In a recent study, Courtenay Haswell and colleagues investigated the ability of children with autism to learn a motor action, to counteract a perturbing force, and to generalize learning to other 'workspaces'.
Participants played a game in which they had to 'catch' animals on a computer display by moving a robotic arm which controlled the cursor on a projected display. Trials varied according to the starting position of the arm ('workspace') and the direction of motion required to capture the animal.
After an extended practice session, a perturbing force was added to the robot arm in one condition only (arm in the LEFT workspace and target capture requiring movement directly ahead). The force dragged the arm anti-clockwise (ie to the participants' left) but they were able to counteract the force and move the cursor towards the animal.
The interesting question was how this adaptation would transfer to other conditions where the arm started in the RIGHT workspace (off to the participants right) where no perturbing force had ever been applied.
Both groups of participants did show generalization, applying more force to the robot arm. However, there was a subtle difference in how the two groups generalized their motor learning. Generalization was strongest when target capture required the same joint movement and, in fact, participants with autism showed stronger generalization than control participants (ie applying more force to the robot arm). There was also weak generalization when target capture required different joint movement but the same hand motion (directly ahead) as it did in the LEFT workspace. Here, the roles were reversed, with the control group showing stronger generalization (more force to the robot arm) than those with autism.
The authors concluded that participants with autism are more likely to encode their movements in terms of intrinsic coordinates (ie their own muscle movements) rather than extrinsic coordinates (ie the task setting).
The authors also discussed the possible neural underpinnings of their findings. Studies of monkeys performing similar tasks suggest that intrinsic coordinates are encoded in primary motor cortex, whereas extrinsic coordinates are encoded in pre-motor areas. Primary motor cortex has short-range connections to somatosensory cortex, whereas premotor cortex has longer-range connections to the posterior parietal cortex. The authors argued, therefore, that the tendency of individuals with autism in their study to generalize more in terms of joint movement than hand direction was a function of impaired long-range connectivity.
However, it's important to remember that no brain activity was recorded in this study, and that several leaps have to be made to arrive at this conclusion. That said, it does fit nicely with recent studies looking at brain connectivity in autism.
The authors also conducted a number of correlation analyses. Recent research using diffusion tensor magnetic resonance imaging has found a correlation between the the integrity of different white matter connections in the brain and specific symptoms of autism. If anomalous brain connectivity is at the root of the atypical motor generalization in autism, we might expect there to be strong associations between measures of motor generalization and specific symptoms.
Three significant correlations are reported:
- Within the autism group, participants with higher scores (more severe impairment) on the social interaction scale of the ADOS-G applied more force in the intrinsic generalization condition. It's unclear whether the same measure correlated with other scales of the ADOS-G and therefore whether this merely an index of severity. There are also concerns about using the ADOS-G subscales as ordinal scales when they are designed to provide a YES/NO cut-off.
- Across the two groups, participants with higher scores (more severe impairment) on the Social Responsiveness Scale applied more force in the intrinsic generalization condition. However, the correlation is entirely driven by the group differences in the two measures (ie people with ASD have higher SRS scores and less generalization but there's no evidence for a relationship between the two measures within each group). As such, this analysis doesn't tell us anything that we don't already know from the group comparisons.
- Finally, across the two groups, performance on an imitation task was correlated with a measure derived from the difference between force applied in the two generalization conditions (ie Intrinsic - Extrinsic). Again, this correlation is entirely driven by group differences. In other words, kids with autism are (unsurprisingly) impaired at imitation and, as we already know, show different patterns of generalization. This gives us our correlation, but doesn't tell us anything about the link between the two variables within either group. It's also unclear why the authors chose this difference index for this analysis, but used the intrinsic measures for the other correlational analyses.
Haswell, C.C., Izawa, J., Dowell, L. R., Mostofsky, S. H., & Shadmehr, R. (2009). Representation of internal models of action in the autistic brain. Nature Neuroscience, 12 , 970-972.