Archaeoacoustics- a Perceptual Study

Archaeoacoustics: A Perceptual Study

Kaan Shenhuy

ABSTRACT: Existing studies show that there are strange acoustical effects at the prehistoric sites. The resonances found at European prehistoric sites are generally in the frequency range of 90 – 120 Hz. The effects of listening to tones on regional brain activity at this frequency range have been studied before. This study is an investigation into the perceptibility of the resonances in the aforementioned frequency range, using a technique called auralisation. Bass response of the auralisations of a prehistoric hypogeum (Hal Saflieni Hypogeum, Malta) and a modern cathedral (Saint-Etienne, France) is analysed using spectral audio analysis methods. An ABX listening test was designed and carried out to test the hypothesis. The quantitative results of the subjective listening tests showed that subjects could differentiate between the two audio files. Combined with the findings from spectral analysis of the audio files, the study is concluded that the resonances found in the auralisation of Hal Saflieni Hypogeum are perceptible, thus, perceptual studies about the effects and possible use of the prehistoric sites need to be urged in order to discover the forgotten cultures and knowledge. Recommendations for further work are stated.

KEYWORDS: Archaeoacoustics, resonance, auralisation, perception.

1.0 Introduction This paper offers the background information related to This paper presents an investigation into the acoustical phenomena found at the prehistoric world perceptibility of resonance phenomena found in and their effects on human perception and brain Neolithic architecture. Researchers have discovered activity. An acoustical comparison of a modern strange acoustic response at prehistoric sites like Hal- auditorium and a Neolithic hypogeum will be made in Saflieni Hypogeum in Malta and Stonehenge in the UK. order to validate the perceptibility of resonance The architecture of these sites plays a major role in the phenomena. Available methods are analysed and design reproduction of the acoustical phenomena. The sonic of the final comparison methodology is justified. The interference patterns caused by the interior results of conducted subjective listening tests and architecture yield a world of nodes and antinodes spectral analysis of the test audio files are discussed. throughout the building. These nodes and antinodes are the points in the building where the constructive and destructive interferences form respectively when a 2.0 Background sound at the right frequency band is being produced. Long before man made Neolithic structures, our These effects are said to be aurally rich and unique, ancestors sheltered in natural formations of rock and enabling the visitor to experience an aural sensation cave systems. Recent studies at European rock art sites designed by their ancestors. These buildings are show that 40,000 years ago modern humans gathered constructed during the Neolithic period when the in the caves of Europe (Douglas, 2014). The earliest agriculture was initiated around the globe. Some theories on the content and the pattern of the paintings researchers also discovered that various Palaeolithic on cave walls were often unsatisfactory (Waller, 1994) cave paintings appear on the rocks where echoes can be until a study by Iegor Reznikoff and Michel Dauvois has perceived as if they are coming from the inside of the shown that the locations of deep cave art in France rock walls. correspond to the areas where resonances of musical

1 notes are found (Reznikoff and Dauvois, 1988, cited in 120 and 130 Hz while their brain activity monitored by Waller, 1994, p.3). EEG. The central finding of their study was that listening to tones at 110 Hz, subjects’ brain activity differed from Till (2014) suggests that music archaeology research all other neighbouring frequencies. They explain this has started 40 years ago where their work was mostly difference as “This might be interpreted as a relative focused on the remains of prehistoric sound-making deactivation of language centres in the brain to allow tools. But it was the rise of digital technology where it other mental processes to become more prominent‘’ helped archaeology researchers to examine the aural (Cook et al., 2008, pp. 95–104). They have also found characteristics of the sites by being able to record out that areas associated with musical pitch processing acoustic fingerprints of the sites. and emotional processing to be active at 110 Hz tone.

The resonances are natural properties of caves where SB Research Group (SBRG) carried out a more recent some parts have great resonances while the other parts study that focused on the same frequency range with 5 have no resonances at all. Iegor Reznikoff (2014; 2008) Hz intervals where subjects listened to the tones and a explains that in most of the painted caves in France, mantra after each tone with the same frequency location of the most resonant areas and the location of (Debertolis et al., 2014b). The main difference of this the paintings coincide. Steven Waller has analysed the study was that all subjects had their own frequency of rock art sites on three continents (North America, activation. They also found out that “...those volunteers Europe and Australia) and has found out that virtually with a frontal lobe prevalence during the toning every site has significant acoustical characteristics at received ideas and thoughts similar to what happens rock art sites; over 60 locations spread over three during meditation, whilst those with an occipital lobe continents (Waller, 1994). prevalence during the toning visualized images’’ (Debertolis et al., 2014b). Later on, Neolithic structures like Stonehenge are studied for their acoustical significances (Till, 2014). Later on, on-site research by SBRG at Cividale del Friuli Watson and Keating (1999) - in the conclusion of their hypogeum was carried out. This time TRV (variable- study of 2 prehistoric sites in the UK - state that “While resonance camera) technology has been used to it cannot be demonstrated that the architecture of monitor micro-mobility of human body where monuments was deliberately configured to enhance dedicated software processes micro-movements and acoustic performance, the behaviour of sound would allows the analysis of emotional/mental balance have been an unavoidable factor in their use’’ (Watson (Debertolis et al., 2014a). The images processed by the and Keating, 1999). In a study made earlier Jahn et al. software showed that a subject exposed to 8 minutes of (1995) have analysed the acoustics of 6 ancient man mantra at the resonant frequency, became in tune with made structures around the England and Ireland. They the environment showing change of the state of have found out that each site sustain strong acoustical consciousness (Debertolis et al., 2014a). resonances at a frequency between 90 and 120 Hz (Jahn et al., 1995). The frequency range falls in the male vocal Following the studies made by SBRG (2014b) and Cook range where researchers propose that the one function et al. (2008), the effects of exposing a person to the of these Neolithic sanctuaries could be for supporting simple tones (90-120 Hz) are clear. Given the human chanting and/or drumming in sacred rituals background information about the neurophysiological (Cook et al., 2008). significance of acoustical resonances discovered at prehistoric sites, their subjective perceptibility to a Although there have been a number of studies about the virtual visitor is not studied before. Over a nearly 20 psychological and physiological effects of infra sounds years of archaeoacoustic research on the phenomena at (�! < 20 ��) (Johnson, 1980; Qibai and Shi, 2004; man made ancient sites, intentionality of their builders Debertolis et al., 2014a) such emphasis on the effects of to achieve those effects is the main question for some of low-frequency sound were not studied until 2008. On the academics in the field. As the ritual use of the sites the basis of Jahn et al.’s findings, a team under are certain (Cook et al., 2008) and if the perception of supervision of Ian A. Cook (2008) at UCLA (USA) the resonances is verified, there should be no question investigated the effect and correlation of those resonant about their presence in the multisensory experience of frequencies on brain activity by audiometric tests using the inhabitants during rituals. Thus, the number of sites electroencephalography (EEG) to monitor regional containing resonances at the aforementioned frequency brain activity. Subjects listened to tones at 90,100, 110, range should no longer be perceived as ‘coincidence’

2 but a form of fulfilling a constructional specification to content holds. Another helpful visualisation method is build a sanctuary. The evidence may implicate that a the spectrogram where the same relation is shown as sort of shamanistic culture or religion might be one of well as decay over a period of time. the factors of discovering similar resonances in the sanctuaries throughout Europe including Malta, Italy, The main experiment of this study is the subjective Bosnia, the UK and Ireland. listening test carried out at the Listening Room in Acoustics Research Centre (Salford). The method selected for this test is the ABX method. An instructions 3.0 Experimental Design sheet has been written with relevant information about the test theme without revealing the hypothesis and the 3.1 Strategy research question. A pilot experiment has been carried out to test the initial procedure. Pilot experiment showed some constraints with the time limit and the clarity of instructions sheet. As a result, duration of the individual sessions was extended from 20 minutes to 30 minutes and instructions sheet has been improved in terms of clarity. In 4 days of testing a total of 13 subjects has attended. Subjects were asked to read the improved instructions sheet and to complete the test using Figure 1 Research Strategy Lacinato ABX software on a PC running Windows. They were asked to complete 10 trials on their own without In order to validate the perceptibility of resonance the help of assessors. phenomena, Hal-Saflieni Hypogeum in Malta is selected as the subject of the experiment. A modern cathedral with a similar reverberation time (Saint-Etienne, 3.2 Generation of the Source Audio Sample France) is selected as a control to this study. The hypothesis of this study is that the resonance Design of the test requires a source audio sample where phenomena found in Hal-Saflieni Hypogeum is it is auralised with the Hal-Saflieni Hypogeum and perceptible when compared to a modern cathedral with Saint-Etienne Cathedral to make up two final test a similar reverberation time. The research strategy for samples to be used in the designed experiments. One of this study is shown in fig. 1. the main considerations when choosing the right source audio sample was to avoid distractive sounds and The approach taken for this study is to design a set of eliminate musical styles that might cause bias. For this experiments to test the hypothesis based on a technique reason, the other main consideration was to obtain an called auralisation, where acoustic fingerprint of a space audio sample with equal power over the frequency (impulse response) is used for the process. The impulse band. This is due to Hal-Saflieni Hypogeum having response (IR) recorded from a building holds the multiple resonant frequencies. The best way to trigger absolute acoustical response of the building and can be those resonances with equal power is to use a sine digitally processed with any sound to emulate the sweep. The use of a sine sweep will allow the subjects to buildings acoustical response (auralisation). For this hear the resonances at different frequencies study, extensive IR library of Altiverb convolution individually, starting from a low frequency sweeping reverb plugin is found to include IR’s of Hal-Saflieni through the pre-mentioned frequency band. Hypogeum and Saint-Etienne Cathedral. For this reason Altiverb plugin is used as the auralisation tool to Review of existing literature showed the effects of produce the audio test samples. frequency band 90-120 Hz, studied under EEG (Debertolis et al., 2014b; Cook et al., 2008). This study is Source audio file is selected to be a sine sweep of an based upon the previously researched frequency band. octave (C1 - C2) ranging from 65 Hz to 130 Hz. The For the generation of audio samples an octave (C1 - C2) source audio file is auralised by Altiverb plugin to is selected where the sweep starts from 65 Hz and ends create the final audio test samples. The spectral data of at 130 Hz. the final audio files produced are analysed to detect any peaks related to resonant phenomena. The spectra of the audio files show the amount of energy frequency

3.3 Auralisation 3.5 Testing Methodology

AudioEase Altiverb 6.1.2 universal audio plugin is used Discrimination test methods are widely used in sensory for auralisation of the signals. The impulse responses of analysis to detect any perceptible difference between the test buildings are available in the IR library two products and it is assumed to be involved in a integrated inside the plugin. Auralisation configurations binomial experiment (Bi, 2005; Lawless, 2013). of the plugin with the selected IR’s are shown below in Binomial experiment is made up of Bernoulli trials fig. 2 and fig. 3. The source audio samples are auralised where the outcome of each trial is either success or by 100% dry/wet ratio in order to maintain the full failure. Thus, the number of s successes in a number of n sonic responses. trials is called a binomial variable that follows a binomial distribution (Bi, 2005).

In order to test the perceptibility of small impairments in audio, a range of discrimination methods have been analysed. Among those methods ABX test has been found to be the most suitable. ABX test is a prototypical discrimination test for the assessment of categorical perception (Gerrits and Schouten, 2004). The categories set for this study are the auralisations of Hal Saflieni Figure 2 Altiverb Hal Saflieni Hypogeum Configurations Hypogeum and St Etienne Cathedral. In an ABX listening test subject are presented with samples A, B and X

where X is randomly set to either sample A or sample B. Subjects are required to state either X=A or X=B in each trial.

Statistical significance of responses from ABX test can be figured out by binomial experiments. Using binomial experiments yields a p-value, which represents the probability of randomly getting same or more correct

identifications with the same conditions. Thus, Figure 3 Altiverb Saint-Etienne Cathedral Configurations confidence level of each subject’s response or the total number of trials can be calculated. Typically, a 95% 3.4 Initial Conditions confidence level is sufficient for psychoacoustic experiments (Boley and Lester, 2012; Meyer, 1990) The summary of initial conditions set for this study is stated under this topic. Regarding the scope of this Listening tests consist of 10 trials stretched over no study, dependent test variable is the subject response more than 30 minutes in order to prevent hearing and controlled independent variables are the fatigue and unexpected variability (Harris and Holland, auralisation categories. In order to meet the 2009). With the chosen method for the study, regarding specifications of the chosen test method (Lawless, to 95% confidence level, 7 successes out of 10 trials 2013), experienced critical listeners are called for the yields a 88% confidence level and 8 successes out of 10 study from MSc Digital Media: Audio Production and trials yields 95% confidence level. Thus a majority of at MSc Digital Media: Video, Audio and Social Technologies least 8 successes are needed to disprove the null courses at the University of Salford. The chosen sample hypothesis on individual level. group has an age distribution of 22-25 years old students with no reported hearing impairments. Initial Control of variables is important in order to get statistical test with dummy data has shown that testing unbiased response. In order to avoid unexpected of 13-16 subjects with 10 trials each yields a sufficient variability test procedure, environment, instructions statistical significance to reject the null hypothesis. and equipment have been kept consistent throughout the experiment.

3.6 Spectral Analysis 3.8 Listening Test Equipment

As a control for the subjective tests, frequency spectrum The desktop computer available in the listening of the test audio samples are analysed using Sonic environment running Windows 7 has been used in the Visualiser and Audacity software. Sonic Visualiser has listening tests. The available headphones’ frequency been used for obtaining spectrograms of the signals responses are critically analysed and AKG k240 MKII where amplitude of the signals are visualised with a model has found to be the best candidate. Graph below user-specified colour theme on a frequency versus time shows the frequency response of the AKG k240 MKII. graph. This visualisation method is useful for detecting There is a 4 dB increase in the amplitude of the tested resonances in the auralised samples where amplitude frequency band (65 Hz - 130 Hz) starting from 60 Hz. differences are shown over the time period. For a more quantitative analysis of the peaks in the spectrums of the test audio samples, Audacity has been used for obtaining frequency spectrum plots. This view of the signal shows amplitude versus frequency without consideration of time.

3.7 Listening Test Software

In order to eliminate any bias that may originate from the assessors and to avoid unexpected variance due to randomisation of the samples, available ABX software have been researched. Available software were Figure 5 Signal flow diagram of the listening test equipment critically analysed in terms of their ease of use and intuitive user interface. The analysed software are Signal flow diagram is shown in fig. 5. Computer ABC/Hidden Reference Audio Comparison Tool, running Lacinato ABX software is connected to M Audio WINABX and Lacinato ABX. Lacinato ABX software has MobilePre via USB 2.0 cable. Headphones AKG k240 is been qualified as most satisfying of all, offering simple connected to headphone output of MobilePre via 1/4’’ graphical user interface (GUI) as well as accuracy and TRS Jack. Subjects are asked to set their preferred confidence scores at the end of each test. The subjects playback volume and maintain a consistent level operate the software using a simple user interface and throughout the experiment. A picture taken from the complete the required number of trials set by the testing environment is shown in fig. 6. assessor. Figure 4 shows a screenshot from Lacinato

ABX software GUI.

Figure 6 Picture taken from testing environment Figure 4 Lacinato ABX Software GUI

3.9 Listening Test Environment

In order to avoid any interference from room acoustics that may affect the perception of sound, listening tests are carried out by use of headphones. However, background noise and sound insulation specifications of the test environment is still of high importance. The environment chosen for this experiment due to considerations stated above is the purpose-built Listening Room at the Acoustics Research Centre in Salford. The Listening Room meets the requirements of ITU-R BS 1116-1 for subjective assessments of small impairments in audio systems. The background noise level is at 5.7 dBA and the reverberation time for mid- Figure 8 Spectrogram of Saint-Etienne Cathedral frequencies is 0.27 seconds. auralisation

Figure 9 shows the spectrogram of the audio signal 4.0 Results auralised with Hal Saflieni Hypogeum. There is a set of 4 strong resonances and a set of 5 less strong resonances In this section, data gathered from listening tests and visible in the spectrogram. These resonances are spectral analysis of the audio samples are shown. represented by energy branching out of the main sweep coloured in red where less energy corresponds to green 4.1 Audio Signal Analysis Results areas. These are the resonant frequency points where sound keeps resonating for a small but reasonable time, A wave representation is shown below in fig. 7, from the forming branch-like structures. analysis of two files in Audacity software. First audio track contains audio sample representing Hal Saflieni Hypogeum and the second track contains the audio sample representing Saint-Etienne Cathedral. Huge difference in the first third of the audio signal is clearly visible. This region corresponds to the 65 - 90 Hz region in the generated sine sweep.

Figure 7 Waveforms of the audio files on Audacity: 1) Hal Saflieni Hypogeum auralisation 2) Saint-Etienne Cathedral auralisation. Figure 9 Spectrogram of Hal Saflieni Hypogeum auralisation

Frequency spectrum plots and spectrograms of the In the frequency spectrum plots of audio samples audio files are shown below. Figure 8, shows the analysed in Audacity, exact resonant frequencies are spectrogram of audio signal auralised with Saint- more visible. These figures show more quantitative Etienne Cathedral. Energy decay rates per frequency bin results regarding to amplitude measurements of each are visible with uniform distribution over the frequency resonant frequency. band.

4.2 ABX Test Results

This sub-section presents the findings from ABX tests carried out. A total of 13 subjects were tested in three days at the Listening Room in Acoustics Research Centre. The raw data (subject responses) from the experiments can be found in the Appendices. The figure 12 shows confidence levels per subject calculated by using BINOMDIST function in MS Office Excel 2011. The majority of subjects have scored 99.9% confidence where two subjects are below the 95% confidence level. However, the overall accuracy of the total number of completed trials is 119/130. With 100% confidence the overall score is not by chance.

Figure 10 Frequency spectrum of Saint-Etienne Cathedral auralisation Confidence Levels per Subject Response Figure 10 shows the frequency spectrum of the audio Conidence Levels (%) sample auralised with Saint-Etienne Cathedral. There are a series of peaks in the spectrum with small differences in the amplitude, biggest of which is 13 dB, 13 99.90% corresponding to around 125 Hz (highest peak) compared to first peak at 68 Hz. 12 99.90%

Figure 11 shows the frequency spectrum of the audio 11 99.90% signal auralised by Hal Saflieni Hypogeum. First peak appears at around 65 Hz where the second and the highest peak in the spectrum is at 77 Hz with 22 dB 10 99.90% difference. This is the strongest resonance appearing in the tested frequency band. Series of other peaks are 9 99.90% visible at 86, 97, 108, 120, 125 Hz. The amplitude differences of the peaks in fig. x compared to fig. y are 8 75% significant.

7 99.90% Subject ID 6 99%

5 99.90%

4 99.90%

3 95.60%

2 88%

1 99.90%

0.00% 20.00% 40.00% 60.00% 80.00% 100.00%

Conidence Level Figure 11 Frequency spectrum of Hal Saflieni Hypogeum auralisation Figure 12 Confidence levels per subject response.

7 The figure 13 shows the binomial probability of least 4 dB difference with the highest response of the occurrence of the subjects’ responses individually. This SE. figure is a contrast view of the subject confidence levels. When looking at these values, critical level of 5% The differences in the acoustical responses of the Hal binomial probability of occurrence is the threshold for Saflieni Hypogeum and Saint-Etienne Cathedral are psychoacoustic significance. As also seen in the figure purely the result of their architecture. Saint-Etienne 12 subjects 2 and 8 are under the predefined threshold. Cathedral has a high ceiling with rectangular shaped architecture. On the other hand Oracle Chamber of Hal Saflieni Hypogeum (where the IR was recorded by AudioEase staff) has a very low ceiling with smoothed corners acting as a waveguide. There is a small hole on the wall in Oracle Chamber (fig. 14) that is hypothesised to be the spot for ritual chanting.

Figure 13 Binomial probability of occurrence per subject response

Overall, 11 subjects were able to differentiate the two audio files with 9 subjects scoring 100% accuracy with 99.9% confidence. Figure 14 (Spiral Roof n.d.). The Oracle Chamber in Hal 5.0 Discussion Saflieni Hypogeum Moving to the results of the ABX listening tests, the For its convenience, audio samples auralised with Hal- overall score of the total completed trials show decisive Saflieni Hypogeum and Saint-Etienne Cathedral will be results. 113 positive answers in 130 trials correspond called HS and SE respectively. to a 100% confidence level that the subjects weren’t guessing. On individual levels, 9 out 13 subjects has Differences in spectral views of the audio samples scored 10/10 accuracy where two of the remaining presented in 4.1 are clearly visible at the first sight. subjects scored 8/10 and 9/10 with over 95% Spectrogram view of the SE shows a uniform decay rate confidence level. The subjects who didn’t pass the 95% throughout the frequency band where spectrogram of level are subjects 2 and 8, with confidence levels 88% HS shows 4 strong resonances with slow decay rates and 75% respectively. The results of this test shows that and 5 less strong resonances with less sustain. The the majority of the subjects could differentiate between strongest of these resonances is appearing around 77 the two audio files. Hz. These resonances are clearer to spot at the frequency spectrum of the HS. The resonance at 77 Hz is The results of the ABX tests and the analysis of the at -18.2 dB. This resonance in the first third of the sine audio files suggest that there are significant differences sweep is also clearly visible at the waveform in fig. 7. In in the bass frequency response of Hal Saflieni contrast the highest peak in the SE appears at 115 Hz at Hypogeum and Saint-Etienne Cathedral. The main 29.6 dB. difference in this bass region is due to strong resonances of Hal Saflieni Hypogeum. The null If the uniform auralisation procedure that was kept for hypothesis has been disproven. both audio files is taken into consideration, 12.6 dB difference at the resonant peaks shows the significance The studies on the effects of listening to the tones in this of the resonance phenomena at Hal Saflieni Hypogeum. frequency range on regional brain are reviewed in Lower peaks of the HS (86, 120, 125 Hz) are also quite section 2.0. Exposed to tones from 90 – 120 Hz, changes significant with amplitudes over -25 dB, showing at of the state of consciousness of the subjects are studied

8 previously (Cook et al., 2008; Debertolis et al., 2014). 7.0 Recommendations for further work Combined with the results of earlier studies, the perceptible resonances at the Hal Saflieni Hypogeum at This study is small-scale project that can be improved to 77, 86, 120, 125 Hz (previously studied frequency find more about the perceptual significance of old range) forms a link. There is clear evidence that the temples/sanctuaries. For a detailed analysis of ancient visitors of the hypogeum could physically perception, experiments need to be carried out inside induce altered states of consciousness using sound. This the sanctuaries, using prehistoric instruments like frequency range corresponds to male voice range where frame drums and male chanting, monitoring the it could suggest that the initiator of the altered states of emotional state of the subjects with portable consciousness of the rituals taken place at Hal Saflieni equipment. Possible techniques for monitoring Hypogeum was probably a male. On the other hand it emotional state could be previously mentioned EEG and could be feasible to achieve same effects by drumming TRV as well as MEG, skin conductance, muscle activity continuously, which can be played either a male or (EMG), eye movements, etc. For more controlled test female. environments, experiments can be carried out in anechoic chambers with the use of ambisonic speaker The disadvantages of this experiment are: arrays. With the correct positioning of sound source and using spherical microphone arrays, spatial sound • IR files were recorded using different recordings of the sites can be made. This leads to a very equipment by AudioEase staff. The change of detailed real-life acoustical fingerprint of the space. By equipment can affect the recorded IR’s. designing a specialised DSP system, ambisonics can be • The IR files used was recorded in mono used to create a virtual immersive environment where format. the perceptual studies can be accommodated. • The headphones used in the listening tests have an acceptable bass frequency response, 8.0 References with 5 dB difference in the tested frequency range. (2014) Spiral Roof, SNIPERSHOT. Hypogeum of Hal Saflieni. Available at: http://snipershot.org/shot/?p=794 (Accessed: 14

January 2015). 6.0 Conclusions Bi, J. (2005)Sensory Discrimination Tests and In the Background section, the existing literature and Measurements: Statistical Principles, Procedures, and studies have been presented about the evidence of Tables. 1st edn. United States: Wiley, John & Sons, resonance phenomena found throughout Europe’s Incorporated. prehistoric architecture. Although the intentionality of their builders is the main question for mainstream Boley, J. and Lester, M. (2012) ‘Statistical Analysis of academics, lack of a quantitative study on subjective ABX Results Using Signal Detection Theory’, in Audio perceptibility of these resonances formed the Engineering Society127th Convention 2009 October 9– motivation for this study. Two types of data are 12 New York, NY, USA. gathered from the designed experiments (Section 3.0) and presented in section 4.0. Spectral analysis of the Cook, I., Pajot, S. and Leuchter, A. (2008) ‘Ancient auralisations showed that there are quantitative Architectural Acoustic Resonance Patterns and Regional differences within the acoustical responses of Brain Activity’,Time and Mind, 1(1), pp. 95–104. doi: prehistoric hypogeum and a modern cathedral. Results 10.2752/175169608783489099. from the listening tests suggest that the majority of the subjects could differentiate between the two Debertolis, P., Gullà, D. and Richeldi, F. (2014) auralisations, thus, disproving the null hypothesis. The ‘Archaeoacoustic analysis of an ancient hypogeum using central finding and the suggestion of this study is that new TRV camera (Variable Resonance Camera) the perceptually significant acoustical effects at technology’, inScientific Conference. prehistoric sites like Hal Saflieni Hypogeum should be studied collaboratively with a wide perspective, Debertolis, P., Tirelli, G. and Monti, F. (2014) ‘Systems of incorporating neurophysiology, acoustics, and acoustic resonance at ancient sites and related brain archaeology methodologies. Archaeoacoustics offers a activity’, OTSFArchaeoacoustics: The Archaeology of multisensory perspective to our history. Sound.

9 FinlandMonographs of the Archaeological Society of Douglas, K. (2014) ‘Foreword’, in C. Eneix, Finland, 2. L.Archaeoacoustics: The Archaeology of Sound. Till, R. (2014) ‘Sound Archaeology: An Interdisciplinary Eustace, D. (no date)Listening Room, University of Perspective’, in C. Eneix, SalfordAcoustics Research Centre. University of Salford | L.Archaeoacoustics: The Archaeology of Sound. A Greater Manchester University. Available at: http://www.acoustics.salford.ac.uk/facilities/?content= Watson, A. and Keating, D. (1999) ‘Architecture and listening (Accessed: 9 January 2015). sound: an acoustic analysis of megalithic monuments in prehistoric Britain’, AntiquityAntiquity, 73. Gerrits, E. and Schouten, M. (2004) ‘Categorical perception depends on the discrimination 9.0 Appendices task’,Perception & Psychophysics. Springer, 66(3), pp. 363–376. doi: 10.3758/BF03194885. 9.1 AKG k240 frequency response

Harris, L. E. and Holland, K. R. (2009) ‘Using Statistics to Analyse Listening Test Data: Some Sources and Advice for Nonstatisticians’, Institute of AcousticsProceedings of the Institute of Acoustics, 31.

HeadRoom (2015) ‘AKG k240 Frequency Response’.

Jahn, R. G., Deveroux, P. and Ibison, M. (1995) ‘Acoustical Resonances at the Assorted Ancient Structures’,ICRL Report, 95002. Figure 15 AKG k240 frequency response graph

Johnson, D. L. (1980) ‘The Effects of High Level Infrasound’, inConference on Low Frequency Noise and 9.2 ABX Scores HEaring.

J Waller, S. (1994) ‘Acoustical Studies of Rock Art Sites on Three Continents’, in ChakravertyFrom Rock Art to Tribal Art: A Global Perspective.

Lawless, H. (2013) ‘Comparison of Discrimination Test Methods’, in Lawless, H.Laboratory exercises for sensory evaluation. Volume 2. United States: Springer.

Table 1 ABX test scores of the subjects Meyer, E. B. (1990) ‘ABX Testing’, Boston Audio SocietyBASS, 19. 9.3 Test Audio Files

Qibai, C. Y. H. and Shi, H. (2004) ‘An Investigation on the Auralised audio files can be found at: Physiological and Psychological Effects of Infrasound on Persons’,Low Frequency Noise & Vibration and Active • https://soundcloud.com/limbovectors/st- Control, 23(1), pp. 71–76. doi: etienne-cathedral-sine-sweep 10.1260/0263092041456828. • https://soundcloud.com/limbovectors/hal- saflieni-sine-sweep Reznikoff, I. (2008) ‘Sound resonance in prehistoric times: A study of Paleolithic painted caves and rocks’, inAcoustics 08 Paris.

Reznikoff, I. (2014) ‘On the Sound Related to Painted Caves and Rocks’, Archaeological Society of