This study examines the effects of the sequence of increasing task complexity in different modalities on the learning of the English past tense of Korean secondary learners. Robinson’s (2007) Cognition Hypothesis argued that learners pay more attention to grammatical forms in complex tasks than in less complex tasks. He suggested that tasks should be sequenced in such a way that resource-dispersing dimensions are first increased in complexity followed by an increase in the complexity of resource-directing dimensions. However, little empirical research has been done on how tasks are sequenced according to their cognitive complexity and how task modality affects second language development in the sequence. Fifty-four learners were divided into an integrated (writing with oral interaction) task group (EG 1), an oral-only task group (EG 2) and a comparison group (CG). After the sequence of six tasks was completed, one-way ANOVA revealed the EGs outperformed the CG significantly on the posttest. The mean score of EG 1 was the highest, while the improvement rate of EG 2 was the highest among the three groups. It is hoped that this result will contribute to building a solid basis on which practitioners can make decisions about sequencing tasks and implementing task modality.