Spatial navigation in large-scale virtual environments: Gender differences in survey tasks

Abstract

Most of the studies on gender differences in spatial abilities have focused on traditional paper and pencil cognitive tests, while these differences have been less investigated in navigational tasks carried out in complex virtual environments (VEs). The aim of the present study has been to evaluate gender differences in route and survey knowledge by means of specific tasks (route-learning, pointing, landmark-placing) carried out in two separate VEs. In addition the male and female participants were subjected to a battery of spatial abilities tests and specific self-report questionnaires. The results showed a significant difference favouring males in the survey tasks, as well as in the spatial abilities tests; on the contrary, no gender differences were found in the route task. Moreover, a different pattern of correlations among the measures were found in the male and female sub-groups.

Introduction

As demonstrated by Sandstrom, Kaufman, and Huettel (1998) most of the studies on spatial ability have not provided for navigational tasks. By navigational tasks, we mean tests which imply the real or “virtual” movement of subjects within large-scale environments, geometrically defined and with landmarks. In such environments the subjects do not have the possibility to comprehend all the spatial information from a single point of view.

In order to evaluate possible differences over and above “simple” spatial abilities, several authors have used bi-dimensional maps (Galea and Kimura, 1993, Holding and Holding, 1989). In a study conducted by Galea and Kimura, the subjects were asked to take a new route on a 2D map of a city (Galea & Kimura, 1993). The male participants made fewer mistakes in learning the route than the female ones and took less time to do the test. In addition, while the females recalled more landmarks than the males, the latter demonstrated more knowledge of the geometric properties of the map.

Despite the fact that the various elements in 2D can easily be controlled and manipulated, it is not clear how far the differences noted in finding streets using 2D maps are representative of the differences found in real navigation. The subjects are only able to comprehend all the environmental information from a single point of view, thus lacking one of the key characteristics of spatial navigation. In addition, as demonstrated by Moffat, Hampsom, and Hatzipantelis (1998) and Sandstrom et al. (1998), the subjects see and understand an abstract representation of the environment, rather than moving inside it. Such structural differences leave open the possibility that spatial abilities subtended to learning through 2D maps and those subtended to real navigation do not completely overlap. This said, 2D maps have the advantage to make easier the acquisition of survey knowledge with respect to sequencial navigation, since maps provide a single view on a complete environment. A study by Golledge, Doughery, and Bell (1995) compared route-based to survey-based acquisition of spatial knowledge by means of two different spatial learning conditions, map-based vs. navigational-based. The results of the study showed some evidence of the advantage of map-based with respect to route-based learning in understanding spatial relations within an environment. On the basis of this finding the authors suggested that exposure to maps positively adds to the understandings we acquire from travelling through an environment. In addition, as far as gender differences were concerned, the present study did not evidence a difference between male and female across all tasks, but only in certain specific spatial task in which female outperformed male.

In recent years some studies have investigated gender differences in spatial navigation tasks through virtual reality methodology (VR), by constructing virtual environments (VEs).

In the wake of studies conducted on rats, some authors have investigated these differences using computerised versions of uncomplex environments similar to the Morris water maze (MWM) or the radial arm maze (RAM). In the standard version of these tasks the animals have to look for a hidden platform in a large pool of water (MWM) or find rewards in the form of food (RAM) (Roof, 1993, Roof et al., 1993, Williams and Meck, 1991). What consistently emerges from such studies is an advantage in favour of male rats, who demonstrate, for example, greater speed in acquiring knowledge of new environments. Male rats appear to rely more on the geometric information present in the environment, while information regarding landmarks also seems to be of particular importance to female rats (Williams et al., 1990, Williams and Meck, 1991).

Also in studies on man, carried out using methodology similar to MWM, we again find an advantage in favour of males (Astur et al., 1998, Sandstrom et al., 1998). Sandstrom and his team tested 48 subjects, 24 male and 24 female, in a MWM-type VE (Sandstrom et al., 1998). In the training phase the subjects were asked to find a hidden target, starting each time from different starting points. The labyrinth used provided both geometric information (trapezoidal room) and landmark information (four different landmarks placed in the labyrinth). The testing phase was sub-divided into four sub-phases. In the “stable landmark environment”, in which the geometric information was varied, meaning the shape of the room, while keeping the position of the landmarks unchanged, the performance of the females substantially overlapped the males’. On the contrary, in the two conditions in which information from the landmarks could not be used to find the target (an environment without landmarks and an environment with the landmarks positioned at random compared to the training phase) the performance of the males turned out to be better than the females’. The males tend to use both geometric and landmark information, while the females selectively use landmark information to find their way in new environments. Consequently varying the position of landmarks in an environment has more effects on the performance of females than on males.

Similarly, Astur and his team used a computerised version of the MWT to measure these types of differences in navigational abilities (Astur et al., 1998). The male and female participants were instructed to navigate towards a hidden platform. The males carried out the task better, taking less time to find the hidden platform. They also spent more time swimming in the quadrant of the swimming-pool where the platform was placed. Despite the fact that the male participants declared having had more experience of video games than the female, these differences did not statistically account for the different performances between the two sexes which emerged from the MWM.

Such studies reveal how virtual environments like the MWM represent a useful instrument in the study of navigational abilities. However, such VEs do not appear to be sufficiently complex. Environments such as the MWM may appear not to be very ecological compared to more complex virtual environments and therefore more similar to real ones.

A study which has analyzed these types of differences in a more complex virtual environment was conducted by Moffat et al. (1998). The navigational environment consisted of a labyrinth with perpendicular intersections and without landmarks. The participants were asked to find a way out in two different labyrinths having a similar level of difficulty. Five different trials were carried out in each of the two experimental labyrinths. The time needed to complete each trial, the route taken and the number of mistakes made were recorded. The results showed how the performance of the males was globally more thorough and quicker than the females: the males took significantly less time and made fewer mistakes in completing the different trials. The authors also excluded the possibility that the major experience of video games reported by the males in the self-report questionnaires might account for the differences found in the navigational trials. For both sets of measurements made during the navigational trials, the difference found remained substantially unchanged when the experience of video games was imposed as a covariant. As expected, the males scored significantly higher in the spatial cognitive tests. There were also significant correlations between the scores obtained in the cognitive tests and the trials carried out in the labyrinth, thereby confirming the results of other studies (Darken and Silbert, 1996, Waller, 2000). Another recent study investigated spatial knowledge, in a battery of tasks including wayfinding, estimated of direction and distance and a map placement task, by means of a complex VE (Tlauka, Brolese, Pomeroy, & Hobbs, 2005); the results showed significant differences favoring males in the majority of the tasks carried out in the virtual shopping centre.

To sum up, the desktop VEs may be considered valid and effective instruments in the study of spatial navigation abilities since they allow a systematic modification of the environmental features and a systematic recording of behaviour (Peruch et al., 1995, Ruddle et al., 1997, Tlauka and Wilson, 1994, Waller, 2000).

A recent work by Saucier and Green tried to establish whether the differences frequently observed between males and females in navigational tasks were somehow linked to the use of different strategies of the two sexes or to different basic abilities (Saucier & Green, 2002). Starting from published data (Dabbs et al., 1998, Lawton, 1994), of a greater focussing on the geometric properties of the environment by the men against a greater reliability on landmarks inside the environment by the women, the authors controlled the experimental conditions by varying the navigational instructions given to the subjects. One half of the experimental sample was asked to take certain routes, towards unknown destinations, following landmark-based instructions, while the other half were given Euclidean-based instructions (north–south–east–west) to reach the same destinations. The results indicated that the females who followed the Euclidean-based instructions made significantly more mistakes than both the females who followed the landmark-based instructions, and the males who followed the landmark and the Euclidean-based instructions. The same result pattern emerged after analyzing the time taken to complete the routes, with the females who followed the Euclidean-based instructions being significantly slower than the other three subject groups.

The same experiment was subsequently repeated by the authors using a 2D matrix as the “navigational environment”, that is to say, a new environment unfamiliar to the participants, in place of the real environment of the previous experiment. This was done to keep the familiarity with the environment variable under greater control. The results overlap with those in the first experiment: the males appear to do better when they use Euclidean information, while the females turn out to be more suited to landmark-based instructions. Such differences have been interpreted by the authors as confirmation that these observed differences reflect dimorphic ability in using these two types of spatial information.

A further interesting observation emerging from the Saucier study (Saucier & Green, 2002) is the significant correlation between a mental rotation test and the number of mistakes made by the subjects who had followed the Euclidean instructions. The correlation does not appear to be significant for the subjects who had navigated with landmark-based instructions; this correlation accounts for more than 50% of the variance of the Euclidean-based trial. It may be hypothesised that most of the mistakes made by the females in the “Euclidean” trial partly derive from their reduced ability to maintain bearings with respect to the four cardinal points, attributable to their reduced ability in the mental rotation test.

A further factor that seems to play some role in navigational tasks is “spatial anxiety” (Coluccia & Louse, 2004). Lawton developed the “Spatial Anxiety Scale” (SAS) in order to measure the levels of anxiety that subjects experience in situations that require spatial navigation (Lawton, 1994). As Schmitz has shown, emotional factors must be considered as one of the mediators in the learning and development of navigational competences (Schmitz, 1997). To date, few studies have analyzed the potential effect of such a phenomenon on spatial learning in new environments. In the study of Saucier et al. there emerged a greater level of anxiety in the females than in the males (Saucier & Green, 2002). The level of anxiety found in girls was significantly higher than in boys in the study carried out on young people by Schmitz (1997). She also demonstrated the presence of significant correlations between anxiety levels and some navigational measurements, such as the time and number of mistakes in the first experimental trial. Correlations between spatial anxiety, greater in females, and performance in spatial/navigational tasks have been demonstrated by other studies (Bryant, 1982, Lawton, 1994, Lawton, 1996).

Lawton showed that the males appear more inclined to navigate in environments by relying on absolute geographic co-ordinates (orientation strategy), while environmental features, landmarks, especially those situated near decisional points, assume particular relevance for the females (route strategy) (Lawton, 1994). In the same study there was a superior performance by the males in the mental rotation test and in a test on spatial perception, while the females obtained significantly higher scores in the SAS (Lawton, 1994). Another interesting observation is the significant correlation between spatial anxiety and the orientation strategy subscale, while there was no relationship with the route strategy subscale.

A study conducted by Waller (2000) tried to measure, using multivariant analysis, the relationship between the various factors involved in spatial navigation in virtual reality environments: the classic paper and pencil spatial tests, spatial representations acquired in real environments, the gender, experience and the attitude towards using the computer, competence in the use of interfaces involved in VE experiments and the ability to acquire and transfer knowledge from a VE to a real one. In the model tested by the author spatial abilities in cognitive tests and ability in using interfaces turned out to be the most powerful predictors of the spatial performances in VE. According to the author, the gender, even though representing a significant predictor of the ability to acquire knowledge from a VE, would not become one primarily, but through its association with other factors, such as the visual–spatial abilities and ability in the use of interfaces.

The results of these studies demonstrate that spatial navigation is a complex ability involving different heterogeneous factors.

Different classifications of the various components and stages involved in the acquisition of spatial knowledge have followed one another through the years.

As Schmitz demonstrated (Schmitz, 1997), the first theories in the field of spatial knowledge followed a sequence of stages, acquired during development mostly in a serial way (Hart and Moore, 1973, Piaget and Inhelder, 1967): the egocentric stage, the allocentric stage, and the geocentric stage. In the first stage spatial knowledge is strictly bound up in self, while in the allocentric stage we begin to acquire spatial knowledge independently of self, on the basis of the spatial properties of the environment. In the final stage, geocentric, spatial knowledge no longer linked to external reference points is formed, but with abstract and absolute co-ordinates, like the Euclidean properties.

Starting with such a classification, literature refers to two types of strategy/knowledge which individuals rely on in the exploration/learning of environments: route knowledge and survey knowledge (Maguire et al., 1998, O’Keefe and Nadel, 1978, Russell and Ward, 1982, Siegel and White, 1975). According to some authors, starting with the theorisations of Siegel and White (1975), a further preliminary stage of knowledge is represented by landmark knowledge, in other words, a stage in which the landmarks present in an environment are simply noted and confined to memory.

In route knowledge the prospective and viewpoint adopted by the subject are centred on self and factors such as sequentialising, the variation of body position compared to the environment and visual information directly accessible to the subject (street intersections and identity of landmarks) will all acquire importance. Instead, survey knowledge implies the formation of a map of the environment by the observer; a “bird’s eye view” where all the environmental information is potentially available. Route knowledge therefore involves a sequence of instructions about how to get from one point to the next, whereas survey knowledge implies the construction of a cognitive map of the environment, a map which “integrates” the different routes by means of the relationships between the different locations (Lawton, 1994). In addition, despite the fact that both types of knowledge allow adequate navigation within the environments, route knowledge, being more rigid and sequential, can more easily lead to mistakes and disorientation whenever any of the segments of the navigational sequence are changed (Lawton, 1994). Survey knowledge is considered, instead, to be more elastic and flexible, allowing for diversion from pre-established routes and greater ability in adapting to obstacles and changes to the environment (O’Keefe & Nadel, 1978).

In the light of these theoretical differences, much experimental evidence can be traced to a difference in the use of these two types of knowledge: the females appear to prefer a route strategy, while the males to rely more on survey strategy. Several studies have shown how both environmental and behavioural factors may favour the formation of one or the other type of knowledge/representation, and have an influence on spatial navigation ability (Coluccia and Louse, 2004, Jacobs et al., 1997, Jacobs et al., 1998, Maguire et al., 1999, Steck and Mallot, 2000).

One of the fundamental factors in distinguishing between the two types of strategies is the familiarity of the subject with the navigational environment. In fact, a greater knowledge of the environment would seem to favour the formation of survey representations (Aginsky, Harris, Rensink, & Beusmans, 1997). In reality, as Aginsky has pointed out, not all the studies agree on this point (Aginsky et al., 1997, Montello, 1998). In the model proposed by Montello, there is not a qualitative transition sequence between different stages but a quantitative increase in the accuracy and completeness of spatial knowledge: a quantitative rather than a qualitative shift. In this framework there is no stage at which pure landmark or route knowledge exists and survey knowledge begin to be acquired on first exposure to a novel environment (Montello, 1998). Also in the study under discussion the author suggests that the spatial navigational tasks can be solved by using a survey strategy right from the start, without necessarily having to go through the landmark and route stages, according to the sequence suggested by Siegel and White (1975) (landmark, route, survey).

The familiarity variable appears to be a problem whenever the experiments are conducted in real environments, already familiar to the subjects, since it is not possible to control the various degrees of familiarity of the subjects.

Another important aspect is the presence of local and global cues (landmarks) which would seem to favour the formation respectively of spatial route and survey knowledge/representations (Aginsky et al., 1997, Gale et al., 1990, O’Keefe and Nadel, 1978, Steck and Mallot, 2000). Global landmarks, being potentially visible from any point within the navigational environment and so from a great distance, become absolute points of reference, favouring orientation strategies in survey terms. On the contrary, local landmarks, only visible from certain points and at reduced distances, often assume the function of relative and sequential reference points, leading to inferences such as “ when I see the shop, I have to turn right”. As O’Keefe and Nadel have suggested (O’Keefe & Nadel, 1978), global landmarks provide world-centred directions which determine a global reference setting that does not change when the subject moves short distances. The authors themselves, instead, define navigation guided by local landmarks as navigation relying on intermediate stage sequencing outlined by the same local landmarks. For these reasons they are closely linked to route navigation (Steck & Mallot, 2000). Such assumptions should not however be taken strictly, since some objects which function as global landmarks in some of the navigation phases might subsequently be used as local landmarks (Steck & Mallot, 2000). Individual differences in the use of the different strategies, like other features of navigational environments, interact with each other in determining the use made of the different types of landmark.

Despite the growing interest in the study of navigational skills, up to now few studies have examined the differences between males and females in relation to the distinction between route and survey representations with the possibility of experimental control offered by VR. Furthermore, few studies have tried to make a thorough study of the relationship between navigational and spatial abilities measured by the classic paper and pencil spatial tests.

The aim of the study has been to evaluate gender differences in route and survey navigational tasks carried out in complex VEs, and their relationship with the cognitive tests on basic spatial abilities and specific self-report scale. To examine route knowledge we used a route-learning task, where the subjects were asked to go back the same way they had come, after just one exploration, in a VE without global landmarks. A previous study had used a similar test of route-reversing as an indicator of route knowledge (Pick, 1993). The acquisition of survey knowledge was measured by using a pointing task and a landmark-placing task on a 2D map of the VE; both tests are considered reliable indicators of survey knowledge (Waller, 2000). Two important factors favouring spatial survey knowledge were examined: familiarity and global landmarks; a prolonged phase of familiarisation with the VE was in fact envisaged, and this was also characterised by the presence of global landmarks. We hypothesised that the males would perform better than the females in the survey tasks.