A semi-conditional grammar is a form of regulated rewriting system. Each rule consists of a context-free core rule A → w and two strings w+, w−; the rule is applicable if w+ (the positive condition) occurs as a substring of the current sentential form, but w− (the negative condition) does not. The maximum lengths i, j of the positive or negative conditional strings, respectively, give a natural measure of descriptional complexity, known as the degree of such grammars. Employing several normal form results on phrase-structure grammars as derived by Geffert, we improve on previously obtained results by reducing the number of nonterminals of semi-conditional grammars of a given degree (i, j) while maintaining computational completeness of the said mechanisms. © 2018, Springer International Publishing AG, part of Springer Nature.