My main interest is in associative learning, a process that seems to be designed to let us learn about the causal structure of the world around us, allowing us to predict and control our environment. It is clear that this type of learning is important for the most basic aspects of our daily life, and it is found throughout the animal kingdom.
show/hide full profile...
Some psychologists also argue that networks of associations might underlie more complex types of learning in human subjects, such as concept formation and language. If this view is correct, then it is important for us to have a thorough understanding of associative processes and the limits of what they can explain. My research is (paradoxically) mainly concerned with types of learning that are assumed not to be easily explained in terms of association formation: the aim is to examine whether modifications of associative theory could allow it to accommodate these apparently nonassociative phenomena.
My previous and current work addresses the following topics:
In some circumstances a particular item can have different associates depending on the conditions in which it is presented: for example, for a bilingual person the sight of a sheep (say) is going to be associated with different verbal labels depending on the language that they are speaking. It appears as though the context is switching between the different associations that have been formed. This process, known as conditional learning (or occasion setting) cannot be explained in terms of simple associations. I have developed an associative-type account of conditional learning, that makes specific predictions about how conditional cues function. Some of my current research is aimed at testing this hierarchical account of conditional learning.
When a stimulus of a particular duration (e.g. 20 seconds) predicts a particular outcome, association formation allows learning that the stimulus predicts the outcome - but not learning when (i.e. outcome will occur 20 seconds after stimulus onset). Learning about time in this way does occur, but most associative theories cannot easily explain it, and make relatively few predictions about the effects of the temporal characteristics of cues on associative learning phenomena. In a collaboration with Dr Dómhnall Jennings (University of Newcastle), we are investigating the way in which associative theories can represent temporal cues, with a view to developing an associative account of timing effects. We are also, in collaboration with Dr Eduardo Alonso (City University) and Dr Esther Mondragón (Centre for Computational and Animal Learning Research), using simulations to evaluate whether our findings can be accommodated by associative theories.
Hippocampus and learning
Damage to the hippocampus produces a complex range of cognitive deficits, for example in spatial learning; however, association formation is usually said to be intact, implying that associative learning does not underly the skills that are affected by hippocampal damage. However, our recent work suggests that certain sorts of associative learning can be impaired after hippocampal damage, implying that an associative learning impairment could mediate the effects if hippocampal damage on, for example, spatial learning. We aim to investigate those aspects of associative learning that are impaired by hippocampal damage, and the extent to which this can explain the other, more complex effects of such damage (p conducted with Dr Eric Tam, now at the University of Oxford).
When learning goes wrong Alzheimer's disease and schizophrenia
In some conditions learning is impaired, and we can use what we know about associative theory to analyse these deficits. For example, schizophrenia is often characterised by high impulsivity, which might be evident as a deficit in inhibitory learning; in a collaboration with Dr Helen Cassaday we confirmed this possibility (work conducted with Dr Zhimin He); related effects were observed in participants with certain types of personality disorder associated with impulsivity deficits. Schizophrenia is also said to be characterised by a deficit in performance on tasks employing task-setting cues - which are formally equivalent to conditional cues (see above); this leads to the question whether such deficits are also evident in subjects with high schizotypy. Finally, Alzheimer's disease is characterised by a wide range of cognitive impairments, and yet many of these have not been precisely characterised in associative terms. In collaboration with Dr Marie-Christine Pardon (School of Biomedical Sciences) and Mr Paul Armstrong, I am currently conducting a series of studies whose long term aim is to analyse the cognitive deficits in a genetically modified strain of mouse that is regarded as a translational model of Alzheimer's disease, and examine their underlying neurobiological correlates.
Learning and Addiction
Human drug seeking has been analysed in terms of classical conditioning: the ability of environmental cues to become associated with the effects of the drug can make them provoke drug-seeking behaviour. The mechanism underlying this process has been modelled by an effect called Pavlovian-instrumental transfer (PIT): if you have two outcomes, chocolate and tobacco, each produced by a different (drug-seeking response), then a conditioned stimulus that signals e.g. chocolate, will increase the level of the chocolate-seeking response more than the tobacco-seeking response (and vv). However, there is still relatively little understanding of how this effect is mediated, and this project would address this.