| Popis: |
Background: Bibliotherapy is under-theorized and under-tested: its purposes and implementations vary widely, and the idea that ‘reading is good for you’ is often more assumed than demonstrated. One obstacle to developing robust empirical and theoretical foundations for bibliotherapy is the continued absence of analytical methods capable of providing sensitive yet replicable insights into complex textual material. This pilot study offers a proof-of-concept for new quantitative methods including VAD (valence–arousal–dominance) modelling of emotional variance and doc2vec modelling of linguistic similarity. Methods: VAD and doc2vec modelling were used to analyse transcripts of reading-group discussions plus the literary texts being discussed, from two reading groups each meeting weekly for six weeks (including 9 participants [5 researchers (3 authors, 2 collaborators), 4 others] in Group 1, and 8 participants [2 authors, 6 others] in Group 2). Results: We found that text–discussion similarity was inversely correlated with emotional volatility in the group discussions (arousal: r = -0.25; p = ns; dominance: r = 0.21; p = ns; valence: r = -0.28; p = ns), and that enjoyment or otherwise of the texts and the discussion was less significant than other factors in shaping the perceived significance and potential benefits of participation. That is, texts with unpleasant or disturbing content that strongly shaped subsequent discussions of these texts were still able to sponsor ‘healthy’ discussions of this content, as evidenced by the combination of low arousal plus high dominance despite low valence in the emotional qualities of the discussion. Conclusions: Our methods and findings offer for the field of bibliotherapy research both new possibilities for hypotheses to test, and viable ways of testing them. In particular, the use of natural language processing methods and word norm data offer valuable complements to intuitive human judgement and self-report when assessing the impact of literary materials. |
