An experimental design was adopted in the present study to cross-compare the effect of two formative assessments, namely peer assessment (PA) and teacher assessment (TA), with summative assessment (SA) on the improvement of language learners' writing skill and self-efficacy. Writing excellence was operationalized by structural (grammar), lexical (word choice), and discoursal (cohesion and coherence) well-formedness. Furthermore, focused group interview was run to elicit the learners' attitudes and preferences toward type of assessment. Three cohorts of proficiency-matched language learners sat for a semester-long experimentation (20 sessions × 90 min = 30 h) and three testing sessions. The results revealed that PA group significantly improved in the writing skill, but TA and SA groups did not. Self-efficacy was not enhanced in any of the assessment groups under study. The interviews and follow-up member checkings indicated that the pupils did not have positive attitudes toward PA and preferred TA. Among the sources of this disfavor, they referred to misunderstanding, lack of knowledge or attention, tiredness, poor handwriting of the assesse, shyness, friendship bias, and even feeling not qualified enough to judge their peers' tasks. Overall, then, the study reported a mismatch between the students' actual performance and attitudinal beliefs under different assessment conditions, implying that (a) Iranian students are more inclined toward teacher-dominated pedagogies and (b) triangulation of assessment modes provides more reliable optimal results.