Skip to main content

Perceptual Learning

Periodic Reporting for period 1 - Perceptual Learning (Perceptual Learning)

Reporting period: 2017-09-01 to 2019-08-31

Even though perceptual learning is well established as a phenomenon, and we have some idea how it works, we do not know a great deal about the mechanisms involved or how to modulate them. The results from this project help us to understand the mechanisms controlling perceptual learning and give us some insights into:
1) The neural structures and processes that allow us to benefit from experience and improve discrimination.
2) How to control and modulate these processes using brain stimulation techniques.
3) A better understanding of perceptual learning’s role in face recognition skills.
The main findings from the project provide evidence in support of a specific experimental brain stimulation method, that allows us to selectively modulate perceptual learning indexed by a robust cognitive phenomenon known as the inversion effect (better recognition performance for upright vs inverted stimuli). We have shown that a brief brain stimulation manipulation can either decrease or increase the inversion effect by reducing or enhancing recognition performance for upright stimuli. Furthermore, these brain stimulation-induced effects have their electrophysiological correlates on a specific event-related potential (ERP) component. Importantly, we also found how certain brain structures are more activated in response to brain stimulation while participants perform a perceptual learning task.
Perceptual learning in this project was indexed by the face inversion effect. This refers to reduced performance when we try to recognise a familiar face turned upside down compared to its usual upright orientation. The main results from the project can be grouped by the key findings that emerged from the series of studies we conducted.

a) Transcranial Direct Current Stimulation modulates perceptual learning indexed by the inversion effect

TDCS uses a constant low current delivered to the brain area of interest via electrodes on the scalp. When anodal stimulation is delivered, the current causes a depolarisation of the resting membrane potential which modulates neural excitability. Sham (control) stimulation lasts for a brief time (30sec) causing no change to neural excitability, but leaving participants believing they were stimulated.
We are now able to show that a particular tDCS paradigm (anodal stimulation delivered for 10 mins at 1.5mA over the left DLPFC at Fp3) can affect perceptual learning by reducing the inversion effect for chequerboards and faces by impairing recognition for upright stimuli. Our findings suggest that our tDCS procedure can affect the perceptual learning acquired for upright stimuli drawn from a familiar category that gives us a benefit when asked to recognise and discriminate between new upright stimuli taken from that category. Critically, no effect of tDCS is found for inverted stimuli, suggesting that after inversion there is no perceptual learning to be lost (we do not usually experience faces turned upside down) and so the tDCS is ineffective. Importantly, the tDCS paradigm is able to entirely eliminate the inversion effect for chequerboards, but that’s not the case for faces. Despite the face inversion effect being significantly reduced in the anodal condition compared to sham, the inversion effect is still significantly present. This is a very important result because it suggests that the face inversion effect may be due to two main components. One component is based on perceptual learning and we can modulate and disrupt it by applying our tDCS paradigm. The other component may have to do with face “specificity” and is unaffected by tDCS.

b) TDCS combined with EEG and fMRI can help us to control and influence perceptual learning and face recognition skills

In order to investigate further the neurocognitive bases of perceptual learning and the mechanisms that control this phenomenon, we then investigated the inversion effect using the Electroencephalography (EEG), and then combined this with our tDCS procedure.

An ERP is a measured brain response to a specific event/stimulus. ERPs are calculated by averaging EEG activity time-locked to the presentation of a stimulus. The N170 ERP peak component is a negative deflection occurring at 170ms after a stimulus is presented and it is usually found to be largest at parietal-occipital regions. It is reliably found with faces and is significantly affected by inversion which results in a larger peak amplitude and a delayed latency.

The most striking finding from our work on the N170 is that obtained by using the combination of tDCS and EEG to investigate the inversion effect on the N170. Importantly, we provided the first evidence in the literature for tDCS being able to modulate the electrophysiological correlates of perceptual learning (indexed by the inversion effect on the N170).
Furthermore, we combined tDCS and fMRI to explore the brain areas activated by the specific tDCS stimulation. For this specific study we used chequerboard stimuli so as to fully control the development of expertise during the pre-exposure phase (a categorization task). Through a region of interest analysis, we found that brain areas in the Precentral Gyrus, Brodmann Area 6 (BA6), showed significantly higher activation in the anodal condition vs sham.

c) TDCS can improve face recognition skills by eliminating the negative effects of generalisation

We demo
The key findings from the project show us how we are now able to modulate perceptual learning to systematically reduce or increase face recognition skills. Furthermore, we now have some evidence indicating the electrophysiological correlates of perceptual learning as well as the brain structures involved in the mechanisms at the basis of this phenomenon. Moreover, we have provided more support to the idea that perceptual learning is a key factor involved, at least in part, in the face inversion effect.

In the future, the results of this project may be extended to real-life situations aiming to improve performance in individuals who make difficult discriminations and for whom we rely on for our safety (e.g. airport personnel doing x-ray screening of baggage). Also, being able to control and influence perceptual learning could be used to help the construction of agents or applications to perform several kinds of tasks for us (e.g. machine learning for recognition of faces, fingerprints, etc.).
EPS Talk (London, 2018)
Psychonomic Talk (Amsterdam, 2018)
Cognitive Science Poster (Montreal, 2019)
Perceptual Learning Workshop (Exeter, 2019)