Detection Thresholds for Combined Infrasound and Audio-Frequency Stimuli

This study investigated whether the presence of audio sound (20Hz < frequency f < 20kHz) inﬂuences the detection threshold for infrasound ( f < 20Hz), and, vice versa, whether the presence of infrasound inﬂuences the detection threshold for audio sound. Monaural detection thresholds of thirteen otologically normal listeners were repeatedly determined for infrasound stimuli (sinusoids at 5Hz and at 12Hz) and for audio sound stimuli (sinusoids and bandlimited pink noise), separately and in presence of the respective other sound type. The measurements were performed with an adaptive 1-up-2-down 3-alternative forced-choice (3-AFC) procedure. Threshold levels for infrasound stimuli were not a ﬀ ected by audio sound at + 5dB sensation level (SL), but they were signiﬁcantly increased by the presence of some of the audio sound stimuli presented at + 50dBSL. For example, thresholds for the detection of infrasound increased on average by around 5dB when simultaneously presented with a pink-noise stimulus (frequency range: 250Hz–4000Hz). On the other hand, the presence of infrasound with levels up to + 10dBSL did not cause any signiﬁcant change in the detection thresholds for audio sound. This could be an indication that infrasound might even be more annoying in a quiet environment.


Introduction
It is noww ell established that humans are able to perceive infrasound (abbr. IS, frequency f < 20 Hz)a tl east down to 2Hz( see, e.g. the reviewp aper by Møller and Pedersen [1]). However, the sensation of IS differs from that of sound in the common audio-frequencyr ange (audio sound, abbr. AS, 20 Hz < f < 20 kHz)i naw ay that the sensation has arather discontinuous character accompanied by afeeling of pressure instead of atonal sensation [1]. Moreover, there is as teep increase of human detection thresholds for frequencies below100 Hz so that high sound pressure levels are needed for humans to detect IS [1,2]. In addition, the distances between equal-loudnesscurves [1,3,4] and equal-annoyance curves [5,6] are smaller for IS than for AS stimuli. Therefore, as mall increase in SPL(sound pressure level) for IS can result in a significant increase in perceivedloudness and annoyance. Furthermore, there are no auditory filters tuned to infrasonic frequencies, since the lowest centre frequencyofan auditory filter lies between 40 Hz or 50 Hz as estimated by psychoacoustic tuning curves [7].
The decrease in sensitivity towards infrasonic frequencies is caused by different processes acting as ahigh-pass filter for the sound transferred inside the human ear.This includes the middle-ear attenuation (decrease in SPLb y 6dB/octave for frequencies belowabout 1000 Hz [8])and the shunt mechanism of the helicotrema (decrease in SPL by 6dB/octave for frequencies below4 0Hz [ 9]). In addition, the sensitivity of the inner hair cells decreases by 6dB/octave,s ince theya re mechanically excited dependant on the velocity of the basilar membrane [10]. This decreased sensitivity of the inner hair cells is assumed to be to some extent compensated by the outer hair cells, since the outer hair cells, which contact the tectorial membrane, are sensitive to the large basilar membrane displacements caused by IS [11].
At manyimmission sites, humans are exposed to noise with frequencyc omponents both in the AS and in the IS frequencyr ange. Therefore, the question arises whether interactions between IS and AS may influence the quality of the auditory perception.
An important potential interaction of IS and AS is the effective amplitude modulation of AS caused by IS, since the hearing system of humans is especially sensitive to amplitude modulation (AM) at frequencies between 2Hz and 5Hz [ 12,13]. When IS and AS are simultaneously present, effective amplitude modulation may be generated on ap hysiological level: Experiments with guinea pigs had demonstrated that IS can modulate AS by cyclically changing the cochlear amplifier gain [14]. This is in line Figure 1. Schematic viewofthe setup for the detection threshold measurements for IS and AS stimuli using the insert earphone sound source system [19].
with the results of psychoacoustic experiments showing that human listeners can hardly distinguish between AS combined with IS and AS that is amplitude modulated at infrasonic frequencies [15].
Another potential interaction of IS and AS may be masking effects, meaning that threshold levels for IS or for AS stimuli are increased when simultaneously presented with the other sound type as masker stimulus. It is well known for stimuli in the audio-frequencyr ange that threshold levels increase when amasking stimulus is presented in an adjacent frequencyr egion, and that there is as pread of masking towards higher frequencyr egions with increasing masker level [12]. These effects are easily modelled in terms of critical bands in the audio-frequency range (see, e.g. [16]). However, it is reasonable to assume that perception mechanisms for IS differ from those for AS. Therefore, potential masking effects for IS stimuli as masker or target stimuli may also not be the same as those for AS stimuli, and theya re unlikely to be modelled in as imilar waya sk nown from the audio-frequencyr ange. The study of Finck [17] revealed abroad masking effect of high-levelIS(10 Hz at 130 dB SPL)onthresholds for AS tones up to 4800 Hz. Apilot study conducted prior to the actual study presented in this paper indicated small or no effects of IS presented up to +10 dB SL (sensation level) on the threshold levels for AS stimuli, butac onsiderable increase of the individual threshold levelfor the IS stimuli caused by the presence of AS background stimuli [18].
The purpose of this study wast od eliveram ore profound investigation of masking effects for IS combined with AS. In particular,i tw as investigated whether the presence of IS changes detection thresholds for AS and, vice versa, whether the presence of AS changes detection thresholds for IS. One hypothesis was, that the modulation effect caused by IS can reduce the threshold levelfor AS. On the other hand, the presence of IS may cause an increase of the detection thresholds, similar to masking effects. Downward masking effects caused by AS on thresholds in the infrasonic frequencyrange were, however, assumed to be unlikely,unless the AS stimuli are presented within the lowfrequencyrange.

Measurement setup
Monaural detection threshold measurements were performed with as pecially developed insert earphone sound source system (see Figure 1) which is described and characterized in more detail in [19]. The IS and AS stimuli were generated by separate electrodynamic loudspeakers, called the IS source and the AS source, mounted inside twod i ff erent housings. Sound tubes (two polyethylene tubes: 1.5 ma nd 7.5 ml ength, 14 mm inner diameter; one silicone tube: 0.5 ml ength and 3m mi nner diameter)a nd at -piece coupled the sound sources to an audiometric eartip (E-A-RTone/E-A-RLink Standard Insert Foam Eartip)a si llustrated in Figure 1. The audiometric eartip wasinserted into the participant'sright ear canal for monaural presentation of the acoustic stimuli. The con-tralateral ear of the participant waso ccluded with an ear plug. An additional loudspeaker,known as the compensation source, wasmounted belowthe AS source. It delivers IS to the back of the AS loudspeaker membrane to compensate the displacements of the membrane caused by the IS down to an imperceptible level [19].
The waveforms of the IS and AS signals were generated with aM AT LAB-based software framework AFC [20] at 96 kHz sample rate. An external sound card (RME Fireface UC)g enerated analogue output signals (ISa nd AS components in separate channels)t hat were fed to three separate amplifiers (BAA 120 BEAK for AS signals, BAA1 20 TIRA for IS signals). Protective switches were inserted in each signal path to avoid that excessively loud signals could be presented in case of equipment malfunction. The sound sources, the sound tubes, and the participant were located in an anechoic room providing as ufficiently lowb ackground noise levele veni nt he infrasonic range (see Figure 2f or background noise levels) during the listening tests. Ac omputer display and ak eyboard, which were required for the experimental procedure (see Section 2.5 Psychoacousticalmeasurement procedure), were placed in front of the participants. The computer controlling the experiments wasl ocated outside the anechoic room.
The calibration of the AS stimuli wasp erformed in an IEC 60318-4 [22] occluded-ear simulator (Brüel &K jaer 4157, with ear canal extension DB 2012). The IS stimuli were calibrated with a 1 2 low-frequencypressure-field microphone (Brüel &Kjaer 4193, with UC0211)that was placed in acavity having av olume of 1.3 cm 3 ,equivalent to that of the average human ear canal. During calibration, the sound tube from the sound source system wasc onnected to either the cavity or to the ear canal extension, respectively,bymeans of the eartip.

Participants
The detection threshold measurements were performed with fifteen participants (seven females, eight males, between 18 and 30 years old). Twoparticipants were not able to participate in all measurement sessions of this study so that the threshold levels were evaluated for the remaining thirteen participants. None of them had experience in psychoacoustic measurements with IS prior to this study.All participants were otologically normal as confirmed by a questionnaire for hearing testing (AnnexAi nI SO 389-9 [23])and by otoscopic examination. Hearing thresholds were better than 15 dB HL between 125 Hz and 8000 Hz in the right ear as tested by standard pure tone audiometry according to ISO 8253-1 [24], with astep size of 1dB.
The Declaration of Helsinki wasa dhered to in all our measurements and apositive vote of the local ethics committee (PTB ethics application 3/16)was given.

Stimuli
Detection threshold levels were repeatedly determined for IS stimuli and for AS stimuli, both separately and during the presence of the respective other sound type. IS stimuli were sinusoids at 5Hza nd at 12 Hz. The aim wast o investigate whether the effects on the detection threshold measurements are similar for IS stimuli at different frequencies, which are well belowthe lower frequencylimit for AS (between 16 Hz and 20 Hz), and which have adistance of more than one octave to each other,t aking into account that the sound source system is applicable for stimulation with IS down to 4Hz [19]. Three different AS stimuli were applied in the detection threshold measurements: as inusoid in the low-frequencyr ange at 100 Hz, an additional sinusoid at ah igher frequencyo f1 000 Hz, and ab roadband pink-noise stimulus with the frequency range between 250 Hz and 4000 Hz. These stimuli were selected to investigate whether the bandwidth (broadband pink-noise centred at 1000 Hz vs. sinusoid at 1000 Hz)o r the distance between AS and IS stimulus frequencies (sinusoid at 100 Hz vs. sinusoid at 1000 Hz)may lead to different results in the detection threshold measurements.
Stimulus onsets and offsets were time-windowed using acos 2 function providing ramp durations of 200 ms for AS stimuli, 250 ms for sinusoids at 12 Hz, and 600 ms for sinusoids at 5Hz (Table I).Because the frequencyresponse of the AS sound source wasnot sufficiently flat in the frequencyr egion covered by the pink-noise stimuli, the latter were digitally pre-shaped in order to equalise the frequencyr esponse of the AS source. The acoustical spectrum of the pink-noise stimuli wasmonitored with asignal analyser (Norsonic Real Time Analyser 840)a nd turned out to be flat within 3dB. Anew sample of pink-noise was generated prior to each stimulus presentation, in order to minimise apotential effect of random signal peaks on the detection threshold measurements (i.e., "running noise").
In the following, the term target stimulus (TS) defines the stimulus for which the threshold levelw as measured. The term background stimulus refers to the stimulus which wasp resented in addition to the target stimulus in some measurements. The duration of the target stimuli and, thus, thed uration of the intervals within the 3-AFC measurement procedure (see Section 2.5 Psychoacousticalm easurement procedure)w as set to 1000 ms, except for the sinusoid at 5Hz, the duration of which wasset to 2000 ms (see Table I).Inorder to present the background stimulus at its full amplitude during the presentation of the three intervals in each measurement trial, the background stimulus started 200 ms-600 ms (corresponding to the ramp duration, Table I) prior to the presentation of the first interval and ended 200 ms-600 ms after the presentation of the third interval.

Experimental paradigm
The measurement sessions were divided into three experiments. Each experiment wasd ivided in several runs performed in random order with different combinations of target stimulus TS and background stimulus BS (see Table II). Within one measurement session only one IS stimulus, either asinusoid at 5Hzorat12Hz, wasapplied. All sessions started with Experiment 1( see Table II). In this experiment detection thresholds were measured for isolated target stimuli (without background stimulus). The threshold measurement for the IS target stimulus wasa lways performed twice within one session. The intention of the first measurement wast ot rain the participants to correctly identify the IS target stimulus. The twof ollowing experiments, Experiment 2a nd Experiment 3, were performed in random order.T hese experiments comprised threshold levelmeasurements for the same target stimuli as in Experiment 1. In addition to the target stimulus, ab ackground stimulus wasp resented at as pecifics ound pressure level( see Table II, column 3). The sound pressure levels of the background stimuli were adjusted with reference to the individual threshold level for this stimulus (i.e. sensation level, dB SL)d etermined in Experiment 1.
In Experiment 2, threshold levels were determined for IS in the presence of AS (see Table II). The sound pressure levels of the AS background stimulus were set to +5dBSL and +50 dB SL, in order to compare the effect of the stimuli presented at one sound pressure that wasj ust audible and at one sound pressure levelthat wasclearly perceptible, although not too loud.
In Experiment 3, threshold measurements were performed for AS in the presence of IS (see Table II). The IS background stimulus wasp resented at twod i ff erent levels, one below( − 10 dB SL)a nd one above the individual threshold level. As reference for the individual threshold levelfor the IS background stimulus, the second measurement determined for the isolated IS stimulus within Experiment 1was selected. Average detection thresholds for sinusoids have previously been reported as 110 dB SPL at 5Hza nd 90 -93dBSPL at 12.5 Hz [1,2]. Since the dynamic range of the human auditory system between threshold and uncomfortably loud is steeply decreasing with decreasing infrasonic frequency [1,3], the levelabove the individual threshold levelwas set to +10 dB SL for the background stimulus at 12 Hz and it wass et to +5dBSLf or the background stimulus at 5Hz. The intention of this was to apply above-threshold sounds that were sufficiently far away from being perceivedasdiscomfortable. In addition, it waso fp articular interest to investigate the influence of IS presented at al evel belowt he individual threshold on the perception of AS, because levels of IS measured in environmental noise usually are also well belowt he perception threshold, e.g. [25].
Prior to the beginning of the first threshold measurement the participants receivedwritten and oral instructions. After each experiment, the participants were asked in afree interviewtodescribe subjective details of their perception of the stimuli and to report if anyabnormalities or discomfort occurred during the experiment. Foreach participant, anythreshold measurement session covering the set of the three experiments wasperformed within one day.Intotal, one session had ad uration of around 1.5 to 2hours, including breaks. The threshold measurement sessions were repeated twice for each participant and for each IS stimulus on separate days to estimate the reliability of the measurements and therefore gather robust detection threshold data.

Psychoacoustical measurement procedure
The detection thresholds were determined using a3alternative forced-choice (3-AFC)p rocedure. The sound pressure levelo ft he target stimulus wasv aried in accordance with the adaptive 1-up-2-down rule that converges at the 71%-point of the psychometric function [26]. The participants receivedf eedback on the display,w hether their response had been correct or wrong to help them identifying the IS stimuli correctly.
Fore nsuring participants' safety,am aximum sound pressure levelw as implemented digitally in the experimental procedure for all stimuli (limit values, see Table I). These limits have been approvedbythe positive vote of the local ethics committee. The experiment would have been terminated if the stimulus had reached its predefined limit more than four times during one run. However, this has neveroccurred during this study.
At the beginning of each run the levels of the target stimuli were set to avalue that is expected to be easily audible on average (start levels, see Table I).The initial step size of 6dBw as used for AS stimuli and as tep size of 4dB wasu sed for the IS stimulus. The reason for this difference in step size is the much smaller dynamic range of the human auditory system for IS compared to AS [1,3]. The step sizes were reduced to 3dBafter the first upper reversal and to 2dBafter the second upper reversal. Then, the measurement phase began, and it ended upon the completion of the eight following reversals. The detection threshold levelwas then calculated as the median value of all levels at the reversals during the measurement phase.

Statistical analysis
Individual threshold levels L indiv (TS,BS)were measured at least three times for each participant and for each target stimulus TS with background stimulus BS or with no background stimulus (BS = 0) in separate sessions. Individual threshold shifts Δ indiv (TS,BS)w ere calculated for each participant by subtracting the individual threshold levelf or the isolated target stimulus TS, i.e. L indiv (TS,BS = 0),from the individual threshold for the same target stimulus TS in the presence of aspecificbackground stimulus BS, i.e. L indiv (TS,BS = 0) both measured within the same session. Thus, ap ositive threshold shift for at arget stimulus TS, Δ indiv (TS,BS) > 0, indicates that the threshold levelf or this target stimulus TS wasi ncreased when the background stimulus BS was present. Average threshold shiftsΔ indiv were calculated as the arithmetic mean of the threshold shifts across the three repeated measurement sessions for each participant and each stimulus condition.
Three factorial analysis of variance (ANOVA )f or repeated measures were performed, for IS and for AS threshold levels separately.T he factors were (1) session (three repeated measurements for each combination of target stimulus and background stimulus), (2) targets timulus (ANOVA for IS: sinusoid at 5Hza nd at 12 Hz; ANOVA for AS: broadband stimulus, sinusoid at 100 Hz and sinusoid at 1000 Hz)a nd (3) background stimulus (different background stimuli each presented at as pecificl evel including the measurement without background stimulus, i.e. BS = 0).T he intention of the ANOVA wast o investigate the statistical significance (significance level α ANOV A = 0.05)o fthe effect of all factors on the threshold levels, and the interaction between the factors. Posthoc t-tests were performed for all factors.

Three factorial analysis of variance forISand AS thresholds
In general, all participants were able to perceive the stimuli applied in this study and did not report anydiscomfort or abnormalities in the context of this study. The ANOVA for AS threshold levels revealed asignificant effect of targets timulus (p<0 . 001), background stimulus (p<0 . 001)a nd session (p = 0.034), butn o significant interaction between factors. The repeated measures ANOVA for IS showed that there were significant effects of the targetstimulus (p<0.001), background stimulus (p<0.001)and session (p = 0.038)onthe threshold levels for IS. In addition, asignificant interaction between targetstimulus and background stimulus (p<0 . 001)w as found for the IS threshold levels.
Post-hoc t-tests for the factor session using the Bonferroni correction (α session = 0.016)d id not reveal as ignificant difference for AS thresholds. On the other hand, for IS thresholds, there is asignificant difference (p = 0.015) between session 1and session 2with an average decrease of 1.0 ± 0.3dB. However, this difference is even smaller than the minimum step size of 2dBwithin the 3-AFC procedure. It can therefore be concluded that the factor session has, if at all, anegligible effect on the threshold levels. Therefore, average threshold levels were evaluated as arithmetic means across the repeated measurements. The influence of the factors targets timulus and background stimulus on the threshold levels were examined further in selected post-hoc t-tests (see Section 3.3).  . Individual detection threshold levels L indiv (TS,BS = 0) of 13 otologically normal participants for the isolated sinusoid at 5Hzand at 12 Hz generated by the insert earphone sound system. The threshold levels were determined three times for each subject on separate days. Each marker indicates the threshold levels for one participant. The three threshold levels of each participant were arranged on the horizontal axis in ascending order of their arithmetic mean.

Detection threshold levels fori solated IS and AS stimuli
In Figure 3t he results of the threshold levels of each of the thirteen participants determined for an isolated sinusoid at 5Hza nd at 12 Hz are shown. Each marker indicates the threshold levels for one participant. The thresholds were determined three times for each participant on separate days. Average thresholds across all participants and associated standard deviations of the single value were 109 dB SPL± 6dBfor the sinusoid at 5Hzand 92 dB SPL ± 5dBf or the sinusoid at 12 Hz. The standard deviations of the individual threshold levels were on average 2.4 dB for the threshold levels at 5Hza nd 2.1 dB for the threshold levels at 12 Hz. Mean threshold levels for AS stimuli, in total measured six times for each participant, were 29 dB SPL± 3dBf or the sinusoid at 100 Hz, 5dBSPL ± 5dBf or the sinusoid at 1000 Hz and 13 dB SPL ± 3dBfor the broadband stimulus. The standard deviations of the individual thresholds were on average 1.9 dB for 1000 Hz, 3.6 dB for 100 Hz, and 2.0 dB for the broadband stimulus.
Asignificant correlation wasfound between the individual thresholds for the IS stimulus at 5Hza nd the respective thresholds at 12 Hz (Spearman correlation coefficient r 5Hz,12Hz = 0.79, p 5Hz,12Hz = 0.001). Comparing the threshold levels for the isolated AS stimuli, ac orrelation waso bserved between the threshold levels of the broad-  band stimulus and the 100 Hz stimulus (r Broadband,100Hz = 0.78, p Broadband,100Hz = 0.002). However, no correlation wasfound between the threshold levels for the isolated IS and the isolated AS stimuli.

Detection threshold shifts forI Sand AS target stimuli
Figures 4a nd 5s howt he boxplots of the average threshold shiftsΔ indiv of the thirteen participants for IS and AS  target stimuli. The whiskers extend to the most extreme data points that are located inside 1.5 times the interquartile range above the upper quartile and belowt he lower quartile. Data points located outside these limits are indicated by aplus sign.
The significance of the threshold shifts wast ested by means of paired t-tests for each stimulus combination (see Table III). This parametric test waschosen since the paired samples were normally distributed according to Shapiro-Wilk tests (α Shapiro−Wilk = 0.05), except one paired sample (No. 17, Table III, p Shapiro−Wilk = 0.003 < 0.050). The null hypothesis (H 0 )f or each t-test wast hat there is no difference between the paired samples of threshold levels for the target stimulus TS in the presence of aspecific background stimulus in comparison to the thresholds for the same stimulus TS with no background sound. Multiple testing wascompensated for by Bonferroni correction (α t−test = 0.0042). Significant threshold shifts are indicated by asterisks in Figures 4and 5and in Table III. 3.3.1. Threshold shifts for AS target stimuli caused by IS background stimulus Figure 4illustrates that the average threshold shift for AS caused by the presence of IS ranges from −5dBto+6dB. The t-tests indicated that the threshold shifts for AS target stimuli were not significant for all conditions. However, acloser investigation of individual results reveals that two participants had threshold shifts around +5dBfor all three AS target stimuli in the presence of the background stimulus 12 Hz presented at +10 dB SL.

Threshold shifts for IS target stimuli caused by
AS background stimulus The threshold shift for the IS stimulus caused by the presence of AS at the lower intensity of +5dBSLranges from −3t o+ 8dB( see Figure 5).T he threshold shifts for IS stimuli caused by the presence of AS at +50 dB SL showed that some participants were hardly affected in their detection of IS by the presence of AS, whereas others showed large shifts for IS thresholds amounting up to +17 dB in the case of the threshold for 12 Hz caused by the presence of the background stimulus 100 Hz at +50 dB SL. On average, the biggest effects across the sample of more than +3.5 dB were observed for the combination of the sinusoid at 5Hzand at 12 Hz with the AS background stimuli at 100 Hz and the broadband stimulus, both presented at +50 dB SL.
The results of the t-tests showed that the threshold shifts for IS target stimuli were not significant for AS background stimuli presented at the lower intensity of +5dBSL. However, when the AS background stimuli were presented at the higher intensity of +50 dB SL, the threshold shift for the sinusoid at 12 Hz wass tatistically significant for all AS background stimuli and the threshold shift for the sinusoid at 5Hzwas statistically significant in the case of the broadband background stimulus.
It should be mentioned that one participant had ac onsiderably smaller shift than all other participants for the threshold for 5HzI Si np resence of 100 Hz AS presented at +50 dB SL, withΔ indiv (5 Hz,100 Hz presented at +50 dB SL) = −6dB. Treating this as one outlier and excluding the data for this participant from the group analysis, the remaining data for the threshold shift at 5Hzinthe presence of 100 Hz at +50 dB SL is reaching significance (p = 0.001 < 0.004). The average shift across the sample for this stimulus combination would then increase to +4.5 dB.

Threshold shifts fori nfrasound and audio sound
This study found that detection thresholds for IS were significantly increased when some of the AS stimuli were presented at asufficient level. On the other hand, this study did not reveal asignificant effect of infrasound on the detection thresholds for audible sound, in contrast to the initial hypothesis. The reason for the observed effect of AS on the threshold levels for IS is not yet clear.For energetic masking, it would be very unusual that am asker more than four octavesabove the test tone (5 Hz and 12 Hz to 250 Hz for the broadband noise, and 12 Hz to 1000 Hz)c ould have any effect on the test tone, because for audio frequencies, noticeable downward masking requires frequencyspacing of less than twooctaves [12]. Only in the case of the 100 Hz stimulus, an energetic masking effect would be ap ossible reason for the threshold increase. Furthermore, there is no characteristic place anymore on the basilar membrane for 12 Hz and for 5Hzt ones [7]. Therefore, the observed masking is very unlikely to happen in as imilar waya s known from audible sounds in different frequencybands.
The study by Salt et al. [14] showed that infrasonic responses from guinea pigs recorded as endolymphatic potentials were suppressed by AS with increasing sound pressure levelo fA S. This indicates that there might be other physiological effects causing the threshold increase for IS.
On the other hand, the threshold increase for IS in the presence of AS might rather be attributed to an effect of attention, that is, to am ore cognitive effect. When IS is presented simultaneously with AS the attention of the listeners may be shifted towards the perception of AS, which is indeed in accordance with statements from the inter-views conducted after each experiment. Some participants reported that, in general, it wasmore difficult for them to detect IS than AS and that this effect wase venmore pronounced when IS had to be detected in the presence of AS. In addition, some participants showed especially large threshold shifts for IS up to +17 dB, indicating that this effect strongly varies across individuals.
It has to be noted, that some internal physiological noise wasaudible for some participants in the anechoic environment used for the experiments. Sevenparticipants reported that the IS stimuli sounded liketheir ownaudible physiological noise, e.g. the heartbeat, or theyr eported that the interaction of IS and the physiological noise makes it even more difficult to detect IS. The presence of the physiological noise might affect the detection thresholds for isolated IS. This is another possible reason for the similarity of thresholds for isolated IS and for IS in the presence of AS at +5dBSL, because in both conditions there wass ome AS present at al ow level( either audible internal physiological noise or the presented AS background stimulus).

Accuracy of the measurements
Robust detection threshold data were gathered within this study,a st he measurements for all combinations of target stimulus and background stimulus investigated in this study were performed three times for each participant, and there wasn os ystematic effect of the repetitions. However,i ts hould be noted that the variance for the detection thresholds and, therefore, the accuracyofthe inferred threshold shifts is limited by the predefined minimum step size of 2dBw ithin the 3-AFC procedure employed here. Since there are no comparable data of threshold levels for combined IS and AS stimuli from other studies yet, the accuracycan only be estimated for the threshold levels for isolated IS, which serves as the basis for anyfurther calculations, likethe sensation levelofISbackground stimulus and the threshold shifts for IS. The thresholds for isolated IS reported here are consistent with the monaural insertearphone threshold levels from 18 participants determined by Kühler et al. [2], as well as with the second order binaural infrasound-threshold estimation by Møller and Pedersen [1]. Theyreported average detection thresholds around 110 dB SPL for asinusoid at 5Hzand 90 -9 3dBSPL for asinusoid at 12.5 Hz. In addition, the standard deviations reported in this study are similar to the frequencyi ndependent standard deviations around 5dBfor the frequency range between 2Hza nd 1000 Hz reported in [1] and the interquartile-range around 6dBand 9dBfor threshold levels at 5Hza nd at 12 Hz reported in [2]. This underlines the validity of our results for the thresholds for isolated IS stimuli.

Comparison between thresholds fori solated infrasound and audio sound
It wasi nvestigated whether there is al ink between individual threshold levels for IS and AS. The comparison of standard deviations of the thresholds for IS and AS showed that interindividual differences for IS threshold levels were on average bigger than those for AS, which is also consistent with the thresholds for sinusoids at frequencies up to 125 Hz obtained in [2]. Another important finding is that there wasnosignificant correlation between the detection thresholds for IS and AS stimuli, butt here wasas ignificant correlation between the thresholds for the sinusoid at 5Hza nd at 12 Hz. These findings suggest that there are certain individuals being especially sensitive to IS, independent of their hearing status in the AS range. One reason for this might be that the perception of AS and IS are based on different mechanisms. Møller and Pedersen [1] suggested that people with ap articularly narrowo rb locked helicotrema might be especially sensitive to IS, because the high IS pressure inside the cochlea is counterbalanced for them more slowly.

Statements from the interviews
The statements within the interviews give an impression howi solated IS and the combination of IS and AS are being perceived. The perception of IS is described likea pressure or tactile sensation, which is in line with many prior studies, e.g. [1,27]. Moreover, afew participants associated the perception of IS with amodulation effect and other participants suggested that theydetected the IS stimuli through ac hange of the hearing sensation of the AS background stimulus. Both are in line with the results from the study of Marquardt and Jurado [15] who found that IS combined with AS cannot be distinguished from AS that is amplitude modulated at infrasonic frequencies.

Limitations and futurer esearch
At this stage, this study waslimited to young listeners aged between 18 and 30 years with normal hearing. Therefore, it would be desirable to repeat this study with other groups of listeners, likeolder population, hearing-impaired listeners or individuals being particularly sensitive to IS, and to compare the results of these groups with the findings of this study. Based on our results, it can be concluded that modulation effects caused by infrasound [14,15] do not affect hearing thresholds for audible sound, since, contrary to the initial hypothesis, audio sound thresholds were not decreased by the presence of infrasound. If there is any masking effect of IS on the threshold levelfor AS, this effect must be very small. However, this does not imply that there is no influence at all of IS on the auditory perception of AS. It would therefore be desirable to investigate whether the interaction of IS and AS might possibly affect other psychoacoustic quantities likel oudness and/or annoyance. Modulation effects might affect the quality of the auditory perception when both, infrasound and audio sound are presented at al evel well above threshold. The statements within the interviews indicate that there is an audible modulation effect. Further studies should be intended to investigate whether this modulation effect might be ap ossible reason for the large annoyance related to noise with IS components.
Furthermore, this study found that threshold levels for infrasound were increased by the presence of audible sound presented at asufficient level. This could be an indication that infrasound might even be more annoying in a quiet environment.