This is the first empirical study to focus on adjectives complemented by 'that'-clauses. The in-depth analysis of more than 50,000 cases taken from the British National Corpus gives comprehensive insights into hitherto neglected relations of lexis and grammar. The result of this corpus-driven study is a novel classification of adjectives based on co- occurrence patterns and corroborated with the help of statistical means. The inductive analysis of corpus data offers new perspectives on and innovative descriptions of well-known phenomena of English grammar, such as extraposition or the resultative construction 'so…that'. It is based on a new methodological approach, which looks at mutual relations of both lexis and grammar in unprecedented ways.