Acoustic signals - pattern recognition Flashcards
Possible mechanisms of pattern recognition
1) Feature recognition - the same circuit that generates the signal can recognise it. Perhaps a separate detector would take input from the signal generator and from the acoustic signal, and see if rate matched. This would require some overlap between CPG and acoustic processing networks, which has not been found.
2) An oscillator - the first sound pulse heard would set off oscillations of a period that matches the species-specific call. Subsequent pulses, if in phase, would strengthen the signal.
3) Coincidence detectors - the auditory signal is sent directly to the coincidence detector via one neuron, but passes through a delay before reaching the detector via another neuron. This delay would be the period of the signal, and thus must be established via inhibitory mechanisms, since period is usually many msec.
Pattern recognition in bush crickets
In some species of bush cricket, male and females sing a duet of ultrasound signals. In one species, the male will only approach if the female’s response is within 25-35 msec. This limits phonotaxis to only a few metres (take into account sound travel time).
In another species, the male gives a trigger pulse 350msec after the chirp, to which the female responds in 35-40msec. She seems to ‘expect’ the single chirp.
How is this delay built into the neural network?
Thoracic auditory interneurons respond without obvious temporal filter properties - they just copy the song, or respond specifically to the 28Hz female response. As does LBN2.
However, LBN9 interneuron in the brain is inhibited for several hundred milliseconds. The inhibition wanes just in time for the female response.
Pattern recognition in Tettigonia cantans (bushcricket)
They exhibit phonotaxis best to 25kHz patterns, but also a weaker phonotaxis to 12.5kHz.
Shifting the phase of the artificial song showed that the insects oriented most strongly to patterns that were in phase or partially in phase with 25kHz.
This suggests the existence of an oscillatory network, where transient oscillation is set up by the first pulse. Pulses in phase with the oscillation cause the system to resonate and give a stronger response.
Pattern recognition in female crickets - characteristics of local brain neurons, predictions of a coincidence detector, role of each neuron in this.
LN2, 3 and 4 are the brain neurons directly contacted by auditory interneurons, so are the prime candidates for pattern recognition AN1 interneuron (the only one that carries the auditory info to the brain) and LN2 local brain neuron just copy the song being heard, with a slight adaptation common in sensory neurons. LN2 follows AN1 with v short latency. LN3 and LN4 sit close to axonal arborisations of AN1 and LN2. LN3 always shows a stronger EPSP to second pulse than to first. LN4 has hyperpolarisation then depolarisation but rarely a spike to first pulse, then bigger depolarisation and spike to second; it showed a band-pass selectivity for pulse intervals between 15 and 25ms. They both have longer latency, and spiking tuning matches the phonotactic tuning. LN5 does not spike, even from injected current. In response to a pulse, it first hyperpolarises to a maximum -9mV, then depolarises to maximum 5mV. The depolarisation is truncated by the next pulse, so it's only allowed to decay slowly after the last pulse in a chirp. Importantly, LN5 is intensity-independent, so it'll respond even if the song is quiet e.g. far away.
Predictions of coincidence detector - response to 2nd pulse stronger, response best at correct pulse period (34-42 ms for crickets), preference of particular pulse duration.
The behaviour matches that predicted by a coincidence detector model - the female will only orient if the delay between two pulses is a certain length.
The delay line appears to be non-spiking interneuron LN5, which exhibits post-inhibitory rebound.
Overall: AN1 transduces the song. Excitatory synapses directly onto coincidence detector LN3. Also excitatory synapses onto LN2, which inhibits LN5. LN5 has intrinsic post-inhibitory rebound, so then stimulates LN3.
If the delayed input comes into LN3 in phase with a direct input, then the pattern is recognised, and sent on to LN4.
LN4 is a feature detector, that only spikes at the second pulse, and only if the interval was between 15 and 25ms
Processing of pattern recognition
Both directionality and pattern recognition are key to the behavioural response to an acoustic signal
Directionality relies on differences between the signal received at either ear, while pattern recognition is improved by summing responses from both ears to create a stronger signal
Speakers at either side of female crickets have shown that they do in fact sum the info from each ear in pattern recognition
When given a non-attractive song but with easier directionality cue, and an attractive song with harder directionality cue, crickets oriented towards the non-attractive song.
Therefore pattern recognition and localisation are processed in parallel.
Functional adaptations of pattern recognition
Grasshoppers sing between 17 and 40°C. As temperature increases, legs move faster and frequency of song increases. To allow for this and maintain pattern recognition, preferred frequency (for phonotaxis) also increases with temperature!
But narrowing it down, the frequency of pattern generation is altered specifically by temperature of thorax (because metathoracic ganglion is site of CPG), whereas frequency preference is altered by heating the brain
Where olfactory (moth) and visual (bees) cues are involved in orientation, learning is often involved. In insects orienting to auditory cues, it’s more hardwired. Females perform small reactive steering movements towards any sound pattern at all, but hearing the male’s song increases gain of steering within 2-5 seconds, so gain control in the auditory-to-motor pathway may allow insects to follow species-specific cues even when corrupted by background noise etc.
Moiseff et al 1978 - Female flying (tethered) crickets oriented towards their species-specific pattern at 3-9kHz, but away from the same temporal pattern at 30-70kHz. SO whilst pulse:interval ratio is important in recognition, the correct frequency is permissive (since the high frequency mimics bat noises and causes avoidance behaviour)
What part of the song is most important for pattern recognition? How is this recognised?
Many cricket species sing with similar frequencies, so it seems temporal pattern is more important.
Syllable-to-pause ratio - which is also v affected by loss of one hindleg, because the legs don’t simply alternate, they sing out of phase.
Gradually altering syllable to pause ratio gradually altered response.
Neural bases:
AN4 is tonically active. During a normal song, interneuron AN4 exhibits bursting. When a gap is put into the song, AN4 stays inhibited (i.e. even lower than normal tonic activity)
Mosquitos and acoustic communication
Mosquitoes do not have specially evolved songs, or sound generators. But males recognise females based on the sound of their wingbeat, and its frequency is distinct from that of males.
The sound is very quiet because the mosquito is so small compared to the wavelength of sound produced by its wings.
This has led to the male evolving very sensitive antennae for auditory detection, with the same number of sensilla as the human cochlea.
In addition to being v sensitive, the mechanosensory neurons’ response to the sound of a female flyby is nonlinear - the gain in the system is amplitude dependent, and enhances male ability to detect the female.
[another parallel with the cochlea -m frequency dependent inhibition to sharpen tuning curves. also affects spike timing for temporal processing
Auditory recognition without pattern recognition
In some animals, behavioural responses are very similar between species. This indicates that they’ve evolved for the same purpose, e.g. identification of bats in mantids Yager 1999. In these species, we don’t understand why they’re even able to hear frequencies outside this relevant range. Mantids have multiple sensory neurons, but all tuned to the same frequency!
