Response to Willems and Hagoort: an imperfect theory gets you further than random facts

Our article [1] offers a precise and detailed approach to the functional anatomy of Broca’s region and tackles key problems whose solution is a prerequisite for understanding the structure and function of this region (Table 1). We hope to have convinced some readers that empirical testing of current theories of the function(s) of this region is possible, and that among these theories, the syntactic movement account fares best. WillemsandHagoortmake twopoints onwhichweagree: (i) that Broca’s region is (probably) part of a larger set of networks and (ii) that we should think about how multiple functions are organized within Broca’s region. They also aptly observe that we focused on syntax, and argue that therearedata indicating thatBroca’s regionalsohas semantic functions.Wewould have loved to treat these results in a manner similar to those from syntax, however at this point we see no way of doing so. Although the picture of syntactic processing in Broca’s region is relatively clear, studies of ‘semantics’ do not make for such clarity, and Willems and Hagoort, alas, propose no elucidation. Although Willems and Hagoort argue for semantic involvement of Broca’s region, evoked response potential (ERP) patterns in ‘semantic’ experiments with patients sharply contradict this claim: deviation from normal is measured in ‘patients who have suffered strokes in the left temporal lobe or temporoparietal junction’, whereas ‘patients with damage restricted to the frontal lobe have typically shown normal semantic context effects’ [2]. Moreover, in healthy individuals, even when Broca’s region is implicated in semantics, it is one region of many [2,3], as inconsistent results tie an overly rich array of brain loci to ‘semantic’ tasks. This murky picture stands in stark contrast to the onewe drew for syntax.We suspect that, among