False Beliefs

I imagine many readers of this blog will have already seen Paul Bloom’s NYT magazine article on the moral life of babies. There’s a lot of interesting stuff in there, but I wanted to focus on something about false beliefs that surprised me. Here’s what Bloom says.

The new studies found that babies have an actual understanding of mental life: they have some grasp of how people think and why they act as they do. The studies showed that, though babies expect inanimate objects to move as the result of push-pull interactions, they expect people to move rationally in accordance with their beliefs and desires: babies show surprise when someone takes a roundabout path to something he wants. They expect someone who reaches for an object to reach for the same object later, even if its location has changed. And well before their 2nd birthdays, babies are sharp enough to know that other people can have false beliefs. The psychologists Kristine Onishi and Renée Baillargeon have found that 15-month-olds expect that if a person sees an object in one box, and then the object is moved to another box when the person isn’t looking, the person will later reach into the box where he first saw the object, not the box where it actually is. That is, toddlers have a mental model not merely of the world but of the world as understood by someone else.

I think the Onishi and Baillaregon paper he is referring to is Do 15-Month-Old Infants Understand False Beliefs?, which isn’t behind a paywall. It’s a very interesting study. It certainly seems like a decent challenge to the (Rutgers-inspired) view that children don’t understand that beliefs can be false until they are nearly 4.

I’m not an expert on any of this stuff, so I’m probably missing a lot, but here’s a crude summary of what the experiments seem to show. The more difficult you make the task you set an infant in a false belief task, the later they make the correct ‘predictions’. That’s not too surprising, but what is surprising is how much difference this can make. If you just see what a baby expects by tracking where it looks, then 15-month olds have expectations that allow for the falsity of beliefs. If you try to get the child to explain the action of an agent with false beliefs, or complete some other kind of demanding verbal task, they don’t allow for false beliefs until they are well into their 4th year.

This seems to suggest, to me at least, a kind of System 1/System 2 story. The automatic system that controls things like eye movements, surprise reactions and the like, allows for false beliefs from a very early age. But when babies get the capacity to reflectively reason (which I assume is much later than 15 months) their reflective thought seems not to allow for false beliefs. They don’t incorporate false beliefs into their explicit reasoning for years and years after their automatic processes are sensitive to them.

As I said, this is all a guess based on non-expert reading of a few studies. The main thing I wanted to highlight here is that people should be reading the new studies, especially people (like me) who didn’t realised that the experimental data on infant’s theory of mind look a lot more complicated now than they did in the 1980s.

4 Replies to “False Beliefs”

  1. It’s not just eye gaze and looking times. For false belief attribution manifested in helping behavior among 18 month olds, see Buttelmann, D., Carpenter, M., & Tomasello, M. (2009), “Eighteen-month-olds show false belief understanding in an active helping paradigm,” Cognition, 112, 337-42. —System 1 or 2?

    Also, folks might find useful Baillargeon’s very recent TICS survey, since there’s been a fair amount of stuff since 2005. I think that also discusses the executive control hypotheses often invoked to explain failure on verbal false belief tasks, etc.

  2. The new studies are indeed highly interesting, and there is considerable dispute over how they should be understood. Not everyone agrees with Bloom’s reading of Onishi and Baillargeon’s results — that 15-month-olds exhibit an understanding of false belief. Josef Perner and Ted Ruffman have argued that toddlers are not recognizing anything as deep as belief, but are instead tracking some much more superficial relationships between agents, objects and locations — see their 2005 rebuttal to O&B’s Science article on Ruffman’s website.

    It is a mystery why the skills that toddlers exhibit at 15 months are lost by 30 months, but I don’t think it’s that the capacity to reason reflectively is choking out the early ability. Reflective reasoning is not involved in the sorts of false belief tasks that 3-year-olds fail and 5-year-olds pass — our representation of the false beliefs in those cases is automatic and effortless.

    That said, I like your suggestion that there are multiple systems involved in tracking the cognition of others. Ian Apperly has some very interesting work along just those lines, in particular his forthcoming Psychological Review paper (co-authored with Stephen Butterfill), which argues that mental state tracking might well be understood as multi-systematic in the same way that our number skills are. There are some skills we share with non-human animals (counting small sets of numbers, very fast, tracking object-agent-location relationships), and then some much deeper and uniquely human capacities (involving the cardinality principle, or involving full-blown desire and belief ascription, where beliefs and desires can be seen as combining in infinitely rich ways). Seeing these systems as working in tandem with each other then makes it a rather complicated question what should count as the recognition of belief. It also ends up generating a huge mess of questions about the relationships between our underlying concepts of mental states and our actual performance in tracking the cognition of others — including some really interesting work by Apperly’s sometime collaborator Dana Samson. Samson has some great work showing that we automatically devote cognitive resources to calculating what others can see even when doing so interferes with a task we are trying to do. If you read her study on this, you may never want to try to work in a coffeehouse again.

  3. Great topic, and interesting interpretations of the data. An alternative “multi-system” story has been proposed by Mascaro, Sperber and colleagues. They claim that what is responsible for the failure to pass the verbal Sally/Ann type tasks is not the children’s missing (or lost) ability to attribute false beliefs, but rather a lack of epistemic vigilance.

    It’s not that the 2 and 3 year old children don’t attribute false beliefs, it’s just that they are unwilling/unable to go against the authority of other agents and judge communicators as malevolent or incompetent. Their claim is that epistemic vigilance is developmentally and functionally dissociable from Theory of Mind, which, if Bloom, Onishi and Baillaregon, Tomasello et al and all the other studies that go against the earlier received view are right, develops much earlier.

    The results of three studies investigating the hypothesis was published in a 2009 issue of Cognition: http://www.dan.sperber.fr/?p=733

  4. Just a general framing issue, already implicit, perhaps, in some of your remarks and those of other commenters:

    Major cognitive developments aren’t likely to happen quickly. They will be distributed in time. Therefore, we should expect that some equally legitimate tests of the very same ability will show earlier success and some will show later success. This distribution might be very temporally extended for some abilities. Developmental psychologists tend to forget this and privilege their favorite tests as the ones revealing the one age at which children “really” acquire the understanding in question.

    (I’ve written a bit on this, by the way, in “Gradual Belief Change in Children” in _Human Development, 1999, and (with Tori McGeer) in “Disorder in the Representational Warehouse” in Child Development, 2006.)

Leave a Reply