fMRI and its interpretations: an illustration on directional selectivity in area V5/MT Andreas Bartels 1 , Nikos K. Logothetis 1, 2 and Konstantinos Moutoussis 3 1 Max Planck Institute for Biological Cybernetics, 72076 Tu ¨ bingen, Germany 2 Imaging Science and Biomedical Engineering, University of Manchester, Manchester M13 9PT, UK 3 Department of Philosophy and History of Science, University of Athens, 15771 Athens, Greece fMRI is a tool to study brain function noninvasively that can reliably identify sites of neural involvement for a given task. However, to what extent can fMRI signals be related to measures obtained in electrophysiology? Can the blood-oxygen-level-dependent signal be interpreted as spatially pooled spiking activity? Here we combine knowledge from neurovascular coupling, functional ima- ging and neurophysiology to discuss whether fMRI has succeeded in demonstrating one of the most established functional properties in the visual brain, namely direc- tional selectivity in the motion-processing region V5/ MT+. We also discuss differences of fMRI and electro- physiology in their sensitivity to distinct physiological processes. We conclude that fMRI constitutes a comp- lement, not a poor-resolution substitute, to invasive techniques, and that it deserves interpretations that acknowledge its stand as a separate signal. Introduction Functional magnetic resonance imaging (fMRI) produces reliable and reproducible results in various fields of research. Typically, fMRI signals are thought to represent changes in the activity of the neuronal populations respon- sible for the task at hand. This assumption has historical rather than scientific origins, as various degrees of stimulus or task selectivity have long been demonstrated with intracortical recordings from isolated single neurons in experimental animals. fMRI activation is thus pre- sumed to reflect an increase in the spiking rate of those specialized neurons underlying the subject’s behavior. Yet, research using diverse methodologies has also demon- strated that hemodynamic responses are more sensitive to integrative dendro-somatic processes than the spiking of a few stimulus- or task-selective cortical projection neurons. In other words, strong evidence exists that although fMRI undoubtedly provides valuable information regarding regional changes of neural activity, it does so by stressing neural processes that might be different from those reported in invasive animal experiments. In this review, we examine to what extent common fMRI data reflect the functional properties of neurons by con- sidering a special example: the neuroimaging data on neuronal directional selectivity, initially reported in electrophysiological studies, in cortical area V5 (also known as MT). The aim is to show how imaging data can or should be interpreted. Can fMRI be seen to provide a pooled (yet whole-brain) version of signals otherwise provided by electrophysiology? Can clever experimental paradigms allow us to infer neuronal spiking properties from blood-oxygen-level-dependent (BOLD) signals? Can they circumvent the limits given by its spatial resolution? Given the exponential growth of noninvasive neuroima- ging experiments in human, and given their ethical advan- tages over electrophysiology, these questions are central to basic science and to policymaking. With this aim, this review attempts to combine knowledge from neurophysiol- ogy, applied fMRI and neurovascular coupling (NVC) to address the above questions. We use concrete examples from functional studies and focus mainly on the cortical region of V5+/MT+, as it has been one of the most intensely studied regions both in physiology and imaging. We hope that any conclusions reached will have more general implications. Functional properties of area V5 Many years of electrophysiological research have convin- cingly demonstrated that area V5 of the monkey brain contains an abundance of neurons selective for the direc- tion of movement of the visual stimulus (e.g. see reviews in Refs [1–4]). In addition, single-cell recordings and micro- stimulation experiments also suggest that V5 neurons play a direct role in the perception of motion direction and speed, because spontaneous or electrically induced fluctu- ations of activity correlate with behavioral performance (e.g. see Refs [5,6]). In accordance with such data, lesion studies demon- strated that damage in V5+ causes specifically a loss in the ability to perceive motion, while otherwise largely preser- ving normal visual performance in both human and monkey (for reviews, see Refs [3,4]). V5/MT is adjacent to MSTd and MSTv, areas that process more complex types of motion, such as visual flow in the case of MSTd. V5+ or MT+ refers to the group of regions. They project to higher- level regions such as VIP, which are thought to process flow and to integrate signals with cues of other modalities [7,8]. Functional imaging techniques such as positron emis- sion tomography (PET) and fMRI identified the homologue of the V5+/MT+ complex in the human brain functionally Review Corresponding author: Bartels, A. (andreas.bartels@tuebingen.mpg.de). 444 0166-2236/$ – see front matter ß 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tins.2008.06.004 Available online 3 August 2008