Fundamental concepts in physics such as Newtonian mechanics are surprisingly difficult to learn and discover. Over the past decade we have used an educational neuroscience approach to science education to investigate the different ways that scientific concepts are invoked or activated in different contexts. In particular, we have sought to determine how networks of brain regions that are highly sensitive to the context in which they are used are involved in the use of scientific concepts. We have found that some physics concepts that are highly tuned to perception are often inhibited in experts (with increased activations in error detection and inhibitory networks of the prefrontal cortex). Other concepts, such as those involved in perceptual causality, can activate highly diverse brain regions depending on task instructions. For example, when students are shown movies of balls colliding, we find increased activation in the right parietal lobe, yet when the students see the exact same movies and are told that these are positively charged particles repulsing we find increased activations in the temporal lobe that is consistent with the students retrieving semantic information. We also see similar changes in activation patterns in students learning about phase shifts in chemistry classes. A key component of both students and scientists' discourse and reasoning is analogical thinking. Our recent fMRI work indicates that categorization is a key component of this type of reasoning that helps bind superficially different concepts together in the service of reasoning about the causes of unexpected findings. Taken together, these results are allowing us to make
insights into the contextually relevant networks of knowledge that are activated during learning. This work is
allowing us to propose why some educational interventions are more successful than others and why certain types
of educational interventions are appropriate for some contexts, but not others.