Fig. 4.

Memory reports from typical illiterate subjects show smaller LSC compared to literate subjects with same age, and psychotic subjects have a near-random structure. a Independent sample of illiterate adults (N = 14, grey bar), and pre-school children (N = 18, white bar) for comparison with a literate adult sample without school dropout (N = 41, blue bar). For a comparison between groups with no age difference (literate × illiterate adults, p = 0.2267) illiterates produced memory reports with smaller LSC (p = 0.0111), but non-significant difference for Nodes (p = 0.0927), RE (p = 0.0889) or ASP (p = 0.1583). For a comparison between groups with no education difference (pre-schooler children × illiterate adults) the pre-school sample produced memory reports with less lexical diversity (p = 0.0082), and graph size (p = 0.0011), but non-significant differences for RE (p = 0.1141), or LSC (p = 0.0402). Literate adults differ from pre-school children in all attributes (N (p < 0.0001), RE (p = 0.0016), LSC (p < 0.0001), and ASP (p < 0.0001)). Bar plots indicate median values and error bars indicate standard error. * For p < 0.05 corrected for multiple comparisons, (Wilcoxon rank sum test, two-tailed, Bonferroni correction for four comparisons, α = 0.0125). b Graph attributes were calculated for each random graph and averaged to compose the denominator of the ratio shown as normalized graph attribute in the next panel. c The graph attributes of each individual report were normalized by the corresponding mean random value, and the data were sorted according to more or less than 12 yE (years of Education). Typical subjects showed significant differences between subjects below (<) or above (>) 12yE (p for RE = 0.00004, LSC = 1.19e-10, ASP = 8.04e-8), but psychotic subjects did not (p for RE = 0.6128, LSC = 0.3712, ASP = 0.1398). *p < 0.05 corrected for multiple comparisons, n.s. for non-significant differences (Wilcoxon rank sum test, two-tailed, Bonferroni correction for six comparisons, α = 0.0083)