**3. Analysis of the results**

#### **3.1. Average hearing threshold**

**Figure 1** shows the values for threshold of hearing for the left and right ear of the population tested before the experiment. These values have been averaged over results obtained for 276 listeners. It can be easily seen that the threshold of hearing is uniformly shifted by about 7–8 dB. In order to confirm the results, various types of statistical testing have been applied. When a calculated value of particular statistics for a tested factor is less than the critical value, depending mainly on the number of repetitions and the level of significance *α* (usually stated as 0.05), the influence of this factor is not important from statistical point of view, so it can be fairer to say that this factor does not influence the obtained results. In this case, the Bartlett test has been used. This test features the distribution asymptotic to χ2 thus it can be applied even to a small population. This kind of test enables to confirm homogene-

**Figure 1.** The average values of the threshold of hearing shift for the tested population.


**Table 1.** The average values and standard deviations for hearing thresholds for left and right ears measured for all the 276 subjects.

ity of variances of obtained results, with the assumption that they featured a normal distribution. The results of statistical treatment showed that the variances of obtained results were homogenous (χ2 = 24.893 < χ*<sup>α</sup>* 2 = 39.977, at *α* = 0.05) for all frequencies. According to the classification of the Bureau International Audiophonology [19], five types of hearing loss can be distinguished:

• hearing loss up to 20 dB: normal hearing

simply refers to an accuracy increase because the attention of listeners was focused only on the noticeable changes between presented samples, without additional tasks about scaling

In this experiment, the TTS phenomenon for the listeners was also the subject of research. The hearing thresholds were measured after every session of music exposure which enabled observation of the TTS caused by listening of loud musical signals in several periods of exposure. In this case, the thresholds of hearing were measured in the same way that at the beginning of experiment, i.e., by ascending stimuli methods and with the use of continuous

**Figure 1** shows the values for threshold of hearing for the left and right ear of the population tested before the experiment. These values have been averaged over results obtained for 276 listeners. It can be easily seen that the threshold of hearing is uniformly shifted by about 7–8 dB. In order to confirm the results, various types of statistical testing have been applied. When a calculated value of particular statistics for a tested factor is less than the critical value, depending mainly on the number of repetitions and the level of significance *α* (usually stated as 0.05), the influence of this factor is not important from statistical point of view, so it can be fairer to say that this factor does not influence the obtained results. In this case, the Bartlett test has been used. This test features the distribution asymptotic to χ2

it can be applied even to a small population. This kind of test enables to confirm homogene-

**Figure 1.** The average values of the threshold of hearing shift for the tested population.

thus

sinusoidal signals with steps of 2 dB. These measurements were repeated twice.

and identifying the reason of the differences [23, 24].

**3. Analysis of the results**

172 Advances in Clinical Audiology

**3.1. Average hearing threshold**


According to this classification, the tested young people belong to the group of normal hearing, but the shift in the threshold of hearing points with the slow tendency to begin a permanent damage of hearing which is caused by a long‐term work with loud music (see Section 3.3). These values, however, are the average ones and the greatest hearing losses can be balanced by the results for the people with otological normal values that is shown in the **Table 1** as the values of standard deviations, especially for higher frequencies. Thus, it was decided to divide the whole group into the categories which could influence the obtained results and reflect the hearing loss for some specific conditions of working activity as well as kinds of equipment used by the people.

#### **3.2. The influence of different kind of headphones on the threshold of hearing**

In this section, results of pure tone audiometry for users of different types of headphones are presented. These results present "the worse" ear (left or right) for each subject, and these values have been averaged over the people who declare to use particular types of headphones. They are shown in **Figure 2**.

It can be seen that the type of headphones used has a major impact on the threshold of hearing values. On the basis of analysis of variance, it was found that for all tested groups of people using different types of headphones and particular frequencies there was a good convergence between all the subjects' notes and thresholds did not depend on the listener in all cases at

**Figure 2.** The influence of different kind of headphones on the threshold of hearing. Standard deviation values are presented as vertical lines on the tops of the bars.

the 95% of confidence (*p* < 0.02). It was decided to use the *F‐*Snedecor statistics because of nonequal numbers of particular groups of users declaring the specific kinds of earphones. It turned out that except for frequency of 4 kHz there is no relation between the types of preferred headphones and the shift of hearing threshold (*F* < *F*α = 2.75, where, *F* and *F*α are calculated and critical values of *F*‐Snedecor test, respectively, at *α* = 0.05). For the frequency equal to 4 kHz, the influence of the headphone types on the threshold values was observed (*F* = 3.35 > *F*α). It means that the most unfavorable for the hearing are the in‐earphones, especially at high frequencies to which our hearing system is the most sensitive. The air in the ear canal is a natural protection from high‐sound pressure. Using inside earphones the length of the channel is reduced, through which natural protection becomes less effective. A good alternative are semi‐open headphones that in a small way can isolate us from the outside noise. They additionally ensure good hygiene of the ear and by their design, they protect from very high‐sound pressure acting directly on the ear membrane. The results of upward threshold shifts obtained for the 4 kHz frequency are presented in **Table 2**.

In order to determine how the particular kinds of headphones are injurious for hearing conditions, the structure index test as a statistical treatment was applied. This test allows to classify the groups of results as influenced by a particular factor, the kind of headphones and its influence on the hearing threshold values in this case. The results of such testing for these series reflect the degree of hearing damage caused by the type of used headphones, with *u*α *=* 1.96 at *α =* 0.05*.* It turned out that for the frequency of 4 kHz the most dangerous type of headphone for the hearing threshold is the in‐ear one (|*u*| = 4.73), while an influence of the semi‐open is inessential statistically (|*u*|= 1.05 < *u*α). The degree of injury for hearing damage obtained for the open and the closed headphones are lower than for the in‐ear headphones (|*u*|= 2.52 and |*u*|= 2.12, respectively).


**Table 2.** The average values for upward shift of hearing thresholds at 4 kHz for various types of headphones used by investigated subjects (in dB).

#### **3.3. Threshold of hearing in terms of professional work**

the 95% of confidence (*p* < 0.02). It was decided to use the *F‐*Snedecor statistics because of nonequal numbers of particular groups of users declaring the specific kinds of earphones. It turned out that except for frequency of 4 kHz there is no relation between the types of preferred headphones and the shift of hearing threshold (*F* < *F*α = 2.75, where, *F* and *F*α are calculated and critical values of *F*‐Snedecor test, respectively, at *α* = 0.05). For the frequency equal to 4 kHz, the influence of the headphone types on the threshold values was observed (*F* = 3.35 > *F*α). It means that the most unfavorable for the hearing are the in‐earphones, especially at high frequencies to which our hearing system is the most sensitive. The air in the ear canal is a natural protection from high‐sound pressure. Using inside earphones the length of the channel is reduced, through which natural protection becomes less effective. A good alternative are semi‐open headphones that in a small way can isolate us from the outside noise. They additionally ensure good hygiene of the ear and by their design, they protect from very high‐sound pressure acting directly on the ear membrane. The results of upward threshold

**Figure 2.** The influence of different kind of headphones on the threshold of hearing. Standard deviation values are

In order to determine how the particular kinds of headphones are injurious for hearing conditions, the structure index test as a statistical treatment was applied. This test allows to classify the groups of results as influenced by a particular factor, the kind of headphones and its influence on the hearing threshold values in this case. The results of such testing for these series reflect the degree of hearing damage caused by the type of used headphones, with *u*α *=* 1.96 at *α =* 0.05*.* It turned out that for the frequency of 4 kHz the most dangerous type of headphone for the hearing threshold is the in‐ear one (|*u*| = 4.73), while an influence of the semi‐open is inessential statistically (|*u*|= 1.05 < *u*α). The degree of injury for hearing damage obtained for the open and the closed headphones are lower than for the in‐ear headphones (|*u*|= 2.52 and

shifts obtained for the 4 kHz frequency are presented in **Table 2**.

presented as vertical lines on the tops of the bars.

174 Advances in Clinical Audiology


Some of tested people have been working in the profession for 7 years. By analyzing these data, it can be concluded that even 3–4 years of working in the entertainment industry, especially as the front‐of‐house engineers may cause a slight loss in hearing ability. By comparing other professional groups, it can be assumed that the results coincide in a large extent and the type of work (noise level) has no longer such effect on the threshold of hearing. In **Figure 3**, hearing thresholds are presented depending on the profession. In **Figure 3**, there are also results for the ordinary user of portable equipment – there are the subjects nonpracticing in any kind of sound‐engineering profession as well as musicians. These results present "the worse" ear (left or right) for each subject, and these values have been averaged over the people within the particular group of profession as well as "amateur" listeners.

On the basis of analysis of variance, it turned out that for frequency values of 500 Hz, 1 kHz as well as 4 kHz the influence of working activity on the threshold of hearing has been observed (*F* > *F*α = 3.29, where *F* and *F*α are calculated and critical values of *F*‐Snedecor test, respectively, at *α*= 0.05). For the other frequencies, there is no relation between the profession of work and the shift of hearing threshold values. As it was mentioned in previous chapter, the hearing loss at 4 kHz can be interpreted as the beginning of permanent hearing damage resulting from the exposure to the sound at high levels while the upward threshold shifts that appeared

**Figure 3.** Thresholds of hearing depending on the profession. Standard deviation values are presented as vertical lines on the tops of the bars.

for lower frequencies (500 and 1000 Hz) are the results of the exposure to hyper‐compressed musical sounds in these frequency bands, especially occurring on stage in order to increase the total loudness impression.

#### **3.4. Detection of spectral changes vs. auditory fatigue**

In **Figure 4**, results of this part of the research are presented. They are expressed as the percentage of correct answer number obtained before and after the loud music exposure. Subjects listened to the test trials containing the introduced several spectral modifications and have to denote if they perceived them. Thus, results may be expressed as a percentage of correct answers in a dependence of degree of introduced corrections for several noise‐like exposures. For statistical treatment, the Bartlett's test was applied allowing the confirmation of homogeneity of variances of obtained results. On the basis of this test, for every exposure, the results were homogeneous (*χ*<sup>2</sup> = 4.922 < χ*<sup>α</sup>* 2 = 28.869, at *α* = 0.05). Thus, the obtained results may be averaged over the total number of subjects and over the all styles of musical material. It is clearly noticeable that the differences before and after exposure for particular frequency are significant (χ2 = 9.103 > χ*<sup>α</sup>* 2 = 5.986, at *α* = 0.05).

It can be noted that the decrease in ability to detect the spectral changes for longer noise exposure has been observed particularly for lower changes and all frequency regions. Moreover, the number of false alarms (i.e., the case when the subjects signalized that some correction had been introduced, but no spectral changes have been really done) is less than 5% of the number of total answers at a specific condition which means that listeners mostly have not perceived the small changes in spectra. The changes of ±1.5 dB are perceived with detection ability higher than 70% only at the beginning of the test for middle and higher frequency regions. When subjects are exposed to noise for a longer period, their ability to detect changes in the spectrum of musical signals is less effective. For the noise exposure longer than 1 h the ability gets worse, especially for 8 kHz octave band where the only larger (±6 dB) equalization of the musical sounds may be perceived properly. This fact can be explained by the nature of frequency analysis made by the hearing system: this range of frequency is responsible for the proper reproduction of temporal structure of transient sounds [25], and the influence of rise time, especially, for the loudness impression has been reported [26]. The loudness changes may be perceived effectively when the "carrier" sound levels are higher than the hearing threshold of 10–20 dB [27]. When the changes of spectra in this frequency region do not exceed ±3 dB the difference of spectrum could be detected less effectively than in other investigated frequency bands because of the lower loudness impression in this region after several times of sound exposure. For octave bands of 125 Hz as well as for 1 kHz, the perceived spectral changes at the level of 70% have been noted for ±3 dB, or greater. It may suggest that the hearing system gets tired for the region of higher frequencies faster than for other bands after listening to a loud music. It can be shown that the trend is almost the same for every frequency of notched/boosted bands: when the attenuation, or amplification in a particular octave band increases, the percentage of correct answer reflecting the ability of detection of changes in the spectrum of musical signals also increases. It can also be observed that the differences between obtained values for increasing time of loud music exposure gets lower when the changes in spectra increase: the difference of ability of perception of spectrum

for lower frequencies (500 and 1000 Hz) are the results of the exposure to hyper‐compressed musical sounds in these frequency bands, especially occurring on stage in order to increase

In **Figure 4**, results of this part of the research are presented. They are expressed as the percentage of correct answer number obtained before and after the loud music exposure. Subjects listened to the test trials containing the introduced several spectral modifications and have to denote if they perceived them. Thus, results may be expressed as a percentage of correct answers in a dependence of degree of introduced corrections for several noise‐like exposures. For statistical treatment, the Bartlett's test was applied allowing the confirmation of homogeneity of variances of obtained results. On the basis of this test, for every exposure, the results

averaged over the total number of subjects and over the all styles of musical material. It is clearly noticeable that the differences before and after exposure for particular frequency are

It can be noted that the decrease in ability to detect the spectral changes for longer noise exposure has been observed particularly for lower changes and all frequency regions. Moreover, the number of false alarms (i.e., the case when the subjects signalized that some correction had been introduced, but no spectral changes have been really done) is less than 5% of the number of total answers at a specific condition which means that listeners mostly have not perceived the small changes in spectra. The changes of ±1.5 dB are perceived with detection ability higher than 70% only at the beginning of the test for middle and higher frequency regions. When subjects are exposed to noise for a longer period, their ability to detect changes in the spectrum of musical signals is less effective. For the noise exposure longer than 1 h the ability gets worse, especially for 8 kHz octave band where the only larger (±6 dB) equalization of the musical sounds may be perceived properly. This fact can be explained by the nature of frequency analysis made by the hearing system: this range of frequency is responsible for the proper reproduction of temporal structure of transient sounds [25], and the influence of rise time, especially, for the loudness impression has been reported [26]. The loudness changes may be perceived effectively when the "carrier" sound levels are higher than the hearing threshold of 10–20 dB [27]. When the changes of spectra in this frequency region do not exceed ±3 dB the difference of spectrum could be detected less effectively than in other investigated frequency bands because of the lower loudness impression in this region after several times of sound exposure. For octave bands of 125 Hz as well as for 1 kHz, the perceived spectral changes at the level of 70% have been noted for ±3 dB, or greater. It may suggest that the hearing system gets tired for the region of higher frequencies faster than for other bands after listening to a loud music. It can be shown that the trend is almost the same for every frequency of notched/boosted bands: when the attenuation, or amplification in a particular octave band increases, the percentage of correct answer reflecting the ability of detection of changes in the spectrum of musical signals also increases. It can also be observed that the differences between obtained values for increasing time of loud music exposure gets lower when the changes in spectra increase: the difference of ability of perception of spectrum

= 28.869, at *α* = 0.05). Thus, the obtained results may be

the total loudness impression.

176 Advances in Clinical Audiology

were homogeneous (*χ*<sup>2</sup>

= 9.103 > χ*<sup>α</sup>*

significant (χ2

**3.4. Detection of spectral changes vs. auditory fatigue**

= 4.922 < χ*<sup>α</sup>*

2

2

= 5.986, at *α* = 0.05).

**Figure 4.** Detection of spectrum changes for frequencies of 125 Hz (a), 1 kHz (b) and 8 kHz (c) for different values of level changes in particular octave band.

**Figure 5.** Average values of TTS after noise exposure of 1 h, 1.5 h and 2 h.

changes between fresh‐ear listening and perception after 2 h‐exposure takes 20% for ±1.5 dB spectrum modification, and then decreases to about 10% for ±6 dB attenuation/amplification. These results are convergent to the ones obtained in experiments on the profile analysis [28–30]: the values reported in a literature are equal to 2–3 dB for similar frequency regions, which can be compared to the values obtained for detection ability at 70% of correct answer number measured before the exposure to the music treated as a disturbing noise.

The obtained results can also be discussed in the light of TTS values presented in **Figure 5**. They have been averaged over all listeners. As it can be seen, the greatest values of TTS have been obtained for 1 kHz (about 9.5 dB, after 120 min of exposure) but the way of change is monotonic for all investigated frequencies. Moreover, the differences between the TTSs after the loud music exposure of 1 and 2 h are about 4 dB, for all frequencies. These values are


**Table 3.** Standard deviation values of percentage of correct answer for spectra changes of musical samples equalized at 125 Hz, measured at different times of loud music exposure (in %).


**Table 4.** Standard deviation values of percentage of correct answer for spectra changes of musical samples equalized at 1000 Hz, measured at different time of loud music exposure (in %).

greater than those resulting from the detection ability presented in **Figure 4** because of the different stimuli used in both tests, although the character of changes is similar.

For a quality of the work activity in this particular profession it is important to detect these changes as accurate as possible, especially at a work as a studio recording engineer. However, the long exposure to the noise causes the worsening of attention, or listening fatigue. This phenomenon may be expressed as the standard deviations values of obtained results which is presented in **Tables 3**–**5**. These values may show that after every sound exposure, the attention of listeners gets lower causing an increase of uncertainty during evaluation of musical samples.

changes between fresh‐ear listening and perception after 2 h‐exposure takes 20% for ±1.5 dB spectrum modification, and then decreases to about 10% for ±6 dB attenuation/amplification. These results are convergent to the ones obtained in experiments on the profile analysis [28–30]: the values reported in a literature are equal to 2–3 dB for similar frequency regions, which can be compared to the values obtained for detection ability at 70% of correct answer

The obtained results can also be discussed in the light of TTS values presented in **Figure 5**. They have been averaged over all listeners. As it can be seen, the greatest values of TTS have been obtained for 1 kHz (about 9.5 dB, after 120 min of exposure) but the way of change is monotonic for all investigated frequencies. Moreover, the differences between the TTSs after the loud music exposure of 1 and 2 h are about 4 dB, for all frequencies. These values are

**‐1.5 dB +1.5 dB ‐3 dB +3 dB ‐6 dB +6 dB**

number measured before the exposure to the music treated as a disturbing noise.

*σ*<sup>0</sup> 18.2 14.0 8.4 6.6 4.8 3.9 *σ*<sup>60</sup> 27.3 22.8 20.3 18.6 15.0 11.5 *σ*<sup>90</sup> 31.1 33.2 24.7 24.3 13.2 12.7 *σ*<sup>120</sup> 36.0 34.3 25.3 24.2 16.5 15.8

**Table 3.** Standard deviation values of percentage of correct answer for spectra changes of musical samples equalized at

**Figure 5.** Average values of TTS after noise exposure of 1 h, 1.5 h and 2 h.

125 Hz, measured at different times of loud music exposure (in %).

**Spectral change/ standard deviation**

178 Advances in Clinical Audiology

It can be seen that precision in spectral changes detection increases when these changes are greater (±6 dB, in this case). Another interesting fact is that after every acting noise (ranging from 60 to 120 min) the standard deviation values increase, but this change is not monotonic: sometimes exposure time does not influence the value of standard deviation of the obtained results which was confirmed by Bartlett test (*χ*<sup>2</sup> = 3.427 < *χ*<sup>α</sup> 2 = 5.986, at *α* = 0.05), and sometimes this influence is significant (as for 125 Hz band, where *χ*<sup>2</sup> = 11.886 > *χ*<sup>α</sup> 2 = 7.802, at *α* = 0.05). This means that the uncertainty for sound color evaluation for small differences of spectra is relatively high when some masking sounds appear simultaneously which increases the hearing system fatigue. For the lowest investigated equalization (±1.5 dB) the standard deviation for results after listening to loud music takes values greater than those presented for ±3 dB correction. Without the noise‐like signal exposure, the standard deviation is almost the same as for ±3 dB (before listening to loud music) and this is in good agreement with previously reported


**Table 5.** Standard deviation values of percentage of correct answer for spectra changes of musical samples equalized at 8 kHz, measured at different time of loud music exposure (in %).

research [31, 32] for amateurs as well as for professional sound engineers. Taking into account the obtained values for all kinds of spectral modification at given octave bands it can be clearly seen that longer exposure to loud signals causes greater uncertainty of sound color assessment but the relation is not proportional: the great increase has been noted when time exposure is 90 min and further prolongation of noise exposition up to 2 h does not influence the standard deviation values for lower and higher frequency regions, so it might be said that the concentration is kept at the same level. It should be also noted that the values of standard deviation are higher for 125 Hz for a modified frequency band than for higher frequencies which clearly means that uncertainty of spectrum change detection is worse for lower frequencies.

#### **4. Conclusions**

The audibility of timbral modifications depends on the frequency of modified region, the amplitude of peak (or notch) as well as the bandwidth. As it is reported in the literature, changes in sound quality, for example, made by introducing resonances or notches depend on musical material used in audition, the listening environment and reverberation used at a recording process [30]. The most important result of present experiment is that the audibility of spectral changes depends on the level of this modification as well as on the time of disturbing loud music exposure. Moreover, with discontinuous, irregular impulsive, or transient sounds characteristic for speech and musical signals, the test material is less resistible in comparison to the steady sounds. Obtained results are in good agreement with the ones reported in the literature as results of profile analysis [29] as well as the "classical" view on the timbre change perception [28]. It should be noted here that so called traditional view on the timbre perception is based on the intensity discriminations in particular frequency bands while the basic assumption of the profile analysis is that discrimination of the spectral changes is based on the evaluation of the overall spectrum shape involving the memory and interstimuli intervals. The results of experiments provided by both methods are similar in a case of such signals as used in our research. According to this, the ability of the distinguished changes in spectrum are 2–3 dB for listeners with normal hearing. It may be assumed that this fact takes place at the beginning of experiment (before exposure to the loud musical material). For the people with relatively small hearing loss (up to 20 dB) the predicted results of the peak or notch of spectrum modification may be shifted up to 5–6 dB which coincides with our results: the attenuation/amplification must be at 6 dB to be perceived with the greatest accuracy after longer (more than 1 h) presentation of loud music.

On the basis of the obtained results, it may be stated that the temporary threshold shift phenomenon is the important factor that determines perceptibility of changes in spectral and amplitude domains of musical signals. This conclusion results from the way of changes in obtained values for different time of loud music exposure. This is a usual phenomenon especially for 1 kHz because this range of frequency is the most sensitive for human hearing [33] and this fact can help the listeners to take a good decision during sound evaluation. Results of spectral changes detection are convergent with results reported in the literature. According to these results, the TTS measured immediately after loud music exposure ranges from 10 to 30 dB, depends on the level, time, and the temporal and spectral structure of noise or loud music [15, 18]. Moreover, if one can assume that TTS phenomenon causes similar effects that may be characteristic for the hearing loss, the decrease of sensitivity of the hearing system affects the perception of auditory signals in all their dimensions, that is, temporal and frequency resolution as well as loudness perception may be distorted or deteriorated. This effect may be observed in the discotheque attendants or in the people who are exposed to the noise level greater than 90 dB [17].

The results may also be influenced by the mental fatigue which occurred after permanently playing loud sounds for several time durations, together with demanding tasks. Such conditions involving the mental engagement in a noisy environment which is natural in the studio can significantly reduce the time of exhaustion which causes the decrease of accuracy in solving several tasks [9].

Nowadays portable players are getting cheaper, smaller, and offer more and better sound quality. Everything would be fine, if not the fact that listening to the loud music does not hurt. These devices induce young people to listen louder and louder, applying that noise directly on themselves. It is very easy to meet someone on the tram, the bus, or on the street with headphones in their ears and the music is reproduced so loud that it is possible to recognize songs that are played being in a distance from the listening person. The body does not give us a sign that the process of destroying the hearing has just began, and once damaged, hair cells would never regenerate. The results of the research conclude that if a person listens to loud music on MP3 player for 5 years for an hour a day it is enough to ruin a hearing system permanently. Thus, it should be noted that the tendency observed in young people to listen loud music in order to be isolated from the environment is still actual which will cause not the TTS but PTS. The most dangerous factor influencing the human hearing system reported in the literature [8, 10, 18] is the type of headphones used for every day listening. Most of young people listen to the music through inside earphones which causes that the reduction in the length of outer ear channel, and as a consequence, a natural protection becomes less effective. From sociological point of view, the young people like this kind of earphones because they take up little space and can be always carried in a pocket, but on the other hand, they are the worst for our hearing. Research has shown that 2–3 years of using this type of headphones leads to a slight hearing damage resulting with incomprehensibility of a whisper or a quiet voice. Listening to music is becoming an addiction primarily among young people, but unfortunately this fact is ignored in the mainstream media.
