ABSTRACT Readers are expected to construct balanced mental representations of socioscientific issues discussed across controversial documents. However, readers tend to be biased toward documents that present belief-consistent perspectives and tend to refute documents that argue against their stance (text-belief consistency effect). Published studies on text-belief consistency effects have used imbalanced designs with all participants typically endorsing one standpoint in the controversy. The present experiment used a balanced design to examine the text-belief consistency in Iranian students of English as a foreign language and to investigate the extent that prior knowledge moderates the effect. Eighty-two students read two texts on an applied linguistics issue (native vs. non-native speakers as English as a foreign language teachers). Based on their performance on a prior beliefs measure, participants were assigned to three groups that varied in agreement to the stance of the texts. A recognition task was used to measure their situation-model strength and text-base strength. The results revealed a large text-belief consistency effect. Participants constructed stronger situation models for the text that communicated belief-consistent information compared with those who read the text that communicated belief-inconsistent information. No difference was found for text-base representation. Although prior knowledge was found to exert a significant positive effect on the strength of participants’ situation-model representations, it did not moderate the text-belief consistency effect.