Meta-analysis of the functional neuroanatomy of single-word reading: method and validation

Intersubject variability and subtle differences in experimental design can lead to variable results in studies of cognitive processes such as reading. To accurately identify the neural processes associated with cognition and sensorimotor processing, meta-analytic methods capable of identifying areas of consistent activation among studies are useful. This paper describes a novel approach for combining published neuroimaging results from multiple studies, designed to maximize the quantification of interstudy concordance while minimizing the subjective aspects of meta-analysis. In this method, a localization probability distribution was modeled for each activation focus obtained from 11 PET studies of reading single words aloud, and the union of these distributions was taken to yield an activation likelihood estimate map for the brain. Significance was assessed via permutation analysis of randomly generated sets of foci. Regions of significant concordance were identified in bilateral motor and superior temporal cortices, pre-SMA, left fusiform gyrus, and the cerebellum. These meta-analytic results were validated by comparison with new fMRI data on aloud word reading in normal adult subjects. Excellent correspondence between the two statistical maps was observed, with fMRI maxima lying close to all meta-analysis peaks and statistical values at the peaks identified by the two techniques correlating strongly. This close correspondence between PET meta-analysis and fMRI results also demonstrates the validity of using fMRI for the study of language tasks involving overt speech responses. Advantages of this automated meta-analysis technique include quantification of the level of concordance at all brain locations and the provision for use of a threshold for statistical significance of concordance.