Processing of transient signals in the visual system of the European starling (Sturnus vulgaris) and humans

Vision Research 51 (2011) 21–25

Contents lists available at ScienceDirect

Vision Research journal homepage: www.elsevier.com/locate/visres

Processing of transient signals in the visual system of the European starling (Sturnus vulgaris) and humans Arne Feinkohl, Georg Klump ⇑ Department for Animal Physiology & Behaviour, C.v.O University Oldenburg, PO Box 2503, D-26111 Oldenburg, Germany

a r t i c l e

i n f o

Article history: Received 11 January 2010 Received in revised form 27 July 2010

Keywords: Double-pulse resolution Visual temporal processing Bird vision

a b s t r a c t The double-pulse resolution (DPR) measures the processing performance for transient visual signals as the threshold duration for detecting a temporal gap between two light flashes in relation to gap duration. The DPR of four European starlings (Sturnus vulgaris) and four humans was measured in an operant Go/ NoGo procedure. We applied the method of constant stimuli and determined thresholds using signaldetection theory. The starling DPR (22.2 ms ± 2.3 ms SE) was significantly shorter than human DPR (35.2 ms ± 1.3 ms SE; p < 0.01, t-test). The difference suggests that starlings have a higher temporal resolution for transient visual signals than humans. Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction

2. Materials and methods

In order to determine the visual temporal resolution of an animal, two different measurement methods must be distinguished: The Critical flicker-fusion frequency (CFF) represents the temporal resolving power for ongoing visual signals. The double-pulse resolution (DPR), on the other hand, depicts the performance of an animal’s visual system for processing transient stimuli such as suddenly appearing obstacles encountered while moving through the natural environment. It measures the detection threshold of a temporal gap between two flashes of light. In this study we determined the DPR of a songbird species in a behavioural experiment in which the subjects’ task was to detect double-flashes (DF) of light in a repeated single-flash (SF) background. Since the DPR of humans depends on factors like flash duration (Mahneke, 1958; Treutwein & Rentschler, 1992), the retinal area of stimulation (Lotze, Treutwein, & Roenneberg, 2000; Poggel & Strasburger, 2004; Treutwein & Rentschler, 1992), the age (Poggel & Strasburger, 2004) and the circadian rhythm (Lotze et al., 2000), the human thresholds in the literature may not have been obtained under comparable conditions to our experiments measuring the starling’s DPR threshold. To compare the results with those of human observers, it was necessary to measure human thresholds under comparable conditions. In relation to findings on the CFF of birds and humans, our hypothesis was that starlings have a better visual temporal resolution for transient stimuli.

2.1. Starlings’ DPR determination

⇑ Corresponding author. Address: Carl von Ossietzky Universität Oldenburg, Fak V, IBU, Department for Animal Physiology & Behaviour, Carl von Ossietzky Str. 9-11, PO Box 2503, D-26111 Oldenburg, Germany. Fax: +49 (0)441 798 5615. E-mail address: [email protected] (G. Klump). 0042-6989/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.visres.2010.09.020

2.1.1. Subjects Four adult, wild-caught and experimentally naïve European Starlings (Sturnus vulgaris; 2 males, le and ma, and 2 females, wi and no) participated in this study. Their ages ranged from at least 2 years (no, ma) and at least 10 years (wi) to at least 13 years (le). They were kept in individual cages (0.8  0.4  0.4 m) with water ad libitum. Daylight lit the room, and in addition it was artificially illuminated for 15 h per day. The food consisted mainly of duck-food pellets (Agravis Geflügelfinisher), and was supplemented by lettuce and oat flakes. Additionally, the birds were occasionally provided with cod liver oil and vitamins. Food was restricted, during the experiments the birds were kept at 91–95% of their free-feeding weight. They could feed in the evening after the test sessions, and we weighed them daily. At least once a week we provided them with the possibility to take a bath. A wire cage (0.2  0.2  0.3 m) which the birds entered voluntarily allowed us to transport the birds from their cages to a soundproof test chamber that was located in the neighbouring room. The care and treatment of the birds were in accordance with the procedures of animal experimentation approved by the Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, Germany. 2.1.2. Setup The experiments took place in a custom-built soundproof chamber (inside dimensions: 0.9  0.5  0.7 m). Inside the chamber was a wire cage (0.3  0.2  0.3 m) with two perches and an opening (0.2  0.1 m) on the front in which a black cardboard box presenting the stimuli was mounted (see Fig. 1). The rear perch

22

A. Feinkohl, G. Klump / Vision Research 51 (2011) 21–25

Fig. 2. Properties of the flash stimuli for the starlings’ DPR determination: (a) reference signal: single-flash. Each flash varied in duration between 11 and 70 ms. (b) Test signal: double-flash. Each of the two flashes had a duration of 5 ms. The seven different gap durations varied between 1 and 60 ms.

Fig. 1. Main components inside the soundproof chamber: (a) wire cage, (b) waiting perch, (c) light barrier, (d) reward perch, (e) automatic feeder, and (f) cardboard box.

was equipped with a light barrier (Conrad Electronic, Germany) indicating when the bird sat on that perch. The front perch permitted access to a custom-built automatic feeder which contained 48 mealworm halves (Tenebrio molitor), i.e., the primary reinforcer delivered if the birds responded correctly to the given task (see below). Two LEDs (Lumitronix Highpower LED Spot, Germany; colour temperature: 6300 K) lit the chamber from the rear of the wire cage. Each LED contained a diffuser lens (Lumitronix), and the illuminance of the chamber reached approx. 340 lux at the point where a bird’s head would be during the experiment. An additional reward light, an LED spot (Paulmann, Germany, High efficiency LED) on the front of the wire cage provided a secondary reinforcer. A camera module (Conrad Electronic) connected to an external monitor (Santec, Germany, 1200 B/W CCTV Monitor) allowed to observe the bird’s behaviour during the experiment. The custom-built black cardboard box produced calibrated flash stimuli. On the front it had a round diffuser patch (Ø 2 cm) that was illuminated by nine LEDs (Dotlight LCFW520140; colour temperature: approx. 10,000 K). During the experiments, the cardboard box was at eye level of the starlings with a viewing distance of 20 cm. A Linux-operated computer controlled the experiment operating a Parallel Interface (PI2, Tucker Davis Technologies, USA) and an Enhanced Real-Time Processor (RP2, Tucker Davis Technologies). The latter generated the flash signals, which drove the cardboard box via a custom-built power amplifier. 2.1.3. Stimuli The stimuli consisted of DF and SF signals (see Fig. 2) and the birds were trained to discriminate between them. A DF consisted of two flashes with a duration of 5 ms each, and a temporal gap with a predefined gap duration (GD). The GD had values of 1.0, 3.0, 11.6, 20.6, 35.4 or 60.0 ms, chosen at random. The duration of the SF stimuli varied between 11 and 70 ms, and the range of durations of the SF stimuli was equal to the range of the total duration of the DF including the gap. This way, the total stimulus duration could not be used as a cue for discriminating the SF and the DF (see Kietzman & Sutton, 1968). The luminance of the DF had random values of 200, 350, 500, 650 or 800 cd/m2. The two flashes comprising a DF had the same luminance. Bloch’s law (Bloch, 1885) states that the detectability of a stimulus depends on the product of its energy and its duration. DF and SF luminance varied independently from each other between the stimuli. In order to rule out the flash luminance as a discrimination cue of DF from SF, the SF luminance of the stimuli was uniformly distributed between 50 and 800 cd/m2.

2.1.4. Procedure We trained the subjects to discriminate DF from SF in a Go/ NoGo paradigm, and the DF were presented with different GD according to the method of constant stimuli. The birds were trained to sit on the rear perch in order to start a trial. In test trials, the SF repeated every 1.5 s and was replaced by a single DF 4–10 s after the beginning of a trial. During the report phase, a period of 2 s starting with the DF, the birds were to jump off the perch onto the front perch in order to report the perception of the DF. Such Hits led to a reward in 80% of all cases, and always the reward light lit for 5 s. The Miss of a DF (i.e. the bird remaining on the rear perch) initiated the beginning of the next trial. If the birds jumped off the perch although no DF was presented the chamber light was switched off for 5 s and the next trial started afterwards. Sham trials in which the test stimulus was a SF provided an indicator for the false-alarm rate. If the birds jumped off the perch during the report phase in such sham trials, the chamber light was switched off for 5 s as well. A session began with 10 warm-up trials that comprised DF stimuli with the longest GD, i.e., the test stimuli that the birds could presumably detect the easiest. Those first 10 trials were not included in the data analysis. The following 100 test trials consisted of two blocks with 50 trials each. In each block any combination of the seven GD and five luminances was tested once, and sham trials represented the remaining 15 trials. The sequence of all trials in a block was completely randomised. The subjects could complete a session in approx. 30 min. Each starling conducted up to four sessions per day, while there was a minimum break of 1 h between the end of the last and the beginning of the next session. 2.1.5. Data analysis In order to assure an accurate threshold we applied specific criteria to decide whether a session was valid and hence included in the data analysis. It was included if the subjects responded to no more than 10% of the sham trials and to at least 70% of the DF with the two longest GD. If a session had to be rejected, it was repeated. Each bird had to complete a total of 10 valid sessions, i.e., each bird’s psychometric function combined the data from 1000 trials. The sensitivity measure d0 (e.g. see Green & Swets, 1966) indicated each subject’s discrimination performance between SF and DF, and this value was calculated separately for each GD. The d0 criterion for the DPR threshold was 1.8, which corresponds to a hit rate of 56% at a false-alarm rate of 5%. We linearly interpolated the d0 values of neighbouring GD to determine the gap duration threshold. In addition to the d0 values, psychometric functions were calculated by analyzing the birds’ hit rates for each GD according to Lam, Mills, and Dubno (1996). 2.2. Human DPR determination In addition to the starling DPR experiment, we conducted an experiment on the human DPR providing similar conditions in order to compare the performance of both species. Only the differences between the starling and the human DPR determination will be

23

A. Feinkohl, G. Klump / Vision Research 51 (2011) 21–25

depicted here. Four subjects participated in the experiment, two men and two women. Their age ranged from 25 to 29 years, and all had normal vision. The experiments were undertaken with the understanding and written consent of each subject, following the Code of Ethics of the World Medical Association (Declaration of Helsinki). The experiment took place in a soundproof chamber (Industrial Acoustics Company GmbH Mini 250; 195  73  99 cm) designed for studies in humans. The chamber light LEDs hung above a chair on which the subjects were seated, and the illuminance in the chamber reached approx. 260 lux. The LED spot that lit up during reinforcements was located in the front right corner of the booth away from the cardboard box. The cardboard box was attached to the chamber door at eye level of the subjects, with a viewing distance of 47 cm. All stimulus attributes were equal to those used in determining the gap duration threshold in the starlings, except for the GDs of the DF. These comprised 25.0, 30.8, 36.7, 42.5, 48.3, 54.2 and 60.0 ms. A pushbutton enabled the subjects to report the perception of a DF, and each subject had to complete 4 valid sessions with 100 trials each (not including the 10 warm-up trials). 3. Results 3.1. Starling DPR

3.2. Human DPR In the human DPR threshold determination, all the four sessions per subject were valid, the mean false-alarm rate was 1.7%. Fig. 4 shows the d0 values of the four subjects for all seven GDs of the DF stimuli, the horizontal line represents the d0 threshold of 1.8. The mean DPR threshold of the four human subjects was 35.2 ms (±1.3 ms SE). As for the starling experiment, we evaluated the data of the human threshold determination in a General Linear Mixed Model ANOVA. The probability of response was the dependent variable, and subjects were included as a random effect. GD was treated as a fixed main effect in the analysis, and DF luminance represented a covariate. The probability of response did not differ between the subjects (p > 0.05, GLMM ANOVA). Not only did the GD influence the human response rate (p < 0.001, F = 20.5, GLMM ANOVA), the subjects’ probability of response decreased linearly with the DF luminance (p < 0.001, F = 18.1, GLMM ANOVA). Furthermore, the interaction of GD and DF luminance on the subjects’ probability of response was significant (p = 0.002, F = 3.7, GLMM ANOVA). 3.2.1. Comparison of starlings’ results and human subjects’ results The four starlings’ DPR thresholds differed significantly from the DPR thresholds of the four human subjects (p = 0.005, t-test for independent samples). Fig. 5 shows the response probabilities of humans and starlings with fitted sigmoidal psychometric functions; see Lam et al. (1996). The corresponding d0 values have been illustrated in Figs. 3 and 4. 4. Discussion In this study, we investigated the temporal processing of transient visual signals of a songbird species, the European starling, in a behavioural experiment measuring the DPR. Using a comparable setup, we additionally investigated the DPR of human subjects. We suggest that starlings possess a better visual temporal resolution for transient signals than humans since starlings manoeuvre

d'

On average, each bird carried out 25 sessions to reach a threshold. The major reason for invalid sessions was an insufficient amount of Hits for the two longest GD (61.0%). The mean falsealarm rate of the valid sessions was 4.8%. Fig. 3 shows the d0 values of the four birds for all seven GD of the DF stimuli, the horizontal line represents the d0 threshold of 1.8. The mean DPR threshold of the four starlings was 22.2 ms (±2.3 ms SE). We evaluated the data in a General Linear Mixed Model ANOVA. The birds’ probability of response was the dependent variable, and the birds were included in the analysis as a random effect. GD was treated as a fixed main effect in the analysis, and DF luminance represented a covariate. The probability of response did not differ between the birds (p > 0.05, GLMM ANOVA). Not only the GD influenced the birds’ probability of response (p < 0.001, F = 38.9, GLMM ANOVA), there was also a linear relationship between their probability of response and the DF luminance (p < 0.001, F = 27.7, GLMM ANOVA). The linear decrease in the starlings’ probability of response with an increase in DF luminance was present even for sub-threshold GDs. In addition, the interaction of GD and luminance on the birds’ probability of response was significant (p = 0.002, F = 3.6, GLMM ANOVA).

The luminance of the SF was uniformly distributed between 50 and 800 cd/m2. The luminance of the SF in the sham trials to which the birds responded differed significantly from a uniform distribution (p < 0.001, Kolmorgorov–Smirnov-test), though. The starlings responded significantly more often to SF with luminances between 50 and 375 cd/m2 in comparison to SF with luminances between 375 and 800 cd/m2 (p = 0.014, Binomial test).

le wi no ma d' = 1.8

Gap duration of the double-flash [ms] 0

Fig. 3. d Values in relation to the seven gap durations of the double-flash, shown separately for each starling. Each gap duration was tested 100 times. The curves show the linearly interpolated neighboured data points for each bird. The horizontal line represents the threshold criterion of d0 equalling 1.8.

A. Feinkohl, G. Klump / Vision Research 51 (2011) 21–25

d'

24

ah fs ss af d' = 1.8

Gap duration of the double-flash [ms] Fig. 4. d0 Values in relation to the seven gap durations of the double-flash, shown separately for each human subject. Each gap duration was tested 40 times. The curves show the linearly interpolated neighboured data points for each subject. The horizontal line represents the d0 threshold criterion of 1.8.

Probability of response

1

0.8

0.6

0.4 Starlings, n=4

0.2

Humans, n=4

0 0

10

20

30

40

50

60

70

Gap duration of the double-flash [ms] Fig. 5. Mean probability of response of starlings and humans for the different gap durations. The x-axis shows the gap duration of the double-flash, the y-axis represents the subjects’ mean probability of response. The bars of the data points represent their standard errors. The curves show the psychometric functions of the starlings and the human subjects with a sigmoidal fit; see Lam et al. (1996). The adjusted R2 is 0.998 and 0.997, respectively.

faster through their environment. In line with this hypothesis we demonstrate a better DPR in starlings than in humans. A study reviewing video image stimulation in animal behavioural experiments (D’Eath, 1998) suggested a similar relationship for an animal’s CFF and its maximal propagation speed through the natural environment. The CFF of humans equals approximately 50 Hz for bright stimuli (Hecht & Verrijp, 1933). Studies on the CFF of bird species other than the European starling (Ginsburg & Nilsson, 1971; Jarvis, Taylor, Prescott, Meeks, & Wathes, 2002; Nuboer, Coemans, & Vos, 1992; Powell, 1967) show that the maximal avian CFF ranges from 71.5 Hz in the chicken (Gallus gallus domesticus) to approximately 145 Hz in the pigeon (Columba livia) which indicates that birds tend to have a better CFF than humans. A study by Greenwood et al. (2004) provided indirect evidence that the starling’s CFF may be higher than 100 Hz. They found that starlings favoured surroundings that were illuminated with the very high modulation rate of >30 kHz over those that were illuminated with lights flickering with 100 Hz (although a similar effect has been described for humans favouring 32 kHz over 100 Hz environments, see Wilkins, Nimmo-Smith, Slater, & Bedocs, 1989). Since the human DPR threshold depends on flash duration (Mahneke, 1958; Treutwein & Rentschler, 1992), we will compare the results from our experiment on the human DPR only to studies

with similar flash durations and comparable methods. Mahneke (1958) investigated the monocular human DPR with many combinations of different flash durations at a luminance of 850 cd/m2. He used the method of limits to determine the threshold. If both flashes in a DF stimulus had durations of 5 ms the DPR threshold was 42.3 ms which is slightly larger than the human DPR observed in the present study. The small difference can be explained by dissimilarities of the visual stimuli or by individual factors, which have both been shown to affect the CFF (Landis, 1954). The starlings’ response probability linearly decreased with an increase in DF luminance not only for GDs in the threshold range, but also for sub-threshold GDs. This suggests that the starlings in general responded more to flash stimuli with low luminance. The analysis of the SF luminance in the sham trials to which the birds responded confirmed this assumption: The starlings responded more often to SF stimuli with low luminance than to SF stimuli with high luminance. It is unlikely, however, that this change in the response probability with luminance may have affected the DPR. Firstly, the signal-detection thresholds are based on both the responses in sham trials and the responses in test trials. Secondly, DPR thresholds were determined averaging the responses obtained with the total range of luminance values. The lack of an effect of the difference in stimulus luminance renders it also unli-

A. Feinkohl, G. Klump / Vision Research 51 (2011) 21–25

kely that the difference in viewing distance and ambient illumination has a large effect on the difference in DPR between humans and starlings studied under relatively similar environmental conditions. We would like to note, however, that the anatomical differences in eye structure related to pupil size and focal length in addition to differences in the physical illumination provided by the experimental setup lead to large differences in retinal illumination that can be estimated to be more than an order of magnitude larger in humans than in starlings (see Martin (1982, 1986) for anatomical data). Starlings feature retinal asymmetries in that the nasal parts of their eyes are shorter than the temporal parts (Martin, 1986) and the photoreceptor distribution is unequal between the left and the right eye (Hart, Partridge, & Cuthill, 2000). This may lead to a different perception of the stimulus depending on the starling’s angle of view and the involved eyes. The Ferry-Porter law (Ferry, 1892; Porter, 1902) states that the human CFF improves proportional to the logarithm of the stimulus luminance. In our experiments on the DPR of human subjects, an increase in flash luminance led to a linear decrease in the probability of the response when the GDs were in DPR threshold range. This suggests that for transient visual stimuli an increase in stimulus energy does not necessarily improve the processing, but it can even have a deteriorating effect on the performance. Other studies on this topic did not find an effect (Kietzman, 1967; Lotze et al., 2000; Venables, 1963) or presented evidence for an improvement of the double-pulse resolution (Lewis, 1967). Short glances at predators or suddenly appearing obstacles should provide sufficient visual information for deciding about the immediately taken actions. In summary, by determining the DPR of both species we could show that starlings possess a better visual temporal resolution for such transient processing tasks than humans, a finding that is line with the results of studies on the CFF of bird species and humans. Acknowledgments We thank Holger Dierker for technical support during the preparation of the experiments. The study was partly funded by the Deutsche Forschungsgemeinschaft (TRR 31). References Bloch, A. M. (1885). Expérience sur la vision. Comptes rendus de séances de la Société de Biologie (Paris), 37, 493–495.

25

D’Eath, R. B. (1998). Can video images imitate real stimuli in animal behaviour experiments? Biological Reviews, 73(3), 267–292. Ferry, E. S. (1892). Persistence of vision. American Journal of Science, 44, S. 192–S. 207. Ginsburg, N., & Nilsson, V. (1971). Measuring flicker thresholds in the budgerigar. Journal of the Experimental Analysis of Behavior, 15(2), 189–192. Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics. New York, London, Sydney: John Wiley & Sons. Greenwood, V. J., Smith, E. L., Goldsmith, A. R., Cuthill, I. C., Crisp, L. H., Walter-Swan, M. B., et al. (2004). Does the flicker frequency of fluorescent lighting affect the welfare of captive European starlings? Applied Animal Behaviour Science, 86(2), 145–159. Hart, N. S., Partridge, J. S., & Cuthill, I. C. (2000). Retinal asymmetry in birds. Current Biology(10), 115–117. Hecht, S., & Verrijp, C. D. (1933). Intermittent stimulation by light III. The relation between intensity and critical fusion frequency for different retinal locations. The Journal of General Physiology, 17, 251–268. Jarvis, J. R., Taylor, N. R., Prescott, N. B., Meeks, I., & Wathes, C. M. (2002). Measuring and modelling the photopic flicker sensitivity of the chicken (Gallus g. domesticus). Vision Research, 42(1), 99–106. Kietzman, M. L. (1967). Two-pulse measures of temporal resolution as a function of stimulus energy. Journal of the Optical Society of America, 57(6), 809–813. Kietzman, M. L., & Sutton, S. (1968). The interpretation of two-pulse measures of temporal resolution in vision. Vision Research, 8(3), 287–302. Lam, C. F., Mills, J. H., & Dubno, J. R. (1996). Placement of observations for the efficient estimation of a psychometric function. The Journal of the Acoustical Society of America, 99(6), 3689–3693. Landis, C. (1954). Determinants of the critical flicker-fusion threshold. Physiological Reviews, 34(2), 259–286. Lewis, M. F. (1967). Two-flash thresholds as a function of luminance in the darkadapted eye. Journal of the Optical Society of America, 57(6), 814–815. Lotze, M., Treutwein, B., & Roenneberg, T. (2000). Daily rhythm of vigilance assessed by temporal resolution of the visual system. Vision Research, 40(25), 3467–3473. Mahneke, A. (1958). Foveal discrimination measured with two successive light flashes: A psychophysical study. Acta Ophthalmologica, 36(1), 3–11. Martin, G. R. (1982). An owl’s eye: Schematic optics and visual performance in Strix aluco L. Journal of Comparative Physiology A, 145(3), 341–349. Martin, G. R. (1986). The eye of a passeriform bird, the European starling (Sturnus vulgaris): Eye movement amplitude, visual fields and schematic optics. Journal of Comarative Physiology A, 159(4), 545–557. Nuboer, J. F., Coemans, M. A., & Vos, J. J. (1992). Artificial lighting in poultry houses: Do hens perceive the modulation of fluorescent lamps as flicker? British Poultry Science, 33(1), 123–133. Poggel, D. A., & Strasburger, H. (2004). Visual perception in space and time – mapping the visual field of temporal resolution. Acta Neurobiologiae Experimentalis, 64(3), 427–437. Porter, T. C. (1902). Contributions to the study of flicker: II. Proceedings of the Royal Society of London, 70(463), 313–329. Powell, R. W. (1967). The pulse-to-cycle fraction as a determinant of critical flicker fusion in the pigeon. The Psychological Record, 17, 151–160. Treutwein, B., & Rentschler, I. (1992). Double-pulse resolution in the visual field: The influence of temporal stimulus characteristics. Clinical Vision Science, 7(5), 421–434. Venables, P. H. (1963). The relationship between level of skin potential and fusion of paired light flashes in schizophrenic and normal subjects. Journal of Psychiatric Research, 1(4), 279–287. Wilkins, A. J., Nimmo-Smith, I., Slater, A. I., & Bedocs, L. (1989). Fluorescent lighting, headaches and eyestrain. Lighting Research and Technology, 21(1), 11–18.