The author explored text difficulty for third-grade bilingual students reading below grade level. In six small-group discussions, students read and discussed informational texts matched to their reading levels, and in six other sessions they read and discussed texts one year ahead of their reading levels. The alternating treatment design and qualitative analysis of transcripts revealed different text levels have both advantages and drawbacks. Findings included the following: Group 1 had more inferential or interpretive responses with matched texts and Group 2 had more inferential or interpretive responses with difficult texts. Most students participated evenly regardless of text difficulty. However, two students talked more when discussing matched texts. Text difficulty did not affect comprehension for anyone except one student, and for him it only had an effect on three of 12 days. Half of the students' fluency scores benefited from matched texts. The other half read difficult texts with similar fluency to matched texts.