How can we know what babies are thinking?

My son Jim loved his bottle when he was a baby. By about 15 months of age, he loved baby formula so much that he was going through over a hundred dollars' worth a week -- more than the rest of the food budget for the entire family! (Yes, we were buying the powdered stuff, not pre-made formula.) There were weeks when we completely exhausted the local grocery store's supply.

Needless to say, soon his pediatrician pointed out he was gaining weight too quickly, and we should cut his rations down to, say, three bottles a day. It was a painful transition. Previously, all Jim would have to do was say "bottle" (which he actually pronounced "BAW-bull"), and one would be provided. Now we gave him one for breakfast, one at lunchtime, and one at bedtime -- no exceptions. Jim would put on his best adorable face, pleading "BAW-bull? BAW-bull?" and we'd say "no, Jimmy, no bottle." He'd grab one of our hands, pull us into the kitchen, and point up to the cupboard where he knew the formula was stored: "BAW-BULL!" It was heart-wrenchingly pathetic. He seemed to be thinking we simply didn't understand what he was asking for.

But figuring out what's going on inside of someone else's head is a complicated proposition, especially when one of you has limited language skills -- just as Jimmy had trouble figuring out what we were thinking, psychologists have had to resort to innovative methods to understand what babies are thinking. Probably the most common method for studying infant cognition is simply observing what babies are looking at (we've discussed several such studies), with the presumption that they look at surprising, interesting, or different stimuli longer. But perhaps such research is overreading the looking behavior. It's possible, for example, that babies simply look longer at things that are more arousing, with no deeper interest.

In response to these criticisms, in 1997 a team led by Yuko Munakata developed a different sort of test on 7-month-olds. They built a device that displayed a toy on a ledge, out of the baby's reach. A button in front of the child caused the ledge to drop, and the toy to slide down a ramp to the baby. Are babies smart enough to know that pushing the button would allow them to retrieve the toy? The experimenters designed an experiment with four conditions: with and without a toy perched on the ledge, and with a transparent or opaque screen in front of the ledge, which revealed or hid the toy (or empty ledge) from view. Here are the results:

The babies pushed the button more often when they could see the toy, and less often when they could see there was no toy. Munakata et al. interpreted these results to mean that babies are able to use one thing (the button) in order to get another (the toy) -- a much more sophisticated line of reasoning than simply grabbing a toy.

But even this research was criticized. Perhaps babies had simply been trained to push the button: they pushed the button most when they were more aroused -- when they could see a toy.

Munakata went back to work, with a new team of researchers. This time, they modified their device so that pressing the button did not drop the ledge and instead lit a string of lights -- everything else about the experiment was the same. Now that they weren't rewarded with a toy, instead of pushing the button more often when a toy was visible, babies pushed the button at the same frequency, regardless of whether a toy was present.

Next, they conducted a new version of the study. In this version, babies were trained to pull a string which caused the button to lower and drop the ledge, releasing the toy. They were seated too low to reach the button. Though they could see the string lowered the button, they weren't trained specifically on button-pushing. Next their chairs were raised so they could reach the button, and their parents demonstrated that they could push the button and retrieve the toy. Finally, the original 1997 experiment was repeated, with these results:

As in the original study, and even without specific button-pushing training, when babies could see the toy, they pushed the button significantly more often than when the toy was not visible. When the screen was opaque, the difference between button-pushing with a toy and with no toy was not significant.

Munakata et al. interpret these results as strong evidence that babies as young as 7 months old can use one object as a means to obtaining another, rather than merely responding more frequently when they see an arousing object. They don't go so far as to endorse the preferential looking method used in so many other experiments, however, since this method can often be open to other potential interpretations. I should point out that this is not always so: in many cases, the purpose of the experiment is simply to determine if the infant is capable of distinguishing between different visual stimuli. In others, preferential looking can be used as one piece of evidence, perhaps not conclusive, but also possibly the only way of studying a phenomenon.

Manakuta, Y., McClelland, J.L., Johnson, M.H., & Siegler, R. (1997). Rethinking infant knowledge: Toward an adaptive process account of successes and failures in object permanence tasks. Psychological Review, 104, 686-713.

Manakuta, Y., Bauer, D., Stackhouse, T., Landgraf, L., & Huddleston, J. (2002). Rich interpretation vs. deflationary accounts in cognitive development: The case of ends-means skills in 7-month-old infants. Cognition, 83, B43-B53.