Immediate Effect of Alcohol on Voice Tremor Parameters and Speech Motor Control.
The effect of long term intake of high dosage of alcohol on health has been the trend in medical and psychological research over many decades. Many shocking reports of health and wealth issues have emerged from these studies (Drug and alcohol services South Australia, 2010). But the most immediate and overt symptoms of alcohol intoxication are the changes in speech, language and motor characteristics of an intoxicated individual. The alcohol interferes with the normal functioning of cerebellum and leaves the individual in an excited state of body and mind leading to uninhibited speech and thoughts which are sometimes socially and culturally inappropriate. The effect of alcohol intoxication on communication behaviour (llouck & Moyers, 2015), language formulation (Alexandrov, Sams, Lavikainen, Reinikainen & Naatanen, 1998; Andrews, Cox & Smith, 1977; Jaaskelainen et.al., 1995; Natale, Klanzer, Jaffe & Jaffe, 1980; Pisoni & Martin, 1989) and motor skills (Knapp, 2006) is well known. Speech being a complex interplay of fine motor control within and across speech systems may be one of the first to reveal alcohol intoxication. Superficially, slurred speech is a reliable indication of alcohol intoxication (Van Wyden, online article). Recent literature on the relation between alcohol and speech characteristics has yielded inconclusive results (Table 1).
From the literature, some but varied effects of alcohol intoxication on speech parameters are confirmed and these effects are identifiable, measurable and reliable if variables are controlled. If the various levels of intoxication can produce a specific set of changes in speech, the reverse can be estimated and a model for estimating levels of intoxication from speech samples can be developed. This can then be a reliable and non-invasive substitute to estimate levels of intoxication for purposes of law enforcement. The study by Lee, Bae and Bae (2015) and Lee and Bae, (2016) was conducted in similar directions using phonation samples. But very few studies in literature have controlled the level of intoxication for studying the changes in speech parameters in their subjects. Oueller and Hamsberger (2010) studied the vowel space, speaking rate and fundamental frequency measures with increasing levels of alcohol intoxication.
Toward Surreptitious Remote Sensing of Blood Alcohol Concentration: Results with Integrated Near-Infrared Spectral Imaging and Laser Speech Detection.
To date, the published reports suggest that the effect of alcohol on the acoustic parameter of speech is variable due to the modest amount of research on possible intoxication-speech relationships. Previous research has also shown the importance of setting strict dosage, using the same subjects throughout all procedures and rigorously controlling intoxication levels, recording speech samples in a precise manner each time, making precise speech measurements, managing strict control over participant selection and the use of technology to make measurements (Chin & Pisoni, 1997; Pisoni, Hathaway & Yuchtman, 1985). Thus, there is a need to replicate the study of parameters of speech on a selected set of participants following intake of a standard dose of alcohol with the advantages of the latest digital technology. This study was aimed at this direction by studying the effect of a pre-determined level of alcohol intoxication on speech motor control and voice tremor parameters as assessed using an automated protocol. The specific objective of the study was to compare the performance of participants on measures of Diadochokinesis, Voice Tremor and Formant Transition before and after a pre-determined level of intoxication.
A total of 38 individuals who consume alcohol were interviewed for their habitual drinking habits. They were included in the participant list if they satisfied the following inclusion criteria:
1. Consume alcohol at least once in a week
2. No neuro-muscular or cognitive communication disorders reported
3. Willing to participate in the study
Participants were excluded if the habitual beer dosage was less than the required alcohol dosage (Detailed in the procedure section below) or if they consumed more dosage of alcohol than the required alcohol dosage. If otherwise, the participants were enrolled into the study procedure.
A total of 10 males of 20-30 years of age satisfied the inclusion and exclusion criteria participated voluntarily in the study. All participants were explained the procedure of the study after which consent was taken for their participation.
1. An Olympus WS-400S DNS Digital Voice Recorder for recording the data sample.
2. Omron HN-286 digital weighing machine for body weight measure of participants.
3. Computerized Speech Lab 4500 with its module Motor Speech Profile (Kay Pentax, USA) loaded onto a desktop for obtaining the speech-motor measures and voice tremor measures of the data sample in a laboratory setting.
The body weight of each participant was measured using a digital weighing machine (Omron HN-286) and the amount of beer intake required for an intermediate level of intoxication, i.e. 0.80g/kg of Blood Alcohol Level (Bauer, Schumock, Horn & Ophcim, 1992; Hollien, Hamsberger, Martin, Hill & Alderman, 2009; Nagoshi & Wilson, 1989) calculated using Widmark's formula as given below:
BAL (%) = mass of alcohol (g) / [reduction factor (0.73 for men) * body weight(kg)].
Each participant was given three tasks to perform before and after alcohol intoxication: (1) diadochokinesis [repeated utterance of/pa/, /ta/, and /ka/(AMR) and repeated utterance of /ptk/ in the maximum rate possible (SMR tasks)] (2) Sustained phonation [sustained phonation of /a/] and (3) Formant transition task [repeated utterance of/i-u-i/]. Three trials of each task were recorded from each participant before and after alcohol intoxication on a portable digital voice recorder (Olympus WS-400S DNS Digital Voice Recorder). The data were recorded and stored for later analysis at a laboratory setting using Motor Speech Profile (Computerized Speech Lab 4500, Kay Pentax, USA).
The specific protocols for Diadochokinesis, Voice tremor and Formant Transition Protocols of Motor Speech Profile (MSP) were run on the data. The protocol specific analysis procedure followed for the study is detailed below:
1. Diadochokinesis protocol:
The recorded data on Didochokinesis task was opened on the Motor Speech Profile Platform with a threshold of detection at 55dB. The protocol was run and the detected DDK peaks were visually verified. If less than 90% of peaks were detected, the threshold was raised higher until the 90% level was obtained. The DDK rate, period, consistency and accuracy measures were obtained and tabulated for statistical analysis.
2. Voice Tremor Protocol:
The recorded sustained phonation data were opened on the Motor Speech Profile Platform. The waveform was inspected for stable amplitude and an 8-sec data with good amplitude was trimmed for analysis. The protocol on voice tremor was run on the trimmed data and measures were tabulated for statistical analysis.
3. Formant Transition Protocol:
The recorded data on formant transition (repetition of /i-u/) was opened on the Motor Speech Profile Platform. The waveform was inspected for amplitude level and an 8-sec data with good amplitude was trimmed for analysis. The protocol on formant transition was run on the trimmed data with a filter setting set per manufacturer's guideline. However, the protocol could not be completed due to breaks and low signal to noise ratio. This protocol was eliminated from the study due to high error rates.
RESULTS AND DISCUSSION
The present study was aimed at comparing the Diadochokinesis (DDK), voice tremor (VT) and formant transition (FT) measures obtained from participants before and after alcohol intoxication. Due to technical issues, the forniant transition protocol had to be eliminated from the data sample. Thus, the statistical analysis was run only on DDK and VT measures. The tabulated data on DDK and VT measures were subjected to descriptive statistical analysis for obtaining the mean and standard deviation of DDK and VT measures (Table 2).
From Table 2, a general trend of change in measures of DDK and VT, pre- and post-intoxication could be observed. The DDKavp decreased or remained comparatively stable post intoxication for AMR tasks but increased for SMR tasks. But DDKavr did not show significant change post-intoxication in AMR and SMR DDK tasks. The other measures of DDK such as DDKcvp and DDKcvi decreased post-intoxication. Interestingly, the standard deviation of DDK measures tend to decrease post-intoxication. In VT measures, v[F.sub.0] showed a significant increase post intoxication. Other measures of VT remained stable across the two conditions.
To understand the descriptive data further and to study the changes in DDK and VT measures pre-and post-intoxication in the participants, the data were subjected to non-parametric statistical analysis using Wilcoxon's Sign Rank test. Further, a Friedman test was run to understand the differences in performance across tasks in the pre- and post-intoxication tasks. The result of this statistical comparison is presented and discussed below under the following sections:
Effect of intoxication on rate and consistency Measures of DDK:
Wilcoxon's Sign Rank test revealed significant differences in the DDKavp (Z= 3.557, p < 0.05) and DDKavr (Z= -3.207, p < 0.05) for repetition of the syllable /p/ but not /t/, /k/ or /ptk/. The DDKavp decreased and the DDK avr increased for the repetition of /pa/ post-intoxication. There was no statistically significant effect of intoxication on DDKcvp, DDKjit and DDKcvi for repetition of /t/, /k/ or /ptk/ or alternate motion for /ta/, and /ka/. Interestingly, there was no significant difference in the DDK measures for sequential movement for /ptk/ before and after alcohol intoxication.
The Friedman test run across tasks in the pre-intoxicated and post-intoxicated conditions revealed mixed results. Statistically, significant differences were obtained for DDKavp ([chi square](3) = 8.99, /; < 0.05), DDKavr ([chi square](3) = 6.01, p < 0.05) and DDKjit ([chi square](3) = 14.95, p < 0.05) in the pre-intoxicated condition. Similarly DDKavp and DDKjit showed statistically significant changes post-intoxication also ([chi square](3) = 4.08, p < 0.05; [chi square](3) = 3.787, p < 0.05 respectively) few other DDK measures such as DDKcvp ([chi square](3) = 3.495, p < 0.05) and DDKcvi ([chi square](3) = 3.57, p < 0.05) showed significant differences post-intoxication. A post(add "-")hoc analysis revealed that the average DDK period for /pa/ was significantly different from /ka/ before and after alcohol intoxication but the significant difference that existed between /pa/ and /ta/ before intoxication was absent after intoxication. The Average DDK period of /pa/ was reduced to that of /ta/ suggesting an increased rate of repetition of /pa/ following intoxication. A similar change was noticed in the measure of DDKjit which measures cycle to cycle variation in a period of DDK. The cycle to cycle variation of period in production of /pa/ increased to that of /ta/ following intoxication.
The rapid production of the bilabial syllable /pa/ may suggest an excitatory effect of alcohol on speech. An increase in the cycle to cycle variation of period suggests the reduced ability of individuals to maintain stable CV combinations probably due to the effect of alcohol on the central speech motor control system. The increased rate of CV production may be achieved by the compromised precision of articulatory movements but this needs to be established with further research.
It is interesting to learn that the syllable /pa/ is more affected by alcohol intoxication than the syllables /ta/ and /ka/. This may be related to the mass of the structure that produces these syllables. The tissue mass of lips is greater than the mass of tongue tip that produces /ta/ or a posterior portion of the tongue that produces /ka/ and a greater sensorimotor planning for precise execution of bilabials (Smith, 1978). Alcohol intake may interfere with this planning and can affect the finer control leading to variation in these acoustic parameters.
Though there was an increase in the rate of production of bilabials after alcohol intoxication, this increased rate cannot be generalized to all syllables and in all context. A previous study by Chin, Large and Pisoni (1996-1997) found an increased duration of word segments following alcohol intoxication suggesting a reduced rate of speech. However, their study did not mention the number of word segments that consisted of bilabial sounds in their speech sample.
Effect of intoxication on Measures of Voice Tremor
Wilcoxons Sign Rank test revealed significant difference across the two intoxicated states in the measure of MFTR (Z= -2.240, p = 0.05). None of the other measures revealed significant changes after considerable intoxication. From Table 2, it has been previously observed that the standard deviation of VT measures increased after intoxication, probably indicating a wide individual variability in the performance of the task. This might have interfered with proving the observation statistically for difference in performance, post alcoholic intoxication. The Friedman test and post-hoc was not run on the data as the data included only one task in two intoxicated conditions.
The magnitude of frequency tremor increased after alcohol intoxication. This measure suggests the ability of a person to maintain a constant fundamental frequency during every cycle of vocal fold vibration. An increased MFTR suggests the effect of alcohol intoxication on the laryngeal control for stable sustained phonation. There are some supporting studies that states that lower alcohol levels can affect laryngeal physiology (Klingholtz et.al., 1988). Studies using high speech motion pictures have found counter motion of mucous membrane in intoxicated subjects. A physiological reason behind this incoordination between cycles of vocal vibration may be the effect of alcohol on the cerebellar processes that control laryngeal behaviors. However, a lowering of [F.sub.0] that was revealed in their study could not be replicated here. A higher standard deviation of F0 in the post alcoholic condition may have masked this difference statistically. However, the periodicity of tremor could not be derived due to technical reasons. Increased standard deviation of acoustic measures post alcohol intoxication suggests that the effects of alcohol on the acoustic parameters of speech are variable. This calls for further research in this area for understanding the effect of various amounts of Blood Alcohol Level on acoustic parameters of speech taking into account the inter subject variability in the susceptibility to effects of alcohol intoxication.
Overall, the effect of blood alcohol level of 0.80g/kg which is the permissible limit of intoxication for legal traffic in USA made the acoustic measures of speech susceptible to changes. In India, this limit is lower (0.30g/ kg). From the present study, changes in motor control for articulator movement and voice production was indicated. This opens a new arena of research towards understanding the effect of various levels of alcohol on speech performance and its application on detecting levels of blood alcohol level without invasive methods. If such speech patterns could be derived, a new non-invasive model for alcohol level that substitutes the present methods of blood alcohol/ breath alcohol estimation methods can be developed. A replication of this study with a larger sample size, various levels of BAL and a larger number of acoustic variables is demanded and is being undertaken by the investigators.
Correspondence concerning this article should be addressed to: Ms. Gayathri Krishnan, Clinical Assistant, All India Institute of Speech and Hearing, Mysuru, Karnataka, India; Email: firstname.lastname@example.org.
Alexandrov, Y., Sams, M., Lavikainen, J., Reinikainen, K., & Naatanen, R. (1998). Differential effects of alcohol on the cortical processing of foreign and native language. International Journal of Psychophysiology, 28(1), 1-10.
Andrews, M. L., Cox, W. M., & Smith, R. G. (1977). Effects of alcohol on the speech of non-alcoholics. Central States Speech Journal, 28, 149-143.
Behne, D. M., & Rivera, S. M. (1990). Effects of alcohol on speech: Acoustic analysis of spondees. Research on Speech Perception, 16, 263-291.
Camargo, C. A., Stampler, M. J., Glynn, R. J., Gaziano, J. M., Manson, J. E., Goldhaber, S. Z., & Hennekens, C. H. (1997). Moderate alcohol consumption and risk of angina pectoris or myocardial infarction in U.S. male physicians. Annals of Internal Medicine, 126(5), 372-375.
Chin, S. B., & Pisoni, D. B. (1997). Alcohol and speech. San Diego, CA: Academic Press.
Chin, S. B., Large, N. R., & Pisoni, D. B. (1996-1997). Effect of alcohol on production of words in context. Research on Spoken Language Processing, Indiana University. http://www.iu.edu/~srlweb/pr/PR21_p403_Effects_of_Alchohol_on_the_Production_of_Words_in_Context_A_First_Report.pdf.
DeJong, G., Hollien, H, Martin, C, & Alderman, G.A. (1995). Speaking rate and alcohol intoxication. Journal of the Acoustical Society of America, 97, 3364.
Dodge, R. & Benedict, F.G. (1915). Psychological effects of alcohol: An experimental investigation of moderate doses of ethyl alcohol on a related group of neuromuscular processes in man. Washington, DC: Carnegie Institute of Washington.
Doll, R, Peto, R., Hall, E., Wheatley, K., & Gray, R. (1994). Mortality in relation to consumption of alcohol 13-year observations on male British doctors. British Medical Journal, 309, 911-918.
Drug and Alcohol Services South Australia. Article online retrieved from http:// www.dassa.sa.gov.au/site/page.cfm?u=122
Fairbairn, C. E, Sayette, M. A., Amole, M. C, Dimoff, J. D., Cohn, J., & Girard, J. M. (2015). Experimental and Clinical Psychopharmacology, 25(4), 255-264. http://dx.doi.org/10.1037/pha0000021
Hollingworth, H. L., (1923). The influence of alcohol. Journal of Abnormal Psychology and Social Psychology, 18, 204-237.
Houck, J. M., & Moyers, T. B. (2015). Within-session communication patterns predict alcohol treatment outcomes. Drug and Alcohol Dependence, 157, 205-209.
Jaaskelainen, I. P., Pekkonen, E., Alho, K., Sinclar, .J. D., Sillanaukee, P., & Naatanen, R. (1995). Dose related effect of alcohol on mismatch negativity and reaction time performance. Alcohol, 12(6), 491-495.
Jetter, W. W., (1938). Studies in alcohol 1: The diagnosis of acute alcoholic intoxication by a correlation of clinical and chemical findings. American Journal of the Medical Sciences, 196, 411-487.
Johnson, K., Pisoni, D.B., & Bernacki, R. H. (1989). Report to the NTSB: Analysis of speech produced by the captain of the Exxon Valdez. Research on Speech Perception, 15, 1-9.
Keil, U., Chambless, L. E., Doring, A., Filipiak, B., & Stieber, J. (1997). The relation of alcohol intake to coronary heart disease and all-cause mortality in a beer-drinking population. Epidemiology, 8(2), 150-156.
Kitamura, A., Iso, H., Sankai, T., Naito, Y., Sato, S., Kiyama, M., Okamura, T., Nakagawa, Y, Iida, M., Shimamoto, T., & Komachi, Y. (1998). Alcohol intake and premature coronary heart disease in urban Japanese men. American Journal of Epidemiology, 147(1), 59-65.
Klingholz, F., Penning, R., & Liebhardt, E. (1988). Recognition of low-level alcohol intoxication from speech signal. Journal of the Acoustical Society of America, 84, 929-935.
Knapp, S. (2006). Brain study considers motor function, cognition with alcohol consumption [Press release] retrieved from http://www.dart-mouth.edu/~news/releases/2006/04/24.html
Lee, W. H., Bae, S. G., & Bae, M. J. (2015). A Study on Improving the Overloaded Speech Waveform to Distinguish Alcohol Intoxication using Spectral Compensation. International Journal of Engineering and Technology, 7(5), 1957-1964.
Lee, W. H., & Bae, M. J. (2016). A study on drinking judgment using phonation characteristics in speech signal. The Journal of the Acoustical Society of America, 139(4), 2223-2223.
Lester, L., & Skousen, R. (1974). The phonology of drunkenness. In A. Bruck, R. Fox, & M.W. LaGaly (Eds.), Papers from the Para session on Natural Phonology (233- 239). Chicago, IL: Chicago Linguistic Society.
Levit, M., Huber, R., Batliner, A., Noeth, E. (2001). Use of prosodie speech characteristics for automated detection of alcohol intoxication. Retrieved from http://www.isca-speech.org/archive.
Mowsowitz,H., & Roth, S. (1971). Effect of alcohol on response latency in object naming. Quarterly Journal of Studies on Alcohol, 32, 969-975.
Natale, M., Kanzler, M., Jaffe, J., & Jaffe, J. (1980). Acute effects of alcohol on defensive and primary-process language: Results with three human volunteers. The International Journal of the Addictions, 15(7), 1055- 1067.
Ouellett, M. C, & Hamsberger, J. D. (2010). Estimating intoxication level from speech. Honors submitted in university of Florida. Retrieved from www.honors.ufl.edu/apps/Thesis.aspx/Download/442.
Pisoni, D. B., & Martin, C. S. (1989). Effects of alcohol on the acoustic-phonetic properties of speech: Perceptual and acoustic analyses. Alcoholism: Clinical and Experimental Research. 13 (4), 577-587. DOI: 10.1111/j.l530-0277.1989.tb00381.x
Pisoni, D. B., Hathaway, S. N., & Yuchtman, M. (1985). Effects of alcohol on the acoustic-phonetic properties of speech: Final report to GM Research Laboratories. (SRL Technical Note No. 85-03). Bloomington: Indiana University.
Rehm, J. T., Bondy, S. J., Sempos, C. T., & Vuong, C. V. (1997). Alcohol consumption and coronary heart disease morbidity and mortality. American Journal of Epidemiology, 146(6), 495-501.
Smith, B.L. (1978). Temporal aspects of English speech production: A developmental perspective. Journal of Phonetics, 6, 37-67.
Sobell, L. C., & Sobell, M. B. (1972). Effects of alcohol on the speech of alcoholics. Journal of Speech and Hearing Research, 15, 861-868.
Sobell, L. C, Sobell, M. B., & Coleman, R.F. (1982). Alcohol-induced dysflucncy in nonalcoholics. Folia Phonatrica, 34, 316-323.
Van Wyden, G. Retrieved from http://www.ehow.com/facts_5032816_effccts-alcohol-speech.htm l#ixzzlz5zSNaJt.
Gayathri Krishnan All India Institute of Speech and Hearing
Vipin Ghosh P. G. JSS Institute of Speech & Hearing
TABLE 1. Summary of Findings in Studies on Various Parameters of Speech Following Alcohol Intoxication. Behavioural measures Speech abnormalities in Jetter, (1938) intoxicated subjects Increased latency of Andrews, Cox and Smith, 1977; Dodge performance of intoxicated and Benedict, 1915; Hollingworth, participants 1923; Moskowitz and Roth, 1971; Sobell and Sobell, 1972; Sobell, Sobell and Coleman, 1982. Fluency measures Moderate to high doses Andrew, Cox and Smith, 1977; Lester of alcohol intake leads to and Skousen, 1974; Klingholz, dysfluency in speech. Penning and Liebhardt, 1988; Pisoni and Martin, 1989 Articulatory measures Misarticulations of /r/, /l/ and Chin and Pisoni, 1997; Johnson, /s/; incomplete articulations Pisoni and Bernacki, 1989 final devoicing Prosodic measures Reduced speaking rate, lower Chin and Pisoni, 1997, DeJong, [F.sub.0] and increased [F.sub.0] Hollien, Martin and Alderman, 1995; jitter, Increased duration of Johnson, Pisoni and Bernacki, 1989; segments Ouellet & Harnsberger, 2010 Acoustic measures Lowering of Fl and F2 Behne and Rivera, 1990 Increased [F.sub.0], preserved Ouellett & Harnsberger, 2010 vowel space, varied formants Fairbairn et.al., 2015. Overall volume TABLE 2. Mean (Standard Deviation) of DDK and VT Measures Pre- and Post-Alcohol Intoxication. Measures Unit of Pre-Intoxication Post-intoxication Obtained measurement DDK-/p/ DDKavp ms 284.89 (104.44) 158.17 (30.94) DDKavr s 4.29 (1.53) 6.28 (1.16) DDKcvp % 72.04 (33.54) 60.10 (36.49) DDKjit % 18.51 (9.70) 35.36 (22.85) DDKcvi % 20.52 (32.09) 5.87 (3.86) DDK-/t/ DDKavp ms 178.69 (45.48) 179.23 (25.93) DDKavr s 5.70 (1.43) 5.70 (0.87) DDKcvp % 84.19 (57.04) 55.08 (29.74) DDKjit % 47.83 (21.83) 31.57 (18.20) DDKcvi % 6.67 (4.95) 4.88 (3.17) DDK-/k/ DDKavp ms 179.03 (13.53) 175.83 (24.61) DDKavr s 5.50 (0.47) 5.72 (0.86) DDKcvp % 85.04 (36.31) 56.38 (27.50) DDKjit % 46.76 (17.88) 31.32(12.91) DDKcvi % 6.82 (3.67) 5.05 (3.37) DDK-/ptk/ DDKavp ms 156.13 (32.88) 173.07 (60.88) DDKavr s 6.57 (1.49) 6.41 (2.10) DDKcvp % 148.43 (210.47) 93.11 (44.44) DDKjit % 57.00 (14.17) 51.92 (12.70) DDKcvi % 8.88 (4.48) 8.73 (3.63) VT Measures [F.sub.0] Hz 132.14 (7.52) 133.62 (13.83) v[F.sub.0] % 1.79 (1.98) 5.82 (14.47) Fhi Hz 132.84 (14. 64) 142.46 (20.10) Flo Hz 128.84 (7.28) 128.62 (12.80) vAm % 11.67 (3.54) 10.82 (3.92) MFTR % 0.41 (0.12) 0.51 (0.20) RATR Hz 5.57 (1.99) 4.55 (1.12) Note: DDKavp=Average Period of DDK; DDKavr= Average rate ot DDK; DDKcvp= Coefficient of variation of DDK period; DDKjit= Jitter of DDK; DDKcvi= Coefficient of variation in intensity; [F.sub.0] = Fundamental frequency; v[F.sub.0] = Variation in [F.sub.0]; Fhi= Highest [F.sub.0]; Flo= Lowest [F.sub.0]; vAm= Variation in Amplitude of the signal; MFTR= Magnitude of Frequency Tremor; RATR: Rate of Amplitude Tremor)
|Printer friendly Cite/link Email Feedback|
|Author:||Krishnan, Gayathri; Ghosh, Vipin P.G.|
|Publication:||Journal of Alcohol & Drug Education|
|Date:||Aug 1, 2017|
|Previous Article:||Drumming: An Innovative Alternative for Drug Addicted Individuals.|
|Next Article:||Exploring Professional Identity Development in Alcohol and Drug Counselors in the 21st Century.|