The large GitHub dataset sample-size allows a mixed-methodsstudy of the effects of language, while controlling for a number ofconfounds. Through a combination of regression modeling, textanalytics, and visualization, we have examined the interactions oflanguage, domain, and defect type. The data indicates functionallanguages are better than procedural languages; it suggests thatstrong typing is better than weak typing; that static typing is betterthan dynamic; and that managed memory usage is better than un-managed. Further, that the defect proneness of languages in generalis not associated with software domains. Also, languages are morerelated to individual bug categories than bugs overall