The mental representation of language cannot be directly observed but must be inferred and modelled from its effects at second hand. Linguists have traditionally responded to this in two ways, either going for a fairly data-light approach and valuing theoretical creativity, or pursuing just those goals for which data is available and trusting to data-driven descriptive work.

More recently, advances in technology and experimental techniques have made data gathering easier and more accessible, so that a theoretically informed but empirically based approach is rapidly growing in popularity....