This paper argues that the universal categories N/V are not applied to content words before the grammatical markings for reference D(eterminers) and predication I(nflection) have been acquired (van Kampen, 1997, contra Pinker, 1984). Child grammar starts as proto-grammar with language-specific operators and category-neutral content words Xo. The first content words are used as proper names with topic-intention and as brand names with characterizing-intention. The language-specificity of early operators may be illustrated by the use of wh-pronouns (what/where?) in English versus the use of illocution particles (nou/denn?) in Dutch/German child language. Subsequently, deictic operators for topic are regularized as Do and deictic operators for comment as Io. These functional categories are the bootstraps for category assignment <+V> and <+N>, as a subdivision of the lexicon. The generalized conclusion is that language-specific systems are not acquired due to a common UG entrance. They rather are highly frequent language-specific bootstraps that coax the child into an adult system that eventually fits UG principles. The operators in proto-grammar are situation-related, whereas the functional categories Io/D o are highly sensitive to syntactic context. This explains why, in a bilingual situation, a language switch in the context-free operators will be relatively easy. A language switch in functional categories, by contrast, will have a reflex on the syntactic context and be less easy.