A study by d'Ydewalle (1991) indicated that all readers (including illiterate, emergent, and struggling readers) will (both consciously and unconsciously) track available text. In other words, the student’s eyes will register the text, even if the student is not intentionally trying to read the captioning. d'Ydewalle used cameras to track and record eye focus during video activities with and without captioning-- the patterns are similar to models presented below.
This is why 'hard' subtitling is so important. Students subconciously track the text on screen ---- reading practice becomes automatic. The problem is that the subtitling needs to be on media that students will view and view repeatedly --MUSIC VIDEO!