The primary purpose of the present study was to examine the effect of explicit instruction and feedback on the development of lexis and cohesion in second language writing as well as the extent to which these two measures correlated with overall writing quality scores. The participants were two intact classes of first-year Japanese university students who attended class twice a week for 14 weeks. The classes were randomly assigned to an explicit instruction group (n = 34) and a comparison group (n = 31). The explicit group received instruction and feedback on the use of target vocabulary and cohesion, while the comparison group only submitted drafts. Both groups produced three drafts of a comparison-contrast essay for Task 1 and a problem-solution essay for Task 2 during one academic semester. Based on their writing performance, six participants were selected for interviews about their learning experiences to complement the results of the quantitative analyses. Writing samples from each participant were analyzed with TALLES (Kyle & Crossley, 2015), TAALED 1.41 (Zenker & Kyle, 2021), and TAACO 2.0 (Crossley et al., 2016). Each draft was assessed for the number of target words, three lexical complexity metrics— MATTR 50 content words (MATTR), COCA academic frequency logarithm content words (COCAFrq), and COCA academic trigram MI2 (Trigrams) —two cohesion indices—All connectives and Adjacent overlap between paragraphs content words (AdjacentP)—and overall writing quality produced with a writing rubric. Six raters trained in applied linguistics assessed the participants’ essays, and FACETS 3.80.0 (Linacre, 2017) was used to produce interval measures of student ability and rater severity. Five hypotheses were assessed. Hypothesis 1 predicted that the explicit instruction group would score higher than the comparison group on target word and cohesion use, and overall writing quality. The results supported Hypothesis 1 for both tasks. The explicit instruction group increasingly used target words and cohesion based on descriptive statistics. Both groups significantly improved overall writing quality over time. The explicit instruction group significantly scored higher. The comparison group was slower in learning about writing essays during Task 1. Both groups scored higher in Task 2. Hypothesis 2, which predicted that both groups would improve on the lexical indices and that the explicit instruction group would exceed the comparison group, was partially supported. MATTR and COCAFrq significantly improved, but Trigram did not improve. There were significant group differences in COCAFrq and Trigrams, but not in MATTR. The comparison group scored higher on COCAFrq, and the explicit instruction group scored higher on Trigram. In Task 2, MATTR improved significantly, but COCAFrq and Trigrams did not. There were no group differences. More effects were found in Task 1 than in Task 2. Hypothesis 3, which predicted that cohesion would improve over time for both groups and that the explicit instruction group would exceed the comparison group, was mostly supported. The results of Task 1 showed that All connectives and AdjacentP significantly improved. Although there was no group significant difference in All connectives, there was a group difference in AdjacentP, as the explicit instruction group scored higher. The results of Task 2 showed that All connectives did not significantly improve over time, but AdjacentP did. Global cohesion was affected more than local cohesion. Hypothesis 4, which predicted that lexis would be positively correlated with overall writing quality for both groups, was supported for Task 1 and not supported for Task 2. In Task 1, MATTR was significantly positively correlated with overall writing quality. COCAFrq was significantly negatively correlated. Trigrams were not correlated. In Task 2, none of lexical indices were correlated. Hypothesis 5 predicted that cohesion would be positively correlated with overall writing quality. This hypothesis was partially supported in Task 1 and Task 2. All connectives were not correlated with overall writing quality, but AdjacentP was significantly and positively correlated with overall writing quality in Task 2. The study contributes to the understanding of the development and assessment of lexis and cohesion using computational automated tools. L2 development is a complex phenomenon, so a further examination of assessment indices offers a wealth of research in future studies. / Applied Linguistics