The relationship between validity and fairness has been heatedly debated in the literature (Kunnan, 2010). The orthodoxy is that test fairness should be subsumed under validity; that is, a valid test ensures fairness. Tracking the test retrofits of a high stakes national language test in Iran, known as Specialized English Test (SET) used to determine admission to tertiary English language programs, and collecting data on test takers’ language learning experiences, this article argues against the established view that more valid tests would necessarily promote fairness. Being an achievement measure based on secondary school English curriculum, the previous version of the SET was widely criticized for its construct underrepresentation (Farhady & Hedayati, 2009) and fizziness. In its current version, the SET is more construct representative, for it goes beyond the high school curriculum and covers more areas of communicative competence. Data collected from 173 undergraduate students of English translation and literature in three national universities across the country revealed that an overwhelming majority of students come from families with high socio-economic status, with poorer students represented only in low-tire university student population. This finding indicates that the improvement in validity has come with a cost in fairness and social mobility; hence, reproducing existing social order by denying underprivileged applicants access to quality tertiary education language programs. The paper further discusses issues of test validity and fairness and calls for a broader understanding of test consequences within a larger sociocultural perspective.