This is the first demonstration that the startle reflex leads to sensitisation of an extreme avoidance response which constitutes a rare example of a sustained sensitisation process tied to a simple reflex. Sensitisation in the sense of an increased responsiveness to the stimulus was found in a variety of response variables i.e. increased haulout time and reduced time spent close to the feeder in later playback sessions. Similarly, the likelihood of flight responses occurring increased rapidly in later playback sessions which is another example of response sensitisation. Treatment with long rise time stimuli on the other hand led to a waning of avoidance responses indicating habituation in experiment 3. This shows that it was the startle reflex and not the defense reflex that caused sensitisation of flight behaviour. Long rise time, high intensity stimuli can elicit the defence reflex [2, 21], which has been interpreted as part of the fight and flight reaction of animals . However, using such stimuli, the animals showed a quick habituation (decreased responsiveness) of avoidance behaviour and decreasing frequency of flight responses. These results are also consistent with our earlier study that tested grey seal avoidance behaviour in response to non-startling, longer duration sound types in which seals were found to habituate rapidly i.e. flight responses waned, animals spent increasing amounts of time close to the feeder and never hauled out during the experiment . Three of the test subjects that sensitised in experiment 1 where also used in our previous study in which they habituated to all stimuli i.e. avoidance behaviour waned. Thus, while the defence reflex might be involved in initial flight responses, in our study only elicitation of the startle reflex resulted in sensitisation of avoidance responses and increased the likelihood of flight responses.
A comparison of the startle threshold from this study with previous studies showed that the startle threshold expressed in units of sensation levels (dB above hearing threshold) is similar to the sensation level required to induce startle in rats  and humans  (Figure 3 and Table 3). Thus, the startle threshold may be fairly universal and conserved among mammals in spite of specific adaptations to aquatic hearing in seals (Table 3). The two seals that did not show sensitisation in subsequent avoidance behaviour in experiment 1 also never showed an observable startle response, not even at the highest tested received level of 180 dB re 1 μPa (experiment 2) This suggests that elicitation of the startle reflex was necessary to evoke sensitisation of avoidance responses in these animals. We suspect that the two non-startling animals had impaired hearing since they were among the oldest animals tested and because in mice the sound pressure level required to elicit a startle response increases with hearing loss . While the exact threshold in mammals depends not only on the received level of the sound but also on stimulus duration and rise time , the sensation level value typically lies at about 90 dB above the hearing threshold if rise times of about 5 ms are used and the duration is kept constant (Table 3). This sensation level remains similar to the original level even when age-related hearing loss sets in (i.e. absolute startle thresholds rise with increasing hearing loss ). One study on mice  found a lower startle threshold (Table 3) but used stimuli with an almost instantaneous rise time which is known to lower the required intensity threshold for a startle response. The majority of the animals tested in our study were females. Even though the only male that we tested also sensitised, it would be interesting to explore sex differences in these responses in more detail. In mice, males have higher startle magnitudes and more pronounced long-term habituation of startle magnitudes than females . Gonadal hormones such as estradiol- and dihydrotestosterone on the other hand can cause a decrease in startle magnitude in rats with gonadectomy . However, it is important to note that it is unclear whether the magnitude of the startle reflex (strength of muscular contraction) is in any way related to the aversive follow-up response (avoidance, flight) observed in our experiments. Furthermore, startle modulation as a result of sex or hormonal differences is unlikely to explain the lack of observable startle responses found in two of the females that habituated.
The behavioural responses observed in experiment 1 and 3 are remarkably similar to those observed in studies that involved electric stimulation of the brain. Repeated electrical stimulation of the amygdala or the defence circuitry in the superior colliculus leads to long-term sensitisation resulting in anxiogenic-type consequences and pronounced flight response to subsequent stressors [29, 30]. Similarly, stimulation of the acoustic pathway in the inferior colliculus at increasing intensities first leads to freezing and then ultimately to escape behaviour  and sensitisation . Although the primary startle pathway is thought to be mediated by the cochlea nucleus which projects into the pontine reticular formation [4, 5], the latter structure also receives indirect acoustic input from the inferior colliculus . Furthermore, previous studies have shown that the magnitude of the startle reflex can be increased by fear-inducing experiences  and startle has long been used as an indicator of fear  and emotional state . Our data showed that a startle stimulus can act as an unconditioned stimulus in a fear conditioning paradigm, as also suggested by an ethically questionable experiment on one human baby . Thus, the startle reflex is not only influenced by emotional state  but repeated exposure to startling stimuli appears to cause fear. This indicates the presence of an afferent input from the startle pathway to brain areas related to emotional processing such as the amygdala and shows that the mammalian startle reflex evolved most likely in the context of general predator avoidance. Interestingly, the projection from the startle pathway to the amygdala  and its effects has received little research attention while the efferent connection from the amygdala to the startle pathway is of great significance in major research efforts using fear-potentiated startle as an indicator of fear conditioning through other stimuli and as a measure of emotional valence of such stimuli [6, 36, 37].
The startle reflex is commonly used as a measure for emotional processing in studies on human anxiety disorders . Patients with panic disorder, post-traumatic-stress disorder (PTSD) or obsessive compulsory disorder (OCD) generally show elevated baseline startle magnitudes . Our study indicates that repeated startling influences emotional processing. Thus, the potential role of repeated startle elicitation in the development of post-traumatic stress disorder should be considered. The main behavioural categories thought to characterize post-traumatic stress disorder (PTSD) in animal models are "conditioned behaviours" (i.e. fear conditioning) and "sensitised behaviours" (e.g. hypervigilance) . In our study, we found evidence for both "conditioned" (fear conditioning) and "sensitised behaviour" (increasing flight responses, reluctance to approach feeder) as a result of exposure to repeated startling stimuli. While the severity of the behaviour patterns observed in this study is probably less strong than in the PTSD model, our data show that long-term exposure of humans to pulsed noise should be critically evaluated.
We think it is likely that the original function of the startle reflex is associated with increasing an animal's propensity for flight as required in a predator avoidance scenario . If the biological function of the startle reflex was primarily associated with injury prevention through increased muscle tonus  we would have expected an absence of flight and avoidance responses as a result of startle elicitation. Many startling sounds indicate serious threats caused by predators. These include sounds of breaking tree branches, falling rocks or the sudden impact noise of a predator attacking a conspecific. A sensitisation to startle sounds as observed in our study would be beneficial not only by enabling a rapid predator avoidance response but also by preventing an animal from moving into an area with an increased threat level where startling sounds are encountered repeatedly. Interestingly, animals may also exploit the startle reflex to manipulate conspecific, prey or predator behaviour. For instance, cod were found to produce potentially startle eliciting clicks before prey capture attempts by seals . Bottlenose dolphins produce high-intensity jaw pops as a threat display during courtship which could potentially startle conspecifics . Future research will be needed to address the question whether basic reflexes like the startle have shaped the evolution of communication signals.
In contrast to most neuro-physiological studies, the animals we tested here were captured from a wild population, where they had spent time in their natural habitat prior to the experiments. They also belonged to a taxon that is not closely related to any of the standard model systems. Our reason to choose the grey seal as a test species was partly a concern over observed detrimental responses of marine mammals to noise pollution. There are many anthropogenic noise sources in use that can cause startle responses. Gun shots and some industrial noise are examples in air. However, most pulsed noise caused by human activity can be found in the marine environment such as in underwater explosions, pile-driving activities, acoustic deterrent devices, sonar pulses and seismic air guns. Marine mammals have been found to abandon areas of high noise pollution  and even strand as an extreme behavioural avoidance response to military sonar [15–18]. While the role of the startle reflex in these reactions needs further study, it is notable that sonar sounds often have a very rapid onset and high source level. Our results showed that a simple oligo-synaptic reflex arc is responsible for extreme avoidance responses to sudden-onset, pulsed sounds. Thus, impact ratings of anthropogenic noise sources in air and in water should be re-evaluated and rise times of loud noise pulses should be increased to mitigate their effects on humans and animals alike.