Metarepresentations Explain Human Uniqueness
Humans alone fluently understand the mental states of others. Humans alone rely on an open-ended system of communication. Humans alone ponder the reasons for their beliefs. For each of these feats, and for others too, humans rely on their most special gift: the ability to represent representations—the ability to form metarepresentations. Hidden behind such mundane thoughts as "Mary believes that Paul believes that it's going to rain" is the explanation of human uniqueness.
There are two ways to represent representations: one immensely powerful, the other rather clumsy. The clumsy way is to create a new representation for every representation that needs to be represented. Using such a device, Mary would have to form a representation "Paul believes that it's going to rain" completely independent of her representation "it's going to rain." She would then have to learn anew all of the inferences that can be drawn from "Paul believe it's going to rain," such as the negative impact on the willingness to go for a jog or the increased probability to fetch an umbrella. This cumbersome process would have to be repeated for each new representation that Mary wishes to attribute, from "Peter things the weather looks lovely" to "Ruth is afraid that the Dow Jones is going to crash tomorrow." Such a process could not possibly account for humans' amazing abilities to attribute just about any thought to other people. How can we account for these skills then?
The explanation is that we use our own representations to attribute thoughts to others. When Mary wants to attribute to Paul the belief "it's going to rain," she 'simply' uses her representation "it's going to rain" and embeds it in a metarepresentation: "Paul thinks "it's going to rain."" Because the same representation is used, Mary can take advantage of the inferences that she could draw from "it's going to rain" to draw inferences from "Paul believes that "it's going to rain."" This trick opened for humans the doors to an unparalleled understanding of their social environment.
Most of the beliefs we form about others are derived from communication: people keep telling us what they believe, want, desire, fear, love… Here again, metarepresentations play a crucial role, since understanding language requires going from utterances—"It's going to rain"—to metarepresentations—"Paul means that "it will soon rain here.""
Mentalizing (attributing thoughts to others) and communicating are the most well known uses of metarepresentations, but they are not the only ones. Metarepresentations are also essential for people to be able to think about reasons. Specific metarepresentations are relied on when people produce and evaluate arguments, as in: "Mary thinks "it's going to rain" is a good argument for "we should not go out."" Again, Mary uses her representation "it's going to rain" but, instead of attributing it to someone else, she represents its strength as a reason to accept a given conclusion.
Several other properties of representations can be represented, from their esthetic value to their normative status. The representational richness made possible by recycling our own representations to represent other people's representations, or to represent other attributes of representations, is our most distinctive trait, one of these amazingly brilliant solutions that natural selection stumbles upon. However, if it is indeed much simpler to rely on this type of metarepresentations than on the cumbersome solution of creating new representations from scratch every time, we still face a complex computational task.
Using the example of mentalizing, it is apparent that even when we use our own representations to attribute representations to other people, a lot of work remains to be done. It cannot be metarepresentations all the way down: at some point, other inputs—linguistic or behavioral cues—have to be used to attribute representations. Moreover, when a representation is represented not all of the inferences that can be drawn from it are relevant. When Mary believes that John believes it's going to rain, some of the inferences that she would draw from "it's going to rain" may not be attributable to John—maybe he doesn't mind jogging in the rain for instance. Other inferences Mary may not spontaneously draw—maybe John will be worried because he has left his book outside. Still, without a baseline—Mary's own representation—the task would jump from merely difficult to utterly intractable.
Probably more than any other cognitive trait, the ability to use our own representations to represent representations is what explains humans' achievements. Without this skill, the complex forms of social cognition that characterize our species would have been all but impossible. It is also critical for us psychologists to understand these ideas if we want to continue our forays into human cognition.
I leave the last word to Dan Sperber who, more than any other cognitive scientists, has made of metarepresentations the most central explanation of humans' unique cognition: "Humans have the ability to represent representations. I would argue that this meta-representational ability is as distinctive of humans, and as important in understanding their behaviour, as is echolocation for bats."