Introduction

Foraging decisions of animals that feed on nectar are influenced by many physiological and environmental factors. Without doubt, the amount and quality of the nectar reward play a dominant role as one of these factors (Shettleworth 1998). Bumblebees, for example, readily return to patches of flowers with high nectar secretion rates, and thereby avoid poorer patches (Cartar 2004). Thus, the ability to perceive differences in available nectar volumes is an important prerequisite for foraging decisions of nectarivores. Underlying this must be a general ability to judge the quality of a flower by estimating nectar content (Thomson 1986; Garrison and Gass 1999; Robertson et al. 1999; Stout and Goulson 2002).

The ability of an animal to discriminate between differing physical stimuli (here nectar volume) is often approximated by the Weber–Fechner law, which states that the just noticeable difference between two physical stimuli increases in proportion to the intensity of the two single stimuli. One way to determine the point where two stimuli can be clearly distinguished from each other is fitting a psychometric function to discrimination data from an alternative forced choice paradigm. The psychometric function relates the psychological reaction of the animal to the physical stimulus intensity. Here, the threshold at the highest slope of the function denotes the stimulus intensity at which bats can clearly distinguish between two stimuli.

We determined the psychometric function for perceiving nectar volumes for the Neotropical flower-visiting bat, Glossophaga soricina. For this we determined choice preferences for differing volumes of sugar solution in a two-alternative forced-choice paradigm (2AFC), while offering volumes similar to those found under natural conditions (Winter and von Helversen 2001).

This is to our knowledge the first application of a standardised method to quantify the ability to distinguish between two volumes. It results in a quantitative, formal model of underlying perceptive abilities. Such data are also relevant in an ecological context as they permit the evaluation of decision-making during foraging in these bats.

Materials and methods

Animals

Six bats of the species G. soricina (Phyllostomidae) bred in captivity were used for this study. The climatic conditions both in the animal holding facility and in the experimental room were 22°C and approx. 60% relative humidity. The diet in the animal holding facility and during experiments consisted of some dry pollen and 17% honey water, to which Nektar Plus or Nutricomp were added. During experiments the amount of ad libitum honey water was reduced to 2 ml, which was given to the bats before the experimental procedure started. After the experiment had started the only food available to bats was artificial nectar from experimental food dispensers. This was a 17% sugar solution made from sucrose, glucose and fructose in equal parts as found in the nectar of bat visited flowers (Baker et al. 1998). Light conditions were LD 12:12 and all experiments with the echolocating bats were conducted during the dark phase.

Experimental apparatus

During the experiments bats were kept individually in cages (0.7 m × 2.2 m × 1.5 m). Inside the cages two feeders were installed on the back wall. Feeders had a cylindrical PVC opening equipped with a photoelectric barrier to automatically detect visiting bats. For a reward, a valve at the backside of the cage opened with an audible click and a syringe pump delivered a variable amount (see experimental protocol) of odourless nectar to the base of the feeder opening, which bats removed by licking (Winter and von Helversen 2001). Swivel arms mounted above each feeder allowed each feeder to be closed by moving a plastic flap in front of its opening. Details of the experimental apparatus are given in Winter and Stich (2005).

Pre-training

All bats received 2 days of training in their individual cages to accustom them to the experimental surroundings. We helped bats find and use the feeders by applying a drop of honey to the tip of each feeder on the first pre-training day. All bats used in this experiment quickly found the feeders and visited them regularly. During pre-training, feeders delivered 30 μl of sugar water solution on each visit to a feeder. On the second night bats had to visit the feeders in alternation to prevent them from developing a spatial preference to a single feeder. This was done by automatically moving a flap in front of the just visited feeder and simultaneously opening the other feeder by means of rotating the swivel arms.

Experimental procedure

We tested the bats in a modified two-alternative forced-choice paradigm (2AFC). The two feeders in the cage provided the bat with differing amounts of nectar solution. One trial consisted of 50 visits to the feeders and was divided into two phases. In the first phase, the sample phase, bats had to visit feeders in alternation for 20 visits (10 visits to each feeder). In this phase the flaps automatically moved in front of a feeder to ensure that no unrewarded visits to a feeder could occur. In the second phase, the choice phase, bats could choose freely between the two different amounts (feeders). This phase lasted for thirty visits. Trials with different pairings of two amounts followed in direct succession.

We presented bats with a total of 8 combinations of different volumes (Table 1).

Table 1 Sugar solution volumes presented to bats at two feeders
Full size table

Stimulus intensities were calculated by dividing the difference between the two volumes by the average of the two volumes. Preliminary experiments had suggested that the threshold for discrimination should be somewhere below 1, thus we used mainly values below 1, and only two values above 1. Each combination of volumes was offered twice, such that during one trial the higher volume was at the right position within the cage, while during the other trial it was on the left to correct for spatial biases of the bats. For analyses we first calculated the average from the two presentations of the same stimulus pairs. Experiments complied with national laws on animal care.

Psychometric function

For the estimation of the psychometric function, we analysed the data from all six bats separately and applied the algorithm proposed by Kuss et al. (2005). We fitted a Weibull psychometric function to the data following the detailed instructions in Kuss et al. (2005). This recently developed algorithm estimates the psychometric function and the three important parameters of this function with their corresponding confidence intervals. The first parameter (threshold parameter) is the point at which the subject can distinguish between the two stimuli. This is the point on the psychometric function with the steepest slope. The second parameter is the slope at this point, which is a measure for the reliability of sensory performance (Treutwein and Strasburger 1999). The third parameter is the lapse rate, which is inferred from the difference between perfect performance and the actual behaviour of animals at high stimulus intensities. It serves as a measure for the errors that are not of perceptional nature but are made due to lapses in attention or motivational problems. Markov Chain Monte Carlo (MCMC) sampling is applied for an estimation of these parameters. To find the parameter estimates in this Bayesian approach, the investigator has to state his/her prior beliefs about the parameter location in form of prior distributions. As prior functions we chose a beta distribution (2;50) for the lapse rate, normally distributed priors for the threshold, and the slope with a mean of 1 and a standard deviation of 0.5 and a log-normal prior for the slope with a mean of 2 and a standard deviation of 1. For the MCMC sampling we performed 5,000 runs with 100 leapfrog steps each. The leapfrog step size for the three parameters were estimated by the heuristic described by Kuss et al. (2005). From this we received psychometric functions for each individual. We calculated the mean and standard deviation for the three parameters across animals. We reran the process, but this time we pooled the data of all six animals by summing up the choices for the high volume feeder across all animals. We received one psychometric function for all animals, which is given in Fig. 1. This was done for visualisation purposes only. And we want to point out that, even though the threshold parameter of the pooled function differs only slightly from the mean threshold parameter of the non-pooled data, the means across subjects of the three parameters of the individual psychometric functions represent the proper estimates.

Fig. 1
figure 1

Reaction of bats to different stimulus intensities. Abscissa gives the stimulus intensities from Table 1. The circles denote the median across the six subjects of the proportion of choices for the feeder with the higher volume. Error bars give median absolute deviation calculated across the six subjects. The line shows the psychometric function fitted to the pooled data. The box shows the threshold of the pooled data (here at 0.63) with a 95% confidence interval

Full size image

Results

Individual bats completed 16 trials of 8 different volume pairs (Table 1) within two experimental nights. With 30 free choices per trial each bat made a total of 480 choices. The three parameters of the resulting psychometric functions for each bat are given in Table 2. The threshold, which gives the stimulus intensity at which bats could clearly discriminate between the two stimuli, ranged from 0.35 to 1.04 with a mean of 0.69 ± SD 0.21. The slope ranged from 0.27 to 3.08. The lapse rate was given with a mean of 0.05 ± SD 0.02. When pooling all bats, they showed an overall increasing preference for the larger volume as the stimulus intensity increased (Table 1, Fig. 1).

Table 2 Threshold, slope, and lapse parameters for the psychometric functions of each individual
Full size table

To examine whether the Weber–Fechner law was preserved within the investigated range of volumes, we compared the reactions at 0.4 stimulus intensity of two different volume combinations, one combination of low volumes (6 and 9 μl) and one combination of high volumes (12 and 18 μl). There was no significant difference between the reaction strengths to the two combinations (Table 1, paired Wilcox test: W = 30.5; N = 6; P = 0.48).

Discussion

We used a 2AFC paradigm to determine psychometric functions that describe the ability for fluid volume discrimination. Such an approach, to our knowledge, has not been used before for this purpose.

One of the assumptions underlying our paradigm for estimating a psychometric function is the preservation of the Weber–Fechner law. When changing stimulus intensities such that the relative difference is kept constant there should be no difference in the reaction to the two stimulus pairs. In our experiments, this relation was preserved as the reaction of the bats to the pairs 6, 9 μl and 12, 18 μl of sugar water solution did not differ significantly (both pairs had a stimulus intensity of 0.4). Thus, over the range of intensities examined, bats behaved in accordance with Weber–Fechner.

Estimates for threshold and slope could have been influenced by the motivation of the bats to discriminate between the two stimuli. The motivation, or more general, the non-perceptional error made by the bats was inferred from the estimate for the lapse rate. The lapse rate in this experiment was relatively low, since only an estimated 5% of the visits did not mirror a decision that was influenced solely by the difference between the two stimuli. Bats had only ten sampling visits to each feeder, which shows that bats quickly adopted to the new stimulus pairs. Thus, this paradigm seems well suited to assess perceptional constraints, which influence decision-making in animals (Dukas 2004).

The estimated threshold value in the experiments was 0.69. The following calculation serves to illustrate which nectar volume a bat can discriminate to be larger than a previously consumed 3 μl amount (see also calculation of stimulus intensities in Materials and methods section):

$$ \frac{{x - 3\,{\text{ $ \mu $ l}}}} {{\frac{{x + 3\,{\text{ $ \mu $ l}}}} {{\text{2}}}}} = 0.69. $$

This results in an amount of at least 6.2 μl required at the second flower in order for the bat to clearly discriminate between the two flowers.

There was high variation between individuals in their choice behaviour at certain stimulus intensities (see error bars in Fig. 1). One reason for this, besides the above mentioned lapses, were the different thresholds estimated for individual animals. While animal 1 did not discriminate between between 9 and 15 μl, animal 4 was able to do so. Error bars were reduced when all animals could easily distinguish stimuli above intensities of 1. However, four animals had very similar thresholds from 0.5 to 0.8 so that the threshold mean of 0.69 given in this study may be a conservative estimate of the discrimination ability of G. soricina for volumes.

The ability to discriminate nectar volumes may be influenced by differences in sugar composition and concentration as both parameters may influence the sensory pathway. Nonetheless, the value for the threshold for bats seems to be lower than for honeybees (Apis mellifera). Bees could not discriminate between 0.4 and 1.2 μl of a 1.5 M (34% w/w) sucrose solution, but could discriminate between 0.4 and 0 μl (34% sugar solution w/w, Shafir et al. 2005). This equals a Weber coefficient of 2 for the bees, a value far higher than the estimated 0.69 for bats. One implication from these results is that future behavioural choice experiments with G. soricina that involve different nectar volumes should adopt their design to the boundaries of the sensory abilities of this species. Moreover, the results will also help in evaluating foraging decisions in this bat species.