A neurocognitive investigation of the role of reinforcement learning in updating dysfunctional self-schema in depression: A putative mechanism for antidepressant action?
: (Alternative Format Thesis)

  • Catherine Hobbs

Student thesis: Doctoral ThesisPhD


Depression is one of the leading mental health problems experienced worldwide. Whilst treatments are available for depression, individual treatment response is varied. The intersection between self, emotion, and reward, referred to in combination as self-referential affective processing, may be a key cognitive vulnerability in depression and a sensitive target for intervention. I aimed to: (1) evaluate whether depression is associated with differences in processing of self-related information, and whether this varies depending on the emotional valence or rewarding nature of information; (2) understand whether antidepressants are associated with change in self-referential affective processing.

In chapter 1, I addressed my first aim by conducting a cross-sectional study where adults experiencing varying depression (n = 144) completed cognitive tasks measuring self, emotion, and reward processing, occurring independently and in combination. Depression was most reliably associated with self-referential processing occurring in interaction with reward and emotion. Participants with greater depression were worse at learning positive versus negative social evaluations about the self in a reinforcement learning task (β = 0.13, 95% CI: 0.06, 0.20, p < .001). In chapter 2, I validated these findings using an individual participant dataset collated from studies previously conducted within this research group (n = 552) and a second independent dataset of participants recruited online (n = 807). Replicating chapter 1, participants with greater depression were again worse at learning positive versus negative social evaluations about the self (dataset 1: β = 0.27, 95% CI: 0.20, 0.35, p < .001; dataset 2: β = 0.10, 95% CI: 0.02, 0.17, p = 0.009). Treatments that increase learning of positive social evaluations about the self may therefore be helpful in addressing depressive symptoms.

In the second half of my thesis, I subsequently investigated the effect of antidepressants on self-referential affective processing. In chapter 3 I conducted a systematic review, narrative synthesis (k = 82), and meta-analysis (k = 28) of behavioural evidence for change in emotional processing following antidepressant administration. I did not find reliable evidence that current behavioural measures of emotional processing were altered following antidepressant administration (positive: SMD = 0.08, 95% CI = -0.05, 0.22, p = 0.230; negative: SMD = -0.04, 95% CI = -0.12, 0.04, p = 0.341). In chapters 4 and 5, I therefore investigated whether change in self-referential affective processing may be a more sensitive measure of early antidepressant action. I did not find evidence that learning social evaluations about the self differed in healthy volunteers randomised to acute citalopram versus placebo (chapter 4: b = 1.95, -2.25, 6.16), or that this was associated with change in depression over the first eight weeks of antidepressant treatment in a prospective observational cohort of primary care patients (chapter 5: β = 0.15, 95% CI: -0.10, 0.39, p = 0.239). However, acute administration of an antidepressant increased positive social behaviours (chapter 4; b = 20%, 95% CI: 2%, 37%, p = 0.030). Additionally, exploratory analyses also indicated that antidepressants increased positive affective biases towards familiar others (chapter 4: b = 4.06, 95% CI: 0.88, 7.24), which weakly predicted a reduction in a secondary measure of depression, the BDI-II, in primary care patients (chapter 5: β = 0.24, 95% CI: -0.01, 0.49, p = 0.069). Exploratory analyses in chapter 5 also indicated stronger evidence that increased learning of positive versus negative evaluations about both the self (β = 0.34, 95% CI: 0.14, 0.55, p = 0.002) and a friend (β = 0.36, 95% CI: 0.16, 0.56, p = 0.001) were associated with a reduction in anxiety.

Due to weak evidence for the role of self-referential affective processing in antidepressant treatment found in chapters 4-5, as well as chapter 3 highlighting the need for novel psychological measures of antidepressant effects, in chapter 6 I evaluated a different area of cognition: optimistic belief updating. Whereas healthy participants updated their beliefs regarding negative life events more following good versus bad news, individuals experiencing depression lacked this bias (β = 0.71, β 95% CI: 0.24, 1.18, p = 0.004). In line with my previous findings of reduced learning of positive social evaluations, treatments attempting to increase the accommodation of positive information may be beneficial in remediating negative beliefs associated with depression.

Overall, whilst depression was reliably associated with reduced self-referential affective biases, this did not appear to be important in early antidepressant treatment. However, I identified potential areas of interest for future research on psychopharmacological antidepressant effects. Firstly, I found evidence that antidepressants may operate by increasing sensitivity to positive information about familiar others and increasing positive social behaviours. This may treat depression by increasing enjoyment and engagement in social interactions. Secondly, exploratory analyses suggested that change in affective biases may be important in remediating anxiety rather than depression. It is therefore possible that antidepressants operate in part by remediating threat-related biases that maintain anxiety symptoms.
Date of Award14 Sept 2022
Original languageEnglish
Awarding Institution
  • University of Bath
SupervisorKatherine Button (Supervisor) & Jie Sui (Supervisor)


  • depression
  • antidepressant
  • self
  • emotion
  • reward
  • cognitive biases
  • affective biases

Cite this