Sound Source Segregation in the Acoustic Parasitiod Fly Ormia Ochracea

Sound Source Segregation in the Acoustic Parasitiod Fly Ormia ochracea

Norman Lee

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Cell and Systems Biology University of Toronto

Sound Source Segregation in the Acoustic Parasitoid Fly Ormia ochracea

Norman Lee

Doctor of Philosophy

Cell and Systems Biology University of Toronto

2012 Abstract

Sound source localization depends on the auditory system to identify, recognize, and segregate elements of salient sources over distracting noise. My research investigates sensory mechanisms involved in these auditory processing tasks of an insect hearing specialist, to isolate individual sound sources of interest over noise. I first developed quantitative methods to determine signal features that the acoustic parasitoid fly Ormia ochracea (Diptera: Tachinidae) evaluate for host cricket song recognition. With flies subjected to a no-choice paradigm and forced to track a switch in the broadcast location of test songs, I describe several response features (distance, steering velocity, and angular orientation) that vary with song pulse rate preferences. I incorporate these response measures in a phonotaxis performance index that is sensitive to capturing response variation that may underlie song recognition. I demonstrate that Floridian O. ochracea exhibit phonotaxis to a combination of pulse durations and interpulse intervals that combine to a range of accepted pulse periods. Under complex acoustic conditions of multiple coherent cricket songs that overlap in time and space, O. ochracea may experience a phantom source illusion and localize a direction between actual source locations. By varying the temporal overlap between competing sources, I demonstrate that O. ochracea are able to resolve this illusion via the precedence effect: exploitation of small time differences between competing sources to selectively localize the leading over lagging sources. An increase in spatial separation between cricket song and masking noise does not reduce song detection thresholds nor improve song localization accuracy. Instead, walking responses are diverted away from both song and noise. My findings support the idea that the ears of O. ochracea function as bilateral symmetry

ii detectors to balance sound intensity, sound arrive time differences, and temporal pattern input to both sides of the auditory system. Asymmetric acoustic input result in corrective turning behaviour to re-establish balance for successful source localization.

iii

Acknowledgments

I owe my deepest gratitude to my supervisor Dr. Andrew Mason for his contributions, and endless support in nurturing me as an independent scientist. Andrew provided me with the opportunity and resources to explore my own research interests independently, but was always available to provide invaluable input critical to the success of this dissertation. I would like to thank my thesis advisory committee members Drs. Maydianne Andrade and John Peever for their experimental advice and for pointing out my weaknesses so that I may grow as a scientist. I thank Andrew and Maydianne for being sources of inspiration that has sparked my passion for pursuing academia. I am grateful for Dr. Thomas Walker who has provided me with support for field experiments and fly collection. I would also like to thank Drs. Damian Elias, Michael Kasumovic, Mark Fitzpatrick, Kevin Judge, Jeff Stoltz, Patrick Guerra, Paul De Luca, Daniel Howard, Carrie Hall, and Fernando Montealegre for helpful discussions and feedback on projects. I thank my fellow graduate students Dean Koucoulas, Jenn Van Eindhoven, Matt Jackson, Sen Sivalinghem, Maria Modanu, Emily MacLeod and Luciana Baruffaldi for their support. I thank Michelle Leung and numerous Mason Lab volunteers that have contributed to maintaining a stable colony of Ormia ochracea for this research endeavor to be possible. I am grateful for the support from my closest friends that include: David Pham, Gary Yan, Fang Zhao, Paul Pan, Steven Wong, Takeshi Ishii, Jeongkyo Jang, Eunji In, and Min Ok Choi. I would like to show my sincere gratitude to John Ko and Phoebe Choe for their love and generosity in providing me with a place to stay during my transition to work in the laboratory of Dr. Mark Bee at the University of Minnesota. I thank John and Phoebe for their provisions that have assisted me in the final stages of completing my dissertation. I would also like to thank my mother Tung Moy Lee, and my father Chee Wing Lee, for teaching me to be diligent at an early age. I am indebted to my wife Mijung Kim and our ‘Somang’ for their love, encouragement, and sacrifices that they have made to provide me with the opportunity to pursue my dreams. Lastly, I would like to thank my mother-in-law Heo Nam Soon and our family in Busan, South Korea for believing in me and entrusting their precious Mijung to my care.

Table of Contents

Contents

Abstract ...... ii

Acknowledgments...... iv

Table of Contents ...... v

List of Tables ...... ix

List of Figures ...... x

Chapter 1 General Introduction ...... 1

1.1 Evaluating Features of Acoustic Signals ...... 1

1.2 Spectral Features of Acoustic Signals Evaluated for Species Recognition ...... 1

1.3 Temporal Features of Acoustic Signals Evaluated for Species Recognition ...... 2

1.4 Directional Hearing in Small Insects ...... 3

1.5 Hearing in Complex Acoustic Conditions ...... 4

1.6 Acoustically Orienting Parasitoid Flies (Diptera: Tachinidae) ...... 5

1.7 Directional Hearing in O. ochracea ...... 6

1.8 Thesis Outline ...... 8

References ...... 10

Chapter 2 Deriving a Sensitive Measure of Walking Phonotaxis Performance in the Acoustic Parasitoid Fly Ormia ochracea ...... 14

2.1 Abstract ...... 14

2.2 Introduction ...... 15

2.3 Materials and Methods ...... 18

2.3.1 Animals ...... 18

2.3.2 Acoustic Stimuli...... 18

2.3.3 Spherical treadmill ...... 18

2.3.4 Protocol ...... 19

2.3.5 Data Analysis ...... 19

2.4 Results ...... 20

2.5 Discussion ...... 23

2.6 Acknowledgements ...... 27

References ...... 28

Figure Legends...... 31

Chapter 3 Convergent Temporal Features Evaluated for Song Recognition in a Host- Parasitoid Relationship ...... 39

3.1 Abstract ...... 39

3.2 Introduction ...... 40

3.3 Materials and Methods ...... 43

3.3.1 Animals ...... 43

3.3.2 Acoustic Stimuli...... 43

3.3.3 Experimental Apparatus...... 43

3.3.4 Protocol ...... 44

3.3.5 Data Analysis ...... 44

3.4 Results ...... 45

3.5 Discussion ...... 46

3.6 Acknowledgements ...... 50

References ...... 51

Figure Legends...... 54

Chapter 4 A Precedence Effect Resolves Phantom Sound Source Illusions in the Parasitoid Fly Ormia ochracea ...... 59

4.1 Abstract ...... 59

4.2 Introduction ...... 60

4.3 Materials and Methods ...... 62

4.3.1 Animals ...... 62

4.3.2 Field study ...... 62

4.3.3 Laboratory Study ...... 63

4.3.4 Data Analysis ...... 63

4.4 Results ...... 64

4.4.1 Field trials ...... 64

4.4.2 Laboratory Trials ...... 65

4.5 Discussion ...... 68

4.6 Acknowledgements ...... 71

References ...... 72

Figure Legends...... 76

Chapter 5 Distracting Noise Interferes with Sound Localization Accuracy in the Acoustic Parasitoid Fly Ormia ochracea ...... 85

5.1 Abstract ...... 85

5.2 Introduction ...... 86

5.3 Material and Methods ...... 88

5.3.1 Animals ...... 88

5.3.2 Acoustic Stimuli...... 88

5.3.3 Experimental Apparatus...... 88

5.3.4 Protocol ...... 89

5.3.5 Data Analysis ...... 90

5.4 Results ...... 91

5.4.1 Walking responses to song and noise in isolation ...... 91

vii

5.4.2 Effects of SNR and source separation on response thresholds ...... 91

5.4.3 Effects of SNR and source separation on response latency and walking distance ...... 91

5.4.4 Effects of SNR and source separation on walking direction ...... 92

5.5 Discussion ...... 93

5.6 Acknowledgements ...... 96

References ...... 97

Figure Legends...... 100

Chapter 6 General Discussion ...... 109

References ...... 115

viii

List of Tables

Table 2.1: Median Angular Heading to changes in broadcast location of songs

that varied in pulse rate………………………………………………………………...... 38

List of Figures

Figure 2.1: Effects of song pulse rate on walking distance………………………………………33

Figure 2.2: Effects of song pulse rate on steering velocity in tracking the switch in song location……………………………………………………………...... 34

Figure 2.3: Effects of song pulse rate on median angular heading in tracking the

switch in song location……………………...……………………………………35

Figure 2.4: Effects of song pulse rate on the error in median angular orientation………………………………………………………………………………………..36

Figure 2.5: Effects of song pulse rate on the phonotaxis performance index……………………37

Figure 3.1: Effects of varying pulse duration and interpulse interval on walking distance……...55

Figure 3.2: Effects of varying pulse duration and interpulse interval on peak steering velocity..56

Figure 3.3: Effects of varying pulse duration and interpulse interval on accuracy in tracking source locations………………………………………………………………….57

Figure 3.4: Using the phonotaxis performance index to measure response variation to varying pulse duration and interpulse interval…………………………………………...58

Figure 4.1: Experimental arena for field phonotaxis trials……………………………………....78

Figure 4.2: Phonotaxis to simultaneous and single sources under field conditions……………...79

Figure 4.3: Phonotaxis to simultaneous sources under laboratory conditions…………………...80

Figure 4.4: Phonotaxis to interdigitated (10 ms onset delay) sources…………………………...81

Figure 4.5: Phonotaxis to overlapping sources with small onset time differences

(5 and 0.2 ms onset delay)…………………………………………………….…82

Figure 4.6: Phonotaxis to overlapping sources with large onset delays

(100ms, 200ms, and 500ms)………………………………..……………………83

Figure 5.1: Experimental setup for tethered-walking phonotaxis………………………………103

Figure 5.2: Estimated song intensity response thresholds……………………………………...104

Figure 5.3: Forward velocities from phonotaxis………..………………………………………105

Figure 5.4: Steering velocities from phonotaxis………………………………………………..106

Figure 5.5: Virtual walking trajectories from phonotaxis ……………………………………...107

Figure 5.6: Current model of directional noise avoidance in Ormia ochracea………………...108

xi 1

Chapter 1 General Introduction

Behavioural decisions are often made based on evaluating information present in the environment. For animals that rely on acoustic signals to form these decisions, their auditory system must correctly identify and recognize sound sources of interest amongst a mixture of sound in the acoustic environment. Two common challenges associated with the task of sound source perception are: 1) grouping together spectral and temporal elements that correspond to the same source identity, and 2) separating relevant sound sources from irrelevant masking noise. These processing tasks are ubiquitous to animals of different taxa that depend on hearing to identify conspecific communication, locate prey, and to avoid predators. Investigating sensory mechanisms involved in these processing tasks may elucidate fundamental principles in common solutions, or divergent solutions that have been shaped by different evolutionary constraints.

1.1 Evaluating Features of Acoustic Signals

Acoustic signals play a crucial role in male-female pairing for reproduction (Gerhardt and Huber, 2002). Males of many anurans (frogs and toads) and orthopterans (crickets, katydids, grasshoppers) form large aggregations that produce conspicuous advertisement calls to attract receptive females (Gerhardt and Huber, 2002). These advertisement signals can be tonal or broadband sound pulses that are repeated at a species-specific pulse rate, and organized into short chirps or longer trills (Gerhardt and Huber, 2002). Females responding to the long range calls of males are faced with the task of identifying conspecific signals, assessing their level of attractiveness, and their location. The two main features of acoustic signals associated with song recognition are the: 1) spectral content, and 2) temporal pattern of signals (Hennig et al., 2004; Wilczynski and Ryan, 2010).

1.2 Spectral Features of Acoustic Signals Evaluated for Species Recognition

Unlike hearing in birds and mammals (Bradbury and Vehrencamp, 1998), most anurans and insects have a limited capacity to process spectral information (Gerhardt and Huber, 2002). In crickets that disperse by flight, hearing may be sensitive to two different categories of sound that allow for the categorical perception of conspecific calls and heterospecific echolocation of

insectivorous bats (Moiseff et al., 1978; Wyttenbach et al., 1996). In tethered flight, females steer towards sound with the carrier frequency of conspecific calls (positive phonotaxis) and steer away from sound with ultrasonic frequencies (negative phonotaxis) that are characteristic of echolocation calls (Moiseff et al., 1978). Preferences for particular carrier frequencies are most parsimoniously explained by frequency tuning of the peripheral auditory system to match the frequency content of conspecific signals (Gerhardt and Huber, 2002). In some instances, frequency preferences may contribute to the recognition of conspecific signals (Gerhardt, 1981; Hennig and Weber, 1997; Deily and Schul, 2006). In two closely related field cricket species, Teleogryllus oceanicus and T. commodus, calling songs are characterized by different temporal patterning of tonal sound pulses with different mean carrier frequencies (T. oceanicus: 4.5 kHz; T. commodus: 3.5 kHz). Both species show frequency-dependent preferences for the conspecific song temporal pattern (Hennig and Weber, 1997). The frequency preference function of T. commodus encompasses the carrier frequencies of T. commodus and T. oceanicus calling songs, and thus respond reliability to both song types (Hennig and Weber, 1997). However, T. oceanicus fail to respond to a conspecific song pattern that is presented at the mean carrier frequency of the T. commodus calling song as this is outside of the frequency preference function of T. oceanicus (Hennig and Weber, 1997). In this scenario, differences in song carrier frequency may assist T. oceanicus in song recognition, but certainly does not contribute to song recognition for T. comodus (Hennig and Weber, 1997). Most studies, instead, have demonstrated the importance of song temporal pattern in song recognition (Pollack and Hoy, 1979; Pollack, 2001; Hennig et al., 2004).

1.3 Temporal Features of Acoustic Signals Evaluated for Species Recognition

In most invertebrates, temporal features of acoustic signals are an important carrier of species- specific information (Hennig et al., 2004). In the duetting European grasshopper Chorthippus biguttulus, for example, males produce calling songs with asynchronous movements of a pair of hind legs across a vein on the forewings to generate broadband sound pulses. The onset of these sound pulses are mainly obscured by pulses produced with alternating leg movements; one- legged males produce songs with abnormally large gaps between sound pulses (von Helversen and von Helversen, 1997). Females evaluate these gaps and reject songs with gaps above a minimum threshold of 1.5-2 ms (von Helversen and von Helversen, 1997). Studies in other

orthopterans demonstrate other temporal features associated with song recognition (Hennig et al., 2004). For example, the field crickets Gryllus campestris and G. bimaculatus evaluates the pulse rate of songs for recognition (Thorson et al., 1982; Poulet and Hedwig, 2005). Preferences for certain pulse rates may be due to specific pulse durations, interval between pulses, or pulse periods (Hennig et al., 2004). Some evidence suggests that closely related species may not necessarily evaluate the same temporal features for song recognition. T. oceanicus selects for the appropriate pulse period, while T. commodus prefers a particular pulse duration of calls (Hennig, 2003).

1.4 Directional Hearing in Small Insects

Once a relevant sound source has been detected and recognized, the next task is to determine the source location. For animals with binaural hearing (two ears), sound direction may be determined by measuring two types of acoustic cues that vary with incident sound direction: an interaural time difference (ITD) that is a difference in sound arrival time at both ears; and an interaural level difference (ILD) that is a difference in sound amplitude between the two ears (Michelsen, 1998) (Fig. 1). ILD’s are generated by diffraction of an incident sound by an intervening body that is at least greater than one-tenth the wavelength of the sound; whereas ITD’s result from the additional time taken for sound to propagate to the contralateral ear after having reached the ipsilateral (Michelsen, 1998). Thus the magnitude of both cues depends on the size of the animal relative to the wavelength of sound to be localized.

A common problem encountered by small animals is their physical size severely limits the magnitude of binaural cues available for directional hearing. Some insects have independently derived different innovations for solving this sensory challenge in directional hearing. The tympanal organs (ears) of crickets are located on the proximal tibia of their forelegs and are separated by approximately 12 mm, resulting in minimal diffraction of the 7 cm wavelength cricket song (Michelsen et al., 1994b). These tympanal organs behave as pressure- difference receivers with directionality resulting from sound directly driving the external surface and indirectly driving the internal surface of the tympanal membrane (Michelsen, 1998). In crickets, this is made possible with tracheal tubes that transverse the prothoracic segment and are interconnected by a central membrane (Michelsen et al., 1994a). Auditory input from the ipsilateral and contralateral acoustic spiracles are conducted through this acoustic trachea to

reach the inner surface of the contralateral tympanum (Michelsen et al., 1994a). Measurements of changes in amplitude and phase for a given tympanum as the result of independent acoustic stimulation at each auditory input reveal that forces acting on the ipsilateral tympanum, ipsilateral acoustic spiracle, and contralateral acoustic spiracle contribute to tympanal displacement (vibration) while the contralateral tympanal input is negligible (Michelsen et al., 1994a). One critical feature of the cricket’s auditory system that may contribute to auditory directionality is the role of the central membrane in generating the appropriate time delay for sound input from the contralateral acoustic spiracle, to reach the ipsilateral tympanum (Michelsen et al., 1994a). The contributions of each auditory input in the displacement of the tympana were incorporated into a model to predict auditory directionality (Michelsen et al., 1994a). This model only matched auditory directionality derived from laser measurements of tympanal vibrations (Michelsen et al., 1994a), and auditory nerve recordings (Boyd and Lewis, 1983) when the appropriate delay was introduced to the contralateral acoustic input (Michelsen et al., 1994a). Directional hearing in crickets is based on measuring ILDs and is the result of this auditory directionality (Michelsen et al., 1994a; Hedwig and Poulet, 2004).

1.5 Hearing in Complex Acoustic Conditions

Hearing in noisy environments can be especially challenging when multiple acoustic signals with similar frequency content overlap spatially and temporally. One well known example of this situation is the so-called ‘cocktail party problem’ that refers to our difficulty in understanding speech in noisy environments such as acoustic conditions encountered in restaurants and other social settings (Cherry, 1953). Insects and anurans are often found to advertise in dense choruses (Gerhardt and Huber, 2002) that are characterized by similar acoustic conditions where overlapping signals may increase signal detection thresholds and mask song features relevant to recognition and localization (Romer et al., 1989; Brumm and Slabbekoorn, 2005; Bee, 2008). Choruses may be more or less structured in different species such that the timing of competing signals range from arbitrary (random) to synchronous (Greenfield, 1994). Although this synchrony is commonly found to be imperfect, as females are found to prefer leaders rather than followers (Gerhardt and Huber, 2002), and synchrony arises from competition among signalers to produce the leading call.

This auditory environment can be further complicated by acoustic clutter that generate reverberation (echoes) leading to ambiguous directional information. In these acoustic conditions, the first sound wave that arrives at the location of the receiver may correspond to the original source location while lagging sound waves may arise from acoustic reflections that travel by an indirect path to the receiver (Litovsky et al., 1999). Another form of ambiguous directional information may arise when two coherent sound sources arrive at the receiver within a short temporal disparity (Blauert, 1997). This results in auditory phantom source illusions where a single sound source is perceived midway between two actual source locations (location summing) (Blauert, 1997). Within a small temporal window of attention, the initial sound source has precedence over lagging sources in overall the perceived source location. Beyond this temporal window of attention, leading and lagging sound sources are perceived as distinct sources emanating from separate locations (Litovsky et al., 1999). This precedence effect is a sensory mechanism that may allow for resolving directional ambiguity caused by echoes, phantom source illusions (Wyttenbach and Hoy, 1993; Litovsky et al., 1999), and may also contribute to female preferences for leaders rather than followers in choruses (Greenfield, 1994; Siegert et al., 2011).

The primary goal of my thesis is to investigate auditory sound source perception in an acoustic parasitoid fly (Ormia ochracea) that depend on their ability of directional hearing to recognize and localize the calling songs specific of host cricket species for reproduction (Cade, 1975). I seek to determine the features of calling songs used in the recognition of host species, and sensory mechanisms that may allow acoustic parasitoid flies to parse out individual host crickets amongst concurrent noise in the acoustic environment.

1.6 Acoustically Orienting Parasitoid Flies (Diptera: Tachinidae)

Several acoustically orienting tachnids (Diptera: Tachinidae), have evolved to exploit the acoustic signals of orthopterans to localize host species for the purpose of providing a resource for the development of their larval young (Zuk and Kolluru, 1998). Ormia ochracea was discovered to be attracted to sound traps broadcasting the calling songs of the field cricket Gryllus rubens (Cade, 1975). Upon host localization, gravid female O. ochracea deposit first instar larvae that burrow into crickets to consume fat body and muscle tissue, and this often leads to a high probability of cricket death (Cade, 1975; Wineriter and Walker, 1990; Adamo et al.,

1995). Sound localization behaviour in O. ochracea is remarkably precise (Mason et al., 2001), even if crickets adopt the strategy of fragmented singing to avoid parasitism (Muller and Robert, 2002). Several other species of parasitoid flies eavesdrop on singing katydids (Allen, 1995; Lehmann et al., 2001), mole crickets (Parkman et al., 1996), and cicadas (Kohler and Lakes- Harlan, 2001) for host localization. In the sarcophagid fly Emblemasoma auditrix, gravid females perform phonotaxis to the calling songs of male cicadas (Okanagana rimosa) (Soper et al., 1976). Upon host localization, E. auditrix squeeze underneath the wings of the cicada to locate the timbal (sound producing structure) and forcefully cut through this structure to deposit larvae (Schniederkotter and Lakes-Harlan, 2004). This highly specialized behaviour depends on an auditory system that can correctly identify and localize specific host species.

Acoustically orienting tachnids possess hearing organs that are found on the prosternum, between the first pair of legs and the base of the neck (Robert et al., 1996b; Lakes-Harlan et al., 2007). In O. ochracea, the two tympana are separated by 0.05 cm (Robert et al., 1992) and results in no diffraction of the 5 kHz, 7 cm wavelength cricket song. Microphone recordings of the sound field confirm that ILDs are non-existent (0 dB) and ITDs are incredibly small (1.5 µsec maximum for a lateral sound source) (Robert et al., 1996a). Based on the principles of physics, directional hearing in O. ochracea should be impossible, but behavioural evidence demonstrates otherwise (Mason et al., 2001). O. ochracea are capable of localizing sound sources to the accuracy of 2° azimuth (Mason et al., 2001), a level of precision that is on par with human directional hearing (Blauert, 1997).

1.7 Directional Hearing in O. ochracea

Directional hearing in O. ochracea is made possible with tympanal membranes that are mechanically coupled by a flexible intertympanal cuticular bridge (Miles et al., 1995; Robert et al., 1996a). For an incident sound at 90° azimuth relative to the midline, the ipsilateral tympanum responds approximately 50 µsecs earlier, and with amplitudes that are 10-12 dB greater than the contralateral tympanum (Robert et al., 1996a). These mechanical timing and intensity response differences are the result of the intertympanal bridge that introduces a delay to produce asymmetrical oscillations of the tympanal membranes (Miles et al., 1995; Robert et al., 1996a, 1998). As a consequence, this mechanical coupling amplifies minute sound field ITDs to

greater mechanical ITDs and ILDs that vary as a function of incident sound direction (Robert et al., 1996a).

O. ochracea also possess primary auditory afferents that are highly specialized to process temporal information essential for hyperacute directional hearing. Inside the air-filled chamber of the prosternum is a pair of auditory organs (bulba acoustica) with an apodeme that inserts into the tympanal membrane (Robert et al., 1992). Each bulba acoustica contains approximately 100 receptor cells (Robert and Willi, 2000). Neural measurements from the auditory nerve and individual receptor cells reveal that their sensitivity is tuned to the carrier frequency of field cricket calling songs (Robert et al., 1992; Oshinsky and Hoy, 2002). A majority of the primary auditory afferents respond with a single spike precisely at the onset of a sound pulse that is independent of stimulus intensity or duration (Mason et al., 2001). Instead, response latencies vary as a function of sound intensity such that shorter latencies occur for greater intensities (Oshinsky and Hoy, 2002). There is little variation (low jitter) in the latency of spike timing to the onset of sound pulses (~75 µsec). Two putative neural codes have been proposed for coding sound direction: 1) an interaural latency difference between the two ears that vary as a function of incident sound direction (Mason et al., 2001; Oshinsky and Hoy, 2002), and 2) a population level code that involves direction-dependent range fractionation of the auditory afferents (Oshinsky and Hoy, 2002). Currently, the processing of sound direction beyond the level of the primary auditory afferents is unknown.

O. ochracea may encounter complex acoustic situations when they are within earshot of multiple attractive host crickets advertising in dense aggregations (Cade, 1981). In addition to ambient noise, the calling songs of other orthopterans sharing the same acoustic space may be a source of distraction. O. ochracea occur in several geographic regions in the southern United States (Florida, Texas, and California) and in Hawaii where they are found to exploit different species of field crickets within each region (Cade, 1975; Walker, 1993; Zuk et al., 1995; Wagner, 1996; Gray et al., 2007). Field sound trap experiments demonstrate that O. ochracea are able to discriminate between songs of different cricket species (Walker, 1993), and they prefer songs of their primary local host (Gray et al., 2007). Successful host song recognition and localization may demand the auditory system of O. ochracea to segregate and process song temporal patterns that correspond to a correct source identity over noise.

1.8 Thesis Outline

In my thesis I seek to elucidate sensory mechanisms involved in auditory source perception in Floridian O. ochracea. Our current understanding of song recognition in O. ochracea is based on song preferences established from field sound trap experiments that provide capture rates for different song variants (Walker, 1993; Gray et al., 2007). These experiments fail to reliably demonstrate specific song temporal parameters evaluated for recognition and cannot reveal dynamic features (and response variation) of phonotactic responses that may underlie song preferences. In Chapter 2, I employ a highly sensitive spherical trackball system to record O. ochracea walking phonotaxis (Lott et al., 2007). Using a no-choice paradigm I test the ability of O. ochracea to track a switch in broadcast location of test songs that only varied in song pulse rates (constant duty cycle). I assessed three different response measures (walking distance, steering velocity, and orientation accuracy) for their suitability in capturing response variation that may describe preferences for a range of pulse rates. I incorporate these measures of phonotaxis into deriving an index that describes phonotactic performance (Chapter 2).

Different combinations of pulse durations and interpulse intervals may combine to form a given pulse rate. The goal of Chapter 3 was to determine the specific temporal parameter (pulse duration, interpulse interval, or pulse period) evaluated for song recognition. Using a similar experimental paradigm as in Chapter 2, but with test songs that vary in different combinations of pulse durations and interpulse interval, I quantify the ‘temporal pattern recognition space’ of Floridian O. ochracea.

The goal of Chapter 4 was to determine the role of directional hyperacuity in O. ochracea in resolving the location of a single sound source amongst multiple attractive sources in the environment. In field-based free-walking phonotaxis experiments, I examined if O. ochracea were able to separate two simultaneous attractive sound sources to successfully localize a single source. Under acoustically controlled laboratory conditions, I test the specific hypothesis that exploitation of small time differences allows for the precedence effect to resolve phantom sound source illusions. In these experiments, I quantified free-walking phonotactic responses to the broadcast of two competing sound sources that varied in the amount of temporal overlap.

In Chapter 5, I test the hypothesis that spatial release from masking contributes to successful host localization in O. ochracea. In tethered walking phonotaxis experiments, I

9 subjected O. ochracea to the simultaneous broadcast of band-limited noise and cricket song at different signal-to-noise ratios, from two speakers that were spatially grouped or separated. I predict that increased spatial separation between cricket song and noise will allow O. ochracea to selectively localize the target signal over noise, with better localization accuracy for greater signal-to-noise ratio.

References

Adamo, S. A., Robert, D. and Hoy, R. R. (1995). Effects of a tachinid parasitoid, Ormia ochracea, on the behavior and reproduction of its male and female field cricket hosts (Gryllus spp). Journal of Insect Physiology 41, 269-277.

Allen, G. R. (1995). The biology of the phonotactic parasitoid, Homotrixa sp. (Diptera, Tachinidae), and its impact on the survival of male Sciarasaga quadrata (Orthoptera, Tettigoniidae) in the field. Ecol. Entomol. 20, 103-110.

Bee, M. A. (2008). Finding a mate at a cocktail party: spatial release from masking improves acoustic mate recognition in grey treefrogs. Animal Behaviour 75, 1781-1791.

Blauert, J. (1997). Spatial Hearing: The Pyschophysics of Human Sound Localization. Cambridge, MA: The MIT Press.

Boyd, P. and Lewis, B. (1983). Peripheral auditory directionality in the cricket (Gryllus campestris L., Teleogryllus oceanicus Le Guillou). Journal of Comparative Physiology 153, 523-532.

Bradbury, J. and Vehrencamp, S. (1998). Principles of animal communication. Sunderland, Mass: Sinauer Associates.

Brumm, H. and Slabbekoorn, H. (2005). Acoustic Communication in Noise. In Advances in the Study of Behavior, vol. Volume 35, pp. 151-209: Academic Press.

Cade, W. (1975). Acoustically orienting parasitoids - fly phonotaxis to cricket song. Science 190, 1312-1313.

Cade, W. H. (1981). Field cricket spacing, and the phonotaxis of crickets and parasitoid flies to clumped and isolated cricket songs. Zeitschrift Fur Tierpsychologie-Journal of Comparative Ethology 55, 365-375.

Cherry, E. C. (1953). Some experiments on the recognition of speech, with one and with two ears. J. Acoust. Soc. Am. 25, 975-979.

Deily, J. A. and Schul, J. (2006). Spectral selectivity during phonotaxis: a comparative study in Neoconocephalus (Orthoptera : Tettigoniidae). Journal of Experimental Biology 209, 1757-1764.

Gerhardt, H. C. (1981). Mating call recognition in the barking treefrog (Hyla gratiosa): responses to synthetic calls and comparisons with the green treefrog (Hyla cinerea). Journal of Comparative Physiology 144, 17-25.

Gerhardt, H. C. and Huber, F. (2002). Acoustic Communication in Insects and Anurans: Common problems and diverse solutions. Chicago: The University of Chicago Press.

Gray, D. A., Banuelos, C., Walker, S. E., Cade, W. H. and Zuk, M. (2007). Behavioural specialization among populations of the acoustically orienting parasitoid fly Ormia ochracea utilizing different cricket species as hosts. Animal Behaviour 73, 99-104.

Greenfield, M. D. (1994). Synchronous and alternating choruses in insects and anurans: common mechanisms and diverse functions. American Zoologist 34, 605-615.

Hedwig, B. and Poulet, J. F. A. (2004). Complex auditory behaviour emerges from simple reactive steering. Nature 430, 781-785.

Hennig, R. M. (2003). Acoustic feature extraction by cross-correlation in crickets? J. Comp. Physiol. A -Neuroethol. Sens. Neural Behav. Physiol. 189, 589-598.

Hennig, R. M. and Weber, T. (1997). Filtering of temporal parameters of the calling song by cricket females of two closely related species: A behavioral analysis. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 180, 621-630.

Hennig, R. M., Franz, A. and Stumpner, A. (2004). Processing of auditory information in insects. Microsc. Res. Tech. 63, 351-374.

Kohler, U. and Lakes-Harlan, R. (2001). Auditory behaviour of a parasitoid fly (Emblemasoma auditrix, Sarcophagidae, Diptera). Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 187, 581-587.

Lakes-Harlan, R., Jacobs, K. and Allen, G. R. (2007). Comparison of auditory sense organs in parasitoid Tachinidae (Diptera) hosted by Tettigoniidae (Orthoptera) and homologous structures in a non-hearing Phoridae (Diptera). Zoomorphology 126, 229-243.

Lehmann, G. U. C., Heller, K. G. and Lehmann, A. W. (2001). Male bushcrickets favoured by parasitoid flies when acoustically more attractive for conspecific females (Orthoptera : Phanopteridae/Diptera : Tachinidae). Entomol. Gen. 25, 135-140.

Litovsky, R. Y., Colburn, H. S., Yost, W. A. and Guzman, S. J. (1999). The precedence effect. J. Acoust. Soc. Am. 106, 1633-1654.

Lott, G. K., Rosen, M. J. and Hoy, R. R. (2007). An inexpensive sub-millisecond system for walking measurements of small animals based on optical computer mouse technology. Journal of Neuroscience Methods 161, 55-61.

Mason, A. C., Oshinsky, M. L. and Hoy, R. R. (2001). Hyperacute directional hearing in a microscale auditory system. Nature 410, 686-690.

Michelsen, A. (1998). Biophysics of Sound Localization in Insects. In Comparative Hearing: Insects, eds R. R. Fay and A. N. Popper), pp. 18-62. New York: Springer-Verlag.

Michelsen, A., Popov, A. V. and Lewis, B. (1994a). Physics of directional hearing in the cricket Gryllus bimaculatus. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 175, 153-164.

Michelsen, A., Heller, K. G., Stumpner, A. and Rohrseitz, K. (1994b). A new biophysical method to determine the gain of the acoustic trachea in bush-crickets. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 175, 145-151.

Miles, R. N., Robert, D. and Hoy, R. R. (1995). Mechanically coupled ears for directional hearing in the parasitoid fly Ormia ochracea. J. Acoust. Soc. Am. 98, 3059-3070.

Moiseff, A., Pollack, G. S. and Hoy, R. R. (1978). Steering responses of flying crickets to sound and ultrasound: mate attraction and predator avoidance. Proceedings of the National Academy of Sciences of the United States of America 75, 4052-4056.

Muller, P. and Robert, D. (2002). Death comes suddenly to the unprepared: singing crickets, call fragmentation, and parasitoid flies. Behavioral Ecology 13, 598-606.

Oshinsky, M. L. and Hoy, R. R. (2002). Physiology of the auditory afferents in an acoustic parasitoid fly. Journal of Neuroscience 22, 7254-7263.

Parkman, J. P., Frank, J. H., Walker, T. J. and Schuster, D. J. (1996). Classical biological control of Scapteriscus spp (Orthoptera: Gryllotalpidae) in Florida. Environ. Entomol. 25, 1415- 1420.

Pollack, G. S. (2001). Analysis of temporal patterns of communication signals. Current Opinion in Neurobiology 11, 734-738.

Pollack, G. S. and Hoy, R. R. (1979). Temporal pattern as a cue for species-specific calling song recognition in crickets. Science 204, 429-432.

Poulet, J. F. A. and Hedwig, B. (2005). Auditory orientation in crickets: Pattern recognition controls reactive steering. Proceedings of the National Academy of Sciences of the United States of America 102, 15665-15669.

Robert, D. and Willi, U. (2000). The histological architecture of the auditory organs in the parasitoid fly Ormia ochracea. Cell Tissue Res. 301, 447-457.

Robert, D., Amoroso, J. and Hoy, R. R. (1992). The evolutionary convergence of hearing in a parasitoid fly and its cricket host. Science 258, 1135-1137.

Robert, D., Miles, R. N. and Hoy, R. R. (1996a). Directional hearing by mechanical coupling in the parasitoid fly Ormia ochracea. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 179, 29-44.

Robert, D., Miles, R. N. and Hoy, R. R. (1998). Tympanal mechanics in the parasitoid fly Ormia ochracea: intertympanal coupling during mechanical vibration. J. Comp. Physiol. A - Neuroethol. Sens. Neural Behav. Physiol. 183, 443-452.

Robert, D., Edgecomb, R. S., Read, M. P. and Hoy, R. R. (1996b). Tympanal hearing in tachinid flies (Diptera, Tachinidae, Ormiini): The comparative morphology of an innovation. Cell Tissue Res. 284, 435-448.

Romer, H., Bailey, W. and Dadour, I. (1989). Insect hearing in the field III. Masking by noise. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 164, 609-620.

Schniederkotter, K. and Lakes-Harlan, R. (2004). Infection behavior of a parasitoid fly, Emblemasoma auditrix, and its host cicada Okanagana rimosa. J Insect Sci. 4, 7.

Siegert, M. E., Römer, H., Hashim, R. and Hartbauer, M. (2011). Neuronal correlates of a preference for leading signals in the synchronizing bushcricket Mecopoda elongata (Orthoptera, Tettigoniidae). The Journal of Experimental Biology 214, 3924-3934.

Soper, R. S., Shewell, G. E. and Tyrrell, D. (1976). Colcondamyia auditrix nov. sp. (Diptera: Sarcophagidae), a parasite which is attracted by the mating song of its host, Okanagana rimosa (Homoptera: Cicadidae). Canadian Entomologist 108, 61-68.

Thorson, J., Weber, T. and Huber, F. (1982). Auditory behaviour of the cricket. II. Simplicity of calling-song recognition in Gryllus, and anomalous phonotaxis at abnormal carrier frequencies. Journal of Comparative Physiology 146, 361-378. von Helversen, D. and von Helversen, O. (1997). Recognition of sex in the acoustic communication of the grasshopper Chorthippus biguttulus (Orthoptera, Acrididae). Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 180, 373-386.

Wagner, W. E. (1996). Convergent song preferences between female field crickets and acoustically orienting parasitoid flies. Behavioral Ecology 7, 279-285.

Walker, T. J. (1993). Phonotaxis in female Ormia ochracea (Diptera, Tachinidae), a parasitoid of field crickets. J. Insect Behav. 6, 389-410.

Wilczynski, W. and Ryan, M. J. (2010). The behavioral neuroscience of anuran social signal processing. Current Opinion in Neurobiology In Press, Corrected Proof.

Wineriter, S. A. and Walker, T. J. (1990). Rearing phonotactic parasitoid flies (Diptera, Tachinidae, Orminii, Ormia spp.). Entomophaga 35, 621-632.

Wyttenbach, R. A. and Hoy, R. R. (1993). Demonstration of the precendence effect in an insect. J. Acoust. Soc. Am. 94, 777-784.

Wyttenbach, R. A., May, M. L. and Hoy, R. R. (1996). Categorical perception of sound frequency by crickets. Science 273, 1542-1544.

Zuk, M. and Kolluru, G. R. (1998). Exploitation of sexual signals by predators and parasitoids. Q. Rev. Biol. 73, 415-438.

Zuk, M., Simmons, L. W. and Rotenberry, J. T. (1995). Acoustically-orienting parasitoids in calling and silent males of the field cricket Teleogryllus oceanicus. Ecol. Entomol. 20, 380-383.

Chapter 2 Deriving a Sensitive Measure of Walking Phonotaxis Performance in the Acoustic Parasitoid Fly Ormia ochracea

2.1 Abstract

Sensitive methods for quantifying behaviour are essential to understand the function and mechanistic basis of behaviour. The acoustic parasitoid fly Ormia ochracea engages in positive phonotaxis to field cricket calling songs solely for the purpose of reproduction. O. ochracea occur in several geographically distinct populations in the United States, with each population showing specialized preferences to different cricket species that produce calling songs that differ in the temporal patterning of sound pulses. In Florida, the field cricket Gryllus rubens produce calling songs with an approximate pulse rate of 58.6 pulses/sec (pps) and is the preferred host cricket species of Floridian O. ochracea. Directional hearing in O. ochracea has been the subject of previous study, but much less is known about song features evaluated for cricket species recognition. In this study, we use a spherical treadmill system to record open-loop walking phonotaxis from Floridian O. ochracea in response to cricket song models that differed in pulse rates. Song recognition was examined by measuring tracking performance with the switch in song broadcast location between two speakers positioned at ±45° azimuth. Walking distance, peak steering velocity, mean angular heading, and error in angular orientation varied with song pulse rate. All of these measures were incorporated into a phonotaxis performance index that compared walking responses to test songs relative to the standard 50 pps G. rubens song. According to the index, O. ochracea are behaviourally tuned to a range of song pulse rates between 50-70 pps. This range of preferred pulse rates are in direct agreement with previous field sound trap experiments that demonstrate a preference for the calling songs of G.rubens (58.6 pps) over G. texensis (78.6 pps). No-choice paradigms that utilize a switch in song broadcast location demands the animal to change their course of response, and this provides the advantage of measuring response variation underlying signal preferences and song recognition.

2.2 Introduction

Advances in the study of animal behaviour depend on reliable methods for quantifying behaviour to elucidate their function and mechanistic basis. Studies in phonotaxis (oriented locomotor responses to sound), can exploit the natural behavioural repertoire of animals to provide insights into signal function, the information content conveyed in signals, and sensory mechanisms used to process signals that form the basis of behavioural decisions. In the acoustic parasitoid fly Ormia ochracea, gravid females utilize directional hearing (Robert et al., 1996) to perform flying (Cade, 1975; Walker, 1993; Muller and Robert, 2001) and walking phonotaxis (Mason et al., 2005) to the calling songs of field crickets for host localization (Cade, 1975). After arriving in close proximity to crickets, O. ochracea deposit first instar planidia (larvae) that burrow into crickets where they feed on fat body and muscle tissue for development (Cade, 1975; Wineriter and Walker, 1990; Adamo et al., 1995). In flight, O. ochracea can land within 8- 10 cm of a sound source (Muller and Robert, 2001) and walking phonotaxis experiments reveal that O. ochracea can localize sound sources to the accuracy of 2° azimuth (Mason et al., 2001). What specific features of cricket song are evaluated for host recognition is currently unknown. Studies from both anurans and other insect species have measured phonotaxis to determine song features evaluated for species recognition (von Helversen and von Helversen, 1995; Schul, 1998; Schul et al., 1998), acoustic cues and sensory processes in computing sound direction (Mason et al., 2001; Pollack, 2003), acuity in directional hearing (Rheinlaender and Blatgen, 1982; Klump and Gerhardt, 1989; Mason et al., 2001; Schöneich and Hedwig, 2010), and even how these organisms may solve fundamental hearing problems such as source detection in noisy environments (Romer and Krusch, 2000; Bee and Micheyl, 2008; Lee et al., 2009).

Phonotaxis can be measured from freely walking (Murphey, 1972; Mhatre and Balakrishnan, 2007; Lee et al., 2009) or flying (Straw et al.; Muller and Robert, 2001, 2002) animals in a closed-loop approach with the advantage of recording behaviour under naturalistic sensory conditions. Unrestrained animals are able to move freely to experience changes in sensory input (i.e. exploitation of sound intensity gradients) (Ulagaraj and Walker, 1973). However, the closed-loop approach may not allow for one to establish strategy in sound localization (lateralization versus localization) or preferences to single signal parameters when movement to a source introduces the confounding effects of changing multiple stimulus conditions. The open-loop experimental paradigm eliminates sensory feedback by cessation of

16 stimulus broadcast before locomotion or restraining animals at a fixed point relative to stimulus sources (Klump, 1995). These measurements can yield valuable information in understanding sound localization and song recognition. Early studies of song recognition in the field cricket Gryllus campestris utilized a spherical treadmill system to measure the ability of crickets to track song models that varied in pulse rates (Weber et al., 1981; Thorson et al., 1982). Open-loop conditions were achieved with walking measurements that controlled servomotors in rotating the treadmill such that “free-walking” animals were repositioned to experience constant acoustic conditions (Weber et al., 1981). This compensated walking revealed that song tracking occurred to a range of pulse rates between 20-40 pps and tracking performance declined above and below this range (Thorson et al., 1982). Best phonotaxis performance was observed at 30 pps forming the ’30 Hz hypothesis’ of cricket song recognition (Thorson et al., 1982). Modern high-temporal resolution treadmill systems have been developed with the capacity to transduce walking responses at data rates that resolve single stepping cycles and this has revealed reactive steering responses of crickets to individual sound pulses within a song (Hedwig and Poulet, 2004; Hedwig and Poulet, 2005; Lott et al., 2007). Furthermore, the onset of phonotaxis occurred within 50-60 msec, which precluded measurements of song temporal pattern for initiation of phonotaxis. Instead, song recognition appeared to modulate the strength of phonotaxis while reactive pulse by pulse steering contributed to overall walking direction (Hedwig and Poulet, 2004; Hedwig and Poulet, 2005).

Song recognition in O. ochracea has received less attention than compared to sound localization. Most studies have used sound traps in the field to examine preferences for songs that differed in the duration or the amount of chirps/trills rather than specific temporal features that may be evaluated for recognition (Wagner, 1996; Zuk et al., 1998; Gray and Cade, 1999; Wagner and Basolo, 2007). These studies generally show that O. ochracea prefer louder songs, with longer chirps at higher chirp rates (Wagner, 1996; Zuk et al., 1998; Wagner and Basolo, 2007). Geographically distinct populations (Florida, Texas, California, and Hawaii) of O. ochracea parasitize different cricket species that produce songs with species-specific temporal patterns (Gray et al., 2007). Other studies have deployed sound traps that presented songs of different cricket species to examine host song preferences and O. ochracea appear to prefer songs of the primary local host (Walker, 1993; Gray et al., 2007), suggesting that some unknown song feature must be used for recognition and host species discrimination. Sound trap

experiments reveal overall population level preferences for test songs by capture rates, but not variation in responses that may underlie song preferences. A measure of phonotaxis that quantifies variation in responses over a range of stimulus parameters, rather than average attractiveness of alternative stimuli would allow more insight into the sensory mechanisms underlying song recognition. Several studies have combined several response measures to derive an index of phonotaxis (Schul, 1998; Bush et al., 2002; Bee, 2007).

Previous work using phonotaxis indices has incorporated all or some of the following response measures to quantify performance: response probability, response latency, duration to reach a sound source, distance travelled, meander in walking path, and accuracy (angular orientation and error) (von Helversen, 1984; Schul et al., 1998; Bush et al., 2002; Schul and Bush, 2002). Response probability measures the presence/absence of phonotaxis or the proportion of choice to a sound source (von Helversen, 1984). Both response probability and response latency can vary with the ability to detect a source or the motivation to respond (or both) (Bush et al., 2002; Schul and Bush, 2002). Response duration, distance traveled, meander in walking, and accuracy differ based on the approach to a source location (Schul et al., 1998; Bee, 2007). The relative difference between controls and alternative test conditions for these response measures are combined to derive a phonotaxis performance index that range from 0 (poor performance), to 1 (performance equivalent to control responses) (Schul, 1998). Many anurans and orthopterans perform walking phonotaxis using a zig-zag approach by turning to the ear most stimulated (sound source lateralization) and this introduces a considerable amount of localization error (Rheinlaender et al., 1979). Prior to making subsequent responses, time is required to determine changes in source location and corrective turns introduce meander that results in greater walking path lengths (Rheinlaender et al., 1979; Rheinlaender and Blatgen, 1982). These previous phonotaxis indices are most sensitive to measuring differences in walking path length.

In this study, we derive a phonotaxis index that has several advantages. This index incorporates total walking distance and peak steering velocity to measure response motivation and error in angular orientation to assess localization accuracy. We demonstrate that the index is suited for use in no-choice paradigms by measuring the ability of animals to track changes in song broadcast location. Furthermore, this index works independently of localization behaviour

18 characterized by a zig-zag approach, and is sensitive to detecting response variation that may underlie song preferences and recognition.

2.3 Materials and Methods 2.3.1 Animals

Experiments were conducted on lab-reared gravid female Ormia ochracea derived from a population originally collected in Gainesville FL. Flies were maintained in environmentally controlled chambers (Power Scientific, Inc. Model DROS52503, Pipersville PA) at 25° C and 75% humidity on a 12-h:12-h light:dark regime and fed nectar solution (The Birding Company, Yarmouth MA) ad libitum.

2.3.2 Acoustic Stimuli

The standard control song, modeled after the calling song of Gryllus rubens, was a trill constructed from 10 msec duration 5 KHz tone sound pulses (1ms on/off ramps) separated by 10 msec interpulse intervals and repeated at 50 pulses/sec for a total stimulus duration of 1 second. All test stimuli (model songs) were 1 second in duration. Pulse rate preferences were examined with songs that ranged from 10 to 90 pulses/second in 10 pulses/second increments. Songs that differed in pulse rates were constructed by adjusting pulse durations and interpulse intervals in equal portions to maintain a constant 50% duty cycle.

Stimulus waveforms were synthesized in Matlab (R2009b, The MathWorks Inc. USA) with custom software and converted to analog signals using National Instruments data acquisition hardware (NI USB-6251, 44100 Hz), amplified with Radio Shack Realistic (SA-10 Solid State Amplifier MOD-31-1982B, Taiwan or NAD S300 amp) and broadcast through piezo electric tweeters. Stimulus levels were controlled with programmable attenuators (Tucker Davis Technologies System 3 PA5) and calibrated using a probe microphone (B&K Type 4182, Denmark) powered by B&K Nexus Conditioning Amplifier (Denmark).

2.3.3 Spherical treadmill

Behavioural measurements were made from tethered flies performing walking phonotaxis on a high resolution trackball system situated equidistant (15 cm) from two test speakers positioned at ±45° azimuth and surrounded by acoustic attenuating foam. The trackball system consists of a

19 light-weight table tennis ball held afloat by a constant airstream above a modified optical mouse sensor (ADNS 2620, Avago Technologies, USA). Walking responses were transduced as rotations of the trackball that actuated the optical mouse sensor to record changes in x and y pixel units at a sampling rate of 2160 Hz (Lott et al., 2007). Pixel units were calibrated to actual walking distances by measuring displacement of points on the ball from highspeed video footage (DRS Lightening RDT, 500 frames per second) synchronized to pixel data captured by the trackball system. Data collection by the trackball system was controlled by custom Matlab software that interfaced with the National Instruments data acquisition system to ensure synchronous sound presentation and the recording of walking traces.

2.3.4 Protocol

A test sequence commenced with a presentation of the standard cricket song from both left (-45°) and right (+45°) speakers synchronously, followed by the presentation of the standard song from the left speaker, then the right speaker in isolation. To control for possible order effects, 1) test responses were primed with the standard cricket song presented from both left and right speakers at an overall intensity of 76 dB SPL, and 2) flies were given 30 seconds of rest between stimulus presentations. Following the priming stimulus, a test song was broadcast from one speaker for 500 msec and switched to the other speaker for the remaining 500 msec of presentation. The same test song was presented for a second iteration, but with the speaker presentation order reversed (i.e. left/right to right/left). This process was repeated for all test songs with the sequence of test songs and the order of first speaker presentation randomized. Experiments ended with the same sequence of standard song presentations that was initially presented in the beginning.

2.3.5 Data Analysis

X and Y coordinates from trackball data traces were collected at 2160 Hz to construct virtual walking trajectories. Steering velocities were calculated as changes in x coordinates over time. The instantaneous angular heading (theta) was calculated by converting Cartesian x and y values to polar coordinates by computing the inverse tangent of y divided by x (instantaneous angular heading = arctan(y/x)). To determine the error in angular orientation, a grand mean angular heading (mean of median instantaneous angular heading) for the first half of the response (response to the initial speaker) and the second half of the response (response to the switched

speaker) was calculated for tracking the standard 50pps song. This value was used as the reference angular heading to each speaker. Error in angular orientation (measure of angular distance to relative to the average reference response) for each response to a test song was determined as the absolute difference between the reference angular heading and the median angular heading to each test speaker. The error in angular orientation measured for the 50pps song condition (Fig. 4) captures the angular response variability to the control song. The phonotaxis performance index incorporated distance travelled, steering velocity, and error in angular orientation and was calculated as:

Phonotaxis Score Dist Steering Velocity Max Steering Velocity Min = × × Dist푇푒푠푡 Steering Velocity Max푇푒푠푡 Steering Velocity Min푇푒푠푡

� 푅푒푓 � � 푅푒푓 � � 푅푒푓 � × Cosine median angular error .

푇푒푠푡 푟푒푠푝 푡표 푖푛푖푡푖푎푙 푙표푐푎푡푖표푛 × Cosine��median angular error . �� 푇푒푠푡 푟푒푠푝 푡표 푠푤푖푡푐ℎ푒푑 푙표푐푎푡푖표푛��

Reference conditions refer to measurements in response to the standard 50 pps song and test conditions refer to measurements in response to test songs. Index values range from 0 to >1. A phonotaxis performance index of 0 indicates poor performance, 1 indicates performance equivalent to responses to the standard song, >1 indicates performance better than responses to the standard song. Repeated responses for the same stimulus conditions were averaged within individuals. Data are given as means ± SEM.

2.4 Results

A total of 141 walking responses were recorded from each of 11 flies. During the first simultaneous presentation of the standard cricket song from both speakers at ±45° azimuth, flies oriented to a forward direction (0.29±1.55°) and covered a mean distance of 5.58±0.47 cm (see (Lee et al., 2009). In the final dual song presentation following a test sequence, flies localized a similar direction (1.11±0.99°) and covered a similar distance (5.09±0.36 cm) that did not differ significantly from initial responses (Distance, Paired T-Test: t(11)=0.97, p=0.36, angular heading, Moore’s Test: R`=0.17, p>0.05).

After priming with the standard song from both speakers, flies reliably performed phonotaxis to all test songs with an average response latency of 82.43±1.28 ms that did not significantly vary with changes in song pulse rate (Friedman Test: χ2(8)=7.394, p=0.495). Mean walking distance varied significantly with song pulse rate (Repeated Measures ANOVA with

Greenhouse-Geisser correction: F(3.037, 30.375) = 12.653, p<0.001, Fig. 1). Increases in pulse rates from 10 to 40 pps caused walking distance to increase from 1.89±0.34 to 3.22±0.33 cm, approached saturation between 50pps to 70pps with distances that ranged from 3.68±0.24 to 3.89±0.29 cm, and decreased with further increases in pulse rate (Fig. 1).

When a test song was initially presented from the left speaker for 500 msec, flies steered towards the left (negative steering values). As the song was switched to the right speaker for the remaining 500 ms of broadcast, flies steered in the opposite direction to track the switch in source location (positive steering values) (Fig. 2a, i). Reversing the order of speaker broadcast produced similar results (Fig. 2a, ii). Peak steering velocities in tracking the location of broadcast

depended on song pulse rate (Repeated Measures ANOVA: F(8,80)= 39.817, p<0.001), but not the location of the initial broadcast (i.e. left speaker leading compared to right speaker leading)

(Repeated Measures ANOVA: F(1,10)= 0.265, p<0.618) nor the switch in song location (Repeated

Measures ANOVA: F(1,10)= 2.846, p<0.123) (Fig. 2b). As pulse rates increased from 10 to 40pps peak steering velocities increased from 1.37±0.204 to 2.69±0.44 cm/sec. Post-hoc analyses with Bonferroni corrected p values indicate that these peak steering velocities were significantly less than that of the standard 50 pps song (-4.09±0.25 cm/sec) (50 vs 10 pps: p < 0.001, 50 vs 20pps: p <0.001, 50 vs 30 pps: p < 0.001, 50 vs 40 pps: p=0.008). Peak steering velocity reached a relative maximum ranging from 4.09±0.25 to 3.58±0.32 cm/sec between 50 to 70 pps, with peak steering velocities at 60 and 70 pps not significantly different from those for the standard 50 pps song (50 vs 60 pps: p = 1.000, 50 vs 70 pps: p = 0.913). When pulse rates were increased further, peak steering velocity decreased to values significantly lower than the standard song (50 vs 80 pps: p = 0.005, 50 vs 90 pps: p <0.001).

Median angular heading depended on song pulse rate and this effect was greater following a change in source location (Fig. 3). Flies generally localized a direction that approached the initial broadcast location for a broad range of pulse rates, although angular distributions tended to be broader for extreme pulse rates (Fig. 3ai,3bi). Flies were more selective, however, in localizing a switch in the broadcast location (Table 1, Fig. 3). Re-

orientation median angular headings to songs with pulse rates between 60-70 pps did not differ significantly from responses to the 50 pps song (Moore’s Test: Left, 50 vs 60 pps: R`=0.23, p>0.05, 50 vs 70: R`=0.44, p>0.05; Right, 50 vs 60 pps: R`=0.16, p>0.05, 50 vs 70 pps: R`=0.86, p>0.05, Fig. 3b,d). At pulse rates above and below this range, flies localized a forward direction that differed significantly from localization responses to the 50 pps song that were directed to the broadcast speaker (Left, 50 vs 10 pps: R`= 1.75, p<0.05, 50 vs 20 pps: R`=1.79, p<0.05, 50 vs 30 pps: R`=1.79, p<0.05, 50 vs 40 pps: R`=1.50, p<0.05, 50 vs 80 pps: R`=1.45, p<0.05, 50 vs 90pps: R`=1.77, p<0.05; Right, 50 vs 10 pps: R`= 1.73, p<0.05, 50 vs 20 pps: R`=1.79, p<0.05, 50 vs 30 pps: R`=1.79, p<0.05, 50 vs 40 pps: R`=1.72, p<0.05, 50 vs 80 pps: R`=1.56, p<0.05, 50 vs 90pps: R`=1.77, p<0.05, Fig. 3).

Results show a significant effect of song presentation order on the error in angular orientation (Repeated Measures ANOVA: F(1,10)= 6.669, p=0.027). With a switch in broadcast location, this error was significantly greater for re-orientation responses than for initial localization responses (Fig. 4). Error in angular orientation was significantly affected by song pulse rate (Repeated Measures ANOVA: F(8,80)= 11.355, p<0.001). Pairwise comparisons with Bonferroni corrected p values reveal that localization error for songs that ranged from 60-70 pps did not significantly differ from responses to the 50 pps song (50 vs 60 pps: p=0.581, 50 vs 70 pps: p=0.689). Error in angular orientation for songs above and below this range were significantly greater than responses to the 50 pps song (50 vs 10 pps: p<0.001, 50 vs 20 pps: p<0.001, 50 vs 30 pps: p=0.001, 50 vs 40 pps: p=0.003, 50 vs 80 pps: p=0.031, 50 vs 90 pps: p<0.001). There was also a significant interaction effect between pulse rate and presentation

order (Repeated Measures ANOVA: F(8,80)= 2.462, p=0.019). Re-orientation responses exhibited significantly smaller localization error for songs with pulse rates between 50-70 pps (Fig. 4).

Peak steering velocity, total distance walked and the error in angular orientation to each speaker were incorporated into the phonotaxis performance index (see methods) to develop a pulse rate preference function for Floridian O. ochracea (Fig. 5). This preference function varied

significantly with changes in song pulse rate (Repeated Measures ANOVA: F(1,10)= 48.30, p<0.001). The performance index increased gradually for pulse rates between 10 to 40 pps, reached a peak for pulse rates between 50 to 70 pps, and decreased with further increases in song pulse rates. Pairwise comparisions with Bonferroni corrected p values reveal that both 60 and 70 pps songs resulted in index values that did not differ significantly from values at 50 pps (50 vs 60

pps: p = 1.000, 50 vs 70 pps: p = 1.000). Song pulse rates above and below this range resulted in significantly reduced index values compared to the 50 pps song (50 vs 10 pps: p <0.001, 50 vs 20 pps: p <0.001, 50 vs 30 pps: p <0.001, 50 vs 40 pps: p =0.005, 50 vs 80 pps: p =0.001, 50 vs 90 pps: p <0.001).

2.5 Discussion

Host cricket song recognition by Floridian O. ochracea was examined using a spherical treadmill system that recorded tethered walking phonotaxis under open-sensory loop conditions. Flies were held at a fixed position relative to broadcasting speakers to maintain constant sensory feedback during stimulus presentation and locomotion. Song preferences were inferred based on measuring the ability of O. ochracea to track a switch in the broadcast location of songs that varied in pulse rates. One caveat of using this approach to study song preferences in O. ochracea is the possibility that preferences may be exerted only during flight, whereas landing and walking to the source is the outcome of such preferences (Wagner, 1998). Muller and Robert studied O. ochracea free-flight phonotaxis using a stereo camera tracking system to reconstruct flight trajectories during sound localization (Muller and Robert, 2001, 2002). In the initial phase of phonotaxis, flies quickly gained altitude during the take-off phase, maintained a steady altitude and approached the source with a slight meandering path during the cruising phase, and engaged in a spiral descent to land within 10 cm of the source location (Muller and Robert, 2001). Landing accuracy was greater in localizing songs with higher chirp rates, and flight trajectories showed less meander when sound pulses were organized in to chirps rather than continuous trills (Muller and Robert, 2002). However, source localization in free-flight is expected to alter perceived signal intensities with the result of confounding interpretation of preferences to particular signal features. To overcome this challenge, insects have been studied in tethered flight where measurements of flight posture may indicate steering tendencies that correspond to signal preferences (Moiseff et al., 1978). Tethered flight phonotaxis has been used to examine the acoustic startle response in field crickets and in O. ochracea. In tethered flight, both crickets and flies adopt steering maneuvers towards stimuli with the temporal pattern and frequency content of cricket calling songs (~4.5 kHz), but steer away from stimuli with ultrasonic frequencies (>20 kHz) that are characteristic of bat sonar. This range of frequencies along a continuum is perceptually categorized with a sharp boundary between 8 to 12 kHz (cricket-like calls, or bat-like echolocation). Frequencies across categories are well discriminated

24 by crickets and flies (Wyttenbach et al., 1996; Rosen et al., 2009). Crickets do not show discrimination between frequencies within a category, and it is not known if O. ochracea show this same discrimination (Rosen et al., 2009). Whether flying or walking, the frequency content and song temporal pattern are important for song recognition in field crickets (Pollack et al., 1984). O. ochracea have been demonstrated to show categorical frequency discrimination in flight, but does not appear to do so while walking (Rosen et al., 2009). In flight, some temporal features of cricket songs are evaluated for song recognition (Walker, 1993; Gray et al., 2007), but the specific temporal features are currently unknown. The goal of this current chapter was to develop a walking phonotaxis index that will allow for quantification of song recognition in O. ochracea.

Song recognition by Floridian O. ochracea was examined with an assay that measured tracking performance to songs that differed in pulse rates when the broadcast location was switched between two different speakers. From these phonotactic responses, measures of walking distance, peak steering velocity, instantaneous angular heading, and angular error in tracking speaker locations were assessed for suitability in describing tracking performance. All of these response features varied significantly to changes in song pulse rates. Walking distance increased for pulse rates between 10-30 pps, reached a level of saturation from 40-70 pps, and decreased with further increases in pulse rate. Mean angular heading to the initial broadcast location was less selective as flies oriented to a direction that approached the broadcast location for a broad range of pulse rates. Re-orientation responses however, were less accurate and flies only re-localized the song broadcast location for pulse rates between 50-70 pps. Flies steered to track the switch in broadcast locations and with the least amount of localization error for this same range of pulse rates when test responses were compared to those of the standard 50 pps song of G. rubens. Above and below the 50-70 pps range, peak steering velocity decreased while orientation error increased. Taken together, distance, peak steering velocity and error in angular orientation were suited for inclusion in the phonotaxis index.

Previously developed phonotaxis performance indices were precise in capturing response variation underlying song recognition in orthopteran insects (Schul, 1998) and anurans (Bush et al., 2002; Bee, 2008). The sound localization behaviour of orthopterans and anurans are characterized by turning responses to lateralize sound sources and this approach causes a meandering walking path about a direct route to the source location. Lateralization causes large

variation in the directedness of walking paths, greater total walking distances required to arrive at a source, and increased time required to locate a sound source, especially if lateralization introduce errors that require a greater number of corrective turns to reorient back on course to the source location. Lateralization responses are often punctuated by brief pauses where animals engaged in scanning behaviour to detect changes in sound direction before making subsequent responses (Murphey, 1972; Rheinlaender et al., 1979; Gerhardt and Rheinlaender, 1980; Rheinlaender and Blatgen, 1982). In the field cricket Gryllus bimaculatus, free-walking phonotaxis experiments demonstrate that sound sources with angles of incidence smaller than 20° azimuth relative to the midline caused consistent turning errors. This lead to the conclusion that meandering walking behaviour occurred to produce large angles of sound incidence beyond the directionally insensitive frontal region (Rheinlaender and Blatgen, 1982). However recent work has demonstrated that G. bimaculatus are able to resolve differences in sound angles of incidence as small as 1-2° under open-loop conditions, rivaling the directional acuity of O. ochracea (Schöneich and Hedwig, 2010). This mismatch between differences in the ability to resolve sound direction while free-walking compared to tethered open-loop conditions is currently unexplained but may be attributed to an artifact of open-loop experimentation (Schöneich and Hedwig, 2010). Several noteworthy differences exist in O. ochracea walking phonotaxis that may render previous performance indices insensitive for detecting response variation underlying song preferences. O. ochracea are able to lateralize source locations to within 1-2 degrees relative to their midline. Furthermore, O. ochracea respond with different angular headings for source locations that differ by only 1-2 degrees on the same side of the midline and thus are able to truly localize sound sources (Mason et al., 2001). This directional hyperacuity may contribute to eliminating walking meander and pauses required to evaluate changes in sound direction. Even for sound sources that dynamically change in the broadcast location O. ochracea appear to detect changes in sound direction instantaneously and engage in a single phonotactic response with turning adjustments that are uninterrupted by pauses. Consequently, the nature of O. ochracea sound localization would inflate performance values of “traditional” indices of phonotaxis and inadequately detect response differences attributed to song preferences (peak steering velocity and localization accuracy).

Choice paradigms involve the simultaneous presentation of two or more alternative stimuli to infer signal preferences according to the dichotomous outcome of whether or not a

signal is chosen. This has been a common strategy to test for discrimination of signals that differ in the relative attractiveness or the role of signals used to facilitate species recognition (Popov and Shuvalov, 1977; Ryan, 1980; Doherty, 1985; Gerhardt and Doherty, 1988; Scheuber et al., 2004). Occasionally, laboratory-based results from choice experiments may be inconsistent with preferences observed in the field (Passmore and Telford, 1983; Gerhardt, 1982). For example, the female painted reed frog displayed clear preferences for songs with a lower dominant frequency under dual choice laboratory conditions, but were observed to be less discriminatory in the field (Passmore and Telford, 1983; Dyson and Passmore, 1988). Inconsistencies may be attributed to the inability of animals to evaluate small signal differences when confronted with multiple simultaneous signals (Gerhardt, 1982). When O. ochracea are presented with identical overlapping stimuli, they localize a phantom sound source that is in between the actual source locations (Lee et al., 2009). It is expected that when presented with alternative signals that differ little in the temporal pattern, O. ochracea would respond similarly. Choice in this situation may depend on stochastic processes that bring subjects closer to a given source rather than biologically relevant preferences and this may lead to the interpretation that no preferences exist. Conversely, adopting the no-choice paradigm to test for song preferences transforms a dichotomous outcome into a graded measure that describes response strength for developing preference functions (Wagner, 1998). No-choice paradigms were originally perceived to provide little information regarding song preferences because animals tend to respond with less discrimination when presented without choice (Doherty, 1985; Wagner, 1998). In support of views presented by Bush and colleagues (Bush et al., 2002), we argue that no-choice paradigms can provide additive information in studying song recognition and signal preferences. In no- choice experiments, Bush and colleagues calculated a phonotaxis index based on measuring the ratio of response time to reach the source location for the control song compared to test songs (Bush et al., 2002). This index revealed variation in response timing to changes in song pulse rates that facilitate song recognition in two closely related species of treefrogs (Hyla chrysoscelis and Hyla versicolor) (Bush et al., 2002). By using a carefully selected phonotaxis task (tracking a switch in source location), our study demonstrates the opportunity to measure variation in tracking performance to develop a preference function that describes song recognition in O. ochracea.

In Floridian O. ochracea, this song recognition/preference function peaks for songs with pulse rates between 50-70 pps. Both G. rubens and G. texensis share overlapping geographic distributions from eastern Texas to western Florida (Milton). Sympatric males cannot be distinguished based on morphology but can be reliably identified based on calling song pulse rates (Walker, 1998, 2000). Both species are bivoltine with a surge in adult abundance in the spring and in the fall. The calling song pulse rates of G. rubens show seasonal variability where males sing at 48.6 pps in the spring and 58.6 pps in the fall (Walker, 1998). G. texensis males produce calling songs at a pulse rate of 77.6 pps in the fall (Walker, 1998). Acoustically orienting O. ochracea have not been found in the spring (Vélez and Brockmann, 2006), but sound trap experiments conducted in the fall demonstrate that O. ochracea prefer the songs of G. rubens over G. texensis (Walker, 1993; Gray et al., 2007). This song preference is a direct parallel to the pulse rate preference function that we describe in our study. If acoustically orienting O. ochracea were present in the spring, the 48.6 pps song of G. rubens would fall just outside of the range of preferred pulse rates. When O. ochracea are active in the fall, the 58.6 pps song of G. rubens is within the range of preferred pulse rates while the 78.6 pps song of G. texensis lie beyond this preferred range.

Although our data shows behavioural tuning to pulse rates, the specific question of the temporal features evaluated for song recognition remains unanswered because varying pulse durations and interpulse intervals in different combinations can produce equivalent pulse rates. The phonotaxis index that we present here will be instrumental to quantifying song recognition in O. ochracea. Experiments are currently underway to measure preference functions to an array of pulse duration and interpulse interval combinations to elucidate specific temporal features used in song recognition. These behavioural studies will guide future work in studying the neural basis of song recognition and sound localization in O. ochracea.

2.6 Acknowledgements

We would like to thank TJ Walker and MJ Kim Lee for assistance in fly collection, M Leung, S Susanto for their meticulous care in fly husbandry duties, MCB Andrade, J Peever, DR Howard, CL Hall, PA De Luca, MJ Kim Lee for helpful discussions. This project was funded by NSERC to AC Mason, OGS, NSERC PGS D3, and SICB grants-in-aid of research to N LEE.

References

Bee, M. A. (2007). Sound source segregation in grey treefrogs: spatial release from masking by the sound of a chorus. Animal Behaviour 74, 549-558.

Bush, S. L., Gerhardt, H. C. and Schul, J. (2002). Pattern recognition and call preferences in treefrogs (Anura : Hylidae): a quantitative analysis using a no-choice paradigm. Animal Behaviour 63, 7-14.

Cade, W. (1975). Acoustically orienting parasitoids - fly phonotaxis to cricket song. Science 190, 1312-1313.

Doherty, J. A. (1985). Phonotaxis in the cricket, Gryllus bimaculatus degeer - comparisons of choice and no-choice paradigms. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 157, 279-289.

Dyson, M. L. and Passmore, N. I. (1988). Two-choice phonotaxis in Hyperolius marmoratus (Anura, Hyperoliidae) - the effect of temporal variation in presented stimuli. Animal Behaviour 36, 648-652.

Gerhardt, H. C. ( 1982). Sound pattern recognition in some north american treefrogs (Anura, Hylidae): implications for mate choice. American Zoologist 22, 581-595.

Hedwig, B. and Poulet, J. F. A. (2004). Complex auditory behaviour emerges from simple reactive steering. Nature 430, 781-785.

Hedwig, B. and Poulet, J. E. A. (2005). Mechanisms underlying phonotactic steering in the cricket Gryllus bimaculatus revealed with a fast trackball system. Journal of Experimental Biology 208, 915-927.

Klump, G. M. (1995). Studying Sound Localization in Frogs with Behavioural Methods. In Methods in Comparative Psychoacoustics, pp. 221-233. Basel: Birkhauser Verlag.

Lee, N., Elias, D. O. and Mason, A. C. (2009). A precedence effect resolves phantom sound source illusions in the parasitoid fly Ormia ochracea. Proceedings of the National Academy of Sciences of the United States of America 106, 6357-6362.

Mason, A. C., Oshinsky, M. L. and Hoy, R. R. (2001). Hyperacute directional hearing in a microscale auditory system. Nature 410, 686-690.

Mason, A. C., Lee, N. and Oshinsky, M. L. (2005). The start of phonotactic walking in the fly Ormia ochracea: a kinematic study. Journal of Experimental Biology 208, 4699-4708.

Muller, P. and Robert, D. (2001). A shot in the dark: The silent quest of a free-flying phonotactic fly. Journal of Experimental Biology 204, 1039-1052.

Muller, P. and Robert, D. (2002). Death comes suddenly to the unprepared: singing crickets, call fragmentation, and parasitoid flies. Behavioral Ecology 13, 598-606.

Passmore, N. I. and Telford, S. R. (1983). Random mating by size and age of males in the painted reed frog, Hyperolius marmoratus. South African Journal of Science 79, 353-355.

Pollack, G. S., Huber, F. and Weber, T. (1984). Frequency and temporal pattern-dependent phonotaxis of crickets (Teleogryllus oceanicus) during tethered flight and compensated walking. Journal of Comparative Physiology 154, 13-26.

Rheinlaender, J. and Blatgen, G. (1982). The precision of auditory lateralization in the cricket, Gryllus bimaculatus. Physiological Entomology 7, 209-218.

Rheinlaender, J., Gerhardt, H. C., Yager, D. D. and Capranica, R. R. (1979). Accuracy of phonotaxis by the green treefrog (Hyla cinerea). Journal of Comparative Physiology 133, 247- 255.

Robert, D., Miles, R. N. and Hoy, R. R. (1996). Directional hearing by mechanical coupling in the parasitoid fly Ormia ochracea. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 179, 29-44.

Rosen, M. J., Levin, E. C. and Hoy, R. R. (2009). The cost of assuming the life history of a host: acoustic startle in the parasitoid fly Ormia ochracea. Journal of Experimental Biology 212, 4056-4064.

Schöneich, S. and Hedwig, B. (2010). Hyperacute Directional Hearing and Phonotactic Steering in the Cricket (Gryllus bimaculatus deGeer). PLoS One 5, e15141.

Schul, J. (1998). Song recognition by temporal cues in a group of closely related bushcricket species (genus Tettigonia). Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 183, 401-410.

Schul, J. and Bush, S. L. (2002). Non-parallel coevolution of sender and receiver in the acoustic communication system of treefrogs. Proceedings of the Royal Society of London. Series B: Biological Sciences 269, 1847-1852.

Schul, J., von Helversen, D. and Weber, T. (1998). Selective phonotaxis in Tettigonia cantans and T-viridissima in song recognition and discrimination. J. Comp. Physiol. A -Neuroethol. Sens. Neural Behav. Physiol. 182, 687-694.

Ulagaraj, S. M. and Walker, T. J. (1973). Phonotaxis of Crickets in Flight: Attraction of Male and Female Crickets to Male Calling Songs. Science 182, 1278-1279.

Vélez, M. J. and Brockmann, H. J. (2006). Seasonal variation in selection on male calling song in the field cricket, Gryllus rubens. Animal Behaviour 72, 439-448.

von Helversen, D. (1984). Parallel processing in auditory pattern recognition and directional analysis by the grasshopper Chorthippus Biguttulus L. (Acrididae). Journal of Comparative Physiology 154, 837-846.

Wagner, W. E. (1998). Measuring female mating preferences. Animal Behaviour 55, 1029- 1042.

Walker, T. J. (1993). Phonotaxis in female Ormia ochracea (Diptera, Tachinidae), a parasitoid of field crickets. J. Insect Behav. 6, 389-410.

Walker, T. J. (1998). Trilling field crickets in a zone of overlap (Orthoptera : Gryllidae : Gryllus). Ann. Entomol. Soc. Am. 91, 175-184.

Weber, T., Thorson, J. and Huber, F. (1981). Auditory behaviour of the cricket: 1. Dynamics of compensated walking and discrimination paradgism on the kramer treadmill. Journal of Comparative Physiology 141, 215-232.

Wyttenbach, R. A., May, M. L. and Hoy, R. R. (1996). Categorical perception of sound frequency by crickets. Science 273, 1542-1544.

Figure Legends

Figure 1. Effects of song pulse rate on walking distance. Bars show walking distance as mean±s.e.m. There was a significant main effect of pulse rate on walking distances largely attributed to significantly shorter walking distances for songs at 10 pulses/sec relative to all other pulse rates. Increases in song pulse rates from 10 to 40 pulses/sec resulted in a general trend of increased walking distances, saturated at pulse rates between 50 to 70 pulses/sec, and decreased with further increases in pulse rates.

Figure 2. Effects of song pulse rate on steering velocity in tracking the switch in song location. Steering velocity is measured as changes in x cm over time. Negative values indicate steering to the left, and positive values indicate steering to the right. A) (i) Flies steered to the left to track the initial 500 msec of song broadcast. This was followed by a steering response to the right to track the switch in broadcast location for the remaining 500 msec of song. The magnitude of steering responses varied with song pulse rate (colours). Steering to the initial broadcast location was less selective while steering to the switched location showed greater selectivity to songs with pulse rates that range from 50 to 70 pulses/sec. Reversed speaker presentation order produced similar results (ii). Values are given as mean steering velocity. B) Bars show peak steering velocity as mean±s.e.m. in response to song broadcast that switched from (i) the left to the right, and (ii) from the right to the left. Peak steering velocity increased for pulse rates between 10 to 40 pulses/sec, reached a peak between 50 to 70 pulses/sec, and decreased with further increase in pulse rates. Flies showed less selectivity (large steering responses) in response to the initial broadcast location, but were more selective in response to a switch in broadcast location with large steering responses only for pulse rates between 50 to 70 pulses/sec.

Figure 3. Effects of song pulse rate on median angular heading in tracking the switch in song location. Coloured circles show means of median angular heading for each fly in response to changes in song broadcast location and song pulse rate (inner ring of black dots: 10 pulses/sec to out ring of red circles: 90 pulses/sec). Coloured lines represent the von Mises kernel density estimates of the of the overall distribution of median angular headings for each song pulse rate. A) Median angular headings to the (i) initial song broadcast (left speaker) and to the (ii) switched broadcast location (right speaker). B) Median angular headings to the (i) initial song broadcast (right speaker) and to the (ii) switched broadcast location (left speaker). Median angular

32 headings to the initial broadcast location were less selective with flies localizing a direct that approached the song broadcast location for a large range of pulse rates. Following a switch in broadcast location, re-orientation responses were more selective. Responses to songs with pulse rates between 50 to 70 pulse/sec approached the broadcast speaker location, but pulse rates beyond this range caused flies to localize a forward direction.

Figure 4. Effects of song pulse rate on the error in median angular orientation. Error in median angular heading was calculated as the absolute difference between the global median angular heading to the 50 pulses/sec song relative to songs at all other pulse rates. Hatched bars represent the error in angular orientation to the initial broadcast location and solid bars represent the error in angular orientation to the switched broadcast location. Flies demonstrated greater localization error following a switch song broadcast location compared to responses to the initial broadcast location.

Figure 5. Effects of song pulse rate on the phonotaxis performance index. Phonotaxis performance index values incorporated peak steering velocity (measurement of tracking motivation) and error in angular orientation (measurement of localization accuracy) to each song broadcast location. The index is computed as the relative difference between responses to the 50 pulses/sec song compared to all other test songs. Index values of 0 indicate poor performance, 1 represent equivalent performance to the 50 pulses/sec song, and >1 indicate performance better than those to the 50 pulses/sec song. Lowest values of the performance index occurred for the 10 pulses/sec song. The performance index gradually increased to a peak for songs between 50 to 70 pulses/sec and decreased with further increases in pulse rates.

Table 2.1

Chapter 3 Convergent Temporal Features Evaluated for Song Recognition in a Host-Parasitoid Relationship 3.1 Abstract

In conspecific communication, sender-receiver co-evolution is expected to match signal characteristics and corresponding sensory mechanisms (Bradbury and Vehrencamp, 1998). In contrast, eavesdropping predators and parasitoids must evolve sensory structures and processing mechanisms to identify salient features in pre-existing signals, and their ability to do so will impose negative selection on exploited signalers. Understanding whether intended receivers and unintended eavesdroppers share convergent sensory mechanisms or arrive at divergent solutions may identify important constraints on the processing of sensory information. A first step to investigating this problem is to determine what features of signals are being evaluated. Male field crickets produce calling songs with species-specific temporal patterns to attract to female crickets in pair formation. Several species of field crickets in the genus Gryllus rely on evaluating song pulse periods for temporal pattern recognition. The acoustic parasitoid fly O.ochracea has evolved to exploit these communication signals to locate hosts for the development of their larval young. Specific temporal features of cricket songs that O. ochracea uses to identify appropriate host species are unknown. In no-choice experiments with Floridian O. ochracea tethered on top of a spherical treadmill system, we measured the ability O. ochracea to track a switch in the broadcast location of test songs that varied in combinations of pulse durations and interpulse intervals. Recorded walking distance showed that flies engaged in walking phonotaxis to a wide range of song types provided that pulse durations were less than 15 ms and interpulse intervals were greater than 5 ms. Flies demonstrated greatest peak steering velocities and best localization accuracy to a narrower selection of song types with pulse durations and interpulse intervals that combine to pulse periods of 14-20 ms. Using our recently developed phonotaxis performance index (chapter 2), flies showed best phonotaxis performance to this same range of pulse periods and thus, Floridian O. ochracea evaluates the pulse period of calling songs to identify appropriate host crickets. Our results demonstrate a convergence of song temporal features evaluated for song recognition.

3.2 Introduction

Receiver mechanisms for exploiting heterospecific communication signals must evolve within the constraints of pre-existing signal features. Sexual signals used in acoustic communication are subjected to intense sexual selection to match the preferences of intended receivers, thus resulting in their often conspicuous nature (Endler, 1993; Andersson, 1994). One consequence of this is that conspicuous signals are often exploited by predators in search of prey, or parasitiods in search of hosts (Zuk and Kolluru, 1998). Unintended receivers must evolve specialized sensory structures and processing mechanisms to exploit these heterospecific signals, but there may be a trade-off between specializations for efficient detection of specific prey or host signals and the ability to detect and process a wider range of possible targets. Both intended receivers and unintended eavesdroppers are required to perform the same tasks to locate signalers, and may even share similar preferences for signal features (Zuk and Kolluru, 1998). Determining whether similar processing mechanisms underlie these preferences may provide valuable insights into understanding constraints placed on the evolution of sensory exploitation by eavesdroppers.

Insect songs often lack melody but instead, are comprised of monotonous sound pulses repeated in a stereotypic temporal pattern (Gerhardt and Huber, 2002). With few exceptions (Fonseca et al., 2000), insect auditory systems also show a limited capacity to perform spectral analyses (Pollack, 2000; Gerhardt and Huber, 2002; Hennig et al., 2004; Mason and Faure, 2004), but are adept at processing temporal information (Hennig et al., 2004; Ronacher et al., 2004). Some of the fine-scale temporal features evaluated for species recognition include detection of appropriate pulse duration (Hennig, 2003), interval between sound pulses (interpulse intervals) (von Helversen and von Helversen, 1997), pulse rate or pulse period (Thorson et al., 1982; Schul, 1998; Hennig, 2003; Poulet and Hedwig, 2005), pulse envelope shape (Vonhelversen, 1993; von Helversen and von Helversen, 1997), and sometimes these temporal features in different combinations (Schul, 1998). In the field crickets Gryllus campestris and Gryllus bimaculatus, best phonotaxis performance of females to conspecific male calling songs occur for a range of pulse rates of 20 to 50 pulses s-1 (Thorson et al., 1982; Poulet and Hedwig, 2005). Above and below this range of preferred pulse rates, a decline in phonotaxis performance is observed (Thorson et al., 1982; Poulet and Hedwig, 2005). Similar pulse rates may arise from a range of pulse duration and/or interpulse interval combinations that sum up to a particular pulse

period (1/rate). Therefore a behavioural preference for a range of pulse rates still allows alternative mechanisms for processing auditory stimulus parameters. In some cases, closely related cricket species appear to evaluate different temporal parameters for song recognition. For example, Teleogryllus oceanicus prefers a specific pulse period while T. commodus prefers a particular pulse duration (Hennig, 2003). As in G. bimaculatus, detection of the correct song pulse rate may increase auditory responsiveness and modulate steering towards a source location (Poulet and Hedwig, 2005). Similar song recognition and preference studies are lacking for unintended receivers that also localize similar targets.

The tachinid fly Ormia ochracea is an acoustic parasitoid of field crickets (Cade, 1975). Reproductive success in O. ochracea depends on finding suitable cricket hosts for the development of their larval young and host localization is based on using directional hearing to pinpoint the location of cricket calling songs (Cade, 1975). Detection of calling songs induce gravid female O. ochracea to perform flying (Cade, 1975; Muller and Robert, 2001) and walking (Mason et al., 2005) phonotaxis to the source location. Upon arrival, first instar planidia (larvae) are deposited near or on top of host crickets where they subsequently burrow inside the cricket to feed on fat body and muscle tissue (Cade, 1975; Wineriter and Walker, 1990; Adamo et al., 1995).

O. ochracea has evolved several auditory specializations that contribute to successful host localization. They possess a pair of tympanal ears that are directionally sensitive and mechanically tuned to respond to the carrier frequency of cricket calling songs (Robert et al., 1996a). Walking phonotaxis experiments reveal that O. ochracea are capable of localizing sound sources to the accuracy of 2° azimuth (Mason et al., 2001). This directional hyperacuity is remarkable given that the small size of O. ochracea severely limits two conventional cues for computing sound direction. With the two tympana separated by only 0.5 cm (Robert et al., 1996b), this results in no diffraction of the 7cm wavelength cricket calling song and thus, sound field interaural intensity (level) differences (ILD) are nonexistent (0 dB) while interaural time differences are immeasurably small (<1.5 µsec) (Robert et al., 1996a). To overcome these impoverished directional cues, O. ochracea has evolved a pair ears that are mechanically coupled and specialized to amplify small sound field ITDs into significantly larger mechanical ILDs and ITDs that are sufficient for neural processing (Robert et al., 1996a, 1998). Further specializations are apparent at the level of the primary auditory afferents. Similar to the

mechanical responses of the tympanal memebranes, these afferents are tuned to the frequency of cricket calling songs (Oshinsky and Hoy, 2002). Suprathreshold stimulation elicits most of the auditory afferents (Type 1) to respond with a single spike time-locked to the onset of sound pulses regardless of changes in intensity or pulse duration (Oshinsky and Hoy, 2002). Response latencies vary as a function of sound intensity such that they decrease to an increase in intensity, and at any given intensity, there is little variation in spike-timing (low jitter) (Oshinsky and Hoy, 2002). Collectively, these response characteristics allow for, 1) the encoding of sound direction as binaural mean latency differences that vary with incident sound direction, and 2) the encoding of song temporal pattern as the primary afferent spike train (Mason et al., 2001; Oshinsky and Hoy, 2002).

In the field, O. ochracea occur in several geographically distinct regions in the United States, with each population demonstrating behavioural specializations for different host cricket species that produce calling songs with different temporal patterns (Gray et al., 2007). O. ochracea prefers to parasitize Gryllus rubens in Florida (Walker, 1993), Gryllus texensis in Texas, Gryllus lineaticeps in California (Wagner, 1996), and Teleogryllus oceanicus in Hawaii (Zuk et al., 1995). Both female G. lineaticeps and gravid female O. ochracea show convergent preferences for male cricket calling songs with higher chirp rates, longer chirp durations, and higher chirp amplitudes (Wagner, 1996). Despite the numerous auditory specializations that O. ochracea possess for the recognition and localization of host crickets, to what extent these specializations impose constraints on the recognition of a broader range of potential host species is unknown. Previous work has demonstrated that Floridian O. ochracea prefer songs with pulse rates that range from 40 to 70 pulses s-1 (Walker, 1993), and this encompasses the 58.6 pulses s-1 calling song of G. rubens (Walker, 1998). At these pulse rates, flies localized songs with greater walking distances, stronger steering responses, and better accuracy compared to songs at higher or lower pulse rates (Lee et. al, unpublished data).

In this study we explicitly test the hypothesis that Floridian O. ochracea song recognition is exclusively based on evaluating the pulse rate or pulse period (i.e. flies will respond to a range of pulse duration and interpulse interval combinations that combine to a specific range of pulse periods). Using our previously developed phonotaxis performance index and a no-choice paradigm, we measure the ability of flies to track a switch in the broadcast location of model songs that vary in different combinations of pulse durations and interpulse intervals. We describe

the acoustic recognition space of Floridian O. ochracea and demonstrate that song recognition is based on evaluating the calling song pulse period.

3.3 Materials and Methods 3.3.1 Animals

3.3.2 Acoustic Stimuli

The standard control song was modeled after the calling song of Gryllus rubens, which are trills constructed from 10 ms duration 5 kHz tone sound pulses (1ms on/off ramps) separated by 10 ms interpulse intervals and repeated at 50 pulses/sec for a total stimulus duration of 1 second. All test stimuli (model songs) were 1 second in duration. Preferences were examined with songs that varied in pulse durations and interpulse intervals (2, 5, 10, 15, 20, 30 mss) in different combinations to result in songs with different pulse rates/pulse periods and duty cycle. Song intensity (dB SPL) was adjusted to account for differences in total acoustic energy.

Stimulus waveforms were synthesized in Matlab (R2009b, The MathWorks Inc. USA) with custom software and converted to analog signals using National Instruments data acquisition hardware (NI USB-6251, 44100 Hz), amplified with Radio Shack Realistic (SA-10 Solid State Amplifier MOD-31-1982B, Taiwan or NAD amp) and broadcast through piezo electric tweeters. Stimulus levels were controlled with programmable attenuators (Tucker Davis Technologies System 3 PA5) and calibrated using a probe microphone (B&K Type 4182, Denmark) powered by B&K Nexus Conditioning Amplifier (Denmark).

3.3.3 Experimental Apparatus

44 light-weight table tennis ball held afloat by a constant airstream above a modified optical mouse sensor (ADNS 2620, Avago Technologies, USA). Walking responses were transduced as rotations of the trackball that actuated the optical mouse sensor to record changes in x and y pixel units at a sampling rate of 2160 Hz (Lott et al., 2007). Pixel units were calibrated to actual walking distances by measuring displacement of points on the ball from high-speed video footage (DRS Lightening RDT, 500 frames per second) synchronized to pixel data captured by the trackball system. Data collection by the trackball system was controlled by custom Matlab software that interfaced with the National Instruments data acquisition system to ensure synchronous sound presentation and the recording of walking traces.

3.3.4 Protocol

A test sequence commenced with a presentation of the standard cricket song from both left (-45°) and right (+45°) speakers synchronously, followed by the presentation of the standard song from the left speaker, then the right speaker in isolation. To control for possible effects of positive/negative priming by a preferred song, 1) test responses were primed with the standard cricket song presented from both left and right speakers at an overall intensity of 76 dB SPL, and 2) flies were given 30 seconds of rest between stimulus presentations. Following the priming stimulus, a test song was broadcast from one speaker for 500 ms and switched to the other speaker for the remaining 500 ms of presentation. The same test song was presented for a second iteration, but with the speaker presentation order reversed (i.e. left/right to right/left). This process was repeated for all test songs with the sequence of test songs and the order of first speaker presentation randomized. Experiments ended with the same sequence of standard song presentations that was initially presented in the beginning (above).

3.3.5 Data Analysis

speaker) was calculated for tracking the standard song (10 ms pulse duration, 10 ms interpulse interval, 50 pulses/sec). This value was used as the reference angular heading to each speaker. Error in angular orientation (measure of angular distance to relative to the average reference response) for each response to a test song was determined as the absolute difference between the reference angular heading and the median angular heading to each test speaker. The error in angular orientation measured for the standard song condition (Fig. 4) captures the angular response variability to the control song. The phonotaxis performance index incorporated distance travelled, steering velocity, and error in angular orientation and was calculated as:

Reference conditions refer to measurements in response to the standard song and test conditions refer to measurements in response to test songs. Index values range from 0 to >1. A phonotaxis performance index of 0 indicates poor performance, 1 indicates performance equivalent to responses to the standard song, >1 indicates performance better than responses to the standard song. Repeated responses for the same stimulus conditions were averaged within individuals. Data are given as means ± SD. Phonotaxis response arrays were created using SigmaPlot (v10.0, Systat Software Inc., USA).

3.4 Results

Response motivation remained consistent throughout the randomized sequences of song models presented. In response to the initial broadcast of the control G. rubens calling song, flies covered a mean distance of 4.61±1.59 cm that did not significantly differ from the mean walking distance of 4.18±1.58 cm to the final presentation of the control song (Wilcoxon Signed Rank Test: N=17, Z=-1.633, p=0.102).

Flies performed walking phonotaxis to a wide range of song types with pulse durations less than 15 to 20 ms and interpulse intervals greater than 5 ms (Fig. 1). A localized region of significantly greater walking distances occurred to a range of pulse duration and interpulse interval combinations that sum up to pulse periods that are approximately 17 to 20 ms.

In tracking the initial song broadcast location, flies displayed greatest peak steering velocities to a small range of pulse duration and interpulse interval combinations with pulse periods from 15- 20 ms (Fig. 2a). Slightly greater peak steering velocities were achieved in response to the initial song broadcast location than compared to responses to the switched broadcast location (more red in Fig. 2a compared to Fig. 2b). Flies generally displayed steering responses to a slightly smaller range of song types in the re-orientation response (Fig 2b).

Flies exhibited best localization accuracy (smallest amount of orientation error) to song types with a pulse period of 14-20 ms (Fig. 3). In response to the initial song broadcast location flies exhibited a considerable amount of orientation error to a broad range of song types (Fig. 3a yellow and green regions). A switch in song broadcast location caused further reduction in orientation accuracy for a greater range of song types (Fig. 3b, blue). Re-orientation responses with the best level of accuracy only occurred to song types with pulse durations less than 15 ms and with interpulse intervals between 5-15 ms that combine to pulse periods around 20 ms (Fig. 3b, red).

With all three response features combined to derive the phonotaxis performance index, flies engaged in best phonotaxis performance to song types with pulse periods between 15 to 20 ms (Fig. 4).

3.5 Discussion

Our results demonstrate the nature of the cricket calling song recognition in Floridian O. ochracea. Walking phonotaxis occurred to a range of song types with pulse durations less than 15 ms and intervals between sound pulses greater than 5 ms. O. ochracea tracked songs with the greatest level of motivation (steering magnitude) and localized sources with greatest accuracy when pulse durations and interpulse intervals combined to form pulse periods between 14 to 20 ms. Using our previously developed phonotaxis performance index, Floridian O. ochracea displayed best phonotaxis to this range of pulse periods.

Inferences on O. ochracea song preferences have mainly been derived from field sound- trap experiments that establish capture rates for different song types. In Florida, these experiments demonstrate that O. ochracea prefer songs with a carrier frequency between 4-5 kHz and amplitude modulated to result in pulse rates between 40 to 70 pulses/sec (Walker, 1993). Other sound trap experiments have examined preferences for G. rubens calling songs compared to songs of other potential host cricket species that occur in sympatry with Floridian O. ochracea. Fly capture rates were greatest for sound traps broadcasting G. rubens calling songs while all other alternatives were negligible (Walker, 1993). To examine geographic variation in host preferences, O. ochracea of a given locale were subjected to songs of their primary local host cricket species compared to songs of other cricket species parasitized in a different part of their range (Gray et al., 2007). Different populations of O. ochracea were found to exhibit specialized preferences for different host cricket species (Gray et al., 2007). In descending order of preferences, Floridian O. ohcracea were most attracted to the calling song of G. rubens, G. texensis, G. lineaticeps, and least attracted to the songs of T. oceanicus (Gray et al., 2007). We argue that this trend in song preferences may reflect a temporal pattern recognition mechanism that is based on the ability of the auditory system to measure the timing of sound pulses occurring within an acceptable range of intervals that make up a range of pulse periods.

Our current understanding of the neural basis of temporal pattern recognition in O. ochracea is limited to the sensory periphery. Physiological response characteristics of primary auditory afferents in O. ochracea may impose neural constraints on possible temporal features evaluated for song recognition. Most of the primary auditory afferents are specialized to exhibit phasic responses with one (type 1), or 2 to 4 spikes (type 2) at the onset of sound pulses and exhibit refractory periods of approximately 4.26 ms (Oshinsky and Hoy, 2002). Primary afferents exhibiting tonic responses to indicate the duration of sound pulses (type 3) have rarely been encountered (Oshinsky and Hoy, 2002). This implies song types with interpulse intervals shorter than 4 ms will be filtered at the sensory periphery. The behavioural results we present here support this view as flies responded to pulse durations as short as 2 ms, but failed to engage in phonotaxis when interpulse intervals were shorter than 4 ms. Evaluating interpulse intervals for song recognition, would require precise indication of the offset of sound pulses and this may be accomplished with primary auditory afferents that spike at the end of sound pulses, or that spike continuously for the duration of sound pulses. Both these response features seem to be

limited in O. ochracea (Oshinsky and Hoy, 2002). Consequently, evaluating pulse periods with minimum gap durations greater than the receptor refractory period, may be the most generalized peripheral song recognition mechanism that adheres to these neural constraints.

The range of accepted pulse rates provides some level of flexibility in the recognition of a range of host species. The calling song of G. rubens is characterized by pulse durations of 9.75 ms and interpulse intervals of 10.25 ms for a total pulse period of 20 ms (Walker, 1998). G. texensis calling songs exhibit similar pulse durations (9.25 ms) but shorter interpulse intervals (4.75 ms) that combine to a shorter pulse period (14 ms) (Walker, 1998). G. lineaticeps calling songs exhibit pulse durations (7 ms) and interpulse intervals (7 ms) that also result in matching pulse periods (14 ms) (Wagner, 1996). Our results show that G. texensis and G. lineaticeps calling songs fall within the accepted range of pulse periods to result in significant phonotactic performance (Fig. 4). Greater preferences for G. texensis calling songs, however, may reflect differences in the continuity of pulse trains and the total acoustic energy of these songs types. This is because the G. texensis calling song consists of approximately 50 pulses organized into trills that are separated by brief pauses. Conversely, the G. lineaticeps calling song is comprised of 7-8 sound pulses grouped into shorter chirps and with chirps separated by longer intervals. A lack of preference for the T. oceanicus song type is likely caused by the significantly longer pulse durations and interpulse intervals that combine to pulse periods (70-90 ms) (Walker and Cade, 2003) and are beyond the recognition space of Floridian O. ochracea.

Female O. ochracea and female field crickets share some striking similarities in their behavioural correlates of temporal pattern recognition. For example, studies in G. campestris and G. bimaculatus reveal that field crickets readily track the location of songs consisting of sound pulses with a carrier frequency of 5 kHz that are repeated at some regular interval (Thorson et al., 1982; Doherty, 1985). Phonotaxis occurs to a range of pulse rates or pulse periods but best performance is only observed to songs with pulse rates near a species-specific value (Thorson et al., 1982; Doherty, 1985; Poulet and Hedwig, 2005). Much like O. ochracea (Walker, 1993; Wagner, 1996), tracking of song location did not depend on whether pulses were organized into chirps or trills (Thorson et al., 1982). When the duty cycle within chirps were altered, crickets tracked chirps with sound pulses as brief as 2 ms and tracking diminished with interpulse intervals shorter than 4-5 ms (Thorson et al., 1982). In addition to sharing convergent preferences for male cricket calling songs with higher chirp rates and longer chirp durations

(Wagner, 1996), our results show that female O. ochracea and female field crickets evaluate the same song features for temporal pattern recognition.

Insect model systems generally show low-level temporal filtering in the auditory periphery as the result of receptor refractory periods (Nocke, 1972; Hutchings and Lewis, 1981). In field crickets, two auditory interneurons (Omega neuron 1: ON1 and, Ascending neuron 1: AN1) in the prothoracic ganglion receive input from primary auditory afferents (Wohlers and Huber, 1985). Further temporal filtering may occur as the result of an intrinsic ON1 response feature. Prolonged excitation of ON1in response to a pulse train leads to an intensity-dependent long-lasting hyperpolarization (Pollack, 1986). Based on measurements of instantaneous spike rates, ON1, this long-lasting hyperpolarization may contribute to a low-pass temporal filter and the encoding of species-specific pulse rates (Nabatiyan et al., 2003). AN1 reliably encode and relay temporal patterns to higher order interneurons in the brain where evidence for temporal pattern filtering emerges (Schildberger, 1984). In the brain, a class of brain neurons called BNC1 show low-, high-, and band-pass characteristics that correspond to behaviour (Schildberger, 1984). In O. ochracea, type 1 afferents are also expected to reliably copy the calling song temporal pattern of T. oceanicus. However, the song recognition space (this study) and field experiments reveal that this song type is not preferred by Floridian O. ochracea (Gray et al., 2007), suggesting that higher order auditory neurons may be responsible for temporal filtering. Neural measurements from interneurons in the thoracic ganglion and in the brain of O. ochracea will be required to establish if, 1) temporal pattern processing in O. ochracea conforms to this idea and, 2) similar neural mechanisms may be responsible for convergent song features evaluated in song recognition in O. ochracea and their host crickets.

Another intriguing question that arises from our current work is whether or not O. ochracea in different parts of their geographic range share identical temporal pattern recognition mechanisms. Walker originally suggested that O. ochracea utilizing different host cricket species may possess different song recognition templates (Walker, 1993). Based on evaluating pulse periods alone, Floridian O. ochracea is expected to recognize the calling songs of primary host crickets in Florida, Texas, and California, but not host crickets parasitized on the Hawaiian Islands. Comparing song recognition between Floridian and Hawaiian O. ochracea will be a direct test of whether song recognition mechanisms remains the same or if these mechanisms have recently diverged.

3.6 Acknowledgements

We would like to thank TJ Walker and MJ Kim Lee for assistance in fly collection, M Leung, S Susanto, for their meticulous care in fly husbandry duties, MCB Andrade, J Peever, DR Howard, CL Hall, PA De Luca, MJ Kim Lee for helpful discussions. This project was funded by NSERC to AC Mason, OGS, NSERC PGS D3, and SICB grants-in-aid of research to N LEE.

References

Andersson, M. (1994). Sexual Selection. Princeton, New Jersey: Princeton University Press.

Bradbury, J. and Vehrencamp, S. (1998). Principles of animal communication. Sunderland, Mass: Sinauer Associates.

Cade, W. (1975). Acoustically orienting parasitoids - fly phonotaxis to cricket song. Science 190, 1312-1313.

Endler, J. A. (1993). Some general comments on the evolution and design of animal communication systems. Philosophical Transactions of the Royal Society of London Series B- Biological Sciences 340, 215-225.

Fonseca, P. J., Munch, D. and Hennig, R. M. (2000). Auditory perception - How cicadas interpret acoustic signals. Nature 405, 297-298.

Gerhardt, H. C. and Huber, F. (2002). Acoustic Communication in Insects and Anurans: Common problems and diverse solutions. Chicago: The University of Chicago Press.

Hennig, R. M. (2003). Acoustic feature extraction by cross-correlation in crickets? J. Comp. Physiol. A -Neuroethol. Sens. Neural Behav. Physiol. 189, 589-598.

Hennig, R. M., Franz, A. and Stumpner, A. (2004). Processing of auditory information in insects. Microsc. Res. Tech. 63, 351-374.

Hutchings, M. and Lewis, B. (1981). Response properties of primary auditory fibers in the cricket Teleogryllus oceanicus (Le Guillou). Journal of Comparative Physiology 143, 129-134.

Mason, A. C. and Faure, P. A. (2004). The physiology of insect auditory afferents. Microsc. Res. Tech. 63, 338-350.

Mason, A. C., Oshinsky, M. L. and Hoy, R. R. (2001). Hyperacute directional hearing in a microscale auditory system. Nature 410, 686-690.

Mason, A. C., Lee, N. and Oshinsky, M. L. (2005). The start of phonotactic walking in the fly Ormia ochracea: a kinematic study. Journal of Experimental Biology 208, 4699-4708.

Muller, P. and Robert, D. (2001). A shot in the dark: The silent quest of a free-flying phonotactic fly. Journal of Experimental Biology 204, 1039-1052.

Nabatiyan, A., Poulet, J. F. A., de Polavieja, G. G. and Hedwig, B. (2003). Temporal pattern recognition based on instantaneous spike rate coding in a simple auditory system. Journal of Neurophysiology 90, 2484-2493.

Nocke, H. (1972). Physiological aspects of sound communication in crickets (Gryllus campestris L.). Journal of Comparative Physiology 80, 141-&.

Oshinsky, M. L. and Hoy, R. R. (2002). Physiology of the auditory afferents in an acoustic parasitoid fly. Journal of Neuroscience 22, 7254-7263.

Pollack, G. (2000). Who, what, where? recognition and localization of acoustic signals by insects. Current Opinion in Neurobiology 10, 763-767.

Pollack, G. S. (1986). Discrimination of calling song models by the cricket, Teleogryllus oceanicus: the influence of sound direction on neural encoding of the stimulus temporal pattern and on phonotactic behaviour. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 158, 549-561.

Ronacher, B., Franz, A., Wohlgemuth, S. and Hennig, R. M. (2004). Variability of spike trains and the processing of temporal patterns of acoustic signals-problems, constraints, and solutions. J. Comp. Physiol. A -Neuroethol. Sens. Neural Behav. Physiol. 190, 257-277.

Schildberger, K. (1984). Temporal selectivity of identified auditory neurons in the cricket brain. Journal of Comparative Physiology 155, 171-185.

Vonhelversen, D. (1993). Absolute steepness of ramps as an essential cue for audiotory pattern recogntion by a grasshopper (Orthoptera, Acrididae, Chorthippus biguttulus L.). Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 172, 633-639.

Wagner, W. E. (1996). Convergent song preferences between female field crickets and acoustically orienting parasitoid flies. Behavioral Ecology 7, 279-285.

Walker, S. E. and Cade, W. H. (2003). The effects of temperature and age on calling song in a field cricket with a complex calling song, Teleogryllus oceanicus (Orthoptera : Gryllidae). Can. J. Zool.-Rev. Can. Zool. 81, 1414-1420.

Walker, T. J. (1993). Phonotaxis in female Ormia ochracea (Diptera, Tachinidae), a parasitoid of field crickets. J. Insect Behav. 6, 389-410.

Walker, T. J. (1998). Trilling field crickets in a zone of overlap (Orthoptera : Gryllidae : Gryllus). Ann. Entomol. Soc. Am. 91, 175-184.

Wineriter, S. A. and Walker, T. J. (1990). Rearing phonotactic parasitoid flies (Diptera, Tachinidae, Orminii, Ormia spp.). Entomophaga 35, 621-632.

Wohlers, D. W. and Huber, F. (1985). Topographical organization of the auditory pathway within the prothoracic ganglion of the cricket Gryllus campestris L. Cell Tissue Res. 239, 555- 565.

Zuk, M. and Kolluru, G. R. (1998). Exploitation of sexual signals by predators and parasitoids. Q. Rev. Biol. 73, 415-438.

Zuk, M., Simmons, L. W. and Rotenberry, J. T. (1995). Acoustically-orienting parasitoids in calling and silent males of the field cricket Teleogryllus oceanicus. Ecol. Entomol. 20, 380-383.

Figure Legends

Figure 1. Effects of varying pulse duration and interpulse interval on walking distance. Heat map show mean walking distance. Flies walked to a broad range of song types for pulse durations less than 15 ms and interpulse intervals greater than 5 ms. Greatest walking distances occurred to song types with pulse durations and interpulse intervals that combine to 14-20 ms (black elliptical outline).

Figure 2. Effects of varying pulse duration and interpulse interval on peak steering velocity. Peak steering velocity is measured as changes in x cm over time. Greatest peak steering velocities were observed for song types with pulse durations and interpulse intervals that approximate to pulse periods from 14-20 ms (black elliptical outlines). Greater peak steering velocities were observed to tracking the initial broadcast location (a) than compared to the (b) subsequent broadcast location. Flies also appear to respond to a smaller range of pulse periods in the subsequent localization response (smaller elliptical outline).

Figure 3. Effects of varying pulse duration and interpulse interval on accuracy in tracking source locations. Orientation error is measured in degrees as the angular distance between the mean angular orientation to the standard song (10 ms pulse duration and 10 ms interpulse interval) and the angular orientation to the test song. Smaller angles (small angular distance, in red) represents greater accuracy in orientation. In tracking the initial song broadcast location (a), flies displayed best accuracy to song types that approximate pulse periods between 14-20 ms (black elliptical outline). In tracking the subsequent broadcast location, orientation accuracy remained high for a smaller range of pulse periods (smaller black elliptical outline), but orientation accuracy to all other song types generally decreased compared to tracking the initial broadcast location.

Figure 4. Using the phonotaxis performance index to measure response variation to varying pulse duration and interpulse interval. The phonotaxis performance index can range from 0 (poor performance) to values >1 (performance better than responses to the standard song). Best performance (red) was observed to pulse periods that range from 14-20 ms.

Chapter 4 A Precedence Effect Resolves Phantom Sound Source Illusions in the Parasitoid Fly Ormia ochracea 4.1 Abstract

The acoustic parasitoid fly Ormia ochracea (Diptera: Tachinidae) relies on accurate directional hearing to localize the 5 kHz pulsatile calling song of their host crickets. Under controlled acoustic conditions, flies localize individual sound sources precisely with ears exquisitely sensitive to sound direction. In nature, however, flies are confronted with multiple attractive sound sources that overlap in time and space, potentially masking temporal information relevant to source segregation and localization. Our findings show that temporal overlap affected sound localization. In field experiments that allow small random acoustic asymmetries between two simultaneous sources, most flies can locate a single source accurately, but some localize a ‘phantom source’ between both speakers. Laboratory experiments under symmetric acoustic conditions reliably elicit such misdirected phonotaxis. Selective localization depends on both the relative timing and location of competing sources. With varied temporal overlap, small time differences between two sources can allow flies to selectively localize the leading over lagging sources. Hyperacute directional hearing in Ormia, by allowing rapid, accurate reorientation to one source, may be a sensory adaptation to resolving individual sources in nature.

4.2 Introduction

A fundamental task in hearing is to identify individual sources despite complex acoustic conditions that may mask relevant information. This is especially true in reverberant environmental conditions where sounds reflecting from acoustic clutter may produce multiple coherent copies that the auditory system must differentiate from the originating source for recognition and localization (Litovsky et al., 1999). For two spatially separated coherent signals of equal intensity that arrive at the ears simultaneously, vertebrate auditory systems experience a psychophysical phenomenon known as summing localization and perceive these inputs as a single “phantom” source located between the actual sources (Takahashi and Keller, 1994; Yin, 1994; Keller and Takahashi, 1996a; Blauert, 1997; Litovsky et al., 1999; Tollin et al., 2004; Bee and Riemersma, 2008). Shifts in timing or loudness of sources result in systematic changes in the perceived phantom source location (Blauert, 1997). With delays between two sources smaller than a minimum time interval a precedence effect is observed and only the leading source is perceived. Summing localization causes both signals to be impossible to localize individually (Takahashi and Keller, 1994; Keller and Takahashi, 1996a; Best et al., 2004) while the precedence effect allows for the leading source to be resolved (Wyttenbach and Hoy, 1993; Blauert, 1997; Litovsky et al., 1999; Tollin and Yin, 2003).

Acoustically orienting insects encounter an analogous situation in nature where the scattering of sound by vegetation may result in multiple acoustic reflections that degrade directional cues for sound localization (Romer and Bailey, 1990). This is further complicated by species that communicate simultaneously and commonly in aggregations (Cade, 1981) rendering signals hard to identify and localize (Romer and Bailey, 1990). We show that the acoustic parasitoid fly Ormia ochracea (Diptera: Tachinidae) may utilize the precedence effect to disambiguate source location under similar conditions. Ormia faces a complex auditory scene, where attractive calls of multiple host crickets may overlap in time and space. Nevertheless, Ormia must accomplish the task of finding a source with a much simpler auditory system compared to vertebrate acoustic specialists (Mason et al., 2001; Oshinsky and Hoy, 2002; Robert and Gopfert, 2002) and under the physical constraints on sound localization facing small animals (Michelsen, 1998). Hearing is critical to reproductive success for Ormia as their larvae must develop as internal parasites in nocturnally singing insects (crickets) that female flies locate by sound (Cade, 1975; Adamo et al., 1995). Ormia thus face two important auditory tasks, (1)

correctly identifying the song of a suitable host, and (2) successfully locating the potential host. Behavioural evidence indicates that flies rely on temporal parameters within songs to identify suitable host crickets (Walker, 1993; Wagner, 1996; Zuk et al., 1998; Muller and Robert, 2002; Gray et al., 2007). Recognition of host calling songs initiates phonotactic behaviour in gravid female Ormia. They make their approach by flight and descend within the vicinity of the attractive sound source (Cade, 1975; Mueller and Robert, 2001; Muller and Robert, 2002). Upon landing, females localize and approach their hosts to deposit larvae which are left to burrow and feed within host crickets (Cade, 1975).

Both localization and recognition of appropriate host species depend on the highly stereotyped characteristics of cricket songs (Oshinsky, 1998; Mason et al., 2001; Oshinsky and Hoy, 2002). Under optimal conditions (single source), the auditory system of Ormia is well- suited to simultaneously processing time differences between consecutive sound pulses to detect species-specific pulse rates (recognition) and small time differences in the arrival of the song between both ears (localization). This is accomplished with specialized auditory receptors that fire precisely and only at pulse onsets, with consistent intensity-dependent firing latencies (Mason et al., 2001; Oshinsky and Hoy, 2002). Acoustic interaural time differences (ITDs), the only directional cue available (Robert et al., 1996), vary according to sound position relative to the midline axis such that they decrease in magnitude with decreasing angular deviations from the midline. The jitter (variation in spike timing) of individual auditory receptors approximate to 70μs but the ability of Ormia to localize a source to 2° azimuthal accuracy corresponds to a mean ITD of just 7μs, a sensory threshold significantly lower than properties of individual auditory receptors (Mason et al., 2001). This directional hyperacuity is derived from the amplification of acoustic ITDs by direction-sensitive tympanal mechanics (Miles et al., 1995) and subsequently translated into increasingly larger interaural latency differences that encode sound direction (Mason et al., 2001). Previous studies have demonstrated the ability of Ormia to accurately localize single sources is based on a strategy of pooling responses of auditory receptors to derive a hyperacute measure of direction-dependent ITDs (Mason et al., 2001; Mason et al., 2005). But under natural conditions this may be difficult as crickets commonly sing in aggregations and within vegetation (Greenfield and Shaw, 1983; Gerhardt and Huber, 2002). Flies are thus expected to be within earshot of multiple overlapping sound sources and acoustic reflections with their auditory system required to segregate individual sound pulses that

correspond to the same source and to measure the temporal disparity between both ears to determine sound direction. Overlapping sound sources and associated reverberations may severely mask both temporal parameters leading to failed recognition and/or source localization.

It is unknown how Ormia respond to acoustic conditions that may mask acoustic cues for recognition and localization. To investigate the role of hyperacute directional hearing in resolving single sound sources we examined walking phonotaxis in Ormia to multiple coherent sound sources that mimic a source from its reflections. We tested the hypothesis that exploitation of small time differences between competing sources allows for a precedence effect, resulting in localization dominance of the leading source to be resolved individually. We first describe the effects of temporal overlaps and whether or not they generate ambiguous directional cues as in summing localization. Secondly, we vary source separation and the relative position of simultaneous sources to examine if directional advantages arise to disambiguate locations of a source from its reflection. Collectively, these results suggest that Ormia experience both summing localization and localization dominance. Furthermore, our results demonstrate the precedence effect as a solution to resolving phantom sound source illusions.

4.3 Materials and Methods 4.3.1 Animals Field experiments were carried out between 1830 to 2100 hours in mid October 2007 in an open pasture in Gainesville, Florida (USA). Laboratory experiments were conducted using gravid female O. ochracea derived from laboratory-reared populations that were originally collected at the same location. Laboratory-reared animals were maintained on a 12h:12h light:dark regime and fed nectar solution ad libitum.

4.3.2 Field study Acoustic Stimuli Acoustic stimuli were simulated G. rubens calls (local host cricket for Ormia) consisting of a continuous trill of 5 kHz, 10 ms tone pulses, presented at 50 pulses/sec for 20 minutes. Acoustic stimuli were synthesized in Matlab (Release 2007b), edited in Sony Soundforge (Version 8.0) and recorded onto an audio CD that was played with a portable CD player (Durabrand CD-566; Edison, New Jersey). Two test speakers were connected to the left and right channel of the CD

player for simultaneous playback from both speakers. Stimuli were amplified with stereo circuit boards removed from powered computer speakers and broadcast from Radio Shack piezoelectric horn tweeters (Fort Worth, TX, USA). Sound level of the lure speaker was calibrated (B&K Type 2231 Sound Level Meter, Type 4139 ¼” microphone) to 104 dB SPL at the centre of the 8 cm circular start region while the two test speakers were calibrated to 94 dB SPL.

4.3.3 Laboratory Study Experiments were conducted under free-field conditions with the test arena surrounded by sound attenuating foam. Acoustic Stimuli Acoustic stimuli were synthesized using Tucker-Davis Technologies (TDT) hardware (System 3) and custom scripts written in Matlab. The synthetic stimuli consist of a train of 5 kHz, 10 ms tone pulses, presented at 50 pulses/s, with 10 pulses/train giving an overall chirp duration of 200 ms. The stimuli were amplified (NAD S300; London, UK), passed through a programmable attenuator (TDT model PA5; Gainesville, Florida, USA) and broadcast from Radio Shack piezoelectric horn tweeters (Fort Worth, TX, USA). Stimulus amplitude and timing were controlled by computer. Stimulus levels and timing were calibrated with a probe microphone (B&K Type 4182; Naerum, Denmark).

4.3.4 Data Analysis

Phonotactic responses were recorded with a standard video camera (Panasonic WV-GP460; Matsushita Electric industrial Co., Osaka, Japan or Sony DCR-HC65; Japan) mounted above the arena and VCR (Hitachi DA4; Tokyo, Japan). Analog video clips were captured and digitized at 15 or 30 frames/sec using Adobe Premiere 6.0 and Cinepak compression codec (Radius). Digitized video clips were imported into motion analysis software (Midas 2.0, Xcitex; Cambridge, MA, USA) to extract distance, velocity and direction of movement. At least 5 phonotactic responses per fly were recorded for all conditions tested. Data from repeated responses for individuals were averaged and individual averages were pooled across flies. Data are given as means +/- SE unless otherwise stated. For comparison of responses under different stimulus conditions, overall response angles were calculated as in Mason et al. (2005). Statistical analyses were carried out using Matlab (version R2007b) and R (version 2.6.1) software.

4.4 Results 4.4.1 Field trials

We attracted gravid female Ormia to land on our experimental arena by broadcasting continuous Gryllus rubens trills upwards from a lure speaker located beneath a 8 cm diameter start circle (Fig. 1). Positioned in front of the arena were three funnel traps located at 0º and +/- 45º azimuths (forward, right and left positions) and test speakers behind the left and right traps that broadcast sound in the direction of the start zone. The arena was divided into three regions, each with lanes and demarcation lines marked with distance measurements in 5 cm increments that extended from the margin of the start circle to the trap. This allowed us to record distance travelled in each direction, distance at which flies crossed demarcation lines (i.e. crossed from one choice lane to another), and final choice region. We tested a total of 41 flies. Flies either landed within the start circle, or elsewhere on the arena and subsequently walked into the start region. No attempt was made to control the precise position or orientation of flies within the start circle so that precise phase relationships between test speakers varied randomly among flies tested. Once a fly was within the start circle the lure stimulus was discontinued, at which point the fly would become stationary, and a single test speaker (single source condition) or both left and right test speakers (simultaneous condition) were turned on to broadcast identical continuous trills. Phonotactic responses were scored as correct if flies reached sound traps behind an active speaker while those that walked to the centre (silent) trap were scored as misdirected phonotaxis.

Single-source trials

All flies that were presented with a single source turned and walked directly to the active speaker (100%) (Fig. 2). In no case were flies unable to resolve source location (N=23 flies).

Simultaneous-source trials

Flies presented with two simultaneous sources showed three response patterns. 1) Direct responses. A majority of flies (23/41, 56.1 %) walked directly to one of the speakers similarly to single source trials (Fig. 2). Flies oriented to both speakers with equal frequency (χ2(1)=0.043, p=0.835). 2) Indirect responses. In a smaller proportion of flies (11/41, 26.83%), initial walking trajectories were towards the centre (silent) trap, but flies then turned towards one of the active sound sources. These flies walked an average of 12.30 +/-2.91 cm toward the centre trap before

deviating to one side and crossed the demarcation line at 17.75 +/- 4.32 cm (mean +/- SD). Flies chose either source with equal frequency (χ2(1)=2.273, p=0.132) but flies always continued in the direction of their initial exit from the centre lane. In these indirect phonotactic responses, flies travelled a significantly greater total distance (indirect response Mdn = 30.9 cm, direct response Mdn = 25 cm, U = 192.00, p = 0.01, r = -0.96) to the source and their walking trajectories were composed of a significantly greater forward-directed component (indirect response Mdn = 12.5 cm, direct response Mdn = 0 cm, U = 0.00, p < 0.001, r = -0.96) compared to direct phonotaxis. 3) Incorrect responses. A smaller proportion of flies (7/41, 17.07%) oriented to the centre “silent” trap and walked to the perimeter of the arena without deviating from the centre lane (Fig. 2).

4.4.2 Laboratory Trials

We examined the flies’ directional responses to competing sources in more detail in the laboratory. We recorded phonotactic responses of freely walking flies to a single presentation of two identical competing stimuli under controlled acoustic conditions. In these trials, flies were placed on top of a piece of clear acetate whereby moving the acetate allowed for precise positioning of flies on a calibrated starting point equidistant (40 cm) from three surrounding speakers. This allowed for control of the precise phase and timing relationships of competing stimuli. Responses to such combined stimuli were tested under two speakers arrangements: 1) both speakers located either symmetrically on either side of the fly (180° angular separation), or 2) one speaker forward (on the midline axis) and the other lateral (90° angular separation).

Simultaneous Stimuli

When flies were presented with synchronous sources (directly overlapping in-phase chirps, Fig. 3 inset) separated by 90º, they did not orient to an individual source location. Walking responses were closely timed with stimulus duration. Following stimulus onset, flies accelerated to a maximum velocity of 0.31 +/- 0.02 cm/sec (N = 10 flies), covered a total distance of 1.77 +/- 0.19 cm, and came to a full stop at stimulus offset. Phonotactic responses were directed to a location between both speakers (“phantom source”: angular heading: 31.32 +/- 2.40º; N=10 flies, 10 run/fly; Fig. 3A).

With a 180º speaker separation, walking responses were similar to sources that were separated by 90º in that responses were closely timed with stimulus onset and offset, flies walked a similar total distance (1.428 +/- 0.18 cm, t(9)= 1.113, p = 0.295), and flies failed to resolve a single source location. Responses were consistently directed between both speakers with an angular heading of 3.34 +/- 2.83º (N=9 flies, 10 run/fly; Fig. 3B).

Asynchronous Stimuli

We varied the delay between broadcasting speakers to create a varying amount of temporal overlap in the combined stimuli. A delay of 10 ms (Fig. 4 inset) between competing sound sources generated an interdigitated temporal pattern where pulses from the leading chirp occurred between pulses from the lagging chirp and the combined stimulus had a doubled pulse rate with alternating pulses from separate speakers. Although flies do not respond to such a stimulus when broadcast from a single source (data not shown), they do respond when the sources are spatially separated. When sources were separated by 90º (one speaker in the forward direction), flies were still unable to localize a single source and walked in an intermediate direction, but showed an overall bias toward the forward direction regardless of which speaker was leading (i.e. delivered the initial pulse: forward source leading: 23.31 +/- 2.17º, lateral (right) source leading: 29.06 +/- 3.06º, N=10 flies, 5 run/fly, Watson’s U = 0.055, p>0.10, Fig. 4A). When interdigitated sources were separated by 180º, flies again directed their walking response toward an intermediate location but with a bias towards the leading speaker (angular headings: 180º speaker separation - left source leading, -30.69 +/- 5.16º, right source leading, 20.54 +/- 4.81º, N=10 flies, 5 run/fly, Watson’s U = 0.425, p<0.001, Fig. 4B).

Shorter time differences, however, allowed flies to resolve source location ambiguity when the leading source was in the forward direction. A5 ms onset delay (Fig. 5 inset), created half-overlapping sound pulses within chirps. When sources were separated by 90º (one speaker in the forward direction), flies correctly oriented to the forward speaker when it was the leading source, but oriented to an intermediate location when the lateral speaker was leading (angular headings: 90º speaker separation - forward source leading: 12.76 +/- 1.59º, lateral source leading: 46.72 +/- 3.05º, N=10 flies, 5 run/fly, Watson’s U = 0.425, p<0.001, Fig. 5A). When both speakers were lateral (180º speaker separation), flies showed a stronger bias towards the leading speaker (compared with interdigitated stimuli), but still oriented to intermediate directions (180º

speaker separation angular headings - left source leading, -29.00 +/- 4.76º, right source leading, 39.59 +/- 3.03º, N=10 flies, 5 run/fly, Watson’s U = 0.4035, p<0.001, Fig. 5B).

With the 90º speaker separation, we also tested flies with a 0.2 ms onset delay (i.e. forward speaker leading, stimuli out of phase by one cycle of the 5kHz carrier frequency). Angular headings were correctly oriented to the leading (forward) speaker and were significantly different from the incorrectly directed responses to simultaneous sources (0.2 delay angular heading: 11.37 +/- 4.89º, mean +/- SE, N=10 flies, 10 run/fly, Watson’s U = 0.2725, p <0.01, Fig. 5C).

When competing sources were delayed by 100 ms (Fig 6 inset), the latter half of the leading train of pulses overlapped (simultaneous with) the first half of the lagging train. An onset delay of 200 ms (Fig. 6 inset) resulted in an overall stimulus with a continuous train of pulses 400 ms in duration that switched speakers halfway through stimulus presentation. An onset delay of 500 ms (Fig. 6 inset) created two distinct chirps separated by 300 ms of silence. For these larger interstimulus delays flies walked initially towards the leading source and they maintained this directional heading until the onset of the second source. Following the onset of the second source, flies changed their directional heading to track the second source (Fig. 6). For 100 and 200 ms interstimulus delays, flies tracked the two sound sources in a single phonotactic response. Responses to a 500 ms interstimulus delay further demonstrate a prominent effect of a forward source. Flies responded to this situation with two separate walking responses where they initially tracked the leading source (90º speaker separation – front source leading: -13.08 +/- 1.85º, right source leading: 88.36 +/- 5.93º , Fig. 6A; 180º speaker separation – left source leading: -83.55 +/- 2.66º, right source leading: 84.09 +/- 5.13º, N=7 flies, 5 run/fly, Fig. 6B), paused during the silent interval (290.61 +/- 6.11 ms, N=7 flies) and subsequently engaged in a second walking response to the lagging source. However, responses were strikingly weaker with smaller turn angles towards the lagging source in the 180º speaker configuration than compared to 90º speaker separation (90º speaker separation – front source leading: 56.88 +/- 4.11º, right source leading: -49.48 +/- 9.33º, Fig. 6A; 180º speaker separation – left source leading: 11.90 +/- 4.38º, right source leading: 18.50 +/- 7.84º. Fig. 6B).

4.5 Discussion We examined the ability of Ormia to localize a single source under conditions that mimic reverberant environments with multiple sources. Temporal overlap of competing signals greatly affected overall walking orientation. Complete overlap (synchrony) of attractive stimuli led to misdirected phonotaxis to a “phantom” source. We observed similar behaviour for stimuli that were simultaneous but with pulses out of phase (interdigitated). More importantly, our study demonstrates the precedence effect as a solution for overcoming conditions (echoes or competing sources) that generate ambiguous directional information. Small positional and temporal asymmetries were sufficient for correct phonotactic responses and are thus important for source localization. Solving the “Phantom source” problem Sensory illusions are well-known in studies of human perception and psychophysics (Bregman, 1990; Blauert, 1997; Seghier and Vuilleumier, 2006). Illusions are more difficult to demonstrate in non-human animals, but have also been used to study sensory mechanisms in a variety of taxa that include monkeys (Von Der Heydt et al., 1984), birds (Nieder and Wagner, 1999; Niu et al., 2006), and honeybees (Horridge et al., 1992). In audition, a well-known sensory illusion is the phenomenon of summing localization, where two spatially separated sources broadcasting identical stimuli elicit the perception of a fused single source located midway between two speakers (Blauert, 1997), a problem that may be due to constraints in neural function for processing simultaneous signals. This phenomenon has also been demonstrated in some non-human animals. For example, barn owls orient to a central location in response to identical lateral sources presented simultaneously (Keller and Takahashi, 1996a, b). This behavioural response has been associated with space-specific auditory neurons that signal the location of a single phantom source between two actual source locations (Takahashi and Keller, 1994). A variety of species including owls (Keller and Takahashi, 1996b), rats (Kelly, 1974), cats (Cranford, 1982; Yin, 1994), crickets (Wyttenbach and Hoy, 1993), and katydids (Greenfield and Roizen, 1993; Snedden and Greenfield, 1998; Greenfield et al., 2004) seem to solve this problem using a common strategy; exploiting small time differences between competing sources such that perception of directional information is predominantly conveyed by the leading source while directional information from lagging sources is suppressed (precedence effect). Several electrophysiology studies suggest that this occurs as neural responses to lagging

sources are suppressed by responses to the leading source (Wyttenbach and Hoy, 1993; Yin, 1994; Litovsky et al., 1999; Tollin et al., 2004; Spitzer and Takahashi, 2006). Our results show similar phenomena in Ormia hearing, although with some differences that may be relevant to the question of how these computations are carried out by simple auditory systems. With synchronous stimuli, flies respond to a phantom source. With a small arrival-time difference between two stimuli, flies selectively localize the leading source (localization dominance). In Ormia, however, the precedence effect appears to operate only within a small time frame when temporal disparities fall within the duration of a sound pulse. Flies orient with a bias to the leading source when pulses are half-overlapping (5ms onset delay), but track the leading source only weakly when disparities are beyond the duration of a pulse (i.e. interdigitated condition). Localization accuracy may be expected to improve with greater temporal separation, but responses to interdigitated stimuli are actually worse than responses to half-overlapping pulses. In addition, selective localization only occurs when the leading source is closer to the midline than the lagging source. When the two sources are symmetrically located relative to the midline axis, the leading source is more attractive but responses to the lagging source are not fully suppressed. Under single-source conditions, Ormia show more vigorous phonotaxis (high walking speed, longer distance walked) for a forward source than for an equivalent stimulus from a lateral source (Mason et al., 2005). Thus source localization in Ormia depends on both the relative timing and location of competing sources. A breakdown (release from suppression) of the precedence effect may occur due to changes in the lagging stimulus (Litovsky et al., 1999). Preference to a leading source can be compensated and even reversed with sufficient increases in sound intensity of the lagging source, as reported in katydids and frogs (Dyson and Passmore, 1988; Snedden and Greenfield, 1998; Fertschai et al., 2007). Sound intensity between competing sources was balanced at the start position of flies tested in our laboratory experiments, but perceived relative intensities are expected to change as flies make their approach to a source. Therefore, observed responses reflect initial differences of lead-lag relationships between sources and reinforcing changes in perceived relative sound intensities. However, time-intensity trading relationships in the precedence effect differ in Ormia. Phonotaxis in tethered Ormia shows a concerted influence of both timing and intensity; flies orient to intermediate directions with a bias proportional to the relative intensities of both sources and not a direct preference to a louder lagging source (unpublished data).

In field experiments the flies’ initial position was allowed to vary between the two sources potentially introducing small intensity and timing disparities. Most flies showed a direct response to a single source, suggesting that small binaural disparities may have permitted dominant representation of a single source. But these disparities were small enough that on average flies were significantly diverted from direct phonotaxis. In our field trials, fly starting positions varied within an 8 cm diameter circle around the centre point, so maximum arrival time differences due to flies being closer to one speaker were on the order of 100µs. In laboratory trials, flies reliably oriented to a forward source when the only cue was a 200 µs time difference. These results suggest that, in addition to sound localization per se, hyperacute auditory directionality in Ormia (Mason et al., 2001) may be an important part of the mechanism for selective responses to competing sources. Rapid, accurate orientation to a sound source places that source on the fly’s midline axis, allowing it to dominate sources at other locations, whereas orientation with larger error angles (as in crickets (Wyttenbach and Hoy, 1997)) would not. Within a small temporal window of attention, flies orient reliably to a leading source while a lagging source has minimal influence on orientation. For very short delays between competing sources, the bias in responses towards the leading source will lead to a more gradual movement of the leading source to the midline position. But the mechanism will be more effective for delays longer than the latency of phonotactic responses. Beyond the temporal window of attention there is an increasing attractive effect of the lagging source, especially if it is more forwardly located. This is demonstrated by our results for longer interstimulus delays (100 – 500 ms, Fig. 6). When both sources were lateral (180° separation), flies initiated phonotaxis to the leading source and showed little reorientation to the lagging stimulus. At the onset of phonotaxis flies orient to face the sound source (Mason et al., 2005) thus placing the leading source in the forward position and the lagging source rearward, resulting in disproportionate attraction to the leading source. Symmetrical stimulation of both ears from a rearward position also give rise to the same directional cues that correspond to a forward location (Miles et al., 1995; Robert et al., 1996, 1998). Consequently, flies sometimes responded to the rearward source by actually walking in the preferred forward direction and away from the lagging source. However, for sources separated by 90°(one forward and the other lateral), orientation to the leading source leaves the lagging source in a position that is still more attractive than a rearward source and consequently flies orient to the lagging source with comparatively stronger walking responses.

Our results suggest both the precedence effect and a forward preference as an effective strategy for isolating sources separated by small time-differences occurring within the duration of a pulse, but flies seem to use the strategy ‘forward is best’ when the precedence effect is weak (greater temporal separation). We have demonstrated this at the behavioural level and suggest hyperacute directional hearing has an important role in selective phonotaxis. Although the range of temporal overlaps between simultaneous sources in this study may not entirely reflect conditions that Ormia encounter in nature (Cade, 1981), our results serve to highlight their behavioural capacity and limitations in isolating individual sources. Further studies on the likelihood of encountering similar temporal overlaps as the combined effect of acoustic reflections and acoustic interactions among singing hosts are required to interpret the ecological relevance of our findings. Fine-grain measurements of auditory directionality to competing sources varying in spatial separation is needed to directly test for the role of temporal disparity and sound direction in source segregation. Furthermore, physiological measurements are needed to test for the sensory basis of these effects.

4.6 Acknowledgements

We are in debt to TJ Walker for his generous help in providing assistance and support for our field collecting trip. We would like to thank M. C. B. Andrade, M. M. Kasumovic, J. A. Stoltz, and M. J. Kim Lee and the Integrative Behaviour and Neuroscience Group (IBN) for discussion and comments on our manuscript. We would like to thank M. Chu, P. Martis, A. Feng and the Mason Lab Group for assistance in animal care. Funding was provided by grants from the Natural Sciences and Engineering Council (238882 241419) to A.C.M., National Institute of Health NRSA (1F32GM076091-01A1) to D.O.E., and Animal Behaviour Society Student Research Award and Natural Sciences and Engineering Council PGS D3 to N.L.

References

Adamo, S. A., Robert, D., Perez, J. and Hoy, R. R. (1995). The response of an insect parasitoid, Ormia ochracea (Tachinidae), to the uncertainty of larval success during infestation. Behav. Ecol. Sociobiol. 36, 111-118.

Bee, M. A. and Riemersma, K. K. (2008). Does common spatial origin promote the auditory grouping of temporally separated signal elements in grey treefrogs? Animal Behaviour 76, 831- 843.

Best, V., van Schaik, A. and Carlile, S. (2004). Separation of concurrent broadband sound sources by human listeners. J. Acoust. Soc. Am. 115, 324-336.

Blauert, J. (1997). Spatial Hearing: The Psychophysics of Human Sound Localization. Cambridge, MA: The MIT Press.

Bregman, A. S. (1990). Auditory Scene Analysis: The Perceptual Organization of Sound. Cambridge, Mass: Bradford Books, MIT Press.

Cade, W. (1975). Acoustically orienting parasitoids - fly phonotaxis to cricket songs. Science 190, 1312-1313.

Cranford, J. L. (1982). Localization of paired soudn sources in cats: effects of variable arrival times. J. Acoust. Soc. Am. 72, 1309-1311.

Dyson, M. L. and Passmore, N. I. (1988). The combined effect of intensity and the temporal relationship of stimuli on phonotaxis in female painted reed frogs Hyperolius marmoratus. Animal Behaviour 36, 1555-1556.

Fertschai, I., Stradner, J. and Romer, H. (2007). Neuroethology of female preference in the synchronously singing bushcricket Mecopoda elongata (Tettigoniidae; Orthoptera): why do followers call at all? Journal of Experimental Biology 210, 465-476.

Gerhardt, H. C. and Huber, F. (2002). Acoustic Communication in Insects and Anurans: common problems and diverse solutions. Chicago: The University of Chicago Press.

Greenfield, M. and Shaw, K. (1983). Adaptive signifcance of chorusing with special reference to the Orthopera. In Orthopteran mating systems. Sexual competition in a diverse group of insects, eds D. Gwynne and G. Morris), pp. 1-27. Boulder, Colorado: Westview Press.

Greenfield, M. D. and Roizen, I. (1993). Katydid synchronous chorusing is an evolutionarily stable outcome of female choice. Nature 364, 618-620.

Greenfield, M. D., Siegfreid, E. and Snedden, W. A. (2004). Variation and repeatability of female choice in a chorusing katydid, Ephippiger ephippiger: An experimental exploration of the precedence effect. Ethology 110, 287-299.

Horridge, G. A., Zhang, S. W. and Ocarroll, D. (1992). Insect perception of illusory contours. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences 337, 59- 64.

Keller, C. H. and Takahashi, T. T. (1996a). Binaural cross-correlation predicts the responses of neurons in the owl's auditory space map under conditions simulating summing localization. Journal of Neuroscience 16, 4300-4309.

Keller, C. H. and Takahashi, T. T. (1996b). Responses to simulated echoes by neurons in the barn owl's auditory space map. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 178, 499-512.

Kelly, J. B. (1974). Localization of paired sound sources in rat: small time differences. J. Acoust. Soc. Am. 55, 1277-1284.

Litovsky, R. Y., Colburn, H. S., Yost, W. A. and Guzman, S. J. (1999). The precedence effect. J. Acoust. Soc. Am. 106, 1633-1654.

Mason, A. C., Oshinsky, M. L. and Hoy, R. R. (2001). Hyperacute directional hearing in a microscale auditory system. Nature 410, 686-690.

Mason, A. C., Lee, N. and Oshinsky, M. L. (2005). The start of phonotactic walking in the fly Ormia ochracea: a kinematic study. Journal of Experimental Biology 208, 4699-4708.

Michelsen, A. (1998). Biophysics of sound localization in insects. In Compartive Hearing: Insects, eds. R. R. Hoy A. N. Popper and R. R. Fay), pp. 18-62. New York: Springer.

Miles, R. N., Robert, D. and Hoy, R. R. (1995). Mechanically coupled ears for directional hearing in the parasitoid Ormia ochracea. Journal of the Acoustic Society of America 98, 3059- 3070.

Mueller, P. and Robert, D. (2001). A shot in the dark: The silent quest of a free-flying phonotactic fly. Journal of Experimental Biology 204, 1039-1052.

Muller, P. and Robert, D. (2002). Death comes suddenly to the unprepared: singing crickets, call fragmentation, and parasitoid flies. Behavioral Ecology 13, 598-606.

Nieder, A. and Wagner, H. (1999). Perception and neuronal coding of subjective contours in the owl. Nature Neuroscience 2, 660-663.

Niu, Y. Q., Xiao, Q., Liu, R. F., Wu, L. Q. and Wang, S. R. (2006). Response characteristics of the pigeon's pretectal neurons to illusory contours and motion. Journal of Physiology-London 577, 805-813.

Oshinsky, M. L. (1998). Physiological Correlates of Sound Localization in a Parsitoid Fly, Ormia ochracea. In Department of Neurobiology and Behavior. Ithaca: Cornell University.

Oshinsky, M. L. and Hoy, R. R. (2002). Physiology of the Auditory Afferents in an Acoustic Parasitoid Fly. Journal of Neuroscience 22, 7254-7263.

Robert, D. and Gopfert, M. C. (2002). Novel schemes for hearing and orientation in insects. Current Opinion in Neurobiology 12, 715-720.

Robert, D., Miles, R. N. and Hoy, R. R. (1996). Directional hearing by mechanical coupling in the parasitoid fly Ormia ochracea. Journal of Comparative Physiology A 179, 29-44.

Robert, D., Miles, R. N. and Hoy, R. R. (1998). Tympanal mechanics in the parasitoid fly Ormia ochracea: coupling during mechanical vibration. Journal of Comparative Physiology A 183, 443-452.

Romer, H. and Bailey, W. J. (1990). Insect hearing in the field 2: Male spacing behaviour and correlated acoustic cues in the bushcricket Mygalopsis marki. Comparative Biochemistry and Physiology a-Physiology 97, 443-447.

Seghier, M. L. and Vuilleumier, P. (2006). Functional neuroimaging findings on the human perception of illusory contours. Neuroscience & Biobehavioral Reviews 30, 595-612.

Snedden, W. A. and Greenfield, M. D. (1998). Females prefer leading males: relative call timing and sexual selection in katydid choruses. Animal Behaviour 56, 1091-1098.

Spitzer, M. W. and Takahashi, T. T. (2006). Sound localization by barn owls in a simulated echoic environment. Journal of Neurophysiology 95, 3571-3584.

Takahashi, T. T. and Keller, C. H. (1994). Representation of multiple sound sources in the owls auditory space map. Journal of Neuroscience 14, 4780-4793.

Tollin, D. J. and Yin, T. C. T. (2003). Psychophysical investigation of an auditory spatial illusion in cats: The precedence effect. Journal of Neurophysiology 90, 2149-2162.

Tollin, D. J., Populin, L. C. and Yin, T. C. T. (2004). Neural correlates of the precedence effect in the inferior colliculus of behaving cats. Journal of Neurophysiology 92, 3286-3297.

Von Der Heydt, R., Peterhans, E. and Baumgartner, G. (1984). Illusory Contours and Cortical Neuron Responses. Science (Washington D C) 224, 1260-1262.

Wagner, W. E. (1996). Convergent song preferences between female field crickets and acoustically orienting parasitoid flies. Behavioral Ecology 7, 279-285.

Walker, T. J. (1993). Phonotaxis in female Ormia ochracea (Diptera: Tachinidae), a parasitoid of field crickets. J. Insect Behav. 6, 389-410.

Wyttenbach, R. A. and Hoy, R. R. (1993). Demonstration of the precedence effect in an insect. J. Acoust. Soc. Am. 94, 777-784.

Wyttenbach, R. A. and Hoy, R. R. (1997). Spatial acuity of ultrasound hearing in flying crickets. Journal of Experimental Biology 200, 1999-2006.

Yin, T. C. T. (1994). Physiological correlates of the precedence effect and summing localization in the inferior colliculus of the cat. Journal of Neuroscience 14, 5170-5186.

Zuk, M., Rotenberry, J. T. and Simmons, L. W. (1998). Calling songs of field crickets (Teleogryllus oceanicus) with and without phonotactic parasitoid infection. Evolution 52, 166- 171.

Figure Legends

Figure 1. Experimental arena for field phonotaxis trials. A lure speaker orientated upwards was positioned at the 4 cm circular start region (yellow) underneath the arena. Two test speakers (black triangles) orientated horizontally were separated by 90º and positioned behind sound traps (blue). A third (silent) trap was placed between the two speakers (0º from the start region). Demarcation lines divided the test arena into three choice regions: two correct choice regions (in green) and an incorrect choice region (in red).

Figure 2. Phonotaxis to simultaneous and single sources under field conditions. Direct responses. Percentage of correct responses with direct approach to source one location (Blue). Indirect responses. Percentage of correct responses with initial misdirected walking trajectories to a phantom source (0º heading) (Green). Incorrect responses. Percentage of unsuccessful phonotaxis (Red) with walking trajectories misdirected towards a phantom source location (0º heading). Most flies were able to resolve an individual source while a smaller proportion of flies walked towards a phantom source location.

Figure 3. Phonotaxis to simultaneous sources under laboratory conditions. Two synthetic sound sources broadcast synchronously (0 ms onset delay) to produce overlapping chirps (inset). (A) Average walking trajectory of free-walking flies to sources separated by 90º. (i) Average response (N=10 flies). (ii) Individual responses of a single representative fly. (B) Average walking response of flies to sources separated by 180º. (i) Average response (N=9 flies). (ii) Individual responses of a single representative fly. Red line within individual response plots (ii) indicate average responses. Flies responded to simultaneous sources by directing phonotaxis to a phantom source location between broadcasting speakers (≈45º heading in 90º condition, ≈0º heading in 180º condition).

Figure 4. Phonotaxis to interdigitated (10 ms onset delay) sources. Two synthetic sound sources broadcast with a 10 ms onset delay to produce interdigitated chirps (inset). (A) Average walking response of flies (N=10 flies) to sources separated by 90º. (B) Average walking response of flies to sources separated by 180º (N=10 flies). Filled symbols indicate walking responses to forward (A) or left (B) leading sound sources while open symbols indicate walking response to lateral or right leading sound sources leading. Flies responded to interdigitated sources by directing

phonotaxis to a phantom source location between the broadcasting speakers (45º heading in 90º condition, 0º heading in 180º condition) but with a slight bias towards leading sources.

Figure 5. Phonotaxis to overlapping sources with small onset time differences (5 and 0.2 ms onset delay). Two synthetic sound sources broadcast with a 5 ms onset delay to produce half- overlapping sound pulses (inset,A,B). (A) Average walking responses of flies to sources spatially separated by 90º (N= 10 flies). (B) Average walking responses of flies to sources spatially separated by 180º (N= 10 flies). Two synthetic sound sources broadcast with a 0.2 ms onset delay (inset, C). (C) Average walking response of flies to sources separated by 90º (N=10 flies). Filled symbols indicate walking responses to forward (A, C) or left (B) leading sources while open symbols indicate walking response to lateral (A, C) or right (B) leading sources. Small onset time differences between sources (0.2 ms) allowed flies to resolve source location ambiguity to the leading source.

Figure 6. Phonotaxis to overlapping sources with large onset delays (100ms, 200ms, and 500ms). Two synthetic sound sources broadcast with 100 ms onset delay (half-overlapping pulse train), 200 ms onset delay (continuous pulse train), and 500 ms onset delay (two distinct chirps separated by 300 ms of silence) (insets,A,B). (A) Average walking responses of flies (N=7 flies) to sources spatially separated by 90º with onset delays of 100 ms (black), 200 ms (blue), and 500 ms (red). (B) Average walking responses (N= 7 flies) of flies to sources spatially separated by 180º with onset delays of 100 ms (black), 200 ms (blue), and 500 ms (red). Filled symbols indicate walking responses to forward (A) or left (B) leading sources while open symbols indicate walking response to lateral (A) or right (B) leading sources.Under all temporal overlaps tested, flies initially turned and walked toward the leading source followed by a secondary turning response to the lagging source. Secondary turning responses were weaker for sources separated by 180º.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

This study was published in Proceedings of the National Academy of Sciences (Lee et al. 2009) and is reprinted with permission from the journal.

Chapter 5 Distracting Noise Interferes with Sound Localization Accuracy in the Acoustic Parasitoid Fly Ormia ochracea 5.1 Abstract

Localizing sound sources of interest in noisy environments is a sensory challenge encountered by all organisms that depend on hearing for behavioural decisions. Reproduction in the acoustic parasitoid fly Ormia ochracea (Diptera: Tachinidae) requires successful localization of species- specific calling songs from singing field crickets that serve as hosts for their developing larvae. Both song recognition and sound localization depend on the measurement, and segregation of temporal information corresponding to single sources and this may be compromised by masking noise in the acoustic space of O. ochracea. In walking phonotaxis experiments with tethered females, we examined the ability of O. ochracea to separate and localize song vs. random noise at different signal-to-noise ratios (SNR) and when song and noise were spatially grouped or separated. Surprisingly, our results indicate no support for spatially-mediated release from masking in O. ochracea. Response thresholds to song intensity were significantly increased in the presence of noise. Increasing the spatial separation between song and noise did not decrease response thresholds. Flies walked significantly less in the presence of noise. Greater SNR resulted in improved song detection that modulated walking velocity to the cricket song and resulted in greater walking distances. Phonotactic responses were diverted away from the noise source with the result that they were not accurately oriented toward the attractive stimulus. This effect was greater for more laterally located noise sources with greater spatial separation from the attractive source. Diverted walking responses were mainly attributed to changes in steering velocity that depended on the location of the noise source. Our results suggest that, for the detection of signals in noise, the flies rely on asymmetries (intensity differences) in the sound field that are largely (or entirely) processed via peripheral auditory mechanics. A consequence of this strategy is that localised noise sources can generate interaural differences in the perception of attractive (cricket) stimuli that bias directional information.

5.2 Introduction

Sensory systems are often bombarded with noise: stimuli that may compromise information- processing crucial to behavioural decisions. In audition, deciphering the auditory scene involves identification and localization of individual sound sources (Bregman, 1990; Pollack, 2000) in the environment. Tympanic eardrums typically respond to a mixture of sources that include sounds of interest, biotic, and abiotic noise (Romer et al., 1989; Feng and Ratnam, 2000; Brumm and Slabbekoorn, 2005), leaving the auditory system with the task of extracting relevant information from irrelevant noise. In human hearing, the so-called cocktail party effect describes the ability to isolate a single source when multiple conversations and noise overlap in time and space (Cherry, 1953; Bee and Micheyl, 2008). Under such conditions speech perception is often limited but can be improved by spatial separation between sounds of interests and noise sources (Shinn-Cunningham et al., 2001). This phenomenon has been termed spatially mediated release from masking. Studies from several vertebrate taxa show that increases in spatial separation between a signal of interest and noise can contribute to solving the cocktail party problem (Wakeford and Robinson, 1974; Saberi et al., 1991; Dent et al., 1997; Bee, 2007, 2008; Bee and Micheyl, 2008) through improved signal detection, signal recognition, and localization (Bee, 2007, 2008; Bee and Micheyl, 2008). Neurophysiological measurements from the auditory periphery show reduced signal detection thresholds for spatially separated sources and recordings from neurons in higher processing centres provide support for the involvement of central processing mechanisms in spatial release from masking (Lin and Feng, 2001).

The acoustic parasitoid fly Ormia ochracea (Diptera: Tachinidae) encounters situations that may be analogous to the cocktail party problem. Reproductive success in O. ochracea depends on finding host crickets for their larval young (Cade, 1975). Male field crickets produce calling songs comprised of 5 KHz tone sound pulses repeated at a rate that varies among species. Host parasitisation is species-specific. For example, Floridian O. ochracea prefer the continuous trills of Gryllus rubens over the chirps of Californian Gryllus lineaticeps (Walker, 1993; Gray et al., 2007). In search of hosts, gravid female O. ochracea use directional hearing to home in on the calling song of field crickets to the accuracy of 2° azimuth (Mason et al., 2001). The ability of O. ochracea to localize individual crickets is an astonishing feat. With ears separated by a mere 500 µm, diffraction of the ~7cm wavelength song is insufficient to result in conventional cues for direction hearing. In the sound field, measurements show that intensity differences (interaural

87 level difference: ILD) between sound arriving at the two ears are absent and time differences (interaural time differences: ITD) are too minuscule for neural processing (Robert et al., 1996). However O. ochracea possess a number of auditory specializations to allow the detection and processing of the small ITD’s that are the only acoustic directional cue available to them. Mechanical coupling between the tympana results in asymmetric vibration for directional sound sources, and auditory receptors are specialized for accurate coding of interaural differences correlated with sound direction (Miles et al., 1995). Interestingly, both localization and recognition of attractive sound sources depend on processing temporal cues. While sound localization depends on measuring time differences between ears, song recognition depends on measuring the timing between consecutive sound pulses of a single song.

The highly specialized auditory system of O. ochracea achieves directional hyperacuity with a relatively simple nervous system that displays unique afferent response characteristics for copying song features and coding sound direction (Mason et al., 2001; Oshinsky and Hoy, 2002; Robert and Gopfert, 2002) compared to vertebrate hearing specialists (Popper and Fay, 2005). Whether trade-offs exist for such specialization is unknown. Previous work has shown that O. ochracea can separate simultaneous attractive sources with temporal cues via the precedence effect (Lee et al., 2009), but auditory processing functions to deal with interfering noise have not been examined. Both song recognition and sound localization may improve when a source of interest is spatially separated from noise. Spatial release from masking may occur with an increase in signal-to-noise ratio (SNR) at one ear for spatially separated signal and noise (better ear effect), or it may occur to differences in the ITD of the masking noise at both ears (binaural effects) (Blauert, 1997; Bee and Micheyl, 2008). A signal of interest at 0° azimuth produces equal stimulation of both left and right tympana while masking noise displaced from the signal of interest may result in ITDs not found in the forwardly located source of interest (Bee and Micheyl, 2008). We test the hypothesis that spatially mediated release from masking improves source localization in O. ochracea by examining the ability of O. ochracea in tethered-walking to resolve song location under masking and non-masking conditions. SNR and spatial separation between a cricket song and noise were varied to determine the effects of noise on song localization. If O. ochracea experiences spatial release from masking, flies should localize the cricket song with improved accuracy and at lower SNRs when song and noise are spatially separated than compared to song and noise that are grouped.

5.3 Material and Methods 5.3.1 Animals

5.3.2 Acoustic Stimuli

Attractive (phonotactic) stimuli were 2-sec trills constructed from 10 ms duration, 5 KHz sound pulses with 1ms on/off ramps, repeated at 50 pulses/sec. Noise stimuli were either 2-sec (as the negative control stimulus) or 4-sec (as part of a test stimulus, see below) bursts of unramped band-limited (2-7 KHz) random noise. Stimulus waveforms (Fig. 1A) were synthesized in Matlab (R2009b, The MathWorks Inc. USA) with custom software and converted to analog signals using National Instruments data acquisition hardware (NI USB-6251, 44100 Hz), amplified with Radio Shack Realistic (SA-10 Solid State Amplifier MOD-31-1982B, Taiwan) and broadcast through Skullcandy earbuds (INK’D, China). Stimulus levels were software controlled and calibrated using a probe microphone (B&K Type 4182, Denmark) powered by B&K Nexus Conditioning Amplifier (Denmark). The continuous noise stimulus lacked temporal structure and contained broad spectrum energy that overlapped and masked temporal and spectral features of the 5 KHz cricket song (Fig 1).

5.3.3 Experimental Apparatus

Behavioural measurements were made from tethered flies performing walking phonotaxis on a high resolution trackball system situated equidistant (10 cm) from all test speakers (Fig. 1B) and surrounded by acoustic attenuating foam. The trackball system consists of a light-weight table tennis ball held afloat by a constant airstream above a modified optical mouse sensor (ADNS 2620, Avago Technologies, USA). Walking responses were transduced as rotations of the trackball that actuated the optical mouse sensor to record changes in x and y pixel units at a sampling rate of 2160 Hz (Lott et al., 2007). Pixel units were calibrated to actual walking distances by measuring displacement of points on the ball from highspeed video footage (DRS

Lightening RDT, 500 frames per second) synchronized to pixel data captured by the trackball system. Data collection by the trackball system was controlled by custom Matlab software that interfaced with the National Instruments data acquisition system to ensure synchronous sound presentation and the recording of walking traces.

5.3.4 Protocol

Experiment 1: Estimating Response Thresholds

Response thresholds were estimated using a procedure described by Bee and colleagues (Bee and Schwartz, 2009). A 0.5-second target signal was presented from a speaker directly in front of the fly (0° azimuth) simultaneously with a continuous masker that varied in broadcast location to achieve different song and noise spatial separations (Fig. 1B). A 30-second pause was introduced between stimulus presentations and switching of the continuous masker broadcast location. In the grouped condition, the continuous masker was presented from a speaker adjacent to the target signal (noise either to the left or right of song) giving 6° of angular separation. In spatially separated source condition, the masker was located 90° to the left or right of the target speaker (Fig. 1B). Flies were initially presented with a 76 dB SPL target signal over the 76 dB SPL continuous masker. Using an automated tracking procedure implemented in Matlab custom software, the sound intensity of the target signal was decreased in 3 dB steps for each response greater than 1 cm in total walking distance. When a walking response was less than 1 cm, the target signal sound intensity was increased by 1.5 dB for a final test response. If flies responded with a final walking distance greater than 1 cm, this final test sound intensity was recorded as the upper bound (UB), and the next lowest intensity eliciting a response was recorded as the lower bound (LB) of the response threshold. If flies failed to respond with a walking distance greater than 1 cm in this final test, this final sound intensity was recorded as the LB, and the previous sound intensity eliciting a response was recorded as the UB of the response threshold. An estimate of the response threshold was calculated as:

Experiment 2: Localization Accuracy

The standard 2-second cricket trill (see above) presented at 76 dB SPL in the absence of noise served as the positive control stimulus while the 2-second noise burst broadcast from the forward speaker in isolation served as the negative control. Target signal localization accuracy under two different song and noise spatial separations (spatially grouped or separated) were examined with identical speaker arrangements as in Experiment 1. Test stimuli consisted of a combination of the standard cricket trill broadcast simultaneously with noise at three different levels (70, 76, and 82 dB SPL, C-weighting, fast RMS) to achieve three signal-to-noise ratios (-6 dB, 0 dB, +6 dB SNR) forming a 2 by 3 factorial experimental design. Test stimuli were timed such that the attractive two-second cricket trill occurred midway through the noise presentation (beginning one second after the onset of the noise playback and finishing one second before noise offset) (Fig. 1A). Flies were tested under a repeated measures design such that 3-6 responses were collected for each SNR and separation combination (number of responses differed because response probability decreased at lowest SNR). A test sequence for a fly commenced with three presentations of the negative control (noise stimulus alone), three positive controls (cricket song), a sequence of test stimuli in random order (song and noise at different SNR), three more positive controls, and finally ending with three negative controls.

5.3.5 Data Analysis

X and Y coordinates from trackball data traces were collected at 2160 Hz to construct virtual walking trajectories. Walking velocities were separated into two components: 1) steering velocities and 2) forward velocities were calculated as changes in x or y coordinates respectively over time. Angular heading was calculated as in Mason et al. (2005) (Mason et al., 2005). Responses collected under the same speaker separations but for the reversed position were combined by mirroring responses relative to 0° azimuth. Repeated responses for the same stimulus conditions were averaged within individuals. Data are given as means ± SEM. Latency, walking distance, forward velocity, and steering velocity were analysed using a 2(separation, within-subjects) by 3(SNRs, within-subjects) repeated measures ANOVA. Linear contrasts across levels of SNR were used to test the prediction that increases in SNRs results in an increase in measures of phonotaxis.

5.4 Results 5.4.1 Walking responses to song and noise in isolation

Flies remained quiescent in the absence of stimulus presentation. When presented with the standard 76 dB 2-second cricket trill from 0° azimuth flies responded with a latency of 56.45 ± 2.76 ms, walked at an angular heading of 2.74 ± 1.45°, and covered a total distance of 16.91 ± 0.91 cm. When subjected to band-limited random noise only from the forward speaker (at the same intensity), 37 % of flies did not respond while the remainder responded with a similar latency of 52.19 ± 3.32 ms (Paired Sample T-Test: n = 9 flies t=0.741, P=0.477), a different mean angular heading of 0.76 ± 4.03° (Watson U2 Test: n = 9 U2= 0.256, P=0.01) and traveled a significantly shorter total distance of 2.08 ± 1.01 cm (Paired Sample T-Test: n=9 flies, t=9.983, P<0.001). Treatment order (noise before song or vice versa) did not appear to affect response 2 probability to noise (χ (0.05, 1) = 0.083, P=0.773).

5.4.2 Effects of SNR and source separation on response thresholds

Estimated response thresholds were significantly affected by treatment conditions (Repeated

Measures ANOVA: F(2,22) = 72.36, P<0.001, Fig. 2). In the absence of the 76 dB SPL continuous masker, flies responded to significantly lower mean song intensities (50.80±5.43 dB SPL) compared to conditions with the continuous masker spatially grouped (63.31±2.67 dB SPL,

F(1,11) = 66.39, P<0.001) or separated (63.19±2.96 dB SPL, F(1,11) =89.94, P<0.001). Pairwise comparisons reveal no significant differences in response thresholds for spatially grouped or separated song and noise (P=1.000, Fig. 2).

5.4.3 Effects of SNR and source separation on response latency and walking distance

Response latencies for cricket songs were not significantly affected by SNR (Repeated Measures

ANOVA: F(2,12) = 0.412, P=0.671) or source separation (F(1,6)=0.001, P=0.974). Flies responded with latencies of 59.51±3.51 ms. Total distance travelled did not depend on source separation

(F(1,9)=0.092, P=0.768) but was significantly affected by SNR (F(2,18) = 8.601, P=0.002). A linear contrast revealed that walking distance increased with SNR (linear contrast: F(1,9) =15.456, P=0.03). With an increase in SNR from -6 dB to +6 dB, walking distance increased from 11.35±1.81 cm to 16.36±1.12 cm. Post Hoc contrasts revealed that a SNR of -6dB resulted in

significantly shorter distances walked than at higher SNRs (-6 dB vs 0 dB SNR: F(1, 1)=8.466,

P=0.017; -6 dB vs +6 dB SNR: F(1, 1)=15.456, P=0.003).

Forward velocity profiles reveal a strong but brief walking component in response to noise at a velocity of 5.20±0.40 cm/sec in the forward direction that quickly decayed before song onset (Fig. 3). During song presentation, forward velocity increased as a function of SNR

(Repeated Measures ANOVA: F(2,14)=25.381, P<0.001) but not song and noise separation

(F(1,7)=1.866, P=0.214). A linear contrast revealed that forward velocity increased with SNR

(F(1,7) =31.869, P=0.01). With an increase in SNR from -6 dB to +6 dB SNR, forward velocity increased from 2.26±0.46 cm/sec to 4.67±0.39 cm/sec but all were significantly less than the average forward velocity of 6.03±0.28 cm/sec in response to the song in isolation (Control vs. -6 dB SNR: F(1,9)=56.86, P<0.001, Control vs. 0 dB SNR: F(1,9)=46.79, P<0.001, Control vs. +6 dB

SNR: F(1,9)=26.24, P=0.001). SNR-dependent changes in forward velocity corresponded with SNR-dependent changes in walking distances (above).

5.4.4 Effects of SNR and source separation on walking direction

At noise onset, flies initiated brief steering responses toward noise with durations that depended on noise intensity (Repeated Measures ANOVA: F(2,14)=7.804, P=0.05, Fig. 4.) but not location

(F(1,7)=1.828, P=0.218, Fig. 4). Steering toward noise lasted between 65.00±20.44 ms and 157.01±29.59 ms as noise intensity was dropped from 82 dB to 70 dB SPL. This was followed by a quick transition to steering away from noise.

For spatially grouped song and noise, the presence of noise had a significant effect on walking direction, but this effect did not significantly vary with SNR. Instead, there was a slight tendency for flies to deviate further from the control direction with greater directional variability at worst SNRs (circular dispersions: -6 dB SNR: 0.054, 0dB SNR: 0.010, +6 dB SNR: 0.002; Rao’s Test for Equality of Dispersion: R= 11.25, P<0.05). At the onset of song presentation (1500 ms into stimulus capture), flies steered away from noise with an overall average velocity of 0.37±0.15 cm/sec velocities that did not significantly change to different SNRs (F(2,14)=1.115, P=0.355, Fig. 4A). Flies responded with similar angular headings (-6dB: -13.67±6.34°, 0dB: - 6.88±2.66°, +6dB: -4.99±1.43°; Wheeler Watson Test: n=10, W=4.49, P>0.05) for all SNR treatments and these differed significantly from controls (2.74±1.45°) (Moore’s Test: control vs.

-6 dB SNR, n=10 flies R´=1.2414, P<0.025; control vs. equal dB SNR, n=10 flies, R´=1.6679, P< 0.0001, control vs. +6 dB SNR, n =10, R´=1.7344, P<0.0001, Fig. 5A).

In response to simultaneous broadcast of noise and an attractive signal from spatially separated sources, flies oriented to different mean angular headings at different SNRs (-6dB: - 15.02±7.78°, 0dB: -12.75±4.27°, +6dB: -10.39±2.49°), all of which were significantly different from responses to controls (control vs. -6 dB SNR, n=8 flies R´=1.2372, P<0.025; control vs. equal dB SNR, n=10 flies, R´=1.7299, P< 0.0001, control vs. +6 dB SNR, n =10, R´=1.7341, P<0.0001, Fig. 5B). At -6dB SNR, angular heading did not differ for the two source separations (-6dB grouped vs. separated: R´=0.2468, n = 8, p> 0.2). However, song and noise separation caused a significant shift of -6.41° and -2.53° contralateral to the location of the noise for 0dB and +6dB SNR respectively (0dB SNR grouped vs. separated: R´=1.5975, n = 10, P<0.0001, +6dB SNR grouped vs. separated: R´=1.1049, n = 10, P< 0.0001, Fig. 5AB). Steering velocity

depend on source separation (F(1,7)=9.193, P=0.019) . When song and noise were separated, steering velocity increased to 1.07±0.34 cm/sec and movement was directed away from noise (Fig. 4B).

5.5 Discussion

Ormia ochracea responded with greater walking distances, consistent latencies, and localized the cricket song with greater accuracy in the absence of noise. Broadcasting song over noise compromised the ability of flies to resolve song location at all SNRs tested. Separating song from noise did not decrease song intensity response thresholds or improve song localization but caused even greater displacement of walking responses away from both sources. Therefore, our results provide no support for spatially mediated release from masking in O. ochracea. More interestingly, diversion of walking responses was directional and depended on noise location.

Few behavioural studies in invertebrates have examined the role of spatial release from masking in source segregation, and fewer have documented support for this mechanism. This may be due in part to neural constraints in simultaneously segregating temporal information for recognition and localization. For example, grasshoppers fail to respond to two speakers broadcasting conspecific songs that sum to an incorrect song temporal pattern regardless of whether speakers are grouped or spatially separated (von Helversen, 1984). However, positive phonotaxis occurs for sources that sum to a correct song as if temporal pattern input from both

ears are internally added at the cost of losing directional information (von Helversen and von Helversen, 1995). Ascending auditory interneurons adept at copying species-specific temporal patterns are directionally insensitive while those that are directionally sensitive copy temporal patterns weakly (Stumpner et al., 1991). In the katydid Tettigonia viridissima, recordings from paired bilaterally symmetric omega interneurons (ON1) show neural response patterns that may result in spatial release from masking (Romer and Krusch, 2000). For two laterally positioned speakers presenting different temporal patterns, ON1 copies the ipsilateral source faithfully while spike discharges to the contralateral source are almost absent (Romer and Krusch, 2000). These response patterns hold true even when both speakers are moved frontally until separated by only 7.5° (Romer and Krusch, 2000). It appears that the auditory system of T. viridissima is divided into two hemispheres such that song arriving at any direction in a given hemisphere will illicit the best response on that side of the auditory system (Romer and Krusch, 2000). What remains to be shown is how T. viridissima responds behaviourally under these conditions.

In our study, increases in SNR improved song detection and resulted in greater forward velocities that translated into greater walking distances toward the cricket song. Above -6 dB, SNR seems to have minimal influence on angular heading. Instead, walking direction was significantly influenced by song and noise separation. For spatially grouped song and noise, both steering velocity and angular heading changed minimally with changes in SNR. Increases in song and noise separation caused SNR-dependent changes in steering velocities that resulted in greater diversions away from both song and noise locations. O. ochracea show graded turning responses to small changes in the sound direction of a single source. Directional information in the auditory system is represented by the relative timing of afferent responses in the two ears, with low-jitter, phasic receptors pooled to achieve hyperacute time-coding (Mason et al., 2001). The noise avoidance response we observed may be a consequence of this mechanism of encoding sound source location.

Directional noise avoidance may be driven by directional response properties of the two tympana that alter the perceived cricket song location. For an attractive source located in the forward direction, both left and right tympana respond equally in timing and amplitude (Robert et al., 1996), presumably resulting in identical afferent activity for both sides of the auditory system. A localised noise source positioned off the midline axis will affect the two ears asymmetrically, due to the directional properties of the tymapana – the ipsilateral ear will be

95 more strongly affected than the contralateral. The nature of auditory coding in O. ochracea suggests that an effect of continuous noise would be to decrease the effective (or perceived) intensity of an attractive stimulus. At the neural level most auditory afferents in O. ochracea are comprised of Type 1 afferents that precisely mark the onset of sound with a single spike followed by a refractory period of approximately 4 ms (Oshinsky and Hoy, 2002). This is a coding strategy that exploits the predictable, pulsatile pattern of cricket song by registering the onset timing of successive pulses within the chirp or trill that comprises a cricket song. Receptor latencies, i.e. the precise timing of onset responses within each ear, are a function of stimulus intensity (Oshinsky and Hoy, 2002). Tympanal mechanics results in interaural differences in the timing and amplitude of vibration (Miles et al., 1995; Robert et al., 1996) which are subsequently encoded in auditory receptor responses as interaural latency differences in the phasic responses of receptors (Mason et al., 2001). As continuous noise does not elicit on-going activity in Type 1 receptors, it will instead reduce the effective amplitude of a pulsatile signal (cricket song) as pulse onsets will be detected relative to a higher noise floor. For a localised noise source, this effect will be greater in the noise-ipsilateral ear and will tend to simulate a directional bias in the location of the attractive source (Fig. 5, 6). Under similar acoustic conditions, the field cricket Gryllus campestris responds similarly (Wendler, 1989). Females localize an intermediate location to two attractive sound sources but are diverted away from both sources when subjected to a song and tone (lacking song pattern) combination (Wendler, 1989). The sustained, albeit weak, negative phonotactic responses of flies noise during noise playback (prior to song onset) suggests that sustained (tonic) responses to ongoing stimulation may be present in the flies auditory pathway and underlie the noise avoidance response. Alternatively, the random amplitude envelope of the noise waveform could result in sporadic activity in phasic receptors.

At noise onset, our unramped noise stimulus may have caused type 1 afferents to fire no differently to the onset of sound pulses within a cricket song leading to orientation towards noise. But flies then quickly steered away from noise. The field cricket Gryllus bimaculatus has been shown to reactively steer to individual sound pulses on a timescale that precludes measurements of song pattern (Hedwig and Poulet, 2004; Hedwig and Poulet, 2005), but overall phonotaxis performance is still gated by song recognition (Poulet and Hedwig, 2005). In O. ochracea, we have observed flies to respond weakly to non-preferred songs after being primed with exposure to attractive songs (unpublished data).

We have now documented two conditions which result in mis-directed phonotaxis in O. ochracea. Multiple attractive sound sources overlapping in time and space results in phantom source phonotaxis with flies orienting in a direction intermediate to both source locations (Lee et al., 2009). Under natural conditions, this ambiguity is resolved via small differences in the timing of competing signals via a precedence effect (Lee et al., 2009). We now show that in the presence of continuous unpatterned noise, flies are consistently diverted away from both song and noise. These results suggest that Ormia lack a specialised mechanism for separating signals from noise, but rely instead on an orientation response that favours symmetrical auditory input. Widely distributed background noise would not bias source localisation (though it might reduce detection), whereas localised noise sources elicit negative orientation responses from flies resulting in a localisation response that is a sum of positive phonotaxis and noise avoidance. Experiments are currently underway to test our model of the role of tympanal mechanics and the neural basis of directional noise avoidance behaviour (Fig. 6).

5.6 Acknowledgements

We would like to thank TJ Walker for assistance in fly collection, MJ Kim Lee and D Koucoulas for their meticulous care in fly husbandry duties, MCB Andrade, J Peever, DR Howard, CL Hall, PA De Luca, PA Guerra, JJ Ting, ME Jackson, and J Van Eindhoven for helpful discussion of experimental design. This project was funded by NSERC to AC Mason (238882, 241419), OGS, NSERC PGS D3, and SICB grants-in-aid of research to N LEE.

References

Bee, M. A. (2007). Sound source segregation in grey treefrogs: spatial release from masking by the sound of a chorus. Animal Behaviour 74, 549-558.

Bee, M. A. (2008). Finding a mate at a cocktail party: spatial release from masking improves acoustic mate recognition in grey treefrogs. Animal Behaviour 75, 1781-1791.

Bee, M. A. and Micheyl, C. (2008). The cocktail party problem: What is it? How can it be solved? And why should animal behaviorists study it? J. Comp. Psychol. 122, 235-251.

Bee, M. A. and Schwartz, J. J. (2009). Behavioral measures of signal recognition thresholds in frogs in the presence and absence of chorus-shaped noise. J. Acoust. Soc. Am. 126, 2788-2801.

Blauert, J. (1997). Spatial Hearing: The Pyschophysics of Human Sound Localization. Cambridge, MA: The MIT Press.

Bregman, A. S. (1990). Auditory Scene Analysis: The Perceptual Organization of Sound. Cambridge, Mass: Bradford Books, MIT Press.

Brumm, H. and Slabbekoorn, H. (2005). Acoustic Communication in Noise. In Advances in the Study of Behavior, vol. Volume 35, pp. 151-209: Academic Press.

Cade, W. (1975). Acoustically orienting parasitoids - fly phonotaxis to cricket songs. Science 190, 1312-1313.

Cherry, E. C. (1953). Some experiments on the recognition of speech, with one and with two ears. J. Acoust. Soc. Am. 25, 975-979.

Dent, M. L., Larsen, O. N. and Dooling, R. J. (1997). Free-field binaural unmasking in budgerigars (Melopsittacus undulatus). Behav. Neurosci. 111, 590-598.

Feng, A. S. and Ratnam, R. (2000). Neural basis of hearing in real-world situations. Annu. Rev. Psychol. 51, 699-725.

Hedwig, B. and Poulet, J. F. A. (2004). Complex auditory behaviour emerges from simple reactive steering. Nature 430, 781-785.

Lin, W.-Y. and Feng, A. (2001). Free-field unmasking response characteristics of frog auditory nerve fibers: comparison with the responses of midbrain auditory neurons. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 187, 699-712.

Mason, A. C., Oshinsky, M. L. and Hoy, R. R. (2001). Hyperacute directional hearing in a microscale auditory system. Nature 410, 686-690.

Mason, A. C., Lee, N. and Oshinsky, M. L. (2005). The start of phonotactic walking in the fly Ormia ochracea: a kinematic study. Journal of Experimental Biology 208, 4699-4708.

Miles, R. N., Robert, D. and Hoy, R. R. (1995). Mechanically coupled ears for directional hearing in the parasitoid fly Ormia ochracea. J. Acoust. Soc. Am. 98, 3059-3070.

Oshinsky, M. L. and Hoy, R. R. (2002). Physiology of the auditory afferents in an acoustic parasitoid fly. Journal of Neuroscience 22, 7254-7263.

Pollack, G. (2000). Who, what, where? recognition and localization of acoustic signals by insects. Current Opinion in Neurobiology 10, 763-767.

Popper, A. N. and Fay, R. R. (2005). Sound Source Localization. New York: Springer.

Robert, D. and Gopfert, M. C. (2002). Novel schemes for hearing and orientation in insects. Current Opinion in Neurobiology 12, 715-720.

Romer, H. and Krusch, M. (2000). A gain-control mechanism for processing of chorus sounds in the afferent auditory pathway of the bushcricket Tettigonia viridissima (Orthoptera; Tettigoniidae). Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 186, 181-191.

Romer, H., Bailey, W. and Dadour, I. (1989). Insect hearing in the field III. Masking by noise. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 164, 609-620.

Saberi, K., Dostal, L., Sadralodabai, T., Bull, V. and Perrott, D. R. (1991). Free-field release from masking. J. Acoust. Soc. Am. 90, 1355-1370.

Shinn-Cunningham, B. G., Schickler, J., Kopco, N. and Litovsky, R. (2001). Spatial unmasking of nearby speech sources in a simulated anechoic environment. J. Acoust. Soc. Am. 110, 1118-1129.

Stumpner, A., Ronacher, B. and von Helversen, O. (1991). Auditory Interneurones in the Metathoracic Ganglion of the Grasshopper Chorthippus Biguttulus: II. Processing of Temporal Patterns of the Song of the Male. J Exp Biol 158, 411-430. von Helversen, D. (1984). Parallel processing in auditory pattern recognition and directional analysis by the grasshopper Chorthippus Biguttulus L. (Acrididae). Journal of Comparative Physiology 154, 837-846. von Helversen, D. and von Helversen, O. (1995). Acoustic pattern recognition and orientation in orthopteran insects: parallel or serial processing? Journal of Comparative Physiology a- Sensory Neural and Behavioral Physiology 177, 767-774.

Wakeford, O. S. and Robinson, D. E. (1974). Detection of binaurally masked tones by cat. J. Acoust. Soc. Am. 56, 952-956.

Walker, T. J. (1993). Phonotaxis in female Ormia ochracea (Diptera: Tachinidae), a parasitoid of field crickets. J. Insect Behav. 6, 389-410.

Wendler, G. (1989). Acoustic orientation of crickets (Gryllus campestris) in the presence of two sound sources. Naturwissenschaften 76, 128-129.

100

Figure Legends

Figure 1. Experimental setup for tethered-walking phonotaxis. A) Spectral analysis of the simulated cricket trill (ii top panel, inset show expanded view of the temporal pattern) show sharply tuned peak energy at 5 kHz (i), and random noise (ii lower panel, inset show expanded view of the random noise) with an approximate 10 dB drop in energy content spanning 3-7 kHz range. In test conditions, noise (4 seconds in duration) was broadcasted 0.5 second post data acquisition (ii lower panel), followed by the simultaneous broadcast of cricket song (2 seconds in duration) 1.5 seconds into data acquisition (ii top panel). B) Gravid female O. ochracea tethered ontop of the trackball system and positioned equidistant (10 cm) from surrounding speakers. Cricket song was broadcasted from the forward (red) speaker and noise was broadcasted from an adjacent speaker (light blue – either to the left or right of the song speaker) for the grouped condition, or from a laterally positioned speaker (dark blue - either to the left or right of the song speaker) for the spatially separated condition.

Figure 2. Estimated song intensity response thresholds. Average response thresholds were lowest in the absence of the 76 dB SPL continuous noise masker. Average response thresholds were elevated in the presence of noise. Average response thresholds did not differ for song and noise that were spatially grouped or separated.

Figure 3. Forward velocities from phonotaxis. A) Average forward velocity (solid lines – changes in y units over time) and SEM (dotted lines) to a 76 dB SPL song alone (black trace) or song and noise broadcasted at SNRs of -6dB (orange trace), equal dB (blue trace), +6dB (purple trace) from song and noise that were spatially grouped. B) Average forward velocity and SEM (solid lines – changes in y units over time) and SEM (dotted lines) to a 76 dB SPL song alone (black trace) or song and noise broadcasted at SNRs of -6dB (orange trace), equal dB (blue trace), +6dB (purple trace) from song and noise that were spatially separated. Light pink shaded areas (A, B) indicate the duration of noise broadcast. Dark pink shaded areas (A, B) indicate the duration of simultaneous presentation of song and noise broadcast. At noise onset, flies responded with a brief walking component in the forward direction that quickly tapered off. At cricket song onset, forward velocities (forward movement to forward cricket song) increased to speeds well above responses to the noise onset. Average forward velocities depended on SNR but not song and noise separation.

101

Figure 4. Steering velocities from phonotaxis. A) Average steering velocity (solid lines – changes in x units over time) and SEM (dotted lines) to a 76 dB SPL song alone (red speaker, black trace) or song and noise broadcasted at SNRs of -6dB (orange trace), equal dB (blue trace), +6dB (purple trace) from song (red speaker) and noise (light blue speaker) that were spatially grouped. B) Average steering velocity (solid lines – changes in x units over time) and SEM (dotted lines) to a 76 dB SPL song alone (red speaker, black trace) or song and noise broadcasted at SNRs of -6dB (orange trace), equal dB (blue trace), +6dB (purple trace) from song (red speaker) and noise (dark blue speaker) that were spatially separated. Light pink shaded areas (A, B) indicate the duration of noise broadcast. Dark pink shaded areas (A, B) indicate the duration of simultaneous presentation of song and noise broadcast. At noise onset, flies reactively steered to the noise, followed by steering movements away from noise that was maintained for the remainder of stimulus presentation. Steering responses were minimally affected by SNR but were enhanced at greater song and noise separation.

Figure 5. Virtual walking trajectories from phonotaxis. A) Average walking responses to a 76 dB SPL song alone (red speaker, black trace) or song and noise broadcasted at SNRs of -6dB (orange trace), equal dB (blue trace), +6dB (purple trace) from song (red speak) and noise (light blue speaker) that were spatially grouped. B) Average walking responses to a 76 dB SPL song alone (red speaker, black trace) or song and noise broadcasted at SNRs of -6dB (orange trace), equal dB (blue trace), +6dB (purple trace) from song (red speak) and noise (dark blue) that are spatially separated. Walking distance changed with SNR but not song and noise separation. Noise caused walking responses to be diverted away from the location of song and noise. Increases in source separations diverted walking response even further in a direction contralateral to the location of the noise.

Figure 6. Current model of directional noise avoidance in Ormia ochracea. The forward cricket song results in equal (symmetrical) stimulation to both tympana while the localisable band- limited random noise provides greater (asymmetrical) stimulation to the noise-ipsilateral tympana compared to the noise-contralateral tympana. This may produce binaural effective (perceived) amplitude differences in the detection of the cricket song pulse train above the random noise floor. Effective pulse amplitude differences translate into intensity dependent firing latencies from the left and right populations of type 1 afferents. Population level mean

102 latency differences encode perceived song direction, resulting in biased phonotaxis away from both song and noise.

103

Figure 1.

104

105

106

107

108

109

Chapter 6 General Discussion

Sound source perception demands the auditory system to recognize and localize salient sound sources in acoustic environments that may be complicated by clutter or distracting noise (Brumm and Slabbekoorn, 2005; McDermott, 2009). Resulting acoustic reverberations and noise may overlap spatially and temporally and mask critical signal parameters evaluated for recognition and localization of sound sources (Kuczynski et al., 2010; Bee, 2012). Under these conditions, solving the auditory scene requires the detection, recognition, and segregation of signal elements that arise from the same source identity amongst noise (McDermott, 2009; Bee, 2012). In my thesis, I seek to examine auditory system function involved in the recognition and localization of individual sound sources under complex acoustic conditions that maybe encountered by an invertebrate hearing specialist.

The acoustic parasitoid fly Ormia ochracea provides a unique opportunity to examine auditory specializations involved in sound source localization under single source conditions and in more complex acoustic environments. Directional hearing in O. ochracea has evolved solely in the context of host localization (Cade, 1975). Female O. ohcracea are expected to be within earshot of multiple singing field crickets and possibly other sources of distracting noise (Cade, 1981; Brumm and Slabbekoorn, 2005). Previous work has documented several auditory specializations that O. ochracea possess to overcome the challenges of directional hearing faced by small animals (Robert and Gopfert, 2002). In particular, the two tympanal hearing organs are mechanically coupled and specialized to amplify minute sound field time differences into larger tympanal time and amplitude response differences for the neural processing of sound direction (Robert and Gopfert, 2002). Most of the primary auditory afferents are adept at precisely marking the onset of sound pulses that convey the temporal pattern of field cricket calling songs (Oshinsky and Hoy, 2002), and detecting interaural time differences proposed to encode sound direction (Mason et al., 2001; Oshinsky and Hoy, 2002). How these auditory specializations may possibly constrain solutions to the ubiquitous problem of representing multiple sound sources in noisy environments is unknown.

A first step to studying sound source segregation requires an understanding of particular features in acoustic signals evaluated for sound recognition (Pollack, 2000). Phonotaxis

110

experiments exploit the animals’ natural behavioural repertoire to elucidate signal preferences in no-choice or dual-choice experimental paradigms (Gerhardt, 1995). Both types of choice paradigms confer different sets of advantages and disadvantages. Dual choice experiments may allow simultaneous testing of preferences for two different signal alternatives, but overlapping signals may interfere with the ability of animals to distinguish fine-scale signal features (Romer et al., 1989; Brumm and Slabbekoorn, 2005; Gerhardt, 1982). Conversely, no-choice experiments may allow animals to evaluate fine-scale signal features but animals are often found to be less selective for signals under no-choice conditions (Doherty, 1985; Wagner, 1998) . The goal of Chapter 2 was to derive a phonotactic performance index sensitive to capturing response variation underlying song recognition in O. ochracea. I utilized a high-speed trackball system to record walking phonotaxis (Lott et al., 2007) of flies subjected to a no-choice paradigm. To address the problem of less discriminant behaviour often associated with no-choice paradigms, flies were subjected to test songs that switched in the broadcast location midway through stimulus presentation. This switch in broadcast location forced animals to change the course of their response and they only did so in response to preferred song types (Chapter 2). Flies were observed to be less selective for different song types in response to initial broadcast locations, but showed greater discriminatory behaviour in tracking a switch in song location. I discovered several response features (walking distance, steering velocity, and localization accuracy) that varied with known song preferences (Walker, 1993) and confirm their validity for inclusion in an overall phonotactic performance index to measure variation in song recognition.

In Chapter 3, I applied this phonotaxis performance index to quantify response variation underlying song recognition. Using a range of test stimuli that varied in different combinations of pulse durations and interpulse intervals, I characterized the temporal pattern recognition space of Floridian O. ochracea. Song recognition in Floridian O. ochracea is based on evaluating song pulse period. Flies show preference for a range of pulse period values that include the calling songs of several host cricket species parasitized in the continental United States (G. rubens, G. texensis, G. lineaticeps0, but not in Hawaii (T. oceanicus). These results also demonstrate that specializations in the physiological response properties of primary auditory afferents in O. ochracea confer a temporal pattern recognition mechanism that allow for some level of flexibility in the recognition of several host species.

111

The goal of Chapter 4 was to determine how sound source localization may occur in complex acoustic environments characterized by multiple attractive sources that overlap in time and space. Field crickets are commonly found to advertise in aggregations and overlapping calls may mask the detection of pulse periods for song recognition (Chapter 3). Two synchronous coherent sound sources of equal intensity may give rise to an auditory illusion commonly encountered in binaural hearing (Blauert, 1997). That is, a phantom sound source is perceived mid-way between the two actual source locations. In Chapter 4, I present behavioral results that demonstrate this phantom sound source illusion in O. ochracea (Lee et al., 2009). In field sound trap experiments, flies were able to exploit small time differences to localize a single sound source most of the time, but were occasionally diverted to a phantom source location (Lee et al., 2009). Under tightly controlled laboratory conditions, I varied the relative timing of two competing sound sources and found that within a small temporal window of attention, localization responses were biased in the direction of a leading source over lagging sources (Lee et al., 2009). These experiments demonstrate that flies experience the precedence effect: signals leading in time are selectively localized and lagging signals have little influence in response orientation. With greater time differences between two competing sound sources beyond the temporal window of attention, the two sound sources are perceived as distinct sources.

Segregating sound sources of interest over distracting noise may be improved with spatial separation between target signals and noise (Bronkhorst, 2000; Shinn-Cunningham et al., 2005). When a masker is displaced laterally relative to a forward target signal, sound diffraction results in masker interaural level differences (ILD) and interaural time differences (ITD) that are absent from the target signal (Bee and Micheyl, 2008). Sensory mechanisms involved in spatial release from masking may exploit these differences to isolate target signals from noise. The focus of Chapter 5 was to test the hypothesis that spatially mediated release from masking may contribute to successful song recognition and sound localization in O. ochracea. Behavioural evidence in support of spatial release from masking may include a reduction in detection thresholds and improved localization accuracy for target signals spatially separated from noise. By varying the signal-to-noise ratio and spatial separation between the cricket song and band-limited masking noise, I show that response thresholds did not differ between spatially grouped and separated song and noise. Instead, an increase in spatial separation between song and noise caused flies to

112

divert localization responses in a direction that is away from both song and noise (Chapter 5). These results do not provide support for spatial release from masking in O. ochracea.

Auditory specializations involved in selectively attending to individual signals may come at the cost of the ability to maintain separate streams of information that correspond to different source identities. Song recognition in acoustically communicating invertebrates generally requires the auditory system to extract information from the temporal envelope of song types (Pollack, 2000; Gerhardt and Huber, 2002; Hedwig, 2006). Successful song localization depends on measuring time and/or intensity differences from correctly segregated song features that correspond to a single source. These processing tasks may be organized in two different arrangements that are contingent on the organization of neural networks involved in song recognition and sound localization (von Helversen and von Helversen, 1995). In parallel processing, information relevant to recognition and localization are processed by different information channels (von Helversen and von Helversen, 1995). Neural circuits involved in song recognition are independent of those involved in sound localization (von Helversen and von Helversen, 1995). One example of this information processing scheme is found in grasshoppers (von Helversen and von Helversen, 1995). When two songs, each with incorrect temporal patterns are presented on either side of the animal and timed to result in a correct temporal pattern, grasshoppers are found to integrate the two patterns centrally to perform phonotaxis (von Helversen, 1984). However, two effective patterns that integrate to an incorrect song fail to elicit phonotaxis (von Helversen, 1984). Thus song temporal pattern arriving at both ears are centrally combined and in most cases, preclude the representation of sound direction within the same information channel (von Helversen and von Helversen, 1995). Instead, separate neural circuits seem to be involved in the processing of directional information (Ronacher et al., 1986; Stumpner et al., 1991; Stumpner and Ronacher, 1994). In serial processing, information for recognition and localization are processed sequentially along one information channel (von Helversen and von Helversen, 1995). Separate temporal filters on either side of the auditory system are required to recognize appropriate song patterns that influence the computation of sound direction (von Helversen and von Helversen, 1995). Sound direction in serial processing may be determined by a central comparator that measures differences in song patterns on both sides of the auditory system to drive localization responses to the side with the best pattern (von Helversen and von Helversen, 1995). Song recognition and sound localization in crickets was

113

thought to be a convincing example of serial processing (Pollack, 1986; Stabel et al., 1989). When crickets were subjected to the simultaneous broadcast of cricket song from above (no directional information, but source is recognized) and a tone from an azimuthal location (not recognized but can be localized), phonotactic responses were directed away from the tone (Stabel et al., 1989). Measurements from auditory interneurons AN1 and AN2 involved in temporal pattern processing suggest that crickets are localizing a direction corresponding to the best song temporal pattern (Pollack, 1986; Stabel et al., 1989). More recent work in cricket auditory processing reveal that crickets reactively localized individual sound pules on a timescale that preclude the measurement of song temporal pattern (Hedwig and Poulet, 2004; Hedwig and Poulet, 2005). Reactive steering responses to individual sound pulses contribute to the overall sound direction that is localized (Hedwig and Poulet, 2005). However, temporal pattern recognition appears to modulate the strength of phonotactic responses (Poulet and Hedwig, 2005).

The organization of song recognition and sound localization in O. ochracea are currently unknown. One of the most informative experiments to determine processing strategies depend on independent stimulation of opposite sides of the auditory system with different temporal patterns (von Helversen, 1984). In the most extreme form of the split-song paradigm, every other sound pulse is delivered from opposite directions such that each side of the auditory system receives a temporal pattern that may not be recognized when presented separately, but combine to form a song that elicits phonotaxis (von Helversen, 1984; von Helversen and von Helversen, 1995). In O. ochracea, the pulse period from one half of the song is beyond the range of recognized pulse periods (pulse duration of 10 msec and interpulse interval of 30 msec, pulse period of 40 ms), and does not elicit phonotaxis (Chapter 3). Song temporal patterns that impinge on the ipsilateral tympanum are mechanically transferred to contralateral tympanum at lower amplitudes (Robert et al., 1996). When alternating sound pulses from separate locations are timed to form a complete song, both auditory organs are expected to respond to each sound pulse even if every other sound pulse is lower in effective amplitude. For two complete songs that are presented on opposite sides of the auditory system and timed to form a doubled-pulse rate song, flies recognize and perform phonotaxis to an intermediate location between both sources but with a slight bias in the direction of the leading pulse (Chapter 4). Song recognition may occur as both sides of the auditory system can detect the onset of sound pulses from the ipsilateral source while the lower

114 effective amplitude of the contralateral song has less of an influence and may appear as the interval between pulses that form a complete song. However, when two complete songs are interdigitated to form a doubled-pulse rate song and presented from a common location flies fail to respond (Chapter 2). Thus separate recognizers on the left and right side of the auditory system may contribute to song recognition, while sound localization may depend on responding to a location that corresponds to the best temporal pattern.

Collectively, my findings suggest that the ears of O. ochracea may function as symmetry detectors for sound intensity, timing between competing sound sources, and song temporal pattern. That is, sound localization may be based on measuring a balanced input of the appropriate song temporal patterns at the same time and at equal intensity and this corresponds to a forward source location (Chapter 4). When presented with spatially separated song and noise, O. ochracea are diverted away from both sources in a direction that is expected to equalize the representation of song temporal pattern on both sides of the auditory system (Chapter 5). Asymmetries in timing and/or sound intensity result in unbalanced input to both sides of the auditory system, and turning responses are made in a direction to re-establish symmetrical inputs (Chapter 4). One functional constraint of this bilateral symmetry detection system is that multiple coherent sources in the environment may result in ambiguous directional information (phantom source) that may not correspond to actual source locations.

115

References

Bee, M. A. (2012). Sound source perception in anuran amphibians. Current Opinion in Neurobiology.

Bee, M. A. and Micheyl, C. (2008). The cocktail party problem: What is it? How can it be solved? And why should animal behaviorists study it? J. Comp. Psychol. 122, 235-251.

Blauert, J. (1997). Spatial Hearing: The Pyschophysics of Human Sound Localization. Cambridge, MA: The MIT Press.

Bronkhorst, A. W. (2000). The cocktail party phenomenon: A review of research on speech intelligibility in multiple-talker conditions. Acustica 86, 117-128.

Brumm, H. and Slabbekoorn, H. (2005). Acoustic Communication in Noise. In Advances in the Study of Behavior, vol. Volume 35, pp. 151-209: Academic Press.

Cade, W. (1975). Acoustically orienting parasitoids - fly phonotaxis to cricket song. Science 190, 1312-1313.

Gerhardt, H. C. (1995). Phonotaxis in Female Frogs and Toads: Execution and Design of Experiments. In Methods in Comparative Psychoacoustics, pp. 209-220. Basel, Switzerland: Birkhauser Verlag.

Gerhardt, H. C. ( 1982). Sound pattern recognition in some north american treefrogs (Anura, Hylidae): implications for mate choice. American Zoologist 22, 581-595.

Gerhardt, H. C. and Huber, F. (2002). Acoustic Communication in Insects and Anurans: Common problems and diverse solutions. Chicago: The University of Chicago Press.

Hedwig, B. (2006). Pulses, patterns and paths: neurobiology of acoustic behaviour in crickets. J. Comp. Physiol. A -Neuroethol. Sens. Neural Behav. Physiol. 192, 677-689.

Hedwig, B. and Poulet, J. F. A. (2004). Complex auditory behaviour emerges from simple reactive steering. Nature 430, 781-785.

116

Kuczynski, M. C., Velez, A., Schwartz, J. J. and Bee, M. A. (2010). Sound transmission and the recognition of temporally degraded sexual advertisement signals in Cope's gray treefrog (Hyla chrysoscelis). Journal of Experimental Biology 213, 2840-2850.

Mason, A. C., Oshinsky, M. L. and Hoy, R. R. (2001). Hyperacute directional hearing in a microscale auditory system. Nature 410, 686-690.

McDermott, J. H. (2009). The cocktail party problem. Current Biology 19, R1024-R1027.

Oshinsky, M. L. and Hoy, R. R. (2002). Physiology of the auditory afferents in an acoustic parasitoid fly. Journal of Neuroscience 22, 7254-7263.

Pollack, G. (2000). Who, what, where? recognition and localization of acoustic signals by insects. Current Opinion in Neurobiology 10, 763-767.

Robert, D. and Gopfert, M. C. (2002). Novel schemes for hearing and orientation in insects. Current Opinion in Neurobiology 12, 715-720.

Romer, H., Bailey, W. and Dadour, I. (1989). Insect hearing in the field III. Masking by noise. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 164, 609-620.

Ronacher, B., von Helversen, D. and von Helversen, O. (1986). Routes and stations in the processing of auditory directional information in the CNS of a grasshopper, as revealed by surgical experiments. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 158, 363-374.

Shinn-Cunningham, B. G., Ihlefeld, A., Satyavarta and Larson, E. (2005). Bottom-up and top-down influences on spatial unmasking. Acta Acustica United with Acustica 91, 967-979.

117

Stabel, J., Wendler, G. and Scharstein, H. (1989). Cricket phonotaxis: localization depends on recognition of the calling song pattern. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 165, 165-177.

Stumpner, A. and Ronacher, B. (1994). Neurophysiological aspects of song pattern- recognition and sound localization in grasshoppers. American Zoologist 34, 696-705.

Wagner, W. E. (1998). Measuring female mating preferences. Animal Behaviour 55, 1029- 1042.

Walker, T. J. (1993). Phonotaxis in female Ormia ochracea (Diptera, Tachinidae), a parasitoid of field crickets. J. Insect Behav. 6, 389-410.