In recent years, gloss presentation format or the location where a gloss appears with respect to its related target word has received renewed attention. Research suggested that different gloss presentation formats could have differential effects on reading comprehension and/or vocabulary learning. This study hypothesized that the effectiveness of different computerized gloss presentation formats in reading comprehension could be explained by drawing on the split-attention effect within the framework of cognitive load theory. The effect predicts that when two related sources of information are physically separated (e.g., a target word and the respective gloss), cognitive resources are unnecessarily wasted, and learning is hindered. To test this hypothesis, 39 Persian-speaking L2 learners of English were divided into two experimental conditions, each being exposed to a text enhanced with either in-text or marginal glosses. Two measures of reading comprehension and three measures of cognitive load were employed. The participants' initial differences in terms of grammatical knowledge and vocabulary size were balanced out, and their lookup behavior was also tracked. The results revealed that the participants with access to in-text glosses, compared with those with marginal glosses, experienced lower levels of cognitive load due to the elimination of split-attention and, accordingly, performed better on the reading comprehension measures. Given the participants' L2 proficiency level, the findings suggested that a text enhanced with in-text glosses tends to be instructionally more efficient.