, Volume 98, Issue 8, pp 693–698

Prospective thinking in a mustelid? Eira barbara (Carnivora) cache unripe fruits to consume them once ripened


    • Escuela de BiologíaUniversidad de Costa Rica, Ciudad Universitaria
    • Department of Biological SciencesMacquarie University
  • Isaías Alvarado-Díaz
    • La Selva Biological Station, Organization for Tropical Studies
Original Paper

DOI: 10.1007/s00114-011-0821-0

Cite this article as:
Soley, F.G. & Alvarado-Díaz, I. Naturwissenschaften (2011) 98: 693. doi:10.1007/s00114-011-0821-0


The ability of nonhuman animals to project individual actions into the future is a hotly debated topic. We describe the caching behaviour of tayras (Eira barbara) based on direct observations in the field, pictures from camera traps and radio telemetry, providing evidence that these mustelids pick and cache unripe fruit for future consumption. This is the first reported case of harvesting of unripe fruits by a nonhuman animal. Ripe fruits are readily taken by a variety of animals, and tayras might benefit by securing a food source before strong competition takes place. Unripe climacteric fruits need to be harvested when mature to ensure that they continue their ripening process, and tayras accurately choose mature stages of these fruits for caching. Tayras cache both native (sapote) and non-native (plantain) fruits that differ in morphology and developmental timeframes, showing sophisticated cognitive ability that might involve highly developed learning abilities and/or prospective thinking.


Goal-directed behaviourCachingFuture thinkingTayras


Many species of birds and mammals cache food items or tools that can be used in the future (Smith and Reichman 1984; Mulcahy and Call 2006; Suddendorf 2006). Several aspects of caching behaviour suggest the existence of complex cognitive abilities in these animals. Examples include adjustments in caching behaviour when pilferers are present (Dally et al. 2006; Leaver et al. 2007), the ability to remember the location and type of food cached (Clayton and Dickinson 1998) and the potential to override immediate needs in favour of future ones (Osvath and Osvath 2008). Nevertheless, even complex behaviours can sometimes represent fixed-action patterns or learnt associations that are triggered by environmental stimuli, undergoing limited cognitive processing (Shettleworth 2001; Raby and Clayton 2009).

Whether nonhuman animals think about the future has been debated for centuries (Dennet 1997; Descartes 1998; Suddendorf 2006; Raby and Clayton 2009). Such cognitive abilities can have profound ecological, evolutionary (Dukas 1998; Shettleworth 2001) and ethical implications (Armstrong and Botzler 2003, but see Verhoog and Visser 1997). Rather than tackling this question through an “all-or-nothing” approach, future-oriented behaviour can be placed into categories and is expected to vary between species (Shettleworth 2001; Raby and Clayton 2009). Goal-directed behaviour can be regarded as behaviour that doesn’t need perception of the future (i.e. fixed-action patterns and learnt associations), or behaviour that requires prospective thinking, which even for humans can be episodic (i.e. mental time travel) or semantic, depending on whether the individual thinking about the future is ‘picturing itself’ in that future or not (Raby and Clayton 2009).

The tayra (Eira barbara) is the sole species in its genus, and occurs from southern Mexico to northern Argentina (Janzen 1983; Presley 2000). It is a large (2.7–7 kg), slender mustelid with a muscular body, strong claws and long tail. Its diet is highly omnivorous, consisting of vertebrate and invertebrate prey in addition to feeding on a variety of ripe and rotting fruits, eggs and honey (Janzen 1983; Bisbal 1986; Presley 2000; Bezerra et al. 2009).

Since animals generally cache food that could otherwise be readily consumed (Smith and Reichman 1984, but see Dearing 1997), studies on prospective thinking during caching have typically focused on whether these animals can predict their future motivational state at the moment of caching (Correia et al. 2007; Raby et al. 2007; Raby and Clayton 2009). In this study, we focus instead on the type of food being cached to infer the possibility of prospective thinking. We discuss the singularity of this behaviour within the current conceptual framework.

Material and methods

This study was conducted at La Selva Biological Station in Costa Rica in July 2004 and from September to December 2006. La Selva comprises primary and secondary forests, but also includes small patches of pastures and abandoned plantations in secondary succession (McDade et al. 1994). Direct observations of behaviour from tayras were opportunistic, while carrying out other activities in the field. A total of 35 days were spent in the field, either at the forest or the plantations, during daylight hours.

We observed tayras retrieving cached, ripe plantains (Musa X paradisiaca) at a forestry plantation comprised of Cedrela odorata, Hieronyma alchorneoides and Cordia alliodora planted in alternation with palms (Euterpe precatoria and Euterpe oleracea). There were no shrubs or other undergrowth at this plantation and visibility across the understorey was high. This site was 80 m away from a small (484 m2) plantain plantation. Both plantations were surrounded by forest.

Removal of plantains from 36 fruiting plants was monitored at the plantation from October to December 2006, through the use of cameras and through visual inspection once or twice per week. Plantains were classified in five groups, depending on their degree of maturation. Classes I to IV constituted unripe plantains with a very firm, sticky pulp, which contains high levels of starch (Offem and Njoku 1993), and only class V referred to ripe plantains (Table 1). Six motion-sensitive cameras were placed at different class IV bunches, and rotated every few days.
Table 1

Percentage of plantains at each class of maturation that were left at the plant, taken from the bunch (and presumably cached) or eaten

n refers to the total number of plantains available in each class of maturation, white arrows point at class IV and class V plantains correspondingly

aWhen scars were found in fruit bunches that already contained ripe plantains, it was not possible to determine if the plantains had been taken from the bunch before or after they had ripened (class IV or V respectively). Therefore, these plantains could have been cached when unripe, cached when ripe, or taken from the plant but eaten immediately after on the ground

In contrast to smaller mammals and birds that consumed the plantains, when a tayra took a plantain it detached it from the base, leaving an obvious scar at the insertion point in the bunch (Fig. 1). When scars were found in bunches with unripe plantains, we assumed tayras were the only animals that removed these plantains since the camera traps did not reveal any other animal removing unripe plantains and also because unripe plantains in standing plants cannot be reached by peccaries, which were the only highly herbivorous mammals observed to visit the plantation. Because the plantation was monitored one or two times per week, scars were often observed in bunches where plantains had already ripened. In these cases, it was not possible to determine if plantains were removed when ripe or unripe. Consequently, when reporting the number of unripe plantains that were taken, we only considered scars left on unripe bunches. Tayras consumed ripe plantains in situ (i.e. at the fruit bunch), or took them from the plant and ate them on the ground at the forestry plantation (pers. obs.). Therefore, scars present on ripe bunches could represent plantains that were taken when unripe, plantains that were taken when ripe but eaten at the plantation, or ripe plantains that were potentially cached.
Fig. 1

Scar (indicated by arrow) left in a fruit bunch of plantains (Musa paradisiaca) after a tayra removed one of them, detaching it from its base. The plantain that was immediately to the left of the scar was cut off for clarity

Radio transmitters (AVM Instrument Company Ltd) were fitted inside class IV plantains by cutting a hole (3.5 cm × 0.6 cm × 3.5 cm deep) into the pulp of the fruit. After inserting the transmitter, the hole was covered with duct tape. If the plantain was not taken before it ripened, it was removed and the radio transmitter was placed into another class IV plantain. There were four radio transmitters available that were reused, so that a total of 22 unripe plantains were fitted with a radio transmitter. Plantains that were fitted with radio transmitters and removed by tayras were later located with a hand-held antenna.

To determine if tayras were caching plantains in sites that were less likely to be discovered by pilferers, we hid 138 pieces (each c. 7 cm long) of ripe bananas (Musa acuminata) at two sites. One site was the forestry plantation where tayras had been observed caching plantains and the other site was the surrounding secondary forest. At each site, three parallel transects were laid, separated by 75 m. On each transect, three banana pieces, indicated by a piece of flagging tape tied to nearby vegetation, were hidden every 20 m. Each of these pieces was placed at one of the vertices of an imaginary, 3 m equilateral triangle centred on the transect. Two pieces, one of them covered with leaf litter, were placed on the ground. The third piece was hidden amongst vegetation at 1–2 m height. After 2 days, we counted the number of banana pieces that were still present.


Tayras were observed retrieving cached, ripe plantains at the forestry plantation (n = 4). On three occasions (one in 2004 and two in 2006), a tayra was observed entering the plantation and heading directly towards a tree. Each of these tayras retrieved a ripe plantain from a bromeliad in the tree, at c. 4 m height. On another occasion in 2006, a pair of tayras was observed at the forestry plantation. One of them entered a small hole in the ground and came out with a very ripe plantain in its mouth. The other tayra appeared to follow the first as it entered the plantation (~ 30 s after) and headed towards the same hole, whilst the first tayra disappeared into the forest with the ripe plantain. The second tayra then emerged without any plantains. No other plantains were found after the hole was inspected. The hole was formed by stems and fronds of felled palms and it measured 23.5 cm in height, 19 cm wide and 1.5 m deep. In addition, peels from ripe plantains, containing both old and fresh teeth marks from tayras, were found on the ground at the forestry plantation on three different occasions.

Direct observations and pictures from camera traps show that, at the M. paradisiaca plantation, ripe plantains were readily consumed in situ by several animals, including tayras, coatis (Nasua narica), opossums (Didelphis marsupialis) and toucans (Pteroglossus torquatus). Unripe plantains were never eaten unless the whole plant fell to the ground, leaving the fruit bunch accessible to peccaries (Tayassu tajacu), which usually consumed the plantains within 1 or 2 days.

Inspection of fruiting plants revealed that tayras removed at least 52 unripe plantains from fruit bunches (Online Resource 1). Direct observations and claw marks left at the stems of the plants indicated that tayras frequently visited plants that had unripe plantains at different stages of development, but they only removed unripe plantains that were at an advanced maturation level (i.e. class IV, Table 1). Pictures from camera traps showed tayras removing both unripe (n = 4; Online Resource 1) and ripe plantains (n = 3) from fruit bunches at the plantation. Ripe plantains removed from the bunch could have been eaten on the ground, as was evidenced from whole peels that were occasionally found on the ground at this site. We did not observe tayras caching ripe plantains.

Caching of unripe plantains was directly observed on two separate occasions. In 2006, a tayra was observed caching an unripe plantain (class IV) in the forest, 10 m away from the plantation. The tayra climbed up a Ficus tree to ~10 m height and came back down without the plantain. In 2004, two tayras were observed together at the forestry plantation. One of them stayed on the ground while the other one climbed a tree carrying an unripe plantain (class III or IV) in its mouth. This tayra attempted to cache the plantain in two different bromeliads at c. 4 m height, in different trees within the forestry plantation, by plunging its head into the bromeliad for a couple of seconds, while still holding the plantain in its mouth. The tayras left the plantation together after c. 5 min, without caching the plantain.

Unripe plantains containing radio transmitters were cached within 150 m of the plantation (n = 3). When recovered, these plantains had already ripened and had been recently consumed. One was found on a grassland floodplain, under tall grass that was flattened against the ground. The other two were cached in the forest, one under a thick cluster of fallen bamboo stems and the other one up a dead tree.

We also found that significantly more of the hidden banana pieces were consumed in the forest than in the forestry plantation (Table 2; χ2 = 13.69, df = 1, p < 0.001). In the forest, fruit was less likely to be consumed if it was hidden above the ground (χ2 = 7.50, df = 1, p < 0.01), but there was no difference between fruits that were left exposed or covered with leaf litter (χ2 = 0.35, df = 1, p = 0.55). At the forestry plantation, overall fruit consumption was low and caching site did not affect the probability of a fruit being consumed (χ2 = 1.51, df = 2, p = 0.47).
Table 2

Number of hidden banana pieces that disappeared after 2 days


Hiding location

Not eaten


Total hidden


Above ground




Hidden on ground




Exposed on ground










Above ground




Hidden on ground




Exposed on ground










Foraging in tayras

Tayras are very versatile foragers that feed on both invertebrate and vertebrate prey, in addition to honey and a variety of fruits (Janzen 1983; Bisbal 1986; Presley 2000; Bezerra et al. 2009). At La Selva tayras were observed feeding on ripe plantains, but never on unripe ones. Digestion of unripe plantains would require modifications of the gastrointestinal tract and the presence of bacterial symbionts, which are adaptations that only occur in mammals with a highly herbivorous diet (Stevens and Hume 1998).

Ripe plantains at La Selva appear to be in high demand as they are taken by a variety of animals as soon as they become available and never go to waste. In contrast, unripe plantains are never eaten unless they fall to the ground, making them accessible to peccaries, which regularly pass through the plantation. Tayras were the only animals observed to pluck unripe plantains and take them away from the plantation. These plantains were cached out of sight and in places that were less frequented by competitors. Although we did not observe tayras caching ripe fruits in this study, it is likely that they do (i.e. some of the scars left at fruit bunches with ripe plantains could represent ripe plantains that were taken for caching).

Between 1945 and 1955, tayras were observed picking unripe, but mature, sapote fruits (Pouteria sapota) from trees at Llano Grande de Río Cuarto, approximately 15 km from La Selva (n > 10; Alvarado-Díaz pers. obs.). At this site, both, plantain and sapote fruits (n > 20) were found in cattle-grazing fields, on top of tree stumps that were densely covered by tall grass. These cached fruits were either unripe and mature or ripe. Like plantains, unripe sapote fruits can also continue their ripening process if detached from the plant, but only if done so at a mature stage of development (Arenas-Ocampo et al. 2003; Offem and Njoku 1993); if harvested when mature, these fruits ripen more rapidly than if left at the plant. Tayras might benefit in securing a food source before strong competition takes place, at the same time that they hasten its availability (i.e. by accelerating their ripening process).

Throughout their range, tayras appear to have higher relative abundances than other carnivores and are able to survive in a diversity of environments, often near human habitations, taking advantage of food found in gardens, cane fields, orchards, cacao and banana plantations (Presley 2000; Guiracocha et al. 2001). In Costa Rica, La Selva appears to be a particularly good place in which to observe these animals (Reid 1997). It is possible that particular aspects of cognition in tayras, such as the ability to sequester food from competitors and their propensity to enter disturbed sites for feeding and caching, have provided tayras with an ecological advantage, enabling them to occupy different habitats, including those affected by human settlements.

Caching and future planning

Most caching species store food that could otherwise be readily consumed, such as seeds, meat, invertebrates, and plant material (Smith and Reichman 1984). Caching in tayras differs markedly, given that they store ‘items’ that cannot be identified as food when guided by sensory information alone (i.e. the fruit looks, tastes and smells different). Tayras cache food that is not edible at the moment of caching and this poses the question if tayras are aware that the fruit will change its condition with time. The only other reported example of caching of food that will only become edible in the future is by the North American pika (Ochotona princeps: Lagomorpha; Dearing 1997). Before winter, these mammals cache plant material that they would normally ignore due to their high levels of secondary compounds, and will only consume these caches once their levels of secondary compounds drop during winter (Dearing 1997).

Caching of unpalatable food by tayras and pikas might represent cases of prospective thinking. However, it is still far from clear what nonhuman animals perceive about the future, and in many cases, even future-oriented, complex behaviours can be achieved by associative learning and fixed-action patterns (Shettleworth 2001; Raby and Clayton 2009). At the moment, the most accepted evidence for prospective thinking in nonhuman animals comes from studies in apes and corvids. Chimpanzees and orangutans take tools for using them in the future, even when the tools are immediately unnecessary or incur a cost (Osvath and Osvath 2008; Osvath 2009). Scrub jays cache particular food items in places where those items are likely to be absent in the near future (Correia et al. 2007; Raby et al. 2007; and critics by Suddendorf and Corballis 2008). Also, when caching food, scrub jays can take into account their previous experience as pilferers to avoid loss of their own caches to pilferage from conspecifics (Emery and Clayton 2001).

Tayras cache unripe but mature stages of sapote and plantain fruits. Sapote is native to this region, whereas plantains were introduced several hundred years ago. Sapote and plantain not only look and taste different, but they also have very different developmental periods (Offem and Njoku 1993; Arenas-Ocampo et al. 2003). Nevertheless, caching behaviour of tayras is finely tuned for both types of fruit; tayras never took fruits that were at early stages of maturation and would fail to ripen due to premature harvesting (e.g. Offem and Njoku 1993). It is possible that tayras posses a concept similar to “unripe, mature fruit”, which could have enabled them to transfer the skill of caching unripe fruits from one species of plant to another one. However, it remains unknown whether the same individuals cache both types of fruits and this is the next crucial step for making any concluding remarks about the cognitive abilities of this species.

Other cognitive mechanisms that do not necessarily imply a perception of the future could also explain why tayras cache unripe fruits. For example, if caching of ripe fruits is a product of fixed-action patterns, mistakes involved in this routine could lead to caching of unripe fruits, and such mistakes could be adaptive. However, the fact that only unripe fruits that were mature enough were harvested suggests that these were not mistakes; otherwise, tayras would have cached all unripe stages indiscriminately. Instead, tayras visited plants with unripe fruits but only took the ones that were mature enough for harvesting. It is possible that highly developed associative-learning abilities in tayras might enable individuals to discover independently (or through social learning) that only unpalatable fruits in certain stages of maturation will ripen if picked from the plants. Simple, learnt associations can be goal-directed, and time-place associations (e.g. learning that a particular food item is present in a certain location at a particular time) can apparently occur without a need of awareness about the future (reviewed in Raby and Clayton 2009).

Resolving if tayras have a concept similar to “unripe, mature fruit” appears crucial in understanding what this species might perceive about the future. In this case, the concept has an intrinsic time component (the fruit will change its condition across time), so if tayras experience this concept, it would suggest that they have some sense of awareness about time. The existence of a concept can be supported when individuals are able to transfer a skill into a “conceptually similar but physically novel problem” (Heyes 1993; Shettleworth 2001). Such transference of a skill (e.g. caching of unripe fruits) might have occurred in tayras after plantains were introduced into the American continent and future research should address if the same individuals are capable of harvesting both types of fruits while ruling out the possibility of independent associative learning with each type of fruit.


CR-USA Foundation and the Organization for Tropical Studies provided funding for this study. We are very grateful to Johanna Hurtado, members of TEAM Project at La Selva, Johel Cháves-Campos and Elizabeth Congdon, for loan of equipment and assistance with logistics. We also thank Orlando Vargas and Ricardo Bedoya for plant identification. William Eberhard, Gilbert Barrantes and Joel Alvarado provided valuable feedback during the course of this study. Previous versions of the manuscript were improved by comments from Ximena Nelson, Martin Whiting and three anonymous reviewers.

Supplementary material

114_2011_821_MOESM1_ESM.pdf (11.3 mb)
ESM 1(PDF 11550 kb)

Copyright information

© Springer-Verlag 2011