental assessors marked more harshly than the original assessors, so subsequent findings were restricted to the original assessments. Interrater correlations ranged from -0.010 to 0.67. Patterns of agreement between the pairs of raters were significantly different from what would have been expected as a result of chance. Data showed problems with the components oriented toward analysis of needs and planning for future professional development. The valuing of equal opportunities was shown to be problematic. Assessors' judgments on the topic may vary considerably. (Contains 16 references.) (SM)