Anim Cogn (2009) 12:237–247 DOI 10.1007/s10071-008-0182-z

ORIGINAL PAPER

Evidence against integration of spatial maps in humans: generality across real and virtual environments Bradley R. Sturz · Kent D. Bodily · JeVrey S. Katz · Debbie M. Kelly

Received: 28 March 2008 / Revised: 2 August 2008 / Accepted: 4 August 2008 / Published online: 3 September 2008 © Springer-Verlag 2008

Abstract A real-world open-Weld search task was implemented with humans as an analogue of Blaisdell and Cook’s (Anim Cogn 8:7–16, 2005) pigeon foraging task and Sturz, Bodily, and Katz’s (Anim Cogn 9:207–217, 2006) human virtual foraging task to 1) determine whether humans were capable of integrating independently learned spatial maps and 2) make explicit comparisons of mechanisms used by humans to navigate real and virtual environments. Participants searched for a hidden goal located in one of 16 bins arranged in a 4 £ 4 grid. In Phase 1, the goal was hidden between two landmarks (blue T and red L). In Phase 2, the goal was hidden to the left and in front of a single landmark (blue T). Following training, goal-absent trials were conducted in which the red L from Phase 1 was B. R. Sturz (&) Department of Psychology, Armstrong Atlantic State University, 229 Science Center, 11935 Abercorn Street, Savannah, GA 31419, USA e-mail: [email protected] K. D. Bodily Department of Psychology, Georgia Southern University, P.O. Box 8041, Statesboro, GA 30460, USA e-mail: [email protected] J. S. Katz Department of Psychology, Auburn University, 226 Thach Hall, Auburn, AL 36849, USA e-mail: [email protected] D. M. Kelly Department of Psychology, University of Saskatchewan, 9 Campus Drive, Saskatoon, SK S7N 5A5, Canada e-mail: [email protected]

presented alone. Bin choices during goal-absent trials assessed participants’ strategies: association (from Phase 1), generalization (from Phase 2), or integration (combination of Phase 1 and 2). Results were inconsistent with those obtained with pigeons but were consistent with those obtained with humans in a virtual environment. SpeciWcally, during testing, participants did not integrate independently learned spatial maps but used a generalization strategy followed by a shift in search behavior away from the test landmark. These results were conWrmed by a control condition in which a novel landmark was presented during testing. Results are consistent with the bulk of recent Wndings suggesting the use of alternative navigational strategies to cognitive mapping. Results also add to a growing body of literature suggesting that virtual environment approaches to the study of spatial learning and memory have external validity and that spatial mechanisms used by human participants in navigating virtual environments are similar to those used in navigating real-world environments. Keywords Open Weld · Virtual environment · Human · Spatial · Cognitive map · Integration · Gender diVerences

Introduction A distinction between learning and performance emerged from the seminal work of Tolman (1948). In these nowfamous maze studies with rodents, Tolman oVered converging evidence showing that what is learned is not always what is behaviorally observed. Since that time, focus on a learning/performance distinction has remained at the forefront of psychological investigations. For example, researchers, especially those in the spatial domain, have

123

238

continued to focus their eVorts on determining the contents of spatial memory (Shettleworth 1998). Arguably, the cognitive map has been one of the more controversial concepts concerning the content of spatial memory (see Bennett 1996). Although deWnitions vary (e.g., Gallistel 1990; O’Keefe and Nadel 1978; Thinus-Blanc 1988; Tolman 1948), novel-shortcutting is often taken as evidence in favor of the cognitive-mapping hypothesis. Recently, Blaisdell and Cook (2005) attempted not only to determine the contents of spatial memory but also to determine how they might be integrated into a unitary representational framework. To investigate this issue, the researchers designed a spatial search task for pigeons which allowed them to investigate the processes underlying pigeons’ spatial representation. In an open Weld containing discrete choice locations, pigeons searched for food in the presence of spatially consistent landmarks during two separate training phases. In Phase 1 of training, a blue T-shaped landmark was positioned to the left, and a red L-shaped landmark positioned to the right, of a food reward (see Fig. 1, top panel). In Phase 2 of training, only the blue T-shaped landmark from the Wrst training phase was presented, and now it was located behind and right of the food (see Fig. 1, middle panel). After pigeons were reliably locating the food in each training phase, nonrewarded transfer trials were introduced to test for integration of the spatial information learned in the two training phases. During transfer, the red L-shaped landmark from Phase 1 training was presented alone (see Fig. 1, bottom panel). Three locations were of primary interest: association, generalization, and integration. Responses to the association location (the location to the left of the landmark during testing) were considered indicative of the pigeons forming an association between the red L-shaped landmark and the location of food from the Wrst training phase in which the food was located to the immediate left of the red L-shaped landmark. Responses to the generalization location (the location down and left of the landmark during testing) were considered to have resulted from the generalization of the landmark information from Phase 2 training in which food was located down and left of the blue T-shaped landmark. That is, the red L-shaped landmark may have been treated as the blue T-shaped landmark. Finally, responses to the integration location (down and two locations to the left of the landmark during testing) may have resulted from a combination of the landmark–landmark vector learned in Phase 1 training (T ! L) and the landmark-goal vector learned in Phase 2 training (T ! G) via the common element, the blue Tshaped landmark. As such, in the presence of the red Lshaped landmark pigeons may have inferred the location of the blue T-shaped landmark and responded to the goal location predicted by this inferred landmark.

123

Anim Cogn (2009) 12:237–247

T

L

S

T S

L

3

2

A

1

3

I

G

1

3

2

2

2

3

3

3

3

S Fig. 1 Schematic of the search space from a possible trial from Phase 1 training (top panel), Phase 2 training (middle panel), and transfer (bottom panel). The hashed bin represents the relative location of the goal for that trial, and the S marks the position where participants entered the open Weld and thus started the search task for all training and testing trials. The positions of the landmarks were quasi-randomized across trials (see text for details). In the bottom panel locations for the integration (I), association (A), and generalization bins (G) are marked. Also marked are bin distances from the test landmark

The results from the Wrst of two transfer tests provided no evidence for integration—the pigeons searched at the association and generalization locations more often than at the integration location. As several months had elapsed between Phase 1 training and testing, the pigeons were provided with “reminder” trials (one Phase 1 trial and one Phase 2 trial) prior to testing. After the reintroduction of these Phase 1 training trials, the pigeons’ directed the majority of their searches to the integration and association locations, and these locations were searched more often than the generalization location. This increase in responding to the integration location across tests was interpreted as evidence for integration of spatial maps in pigeons.

Anim Cogn (2009) 12:237–247

Although such evidence for integration of spatial maps by pigeons in an open-Weld search task is also consistent with a recent touch-screen study with pigeons (Sawa et al. 2005) and a water maze study with rats (Chamizo et al. 2006), a conceptual replication and extension of the pigeon open-Weld search task with human participants engaged in a virtual-environment search task suggested that integration may not be the mechanism guiding search behavior (Sturz et al. 2006). SpeciWcally, human participants reproduced the critical Wnding which was taken as evidence for integration of spatial maps in pigeons (i.e., increased responding to the integration location across tests). However, additional Wne-grained analyses of trial-by-trial and within-trial responding, suggested that, across subsequent transfer trials, the participants’ search behavior spread outward from the location of the test landmark—presumably due to a lose-shift strategy. Such an eVect was conWrmed by two control conditions in which either the reminder trials were not presented during the second test (No Reminder) or a novel stimulus was presented in the place of the red Lshaped landmark during testing (Integration Control). The pattern of responding was the same across all three conditions. Seemingly, the outward spread of search behavior produced an increased number of responses to the integration location across tests—the critical comparison taken as evidence for integration of spatial maps. Despite increasing use of virtual environments in spatial research (for a review, see Kelly and Gibson 2007), relatively few direct comparisons have been made between human navigation in real and virtual environments (cf., Klatzky et al. 1998). Hence, the qualitative similarity of mechanisms used by humans to navigate real and virtual environments remains a matter of debate. As a result, we conducted a partial replication of Sturz et al. (2006) using humans in a real-world environment search task to 1) make explicit comparisons of mechanisms used by humans to navigate real and virtual environments and 2) clarify the two competing accounts for the change in search behavior across tests. Although Blaisdell and Cook (2005) found that reintroduction of prior training trials was necessary for pigeons to integrate independently learned spatial maps, the results of Sturz et al. (2006) found that these “reminder” trials did not produce a change for human participants in the pattern of results that were obtained with pigeons (i.e., increase in choices to the integration location across tests). In short, the absence of reminder trials had no eVect on human search performance. Thus, the No Reminder group from Sturz et al. was not included in the current experiment. SpeciWcally, only two groups of participants were tested: Reminder and Integration Control (as in Sturz et al.). The Reminder group served as a conceptual replication of the procedure from the pigeon open-Weld study and the human

239

virtual-environment open-Weld study. The Integration Control group served as a control condition in which the opportunity to integrate spatial maps was eliminated through the presentation of a novel landmark during testing (i.e., a green X). Because a novel landmark could not indicate a trained landmark-goal vector or a trained landmark–landmark vector, its individual or inferential use for determining accurate goal location during testing was eliminated.

Method Participants Fifty-nine University of Saskatchewan students (31 males and 28 females) served as participants at the University of Saskatchewan during the months of June 2006 through January 2007. Nineteen participants (11 males and 8 females) did not meet training criteria (see below) and therefore were excluded from analyses. The remaining 40 participants (20 males and 20 females) were included in all analyses. The mean age of participants that were included in all analyses and opted to provide age information was 19.39 (SEM = 0.49). All participants received extra class credit for participation in the experiment. Apparatus and stimuli All dimensions are length £ width £ height unless otherwise noted. A search space was created by hanging white opaque curtains from the ceiling to the Xoor of an experimental room. A black piece of cloth was aYxed to the curtain on the west wall (i.e., the wall opposite the participant’s starting location) so as to provide an orienting cue. The Xoor of this room (5.55 m £ 3.74 m £ 5.30 m) was covered with shredded paper. Sixteen raised bins (40 cm diameter £ 30 cm height), which also contained shredded paper, were arranged in a 4 £ 4 grid within the room (see Fig. 1). A red L-shaped object (30 cm £ 30 cm £ 56 cm), blue T-shaped object (30 cm £ 30 cm £ 56 cm) and a green X-shaped object (30 cm £ 30 cm £ 56 cm) made of painted styrofoam were used as landmarks. A red cylinder (30 cm diameter £ 56 cm height) and blue cylinder (30 cm diameter £ 56 cm height) made from colored construction paper served as foil landmarks. Procedure Participants were randomly assigned to one of two groups: Reminder or Integration Control. Each group contained a total of 20 participants (10 males and 10 females). Participants were instructed to search for a small red plastic ball that was hidden within the paper shredding in one of the 16

123

240

Anim Cogn (2009) 12:237–247

bins. Participants began each trial at the same starting position located opposite the black cloth (position S in Fig. 1). A choice was deWned as when a participant fully inserted his or her hand into a bin. If the ball was retrieved, a correct choice, participants were required to return the ball to the experimenter and exit the open Weld. Otherwise, participants continued searching until the ball was retrieved. All participants experienced two training phases followed by a testing phase. In each training phase, participants were required to meet the criteria of not more than 2 incorrect bin choices on each of the last two trials to ensure suYcient learning of the landmark-goal spatial relationships. All training and testing trials were conducted in one continuous daily session lasting approximately 1 h. Only the testing phase diVered across the groups of participants. The testing phase consisted of two tests: Test 1 and Test 2. The experimental design and the landmarks used during training and testing are summarized in Table 1. Phase 1 training Phase 1 consisted of eight training trials. For each trial, the reinforced bin was randomly assigned to one of the 16 bin locations. The blue T and the red L were then respectively positioned to the left and right of this bin (Fig. 1, top panel). As a result, the goal location varied from trial-to-trial rendering absolute spatial information uninformative and making the landmarks the only consistent predictor of the goal location. The blue and red cylinders functioned as foil landmarks to ensure participants attended not only to color but also to shape (see Blaisdell and Cook 2005; Sturz et al. 2006) and were respectively positioned to the left and right of a randomly assigned bin (excluding the goal bin and bins directly to the left and right of the goal). Phase 2 training Phase 2 consisted of eight training trials. For each trial, the reinforced bin was randomly assigned to one of the 16 bin locations, with the exception of the row of bins located closest to the west wall because of the landmark–goal relationship: the blue T was positioned up and right of the bin Table 1 Landmark presentations during training and testing by group Training Group

Phase 1

Reminder group testing Test 1 consisted of three blocks with three trials per block. Each block consisted of two training Phase 2 trials followed by one non-reinforced transfer trial. For each transfer trial, a red L was randomly assigned to one of nine locations. The left-most column of bins and the row of bins closest to the starting location (east wall) were excluded as possible locations because assigning the red L to any of these locations would have limited full analyses of participant search strategies. As a result, the location of the landmark varied among the remaining locations across transfer trials. During each non-reinforced transfer trial, participants were permitted to make six choices; after the sixth choice participants were told to stop searching by the experimenter. No foils were presented during testing. All other details of testing were identical to training. The three primary bins of interest were the association, generalization, and integration locations (Fig. 1, bottom panel). Test 2 consisted of two blocks with three trials per block. Each block consisted of one Phase 1 training trial (i.e., the reminder trial), one Phase 2 training trial, and one non-reinforced transfer trial with the red L. All other details were identical to Test 1. Integration Control group testing Test 1 and 2 were conducted in an identical manner to the Reminder group with the exception that a green X (i.e., a novel landmark) was substituted for the red L during transfer trials.

Testing

Phase 2

Test 1

Test 2

Reminder

TL

T

T, T, L

T L, T, L

Integration Control

TL

T

T, T, X

T L, T, X

Phase 1 and Phase 2 consisted of eight trials each. Test 1 consisted of three blocks of three trials and Test 2 consisted of two blocks of three trials. One block from Test 1 and Test 2 is represented. Foils presented during Phase 1 and Phase 2 training are not included

123

containing the ball (Fig. 1, middle panel). As a result, the goal location varied from trial-to-trial rendering absolute spatial information uninformative and making the landmark the only consistent predictor of the goal location. The blue cylinder functioned as a foil landmark to ensure participants attended not only to color but also to shape (see Blaisdell and Cook 2005; Sturz et al. 2006) and was positioned to the left of a randomly assigned bin (excluding the location of the blue T) for the Wrst Wve trials of training (as in Blaisdell and Cook; Sturz et al.). The foil was absent in the remaining three trials. All other details were identical to Phase 1 training.

Results Acquisition Participants’ search behavior rapidly came under the control of the consistent landmark(s) (the blue T and the red L during Phase 1, and the blue T during Phase 2).

Anim Cogn (2009) 12:237–247

241

12

Mean Errors

10 Male Phase 1 Male Phase 2 Female Phase 1 Female Phase 2

8 6 4 2

Proportion of Choices

0.5

Left T Right T Left Foil Other

Males

0.4

0.3

0.2

0.1

0.0

Second

First

Choice 0.5

Females

Left T Right T Left Foil Other

0.4

Proportion of Choices

Figure 2 shows mean search error on each of the eight training trials within the two training phases, averaged across participants in each of the groups for males and females. A three-way mixed analysis of variance (ANOVA) on errors with Trial (1–8), Phase (1, 2), and Gender (male, female) as factors revealed a main eVect of Trial, F7,266 = 73.91, P < 0.001, a main eVect of Phase, F1,38 = 10.41, P < 0.01, and signiWcant Trial £ Phase, F7,266 = 3.59, P < 0.001, Trial £ Gender, F7,266 = 2.21, P < 0.05, and Trial £ Phase £ Gender, F7,266 = 5.17, P < 0.001 interactions. All other main eVects and interactions were not signiWcant (F1,38 = 1.73, P = 0.2 and F1,38 = 0.8, P = 0.38 for Gender and Phase £ Gender respectively). Despite initial diVerences in average search error across phase and gender (as evidenced by the presence of the signiWcant interactions), all participants reached training criteria, performed with few to no errors by Trial 5, and maintained this level of accuracy for the remainder of the training session. The Wrst training trial in Phase 2 provided an opportunity to assess participants’ strategies when presented with the blue T and blue cylinder alone for the Wrst time making this trial similar to a transfer trial. Figure 3 shows the proportion of the Wrst two choices during Phase 2 for males (top panel) and females (bottom panel) collapsed across groups plotted by choice location. The majority of participants initially selected a bin directly to the left or right of the blue T. First choice analysis showed that these two bins were selected signiWcantly more often than would be expected by chance for both males (45% of Wrst choices were made to the left bin and 40% of Wrst choices were to the right bin,
2 1,N=20 = 50.61 and 38.37, respectively, both Ps < 0.001) and females (35% of Wrst choices were made to the left bin and 45% of Wrst choices were to the right bin,
2 1,N=20 = 27.83 and 50.61, respectively, both Ps < 0.001). Second choice

0.3

0.2

0.1

0.0 First

Second

Choice Fig. 3 Proportion of Wrst two choices during Phase 2 for males (top panel) and females (bottom panel) collapsed across groups plotted by choice location

analysis also showed that these two bins were selected signiWcantly more often than would be expected by chance for both males (40% of second choices were made to the left bin and 45% of second choices were to the right bin,
2 1,N=20 = 38.37 and 50.61, respectively, both Ps < 0.001) and females (35% of second choices were made to the left bin and 40% of second choices were to the right bin,
2 1,N=20 = 27.83 and 38.37, respectively, both Ps < 0.001). This suggests that the majority of participants used strategies learned during Phase 1 as they searched either left or right of a consistent landmark (85 and 86%, respectively,
2 1,N=40 = 166.6 and 179.2, both Ps < 0.001). These analyses indicate that the majority of participants’ initial choices were guided by either an association or generalization strategy. Transfer

0 1

2

3

4

5

6

7

8

Trial Fig. 2 Mean errors during acquisition averaged across groups for males and females by phase and trial. Error bars represent standard errors of the mean

Choice type The three bins of primary interest were the association, generalization, and integration locations (see Fig. 1, bottom

123

242

Spatial distribution of choices To more fully illustrate the changes in bin choices across tests, Fig. 5 shows the spatial distribution of choices to each bin for each group and gender (columns) across the Wve transfer trials (rows). Choice distributions are centered at the integration bin (4, 4). As shown, responding was concentrated around the test landmark (L or X) in transfer Trial 1 but began to spread with repeated presentations. Despite

123

0.18

Mean Choices (proportion)

panel). Selection of the association bin would provide evidence that choices were under landmark control from Phase 1 because during this phase the goal was located left of the red L and right of the blue T. Selection of the generalization bin would provide evidence that choices were under landmark control from Phase 2 because of a generalization from the blue T to the red L. Selection of the integration bin would provide evidence that participants were integrating spatial information because they were able to combine the two spatial maps. A four-way mixed ANOVA on mean proportion of choices with Test (1, 2), Choice Type (integration, association, generalization), Group (reminder, integration control), and Gender (male, female) revealed a signiWcant main eVect of Test, F1,36 = 14.7, P < 0.001 and Choice Type, F2,72 = 13.02, P < 0.001 with a signiWcant Test £ Choice Type interaction, F2,72 = 9.41, P < 0.001. There was no main eVect of Group (F1,36 = 0.001, P = 0.98), but there was a main eVect of Gender, F1,36 = 5.23, P < 0.05. All other interactions were not signiWcant (Test £ Group, F1,36 = 0.02, P = 0.88; Test £ Gender, F1,36 = 0.56, P = 0.46; Choice Type £ Group, F2,72 = 0.09, P = 0.92; Choice Type £ Gender, F2,72 = 1.88, P = 0.16; Test £ Group £ Gender, F1,36 = 1.63, P = 0.21; Test £ Choice Type £ Group, F2,72 = 0.68, P = 0.51; Test £ Choice Type £ Gender, F2,72 = 0.16, P = 0.86; Choice Type £ Group £ Gender, F2,72 = 0.21, P = 0.81; Test £ Choice Type £ Group £ Gender, F2,72 = 0.48, P = 0.62). Figure 4 shows mean proportion of choices to the integration, association, and generalization bins collapsed across the three transfer trials from Test 1 and the two transfer trials from Test 2 (as performed by Blaisdell and Cook 2005; Sturz et al. 2006). In Test 1, more choices occurred to the association and generalization bins than to the integration bin. In Test 2, choices decreased to the association and generalization bins. Choices to the integration bin did not diVer across tests. Although groups were performing qualitatively similarly across tests, females were allocating more responses to the three focal bins in comparison to males. Across tests, choice responses signiWcantly decreased to both the association and generalization bins (F1,36 = 9.45 and 36.98, respectively, both Ps < 0.01), whereas choices to the integration bin did not change across tests (F1,36 = 0.01, P = 0.92).

Anim Cogn (2009) 12:237–247

0.15

Male Test 1 Male Test 2 Female Test 1 Female Test 2

0.12 0.09 0.06 0.03 0.00 Integration

Association

Generalization

Choice Type Fig. 4 Mean proportion of choices to integration, association, and generalization bins averaged across groups and trials separated by gender for Test 1 and 2. Error bars represent standard errors of the mean

variability in the direction of the shift in search behavior, males and females in both groups moved away from the test landmark across transfer trials. This movement away from the test landmark could account for the decreased choices to the association and generalization bins from Test 1 to Test 2, as shown in Fig. 4. Distance analyses To clarify the shifts in search behavior from Test 1 to Test 2, the mean response distance (in bins) from the test landmark (i.e., red L or green X) for gender, group, and averaged across groups for males and females was plotted across the Wve transfer trials in Fig. 6. Distances were calculated by counting the number of bins that each bin was displaced from the test landmark (see Fig. 1, bottom panel). As shown, search behavior moved away from the test landmark with the accumulation of choice responses during the Wrst four transfer trials and returned to this landmark during transfer Trial 5. A three-way mixed ANOVA of Trial (1– 5) £ Group (reminder, integration control) £ Gender (male, female) on distance revealed a main eVect of Trial, F4,144 = 40.46, P < 0.001 and Gender, F1,36 = 5.82, P < 0.05. The absence of a signiWcant Group £ Trial interaction (F4,144 = 1.15, P = 0.34) indicated that the Reminder and Integration Control groups performed the same across transfer trials. A trend analysis on the Trial factor revealed signiWcant linear, F1,36 = 32.45, P < 0.001, quadratic, F1,36 = 152.01, P < 0.001, and quartic, F1,36 = 39.21, P < 0.001, components indicating that search moved away from, but then returned to, the test landmark. It is important to note that the integration bin was always located 2 bins from the test landmark yet search behavior was concentrated well below this distance during transfer Trial 1 for both males (M = 1.23, SEM = 0.08) and females

Anim Cogn (2009) 12:237–247

243

MALE: Reminder

8

MALE: Control

FEMALE: Reminder

FEMALE: Control

8

7

7

6

6

X

L

5

L

X

5

4

4

Figure 5.

3

3

2

2

1

1

0

0

8

8

7

7

6

6

5

X

L

L

5

X

4

4

3

3

2

2

1

1

0

0

8

8

7

7

6

6

L

5

X

L

X

5

4

4

3

3

2

2

1

1

0

0

8

8

7

7

6

6

5

L

X

L

5

X

4

4

3

3

2

2

1

1

0

0

8

8

7

7

6

6

L

5

X

L

X

5

4

4

3

3

2

2

1

1

0

0

1

2

3

4

5

6

7

8

0

1

2

3

4

5

6

7

8

Fig. 5 Spatial distribution of choices to each bin plotted for each gender and group (columns) across the Wve transfer trials (rows). Transfer trials 1–5 appear consecutively from top (Trial 1) to bottom (Trial 5). The Wrst three transfer trials formed Test 1 (solid grid lines), and the last two trials formed Test 2 (dashed grid lines). The size of a circle at

0

1

2

3

4

5

6

7

8

0

1

2

3

4

5

6

7

8

0

an intersection is indicative of the mean proportion of choices to that spatial location. The distribution is centered at the integration bin (4, 4). L or X = location of test landmark; Wlled red circle = association bin choices (5, 5); Wlled blue circle = generalization bin choices (5, 4); Wlled purple circle = integration bin choices (4, 4)

123

Anim Cogn (2009) 12:237–247

Mean Distance from Test Landmark (in bins)

244 3.0

Transfer Trial 1 choices

2.5

Collectively, these results oVer evidence that integration was not the mechanism responsible for the changes in bin choices across tests; however, it is possible that participants integrated on their initial choices and then proceeded to shift their search as a result of the consequences of the choice responses. To assess the possibility of such early integration, each choice response during transfer Trial 1 was analyzed. Table 2 shows the number of participants (observed) choosing the integration, association, generalization, and other bins (parsed into distances of one bin or two or greater bins) for the six choices of transfer Trial 1 compared to the expected values. Expected values were calculated based on the probability of choosing a particular bin(s) multiplied by the total number of participants (40). For example, the probability of responding to the integration bin was 1/16 = 0.0625 multiplied by 40 yielding an expected value of 2.5. As shown, more participants than would be expected by chance responded to the association and generalization bins for their Wst choice. This result was conWrmed by separate Chi-Squares,
2 1,N=40 = 8.64, P < 0.01,
2 1,N=40 = 116.2, P < 0.001, respectively. Also as shown, more participants than would be expected by chance responded to the association, generalization, and other 1 bins for their second choice. This result was also conWrmed by separate Chi-Squares,
2 1,N=40 = 30.83, P < 0.001,
2 1,N=40 = 8.64, P < 0.01,
2 1,N=40 = 6.53, P < 0.05, respectively. Although three responses did occur to the integration location during the third choice, this was not more than would be expected by chance,
2 1,N=40 = 0.11, P > 0.7. Moreover, these three responses were the only responses (out of a total 240 responses) to occur to the integration bin during transfer Trial 1. As a

2.0 1.5 Male Reminder Male Integration Control Male Mean Female Reminder Female Integration Control Female Mean

1.0 0.5 0.0 1

2

3

4

5

Transfer Trial Fig. 6 Mean distance from test landmark (in bins) averaged across groups by transfer trial (6 responses per trial) separated by gender. The dashed line represents the distance of the integration bin from the test landmark. Error bars represent standard errors of the mean

(M = 1.13, SEM = 0.07) and well above this distance by transfer Trial 4 for both males (M = 2.69, SEM = 0.06) and females (M = 2.53, SEM = 0.08). These results were conWrmed by one-sample t tests that compared mean distance for males and females in transfer Trial 1, t19 = ¡8.68, P < 0.001, t19 = ¡10.22, P < 0.001, respectively, and transfer Trial 4, t19 = 4.49, P < 0.001, t19 = 3.12, P < 0.01, respectively to the distance of the integration bin (2.0 bins). Of note, both male (M = 1.62, SEM = 0.01) and female (M = 1.58, SEM = 0.04) participants returned to search near the test landmark on transfer Trial 5. Search distance during this transfer trial was also well below the distance of the integration bin as conWrmed by separate one-sample t tests for males and females, t19 = ¡8.64, P < 0.001, t19 = ¡9.75, P < 0.001, respectively.

Table 2 Observed and expected values for the six choices during transfer Trial 1 by choice type

Choice

Integration Bin

Association Bin

Generalization Bin

Observed Expected

Observed Expected

Observed Expected

Other 1 bin Distance Observed Expected

Other 2 bins or > Distance Observed Expected

1

0

2.5

7

2.5

19

2.5

8

10

6

22.5

2

0

2.5

11

2.5

7

2.5

17

10

5

22.5

3

3

2.5

9

2.5

4

2.5

23

10

1

22.5

4

0

2.5

4

2.5

3

2.5

27

10

6

22.5

5

0

2.5

2

2.5

2

2.5

30

10

6

22.5

6

0

2.5

3

2.5

1

2.5

29

10

7

22.5

123

Anim Cogn (2009) 12:237–247

result, it is reasonable to conclude that during transfer Trial 1 participants did not integrate spatial maps but associated from Phase 1 training or generalized from Phase 2 training.

Discussion In the present open-Weld search task, participants rapidly acquired each training map. Subsequent transfer tests indicated that participants seemed primarily controlled by a generalization strategy. Regardless of group or gender, responses to both the association and generalization locations decreased across tests, whereas responses to the integration location did not change. In addition, the search behavior of both males and females in both the Reminder and Integration Control groups was initially concentrated near the test landmark but progressively moved away from this object with its repeated non-reinforced presentation. Moreover, detailed analysis of transfer Trial 1 ruled out the use of an integration strategy early during testing as responses to this location were not signiWcantly more than would be expected by chance. Comparing our current results to those of Sturz et al. (2006) and Blaisdell and Cook (2005), there is a glaring diVerence: responses to the integration location did not increase across tests. One plausible explanation for an absence of this eVect (and, as a result, absence of evidence for integration of spatial maps by humans) is that participants failed to encode the spatial relationship between landmark T and landmark L during Phase 1 training. SpeciWcally, if participants failed to encode the T ! L spatial relationship in Phase 1 training, they would not be able to utilize the L alone in testing to predict the location of the T and hence the location of the “goal”. However such a result is unlikely as all participants were controlled by the landmark–goal spatial relations from training prior to testing. SpeciWcally, all participants reached training criteria in Phase 1 with 35 out of 40 participants (87.5%) not making a single error in the last trial of Phase 1 (M = 0.15, SEM = 0.05). Moreover, participants responded more than would be expected by chance to the right of the T (association bin) for their Wrst choice in Phase 2. As a result, it seems reasonable to conclude that participants also encoded the landmark–landmark spatial relationship but applied some form of the rules learned during training when encountering a novel situation (i.e., Wrst trial of Phase 2 and Wrst transfer trial)—relying exclusively on search strategies learned within the task. Although an absence of an increase to the integration location across tests is problematic for an integration of spatial maps account, as it is the deWnitive evidence, the lack of such evidence does not prove problematic for a non-integration account of participants’ search behavior.

245

SpeciWcally, Sturz et al. (2006) showed that search behavior progressively shifted away from the test landmark with the accumulation of non-reinforced choice responses and suggested that such an outward shift could produce increased responding to multiple locations displaced from this landmark. As a result, an increase to any speciWc location (e.g., the integration location) is not critical as long as overall search behavior moves away from the test landmark. In the present task, the spatial distribution of searches and distance analyses are consistent with a non-integration account as responses were initially concentrated near the test landmark but began to spread well beyond the integration location with repeated testing. Interestingly, as an aside, female participants allocated more responses to locations near the test landmark than their male counterparts. Such a result is seen in multiple analyses in the current experiment including choice type (Fig. 4), spatial distribution of choices (Fig. 5), and distance analyses (Fig. 6). In conjunction, these results suggest a stronger reliance on the test landmark by females compared to males, and this diVerence is consistent with sex diVerences found with spatial tasks in both real (Dabbs et al. 1997; Choi et al. 2006) and virtual environments (Astur et al. 1998; Kelly and Bischof 2005; Sandstrom et al. 1998). It has been suggested that this diVerence is due to the use of topographic navigation strategies by females and Euclidean navigation strategies by males (e.g., Choi and Silverman 1997; Dabbs et al. 1997), and support for such hypotheses is consistent with recent evidence for sex diVerences in neural correlates of spatial navigation (Grön et al. 2000). In comparing present results collected in a real-world environment to those of Sturz et al. (2006) collected in a virtual environment, it seems relevant to discuss possible experiential diVerences between the two environments that may impact participants’ navigation abilities and strategies. Although by no means exhaustive, some potential critical diVerences include: physical eVort required to navigate each environment, availability and amount of proprioceptive and vestibular feedback received during the locomotor task, amount and quality of auditory, olfactory, and visual cues, and visual angle. Seemingly, any of these experiential diVerences have the possibility to produce diVerences in results across tasks; yet despite these diVerences, performance in both the real-world and virtual versions of the task were markedly similar. In short, no evidence was obtained in either task to indicate that the participants learned to integrate spatial maps. Instead, participants in both tasks relied on strategies learned during training; that is, they searched left and right of a landmark. When these learned strategies were not successful participants continued to search for the goal location, and this continued search for the goal location appeared to be guided by a lose-shift strategy.

123

246

Despite these similarities, however, participants in the present task returned to search near the test landmark during transfer Trial 5 after a progressive outward shift in search behavior whereas those in the virtual task did not. Although these expanding search patterns in both environments resemble that of the desert ant (Cataglyphis) attempting to locate its nest at the terminal point of a return trip (Wehner and Srinivasan 1981), suggesting similarity in the search process across species, only participants in the real environment showed evidence of a return to the start of search. It is possible that this diVerence reXects experiential diVerences across environments; however, it seems more plausible that this diVerence reXects procedural diVerences across tasks. SpeciWcally, the participants in the present task completed half as many training trials as those in the virtual task (8 vs. 16, respectively). As participants in both tasks reached asymptotic performance in training by Trial 5, it remains unclear what inXuence the additional training had on participants engaged in the virtual task compared to those in the real-world task. For example, extended training may inXuence the time course of a shift in search behavior or a possible threshold for switches in search strategies. Seemingly, this variable warrants further investigation to dissociate environmental versus training eVects. As suggested by Sturz et al. (2006), there are two possible interpretations of the human data compared to that of the pigeons (Blaisdell and Cook 2005): there exists either a qualitative similarity or diVerence in the way in which pigeons and humans perform the open-Weld search task. The present results lend further support to a view of qualitative similarity across species. SpeciWcally, humans in both real-world and virtual-environment versions of the present task performed qualitatively similarly when subjected to identical Wne-grained analyses, and this qualitative similarity across search tasks is also consistent with the extant comparative literature (for a review, see Cheng et al. 2006). For example, parallel results have been shown from analogous touch-screen and open-Weld search tasks for both humans and pigeons (Spetch et al. 1996, 1997). In conclusion, results from the present real-world environment search task are consistent with those found in a virtual-environment search task by Sturz et al. (2006) suggesting that humans did not integrate spatial maps. These results have comparative implications for the cognitive mapping hypothesis. SpeciWcally, we found no evidence for the use of cognitive mapping during navigation. During critical test trials, participants did not integrate independently learned spatial maps but applied strategies learned during training to novel situations. Although not without exceptions (see Chamizo et al. 2006; Sawa et al. 2005), present results are consistent with the bulk of recent Wndings suggesting the use of alternative navigational strategies to cognitive mapping (e.g., Foo et al. 2005; Gibson

123

Anim Cogn (2009) 12:237–247

2001; Gibson and Kamil 2001; Waller et al. 2000; for a review see Wang and Spelke 2002). The speciWc reasons for these conXicting results for and against the cognitivemapping hypothesis remain unknown; however, potential critical factors may relate to stability and consistency of spatial information. SpeciWcally, it has been shown that landmark stability plays a critical role in spatial learning in that they may be ignored if perceived as unstable (Biegler and Morris 1993, 1996; JeVrey 1998; Learmonth et al. 2001). Seemingly, these eVects are related to the relative predictability of spatial information, and it has been suggested that spatial information may be weighted in inverse proportion to its variance (Cheng et al. 2007). Such hypotheses may explain the absence of integration in spatial task that employ mobile landmark and/or conXicting spatial information with respect to landmark–landmark and/or landmark–goal associations. Perhaps as important, the present results also add to a growing body of literature suggesting that virtual environment approaches to the study of spatial learning and memory have external validity and that spatial mechanisms used by human participants in navigating virtual environments are similar to those used in navigating real-world environments (Arthur et al. 1997; Hartley et al. 2003; Kelly and Bischof 2005; for a review, see Kelly and Gibson 2007). As a result, virtual environments seem to be ideally suited for testing comparative mechanisms of spatial learning, memory, and cognition. Acknowledgments This research was supported by an Alzheimer Society of Canada Grant to Debbie M. Kelly and a National Science Foundation Grant (0316113) to JeVrey S. Katz. This research was conducted following the relevant ethical guidelines for human research. The authors would like to thank Danielle Fontaine and Jim Reichert for their assistance with data collection. The authors also would like to thank three anonymous reviewers for comments on an earlier version of the manuscript.

References Arthur EJ, Hancock PA, Chrysler ST (1997) The perception of spatial layout in real and virtual worlds. Ergonomics 40:69–77 Astur RS, Ortiz ML, Sutherland RJ (1998) A characterization of performance by men and women in a virtual Morris water task: a large and reliable sex diVerence. Behav Brain Res 93:185–190 Biegler R, Morris RGM (1993) Landmark stability is a prerequisite for spatial but not discrimination learning. Nature 361:631–633 Biegler R, Morris RGM (1996) Landmark stability: further studies pointing to a role in spatial learning. Q J Exp Psychol 49B:307–345 Bennett AT (1996) Do animals have cognitive maps? J Exp Biol 199:219–224 Blaisdell AP, Cook RG (2005) Integration of spatial maps in pigeons. Anim Cogn 8:7–16 Chamizo VD, Rodrigo T, Mackintosh NJ (2006) Spatial integration with rats. Learn Behav 34:348–354 Cheng K, Shettleworth SJ, Huttenlocher J, Rieser JJ (2007) Bayesian integration of spatial information. Psychol Bull 133:625–637

Anim Cogn (2009) 12:237–247 Cheng K, Spetch ML, Kelly DM, Bingman VP (2006) Small-scale spatial cognition in pigeons. Behav Process 72:115–127 Choi J, McKillop E, Ward M, L’Hirondelle N (2006) Sex-speciWc relationships between route-learning strategies and abilities in a large-scale environment. Environ Behav 38:791–801 Dabbs JM, Chang E, Strong RA, Milun R (1998) Spatial ability, navigation strategy, and geographic knowledge among men and women. Evol Hum Behav 19:89–98 Foo P, Warren WH, Duchon A, Tarr MJ (2005) Do humans integrate routes into a cognitive map? Map- versus landmark-based navigation of novel shortcuts. J Exp Psychol Learn 31:195–215 Gallistel CR (1990) The organization of learning. MIT Press, Cambridge Gibson BM (2001) Cognitive maps not used by humans during a dynamic navigational task. J Comp Psychol 115:397–402 Gibson BM, Kamil AC (2001) Tests for cognitive mapping in Clark’s nutcrackers. J Comp Psychol 115:403–417 Grön G, Wunderlich AP, Spitzer M, Tomczak R, Riepe MW (2000) Brain activation during human navigation: gender-diVerent neural networks as substrate of performance. Nat Neurosci 3:404–408 Hartley T, King JA, Burgess N (2003) Studies of the neural basis of human navigation and memory. In: JeVery KJ (ed) The neurobiology of spatial behavior. Oxford University Press, New York, pp 144–164 JeVrey KJ (1998) Learning of landmark stability and instability by hippocampal place cells. Neuropharmacology 37:677–687 Kelly DM, Bischof WF (2005) Reorienting in images of a three-dimensional environment. J Exp Psychol Hum 31:1391–1403 Kelly DM, Gibson BM (2007) Spatial navigation: spatial learning in real and virtual environments. Comp Cogn Behav Rev 2:111–124 Klatzky RL, Loomis JM, Beall AC, Chance SS, Golledge RG (1998) Spatial updating of self-position and orientation during real, imagined, and virtual locomotion. Psychol Sci 9:293–298 Learmonth AE, Newcombe NS, Huttenlocher J (2001) Toddlers’ use of metric information and landmarks to reorient. J Exp Child Psychol 80:225–244

247 O’Keefe J, Nadel L (1978) The hippocampus as a cognitive map. Oxford University Press, Oxford Sandstrom NJ, Kaufman J, Huettel SA (1998) Males and females use diVerent distal cues in a virtual environment navigation task. Cogn Brain Res 6:351–360 Sawa K, Leising KJ, Blaisdell AP (2005) Sensory preconditioning in spatial learning using a touch-screen task in pigeons. J Exp Psychol Anim B 31:368–375 Shettleworth SJ (1998) Cognition, evolution, and behavior. Oxford University Press, New York Spetch ML, Cheng K, MacDonald SE (1996) Learning the conWgurations of a landmark array: I. Touch-screen studies with pigeons and humans. J Comp Psychol 110:55–68 Spetch ML, Cheng K, MacDonald SE, Linkenhoker BA, Kelly DM, Doerkson S (1997) Learning the conWgurations of a landmark array in pigeons and humans: II. Generality across search tasks. J Comp Psychol 111:14–24 Sturz BR, Bodily KD, Katz JS (2006) Evidence against integration of spatial maps in humans. Anim Cogn 9:207–217 Thinus-Blanc C (1988) Animal spatial cognition. In: Weiskrantz L (ed) Thought without language. Oxford University Press, Oxford, pp 371–395 Tolman EC (1948) Cognitive maps in rats and men. Psychol Rev 55:189–208 Waller D, Loomis JM, Golledge RG, Beall AC (2000) Place learning in humans: the role of distance and direction information. Spat Cogn Comput 2:333–354 Wang RF, Spelke ES (2002) Human spatial representation: insights from animals. Trends Cogn Sci 6:376–382 Wehner R, Srinivasan MV (1981) Searching behaviour of desert ants, genus Cataglyphis (Formicidae, Hymenoptera). J Comop Physiol A 142:315–338

123