For decades, the standard object-choice test has been a bad day for chimpanzees. An experimenter hides a reward in one of two containers, points at the correct one, and the ape, more often than not, performs at chance. That pattern is the empirical backbone of a strong claim: chimpanzees do not understand that a point is meant to help.

Alicia Melis at University College London, Federico Rossano at the University of California, San Diego, and colleagues report a different result from nine chimpanzees at Ngamba Island Chimpanzee Sanctuary in Uganda. In iScience, the group succeeded when the point came with the chimpanzee’s name and with food grunts. The group did not beat chance when the same point came in silence, or when the experimenter looked and food-grunted without pointing at all.

They call it the attentional deficit account: the animals may fail the classic test because they are not attending to the person, especially when they are in what the paper terms a foraging mindset, rather than because they cannot read a cooperative intention. The new experiment is built to make the person hard to ignore. It does not isolate which sound did the work.

The key that could open only one box

Each trial used two identical hiding boxes, 150 centimetres apart, and a collaboration box that held grapes. The tools needed to get those grapes, a rake and a pusher, were hidden in one of the two boxes. A key sat three metres in front of them, equally far from both. Once the key went into a lock it could not come out, so each trial was a single choice.

Experimenter 2 squatted 130 centimetres from the bars and, depending on the condition, cued the correct box before the chimpanzee entered. In the attention-enhanced condition the experimenter called the subject’s name, looked at the correct box, produced four food-grunt vocalisations, and held a continuous contralateral point, the arm crossing the body midline, with a slight body orientation. In the baseline pointing condition the sounds were stripped out: the same slight body orientation, the same contralateral point, and gaze, nothing audible. In the non-intentional condition the experimenter looked and food-grunted with a slight body orientation and without pointing.

Each chimpanzee completed two sessions of six test trials per condition, twelve trials per condition, 36 test trials in all. Condition order was counterbalanced. Each session began with four warm-up trials in which nothing was hidden and no cues were given, so opening a box in the absence of information was not rewarded. Baiting was done while a caregiver distracted the subject. The paper’s participant table lists six females and three males. Ages in that table run from 20 (Medina) to 40 (Kidogo). The STAR Methods prose says 26 to 41. The table is the named list; the prose range does not match it.

All nine animals already knew the apparatuses from earlier work at the sanctuary. Four of them had been in a previous object-choice pointing study. They are sanctuary-housed, described as only occasionally exposed to humans rather than language-trained or fully enculturated, and they are not a wild sample.

Scores from the nine chimpanzees

As a group, the chimpanzees were correct on 73 percent of attention-enhanced trials, 53 percent of silent-pointing trials, and 46 percent of look-and-grunt trials. Chance is 50 percent. Only the 73 percent figure beat chance on a Wilcoxon signed-rank test (Z = minus 2.38, p = 0.016).

A mixed-effects model with a random intercept for subject found that condition improved fit (chi-squared with two degrees of freedom = 19.17, p less than 0.001). Relative to the attention-enhanced condition, the odds of a correct choice were 62 percent lower in silent pointing (odds ratio 0.38, 95 percent confidence interval 0.21 to 0.69). They were 71 percent lower in the non-intentional condition (odds ratio 0.29, 0.16 to 0.52). Session order, trial number and which side was baited did not predict accuracy.

Individually, on two-tailed binomial tests against 12 trials, three animals were above chance in the attention-enhanced condition: Indi 11 out of 12, Medina 12 out of 12, Pasa 10 out of 12. In silent pointing, only Indi was above chance (11 out of 12). In the non-intentional condition, nobody was (all scores 7 out of 12 or lower). Six of the nine scored nine or more correct in the attention-enhanced condition, against one of nine in silent pointing and none in the non-intentional condition. The authors also treat nine out of 12 as significant on a one-tailed test (they report p = 0.037); that count is not the same as nine chimpanzees each passing a two-tailed test. The group result is not the same as nine chimpanzees each passing the test.

The paper notes that the same individuals who succeeded with the full cue package were at chance when the point was removed. Food grunts and gaze, without an indexical gesture, were not enough. A silent point, without the sounds, was not enough either.

The enculturation argument

A common reading of earlier successes is that only chimpanzees raised in close contact with people learn to treat a human point as information. Melis and colleagues argue that their sanctuary animals are not that population, and that once attention is engaged the point can be used. They suggest the difference between enculturated and non-enculturated chimpanzees may lie more in how readily they attend to a human than in a long history of pairing points with rewards.

They compare the result with a much larger sample (N = 138) reported by Hopkins and colleagues, in which pointing with food grunts and name-calling, and pointing with name-calling, both beat food grunts without pointing. That study used a proximal ipsilateral point, the finger next to the target, so local enhancement is harder to rule out, and most of those animals had long testing histories. Forty-three percent of individuals in that paper were counted as passing, pooled across conditions. The Ngamba setup puts the point at a distance and across the midline, which the authors treat as a stronger guard against simply approaching the nearest hand.

They also put the 73 percent group figure next to published object-choice numbers for human infants: 64 percent at 14 months and 77 percent at 18 months in the comparison they cite. That is a group-to-group juxtaposition, not a claim that a Ngamba chimpanzee is an 18-month-old child.

Limits the nine chimpanzees leave open

The study cannot say whether the name or the food grunts did the lifting. The limitations section is explicit: the cues may have acted together, and a follow-up would have to run pointing with one sound at a time. They already suspect, from earlier work in which name-calling plus pointing did not always work, that the pile of cues is the point.

N is nine. Three animals carry the two-tailed individual result. Four of the nine had seen a pointing object-choice task before. All nine knew the boxes. The sample is adult sanctuary chimpanzees with mixed rescue backgrounds, including the pet trade and entertainment, not a wild community and not a language-trained laboratory group. The paper’s stronger sentence, that chimpanzees understand cooperative communicative intentions, is the hypothesis the results are "in line with." The measurement is a box choice after a multimodal cue.

Kirchhofer and colleagues’ earlier pointing study is in the paper as the prior test four subjects had taken. Colleague pieces on this site have covered chimpanzees pointing to guide a person to hidden food, and bonobos pointing in conversation. Those are different studies. A Science Blog piece from 26 August this year on white sclera in great apes is about the appearance of the eye, not this object-choice design.

When a person pointed in silence, this group was at chance. When the person pointed and made the chimpanzee’s name and a food call part of the same act, the group was not.