Our experiment, therefore, was specifically set up to

define the neural mechanisms operating during the emergence of knowledge about social hierarchies, examine how rank information is coded in the brain, and dissociate the operation of social-specific from domain-general processes. Participants completed training trials, where a pair of adjacent items in the hierarchy was presented (e.g., P1 versus P2, G1 versus G2, where P = person and G = galaxy; Figure 1A): they were required to learn through trial and error, which person had more power (social condition) or which galaxy had more mineral (nonsocial condition). Following each block of training trials, participants completed test trials where they were required to select the higher ranking of the two items presented (e.g., P3 versus P6, G3 versus selleck compound G6; Figure 1B) and rate their confidence in their decision on a scale of 1 (guess) to 3 (very sure). Test trials differed from training trials in two critical ways: nonadjacent items in the hierarchy were presented during test trials (e.g., P3 versus P6), and no corrective feedback was issued. As such, participants were required to use transitive inference to deduce the correct item during test trials (e.g., P3, in a P3 versus P6 trial), by using knowledge of the underlying hierarchy (e.g., P1 > P2 > P3 > P4 > P5 > P6 > P7: see below). In contrast, participants could achieve proficient performance on

training trials by simply memorizing the correct item in each pair (e.g., P1, in a P1 versus this website P2 trial). While the Learn phase paradigm builds on a rich vein of research that has used the transitive inference task across species (Bryant and Trabasso, 1971; Dusek and Eichenbaum, 1997; Greene et al., 2006; Grosenick et al., 2007; Heckers et al., 2004; Hurliman

et al., 2005; Moses et al., 2010; Paz-Y-Miño et al., 2004; Zeithamova et al., 2012), we incorporated several features designed to achieve the specific goals of our experiment: first, we interleaved blocks of training and test trials throughout the time course of the Learn phase in order to chart the development of successful transitive behavior. In contrast, previous fMRI studies have typically included test trials only at the very end of training (Greene et al., 2006; Heckers et al., 2004; Moses et al., 2010). Second, we incorporated a novel measure of test trial performance (i.e., “inference score”), which Resminostat was validated in a separate behavioral experiment (see below and Supplemental Results). The inference score index – which incorporated participants’ assessment of their confidence in their choices, a metacognitive measure typically used to characterize medial temporal lobe dependent memory processes (e.g., (Eichenbaum et al., 2007)—allowed us to track the emergence of knowledge of the linear structure of the hierarchy, and thereby reveal the underlying neural mechanisms. Lastly, our paradigm was unique in affording a direct comparison of social (i.e.