AVA Awards Recipients Report - Cam Smith - 2022

Bradshaw-Eagle Undergraduate Research Scholarship

Modulating visual attention with repetitive motor behaviours

Supervisor: Dr Daniel Baker

University of York

Introduction
Visual attention has been previously demonstrated to be impaired in ADHD and Autism. McAvinue et al. (2021) showed that individuals with ADHD have a specific deficit in sustained visual attention. Townsend et al. (2009) found that individuals with Autism have slower visual attention deployment. As such, we decided that visual attention was an appropriate measure to use for the study, as we aimed to investigate whether attention could be facilitated through stimming. Stimming actions are repetitive movements, also referred to as motor stereotypies or self-stimulatory behaviour. We predicted accordingly that when individuals with ADHD/Autism were allowed to stim, they would show improved attentional deployment relative to when they were not allowed to stim. We also predicted we might see an improvement in attentional deployment from the neurotypical participants when they were allowed to stim, as stimming appears in some neurotypical individuals (e.g., Chadehumbe, 2018).


Stimming behaviours are an important area for research in autism to focus on, as there is a lot of stigma surrounding them. Many studies aim to find ways to reduce these behaviours, such as wearing weighted vests (Fertel-Daly et al., 2001). However, when asked about the purpose of their stimming behaviours, individuals with autism typically report these behaviours relate to emotional regulation (Kapp et al., 2019). This is supported by research linking increases in stimming behaviour with emotional states such as stress, boredom, concentration, or excitement (Mackenzie, 2018). And while most forms of stimming are defined as purposeless, it has been shown many times that one form of vocal stimming known as echolalia facilitates the acquisition of object labels (originally Charlop, 1983; see Pruccoli et al., 2021, for a review). This evidence supports a need to investigate the function of other forms of stimming.


Methods
Participants were recruited through the researchers’ contacts, the University of York’s participant recruitment system (SONA), and ADHD/autism support groups based in York. There were a total of 25 participants recruited, however 3 of them did not complete the experiment. The diagnosis breakdown of those who completed the experiment was 17 with no diagnosis of a developmental disorder, 2 with autism, and 3 with ADHD. 15 reported that they usually engaged in stimming behaviours in their everyday life, while 7 reported that they did not usually stim. Participants first completed a questionnaire which asked them about their diagnosis and their stimming habits. They were fitted with a 64-channel EEG cap (ANT Neuro Waveguard). Then, they completed a Posner cueing task, set up in Matlab. They had to respond to whether the top or bottom half of a grating had a higher contrast, and the location of the grating was cued with 75% accuracy by an arrow in the centre. Participants were instructed to keep their eyes on a central fixation cross throughout the experiment, which was monitored using an eyetracker (Eyelink 1000). They completed four blocks of the experiment, in two of which they were asked to keep still, and in two of which they were asked to stim. The order of the blocks was randomly counterbalanced. Participants were paid a £10 Amazon voucher for their time. The EEG data was pre-processed using Brainstorm. Noisy channels were removed, selected based on a Welch frequency test, and electrodes with poor impedences were noted down when the caps were set up. The signal was band-pass filtered (0.5Hz to 48Hz), and then the data were re-imported using the triggers for left and right arrow cues to define epochs. We normalized each epoch to the average voltage in the 200ms before cue onset.


Results
Our primary measure of interest was the deployment of spatial attention in response to the cue. Previous work has shown that this involves a voltage lateralisation that builds up following cue onset. We used a pattern classification algorithm (a support vector machine) for all blocks, and also separately for condition (stimming/non-stimming). The algorithm was trained on the left vs. right cue data, and then was tested on its ability to differentiate attentional deployment to the left vs. to the right. This ability was compared across the stimming/non-stimming conditions. Greater attentional deployment should result in better ability to differentiate left from right, because the activity will be more polarised.

 

Figure 1: Pattern of scalp voltages 1000ms after cue onset, for left cues (a) and right cues (b). The strong lateralisation provides an index of the deployment of covert spatial attention following the cue.

 

As seen in Figure 1, attentional deployment to the left vs. to the right shows strong hemispheric lateralisation. The pattern classifier can use this to differentiate the direction of deployment of attention, and compare the timecourse between stimming conditions. Figure 2 shows this timecourse of classifier accuracy across all participants. Accuracy rises above chance levels from around 500ms post cue-onset, with significant clusters across all data, and when split by stimming condition. Although the accuracy scores are quite similar across conditions, it appears that the cluster for stimming condition begins substantially earlier (at around 500ms) than the non stimming condition (after 800ms).

 

Figure 2: Average SVM outputs across all participants, overall and divided into stimming/non-stimming conditions. Shaded regions indicate ±1SE across participants (N=22). Horizontal bars around y=75 indicate clusters where classifier performance was significantly above chance.

We also split participants according whether they reported that they usually stim. In Figure 3a, the results of the participants who do not usually stim are shown – there does not appear to be any particular effect of stimming on their attentional deployment. However, in Figure 3b, where the results of the participants who do usually stim are shown, there appears to be a benefit of stimming on attentional deployment – the classifier accuracy is consistently higher during stimming (green), suggesting greater attentional deployment, versus non-stimming (blue). This is consistent with our main hypothesis.

 

Figure 3: Timecourse of classifier accuracy for participants who do not typically stim (a) and who report regularly stimming (b). The format is the same as for Figure 2.

 

Discussion

The results suggest that, for those who typically stim, being able to stim provides a benefit in attentional deployment compared to not being able to stim. This is a very important finding, because, as discussed earlier, stimming behaviours are stigmatised and often interventions aim to eliminate them. Demonstrating that they have a valuable function (benefitting attention) can help to change opinions on these behaviours, promoting their acceptance. This could help individuals to learn better, for example in educational settings. This also opens up a new avenue of research to understand the way in which these behaviours are able to benefit attention.

The use of visual attention as a measure has been very valuable here, as it was possible to create a simple experiment with clearly defined results to demonstrate the value of stimming. In this case, we had to divide the groups based on whether they typically stim in their everyday life, rather than dividing them by their diagnosis as we originally intended. This was due to the very low sample of individuals with autism/ADHD. We were still able to acquire interesting results from this analysis, but future research should aim to replicate the experiment with larger groups of individuals with autism/ADHD, in order to specifically investigate the link between stimming and attention in them. This is especially important as they are a key group where stimming behaviours are seen, and they are the ones who are often targeted with interventions to reduce these behaviours.

Participants reported in the questionnaire what they believed their reasons were for stimming, and one of the main reasons that emerged was coping with stress or promoting relaxation. Accordingly, future research should investigate other potential emotional benefits of stimming.

 

Skills learned:

Running the experiment involved me learning new technical skills, such as setting up and recording from an EEG cap, and using an eyetracker- this included calibrating it, and troubleshooting the software. I was also responsible for recruiting the participants and running the experiment myself.

Analysing the data involved me learning to use Brainstorm (Matlab) to process the EEG data, and R for the eyetracking data.

I am also intending to write up the project in full with Daniel Baker, which will give me experience in writing an academic paper, and having it reviewed and published.