The neural networks that support normal single word reading have been studied extensively with functional magnetic resonance imaging (fMRI) and been heavily meta analysed; yet two caveats persist. Firstly, inclusion criteria are generally broad, allowing for a range of studies with inconsistent methodological practices to become a major source of variance in the data. Secondly, the sophistication of meta analytic techniques are ever improving, calling for a continuation of quantitative summaries for more informed interpretations of the literature. We conducted a novel meta-analysis using the updated activation likelihood estimation (ALE) approach on fMRI data from 345 native English speakers published from 1992 to 2014, comparing neural correlates between single word reading (SWR); and lexical decision tasks (LDT). SWR paradigms activated a highly predictable, left lateralised reading network, whereas LDT, did not; and instead, fewer clusters overall were found in bilateral regions associated with orthographic, semantic and decision making processes, as well as traces of eye and hand movements. A conjunction analysis conducted on shared LDT/SWR activations found left fusiform and right inferior frontal gyral (IFG) regions to be common to both tasks; but an overlapping right inferior, precentral and middle frontal cluster to be specific to LDT. The findings from this meta analysis warn of the implications of having broad inclusion criteria when meta analysing heavily researched phenomena. We demonstrate the importance of the considerations researchers need to make when designing their paradigms as task related activations in single word reading can heavily influence our interpretation of the nature of the reading network.