I present to you the findings which show that we can recognise objects both as literal things and symbols, whereas animals treat all objects in the world literally. It is evidence based on empirical observation of the world and which is thus factual for believers and non-believers alike. It is an outcome of an impressive work of science spanning several continents and nearly a thousand of humans and animals, published this April in the Proceedings of the National Academy of Sciences of France, doi.org/10.1073/pnas.2023123118.

The basis of the paper boils down to a simple idea. Geometric shapes can be broken down into abstract elements – right angles, parallel lines and symmetry. Only humans have the ability to grasp and utilise the immaterial idea of these elements. We subdivide any geometric shape into these more abstracts concepts without even realising. This is analogous in words, numbers and music notes. There is no need to go into more detail over the intricacies and arguments about this conception. The relevant point for the study is that humans do not learn lists of right angles (or any other symbol) by heart. Once we understand one, we can abstract it from any perceived shape and use it as a concept to understand the world. Animals can be taught really well to recognise particular objects by being rewarded with food. They can learn a list of squares or hinges by heart, but only as signs for incoming food. In this respect, a square does not mean anything geometric or symmetric. It is just a visually specific shorthand for food with no further sub-features.

The researchers behind this paper worked with baboon monkeys, who are believed to have one of the most similar visual systems to humans. They taught them to identify geometric quadrilaterals (four-sided) such as rectangles, parallelograms and trapezoids and to tell them apart from other irregular polygons. They were taught that these shapes, being the odd ones out, are a sign for a reward in the form of food. Once the researchers were sure that the baboons were able to identify the various geometric shapes and understood the task of finding the odd one out, they experimentally compared them on the full shape recognition task with humans.

A human and an ape not being similar

Both humans and baboons saw a collection of scattered geometric shapes. They were meant to indicate whether they saw a shape that did not fit into the collection. For instance, a collection of right kites with one lacking a right angle. The wrong shape was always geometrically irregular. An animal was then rewarded with food for a successful identification of an irregular shape. To clarify, this meant that a monkey would identify a wrong shape among those it had already been exposed to in its training. The difference from their original learning sessions was focus on all the used types of quadrilaterals simultaneously, not just one at time. The task was the same for humans, just without a reward. They were also just meant to indicate the wrong shape. Importantly, the researchers manipulated the presence of underlying geometric features by utilising eleven different quadrilaterals ranging from the most regular like squares to those almost irregular like iso-trapezoids. They then went to see whether geometric regularity of the shape predicted performance.

The results were unanimously convincing. The human ability to see geometric patterns was directly related to shape complexity. The more irregular the shape, the worse was the human performance. The monkey’s ability to find any one shape improved over time due to learning to pair correct answers with food but turned out to be the same for any shape no matter its complexity. A square and an iso-trapesoid posed the same challenge. This means that the animals could not abstract any geometric elements from the shapes. They did not use any geometric principles when trying to find the shapes, they just relied on memory of older instances of the same shapes without making any abstraction or connection between them. The humans on the other hand saw the shapes as a collection of geometric features they abstracted to solve the task more effectively for geometrically regular shapes. This way, humans demonstrated their ability to draw on abstraction of geometric rules to tell the shapes apart more effectively.

An objection can arise that the analysis was too linear. Therefore, the authors provided a secondary analysis using machine learning modelling. They tried predicting the behaviour by two artificial intelligence models. One assumed that each shape has a certain amount of regular geometric properties and predicted behaviour based on a simulated brain encoding geometric rules when deciding about the odd shape out. The other model only treated the shape as a collection of dots that was processed in cortical layers from bottom up and was adapted from previous visual perception research. The first model predicted human behaviour. The second was successful with the monkeys.

A common objection to the conclusions can be found in the carpeted world hypothesis. It says that humans are primed since birth to detect right angles, parallel lines and symmetry. We are merely habituated to them instead of having to construct any immaterial abstractions. However, this was addressed as the researchers also used a cohort of people from the Himba tribe which is one of the few still completely pre-industrial societies. Their living environment contains very little geometric regularity. Even these people who had no exposure to mathematical education behaved in a very similar manner to French adults and another control cohort, preschool children. On the other hand, the monkeys used for these experiments were tested and kept in proximity to facilities built in an industrial fashion. These included plenty of right angles and parallel lines in much greater abundance than a Himba village. It becomes quite apparent that the objection cannot hold.

There is another objection about making two fake categories of symbolic and non-symbolic thought. This objection can be supported by the claim that there was a degradation in this effect of complexity of the geometric shape on success in humans. It was true that the children were not affected by complexity as much as the adults and the performance of the Himba people was even poorer than that of the children, seemingly approaching the monkeys. But there was still a massive leap between the abilities of the tribesmen and the smartest of the monkeys. Some would then make the case that it would have been interesting to see whether this degradation continued beyond the baboons, but that would be missing the point of how massive and nonlinear the human-animal leap was. A degradation relative to the baboons would be relevant only for the animal kingdom and purely indicate anatomic deficiency in the visual processing system of the brain, but the principle of difference in symbolic thought would remain.

The outcomes of this experiment are twofold. On the one hand, it provides some powerful and strong evidence which can be kept at the back of one’s mind when discussing the issue of human uniqueness. But it also points to a deeper philosophical reality. Atop of the abilities of animals, our existence is characterised by the ability to understand and reason using immaterial truths. Reversing the sentence, immaterial truths make us human.