ArticlePDF Available

Spatial Learning and Memory in the Tortoise (Geochelone carbonaria)


Abstract and Figures

A single tortoise (Geochelone carbonaria) was trained in an eight-arm radial maze, with the apparatus and general procedures modeled on those used to demonstrate spatial learning in rats. The tortoise learned to perform reliably above chance, preferentially choosing baited arms, rather than returning to arms previously visited on a trial. Test sessions that examined control by olfactory cues revealed that they did not affect performance. No systematic, stereotyped response patterns were evident. In spite of differences in brain structure, the tortoise showed spatial learning abilities comparable to those observed in mammals.
Content may be subject to copyright.
This is a repository copy of Spatial learning and memory in the tortoise (Geochelone
White Rose Research Online URL for this paper:
Wilkinson, A., Chan, H.M. and Hall, G. (2007) Spatial learning and memory in the tortoise
(Geochelone carbonaria). Journal of Comparative Psychology. pp. 412-418. ISSN
Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless
indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by
national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of
the full text version. This is indicated by the licence information on the White Rose Research Online record
for the item.
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing including the URL of the record and the reason for the withdrawal request.
promoting access to White Rose research papers
White Rose Research Online
Universities of Leeds, Sheffield and York
This is an author produced version of an article published in Journal of
Comparative Psychology. This article may not exactly replicate the final version
published in the APA journal. It is not the copy of record.
White Rose Research Online URL for this paper:
Published paper
Wilkinson A, Chan HM, and Hall G. (2007) Spatial learning and memory in the
tortoise (Geochelone carbonaria) Journal of Comparative Psychology 121 (4):
Tortoise spatial learning 1
Spatial Learning and Memory in the Tortoise (Geochelone carbonaria)
Anna Wilkinson, Hui-Minn Chan, Geoffrey Hall
University of York
Short title: Tortoise spatial learning
Contact address:
Anna Wilkinson
Department of Psychology
University of York
YO10 5DD
Tortoise spatial learning 2
A single individual of the species Geochelone carbonaria was trained in an
eight-arm radial maze, with the apparatus and general procedures modelled
on those used to demonstrate spatial learning in rats. The tortoise learned to
perform reliably above chance, preferentially choosing baited arms, rather
than returning to arms previously visited on a trial. Test sessions that
examined control by olfactory cues revealed that these did not affect
performance. No systematic, stereotyped response patterns were evident. In
spite of differences in brain structure, the tortoise showed spatial learning
abilities comparable to those observed in mammals.
Key words: tortoise, Geochelone carbonaria, spatial learning, radial maze
Tortoise spatial learning 3
Spatial Learning and Memory in the Tortoise (Geochelone carbonaria)
Nonavian reptiles, birds, and mammals all evolved from a common
amniotic ancestor and it is therefore possible that these classes share
common behavioral traits and capabilities. Equally, since the putative
common ancestor lived as long as 280 million years ago, there is ample time
for evolutionary paths to have diverged and for quite different capacities and
mechanisms to have evolved in the different classes. Certainly, brain
structures appear to differ in important respects Ð for example the forebrain of
the reptile, with its thin cortical layer is very different from the multilayered
structure seen in mammals.
The study of spatial learning in chelonia (turtles, terrapins, and
tortoises) has a long history (for a review see Burghardt, 1977). It started early
with Yerkes (1901), who demonstrated that the speckled turtle (Clemmys
guttata) could learn a multiunit maze Òwith surprising quicknessÓ (quoted by
Macphail, 1982), a result confirmed for the common wood turtle (Clemmys
insculpta) by Tinklepaugh (1932). Acquisition and reversal of a T-maze task
by the terrapin Chrysemys picta picta was demonstrated by Kirk and
Bitterman (1963), and the ability of this species to show serial reversal
improvement in a (slightly different) spatial task was confirmed by Holmes and
Bitterman (1966). What these various studies do not reveal is whether
chelonians are capable of forms of spatial learning shown by mammals. In
mammals, some forms of spatial learning are thought to be dependent on the
hippocampus (a structure that reptiles lack). It is possible, then, that the
chelonians learned the mazes using a system (e.g., by learning to make a
Tortoise spatial learning 4
given turn or sequence of turns) different from some more advanced,
hippocampally dependent, navigational system used by mammals.
This issue has been addressed directly in a series of experiments by
L—pez and his colleagues. L—pez et al. (2000), working with the terrapin
Pseudemys scripta, showed that this species could learn, in a T-maze, to
approach a given location in space regardless of which of the other two arms
they started from. The animals maintained this performance even when the
entire maze was rotated, so that the starting point was some quite novel
location. The ability appeared to depend on navigation by means of
extramaze (room) cues, in that it was disrupted by the introduction of shielding
curtains around the maze. L—pez et al. suggested that the turtles were using a
Òcognitive mapÓ of the sort postulated for mammals. They went on to show
that lesions of a forebrain area, the medial cortex, taken on anatomical
grounds to be a parallel of the mammalian hippocampus, disrupted
performance on these tasks (L—pez, Vargas, Gomez, & Salas, 2003).
The results just described encourage the view that chelonians (with an
intact medial cortex) should be capable of coping successfully with other tasks
that have been used to demonstrate the spatial learning abilities of mammals.
To this end we have studied the performance of a red-footed tortoise
(Geochelone carbonaria) in an eight-arm radial maze (Figure 1).
This species is a land-dwelling chelonian, unlike the semi-aquatic
terrapins that were tested in the experiments just described. Previous work
with the desert tortoise (Gopherus agassizii) by Fink (1954, cited by Burghardt
1977) has shown that the performance of this species on a spatial reversal
task is comparable to that of terrapins. However, its behavioural ecology is
Tortoise spatial learning 5
different from that of the red-footed tortoise; the latter species eats fallen fruit,
and flowers whereas the desert tortoise is largely a grass grazer. It is possible
that the differences in their foraging strategy may have more influence on their
performance than evolutionary proximity. The red-footed tortoise is a relatively
active species, and is capable of travelling up to 85 meters an hour
(Moskovits, 1985, cited by Strong & Fragoso 1987). This liveliness, in addition
to their foraging behavior makes this species an ideal subject for our study.
The radial arm maze was pioneered for use with rats by Olton and
Samuelson (1976) and consists of a central area from which eight arms
radiate. Food is available at the end of each arm. A well trained rat will visit
each arm to collect the food, and rarely return to arms that it has previously
visited, exhibiting an ability to discriminate among the various spatial
locations, and remember which places have been visited on a given trial. The
procedure provides an excellent test of an animalÕs spatial learning capacities
and its working memory. It can readily be adapted for use with many species
and provides a useful tool for making direct comparisons across species.
In the present experiment we examined the performance of a tortoise
in the radial maze asking, first, whether this animal could achieve efficient
performance. To anticipate, we found that he could. We then went on to
investigate the nature of the mechanisms responsible for its performance by
carrying out a series of tests designed to exclude the contribution of non-
spatial factors. We hoped to reveal the extent to which the tortoiseÕs behavior
is comparable to that of a mammal.
Tortoise spatial learning 6
The subject (named Moses) was a male captive-bred red-footed
tortoise (Geochelone carbonaria). He was approximately 2 years old and his
plastron (the base of his shell) measured 7.5 x 6 cm at the start of the
experiment. He was experimentally na•ve. During the study Moses lived in a
tank in an office adjacent to the experimental room. The office was kept on a
daily 12L:12D cycle (light on: 0800-2000). The tank measured 61 x 30 x 30
cm and was maintained at 29oC (+/- 4oC); humidity within the tank was
maintained at 50%. The tortoise was given access to food (fruit and
vegetables) for 60 min each day, approximately 30 min after that dayÕs
experimental procedures had been completed.
The apparatus was an eight-arm radial maze made of opaque black
Perspex (see Figure 1). Each arm was 10 cm wide, 20 cm long, and the sides
and one end had walls 7 cm high. The arms radiated from a hexagonal central
platform, 24 cm across. Removable guillotine doors could be placed at the
junction between the arm and central platform. During the training and testing
phases a white plastic food cup, 3 cm in diameter and 1.5 cm high was placed
in a central position at the end of each arm. The maze was positioned on a
table in a small experimental room, that was lit by two 60-W ceiling lights and
maintained at approximately 27-29!C. External cues that were, in principle,
visible from within the maze, included shelving on which laboratory equipment
was stored (adjacent to arm 7 of the maze disposition shown on the left of
Figure 1), and a poster on the opposite wall (above arm 3). The wall adjacent
Tortoise spatial learning 7
to arm 1 contained a door, to the left of which (adjacent to arm 8), the
experimenter sat. The experimenter remained in the room for the entire
session. Two experimenters were involved in conducting experimental
sessions. Experimenter 1 observed the tortoise from the beginning of the
experiment up until midway through the training phase ÒAssessing the
influence of odor trailsÓ (see below). The second experimenter completed the
The experiment took approximately 5 months and was conducted from
25th January 2006 - 17th June 2006. Procedures took place in the afternoon,
as this was the time when Moses was most active. He was removed from his
tank and handled for approximately 5 min prior to experimentation. During this
time he was allowed to walk around the office space or the experimenterÕs
lap. This served to increase his activity level. He was then placed in a holding
cage and taken to the experimental room. On each day he received several
trials (detailed below), each separated from the next by an intertrial interval
(ITI), usually of 5 min, spent in the holding cage. The maze was wiped clean
at the end of each day but not between trials.
Familiarization to the maze. Extensive pretraining was needed in order
to ensure that the tortoise would locomote around, and eat readily in, the
maze. The procedures, which involved trial-and-error learning, as much on
the part of the experimenters as on the part of the tortoise, are detailed in
Table 1. By the end of this phase of pretraining the subject would, on most
occasions, visit all 8 arms within a 30-min trial, to obtain the reward (a small
piece of strawberry) that was visible at the end of each arm.
Tortoise spatial learning 8
Basic radial arm maze training. There were 12 daily sessions in the first
phase of training, each consisting of 4 trials. At the start of each trial Moses
was placed on the central platform, facing an arm selected at random. Each
arm of the maze contained a food cup baited with a piece of strawberry.
Choice of an arm was recorded whenever half of Moses had advanced into an
arm, so that half his shell was within the arm. This measure was used as
Moses rarely backed out of the maze once he had entered this far and we felt
that this measure was suitably conservative as to ensure all errors were
included in analysis. The trial ended when all eight rewards had been
consumed, or after 30 min. A record was kept of which arms were entered
and in what order.
Assessing the influence of food odor. This 2-day phase introduced test
trials designed to assess whether or not Moses was using odor cues from
food in the food cups to guide his behavior. Sessions were organised as
before, except that on trials 2 and 4, only four of the arms were baited. On trial
2 of the first test day and trial 4 of the second, these were arms 1, 3, 5, and 7;
on trial 4 of the first test day and trial 2 of the second they were arms 2, 4, 6,
and 8. If performance is guided by odor cues, we might expect Moses to show
a preference for the baited arms on these test trials.
Assessing the influence of odor trails: Training. During this phase,
which lasted 9 days, each trial was divided into two parts. In the first, four
arms (equally often the even- or odd-numbered arms) were blocked by the
guillotine doors, and Moses was allowed to take food from the food cups of all
four of the available arms. (A maximum of 30 min was allowed for this part of
the trial.) He was then removed from the maze and placed in the holding cage
Tortoise spatial learning 9
for 30 s. During this time the doors were removed, and the other four arms
were baited. Moses was then replaced in the maze and allowed to enter any
arm (although he only received reward when he entered an arm not visited
during the first part of the trial). The trial was terminated when all four rewards
had been eaten or after 60 min. The procedure of removing the tortoise and
then replacing him proved to be somewhat disruptive and on occasion Moses
failed to take all rewards in the time allowed. Over the course of the 9 days
Moses successfully obtained all the rewards on 17 of the 36 trials.
This procedure was introduced principally to provide a baseline against
which the effects of the test procedure, to be described next, could be
assessed. But it also allows the possibility of testing the animalÕs memory Ð
efficient performance in the second part of the test requires that information
acquired in the first part should survive the retention interval and the
disturbance it involved.
Assessing the influence of odor trails: Test. This test lasted 9 days and
consisted of four retraining days interspersed among which were 5 test days.
On retraining days the procedure was identical to that described above for the
odor trail training phase. Test trials were similar except that, when Moses was
removed from the maze having visited the four available arms, the maze was
rotated through 45 degrees (clockwise on half the trials, anticlockwise on the
rest) with the result that an arm not previously visited was now in the same
spatial location as one visited in the first part of the trial (see Figure 1). Food
was made available only in arms in spatial locations that had not previously
been visited (i.e., in order to perform efficiently Moses needed to return to
arms that he had traversed in the first part of the trial). This procedure allows
Tortoise spatial learning 10
us to test whether the tortoise had learned a strategy of avoiding arms that he
had previously visited, and had perhaps marked by means of some sort of
odor. (Such a strategy would result in poor performance in the rotated maze.)
This procedure also constitutes a test of the extent to which the animalÕs
behavior is controlled by extra-maze cues. We take up these matters in the
final Discussion.
Basic radial arm maze behavior. In spite of the extensive
familiarization, the tortoiseÕs movement around the maze was often slow; on
15 of the 48 trials of the first training phase, the time limit was reached and
testing was terminated before the animal had made eight choices. Our
analysis will be confined to the remaining 33 trials, on each of which at least
eight choices were made. According to Olton (1978), the number of correct
responses in the first eight choices, to be expected on the basis of chance
performance, is 5.3. (A correct response is entering an arm that had not been
entered previously; chance performance is computed assuming that every
choice is made at random, without replacement). The mean number of correct
responses in the first eight choices of the 33 trials available for analysis was
5.88 (SEM = 0.16; range 4-8). A one-sample t test comparing this score
against chance expectation revealed a significant effect, t(32) = 3.59, p < .01.
To obtain a more detailed picture of MosesÕ performance we focused
our analysis on those trials on which he successfully visited all eight arms.
There were 18 such trials; the first occurred on day 2 of training and there was
at least one such trial on all succeeding days. Table 2 presents a full list of all
the choices made on these 18 trials. For each trial we calculated the
Tortoise spatial learning 11
probability that the task would be completed in the number of choices actually
made, on the assumption that choices were made at random and with
replacement. This probability is given in the far right column of the table. This
shows that, although accuracy of performance fluctuated substantially from
trial to trial, it was consistently at a level unlikely to be achieved on the basis
of chance. Particularly, good performance was as likely to be seen on early
trials as on later trials; that is, there was no indication of a gradual acquisition
It is possible that Moses adopted stereotyped response patterns (e.g.,
a pattern of always turning into the next arm on the left would ensure perfect
performance). To examine this we scored each of the responses he made
after the first choice on each of the trials detailed in Table 3. There were 203
of these. Table 3 breaks these down into choices of arms that were 1, 2, or 3,
positions, either clockwise or anticlockwise, from the arm just left, and those
that were choices of the arm directly opposite. A strategy of a sort is
immediately apparent as on none of the trials did Moses reenter the arm that
he had just exited. Random choice among the remaining 7 possible turns
would result in 29 choices of each of these possibilities. Table 3 reveals a
tendency for choices of arms two positions away from that being exited to be
overrepresented, at the expense of choices of arms 3 positions away. A one-
sample chi-squared test on the scores presented in the table showed the
deviation from chance expectation to be significant, chi-squared = 18.96, df =
6, p > .01
Assessing the influence of food odor. This test was designed to assess
if MosesÕ performance was based on odor cues. If it were, we might expect
Tortoise spatial learning 12
him to choose preferentially those arms that were baited on the four test trials
of this phase on which four of the arms were left unbaited. Performance on
these test trials turned out to be very similar to that shown on the four
standard trials with which they were intermixed. Scoring a correct response as
choice of an arm not previously visited (whether it contained food or not)
showed that he made a mean of 6.25 (95% CI = +/-.49) correct responses in
the first eight choices on the standard trials, and a mean of 5.75 (95% CI = +/-
.49) on the test trials. Critically, correct choices on test trials were as likely to
be made by entering unbaited as by entering baited arms; of the total of 23
correct responses under consideration, 10 were to unbaited arms and 13 to
baited arms, chi-squared = 0.39, df = 1, p >.50.
Assessing the influence of odor trails: Training. In this phase of
training, the guillotine doors forced Moses to enter four of the arms before a
30-s interval; after this all eight arms were made available. Moses performed
rather poorly during the forced-choice trials of this procedure and on several
occasions failed to visit the four arms available during the 30 min allowed for
the first part of the trial. This seemed to be caused by the introduction of the
barriers which he spent a large amount time trying to push, a pattern of
behavior that became more pronounced as training proceeded. This behavior
is commonly observed in tortoises. If barriers either have visible gaps or move
when pushed, tortoises spend a great deal of time trying to get through.
These trials were abandoned and were excluded from the analysis. We
analyzed the 17 trials in this phase on which Moses succeeded in visiting all
eight arms. On these trials the first four baits were collected efficiently (the
mean number of choices required was 5.88). When returned to the maze after
Tortoise spatial learning 13
the interval he took a mean of 9.59 (range 5Ð15; see Table 4) choices to
collect the remaining four baits. For comparison we looked at the number of
trials taken to obtain the final four baits on the 18 trials of phase-1 training on
which this was achieved (see above). In this latter case the mean number of
trials required was 7.78. These scores differed significantly, t(33) = 2.07, p <
Evidently performance was rather poor on these trials, but despite
performance being disrupted, it did not decline to a level that might be
expected on the basis of random choice. Table 4 shows the number of
choices required to visit the remaining four arms on each trial of this stage
and, for each such trial, the probability that the task would be completed in the
number of choices actually made, on the assumption that choices were made
at random and with replacement.
Assessing the influence of odor trails: Testing. In this final test the
maze was rotated after the retention interval so that arms that had previously
been visited were now in spatial locations that had not previously been visited.
The score is the number of trials taken, after the retention interval, to visit the
four unvisited spatial locations. The scores were 11, 9, 7, and 15 trials, with a
mean of 10.50. This is not markedly worse than that (9.59) reported for the
second training phase. Had his performance in that phase been based on the
avoidance of the odor of a previously visited arm we would have expected a
total disruption in performance.
Basic radial arm maze behavior. In spite of the extensive
familiarization, the tortoiseÕs movement around the maze was often slow.
Tortoise spatial learning 14
Accuracy of performance fluctuated substantially from trial to trial; it was,
however, consistently at a level unlikely to be achieved on the basis of
chance. There was no indication of a gradual acquisition process. This is
perhaps not surprising. Although the food was visible during the extensive
familiarization phase, the general procedures used in that phase matched
those used in the basic radial arm maze training. The results of the
familiarization phase revealed a sharp learning curve. It is possible then that,
during pretraining, Moses acquired strategies that he could then use in the
training phase. This would allow the immediate above-chance performance
that we observed, even when the food was hidden from view by the food
The rest of the study was intended to elucidate the nature of the
strategies involved in MosesÕ performance. One possibility was that Moses
adopted stereotyped response patterns (e.g., a pattern of always turning into
the next arm on the left would ensure perfect performance). As we have
noted, our analysis showed that on none of the trials did Moses reenter the
arm that he had just exited. The analysis also revealed a tendency for choice
of arms two positions away from that being exited; rats show a similar pattern
(Olton, Collison, & Wertz, 1977). The factors controlling this behavior in rats
were investigated by Yoerg & Kamil (1982), who manipulated the size of the
central platform of a radial arm maze. They found that this had no effect on
the accuracy of performance, but the use of adjacent arms significantly
increased with a larger platform. They suggested that this could be due to the
increased cost of choosing a nonadjacent arm. However, it is possible (as
they acknowledged) that the sharp angles of adjacent arms in a small maze
Tortoise spatial learning 15
make it hard to negotiate and make it easier to choose a non-adjacent arm in
such a maze. Both of these hypotheses could account for our tortoiseÕs arm
choice behavior. No other simple response patterns were discerned.
Assessing the influence of food odor. If MosesÕ performance was
based on odor cues from the food we might expect him to preferentially
choose those arms that were baited on the test trials over those that were left
unbaited. Performance on these test trials turned out to be similar to that
shown on the four standard trials with which they were intermixed. There is no
evidence, therefore, of control by food odor.
Assessing the influence of odor trails: Training. This phase of training
was conducted in preparation for testing whether Moses learned to avoid his
own odor trails. It also allowed examination of the extent to which information
acquired in the first part of the trial survived the interval (and the disruption
consequent on removal from and return to the maze). Performance on this
part of the task was compared with that of the last four arms of the basic
radial arm maze training. There was some disruption following the retention
interval, however it did not decline to a level that might be expected on the
basis of random choice. This suggests that performance in the second part of
the trial was controlled, to some extent, by memory of the first part of the trial.
Assessing the influence of odor trails: Testing. In this final test the
maze was rotated after the retention interval so that arms that had previously
been visited were now in spatial locations that had not previously been visited.
MosesÕ performance was not markedly worse than that reported for the
training phase. Had his performance in this phase been based on the
avoidance of the odor of a previously visited arm, rotation of the maze (which
Tortoise spatial learning 16
required the animal to return to a location previously visited) would have
produced a total disruption. We tentatively conclude, therefore, that his
performance is based, at least in part, on information about the spatial
location of the maze arms.
Conclusions. The study of a single individual cannot tell us what is
generally true of some larger grouping (such as reptiles, or chelonians, or
members of the species Geochelone carbonaria). It does, however, set some
limits on assertions about what that group is or is not capable of. Our study
allows the conclusion that a tortoise is capable of showing fairly efficient
performance in a radial maze. Its performance is less efficient than that of rats
(see Olton & Samuelson, 1976) (for whatever reason Ð this may reflect an
inadequacy in our procedure rather than a lack of capacity in the animal), but
it is, none the less, above the level to be expected on the basis of chance. As
is true for rats, the performance of the tortoise does not appear to depend on
the acquisition of stereotyped response strategies; nor is it controlled by odor
cues or the following (or avoidance) of odor trails. As for the rats, the evidence
points to an ability to learn about spatial locations, to remember which have
been visited, and to adopt a strategy of going to those that have not been
visited previously (or of avoiding those that have). Exactly what cues control
this ability remains to be determined. It is tempting to suppose that the tortoise
identifies spatial locations by the configuration of extramaze cues that define
them. Direct support for this proposal requires studies in which the
relationship of the maze arms to the extramaze cue is explicitly manipulated.
We can further conclude that hippocampal formation of the mammalian
brain is not essential for adequate performance on this sort of spatial task.
Tortoise spatial learning 17
This may mean that some quite different brain structure is capable of carrying
out the same functions, but perhaps by way of quite a different mechanism.
Alternatively it may be taken to support the view that the reptilian medial
cortex is functionally equivalent, even analogous, to the mammalian
hippocampus. In the latter case, further studies could reveal the operation of
similar mechanisms in reptiles and mammals.
In summarising his study, Tinklepaugh (1932) wrote as follows: ÒThis
report on the maze running of a single turtle is made not because this lowly
subject learned the maze, but rather because of the nature of its behavior
during the processÉÓ (p. 201). The same holds for our report. Tinklepaugh
went on to say: ÒIn my estimation, the learning of the turtle equalled the
expected accomplishment of a rat in the same maze ÉÓ (p.206). We would
not want to make the same claim for our own subject; we have already noted
ways in which his performance fell short of what might be expected of a rat
trained in the same maze. But we would want to say that his performance was
not fundamentally different from that of the rat Ð that any difference appears to
be quantitative rather than qualitative. His movements around the maze may
have been slow, but satisfactory learning was ultimately achieved. To that
extent we can endorse the conclusion reached by Tinklepaugh, that ÒÉthe
physical sluggishness and awkwardness of the turtle may have earned him an
undeserved reputation for stupidityÓ (p. 206).
Tortoise spatial learning 18
Burghardt, G.M. (1977). Learning processes in reptiles. In C. Gans & D.W.
Tinkle (Eds.), Biology of the Reptilia (pp. 555-681). London: Academic
Holmes, P.A., & Bitterman, M.E. (1966). Spatial and visual habit reversal in
the turtle. Journal of Comparative and Physiological Psychology, 62, 328-
L—pez, J.C., Rodr’guez, F., G—mez, Y, Vargas, J.P., Broglio, C., & Salas, C.
(2000). Place and cue learning in turtles. Animal Learning & Behavior, 28,
L—pez, J.C., Vargas, J.P., G—mez, Y., & Salas, C. (2003). Spatial and non-
spatial learning in turtles: The role of the medial cortex. Behavioral Brain
Research, 143, 109-120.
Macphail, E. M. (1982). Brain and intelligence in vertebrates. Oxford:
Clarendon Press.
Olton, D.S. (1978). Characteristics of spatial memory. In S.H. Hulse, H.
Fowler, & W.K. Honig (Eds.), Cognitive processes in animal behavior (pp.
341-373). Hillsdale, NJ: Erlbaum
Olton, D.S., Collison, C., & Wertz, M. (1977). Spatial memory and radial arm
maze performance of rats. Learning and Motivation, 8, 289-314.
Olton, D.S, & Samuelson, R.J. (1976). Remembrance of places passed:
Spatial memory in rats. Journal of Experimental Psychology: Animal
Behavior Processes, 2, 97-116.
Tortoise spatial learning 19
Strong, J.N., & Fragoso, J.M.V. (2006). Seed dispersal by Geochelone
carbonaria and Geochelone denticulate in Northwestern Brazil. Biotropica,
38, 683-686.
Tinklepaugh, O.L. (1932). Maze learning of a turtle. Journal of Comparative
Psychology, 13, 201-206.
Yoerg, S.I., & Kamil, A.C. (1982). Response strategies in the radial arm maze:
Running around in circles. Animal Learning & Behavior, 10, 530-534.
Tortoise spatial learning 20
Author Note
Anna Wilkinson, Hui-Minn Chan, and Geoffrey Hall, Department of
Psychology, University of York, York, United Kingdom. We thank J. Thijssen,
R. Gillett, K. Kirkpatrick and S. Franklin for statistical advice and assistance,
and Tiffany Galtress for her help in caring for the tortoise.
Correspondence concerning this article should be addressed to Anna
Wilkinson, Department of Psychology, University of York, York, YO10 5DD,
UK. E-mail:
Tortoise spatial learning 21
Table 1
Overview of the experimental procedure
Total no.
Total no.
1 arm open. Tortoise placed in the center;
needs to enter the arm to get food.*
5 min or
No arms open. Dandelion and strawberry
placed in central platform.
30 min or all
food eaten
1 arm open. Tortoise placed in the center;
needs to enter the arm to get food.
15 min or
All arms open. Food visibly available at the
end of each arm.
30 min or
All arms open. Food available, but hidden in
cups at the end of each arm. Tortoise
allowed to accustom himself to eating from
food cups.
30 min or
Basic RAM
All arms open. Food available, but hidden in
cups at the end of each arm.
30 min or
Food Odor Test
All arms open. Food available in four of the
arms. Intermixed with normal training trials.
30 min or
4 test
4 retraining
Odor Trails
a: 4 arms open, other 4 arms blocked.
b: 30-s retention interval.
c: All arms open, food only available in the
arms not previously visited.
a: 30 min or
c: 60 min or
Odor Trails Test
a: 4 arms open, other 4 arms blocked.
b: 30-s retention interval during which the
maze is rotated by 45o.
c: All arms open; food only available in
unvisited arms spatially.
a:30 min or
c: 60 min of
5 test
4 retraining
Note. *After four days of training on this procedure, Moses was not eating
readily. Phase 1a was include to encourage him to eat in the maze; RAM:
Radial arm maze; ITI: Intertrial interval.
Tortoise spatial learning 22
Table 2
Sequence of choices in the 18 trials of the first phase of training on which all
arms were visited.
No. choices to
[2,5,4,7,1,2,1,8] 7,6,5,1,3
[3,4,8,7,5,6,5,1] 7,5,8,2
[1,6,8,6,8,6,8,7] 5,6,8,2,4,3
[2,6,4,3,1,7,5,7] 8
[2,4,5,8,3,7,1,2] 8,6
[5,8,2,6,1,2,1,7] 4,5,8,7,6,3
[4,7,1,5,7,1,2,8] 6,3
[5,1,8,2,8,6,7,3] 4
[4,8,4,7,1,8,6,8] 2,1,8,5,6,1,3
[7,8,6,7,1,2,8,3] 8,6,5,7,3,8,4
[6,4,8,6,1,7,6,5] 1,2,6,8,6,3
[3,1,3,8,2,6,7,1] 8,4,3,5
[2,7,3,8,2,6,4,2] 4,5,3,8,1
[4,8,3,5,7,6,4,2] 8,2,1
[3,8,4,5,1,5,6,8] 2,5,6,8,7
[4,7,6,7,8,2,1,3] 8,1,3,5
Tortoise spatial learning 23
Note. The numbers in the Choices column refer to the arms of the radial
maze. The first eight choices made are enclosed in square brackets. Choices
in bold indicate errors (returning to an arm already visited). The probability
given for each trial is that associated with the number of choices to completion
assuming that every choice is made at random, with replacement of choices
already made.
Tortoise spatial learning 24
Table 3
Classification of type of turn for choices made in the 18 trials of the first phase
of training on which all arms were visited.
Type of turn
Percentage of total
1 arm anticlockwise
2 arms anticlockwise
3 arms anticlockwise
1 arm clockwise
2 arms clockwise
3 arms clockwise
Tortoise spatial learning 25
Table 4
Number of choices required to complete the task for the 17 trials of the
second phase of training on which this was achieved.
No. choices to
Tortoise spatial learning 26
Note. In this phase of training the subject had received forced trials with four
of the maze arms; choices to completion refers to the number of choices
required to visit the remaining four arms when all eight were made available.
The probability given for each trial is that associated with the number of
choices to completion assuming that every choice is made at random, with
replacement of choices already made.
Tortoise spatial learning 27
Figure Caption
Figure 1. Layout of the maze in the two phases of a trial when testing the
influence of odor paths. Before the turn, guillotine doors blocked access to
four of the arms, allowing access to food only in the other four (those
numbered 1, 3, 5, and 7 in this example). After rotation of the maze the doors
were removed and food was available again only in arms 1, 3, 5, and 7. The
tortoise was therefore required to enter same arms as had been visited
before, these now being in different spatial locations.
Tortoise spatial learning 28
Figure 1
Before the turn
After the turn
... The boards had a hole cut into them to let the light into the maze. trials terminated, we drained the maze and cleaned it with 75% ethanol (e.g., Wilkinson et al., 2007). ...
... In order to train turtles to associate the target arm with food, turtles in Experiment 2 received pieces of thawed fish, circa 1 cm 3 in volume, as reinforcement at the end of the target arm only (e.g. Grosse et al., 2010; see also Kramer, 1989;Avigan & Powers, 1995;López et al., 2000;Wilkinson et al., 2007;Young et al., 2012). To avoid interference by the experimenter, we placed the food reward in the apparatus before the turtle was placed in it. ...
Many species consider both prior experiences and the context of current stimuli when making behavioural decisions. Herein, we explore the influence of prior experience and novel incoming stimuli on the decision-making in the Eastern painted turtle ( Chrysemys picta ). We used a free-choice Y-maze to assess the preferences of turtles wavelength and intensity of light. We then trained naïve turtles to associate one arm of a maze with a food reward, and then tested the relevance of light colour and intensity on the turtles’ decision-making regarding arm choice. Turtles avoided bright light, even when presented on the side of the maze with which they had learned to associate a food. When light intensities of both sides were the same — irrespective of intensity — turtles chose the side they had previously learned to associate with the food reward. C. picta in our study showed a weak attraction to blue light and a strong avoidance of yellow light, a response generally consistent with previous work in sea turtles. Future studies should examine the ecological and evolutionary relevance of these decisions in field-oriented tests.
... Maple and Perdue (2013; p 108) described cognitive enrichment as: "challenging and stimulating an organism's memory, decision-making, judgment, perception, attention, problem-solving, executive functioning, learning and species-specific abilities." A training routine using associative learning (Lopez et al 2001;Wilkinson et al 2007Wilkinson et al , 2009 would provide this type of enrichment and has been proven possible in marine turtles (Mellgren & Mann 1998;Bartol et al 2003). However, since rehabilitation turtles only remain in facilities temporarily, training may not be a worthwhile form of EE due to the potential time investment required for it to be successful. ...
For animals undergoing rehabilitation it is vital to monitor welfare in a way that is feasible, practical, and limits stress to the animal. The industry gold standard is to assess welfare under the Five Domains model, including nutrition, environment, physical health, and behaviour as the first four physical domains and mental domain as the fifth. Feasibility and effectiveness of these domains for assessing welfare of sea turtles undergoing rehabilitation were reviewed and it was determined that the mental state can be best assessed through behavioural changes. A scoping review of the literature was conducted using Scopus and Web of Science to investigate use of environmental enrichment devices (EEDs) as a measure of welfare in sea turtles. Behavioural assessments using EEDs were found to be well-documented; however, most EED studies pertained largely to livestock or zoo animals. Furthermore, studies rarely concentrated on reptiles, and specifically sea turtles. Results also showed that certain welfare assessment methods may be less appropriate for short-term captivity experienced during rehabilitation. Additionally, the hospital environment limits the ability to address some of the domains (ie biosecurity, feasibility, safety of turtle, etc, might be compromised). This review shows that only three of the nine environmental enrichment strategies described in the literature suit the specific requirements of sea turtles in rehabilitation: feeding, tactile, and structural. It is documented that turtles display behaviours that would benefit from EEDs and, therefore, more specific studies are needed to ensure the best welfare outcomes for sea turtles undergoing rehabilitation.
... Interestingly, a substantial minority of reptile owners believed that their reptile exhibited loyalty, knowing when the owner felt bad, and understanding their owner. This, together with growing evidence that reptiles are capable of complex cognitive abilities including spatial, physical, and social learning, and even discriminate between familiar and unfamiliar humans (Burghardt, 2013;Kis et al., 2015;Mueller-Paul et al., 2014;Nagabaskaran et al., submitted;Wilkinson et al., 2007Wilkinson et al., , 2010aWilkinson & Huber, 2012), suggests that more research is needed to refine and elucidate the human-reptile relationship and its variations among different species. In light of these findings, it should not be assumed that poor welfare, experienced by some reptiles, is primarily due to a lack of attachment; rather, it is more likely the result of a lack of knowledge. ...
A growing number of reptiles are being kept as companion animals in private households, and this has resulted in a concomitant rise in welfare concerns. Poor welfare has been linked to a lack of the emotional attachment in other groups. The aim of this study was to investigate owner attachment to pet reptiles, grouped as lizards, snakes, and tortoises. The Lexington Attachment to Pets Scale (LAPS) questionnaire was administered online to a self-selected sample of 2,992 adult reptile owners, including 1,281 owners of both reptiles and dogs. The latter responded first for reptiles and then for dogs. Findings revealed evidence of an emotional attachment toward reptiles that was within the range of those observed in traditional pets. In addition, owners had a stronger attachment to lizards than snakes or tortoises and were also more attached to snakes than tortoises. Owners with both dogs and reptiles showed significantly higher attachment scores to their dogs than to reptiles. However, attachment scores for dogs were higher than those found in other studies. This work reveals strong evidence for a human–reptile attachment and suggests that instances of poor welfare may not result from a lack of attachment but rather a lack of knowledge.
... In the context of map-based navigation, diverse and early diverging vertebrates including fish (Burt de Perera et al., 2016), reptiles (Wilkinson and Huber, 2012;Broglio et al., 2015), turtles (López et al., 2001), amphibians , and tortoises (Wilkinson et al., 2007) all show the ability to build spatial maps of their environment and flexibly generate novel navigational routes to known places (Rodríguez et al., 2002). Model organisms for early bilaterians, such as flatworms seem to navigate only with taxis and perhaps response-based learning (Pearl, 1903;Luersen et al., 2014;Larsch et al., 2015;Gourgou et al., 2021) and show no ability to remember specific un-cued locations. ...
Full-text available
Retracing the evolutionary steps by which human brains evolved can offer insights into the underlying mechanisms of human brain function as well as the phylogenetic origin of various features of human behavior. To this end, this article presents a model for interpreting the physical and behavioral modifications throughout major milestones in human brain evolution. This model introduces the concept of a “breakthrough” as a useful tool for interpreting suites of brain modifications and the various adaptive behaviors these modifications enabled. This offers a unique view into the ordered steps by which human brains evolved and suggests several unique hypotheses on the mechanisms of human brain function.
... Many diverse vertebrates, including those that diverged early such as fish (Burt de Perera et al., 2016), reptiles (Wilkinson and Huber, 2012;Broglio et al., 2015), turtles (López et al., 2001), amphibians (Phillips et al., 1995), and tortoises (Wilkinson et al., 2007) show incredibly sophisticated mapping abilities (Rodríguez et al., 2002a) -capable of learning un-cued locations and capable of flexibly generating new navigation routes. For example, fish can remember specific locations in 3-dimensional space (Karnik and Gerlai, 2012;Lucon-Xiccato and Bisazza, 2017;Wallach et al., 2018) and can generate a correct novel path to specific goal locations from many different starting places (Brown, 2015). ...
Full-text available
This paper presents 13 hypotheses regarding the specific behavioral abilities that emerged at key milestones during the 600-million-year phylogenetic history from early bilaterians to extant humans. The behavioral, intellectual, and cognitive faculties of humans are complex and varied: we have abilities as diverse as map-based navigation, theory of mind, counterfactual learning, episodic memory, and language. But these faculties, which emerge from the complex human brain, are likely to have evolved from simpler prototypes in the simpler brains of our ancestors. Understanding the order in which behavioral abilities evolved can shed light on how and why our brains evolved. To propose these hypotheses, I review the available data from comparative psychology and evolutionary neuroscience.
... The first evidence of use of a cognitive map (the same system used by mammals and birds) from reptiles comes from pound turtles (Pseudemys scripta), that failed in in approaching a goal when the extramaze cues were removed (Lopez et al. 2003). Wilkinson et al. (2007) tried to investigate tortoises' spatial learning abilities coupled with working memory using an eight-arm radial maze. The experiment was set-up in order to test tortoises' ability to both discriminate different locations of food (placed at the end of the eight arms of the maze) and to remember which places they had already visited. ...
... A red-footed tortoise (Chelonoidis carbonaria) adjusted its navigational strategy in a radial-arm maze (a common maze used for testing spatial memory) according to the visual cues available in the surrounding environment. In a cue-rich environment, the tortoises used visual cues to find their way around the maze, avoiding already visited, food-depleted arms (Wilkinson, Chan, & Hall, 2007). By contrast, in an environment with little visual structure (when the maze was surrounded by a curtain), the animal used a response-based strategy by entering the arm adjacent to the last-exited arm. ...
Recently, there has been a surge in cognition research using non-avian reptile systems. As a diverse group of animals, non-avian reptiles [turtles, the tuatara, crocodylians, and squamates (lizards, snakes and amphisbaenids)] are good model systems for answering questions related to cognitive ecology, from the role of the environment on the brain, behaviour and learning, to how social and life-history factors correlate with learning ability. Furthermore, given their variable social structure and degree of sociality, studies on reptiles have shown that group living is not a precondition for social learning. Past research has demonstrated that non-avian reptiles are capable of more than just instinctive reactions and basic cog-nition. Despite their ability to provide answers to fundamental questions in cognitive ecology, and a growing literature, there have been no recent systematic syntheses of research in this group. Here, we systematically, and comprehensively review studies on reptile learning. We identify 92 new studies investigating learning in reptiles not included in previous reviews on this topic-affording a unique opportunity to provide a more in-depth synthesis of existing work, its taxonomic distribution, the types of cognitive domains tested and methodologies that have been used. Our review therefore provides a major update on our current state of knowledge and ties the collective evidence together under nine umbrella research areas: (i) habituation of behaviour, (ii) animal training through conditioning, (iii) avoiding aversive stimuli, (iv) spatial learning and memory, (v) learning during foraging, (vi) quality and quantity discrimination, (vii) responding to change, (viii) solving novel problems, and (ix) social learning. Importantly, we identify knowledge gaps and propose themes which offer important future research opportunities including how cognitive ability might influence fitness and survival, testing cognition in ecologically relevant situations, comparing cognition in invasive and non-invasive populations of species, and social learning. To move the field forward, it will be immensely important to build upon the descriptive approach of testing whether a species can learn a task with experimental studies elucidating causal reasons for cognitive variation within and among species. With the appropriate methodology, this young but rapidly growing field of research should advance greatly in the coming years providing significant opportunities for addressing general questions in cognitive ecology and beyond.
Circumstances surrounding advances in stranding response and veterinary care have created a growing need for the long-term housing of captive sea turtles. However, the difficulty in recreating natural conditions in captive settings places a responsibility on caregivers to offset wild-type behavioural deficits with enrichment programming that is, preferably, commensurate with the limitations of each enclosure. Though standardised, multi-institutional behavioural monitoring programmes are currently lacking for marine turtles, facilities housing (or planning to house) sea turtles for the long-term are strongly advised to include 'wellness' as a fundamental part of their animal care protocol. Here, concepts of wellness and enrichment in sea turtles are reviewed, and a framework for developing longterm behavioural monitoring programming is provided.
The present study examined effects of retention and intertrial intervals on proactive interference in the eight-arm radial maze performance in rats. A trial consisted of a forced choice of four arms in a learning phase, retention interval, and a free choice among eight arms in a test phase. In Experiment 1, rats were given two daily trials with 10 s or 1 min. retention intervals between the learning and the test phases and with 5, 30, or 60 min. intertrial intervals. In the 1 min. retention condition, proactive inference indexed by decline in performance from the first trial to the second trial was observed regardless of intertrial intervals. In contrast, such decline in performance was not observed for all the intertrial interval conditions in the 10 s retention condition. In Experiment 2, rats were tested with a 1 min. retention interval and 5 or 120 min. intertrial intervals. Significant proactive interference was observed again for a 5 min. intertrial interval condition replicating the results of Experiment 1. In contrast, proactive interference was eliminated completely by lengthening the intertrial interval to 120 min. These results suggest that discriminability among memories in current and prior trials in terms of elapsed time is a determinant of proactive interference in the radial maze performance in rats.
Full-text available
Comparative cognition recently advanced towards a wider taxonomic approach evidenced by an increase in non-avian reptile learning studies but our knowledge still exhibits many gaps. In primates, sociality is linked to enhanced cognitive ability. My aim was to investigate if sociality affects cognitive ability in four Australian lizard species. I specifically focused on behavioural flexibility, which is an index of an organism's ability to cope with environmental change at a cognitive level. I applied the ID/ED attentional set-shifting paradigm which includes several colour/ shape discriminations, reversals of these discriminations, an acquisition of a new colour/ shape discrimination and a shift from colour to shape (and vice versa). Moreover, I tested how age affects learning, if behavioural flexibility correlates with unpredictable environmental conditions and how inhibitory control is exercised in different contexts. Finally, I tested if individual differences in learning could be explained by sex utilising a meta-analytic approach. All four tested species discriminated between one-dimensional stimuli, however, only three out of four showed behavioural flexibility and only two species successfully completed the shift stage learning each set of stages like a new problem. Furthermore, juvenile lizards learnt at adult levels, behavioural flexibility was enhanced in the arid-adapted species and lizards showed context specific inhibitory skills. Neither trials to criterion nor the number of successful individuals differed between the tested species belonging to the Egernia group implicating no adaptations based on sociality in the tested context. Furthermore, the fourth tested, non-Egernia species, failed to perform even a single reversal. Importantly, resource predictability predicted learning proficiency in one species suggesting that other species-specific adaptations underlie differences in learning between species. Similarly, in my meta-analysis a sex difference emerged only between species. Overall, my results contribute important new insights into lizard cognition, however, we need more data on a broader range of lizards to make distinct conclusions on how sociality or ecology affect learning.
Full-text available
Full-text available
The effects of the size of the central arena on the use of response strategies by rats on an eight-arm elevated maze were examined. The size of the central arena had no effect on accuracy, but the use of adjacent arms increased significantly with a larger central arena, regardless of the size of arena to which rats were first exposed. These results are interpreted in terms of foraging efficiency.
Two studies were conducted to examine spatial memory of cattle. In Study 1, six heifers were trained and observed in a radial- and parallel-arm maze at two levels of complexity. Grain was placed at the end of each arm, and heifers were released individually and allowed to choose arms freely until all grain was consumed. Incorrect choices occurred when heifers entered a previously entered arm. At the 4-arm level, the mean number of correct choices in the first four entrances was 3.83 and 3.60 for the radial and parallel mazes, respectively. At the 8-arm level, the number of correct choices in the first eight entrances was 7.78 and 7.36, respectively. Heifers were slightly more efficient (P < 0.05) in the radial maze in which directional and distal cues were more pronounced. In Study 2, two sets of monozygous twin steers were trained in a radial-arm maze using similar procedures as Study 1. The mean number of correct choices in the first eight entrances was 7.68. A variable delay interval was then imposed between Choices 4 and 5. Steers rarely made errors after delay intervals from 5 min to 4 h. Performance appeared to decline (P < 0.1) after an 8-h delay interval. Accuracy declined dramatically (P < 0.001) after a 12-h delay interval. The mean number of correct choices in the first eight entrances was 7.63, 7.29 and 5.80 for delay intervals of 4, 8 and 12 h, respectively. Cattle appear to have the ability to associate several locations with food resources and to remember the locations for periods of up to 8 h.
Two groups of mature “painted” turtles were trained in a T-maze. One group had five trials per day, with correct and incorrect turns reversed for each animal whenever it reached the criterion of five errorless trials on a given day. The second group had 10 trials per day, with correct and incorrect turns reversed daily, irrespective of the performance of the animals. Only in the data for the second group did there appear some indication of progressive improvement in habit reversal. The results are considered in relation to those obtained in analogous experiments with other species.
A turtle was run through a five-blind multiple-T maze, which it learned on the fourth trial. It is believed that vision was used in threading the maze and that the exceptionally quick learning was due to intense motivation (the animal was run only on days when it sought food of its own accord) and deliberateness in making choices. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Tested 24 male albino rats on an 8-arm maze in a paradigm of sampling with replacement from a known set of items until the entire set was sampled. Exps I-III demonstrate that Ss performed efficiently, choosing an average of more than 7 different arms within the 1st 8 choices, and did not utilize intramaze cues or consistent chains of responses in solving the task. Exps IV-VI examined some characteristics of Ss' memory storage. There was a small but reliable recency effect with the likelihood of a repetition error increasing with the number of choices since the initial instance. This performance decrement was due to interference from choices rather than just to the passage of time. No evidence was found for a primary effect. The data also suggest that there was no tendency to generalize among spatially adjacent arms. Results are discussed in terms of the memory processes involved in this task and human serial learning. (27 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
The role of red-footed tortoises (Geochelone carbonaria) and yellow-footed tortoises (G. denticulata) as seed dispersal agents was investigated in northwestern Brazil from 5 to 26 January 2002 by analyzing fecal samples for frequency and viability of seed species and estimating daily displacement of tortoises from recaptured and thread-trailed individuals. Fourteen of 19 fecal samples contained a total of 646 seeds represented by 11 plant species. The most abundant species was Ficus sp. (N= 400) with 100 percent of seeds viable, followed by Aechmea sp. (N = 88) with 93 percent of seeds viable, and Genipa americana (N= 59) with 91 percent of seeds viable. Mean minimum retention time of seeds was 1.6 d and mean daily displacement of tortoises based on recaptured (N= 7) and thread-trailed tortoises (N= 2) was 57 m. Thus, the diversity and proportion of viable seeds consumed by tortoises, combined with the seed retention times and daily movements, suggest they may be effective dispersal agents. These preliminary findings warrant further investigation into the ecological role of these tortoises in Neotropical ecosystems and their contribution to the maintenance of species diversity and forest structure.