Abstract
Functional and structural brain abnormalities have been reported in many imaging
studies of depressive illness. However, the mechanisms by which these
abnormalities give rise to symptoms remain unknown. The work described in this
thesis focuses on such mechanisms, particularly with regard to neural predictive error
signals. Recently, these signals have been reported to be present in many studies on
animals and healthy humans. The central hypothesis explored in this thesis is that
depressive illness comprises a disorder of associative learning. Chapter 2 reviews
the brain regions frequently reported as abnormal in imaging studies of depressive
illness, and the normal function of these particular brain regions. It is concluded that
such regions comprise the neural substrate for associative learning and emotion.
However, confidence in this conclusion is limited by considerable variability in the
human imaging literature. Therefore, chapter 3 describes a meta-analysis, which
tests the hypothesis that, consistent with the non-imaging literature, the ventromedial
prefrontal cortex is most active during emotional experience. The results of the
meta-analysis were clearly consistent with this hypothesis. Chapter 4 provides an
introduction to neural predictive error signals from the general perspective of
homeostatic physiological regulation. Both experimental evidence supporting the
error signals, and various formal mathematical theories describing the error signals,
are summarised. This provides the background to chapter 5, which describes an
original fMRI study which tested the hypothesis that patients with depressive illness
would exhibit abnormal predictive error signals in response to unexpected
motivationally significant stimuli. Evidence of such abnormality was found.
Chapter 6 describes a further original study using transcranial ultrasound and
diffusion tensor imaging of the brainstem, which investigated reports of a subtle
structural abnormality in depressed patients. If present, it might give rise to
abnormal error signals. However, no structural abnormality was found. Finally,
chapter 7 discusses the significance of these findings in the context of clinical
features of depressive illness and a wide range of treatments, ranging from
psychotherapy through antidepressants to physical treatments. A number of potential
future studies are identified, which could clarify understanding of depressive illness.
