Sensory drive in acoustic communication of insects

The sensory drive model of sexual selection emphasizes the evolutionary processes and complex interactions between signal production, ecology of transmission, perception and information processing, up to the final decision making. Because sensory systems and brains often have features as a result of a complex evolutionary history, these features will make some signals more efficient in eliciting responses in receivers than others, and, as a consequence, may drive the evolution in a particular direction. The current project will investigate three cases of sensory drive in the communication of acoustic insects (crickets and katydids) at the proximate level, which act at very early stages in auditory processing and range from biophysical to central mechanisms. Central mechanism of gain control: A property in the central nervous system of katydids and crickets enables species with a high duty cycle calling song to selectively listen to only few individuals in a chorus ("selective attention"). However, this very same property does result in negative consequences for signal detection in species with low duty cycle calling song, if they interact acoustically with other species. Proximate mechanisms of preferences for large and symmetrical males: Female field crickets prefer larger and more symmetrical males, based on acoustic cues of carrier frequency and the degree of frequency modulation in the calling song. The variability of tuning of the sensory organ and of interneurons represents a bias for female responses to male calling songs differing in these traits, and thus for their preference. Peripheral mechanism of directional tuning: Due to biophysical reasons and the existence of a "phase-shifter" in the ear of field crickets the directional sensitivity is strongly frequency dependent. Variance in the size of receivers will bias the amount of directional information available from males calling at different carrier frequencies.

Calling songs of male field crickets represent secondary sexual characters and are subject to sexual selection by female choice. Following predictions from the `matched filter hypothesis` we studied the tuning of an identified interneuron in a field cricket, known for its function in phonotaxis, and correlated this with the preference of the same females in two-choice trials. Females vary in their neuronal frequency tuning, which strongly predicts the preference in a choice situation between two songs differing in carrier frequency. A second `matched filter` exists in directional hearing, where reliable cues for sound localization occur only in a narrow frequency range. This second `matched filter` also varies widely in females, and surprisingly, differs on average by 400 Hz from the neuronal frequency tuning. Our findings on the mismatch of the two `matched filters` would suggest that the difference in these two filters appears to be caused by their evolutionary history, and the different trade-offs which exist between sound emission, transmission and detection, as well as directional hearing under specific ecological settings. The mismatched filter situation may ultimately explain the maintenance of considerable variation in the carrier frequency of the male signal despite stabilizing selection. We also studied these two filters in three further field cricket species, and in two of these the mismatch also exists, even to a larger extent. Only on T. commodus, which lives sympatrically with T. oceanicus, both filter a well matched, indicating that the selection pressure to separate the CF of calling songs causes the match. By analysing the activity of the pair of AN1-neurons simultaneously at suprathreshold levels in single stimulus and two-choice paradigms, we were able to correlate behavioural choice and performance with the actual discharge differences. In no-choice trials with stimuli differing in CF there was a high correlation between instantaneous bilateral discharge differences, the lateral steering and IIDs, and these also correlated with the probability of correct decisions based on AN1-activity. In two-choice trials with stimuli differing in CF, the AN1-discharge differences also correlated strongly with lateral steering and with the threshold differences for the alternative frequencies. In trading experiments we quantified the additional SPL necessary to compensate a given advantage in CF in two- choice trials. For most stimulus pairs the amount of SPL necessary for the trading could be predicted from the AN1-tuning. However, we found no evidence that the degree of fluctuating asymmetry, as expressed in the amount of downward frequency modulation in the second part of syllables, can be analysed by females on the basis of stimulus coding in AN1-neurons. The possible role of a sensory bias was also examined with respect to a neuronal mechanism previously described as `selective attention`. First, we investigated whether individuals of the bushcricket genus Mecopoda restricted their attention to louder chirps neurophysiologically, behaviorally and through spacing. We found that louder leading chirps were preferentially represented in the omega neuron but the representation of softer following chirps was not completely abolished. Following chirps that were 20 dB louder than leading chirps were better represented than leading chirps. These and further data let us assume that selective attention is achieved in this bushcricket through spacing rather than neurophysiological filtering of softer signals. In two independent studies we further examined whether `selective attention` is relevant for coding of important stimuli under conditions of increased background noise. We used both playback experiments in the lab, as well as the `biological microphone-approach` with field recordings to develop a `neuronal bat-detector`, which should be able to reliably detect the presence of bats in afferent spike trains. Indeed, the algorithm detected such activity with a probability of more than 90% even in high levels of background. Long-term recordings of omega-neuron activity over the period of several hours served as a data set to study the burst activity in response to background noise, and to compare these bursts with those in response to artificial playback stimuli. Using recently developed unsupervised machine learning techniques we show that burst responses to environmental noise can be well distinguished from responses to a set of artificial acoustic stimuli.