The use of unnatural co-occurrence expressions by second language learners is caused by errors in word selection. The aim of this study is to develop a composition support system that can be used as an online dictionary by Japanese language learners to improve vocabulary knowledge that they tend to lack. We propose a method of providing candidates for appropriate alternative verbs to replace the verb of a learner's Noun-Case Particle-Verb co-occurrence input, or <n,c,v>. A verb of unnatural co-occurrence (error verb) is expected to occur in an environment similar to that of an appropriate verb (corrected error verb). Based on this hypothesis, this method displays candidates for appropriate alternative verbs estimated to be naturally co-occurring with the noun of the learner's co-occurrence input using statistical information obtained from a large-scale Japanese corpus, in descending order of similarity of occurring environments with the original verb. Experimental results show that the rate of test items having appropriate verbs in the top 30 candidates provided in the proposed method output to 260 test items (learners' co-occurrence data including errors in verb selection) was 70%. We discuss the usefulness of such a composition support system for co-occurring expressions based on the proposed method.