Those ideas affect more than academic diagrams. If the mind routinely represents hierarchical groupings and placeholders, then language learners—children, second-language students, and people recovering from brain injury—may rely on abstract scaffolding that goes beyond linear word relationships. That perspective also changes how researchers build models of sentence comprehension and how speech technologies handle questions, relative clauses, and elided material. Testing whether constituency and hidden elements are real helps us choose models that reflect actual human capacities.
Follow the link to see how new evidence weighs in on these competing models and what experiments reveal about movement and unpronounced structure. Understanding which architecture better fits human behavior will shape teaching methods, clinical approaches, and inclusive language technologies that aim to support diverse communicative needs.
Gibson advocates dependency grammar [1], contrasting it with constituency-based accounts that posit movement and phonologically null (unpronounced) syntactic elements [2,3]. Dependency grammar represents sentence structure through word-to-word relations, whereas constituency-based accounts represent sentence structure by grouping words into larger hierarchical constituents. Movement accounts for cases in which a constituent appears in one position but is interpreted in another. For example, in ‘Who did you call?’, ‘Who’ appears sentence-initially but is interpreted as the object of ‘call’; this is analysed as movement from the object position to the sentence-initial position.
