A Madness to the Methods in Cognitive Neuroscience?

Abstract

Since Paul Broca, the relationship between mind and brain has been the central preoccupation of cognitive neuroscience. In the 19th century, recognition that mental faculties might be understood by observations of individuals with brain damage led to vigorous debates about the properties of mind. By the end of the First World War, neurologists had outlined basic frameworks for the neural organization of language, perception, and motor cognition. Geschwind revived these frameworks in the 1960s and by the 1980s, lesion studies had incorporated methods from experimental psychology, models from cognitive science, formalities from computational approaches, and early developments in structural brain imaging. Around the same time, functional neuroimaging entered the scene. Early xenon probes evolved to the present-day wonders of BOLD and perfusion imaging. In a quick two decades, driven by these technical advances, centers for cognitive neuroscience now dot the landscape, journals such as this one are thriving, and the annual meeting of the Society for Cognitive Neuroscience is overflowing. In these heady times, a group of young cognitive neuroscientists training at a center in which human lesion studies and functional neuroimaging are pursued with similar vigor inquire about the relative impact of these two methods on the field. Fellows and colleagues, in their article titled ‘‘Method matters: An empirical study of impact on cognitive neuroscience,’’ point out that the nature of the evidence derived from the two methods are different. Importantly, they have complementary strengths and weaknesses. A critical difference highlighted in their article is that functional imaging by necessity provides correlational data, whereas lesion studies can support necessity claims for a specific brain region in a particular function. The authors hypothesize that despite the obvious growth of functional imaging in the last decade or so, lesion studies would have a disproportionate impact on cognitive neuroscience because they offer the possibility of establishing a causal role for structure in behavior in a way that is difficult to establish using functional imaging. The authors did not confirm this hypothesis. Using bibliometric methods, they found that functional imaging studies were cited three times as often as lesion studies, in large part because imaging studies were more likely to be published in high-impact journals. Given the complementary nature of the evidence from both methods, they anticipated extensive cross-method references. However, they found a within-method bias to citations generally, and, furthermore, functional imaging articles cited lesion studies considerably less often than the converse. To confirm the trends indicated by Fellows and colleagues, I looked at the distribution of cognitive neuroscience methods in the abstracts accepted for the 2005 Annual Meeting of the Cognitive Neuroscience Society (see Figure 1). Imaging studies composed over a third of all abstracts, followed by electrophysiological studies, the bulk of which were event-related potential (ERP) and magnetoencephalogram (MEG) studies. Studies that used patient populations composed 16% of the abstracts. The patient studies were almost evenly split between those focused on understanding a disease (47%), such as autism or schizophrenia, and those in which structure–function relationships were a consideration (53%). These observations do not speak of the final impact of these studies, but they do point out the relative lack of patient-based studies, particularly those addressing basic cognitive neuroscience questions. Fellows and colleagues pose the following question: Despite the greater ‘‘in-principle’’ inferential strength of lesion than functional imaging studies, why in practice do they have less impact on the field? They suggest that sociologic and practical considerations, rather than scientific merit, might be at play. Here, I offer my speculations on the factors that contribute to the relative impact of these methods. These speculations are not intended to be comprehensive. Rather they are intended to begin conversations in response to the question posed by Fellows and colleagues. In my view, the disproportionate impact of functional imaging compared to lesion studies is driven by three factors: the appeal of novelty and technology, by ease of access to neural data, and, in a subtle way, to the pragmatics of hypothesis testing. First, novelty is intrinsically appealing. As a clinician, I often encounter patients requesting the latest medications, even when they are more expensive and not demonstrably better than older ones. As scions of the enlightenment, many of us believe in progress, and that things newer are generally things better. Lesion studies have been around for a century and a half. Any advances made now are likely to be incremental. By contrast, functional imaging is truly a new way to examine the University of Pennsylvania