Perseveration; Perseverative error; Stage IV error

The A-not-B error arises from the difficulty of switching an action directed toward one location toward a new location. The original set-up, called visible displacement, involves hiding a toy in location A while an infant is watching. The child is then allowed to search for it. If they retrieve the toy it is seen as an indication of object permanence – the understanding that objects continue to exist even if they are no longer perceived. After a number of repeats of this hide and seek procedure, the toy is then obviously hidden in a different location, B. Up to a certain developmental stage, infants persevere in searching in location A instead of B. This is called the A-not-B error. This behavior can be observed in reaching tasks, but also in larger spatial set-ups that require detours.

When Piaget published his theory of child development in 1937 (1954 in English), it quickly became absorbed by comparative psychologists because it offered a suitable framework for interspecies comparison. There are two reasons for this: first, Piaget was originally trained as a zoologist (the subject of his Ph.D. thesis was the snail), so his ideas are expressed within a broadly biological frame; and second, he denied that language develops prior to cognition, so his experimental tasks were designed to test cognition without relying on language. Much of his theorizing on early development was based on observable, nonverbal behavior, which makes it ideal for the study of animals (Pepperberg 2002).

According to Piaget, the A-not-B error occurs in Stage IV of the sensorimotor period (0–24 months) of human development. During stages I, II, and III infants acquire fundamental perceptual and motor abilities enabling them to perform search behavior. They are not yet able to search for an object that is no longer visible: it ceases to exist in the perceptual world of the infant and they therefore do not possess the concept of object permanence. Stage IV (from around 8 months) is defined by the ability of the infant to search for an object that has moved (or has been moved) out of sight. Here children are very likely to show the so-called A-not-B error, or Stage IV error, in which they continue to search at a previous hiding place (A) although the object was obviously hidden at a new place (B) (see Diamond 1990, for a review). A number of factors can affect the occurrence of the perseveration error, such as the delay between the hiding and retrieval, the number of possible hiding places, the motor requirements, and obviously the age of the infant (Markovitch and Zelazo 1999).

It was initially thought that children in Stage IV (approximately between 8 and 12 months) simply do not understand the concept of an object as continuing to exist out of sight, and in one place only. In certain set-ups, the error could be caused by associative learning as a repeated behavior is reinforced. At Stage V (from around 12 months), the A-not-B error no longer occurs.

Piaget himself was the first to apply his theory to animal behavior, albeit at a theoretical level (Piaget 1971). The first published experimental Piagetian animal research was done in 1971 by Gruber, Girgus, and Banuazizi who investigated the development of object permanence in cats. And in addition to a large number of studies with primates, object permanence has been studied in dogs and wolves (Gagnon and Doré 1994), chicks (Etienne 1973), hamsters (Thinus-Blanc and Scardigli 1981), and magpies (Pollok et al. 2000). Vauclair (1996) gives a detailed account of published studies in this area, as do Cacchione and Rakoczy (2017).

This order of the developmental stages has been applied to animals as it provides a valid and reliable base for a comparison among species. Especially in the area of object permanence, animals can be grouped by the highest stage they can master.

Etienne (1973) stated that animals react in three different ways to the disappearance of objects: animals like insects, spiders, and some lower vertebrates apply special devices or stereotyped movements to find an object of immediate survival value; most birds and some mammals, like rabbits, possess either learned or unconditional responses they apply to find an object that has disappeared, but with no persistence; some mammalian carnivores, primates, and birds of the family Corvidae show spontaneous and active search behavior in response to the disappearance of an object. This third group’s performance corresponds to at least Stage IV of Piaget’s ranking of object permanence performance.

When considering the current literature on A-not-B errors in nonhuman animals, it has to be taken into account that the task set-up often varies from Piaget’s original one. Most studies use variations of Uzgiris and Hunt’s (1989) sequence of tests where the task for successive visible displacement involves the hiding of the target (treat or toy) out of sight in one location before it is moved by the experimenter in plain sight to a second location. In the original tests by Piaget, the target was hidden and retrieved from location A several times before it is hidden in location B for another retrieval. A third, combined set-up was used by Amici et al. (2008). They hid the target in location A three times, then in the fourth trial they visibly moved it to location B after first hiding it at A. Zucca et al. (2007) repeated the hiding in location A twice before visibly hiding the target in location B. Most other current studies omit the repeating stage, such as Deppe et al. (2009) in their study with lemurs. These rather substantial differences in methodology need to be taken into account when considering and comparing findings. The method without the repeated retrieval excludes the explanation that the error is caused by motor memory. But it needs to be studied further whether errors in the visible displacement task with no repetition involve the same cognitive processes as the A-not-B error as initially investigated. Some authors suggest that social factors could explain the preference for the A location in certain species (Topál et al. 2008). Because the experimenter points to the location, interacts with it, and thereby teaches the animal where to look, it tends to focus on this location, instead of searching for the target independently. As seen in the history of animal cognition (Clever Hans – Pfungst 1911), every human involvement in an experiment can affect the outcome, especially in domesticated species such as dogs. This needs to be taken into account when designing and interpreting experiments, even if it does not apply to all species. Kis et al. (2012), for example, found that the “teaching” factor does not apply to common marmosets (Callithrix jacchus).

Many species tested show the A-not-B error during their early development, but not when they are adults. Their developmental cognitive changes follow the same path as those of humans. The only exception so far are domestic dogs (Canis lupus familiaris) and cats (Felis catus). Neither puppies (Gagnon and Doré 1994) nor kittens show the perseveration error (Dumas and Dore 1991) whereas adult cats do (Dumas and Doré 1989). For adult dogs, the results are inconclusive as different set-ups have been used in different studies. The specific methodology of each study must be taken into account before any comparisons between species can be made.

The A-not-B error involves the perseveration of a movement toward one location, and the inability to switch the movement away from it, despite the fact that the target was obviously moved toward a new location. The testing set-up for infants usually requires reaching for the target. But this behavior can also been seen on a larger scale, in detour tasks. The perseveration error occurs when participants attempt to use a previous route around a barrier despite a visible change in the correct path, thus rendering the attempted route impossible (McKenzie and Bigelow 1986). The basic set-up has the participant facing a straight barrier and they have to move around it through a gap at one end toward a target on the other side. The gap to walk through is in location A. After a few repeats, the barrier is moved so that the gap is now in location B, the other end of the barrier. There is no change in the size of the gap, or the difficulty in the task, only a change in location. This spatial set-up removes the experimenter input of touching or indicating the target locations and therefore excludes social explanations for any errors. Perseveration cannot be explained by simple motor memory, as the detours take several seconds and involve far more complex movements than simple reaching, as in the original task. McKenzy and Bigelow (1986) explored this detour behavior in human infants and found the spatial equivalent of the A-not-B error in young infants (10-months old), but not in 14-month-old children. This indicates that the same cognitive development applies to both the searching and the detour behavior.

Several studies have shown that dogs will follow a previously learned path even if this no longer leads directly to the goal. Buytendijk and Fischel (1932) were the first to describe how dogs persevere with a detour path, even after changes in the set-up would make an alternative path much faster and easier. Clarke et al. (1951) found that 10-month-old puppies showed perseveration errors in detour tasks. Puppies persisted in following the wrong, but previously learned paths even if this resulted in falling from the end of a false ramp in an elevated detour test. These findings were confirmed for adult dogs reared in restricted environments by Thompson (1954).

More recently, the spatial A-not-B error has been explored in dogs (Osthaus et al. 2010) as well as horses, mules, and donkeys (Osthaus et al. 2013). The number of repetitions of the A trial varied and therefore allowed an exploration of the influence of repetition on perseveration. Unlike in the search set-up, adult dogs persevere in the detour task, even after one A trial. Horses perform comparably, whereas mules and donkeys had lower rates of spatial perseverance and showed fewer A-not-B errors. So far, equines have not yet been tested for A-not-B errors involving the search set-up. Further tests show the same error in cats, chickens, and sheep (Osthaus, unpublished data).

The A-not-B reaching error has been shown to exist in most nonhuman species tested so far (Table 1). There are a few exceptions: apes (apart from orangutans), puppies, kittens, wolves, and goats. It must be taken into account that, although there is growing number of studies investigating object permanence in a wide range of species, most of them do not use specific tests to investigate the A-not-B error. In addition, the methodologies vary immensely between studies. Although many studies use the basic set of tests by Uzgiris and Hunt (1989) that explore successive visible displacements, the precise set-up varies between studies and makes comparisons near impossible. It is therefore difficult to compare species with regard to the A-not-B error until a standard test with a defined number of repetitions and movements between targets is implemented.

The A-not-B error is a behavior observed across the animal kingdom and its causes remain undiscovered to date. Only with standardized tests will it be possible to investigate and identify the underlying cognitive processes.