Visual landmark orientation by flying bats at a large-scale touch and walk screen for bats, birds and rodents

Visual landmark orientation by flying bats at a large-scale touch and walk screen for bats, birds and rodents

Journal of Neuroscience Methods 141 (2005) 283–290 Visual landmark orientation by flying bats at a large-scale touch and walk screen for bats, birds ...

234KB Sizes 2 Downloads 25 Views

Journal of Neuroscience Methods 141 (2005) 283–290

Visual landmark orientation by flying bats at a large-scale touch and walk screen for bats, birds and rodents York Wintera,b,∗ , Sophie von Mertena , Hans-Ulrich Kleindienstc a Department of Biology, University of Munich, Germany Max-Planck Institute for Ornithology, Seewiesen, Germany Max-Planck Institute of Behavioural Physiology, Seewiesen, Germany b


Received 24 March 2004; received in revised form 1 July 2004; accepted 2 July 2004

Abstract Orientation depends on multi-modal information about the locally perceptible environment (local view) in many situations. We developed a behavioural paradigm for investigating visual orientation of flying bats based on a large-scale touch screen (1.2 m × 1.8 m). It functions by a grid of rows and columns of infra-red beams just in front of a screen with back-projected visual stimuli. Approaching animals interrupt the beams and thus permit automatic recording of the time and place of an animal’s locational choice. We used it as a vertical touch surface. Installed as a horizontal walk surface, it may also serve as a more natural ‘firm ground’, circular arena analogue to the ‘Morris water maze’ for investigating orientation behaviour and spatial cognition from rodents to birds while offering automatic real-time recording of paths, times and latencies with enhanced possibilities to score details of motor behaviour and to control stimuli interactively. Bats offer a unique possibility to investigate the use of both echo-acoustic and visual information processing pathways for the process of self-localization and orientation. In our first experiment, a bat was presented with five identical targets, one central and four peripheral and had to choose the central target. After task acquisition, the array was shifted by the distance between targets, so that a formerly peripheral landmark was now in the absolute location of the formerly central target. At small inter-target distances, the bat ‘went with’ the array, and chose the new central target (at a new absolute location). With 30 cm or more of inter-target distance (60 cm across the landmark configuration), however, the bat went with absolute location, and chose a peripheral target. In experiment 2, the bat was presented with two landmarks 30 cm apart and an unmarked target located at midline beneath them. On tests, the landmarks either maintained training distance or were expanded to 50 cm apart. On such expansion tests, the bat chose most the location at the correct vector from the right landmark. This showed that the bat first identified a single landmark by the configuration and then applied a previously learnt vector (angle and distance) to locate the target. Glossophaga did not orient by pure angular geometry between landmarks and target. © 2004 Elsevier B.V. All rights reserved. Keywords: Visual orientation; Navigation; Spatial memory; Expansion test; Touch screen; Morris water maze; Walk screen; Glossophaga

1. Introduction Localizing a known goal in space often depends on the processing of exteroceptive (allothetic) information about the local environment for the process of self-localization, orientation and way finding. While exteroceptive environmental information is often multi-modal, the emphasis of research has been on visual information and the corresponding information for orientation has been labelled the ‘local view’ ∗

Corresponding author. E-mail address: [email protected] (Y. Winter).

0165-0270/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jneumeth.2004.07.002

(Collett and Collett, 2002; Redish, 1999). Neuronal pathways and specializations for extracting spatial information from visual images are well known in mammals (Burgess et al., 1999). Bats differ from other mammals in that their primary mode for environmental scanning is echo-acoustic (Neuweiler, 2000). Despite this, they also have a fully developed visual system (Neuweiler, 2000), important for orientation (Chase, 1981, 1983; Chase and Suthers, 1969; Ekl¨of and Jones, 2003; Ekl¨of et al., 2002; Suthers et al., 1969; Suthers, 1970; Manske and Schmidt, 1979; Rydell and Ekl¨of, 2003; Winter, 1999; Winter et al., 2003). The flower- and nectarspecialists among the bats, the glossophagines, use visual


Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290

cues for localizing flower targets. Their ultraviolet sensitivity (Winter et al., 2003) appears to be a specialized adaptation for this task, as flowers of many bat pollinated plant species have a correspondingly high degree of UV reflectance (Burr et al., 1995). Together, this enhances floral contrast perception during the twilight hours of the day when the ambient illumination spectrum shifts towards short wavelengths (Smith, 1982). Bats thus offer a unique possibility to investigate the use of both echo-acoustic and visual information processing mechanisms and pathways for the process of self-localization and orientation. This requires a suitable behavioural testing paradigm. For this study, we developed such a method based on a large-scale touch screen that is also suitable for studying visual orientation by flying animals. A touch screen allows the presentation of (animated) visual stimuli and records the time and spatial distribution of behavioural choices made by an experimental subject (Bhatt and Wright, 1992; Biegler et al., 2001; Bussey et al., 1998; Cheng and Spetch, 1995; Markham et al., 1996; Morrison and Brown, 1990; Spetch, 1995; Spetch et al., 1992, 1996; Stich et al., 2003; Wright et al., 1988). We used such a system to examine the rules applied by Glossophaga bats to infer positional information from arrays of visual landmarks. We were interested in the small scale navigation mechanisms of pinpointing a goal within sensorial reach of the bat. Animals may infer the location of a goal from configurations of landmarks and spatial cues by a variety of possible mechanisms which have been reviewed in depth by a number of authors (e.g. Cheng and Spetch, 1998; Healy, 1998; Kamil and Cheng, 2001; MacDonald et al., 2004; Redish, 1999; Shettleworth, 1998). For our study, we first asked at which spatial scale a Glossophaga bat pays spontaneous attention to the local geometry or relational configuration of an array of visual spatial cues. We examined this question with a visual cue displacement experiment at different spatial scales of landmark configurations. Based on the initial results about the use of local scale geometry, we investigated in a landmark expansion experiment the orientation process whereby a bat locates an unmarked spatial goal from configurations of two landmarks. 2. Methods 2.1. Touch screen components Computer generated symbols were rear projected onto an opaque plexi glass pane (180 cm × 120 cm; Degussa R¨ohm, plexiglas GS white 072; 5.0 mm thick, transmission 40%) by means of a data projector with long-life UHP lamp or with an LCD-panel combined with a daylight projector (the technically less susceptible variant). Touching of the projection panel is detected through two perpendicularly arranged IR-light beam grids (Banner, Mini-Array; Animal Cognition Systems). The basic unit of these sensors consists of a pair of emitter and detector bars that lie opposite one another. The distance between grid beams is 9.53 mm. Two

pairs are employed here: horizontal bars 181 cm long and 192 IR-diodes and detectors (Banner, BMEL and BRML 7232A with a MAC-1 controller and QDC-515C) and vertical bars 120 cm long and 128 diodes and detectors (Banner, BMEL and BRML 7232A with a MAC-1 controller and QDC-515C). The grid beam field is positioned a few centimetres in front of the projection screen. Due to the sideways spread of the IR-lights, light beams can reach the sensor not only by the direct way, but also through reflection from the screen. A blind prevents the sidewise spread of the beams and therewith the unwanted continuous release. 2.2. Programming The single beams are sequentially activated by the controller (MAC-1 and QDC-515C) and interrogated at time intervals of 55 ␮s. A complete cycle for 192 grid beams lasts 10.6 ms. Data transmission to the controlling computer per serial interface lasts 4 ms (at 36 kbaud) and a complete sensing interval approximately 15 ms. Several scan analysis modes are available. In “straight” mode the grid beams are scanned sequentially and only those that lie opposite one another are activated by the LED and sensor. “Skip scan” increases response speed at the expense of decreased sensing resolution because only every second to eighth beam is scanned. This was not necessary for this experiment as 15 ms was a sufficiently fast sensing response for the bats. An increased spatial resolution of better than 9.5 mm is offered by scanning per “interlacing”. This mode alternates a “straight scan” with a slanted-beam scan. Thereby a light beam is registered not by its opposite receiver but instead by the direct neighbour. For this mode the distance between the sender and the receiver should be three times the length of the observation area, because only in the middle-third do the slanted beams come near the middle of where the direct beams lie. In experiments with bats, however, an increased spatial resolution was not necessary, because the animal’s bodies and wings interrupt the 9.5 mm grid dependably. The information on interrupted grid beams follows, in our case, via the “first beam blocked” (FBB) and “contiguous beams blocked” (CBB) modes (Fig. 1B). FBB supplies the number of the first interrupted beam between emitter and detector bar. CBB supplies the number of interrupted grid beams: the widest connected interrupted grid beam distance. Grid beam interruptions are collected in X- and Y-directions independently from one another and transmitted to the control computer asynchronously. From the position of the first interrupted beam and the width of continuously interrupted beams we calculate the midpoint of grid beam interruptions for each single scan. The X- and Y-positions are provided with a time marker and from this time information, corresponding XY-pairs are joined. The ensuing series of XY-coordinates are organised into a single event. If there is no interruption for a period of 200 ms, a touch event is treated as ended. The final location computed for each choice is the mean of all position fixes in X and Y directions that occurred during the sensitive

Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290


Fig. 2. Succession times of light beam interruptions separated for X- (B) and Y (A)-directions of a bat in front of the touch screen. (A) Vertical extension (B) horizontal extension. The black squares connected by a line show the width over which light beams were interrupted during a single scan. The region between the two dotted lines in A and B indicate the target region. The hatched area shows the sensitive time interval, during which the touch event is scored as a choice by the animal. Grid distance is 9.53 mm. Arrows indicate the progression of time during the event.

Fig. 1. General setup for the large format touch screen (1.2 m by 1.8 m). (A) Vertically oriented, C horizontally oriented with walking lattice (g) and a projection mirror (m), touch screen (ts), projector (p). (B) It illustrates the method of determining XY-coordinates. First beam blocked (FFB) contiguous beams blocked (CBB), for further details see text. (D) Circular maze for rodents. A bright circular arena is projected from underneath (as in C) onto the walk screen. XY coordinates of the subject are transmitted in real-time so that on reaching the invisible target site (t) a black target area is projected and the projected bright circle is reduced in intensity or extinguished. IR: infrared beams, b: border of circular arena, t: target. Continuous water circulation over the walk screen prevents deposition of olfactory marks. In A, using a mirror reduces the horizontal projection distance (touch screen components by Banner and by Animal Cognition Systems).

time window (hatched time intervals in Fig. 2). In the case of a flying bat with wings interrupting many light beams in a rhythmic pattern when hovering in front of a screen the accuracy of the computed location fix is probably not much better than ±5 cm, despite the 0.95 cm grid spacing. In our set up the grid of beams laid 10 cm in front of the projection screen. The flower bats that we used flew, as a rule, towards a target point in an upswing from bottom to top. Thereby in approaching the target point on the vertical screen the lower grid beams are usually interrupted before the upper target point is reached. Furthermore, a bat that is interrupting the beams persistently may also move sideways through the beams. This can also lead to the “swiping” problem (Morrison and Brown, 1990). In order to ensure that the primary target point of an approach flight is valued as the animal’s place of choice, our programme is set up to evaluate only those interruptions that lie within a time window of 30 and 200 ms after the first grid interruption (hatched areas in Fig. 2). Subsequent interruptions are registered but not used to evaluate the choice position. Limiting of the sensitive time interval forces the animal to approach the correct position directly. 2.3. Training and shaping procedure Training begins with the animal learning to visit a feeder (here with 17% sugar water), which is temporarily fixed to


Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290

the projection surface of the touch screen. Next, it learns to alternate (continuous strict alternation) between visits to the feeder on the touch screen and an external feeder situated beneath the screen. An electronically controlled flap on the external feeder opens when a reward is available. When the animal alternates 80% correctly, it must learn to approach the touch screen at the projected sign shown behind a feeder at one of three positions. Correct choices are rewarded, the sign is turned off, and after visiting the external feeder, a new sign is projected. The positions of feeders and landmark projections are changed regularly to counteract early place training. To facilitate the learning of the response reward associations an acoustic signal is also given for a correct choice. This is important for later training when no more rewards are given on the touch screen surface. During final training food rewards on the touch screen are gradually reduced and at the same time food amounts at the external feeder are increased. Eventually, the ‘feeders’ on the touch screen are replaced by non-rewarding decoys and those are made successively smaller in size. The originally cylindrically formed feeders are replaced on the projection screen by just the nectar tubes and later by knotted pieces of tubing, which serve as echoacoustically perceivable markers, but do not give food. The initial training is completed when the bat approaches the visual signs on the touch screen without any aid structures. It has then learnt to touch specific visual signs or locations on the screen as an activating switch for the external feeder. 2.4. Experiment I—regional scale geometry We were interested in the geometric rules applied by a Glossophaga bat to infer positional information from arrays of visual landmarks. For this we first asked at which spatial scale a Glossophaga bat pays spontaneous attention to the local geometry of an array of visual spatial cues. We examined this question with a visual cue displacement experiment which we performed at different spatial scales. A single male Glossophaga soricina was trained in a 5 m by 5 m room to operate the touch screen as a switch to activate an external feeder as described above. This required about 6000 choices made by the bat at the touch screen. Subsequently, the bat was trained for experiment I. A trial during experiment I consisted of a learning phase and a test phase. The bat received rewards during the learning phase but not during the test phase. During the learning phase, the bat was shown an array of five equidistant, visual landmarks (circles with 5 cm diameter, Fig. 3A). The touch sensitive goal areas were 10 cm squares centred around the bottom edge of a landmark. When a bat touched one of the five landmarks the projection extinguished. A choice of the central landmark was rewarded. Approaches to the touch screen area outside the landmarks did not terminate the projection. Five seconds after a wrong choice or immediately after the reward had been collected the same projection was repeated. The learning phase for a trial lasted for at least 15 min and was terminated when at least 75% of the last 20 choices had been correct.

Fig. 3. Configural displacement experiment with five visual landmarks (experiment I). (A) An animal was trained to touch the central of five landmarks (black dots); after task acquisition, landmarks were moved by one inter-landmark distance A either horizontally (shown here) or vertically. In the test array (circles) the position of one landmark is at the absolute location (al) as during learning while another landmark occupies the relative position (rp) within the array as during learning (central position). (B) Frequency of choices of landmarks at ‘absolute location’ (black dots) or ‘relative position’ (circles) during tests at different spatial scales of distance between landmarks by a single bat. Dotted line shows expected frequency of random choice. The small symbols connected by dotted lines in b show the results of an equivalent experiment with Glossophaga commissarisi but using echo-acoustic instead of visual cues (data from Thiele and Winter, in press).

A 5 min retention interval followed the learning phase. During the subsequent test phase the presentation of the array was shifted by one inter-landmark distance either vertically or horizontally as compared to the previous learning phase (Fig. 3A). After the bat touched one of the landmarks, the projection extinguished but was shown again after 5 s. No rewards were given during the test phase. A total of six identical presentations were shown during the test phase of a trial. A test phase was followed immediately by the learning phase of the next trial. Distances between landmarks varied between 20 and 50 cm (20, 25, 30, 35, 40 and 50 cm, see Fig. 3B) during different trials. The positions of the landmark array on the projection screen, the distances between landmarks and the direction of shift (vertical or horizontal) were randomised for each single trial, with the

Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290


restriction that successive parameter combinations always differed. Choice frequencies of ‘absolute location’, ‘relative position’ and ‘other’ were tested for deviation from random choice for each different scale using Chi-square tests with a Bonferroni correction of P-values (multiplication by two) to adjust for double testing of the same data set. 2.5. Experiment II—interfloral distance and spatial cues Glossophaga paid attention to configurations of visual spatial cues (experiment I) and evaluated their geometry for the orienting response towards a target when cues were distributed within a 50 cm diameter area. In experiment II, we investigated the geometric rules by which bats find the spatial location of an unmarked target from a configuration of two landmarks. The experiment was performed with the same male used for experiment I. The test period of this experiment was preceded by a learning period. During the learning period two landmarks (5 cm diameter) were projected with a horizontal distance of 30 cm between them (Fig. 4A). The position of the target was 25 cm distant from these landmarks and at midline between them (Fig. 4A). The angle between target and landmarks was 74◦ . During learning presentations the target location was marked by projecting an auxiliary diagonal cross of 1.25 cm in width. Three touch sensitive areas (10 cm squares) were defined as choice areas: two at the displayed landmarks and one at the target point. When a bat touched one of these areas the projection extinguished. For a correct choice of the target area it also received a reward. After each choice the two landmarks were projected at a new location on the screen. Presentations alternated at random between 13 different projection sites. The auxiliary target landmark was not shown during every presentation. Initially, it was shown during 80% of all presentations but this was reduced to 50% at the end of the learning period. During the test period of this experiment, three different presentations were shown to the bat in random order (frequency of presentation given in parentheses): 30 cm landmarks with auxiliary target sign (40%), 30 cm landmarks without auxiliary target sign (40%), and 50 cm expanded landmarks, no target sign (20%). Landmark positions were alternated at random between 13 different positions on the screen. Rewards were only given for choices of the target area during 30 cm landmark projections. During expanded presentations all touch events were recorded but only touch events to a 40 cm × 40 cm area positioned midway between and 10 cm below the two landmarks was scored as a choice and extinguished the projection (Fig. 4C). Within this 40 cm by 40 cm area we defined four sub-areas (10 cm × 10 cm) in line with the predictions of the different hypotheses for possible orientation mechanisms (Fig. 4). The area ‘angle’ was at the corner of the 74◦ angle between the two landmarks. The remaining three areas were 19 cm below the horizontal line of the landmarks: ‘right’ and ‘left’ were defined by the indepen-

Fig. 4. Two-landmark expansion test (experiment II). (A) A bat was trained to touch an area 19 cm distant (74◦ angle) from two landmarks spaced 30 cm apart. During the learning phase, an auxiliary sign was shown at the target location during 50% of the presentations. For the test, landmarks were expanded to 50 cm distance. No auxiliary sign was shown. Potential goal locations defined by angular and/or vector relationships to one or both landmarks are now at different locations of the test screen. (B) Spatial distribution of choices during training trials (30 cm landmarks) without auxiliary goal sign. (C) Spatial distribution of choices during tests with expanded landmarks (50 cm) and without auxiliary goal sign. Rectangles mark sub-areas ‘left’, ‘middle’, ‘right’ and ‘angle’ (see text). Open circles indicate landmark positions during previous training (not shown during tests).


Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290

dent vectors (angle and distance) between the two landmarks and the original target location. ‘Middle’ was midway between the two vector defined positions. For the analysis we tested if number of choices of one of these four areas differed from the expectation of random choice within our 40 cm by 40 cm choice area. We used Chi-square tests with Bonferroni corrections to adjust for multiple testing of the same data set. 3. Results 3.1. Experiment I During the initial trials the bat learnt to always choose the central landmark. For this it needed 100 and 91 choices during the first two trials. Afterwards, a performance of 75% correct out of 20 choices was reached within 22 (±6 S.D.) choices. The behavioural choices by the bat were performed while in flight. Thus, the bat did not ‘wait’ in front of the touch screen for subsequent projections but instead circled in the 5 m by 5 m room and approached the touch screen only when a new projection was shown. Approaches, as observed by video, were along a straight course towards the target ending in an upswing shortly before reach of the target. During training when the bat received rewards at the touch screen the bat would hover briefly (less than 1 s) in front of the screen to imbibe the sugar solution. This corresponds to the commonly observed flight trajectory of glossophagine flower bats while visiting flowers (Winter, 1998). During trials without rewards at the screen the bat did not remain hovering in front of the screen but simply performed a pendulum like upswing followed by immediate retreat. We completed 33 trials with 193 choices made by the bat (four–six choices per test). Landmark locations during test phases were assigned to the three categories ‘absolute location’, ‘relative position’ and ‘other’ (Fig. 3). Choice of spatial category was dependent on the distance between landmarks. The landmark at ‘absolute location’ was chosen above chance level at all distances except for 25 cm (Chi-square: χ2 > 11, d.f. = 1, PBonferroni < 0.01, Fig. 3B). The landmark at ‘relative position’ was chosen above chance level only at 20 and 25 cm distances (Chi-square: χ2 > 11, d.f. =1, PBonferroni < 0.001, Fig. 3B). There was some indication that the observed preference for ‘relative position’ at 20 and 25 cm inter-landmark distances was stronger after vertical displacements than after horizontal displacements. 3.2. Experiment II As during experiment I the bat approached its chosen target directly. We never observed the bat to first approach a landmark and then veer off to the target location, which would have indicated the use of some kind of sensorimotor vector. The bat reached 80% correct choices of the target area without auxiliary sign after 673 presentations from the beginning of the learning period. During the test period, the bat maintained a performance of 82% correct choices during

Fig. 5. Frequency of choices of touch screen sub-areas during expansion tests (experiment II). For placement of sub-areas see Fig. 4. ∗∗∗ P < 0.001 (see text for statistics). Data are from Fig. 4C.

control presentations with landmarks at 30 cm and without auxiliary sign (N = 278, Fig. 4B). We performed 74 test trials with expanded landmarks (Fig. 4C). The area outside our designated four choice areas was selected significantly less often (N = 19) than expected by chance (Chi-square: χ2 35.6, d.f. = 1, PBonferroni < 0.001, Fig. 5). The areas ‘right’ and ‘middle’ were both chosen more often than expected by chance (Chisquare: χ2 > 16, d.f. = 1, PBonferroni < 0.001, Fig. 5), while frequency of choices for ‘left’ and ‘angle’ did not differ from chance expectation (Chi-square: χ2 < 0.5, d.f. = 1, n.s., Fig. 5). 4. Discussion 4.1. Large-scale touch screen With this study, we established a new behavioural paradigm based on a large-scale touch screen for studying visual orientation by flying or other highly mobile animals. It can be used for behavioural research on cognition and memory in various animals. Spatial memory tasks are a popular neuroscientific diagnostic tool for assessing various neurophysiological, pharmacological, etc. effects on memory. We used it to examine visual landmark orientation by flying bats. While echolocation is the dominant mode of orientation in bats, all bat species also have functional eyes and use them for orientation. A bat of the nectar-feeding species Glossophaga soricina needed about 6000 trials to learn the operant task of touching a sign on the screen in order to activate a feeder beneath the projection panel. This was achieved in less than 1 month of training due to these animals’ naturally high rate of activity. Further training for specific tasks was much faster. Here, the bat needed another 191 presentations to learn to touch a single, visible sign within a configuration or 673 presentations to touch an unmarked area defined by its spatial position relative to two landmarks. The 9.5 mm resolution of the touch grid was entirely sufficient as the animal always moved half of its body into the sensitive plane of the grid of infrared beams (Fig. 2). The restriction of valid touch events to a sensitive time window by the operating software, success-

Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290

fully prevented the problem of “swiping” by the test animal (Fig. 2). This instrument can also serve as an alternative to the Morris water maze if used as a touch screen comparable in size to the typical swimming pool task (Fig. 1D). The swim task is not the ideal behavioural task for rats and mice as the required swimming is not natural to these animals. In addition, scoring is typically laborious and crude: latency to reach target, or time spent in quadrants. A touch screen with the size of a typical rodent pool would provide a much richer database: paths, times, and latencies of travel, etc. This can serve to better diagnose alternative explanations such as deficits in motor behaviour. In addition, the easy to obtain real-time recording of position coordinates offers enhanced possibilities for stimulus control. 4.2. Visual orientation in bats Our experimental results from a single animal must be regarded as preliminary. In experiment I the bat showed the ability to use spatial cues from two different categories (Fig. 3). When visual landmarks were distributed over an area of 70 cm in diameter the bat used its own absolute spatial reference derived from the spatial configuration of the experimental room to select the single landmark to be approached as a beacon. When visual landmarks were spaced more closely together, i.e. within an area of 50 cm diameter or less, neighbouring landmarks influenced local orientation. In that situation the bat evaluated the geometry between at least three landmarks to infer the identity of the central landmark. Thus, the bat seems to have relied on its internal representation and guidance structures of the room to locate a target area of approximately 50 cm in diameter. Subsequently and within that target area the bat matched local spatial cues (its local view) with expectations to fine tune goal point selection. Interestingly, this spatial scale for local visual orientation coincides with results from an analogous experiment but using echo-acoustic cues with Glossophaga commissarisi (Fig. 3). Here, within an area of 40 cm diameter choice behaviour was influenced by regional scale geometry of echo-acoustic landmarks but at 80 cm diameter it was not (Thiele and Winter, in press). This identity in influence of regional scale geometry across sensory modalities may indicate a generality of Glossophaga’s orientation ability. These bats appear to rely on the precision of their spatial orientation to guide them within ±25 cm of the target. Only within such a restricted area do they evaluate local spatial cues (both visual and echo-acoustic) to pinpoint the goal point. Such a reliance on their own spatial precision, at least, seems to be the case within the geometry of the laboratory rooms used for our experiments. How does Glossophaga infer goal locations from configurations of landmarks? The expansion experiment (experiment II) was designed to discriminate between a number of possible alternatives. Glossophaga bats evaluate regional scale geometry within goal areas of 50 cm diameter or less. We therefore chose the spatial scale of experiment II such that


after expansion, both landmarks were still within a 50 cm area. The location of a goal is defined by its distances and angles (here, to the gravitational vector) to neighbouring landmarks. When several landmarks are available, both pure distance information and pure angular information are sufficient to define a goal location. Alternatively, also a single landmark within a directional frame of reference is sufficient for determining the endpoint of a goal vector. This latter possibility appears to have been mostly used by the bat. The bat most often touched the area defined by the vector between the right landmark and the goal during the learning phase (Figs. 4 and 5). In order to use this vector for orientation the bat must have followed a two step orientation process of first identifying the right landmark within the configuration and then derive goal location from the learnt vector. We know that the bat did perform the first step in this sequence of choices, select the right landmark, rather than just selecting an area below and to the left of any landmark: the equivalent area to the left of the left landmark was selected only once. The bat also chose the area ‘middle’ more frequently than expected by chance. This area is defined by its correct position on the vertical midline between the landmarks and by the correct distance to the line connecting both landmarks. Choice of this area would be expected if a bat averaged the information from the vectors to both landmarks along the horizontal dimension. The caveat of this interpretation is, that the location ‘middle’ during expansion tests coincided with the absolute locations of the goal during previous learning. Thus, if the bat had been able to remember the 13 absolute goal locations on the screen it might also have preferred one of those locations in the test situation while disregarding local spatial cues. Experiment I has shown that even at small distances to neighbouring landmarks the bat sometimes oriented towards the previously learnt absolute location, thus disregarding neighbouring positional cues (Fig. 3B). Glossophaga used purely visual landmark information for final goal selection. It used this visual spatial information both as goal cues (beacons) and as relative spatial cues. The bat went through up to three successive stages of orienting reactions during target approach in our experimental set up: (i) it used internally generated cues or guidance by the configuration of the room to approach a target area with a diameter of about 50 cm (ii) within that area it evaluated local spatial cues (a minimum of three in experiment I) to identify the goal cue; when the target itself was not visible but instead predicted from a configuration of two landmarks the bat first identified a preferred landmark within the set and then (iii) determined the vector from this single landmark to the target location. The preference to maintain angle and distance to a single landmark during orientation after expansion of landmark arrays as found here, has also been shown for gerbils (Collett et al., 1986), pigeons (Spetch et al., 1996, 1997), young children and marmoset monkeys (MacDonald et al., 2004; Spetch et al., 1996). A tendency to average between the goal locations defined by two independent vectors is known


Y. Winter et al. / Journal of Neuroscience Methods 141 (2005) 283–290

from pigeons (Cheng, 1988, 1989; Cheng and Spetch, 1998) and corvids (Kamil and Jones, 1997). Only adult humans have been shown to maintain a geometry of fixed angles during expansion tests (MacDonald et al., 2004; Spetch et al., 1996). If estimates of angles are less susceptible to error than estimates of distance then animals would benefit from a predominant reliance on angular information for spatial memory. The ‘multiple bearing hypothesis’ (Kamil and Cheng, 2001) proposes such a mechanism to account for the extraordinary performance of Clark’s nutcrackers to relocate stored seeds which is hardly conceivable without multiple-bearing mechanisms. Additional experiments are needed to evaluate the bats’ ability to perceive and utilize angular relationships between multiple landmarks. Acknowledgements Two anonymous referees provided helpful comments. This study was supported by a grant from the Volkswagen Foundation. References Bhatt RS, Wright AA. Concept learning by monkeys with video picture images and a touch screen. J Exp Anal Behav 1992;57:219– 25. Biegler R, McGregor A, Krebs JR, Healy SD. A larger hippocampus is associated with longer-lasting spatial memory. Proc Natl Acad Sci USA 2001;98:6941–4. Burgess N, Jeffery KJ, O’Keefe J, editors. The hippocampal and parietal foundations of spatial cognition. Oxford: Oxford University Press; 1999. p. 490. Burr B, Rosen S, Barthlott W. Untersuchungen zur Ultraviolettreflexion von Angiospermen-Bl¨uten III. Akademie der Wissenschaften und der Literatur. Stuttgart: Mainz Franz Steiner Verlag; 1995 (p. 300). Bussey TJ, Warburton EC, Aggleton JP, Muir JL. Fornix lesions can facilitate acquisition of the transverse patterning task: A challenge for “configural” theories of hippocampal function. J Neurosci 1998;18:1622–31. Chase J. Visually guided escape responses of microchiropteran bats. Anim Behav 1981;29:708–13. Chase J. Differential responses to visual and acoustic cues during escape in the bat Anoura geoffroyi: cue preferences and behaviour. Anim Behav 1983;31:526–31. Chase J, Suthers RA. Visual obstacle avoidance by echolocating bats. Anim Behav 1969;17:201–7. Cheng K. Some psychophysics of the pigeon’s use of landmarks. J Comp Physiol A 1988;162:815–26. Cheng K. The vector sum model of pigeon landmark use. J Exp Psychol: Anim Behav Processes 1989;15:366–75. Cheng K, Spetch ML. Stimulus control in the use of landmarks by pigeons in a touch-screen task. J Exp Anal Behav 1995;63:187–201. Cheng K, Spetch ML. Mechanisms of landmark use in mammals and birds. In: Healy S, editor. Spatial representation in animals. Oxford University Press; 1998. p. 1–17. Collett TS, Cartwright BA, Smith BA. Landmark learning and visuospatial memories in gerbils. J Comp Physiol A 1986;158:835–51. Collett TS, Collett M. Memory use in insect visual navigation. Nature Rev Neurosci 2002;3:542–52.

Ekl¨of J, Jones G. Use of vision in prey detection by brown long-eared bats, Plecotus auritus. Anim Behav 2003;66:000–000 doi:10.1006/anbe.2003.2272. Ekl¨of J, Svensson AM, Rydell J. Northern bats, Eptesicus nilssonii, use vision but not flutter-detection when searching for prey in clutter. Oikos 2002;99:347–51. Healy S, editor. Spatial representation in animals. Oxford: Oxford University Press; 1998. Kamil AC, Cheng K. Way-finding and landmarks: the multiple-bearings hypothesis. J Exp Biol 2001;204:103–13. Kamil AC, Jones EJ. The seed-storing corvid Clark’s Nutcracker learns geometric relationship among landmarks. Nature 1997;390:276–9. MacDonald SE, Spetch ML, Kelly DM, Cheng K. Strategies in landmark use by children, adults, and marmoset monkeys, Learning Motivation 2004. Manske U, Schmidt U. Untersuchungen zur optischen Musterunterscheidung bei der Vampirfledermaus, Desmodus rotundus. Z Tierpsychol 1979;49:120–31. Markham MR, Butt AE, Dougher MJ. A computer touch-screen apparatus for training visual discriminations in rats. J Exp Anal Behav 1996;65:173–82. Morrison SK, Brown MF. The touch screen system in the pigeon laboratory: an initial evaluation of its utility. Behav Res Meth Instrum Comp 1990;22:123–6. Neuweiler G. The biology of bats. Oxford University Press; 2000 (p. 210–23). Redish DA. Beyond the Cognitive Map: From Place Cells to Episodic Memory. Cambridge: MIT Press; 1999. Rydell J, Ekl¨of J. Vision complements echolocation in an aerial-hawking bat. Naturwissenschaften 2003;90:481–3 doi:10.1007/s00114-0030464-x. Shettleworth SJ. Cognition. Evolution and behavior. New York, Oxford: Oxford University Press; 1998. Smith H. Light quality, photoperception and plant strategy. Annu Rev Plant Physiol 1982;33:481–518. Spetch ML. Overshadowing in landmark learning: touch-screen studies with pigeons and humans. J Exp Psychol: Anim Behav Proc 1995;21:166–81. Spetch ML, Cheng K, MacDonald SE. Learning the configuration of a landmark array. I. Touch-screen studies with pigeons and humans. J Comp Psychol 1996;110:55–68. Spetch ML, Cheng K, MacDonald SE, Linkenhoker BA, Kelly DM, Doerkson SR. Use of landmark configuration in pigeons and humans. II. Generality across search tasks. J Comp Psychol 1997;111:14– 24. Spetch ML, Cheng K, Mondloch MV. Landmark use by pigeons in a touch-screen spatial search task. Anim Learning Behav 1992;20:281–92. Stich KP, Denhardt G, Mauck B. Mental rotation of perspective stimuli in a California sea lion (Zalophus californianus). Brain Behav Evol 2003;61:102–12. Suthers RA. Biology of bats. New York, London: Academic Press; 1970. Suthers RA, Chase J, Braford B. Visual form discrimination by echolocating bats. Biol Bull 1969;137:535–46. Thiele J, Winter Y. Hierarchical strategy for relocating food targets in flower bats: spatial memory versus cue-directed search. Anim Behav, in press. Winter Y. Energetic cost of hovering flight in a nectar-feeding bat measured with fast-response respirometry. J Comp Physiol B 1998;168:434–44. Winter Y. Flight speed and body mass of nectar-feeding bats (Glossophaginae) during foraging. J Exp Biol 1999;202:1917–30. Winter Y, L´opez J, von Helversen O. Ultraviolet vision in a bat. Nature 2003;425:612–4. Wright AA, Cook RG, Rivera JJ, Sands SF, Delius JD. Concept learning by pigeons: Matching-to-sample with trial-unique video picture stimuli. Anim Learn Behav 1988;16:436–44.