Integrative studies bridging levels and domains of cognition can elucidate both the normal developmental emergence of higher-order, domain-specific cognitive skills from simpler, more domain-general antecedents, and abnormal cognitive variations in developmental disorders or other clinical neuropsychiatric conditions. Such studies demand suites of behavioural experiments that measure many domains of perceptual and cognitive function, requiring subjects to remain on task throughout large numbers of experimental trials and paradigms. This is especially true in cases of physiological recording, where many repeated trials may be required in order to achieve sufficient signal-to-noise. These lengthy, taxing and tedious experiments can be difficult to complete in contexts of cognitively impaired or paediatric populations, or in stressful environments such as MRI scanners, and can place experimental control at odds with ecological validity. A way between the horns of this dilemma is offered by embedding experiments within the motivating, engaging, yet strongly regular and systematic environment of a computer game. Our group has developed a suite of games that measure motion coherence threshold, go/no-go inhibition, focused and distributed visual and multimodal attention, perceptual disembedding, and first- and second-order “theory of mind.” Game events are transparently logged for offline analysis, and can be synchronised with physiological recordings. The game format is person-centred and event-driven rather than computer-centred and timed, encouraging players to apply skills at their own pace and on their own terms. Predictable and anxiety-minimising perceptual and social environments afford opportunities not only to demonstrate skills but also to develop them; the game can be extended from a vehicle for cognitive testing into a vehicle for cognitive training, with the potential to titrate difficulty and to measure cross-domain consequences of training in one cognitive domain on expression of skills in another, e.g., measuring the effect of attention training on social cognition. We have applied the game in behavioural and physiological studies of autistic cognitive traits both in people with autism spectrum conditions and in the normal population, and will present behavioural and physiological data demonstrating covariation of non-social and social cognitive skills in both these populations.

Discoveries in biomedical science are determined at least as much by the questions that scientists ask as by the data that scientists observe. Complex neuropsychiatric conditions in general, and autism in particular, have been framed as binary distinctions: one either has or does not have the diagnosis. This categorical division neatly separates the population into “patients” and “healthy controls” convenient for experimental constrasts — but is it actually an accurate model? An accumulation of evidence - evidence that began to pile up as early as autism's first published description - indicates that many aspects of autism are most accurately described not by contrasts between levels of a categorical variable, but rather by regression or correlation with dimensional variables whose ranges are continuous throughout the population. Both within and beyond the diagnostic boundaries of the autism spectrum, individual variations in autistic cognitive traits are influenced by individual variations in neurophysiology and genetics. Relations within and between these genetic, physiological and cognitive behavioural levels of expression are difficult to decode because they are not always independent, and not always monotonic — that is, separate factors can synergise, and abnormally low or abnormally high activities at one level can produce one and the same effect at higher-order levels. Thus slight differences in networks of genes, proteins, and neurones lead not only to autism but also to a continuum of human cognitive diversity, and categorical diagnostic distinctions are determined not only by biology but also by the scientific culture within which the diagnosis is framed.

Presenting psychology experiments as games isn't a new idea — but if you've ever played one of these games you know that the underlying science experiment often is only thinly veiled. Unfortunately, the more controlled and repeatable an experimental task is from a scientist's point of view, the more repetitive and tedious the game designed around it can seem from the player's point of view. This tension between experimental control and ecological validity makes not just boring games, but also bad science: a player who isn't engaged with the game doesn't express the same level of cognitive skills that (s)he does in the real world, and the game therefore becomes a poor vehicle for measuring and training these real-world cognitive skills. But what if well controlled science didn't have to be boring? My research group is answering this challenge by embedding experiments in an innovative video game which we use to study autism. Created by neuroscientists and game developers working cooperatively, Astropolis is designed to accommodate and to appeal both to players with autism spectrum conditions and to non-autistic players with whom they can be compared. Although each individual player is different, autistic cognition is typified by emphasis on visual rather than verbal coding, perceptual reliance on local detail rather than context or gestalt, superior focused attention with difficulty shifting and integrating attention to multiple simultaneous events, and untimed performance that may far outpace timed performance. Astropolis consists of core software for event logging and real-time synchronisation, a main game that is themed around the construction of a simulated space colony and which appeals to autistic persons' attraction to strongly rule-based simulations, and a collection of mini-games each of which allows the player to collect resources for the colony or to defend the colony against hazards. Each mini-game has its own plot and its own set of embedded experiments which include measures of perceptual psychophysics, attention, executive function and social cognition. Astropolis is open-source software available and modifiable under the terms of the GNU General Public Licence, and the research data that we are collecting with it will, after anonymisation and archiving, also be public to the autism research community. The longer-term goal is to make Astropolis a platform not only to measure autistic cognitive skills, but also to facilitate and to train autistic cognitive skills by removing barriers to their expression, and to serve as an activity around which social exchange can develop.