Current approaches to assess interoception of respiratory functions cannot differentiate between the physiological basis of interoception, i.e. visceral-afferent signal processing, and the psychological process of attention focusing. Furthermore, they typically involve invasive procedures, e.g. induction of respiratory occlusions or the inhalation of CO2-enriched air. The aim of this study was to test the capacity of startle methodology to reflect respiratory-related afferent signal processing, independent of invasive procedures. Forty-two healthy participants were tested in a spontaneous breathing and in a 0.25 Hz paced breathing condition. Acoustic startle noises of 105 dB(A) intensity (50 ms white noise) were presented with identical trial frequency at peak and on-going inspiration and expiration, based on a new pattern detection method, involving the online processing of the respiratory belt signal. The results show the highest startle magnitudes during on-going expiration compared with any other measurement points during the respiratory cycle, independent of whether breathing was spontaneous or paced. Afferent signals from slow adapting phasic pulmonary stretch receptors may be responsible for this effect. This study is the first to demonstrate startle modulation by respiration. These results offer the potential to apply startle methodology in the non-invasive testing of interoception-related aspects in respiratory psychophysiology.
This article is part of the themed issue ‘Interoception beyond homeostasis: affect, cognition and mental health’.
Interoception, the perception of bodily processes, plays an important role in health and disease. For example, accurate interoception is associated with factors that contribute to mental health, such as emotional experience , emotion regulation  or empathy . The majority of experimental paradigms to assess interoceptive processes focuses on the perception of cardiac sensations, such as the silent counting of heartbeats , or the discrimination of heartbeats and exteroceptive stimuli , which are considered proxies for cardiac interoceptive accuracy. Organ activation, the stimulation of interoceptors and the afferent neural signal transmission provide the physiological basis for visceral sensations. Interoception occurs when attention is directed towards physical sensations [6,7]. ‘Conventional’ methods to assess interoceptive accuracy, however, may have several shortcomings: (i) they cannot differentiate between the physiological basis of interoception and the psychological process of attention focusing and (ii) they rely on subjective reports, which are expected to be affected by various psychological states, e.g. motivation, affect and attention. The differentiation between actual afferent signals from the body and their perception is of particular relevance for the investigation of interoception in mental disorders, as they (e.g. panic or somatoform disorders) may be associated with selective disturbances in one or both of these processes.
An alternative to interoceptive tasks, which confound visceral sensations and their perception, is provided by the cardiac modulation of startle (CMS) [8–13]. CMS assesses afferent neural traffic originating from the cardiovascular system independent of the conscious perception of visceral sensations . More specifically, startle responses to acoustic noise stimuli have repeatedly been demonstrated to be lower during the early cardiac cycle phase (R-wave +230 ms), when the arterial pulse wave is expected to stimulate arterial baroreceptors in large blood vessels, compared with the late cardiac cycle phase (R +530 ms) [8–10,12,13]. Because this CMS effect was largely diminished in individuals with a degeneration of autonomic afferent nerves , this approach could serve as an indirect method to reflect baro-afferent neural signals. Neural traffic from arterial baroreceptors is considered one important source for interoceptive neural traffic from the cardiovascular system .
Afferent signals from other sources than the cardiovascular system may be required to better understand the role of adequate processing of bodily signals for health and disease. Disorders of the respiratory system, such as asthma [15–18] or chronic obstructive pulmonary disease [19–21], are associated with altered interoceptive signal processing from the respiratory system, which may contribute to the development of panic disorder in those patients [22–24]. Furthermore, alterations in breathing can be both a cause for (e.g. respiratory resistance, CO2-enriched air) and a consequence of anxiety, and respiratory interoception could thus be a useful target for the treatment of anxiety disorders . A range of methods for the assessment of respiratory interoception are available, however with certain limitations. Some techniques assess subjects' sensitivity to detect changes in respiratory resistance, thus relying on subjective reports [15,16,26,27]. Other methods involve invasive procedures, e.g. the inhalation of CO2-enriched air [28,29], or voluntary hyperventilation [30–32] to induce interoceptive fear. Respiratory-related evoked potentials are considered psychophysiological indicators of the cortical processing of afferent respiratory signals and thus represent an alternative to behavioural indicators of respiratory interoception; however, their determination is also based on respiratory occlusion events [33–35]. A common characteristic of these approaches is that their assessment provokes a highly uncommon and in many cases uncomfortable situation, thus limiting the ecological validity of findings.
To address these limitations, the aim of this study was to test the capacity of startle methodology to reflect the processing of afferent signals originating from the respiratory system. The respiratory cycle phase may affect the central processing of exteroceptive stimuli, and thus affect psychomotor reaction times. In line with this hypothesis, some studies have reported faster reaction times during inspiration [36–39]. In contrast, others have reported faster reaction times [40,41], increased signal detection  and higher visual event-related potentials during expiration . In anaesthetized cats, discharges in limb and lower intercostal nerves as indicators for startle responses have been shown to be lower during inspiration than during expiration . It could be argued that these effects are due to afferent signals from the respiratory system, presumably transmitted over pulmonary stretch receptors.
In this study, we tested the capacity of a new method to assess interoception of respiratory processes. This new approach is based on the quantification of startle reflex modulation. Reflexive eye blinks were measured in response to startling acoustic noise stimuli, presented at four different time points within the respiratory cycle: (i) at peak expiration, (ii) on-going inspiration (maximum slope), (iii) peak inspiration, and (iv) on-going expiration (maximum slope of inversed respiratory belt signal). Blood pressure is affected by respiratory-induced blood volume shifts within the pulmonary circulation and central control mechanisms integrating neural activity of pressure/stretch receptors located in the lung, the heart and large proximal vasculature. As a result, blood pressure oscillations follow the respiratory cycle with a time lag depending on respiratory frequency. As part of this study, we conducted a pilot study demonstrating that respiratory frequency of 0.25 Hz is associated with similar blood pressure levels at the two time points of ‘on-going inspiration’ and ‘on-going expiration’. This ensures that potential differences in startle responsiveness between these time points are unaffected by differences in blood pressure levels.
Forty-seven healthy, undergraduate students participated in the study and received monetary compensation of €20. Owing to a technical malfunction during data assessment, data of five participants were lost, resulting in a final sample of 42 (30 females; mean age: 23.8 [s.d. = 3.0] years; BMI: 22.5 [5.1] kg m−2). Self-reported physical health status was assessed by customized interview. Exclusion criteria were hearing impairments and tinnitus, regular use of contact lenses, any acute or chronic physical or mental health problems and regular use of medication. All participants provided written informed consent and were made aware of their right to discontinue participation in the study at any time.
Participants were seated in front of an LCD computer display in a comfortable chair. Participants were asked to remove their glasses if applicable, and electromyography (EMG) electrodes and the breathing belt were attached. Headphones (Sennheiser Electronic GmbH & Co. KG, Wedemark, Germany) were placed, and participants were informed about the experimental procedures on the computer display. They were asked to relax, to neither speak nor move, avoid longer periods of eye closure and listen carefully to all acoustic stimuli. Before the beginning of the main study procedure, participants were instructed to inhale a volume of 1, 2 and 3 l of air from customized cylindrical containers. These trials were intended to serve as ‘standardization trials’ for a later transformation of the raw respiratory belt data from arbitrary units into millilitres.
All participants were tested in two experimental breathing conditions: (i) during spontaneous breathing and (ii) during a condition with a 0.25 Hz paced breathing rhythm (inspiration–expiration ratio of 2 : 3). A respiratory frequency of 0.25 Hz was chosen based on results from a pilot study reported at the end of the Methods section. The order of appearance of experimental conditions was counterbalanced across participants. In both conditions, acoustic startle stimuli were presented at one defined time point during each of the four different phases within the respiratory cycle (figure 1): (i) at peak expiration, (ii) at midpoint during on-going inspiration, (iii) at peak inspiration and, (iv) at midpoint during on-going expiration. During each of these four respiratory phases, 10 startle stimuli were presented, resulting in 40 stimuli per breathing condition and 80 stimuli in total. The order of respiratory phases in which startle stimuli were presented was also counterbalanced within participants.
During the spontaneous breathing condition, participants focused on a fixation cross on the LCD screen. Startle stimuli were presented with a jittering interstimulus interval of 8–12 s. In the paced breathing condition, computer-controlled verbal instructions to inhale (‘in’) and exhale (‘out’) were given over the headphones. During startle stimulation, this auditory stimulation had to be interrupted without affecting a participant's breathing rhythm, because the auditory instructions could otherwise act as pre-pulses to the processing of the startle stimulus . The paced breathing condition was, therefore, subdivided into trials consisting of five respiratory cycles each. During the first three cycles, participants were verbally instructed to breath, while they saw a green light on the LCD screen. During cycles four and five, they were asked to continue breathing in the paced rhythm without verbal instructions. This part of the trial was indicated by a blue light on the screen. All startle stimuli appeared during the fifth respiratory cycle if the respective respiratory event was detected, whereas in 20% of those trials, there was no startle stimulus at all (figure 2).
(c) Respiratory pattern detection
For the elicitation of startle stimuli at the defined time points during the four phases within the respiratory cycle, an online analysis of the respiratory signal is required, which was conducted by a DASYLab 8.0-based algorithm developed at the Division of Clinical Psychophysiology, University of Trier. For each subject and each of the four respiratory time points, an individual reference template pattern was defined from a sequence of individual breathing cycles. Depending on the breathing phase and individual differences in respiratory parameters (e.g. breathing frequency, depth, etc.), this reference template pattern had a length of 1000–1500 ms and ended with the exact time point at which the startle elicitation was triggered (figure 1). For example, a section of 1000–1500 ms duration including the post-expiratory break and the first half of the increasing slope serve as template for ‘on-going inspiration’ (figure 1a), whereas the entire increasing slope of the signal until the local maximum serves as template for ‘maximal inspiration’ (figure 1b). Thereafter, the individual respiratory signal was continuously processed and compared with the reference template according to a pattern detection procedure using the following formula where Δ is the difference between the current signal and the reference pattern; s(t) is the signal value at time index t; p(t) is the reference pattern value at time index t; s′(t)/p′(t) is the first derivation of s/p and T is the pattern length.
When the difference between the current respiratory signal and the reference template pattern crosses a fixed threshold and a local minimum of this function occurs, the pattern is considered as ‘detected’. In a pattern calibration phase prior to each of the breathing conditions (spontaneous/paced), the reference template pattern and the threshold were defined individually for each participant. The threshold criterion was adjusted to achieve a sensitivity of at least 90% (nine out of 10 detected patterns in subsequent respiratory cycles) and a specificity of 100%. Offline evaluation showed that 13.8% (s.d. = 10.0%) of the breathing patterns in the spontaneous breathing condition and 6.3% (s.d. = 5.6%) of those in the paced breathing condition were not accurately detected, leading to an omission of startle stimulation in the respective trial. In this case, the stimulus was presented in the next trial, in which the pattern was detected (usually the subsequent trial).
(d) Technical parameters
EMG electrodes (Tyco Healthcare H124SG), which have a sensor diameter of 10 mm as recommended by Blumenthal et al. , were attached below the left eye with an interelectrode distance of 1.5 cm to assess the activity of the orbicularis oculi muscle. Startle stimuli consisted of acoustic white noise probes (105 dB, 50 ms duration, instantaneous rise time, binaural stimulation). Psychophysiological data were recorded on hard disk with a Biopac MP150 amplifier system (Biopac Systems, Inc., Goleta, CA) at 16-bit resolution and 1 kHz sampling rate. Hardware band-pass filter settings were 10–500 Hz, followed by a 28 Hz software high-pass filter . The raw signal was rectified and integrated online with a time constant of 10 ms . Respiratory activity was measured via a thoracic respiratory belt (James Long Company, Inc., Caroga Lake, NY) placed between the fifth and eighth ribs. The signal was band-pass filtered (0.05–10 Hz) before it was digitized. Electrocardiography (ECG) electrodes (Tyco Healthcare H34SG, Ag/AgCl) were placed according to the Einthoven lead II configuration. The ECG signal was high-pass filtered (0.5 Hz).
(e) Psychophysiological data analysis
A customized C++-based semi-automated PC program was used on a WinXP platform to analyse EMG responses offline. The algorithm identified response peaks in the rectified and integrated signal in the time interval of 20–150 ms after the startle probe onset. The baseline period was defined by a 50 ms interval prior to acoustic stimulation. All response data were manually inspected. Signals with electrical and physiological artefacts, such as coinciding blinks or excessive noise from other facial muscular activity, were rejected from analysis and defined as missing. If a participant's responses were not visible in their individual response latency range, the response amplitude was set to zero. Zero response data were included in the averaging procedure, with startle response magnitude as the final output measure . Averaging was carried out per subject separately for each respiratory phase and breathing instruction condition. Within each participant, startle response magnitudes were T-scored over all four respiratory phases and both breathing instruction conditions.
Respiratory cycles were automatically detected with WinCPRS software (Absolute Aliens Oy, Turku, Finland) and manually confirmed, from which breathing frequency data were derived. Arbitrary units as results of the raw respiratory belt signal were linearly transformed into millilitres based on individual values in standardization trials at the beginning of each session. Tidal volume as derived from these transformed values is denoted as ‘adjusted’ tidal volume.
(f) Distribution of startle stimuli across the cardiac cycle (control analyses)
In previous studies, we observed lower startle responses during the early when compared with the late cardiac cycle phase (i.e. CMS) [8–13]. To rule out that this CMS effect may have affected the hypothesized respiratory modulation of startle, we investigated in separate analyses any interaction effects of cardiac cycle and respiratory cycle phase on startle responses. While in earlier studies on CMS precise time points relative to the cardiac R-wave (e.g. 230 and 530 ms after R-wave) were used [9,10,12,13], in this study, we identified specific events during the respiratory cycle. As it is very unlikely that events in both physiological signals (e.g. peak inspiration and R-wave +230 ms) occur exactly at the same point in time, it is impossible to present startle stimuli dependent on events in both physiological signals simultaneously. To overcome this problem, we presented startle stimuli only dependent on the respiratory cycle, but evaluated the latency to the preceding R-wave offline to determine the cardiac cycle phase of each startle stimulus. Based on the summary by Edwards et al.  on the timing of arterial baroreceptor stimulation by the arterial pulse wave, we considered the time period of 90–390 ms after the R-wave the ‘early cardiac cycle phase’, while the remaining interval was considered the ‘late cardiac cycle phase’. We evaluated the number of startle stimuli appearing in the early and in the late cardiac cycle phase and the average latency of startle stimuli to the preceding R-wave for each of the respiratory cycle phases. In agreement with this period as suggested by Edwards et al. representing arterial baroreceptor stimulation, we previously observed lower startle stimuli in a largely overlapping interval (100–400 ms after the R-wave) when compared with the late cardiac cycle phase . Furthermore, we introduced cardiac cycle phase as additional factor in our analysis to investigate whether it interferes with the respiratory cycle phase in its effect on startle.
(g) Statistical analysis
To evaluate the impact of the respiratory phases on startle, we employed a repeated-measurements 4 × 2 ANOVA with the T-scored startle response magnitude as a dependent variable. The factors were ‘respiratory phase’ (peak expiration, midpoint during on-going inspiration, peak inspiration, midpoint during on-going expiration) and ‘breathing instruction’ (spontaneous versus paced breathing). The same 4 × 2 ANOVA was calculated to evaluate (i) the number of startle stimuli, which had occurred by chance during the early and (ii) the late cardiac cycle phase, and (iii) the average latency between startle stimulus and preceding R-wave for each of the respiratory phases and breathing conditions. A separate 4 × 2 × 2 repeated-measurements ANOVA was employed with the additional within-subjects factor ‘cardiac cycle phase’ and the T-scored startle responses as dependent variable to determine if the cardiac cycle phases systematically affected startle responses across respiratory conditions. Respiratory parameters (tidal volume, breathing frequency) across conditions with different breathing instructions were compared using a one-factorial two-levelled repeated-measurements ANOVA comprising the factor ‘breathing instruction’. Critical alpha-level was set to 0.05 in all analyses. All p-values of within-subject factors with more than two conditions are reported after Greenhouse–Geisser correction. Post hoc analyses of simple main effects within the ANOVA-models were performed with dependent t-tests.
(h) Analysis of pilot data
Beat-to-beat inter-beat interval (IBI), blood pressure (assessed with a Finapres System, Ohmeda Corp., USA) and thoracic impedance derived respiratory data reflecting lung volume changes of 10 healthy control subjects with a distribution of age and sex comparable to those in the main study (4 m/6 f; mean age: 25.9 years) were taken from a previous study, in which paced respiration was performed at various frequencies, ranging between 0.15 and 0.4 Hz . IBI, blood pressure and respiratory data were resampled with 5 Hz and adjusted to individual averages. Comparisons according to four time points of the respiratory cycle (figure 3) were performed. At a paced breathing frequency of 0.25 Hz, the difference between systolic blood pressure levels at ‘on-going inspiration’ (breathing phase ‘II’; 0.27 [95% CI: −0.28 to 0.83] mmHg) and ‘on-going expiration’ (breathing phase ‘IV’; −0.22 [95% CI: −0.79 to 0.35] mmHg) was minimal. This finding is in line with Laude et al.  demonstrating a synchronization of respiration and blood pressure in higher breathing frequencies. We, therefore, used a paced breathing condition of 0.25 Hz to minimize blood pressure differences between on-going inspiration and expiration.
(a) Startle eye blink responses
Startle response magnitudes did not differ between spontaneous and paced breathing conditions (F1,41 < 1). Nevertheless, we observed a significant main effect of ‘respiratory phase’ on startle responses (F3,123 = 6.17; p = 0.001; η2 = 0.13; figure 4). Post hoc analyses revealed that startle response magnitudes were higher during on-going expiration (M = 52.99 [s.e.m. = 0.83]) compared with peak inspiration (47.60 [0.77]), peak expiration (49.41 [0.72]) and on-going inspiration (50.00 [0.80]; all ps < 0.01). There was no interaction effect of ‘respiratory phase’ and ‘breathing instruction’ on startle responses (F3,123 < 1). In an exploratory analysis, we additionally entered ‘sex’ as between-subjects factor into the ANOVA model. While the main effect of ‘respiratory phase’ on startle responses remained (F3,120 = 7.54; p < 0.001; η2 = 0.16), there was no significant interaction of the factor ‘sex’ with any other variable. The main effect also remained significant when the ANOVA was re-calculated with raw startle eye blink responses (F3,123 = 4.03; p = .01; η2 = 0.18; see electronic supplementary material, S2).
(b) Control analyses of cardiac cycle phases
Across the four respiratory phases and both breathing conditions, the number of startles presented in the early and in the late cardiac cycle phase did not differ (main effects and interaction effect: all ps > 0.10; peak expiration: 3.6 early/6.4 late [0.18]; on-going inspiration: 4.0/6.0 [0.20]; peak inspiration: 4.1/5.8 [0.19] and on-going expiration: 3.9/6.1 [0.18]). There was a main effect, however, of ‘respiratory phase’ on the average latency to the preceding R-wave (F3,123 = 5.71; p = 0.002; η2 = 0.12). The post hoc analysis showed that this effect was due to a significantly shorter latency at ‘peak inspiration’ (374.0 [9.1] ms) compared with all other respiratory phases (peak expiration: 418.7 [12.5]; on-going inspiration: 413.0 [14.6] and on-going expiration: 402.6 [12.1] ms; all ps < 0.05). When introducing ‘cardiac cycle’ (early/late) as additional factor in the ANOVA to evaluate the impact on startle responses, we again found the main effect of the ‘respiratory phase’ on startle (F3,123 = 4.09; p = 0.01; η2 = 0.09). Neither a main effect of the ‘cardiac cycle phase’ on startle nor an interaction of the cardiac cycle phase with any other factor was observed (all ps > 0.05).
(c) Respiratory parameters
Adjusted tidal volume was higher in paced breathing (421.3 [29.9] ml) than during spontaneous breathing (326.9 [21.1] ml; F1,41 = 15.87; p < 0.001; η2 = 0.28). Breathing frequency did not differ between the two breathing conditions (spontaneous: 0.261 [0.003] Hz; paced: 0.259 [0.002] Hz; F1,41 = 1.37; p = 0.25; table 1). Additional analyses, using tidal volume as a covariate in the statistical model on startle eye blink, are reported in electronic supplementary material, S3.
The aim of this study was to investigate whether respiration affects human startle responsiveness. Indeed, higher startle responses were observed during on-going expiration in comparison with all other time points investigated. This effect was independent of spontaneous or paced breathing conditions, although higher adjusted tidal volumes were found during paced compared with spontaneous breathing.
These findings clearly demonstrate that respiration modulates startle responses elicited by acoustic noise bursts. This effect was independent of the breathing condition, e.g. whether breathing instructions were applied, or not. During spontaneous breathing, the generation of breathing rhythms is mostly automatically operated by brainstem centres in the pons, rostral and ventral medulla (e.g. pre-Bötzinger and Bötzinger complex) [52,53], whereas in paced breathing, higher cortical structures (e.g. supplementary motor area, cortico-subcortical network) are taking control over these brainstem centres [54,55]. The observed independence of the respiratory startle modulation from breathing conditions suggests, therefore, that neither automatic nor consciously controlled efferent motor code generation are a prerequisite of the observed effect. Two possible processes may contribute to the effect of respiration on startle responsiveness: (i) afferent neural input from slow adapting pulmonary stretch receptors and (ii) an indirect effect of respiration on cardiovascular haemodynamics and respiratory sinus arrhythmia.
In analogy to the CMS (which is associated with afferent input from arterial baroreceptors ), respiratory modulation of startle may also be associated with afferent input from a specific receptor type. One candidate is the slow adapting pulmonary stretch receptor (SAR), which responds to transient changes in lung volume and to maintained inflation , and is mainly located in the trachea and main stem bronchi . SARs may be further subdivided into low-threshold or tonic SARs, which respond to the absolute lung volume and are thus active at inspiration and expiration, and high-threshold or phasic SARs, which are sensitive to changes only during inspiration [56,58,59]. The stimulation of phasic SARs starts with inspiration and is maintained until peak inspiration is reached. The discharge of tonic SARs is linearly related to the total lung volume, which implies comparable stimulation during on-going inspiration and on-going expiration, highest stimulation at peak inspiration and lowest stimulation at peak expiration [58,59]. Because we observed higher startle response magnitudes during on-going expiration when compared with all other time points, the explanation for this result may lie in a process that is specific for the midpoint during expiration. On the one hand, it could be argued that the relative unloading of phasic SARs, which is uniquely observed during on-going expiration, is responsible for the observed effects. On the other hand, the interaction of tonic and phasic SARs may account for the respiratory modulation of startle. In particular, the stimulation of tonic SARs may only enhance startle modulation if there is no additional discharge of phasic SARs (i.e. on-going expiration), in contrast to time points, during which a discharge of phasic SARs can be observed (i.e. on-going inspiration, peak inspiration).
Afferent neural traffic from SARs is transmitted over the vagus nerve to the nucleus tractus solitarius (NTS) that connects with the nucleus ambiguus and nucleus paraambiguus to control motor output of respiratory muscles and airway smooth muscles . The NTS projects onto the parabrachial nucleus and nucleus coeruleus, from where higher limbic structures are reached . At a cortical level, the sensory cortex and the association cortex are involved in the processing of afferent signals from the respiratory system [20,33]. The eye blink startle response is a defence reflex that can be observed when an organism is confronted with an intense stimulus (e.g. loud noise) and whose amplitude is modulated by a number of psychological (e.g. emotion , attention ) and physiological factors (e.g. neural  and endocrine signal input ). These factors affect neural transmission of the primary acoustic startle circuit at the nucleus reticularis pontis caudalis (NRPC) mediated over limbic and tegmental brain structures . In the earlier established CMS effect, we proposed that visceral-afferent information from the NTS may be transmitted to the NRPC . Because afferent information originating from the cardiovascular and respiratory system share neural structures, it is plausible that this possible connection between the NTS and the NRPC may also be involved in the respiratory modulation of startle. We conclude that startle methodology is useful to examine afferent neural signal transmission from the respiratory system.
Respiratory modulation of startle may also be explained by respiratory effects on the cardiovascular system. During on-going inspiration, the parasympathetic output via the vagal nerve is reduced in order to increase pulmonary blood circulation and gas exchange. Vice versa, vagal output during expiration is increased. Because these oscillations in vagal output induce an increase of heart rate during inspiration and a decrease during expiration, this mechanism is called ‘respiratory sinus arrhythmia’ . The increase of heart rate during inspiration is then followed by an increase in systolic and diastolic blood pressure [51,66,67]. As there is a time delay of some seconds between the respective maxima in heart rate and blood pressure, breathing frequency largely determines the precise temporal location of the blood pressure maximum within the respiratory cycle. Interestingly, as indicated by the results of our pilot data, at 0.25 Hz breathing there is no difference in blood pressure between on-going inspiration and expiration. This finding suggests that differences in blood pressure and possibly arterial baroreceptor stimulation do not play a role for the higher startle responses during on-going expiration.
The question remains if startle stimuli were not distributed equally across the cardiac cycles in the different respiratory phases and breathing conditions. On the one hand, the results show average latencies to preceding R-waves to be lower at maximal inspiration when compared with all other conditions, which may be explained with the observed higher heart rates during inspiration. On the other hand, there was no difference between the most relevant condition ‘on-going expiration’ and the other conditions. Interestingly, and in contrast to the shorter latency at peak inspiration, the frequency of startle presentations between the early and late cardiac cycle phase did not differ between respiratory phases and breathing conditions. As opposed to earlier observations [8–10,12,13], our separate analysis on startle responses with ‘cardiac cycle phase’ as an additional factor indicated no effect of the cardiac cycle on startle responses, which may be due to the fact that the number of valid startle responses in the early versus late cardiac cycle varied between 1 : 9 and 9 : 1. Importantly, there was no interaction of cardiac cycle phase with respiratory phase or breathing condition. This suggests that different ratios of duration between the early and late cardiac cycle phases across the four respiratory cycles did not systematically affect startle responses. CMS is, therefore, unlikely to have contributed to the observed respiratory modulation of startle effect, as there was a comparable number of startle stimuli in the early versus late cardiac cycle phase and the observed main effect in latencies to the preceding R-wave was only owing to shorter latencies at ‘peak inspiration’, but not to ‘on-going expiration’.
The precision of detecting specific phases within the respiratory cycle may represent a crucial factor for the validity of findings. Given the current lack of mathematic algorithms, which sensitively and specifically detect respiratory phases, the majority of previous studies has followed one of two strategies: (i) the authors present stimuli without any contingency to respiratory phases, assuming an equal distribution, and assign trials later to respiratory phases in offline analyses depending on predefined criteria [39,40] or (ii) an experimenter continuously monitors respiratory activity and manually elicits the presentation of stimuli depending on current respiratory phase . For future research, the present algorithm may help to increase the precision in the temporal location of stimulus presentation depending on the respiratory cycle.
Owing to the non-invasive character of this study, we did not experimentally manipulate SARs to determine if they play a role in the observed effects. Further investigations of humans with altered SAR function (e.g. in pulmonary disorders) or their experimental manipulation in animals may be required to reveal if they represent a critical factor for the respiratory modulation of startle. The possible involvement of brain structures in the cardiac and respiratory modulation of startle is based on theoretical considerations and may require the additional support of systematic lesion studies in animals.
In the main study, we included only one paced breathing frequency of 0.25 Hz, based on the findings of comparable blood pressure during on-going inspiration and on-going expiration. Moreover, the mean breathing frequency in the spontaneous breathing condition did not significantly differ from 0.25 Hz. These findings, therefore, may be limited to this particular frequency. Future studies are suggested to include multiple breathing frequencies. The concurrent monitoring of EMG, ECG and blood pressure in the main study would have been desirable. We included blood pressure in a pilot sample and decided against using beat-to-beat blood pressure assessment in the main study as this technique implies continuous somatosensory feedback, which may act as pre-pulses for the subsequent processing of startle stimuli. Although the demographics of the pilot sample were almost identical to those in the main study, conclusions on blood pressure responses in the main study remain speculative.
The reliability of the control analysis on cardiac and respiratory modulation of startle may be limited owing to the low number of valid startle responses in some cells.
Finally, the estimation of tidal volume was based on data assessed using a respiratory belt and calibrated with standardized air inhalation of 1, 2 and 3 l. Despite the fact that (i) this procedure was repeatedly described as providing acceptable results for volume estimation [68,69], and (ii) the data as provided by the breathing belt show a strong positive correlation with tidal volume, it needs to be acknowledged that this technique may be less precise when compared with spirometric measures.
The effect of respiratory modulation of startle suggests that startle method is potentially useful for the investigation of afferent neural signal transmission from the respiratory system. The independence of the respiratory startle modulation from the breathing conditions suggests that neither automatic nor consciously controlled efferent motor code generation are a prerequisite of the observed effect.
Study procedures were approved by the local ethics committee at the University of Trier.
Data sheet including startle response magnitudes, breathing frequency, adjusted tidal volume and all control analyses depending on cardiac cycle phases is available as electronic supplementary material, S1.
A.S. and H.S. conceived the study idea and designed the study. A.S. and T.M.S. performed data collection. A.S., C.V., M.F.L. and H.S. analysed the data. A.S., T.M.S., C.V., M.F.L. and H.S. authored the manuscript.
The authors declare that they have no potential conflict of interest.
This project was supported by the Research Funds of the University of Trier (PI: André Schulz). The Research Office of the University of Trier had no further role in the study design, in the collection, analysis and interpretation of data, in the writing of the report, and in the decision to submit the paper for publication.
We thank Christoph Kinzig for the programming of the DASYLab-based algorithm, Dr Immo Curio for employing the algorithm in our laboratory devices, Alexandra H. Gräbener and Maria R. Barthmes for their help in data collection, Miriam-Linnea Hale for her help in data analysis, as well as Dr Paul Davenport for his literature suggestions.
One contribution of 16 to a theme issue ‘Interoception beyond homeostasis: affect, cognition and mental health’.
Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9.figshare.c.3464874.
- Accepted June 21, 2016.
- © 2016 The Author(s)
Published by the Royal Society. All rights reserved.