LanA (Language Atlas): A probabilistic atlas for the language network based on fMRI data from >800 individuals

Abstract

Two analytic traditions characterize fMRI language research. One relies on averaging activations voxel-wise across individuals. This approach has limitations: because of inter-individual variability in the locations of language areas, a location in a common brain space cannot be meaningfully linked to function. An alternative approach relies on identifying language areas in each individual using a functional ‘localizer’. Because of its greater sensitivity, functional resolution, and interpretability, functional localization is gaining popularity, but it is not always feasible, and cannot be applied retroactively to past studies. We provide a solution for bridging these currently disjoint approaches in the form of a probabilistic functional atlas created from fMRI data for an extensively validated language localizer in 806 individuals. This atlas enables estimating the probability that any given location in a common brain space belongs to the language network, and thus can help interpret group-level peaks and meta-analyses of such peaks, and lesion locations in patient investigations. More meaningful comparisons of findings across studies should increase robustness and replicability in language research.

The language network reliably ‘tracks’ naturalistic meaningful non-verbal stimuli

“…First, to account for inter-individual variability in the location of the language network (e.g., Amunts et al, 1999; Tomaiuolo et al, 1999; Fedorenko et al, 2010; Fedorenko & Blank, 2020; Lipkin et al, 2022), we defined language regions in each participant individually using a well-established functional localizer task (Fedorenko et al, 2010). This approach yields greater sensitivity, functional resolution, and interpretability compared to the traditional group-averaging approach, where functional correspondence across participants is assumed to hold voxel-wise (e.g., Nieto-Castañón & Fedorenko, 2012; Fedorenko, 2021).…”

“…In brain imaging studies, these regions exhibit highly selective responses for language processing over diverse cognitive tasks, including those that bear parallels to language and have long been argued to share resources with language. For example, language-responsive regions show little or no response to arithmetic (e.g., Fedorenko et al, 2011; Monti et al, 2012; Amalric & Dehaene, 2019), music (e.g., Fedorenko et al, 2011; Rogalsky et al, 2011; Chen et al, 2021), logic (e.g., Monti et al, 2009), executive control (e.g., Fedorenko et al, 2011; Mineroff, Blank et al, 2018), action/gesture observation (e.g., Pritchett et al, 2018; Jouravlev et al, 2019), Theory of Mind (e.g., Deen et al, 2015; Paunov, 2018; Paunov et al, 2022; Shain, Paunov, Chen et al, in prep. ), and even the processing of computer languages (e.g., Ivanova et al, 2020; Liu et al, 2020).…”

Non-literal language processing is jointly supported by the language and Theory of Mind networks: Evidence from a novel meta-analytic fMRI approach

“…For each network, an activation overlap map was created by overlaying a large (n>100) number of individual binarized activation maps for the ‘localizer’ task targeting each network, as described below (this is the first step in the Group-Constrained Subject-Specific analytic approach, as described in Fedorenko et al, 2010 and Julian et al, 2012). To account for inter-individual variability in the overall level of activation, we selected in each individual the top 10% most localizer-responsive voxels across the brain (fixed-statistical-threshold approaches yield near-identical results; Lipkin et al, 2022). Specifically, we sorted the t -values for the relevant contrast and took 10% of voxels per participant with the highest values.…”

“…A version of this localizer is available from https://evlab.mit.edu/funcloc/download-paradigm, and the details of the procedure and timing are described in Figure 1 and Table 2 . The probabilistic functional atlas used in the current study (Language Atlas (LanA); Lipkin et al, 2022) was constructed using data from 806 participants, and the voxel with the highest network probability had a value of 0.82 (i.e., belonged to the top 10% of most language-responsive voxels in 82% of participants).…”

“…Aggregating group-level peaks in meta-analyses may improve estimates of location, but these estimates are still interpreted via reverse inference. Here, we leverage probabilistic functional atlases (e.g., Dworetsky et al, 2021; Thirion et al, 2021; Lipkin et al, 2022) created from large numbers of individuals who have performed extensively validated functional ‘localizer’ tasks that target particular functional networks. Given that these atlases capture inter-individual variability in the locations of the relevant networks, we can estimate for any given location in the common space the probability that it belongs to each candidate network.…”