Chimps engage with pretend objects, suggesting they have imagination and can engage in pretense

Humans have the ability to do “secondary representations”: that is, to pretend that one object or action is actually different from a real one. This can also be called “pretense”. Examples are children’s “tea parties” in which they use empty pots and toy cups, pretending to drink from the empty cups while knowing they are empty. Then they can pour pretend tea into one of two cups, and when asked to drink will drink from the “pretend full” cup. Or they can have sword fights with sticks, pretending that the sticks are real weapons while knowing they are not.

Secondary representations of states that are only imagined start early (some experiments suggest at 15 months), and the ability to imagine things that haven’t happened, or aren’t real, surely underlie much of human behavior involving planning for the future or imagining what someone might be thinking.  The authors of a new paper in Science (see below) argue that no such abilities to “pretend” or have secondary representations are known from any species save humans. (I am not sure about this. As I recall some birds caching food are known to unhide it and re-cache it elsewhere if they see other birds looking on: something that seems like an ability represent another bird’s state of mind.)

And there is anecdotal evidence that chimps can do this.  For example, female chimps have been seen to hold and carry sticks as if they were their babies; this involves imagining that the stick is a real baby (that only females do this suggests sexually differentiated behavior).  Or if chimps have played with blocks, sometimes they’ve been observed to drag around imaginary strings of blocks.  This and other data suggest that some primates can have imaginative representations, but the existing data, say Bastos et al., don’t rule out other explanations.

They thus did three experiments on a single, human-acclimated male bonobo at a facility in Iowa. The bonobo, named Kanzi, was 43 years old and died the year after the experiment (no, he didn’t pretend to be dead!). Kanzi has his own Wikipedia page, which notes his abilities:

Kanzi is well known for his noteworthy cognitive abilities. He had a very well-documented linguistic understanding of the human language. He is believed to be the first non-human great ape to understand and comprehend spoken English. In addition, he was also heavily documented for his understanding and usage of symbols to communicate, usually through lexigrams and partial ASL. The vast amount of information that researchers gathered from Kanzi created a significant impact for the fields of linguistics and cognitive science. Kanzi’s behavior and abilities have been the topic of research published in scientific journals, as well as reports in popular media. He died in 2025, in Atlanta, Georgia.

Click below to go to the paper (pdf here), or you can read a summary of the study in the NYT, written by Alexa Robles-Gil, here (archived here)

Three experiments were involved, but the second was really a control for the first.

First, Kanzi was prepared for the pretense test by letting him learn about a real object: fruit juice that could be poured into glasses from a pitcher. In 18 trials, real juice was poured into one of two cups from a pitcher. Kanzi, who had been trained to point at what he wanted to have, was then asked, “Where’s  the juice?” He was successful in all trials.

Then the pretense experiment began. The same pouring was done, but from an empty pitcher into both of two empty cups.  Then one of the pretend-filled cups was poured back into the pitcher, so it would be pretend-empty while the other was pretend-full. Again, Kanzi was asked “Where’s the juice?”  In 50 trials, involving no reinforcement of any kind for making the correct choice, Kanzi chose the pretend-full cup 34 times and the pretend-empty cup 16 times, a highly significant deviation from an expectation of 50:50 under the null hypothesis. This showed that Kanzi could track where pretend juice was.

The second experiment used a cup of real juice next to an empty cup, and the empty cup was pretend-filled from an empty pitcher. Then Kanzi was asked, “Which one do you want?” Kanzi wanted the real juice in 14 out of 18 trials, again, a significant deviation from 50:50 under the null hypothesis. This showed that Kanzi didn’t simply believe that there was real juice in the empty cups in the first experiment, for he was able to distinguish real juice from pretend-poured juice.

The third experiment was like the second, except involving grapes. First, Kanzi was “trained to indicate the location of a real grape in one of two transparent jars after observing the experimenter sample a grape from a plastic container and place it into one of the jars and perform a control action on the other jar.” When asked to choose one jar to get the grape, he was successful in every one of 18 trials.

Then Kanzi was given pretend grapes to choose. From the paper;

In probe trials, the experimenter pretended to sample a grape from an empty container, then placed it inside one of the two jars, before repeating the same action on the other side. Then, one of the jars was pretend-emptied, and Kanzi was asked, “where’s the grape?” Kanzi succeeded at this conceptual replication even more quickly than in the first experiment. He correctly indicated the location of the remaining pretend grape above chance, in 31 out of 45 unreinforced probe trials

Again, Kanzi was highly successful at the juice and grape trials, able to recognize a pretend action of pouring and emptying juice, and determining which of two jars containing pretend-grapes had had the grape removed. In other words, he was playing tea party, and highly successfully.

This one chimp, then, was able to conceptualize pretend actions as real ones.

There are a number of possibilities not involving secondary representation that the authors say could be happening here. For example, apes like Kanzi who have been trained to recognize symbols to represent objects (as he was), might be better at communicating their wishes than are wild apes. Or symbol training could actually create the ability to do secondary representation. It’s hard to rule out these possibilities since to do such experiments an ape has to be “enculturated” by interaction with humans, and Kanzi was surely highly enculturated.

But if the authors are right that these experiments show that apes can have secondary representation, playing along with “pretense”, that opens up a world of possibilities of thinking about the cognitive abilities of apes (and other animals). The authors dwell on this at the end:

Secondary representations underlie many other complex cognitive capacities, such as imagining future possibilities (20) and mental state attribution (13). Our results may therefore help to interpret other bodies of data that have been hampered by an apparent logical problem (32). Finding that a bonobo can generate secondary representations in pretense contexts increases the likelihood that these representations are available for other cognitive functions. This finding reinforces growing evidence that apes track decoupled mental states, such as beliefs, rather than simply reading behavior (25, 28, 31). It also increases the likelihood that secondary representations could subserve future-oriented behavior (24, 35, 50–53), whose underlying representations have not yet been established.

In conclusion, our findings suggest that some nonhuman animals can generate secondary representations that are decoupled from reality, and that this capacity was likely within the cognitive potential of our last common ancestor with other apes, which lived 6 to 9 million years ago.

It is no surprise that our closest relative (along with chimps) could do this. As Darwin posited in 1871, our own behaviors and mental states evolved from those present in our common ancestors.

Kanzi died suddenly the year after the experiment, simply collapsing. He apparently suffered a heart attack, as he had a history of heart issues and had previously been obese. You can read about his other training in representing objects with keyboard symbols at the Wikipedia site.

From Wikipedia, here’s Kanzi in 2006 (he died in 2025):