The mystery of instruction following: Computational modeling and EEG studies
People can implement without effort instructions that are typically completely novel (e.g., U+201CPlease give me the blue cupU+201D). How they do so, is still a mystery. In the current project this will be addressed from a conceptual framework that has not been used for this purpose, called binding by synchrony. Binding by synchrony has proven very fruitful in the perception and attention literatures. Binding by synchrony entails that elements (e.g., U+201CblueU+201D and U+201CcupU+201D) are bound together in the brain by letting the neurons corresponding to the respective elements fire at the same time (U+201Cin phaseU+201D). This principle is eminently suited to implement completely novel information, such as instructions, because it allows binding and unbinding of elements extremely rapidly. We will construct a computational model of instruction following that implements the principle. We will also consider how such a model can be trained by reinforcement learning. In a next step, we will perform EEG studies both to test the general principle that instructions are implemented using binding by
synchrony, and also more specific model predictions.