ScienceDaily has a report on a paper by K. Allen, S. Ibara, A. Seymour, and N. Botvinick published in Psychological Science, Abstract structural representations of goal-directed behaviour. (here) They draw parallels between syntax in language and how we understand the actions of others.
There are oceans and oceans of work on how we understand languages and how we interpret the things other people say
the same principle might be applied to understanding actions. For example, if you see someone buy a ticket, give it to the attendant, and ride on the carousel, you understand that exchanging money for a piece of paper gave him the right to get on the round thing and go around in circles. (Researchers used) action sequences that followed two contrasting kinds of syntax — a linear syntax, in which action A (buying a ticket) leads to action B (giving the ticket to the attendant), which leads to outcome C (riding the carousel), and another syntax in which actions A and B both independently lead to outcome C. They were testing whether the difference in structure affected the way that people read about the actions.
People can read sentences faster if they have the same syntax as the preceding sentence. The researchers found the similarly people could read sentences faster if the relationship between actions had the same pattern as the preceding sentence.
This indicates that readers’ minds had some kind of abstract representation of the ways goals and actions relate. … It’s the underlying knowledge structure that kind of glues actions together. Otherwise, you could watch somebody do something and say it’s just a random sequence of actions.
Here is the abstract:
Linguistic theory holds that the structure of a sentence can be described in abstract syntactic terms, independent of the specific words the sentence contains. Nonlinguistic behavior, including goal-directed action, is also theorized to have an underlying structural, or syntactic, organization. We propose that purposive action sequences are represented cognitively in terms of a means-ends parse, which is a formal specification of how actions fit together to accomplish desired outcomes. To test this theory, we leveraged the phenomenon of structural priming in two experiments. As predicted, participants read sentences describing action sequences faster when these sentences were presented amid other sentences sharing the same parse. Results from a second experiment indicate that the underlying representations relevant to observed action sequences are not strictly tied to language processing. Our results suggest that the structure of goal-directed behavior may be represented abstractly, independently of specific actions and goals, just as linguistic syntax is thought to stand independent of other levels of representation.
This seems a somewhat predictable notion but it is nice to have some confirmation of it.