This corpus study investigates well-known violations of the verb second constraint in Old High German main clauses. Starting from the observation that the attested orders are overwhelmingly ambiguous regarding the position of the verb and the preverbal constituents, the paper applies methods for identifying diagnostic datasets which can be clearly distinguished in terms of structure and use. Two types of diagnostic non-verb-second orders are tested, namely verb-final clauses in which no verb movement applies, and verb-third clauses in which the verb as well as all preverbal categories occupy specified positions in the left periphery of the clause. The result is that diagnostic evidence for any of the datasets is sparse in native prose, and that especially diagnostic verb-third orders are not sufficiently distinguishable against their Latin originals. In addition, both diagnostic patterns interfere in use, suggesting that they represent less clearly differentiated, unproductive patterns, used for special stylistic purposes.