Building tasks from verbal instructions
With 66 people under EEG, extra rules in the instructions worsen a new task. Practising the extra rule is not what installs it.
The idea
Configuring a novel task from verbal instructions is effortful, and it tracks general intelligence. Earlier work had found that rules in the instructions can keep influencing behaviour even when they are not needed for the upcoming task. One account is that the practice trials before the experiment write those extra rules into the person's model of the task.
The approach
EEGs were recorded while 66 participants did a multirule task that crossed the number of rules in the instructions with whether practice trials exposed them to the extra rule.
Performance was worse when the instructions specified more rules. Practice with the extra rule did not change that, so an instruction that was never acted on was enough to put the rule into the task model. Instruction complexity came with a phasic rise in frontal theta synchronisation and less posterior alpha and beta desynchronisation, during a stretch of low intertrial phase coherence. The paper reads that transient as people checking the task rules mid-trial.
- 66 participants, EEG during a multirule task
- More rules in the instructions, poorer performance
- Practice on the extra rule is not required to install it
- Frontal theta up, posterior alpha and beta desynchronisation down
- Roberts, Jones, Davis, Ly, and Anderson. Cognitive, Affective, & Behavioral Neuroscience, 14, 1356-1374
The paper: link.springer.com .