Abstract
The objective of this thesis is to examine neural activity in the pigeon brain during a wide range of categorisation tasks. Single-unit electrophysiology will be used to record from neurons in a prefrontal area (NCL) and an intermediary visual area (ENTO), in an attempt to understand how neurons in both of these areas are involved in the categorisation process, and how neurons from these areas use reward/behavioural information and visual information about categories during discrimination. We then present an integrated model of the neural correlates of categorisation in the pigeon brain based on our findings.
This thesis starts with an exploration into the second intermediary visual area and the termination point of the thalamofugal pathway, the visual Wulst (Chapter 3). While we found some interesting evidence of the Wulst being a highly reward-driven visual area, we decided not to continue recording from this area for the remainder of the studies as further research is needed to understand the Wulst’s primary mode of processing.
The first of the categorisation tasks was a semi-symbolic categorisation task, in which birds learned to distinguish several English four-letter words from four-letter nonwords (Chapter 4). Neural activity was recorded while the pigeons discriminated their learned “vocabularies” from an extensive pool of nonwords. We found that there were no differences between activity to words and non-words, except in birds who were potentially still learning the task. We argue that behavioural information such as reward helps guide categorisation in NCL and ENTO before learning but is likely no longer used after learning.
In the second categorisation task, we trained birds to respond to either Monet or Picasso paintings using an S+/S- paradigm (Chapter 5). We recorded neural activity from an additional area of interest known as the mesopallium ventrolaterale (MVL), a higher order visual area. In both NCL and ENTO, we found a significant left-hemisphere dominance for both reward and visual-based categorisation, while no such asymmetry was found in MVL. Furthermore, both NCL and ENTO appear to use reward information during categorisation when there is the opportunity to (i.e. differential rewards), but MVL does not.
The third categorisation task involved more complex stimuli, in which category exemplars are comprised of four smaller stimuli (Chapter 6). Exemplars from both categories share some features (irrelevant features), while other features are unique to each category (relevant features). We found that both excitatory and inhibitory ENTO neurons use visual (colour) and behavioural (relevancy) information to aid categorisation, and we posit that these neurons relay this information onto excitatory NCL neurons.
Inhibitory NCL neurons then use the information encoded and received by ENTO and excitatory NCL neurons to execute a behavioural decision.
The final categorisation task involved the same stimuli as the Monet/Picasso study in Chapter 5, but in a delayed matching-to-category (DMC) paradigm instead of the original S+/S- paradigm (Chapter 7). We present preliminary results indicating that NCL, ENTO, and MVL share information about categories with each other via reciprocal projections.