Abstract
36 Subjects with a diagnosis of schizophrenia (Scz) overweight unexpected
37 evidence in probabilistic inference: such evidence becomes ‘aberrantly salient’. A
38 neurobiological explanation for this effect is that diminished synaptic gain (e.g.
39 hypofunction of cortical N-methyl-D-aspartate receptors) in Scz destabilizes
40 quasi-stable neuronal network states (or ‘attractors’). This attractor instability
41 account predicts that i) Scz would overweight unexpected evidence but
42 underweight consistent evidence, ii) belief updating would be more vulnerable
43 to stochastic fluctuations in neural activity, and iii) these effects would correlate.
44
45 Hierarchical Bayesian belief updating models were tested in two independent
46 datasets (n=80 and n=167, male and female) comprising human subjects with
47 schizophrenia, and both clinical and non-clinical controls (some tested when
48 unwell and on recovery) performing the ‘probability estimates’ version of the
49 beads task (a probabilistic inference task). Models with a standard learning rate,
50 or including a parameter increasing updating to ‘disconfirmatory evidence’, or a
51 parameter encoding belief instability were formally compared.
52
53 The ‘belief instability’ model (based on the principles of attractor dynamics) had
54 most evidence in all groups in both datasets. Two of four parameters differed
55 between Scz and non-clinical controls in each dataset: belief instability and
56 response stochasticity. These parameters correlated in both datasets.
57 Furthermore, the clinical controls showed similar parameter distributions to Scz
58 when unwell, but were no different to controls once recovered.
59
60 These findings are consistent with the hypothesis that attractor network
61 instability contributes to belief updating abnormalities in Scz, and suggest that
62 similar changes may exist during acute illness in other psychiatric conditions.
37 evidence in probabilistic inference: such evidence becomes ‘aberrantly salient’. A
38 neurobiological explanation for this effect is that diminished synaptic gain (e.g.
39 hypofunction of cortical N-methyl-D-aspartate receptors) in Scz destabilizes
40 quasi-stable neuronal network states (or ‘attractors’). This attractor instability
41 account predicts that i) Scz would overweight unexpected evidence but
42 underweight consistent evidence, ii) belief updating would be more vulnerable
43 to stochastic fluctuations in neural activity, and iii) these effects would correlate.
44
45 Hierarchical Bayesian belief updating models were tested in two independent
46 datasets (n=80 and n=167, male and female) comprising human subjects with
47 schizophrenia, and both clinical and non-clinical controls (some tested when
48 unwell and on recovery) performing the ‘probability estimates’ version of the
49 beads task (a probabilistic inference task). Models with a standard learning rate,
50 or including a parameter increasing updating to ‘disconfirmatory evidence’, or a
51 parameter encoding belief instability were formally compared.
52
53 The ‘belief instability’ model (based on the principles of attractor dynamics) had
54 most evidence in all groups in both datasets. Two of four parameters differed
55 between Scz and non-clinical controls in each dataset: belief instability and
56 response stochasticity. These parameters correlated in both datasets.
57 Furthermore, the clinical controls showed similar parameter distributions to Scz
58 when unwell, but were no different to controls once recovered.
59
60 These findings are consistent with the hypothesis that attractor network
61 instability contributes to belief updating abnormalities in Scz, and suggest that
62 similar changes may exist during acute illness in other psychiatric conditions.
| Original language | English |
|---|---|
| Pages (from-to) | 9471-9485 |
| Journal | JOURNAL OF NEUROSCIENCE |
| Volume | 38 |
| Issue number | 44 |
| DOIs | |
| Publication status | Published - 31 Oct 2018 |
Keywords
- Schizophrenia
- data-gathering
- computational modelling