My work on formal language theory, theoretical syntax, and the formal foundations of natural language grammars stems from the necessity to provide consistent theoretical accounts for a range of phenomena observed in natural languages, mainly English and Spanish, that have proven challenging for theories based on structural uniformity through proof theoretic derivations. The core of my work argues for a ‘mixed computation’ approach to natural language syntax, whereby sentences are assigned structural descriptions whose computational properties vary locally as a function of compositional semantic interpretation: changes in computational dependencies delimit local syntactic cycles. For example, in a string like some old old men, the adjective iteration can be characterised by a regular grammar without loss of information (since none of the adjectives modifies or has scope over another) but the relation between the quantifier and the noun requires a richer structure. This system lends itself to a lexicalised treatment, and led me to explore aspects of procedural structure composition as well as derivation-based and constraint-based ways to capture the data; my recent work has focused primarily on developing an approach that can capture the computational variety exemplified above, while maintaining the advantages of lexicalised (tree adjoining) grammars. I also collaborate actively with researchers in the universities of Verona (Italy) and Reading (UK) on developing Artificial Grammar Learning protocols using Lindenmayer systems, to probe for human sensitivity to hierarchical structure and the mechanisms whereby information is extracted from (non-linguistic) signals.