This study investigated the accuracy of scores assigned by self-, peer-, and teacher assessors over time. Thirty-three English majors who were taking paragraph development course at Vali-e-Asr University of Rafsanjan and two instructors who had been teaching essay writing for at least two years at university, participated in the study. After receiving instructions on paragraph development, participants were trained for a session on how to rate the paragraphs. For three sessions the students were given topics to write about and were asked to rate their own and one of their peers’ papers for mechanics, grammar and choice of words, content development, and organization. The teachers also rated the paragraphs according to the same criteria. Multifaceted Rasch measurement was employed to analyze the data. The results showed different patterns of performance for the subjects rated by different raters at the beginning of the experiment. However, rater bias showed significant decrease across time. The results of the study have useful implications for language teachers especially in portfolio assessment where self and peer assessment provide invaluable help.
