A decade-long series of online studies of intellectual ability has culminated in insights into the positive and negative impacts of COVID-19, after the latest study of 390,000 people coincided with the first wave of the pandemic.
The analysis – which is undergoing peer review – was carried out by Adam Hampshire at Imperial College London, along with Peter Hellyer and colleagues at King’s College, and Sam Chamberlain at Southampton (you are welcome to take part in the online intelligence tests).
People who had recovered from COVID-19, including those free of symptoms, exhibited ‘significant cognitive deficits’ when controlling for other factors, said Hampshire. ‘When we first saw the results, we were quite alarmed, the average change was equivalent to – roughly speaking – around a seven IQ point underperformance.’
‘Some aspects of cognition are associated, not others,’ he says. ‘For example, people show poor performance of tasks measuring attention and higher language faculties but perform normally on other aspects such as word memory.’ Surprisingly, they could also see a deficit in cognitive skills among those who were infected but not hospitalised.
‘We also examined how mental health was affected. People who had been ill showed elevated symptoms of depression, anxiety and post-traumatic stress, but the cognitive deficits seemed to be distinct.’
These insights arise from a series of studies that began with one on 44,600 people conducted a decade ago, when Hampshire worked with me to reach a big audience while I was the editor of the magazine New Scientist.
The findings of our online mass cognitive test, published in 2012 in the journal Neuron, gave a more nuanced view of intelligence that challenged the traditional idea that it should be reduced to a single number, IQ.
The study instead suggested that there were at least three distinct components to human cognitive skills that drew on different brain networks: short-term memory; reasoning; and finally, a verbal component.
This work was controversial. Despite more than a century of study, the subject of intelligence is a battleground, not least because ‘intelligence’ can be defined many ways.
The origins of the most common measure of intelligence, IQ, date back to 1904, when the English psychologist Charles Spearman (1863-1945) observed that cognitive performance was linked across different tasks. In other words, if you have a good memory, you tend to be good at reasoning too, and it is possible to come up with an intelligence quotient (IQ).
The utility of IQ was challenged as too broad brush by our 2012 Neuron paper. Adam Hampshire, who is now based at Imperial College London, has followed this work up by using machine learning methods to analyse both brain and behaviour during the performance of cognitive tasks.
The new insights are even more radical, fitting a model of cognition put forward by Godfrey Thomson (1881 –1955), an English educational psychologist: ‘He had this theory, sampling theory, that proposed that every mental test taps a different combination of “bonds” or “neural arcs” from a large available pool of resources,’ says Hampshire.
In one sense, Thompson’s idea complements Spearman’s view, in that mental tasks that recruit similar combinations of ‘bonds’ will be linked in terms of overall ability. ‘From this perspective, IQ is a summary measure of the capacity of a highly complex system that has many distinct parts,’ says Hampshire.
Thompson’s idea is also consistent with modern ‘network science’ perspectives on how the brain functions, and Hampshire and colleagues set out to test this more formally by applying machine learning methods – a form of artificial intelligence – to analyse brain scans and behavioural data whilst people undertook different cognitive tasks.
This work provided support for the idea of the brain drawing on different networks, depending on the task – and the results reported a few days ago in the journal Nature Communications by Hampshire, with Eyal Soreq, Ines Violante and Richard Daws.
The results back the view that our cognitive skills depend on shifting complex coalitions of brain networks, which change in response to what kind of task a person is doing, rather than being dependent on brain regions that are focused on doing one job. ‘Researchers spent many years trying to map individual functions to individual brain regions but, in reality, the brain is much more complex than this,’ he says.
The brain adopts different states that are optimal for tackling different tasks, and, in each state, it forms transient networks comprising many different brain areas. A century later, armed with modern machine learning and brain imaging technology, ‘we are able to demonstrate that the network mechanisms that underlie intelligence conform remarkably well with Thomson’s sampling theory,’ says Hampshire.
In parallel, Hampshire, has followed up this study with Peter Hellyer and William Trender by using an artificial intelligence website – called Cognitron – to analyse online cognitive tests in work with Wired, the Science Museum and now the BBC to identify a brief set of online tasks that could be used to fractionate different aspects of intelligence, providing detailed profiles of peoples’ cognitive abilities.
These more detailed cognitive profiles are now proving useful for research in a variety of clinical populations, where patients can suffer from different types of cognitive problem. For example, Hampshire put the technology to the test in collaboration with Amy Jolly and David Sharp, fellow Imperial researchers who are studying brain injury.
By applying machine learning to brain imaging data, they demonstrated that the cognitive tasks that patients struggle to perform after a brain injury can be predicted from the combination of network connections that were disrupted.
The impact of the pandemic on cognitive health emerged when this multi-dimensional assay of intelligence was used to gather data from more than 390,000 people who took part last year in a BBC experiment, the Great British Intelligence Test.
Further studies are now underway to try and understand whether cognitive problems relate to changes in the brain after a COVID-19 infection.
The research rests on a wide range of earlier innovations that can be seen in the Science Museum Group’s collection and museums, from the development of computers and body scanners to the server that paved the way for the World Wide Web.