Research on the neurocognitive foundations of reading in typically and atypically developing readers has benefited in recent years from advances in several neuroimaging technologies (see Papanicolaou, Pugh, Simos, & Mencl, 2004, for a review). In this chapter we describe recent studies from our lab, and from others, that were designed to generate data, not only on localization of readingrelated brain activation, but also to examine patterns of interactions and tradeoffs among these distributed reading-related systems. Acquiring data of this sort is necessary, in our view, if we are to begin to construct neurobiological models of word identification that can speak to the complexities and dynamics of reading performance. Indeed, computational models of reading stand or fall, not by their capacity to account for simple group differences on behavioral measures of reading performance or main effects of isolated variables, but rather by whether they can account for complex interactions among them (Harm & Seidenberg, 1999). Ultimately, the same criteria must be applied to neurobiologically grounded models as well. Thus, it is critical that we begin to look beyond simply more fine-tuned localization and consider also a systems-level approach. Research on systems appears to be a realistic possibility at this stage in the development of these new technologies, given the evidence that extant neurophysiologic measures are amenable to sophisticated psycholinguistic designs (Dehaene et al., 2004; Frost et al., 2005; Sandak, Mencl, Frost, Mason, Rueckl, et al., 2004). When we add to all this the clinically oriented goal of better understanding what differences in activation patterns between skilled and struggling readers imply about core deficits and optimal remediation, it becomes all the more pressing to develop a more comprehensive neurobiological account of how component processes interact and change with learning and development (Pugh et al., 2006). We focus here primarily on studies of phonological processing in reading, given that this stage is typically the
most sensitive in discriminating nonimpaired (NI) from reading disabled (RD) cohorts (Pugh, Mencl, Shaywitz, et al., 2000).