Fig. 2: The connectivity of personalised networks of depression-relevant language is associated with individual differences in self-reported depression severity.

a There was a significant positive association between global network strength and depression severity (β = 0.008, SE = 0.003, p = 0.002) (N = 946). b Mean regression coefficients with 95% CIs for individual network node strengths, positive coefficients indicate increased node strength with greater depressive severity. There was a significant association between depression severity and the node strength of Neg. Emo (β = 0.02, SE = 0.007, p = 0.007), Swear (β = 0.02, SE = 0.007, p = 0.01), and Articles (β = 0.01, SE = 0.003, p < 0.001). c Mean personalised network of all participants. Green and red lines indicate positive and negative associations respectively. Line widths represent edge strength between nodes, with larger widths corresponding to greater edge strength. Associations were not affected after adjusting for the number of days, a proxy for network stability, (Table S3) nor were they altered by omitting 3^rd person pronouns from the network (Figure S4). ^aUnadjusted two-sided Pearson correlation, ^bTwo-sided general linear regression unadjusted for multiple comparisons. Source data are provided as a Source Data file. *p < 0.05, **p < 0.01, ***p < 0.001.