Disconnected brains: What is the role of fMRI in connectivity research? Ngoc Jade Thai ⁎, Olivia Longe, Gina Rippon Neuroscience Research, School of Life and Health Sciences, Aston University, UK abstract article info Article history: Received 15 August 2008 Received in revised form 3 December 2008 Accepted 23 December 2008 Available online 10 February 2009 Keywords: Functional connectivity Autism Functional Magnetic Resonance Imaging EEG MEG In this paper we consider how functional Magnetic Resonance Imaging (fMRI) has been used to study cortical connectivity in autism and autistic spectrum disorders (ASD). We discuss those studies that have contributed to the evidence supporting a model of disordered cortical connectivity in autism and (ASD), with a focusing emphasis on the application to research into the underconnectivity model. We note that the analytical techniques employed are limited and do not allow interpretation in terms of effective, or directional connectivity, nor do they provide information about the temporal or spectral characteristics of the functional networks being studied. We highlight how currently the features of neural generators that are being assessed by functional connectivity in fMRI are unclear. In addition, we note the importance in clinical studies of considering the consequences of task choice for the nature of the imaging data that can be collected and also of individual differences in participant state and trait characteristics for the accurate interpretation of imaging data. We discuss how alternative techniques such as EEG/MEG may address the limitations of fMRI in assessing brain connectivity, and additionally consider the potential of multimodal approaches. We conclude that fMRI has made significant contributions towards our understanding of the brain in terms of neural systems but that the conclusions drawn from its application in the sphere of autism research need to be approached with caution. It is important in research of this kind that we are aware of the need to examine the methodological and analytical techniques closely when applying findings in clinical populations, not only when they are used to support the development of theoretical models but also to inform diagnostic or treatment decisions. © 2009 Elsevier B.V. All rights reserved. 1. Connectivity in the human brain Contemporary models of brain function stress the coordinated interconnections within and between networks of neurons, large- and small-scale, often only transiently coupled, with both long-distance and local pathways providing this interconnectivity. Structural con- nectivity measures in the human brain frequently focus on the integrity (or otherwise) of white matter tracts, using techniques such as Diffusion Tensor Imaging (e.g. Hagmann et al., 2007). The main emphasis is on spatial information and the accurate and fine-tuned identification of key structures linked by measured neuronal path- ways. Functional connectivity refers to dynamic relationships between brain regions underpinning specific behavioural processes; this is commonly measured by correlational analyses of varying degrees of sophistication, and may refer to task-specific networks or ‘default mode’ networks, evident in the ‘resting state’ activity of the brain (Raichle et al., 2001; Greicius et al., 2003). Again the emphasis is on spatial information and the accurate location of those structures whose activation is correlated and coordinated in some way. Clearly, there will be a close relationship between structural and functional connectivity, recently demonstrated, for example, by Hagmann et al. (2008). Measures of effective connectivity infer a causal, modulatory relationship between regions or networks and describe directional, temporal interactions. These are potentially the most exciting developments in this area, but increase the analytical demands (Friston, 1994) as will be discussed below. Although spatial informa- tion remains important for the identification of network members, some measure/assumption of temporal ordering is also required to ascertain the sequencing of activation changes and to confirm the feed-forward, feedback aspects of the modulations. More recently, there has been some focus on the role of spectral information in assessing connectivity, with the notion that functional connectivity is, in fact, frequency specific and that the coupling of network members is achieved by synchronization of their particular oscillations (Sauseng and Klimesch, 2008). This approach was initially based on techniques which gave direct access to spectral information such as EEG or MEG; simpler correlational or coherence measures between activated areas were initially applied to provide assessments of connectivity; but more sophisticated processing approaches based on estimations of directed causality between brain areas are now available, such as Granger causality or Synchronization Likelihood (Stam, 2005; Kujala et al., 2008). A recent review based on EEG International Journal of Psychophysiology 73 (2009) 27–32 ⁎ Corresponding author. Neuroscience Research, Vision Sciences Building, School of Life and Health Sciences, Aston University, Aston Triangle, Birmingham B4 7ET, UK. Tel.: +44 121 204 3865; fax: +44 121 204 4090. E-mail address: j.n.thai@aston.ac.uk (N.J. Thai). 0167-8760/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpsycho.2008.12.015 Contents lists available at ScienceDirect International Journal of Psychophysiology journal homepage: www.elsevier.com/locate/ijpsycho