Hydration was linked to an increase in fundamental frequency mode in conversation (p=0.037). No other measures showed a significant effect with change in hydration status. The discussion focuses on the possible explanations for these findings, the clinical implications of the findings and on recommendations for further research in this area.
Relatively few studies have been carried out to date to investigate the effects of hydration status on the voice. Those that do exist would appear to have arisen in response to the need for evaluation of clinical advice that improved hydration status will improve the voice. Studies so far differ in the subjects they use, in the methods they employ to manipulate hydration status and in the aspects of voice production that they measure. Little experimental evidence is available about the effect of improved hydration status on specific vocal parameters. The rationale behind this study, therefore, was to investigate and quantify acoustic changes in the voice with altered hydration status, that is, when mildly dehydrated and then following rehydration.
1.1 Hydration status and the body
The kidneys, under normal hormonal control, regulate water and solute excretion from the body in excess of the obligatory urine loss. Water intake occurs in the form of food and drink, with the sensation of thirst as the primary factor controlling intake. Daily fluid intake in humans is usually in excess of perceived need (Engell and Hirsch, 1991) and water balance is maintained by urinary losses. However, when water loss is greater than water gain the thirst centre in the hypothalamus is stimulated which, under normal circumstances, leads to increased fluid intake and the restoration of normal fluid volume. However, when fluids are unavailable, dehydration may occur to some degree. Usually, dehydration has already occurred to a slight extent before the sensation of thirst is noticed and even mild dehydration affects the body to some degree (Tortora and Grabowski, 1996).
1.2 Hydration status and the voice
Little is known about how the voice may be affected by conditions of dehydration. However, there is evidence that, as the body attempts to stimulate the thirst centre in the hypothalamus, saliva production is decreased, resulting in dryness of the mucosa of the mouth and pharynx (Tortora and Grabowski, 1996). As the vocal folds also consist of mucosa, which is broadly divided into the epithelium and lamina propria (Colton and Casper, 1996), it is likely that their viscosity and colour may change with alterations in total body water (TBW).
Clinicians emphasise the presence of a smooth mucosal wave on the superior surface of the vocal folds as a sign of healthy tissue and an indicator of regular vocal fold vibration, essential for good voice production (Greene and Mathieson, 1989). This ripple-like wave of the mucosa originates in the sub-glottal area and follows the contour of the vocal fold. Under examination, the horizontal excursion across the superior surface of the fold appears as a ripple or can sometimes be tracked as a light reflection. If the mucosal covering becomes more viscous, for example, as water levels decrease, then the surface of the vocal folds may become more sticky and smooth vibration could be impeded (Hiroto, 1981).
When considering how hydration status may affect the voice, it is interesting to note that people presenting in voice clinics are regularly seen with voice and throat complaints without evident pathology or indications of voice misuse (Hemler et al., 1997). In addition, they commonly report low fluid intake, especially of water and other non-caffeinated drinks (Greene and Mathieson, 1989). It has been suggested that there may be a link between low fluid intake and the occurrence of voice disorders (Lawrence, 1981; Verdolini, 1988; Sataloff, 1987). This has not been substantiated but, nevertheless, clinical advice to improve hydration status is widely given.
Vocal hygiene recommendations often include avoiding drying agents and dry environments. This is based on the assumption that these factors will dry out the mucosa, and put an increased strain on the phonatory apparatus (Sundberg, 1986). As a result, the vocal folds may be more susceptible to damage. Actors, singers and other professionals who place great demands on their voice are often advised to avoid dessicating agents such as smoke, alcohol, caffeine and certain medications and they may regularly take hydration treatments such as steam inhalations (e.g. Sataloff, 1987). These measures are widely accepted as being beneficial to the voice, but without any scientific basis.
Clearly, then, any attempt to investigate the effects of hydration status on vocal fold function will have great relevance for speech and language therapists, ENT voice specialists, voice coaches and their clientele.
1.3 Review of previous work in this area
The studies that have investigated changes in the voice with altered hydration status, although few in number, have set the framework for further study in this area.
Much of the work to date has focused on how viscosity changes in the vocal folds may affect vibration. The viscosity of pure water is 0.01 poise (P) and the viscosity of vocal fold tissue is usually higher than this. The theoretical prediction has been that hydration treatments, such as steam inhalations, reduce the viscosity of the vocal folds, by infusing them with low viscosity fluids. As a result, phonation threshold pressure (PTP) should be reduced, as should the likelihood of vocal fold oedema.
Finkelhor et al. (1988) bathed four excised canine larynges in solutions with different osmotic pressures produced by different concentrations of saline. Water transportation through the mucosa was determined by the osmotic gradient. The result was that the greatest pressures were required to produce phonation where water transportation was lowest and the lowest pressures were required where water transportation was greatest. This confirmed the relationship between PTP and vocal fold viscosity. However, the larynges used in the study were canine, not human, and they were not in vivo. Some caution should therefore be exercised when generalising the results to the normal functioning of the human larynx. Direct bathing of the vocal folds with saline may also be rather unrepresentative of how the human larynx is lubricated.
Verdolini-Marston et al. (1990) conducted a study using humans, in whom hydration status was altered through environmental control and administration of drugs. PTPs were estimated from oral pressures for subjects in a well-hydrated and in a dehydrated condition and were compared to a no-treatment control. The results were comparable with those of Finkelhor et al.(1988). PTPs were highest following dehydration and lowest following the hydration condition. PTP, therefore, appears to be related to vocal fold viscosity and is affected by hydration level. Some direct measurement of vocal fold viscosity is still required to fully substantiate this. Unfortunately, the inclusion of women and professional voice users up to age 46 in this study means that some variables were not well controlled. Firstly, changes related to the menstrual cycle are known to cause fluctuations in body water levels (Abitbol et al., 1989) which raises questions about whether all the subjects would have been equally affected by the treatments, as their body fluid levels may have been different to begin with. Secondly, the menstrual cycle is known to cause changes in the voice (Hirson and Roe, 1993) in addition to those caused by the dehydration. Thirdly, evidence exists to suggest that professional voice users may have a different fundamental frequency (Fx) range to non-professional voice users (e.g. Sulter et al., 1995) and it is possible that other vocal parameters may also be different. Consequently, professional voice users may not be very representative of the general clinical population. Fourthly, the effects of ageing may bring about changes in the voice which can appear in subjects older than about 30 years (Stoicheff, 1981; Ramig and Ringel, 1983). While such effects may not be dramatic, and may be less noticeable in professional voice users (Brown et al., 1991), they could be avoided by using subjects between 18 and 30. Finally, choosing just one method of altering body water, rather than adjusting the speaker's environment and administering drugs, would increase the likelihood of any observed changes in the voice being appropriately attributed to either the effects of systemic dehydration or to topical drying of the vocal folds. Furthermore, drugs are not guaranteed to have the same effect on all subjects and may even have additional side effects on the voice that are unrelated to dehydration.
A second study by Verdolini-Marston et al. (1994) investigated the hypothesis that hydration treatments reduce oedema-based laryngeal pathologies. Female subjects with nodules or polyps received a hydration treatment and a control treatment. Measures of phonatory effort and PTPs showed hydration treatments to be of some general benefit in the treatment of nodules and polyps. However, particular benefits were not isolated and no attempt was made to measure these benefits acoustically.
1.4 Aims of this study
The main aim of this study was to assess quantitatively whether poor hydration status affects voice production. Hydration treatments are often recommended to patients and performers and clinical advice to increase fluid intake is widely given, but, as yet, no studies have used objective measures to demonstrate how higher body water levels affect specific vocal parameters or exactly how dehydration affects voice in human subjects. It was hoped that the results of this study would reduce speculation about the vocal benefits of drinking more water by providing experimental evidence that inadequate hydration affects voice production. The second, and perhaps the most important, aim was to investigate the effect of only mild dehydration on the voice and to induce dehydration in a natural way so that the results would be truly representative of the way in which people present in clinic.
Eight subjects were recruited. All were men to avoid the changes in the voice that accompany hormonal changes in the menstrual cycle in women (Abitbol et al., 1989; Hirson and Roe, 1993). The subjects' ages ranged from 18-31 years and the average age was 20.5.
On the day of testing, three subjects revealed that they had been smoking for approximately the last five years. Although it was hoped to exclude smoking as a variable, their inclusion was unavoidable as there was no time to find other subjects. However, they did report that they only smoked between 3-5 cigarettes per day and it was decided that smokers and non-smokers would be separated in the analysis.
Subjects were required to complete a screening questionnaire to identify other variables that could bring about changes in the voice. Questions were divided into three broad areas: general health and use of medication, voice use and diet. Information on the subjects' general health was needed to establish that no medication was being taken that is known to cause dryness and that there was no evidence of past or present voice pathology or hearing loss that may have brought about changes in the voice. Regarding voice use, three subjects reported occasional shouting when playing sport but none of the subjects were professional voice users and none had had any singing or professional voice training. Questions on diet established average daily intake of alcohol and caffeine, which are thought to have both diuretic and drying effects and, when consumed in large quantities, may affect body water content (Taber's Cyclopedic Medical Dictionary, 1981). None of the subjects drank more than the recommended maximum of 3 units of alcohol (Health Education Authority Guidelines) per day. When asked about water intake, reports ranged from 0.25-2 litres per day. According to the British Nutrition Foundation (1997), average daily intake for an adult of medium build and moderate activity level should be at least two litres per day. Half of the subjects considered that they did drink enough water and half reported that they often felt thirsty.
No subjects needed to be eliminated from the study following the screening questionnaire.
All subjects were naive to the experimental hypothesis and were in good health on the day of testing.
2.2 Manipulation of hydration status
Hydration status was manipulated by controlling food and fluid intake alone. No drugs were prescribed to avoid possible side effects on voice. No alterations were made to environmental conditions as these may have caused different systemic changes in addition to those caused by fasting as well as some topical drying of the vocal folds.
On the day before testing, subjects were instructed not to drink alcohol or caffeinated drinks that may cause additional drying of the vocal folds. Subjects therefore had nothing to drink after 4 p.m. and nothing to eat after 6 p.m., which meant that the period over which they were dehydrating lasted approximately 18 hours. On the day of testing, therefore, subjects had nil by mouth prior to the recordings. On the test day, subjects completed a food and fluid intake sheet for the previous 24 hours which showed that no-one had consumed excessive quantities of foods with a high water content, such as fruit or salads.
The environment in the waiting and recording rooms was not controlled and the temperature registered at 20°C. The air conditioning was turned off to avoid any possible topical drying effect.
In order to achieve rehydration, subjects drank two litres of fluid. It is now well documented that the replacement of electrolytes, as well as water, is essential for effective rehydration (Costill and Sparks, 1973; Gonzales-Alonso et al., 1992; Nose et al., 1988). Sodium and potassium enhance the retention of water in the extracellular and intracellular spaces respectively. Subjects were given an electrolytic drink containing these electrolytes, therefore, so that their electrolyte balance was restored at the earliest opportunity. In an eighteen hour period of fasting, body water is only expected to decrease by approximately one litre, as the body appears to have effective regulatory mechanisms to cope with dehydration (Tortora and Grabowski, 1996). In order to ensure full rehydration, therefore, subjects drank twice the amount that they were expected to lose overnight.
2.3 Measurement of hydration status
Bioelectrical impedance analysis (e.g. Heyward and Stolarczyk, 1996) was employed during pilot studies and on the day of testing to measure hydration status, but the results obtained were inconsistent. This technique did not appear to be able to track small changes in TBW and the data were discarded. No further reference will be made therefore to the inclusion of this technique in our study.
2.4 Measurement of voice changes
Listeners' judgements of voice quality and voice pitch (the perceptual correlate of fundamental frequency) can be unreliable (e.g. Montague et al., 1978) and therefore quantitative acoustic analysis of vocal parameters was preferred as the initial method of analysis in this study.
Acoustic analysis in this study was carried out using the Laryngograph (Fourcin and Abberton, 1971) and the accompanying PCLX software (Fourcin et al., 1995). Two simultaneous speech and Lx waveform recordings (one when dehydrated and one when rehydrated) were made for each subject. The recordings were carried out in a sound-proof room using a B&K condenser microphone (type 2231) positioned 22cm away from the subject's mouth, level with the mouth at an angle of 45° from the mid-line. The speech signal was recorded onto a Sony 60 ES digital audio tape (DAT) recorder. The Laryngograph Processor detected the changes in impedance across the glottis during voicing with a pair of gold-plated electrodes placed on the alae of the thyroid cartilage. The output from the Laryngograph was recorded onto the other channel of the DAT recorder.
Direct observation of the vocal folds was achieved using the Laryngostrobe (Fourcin et al., 1995), which incorporates the techniques of stroboscopy and Laryngography. The PC-based system uses a PCLX card, a Machida 70° rigid endoscope (type LY-C30 62B), a Panasonic camera head (type GP-KS162HDE) and Panasonic camera power supply unit (type GP-KS162), a Laryngograph Processor and a Laryngostrobe (prototype 1). The rigid endoscope was used to view the vocal folds and the Laryngograph electrodes were placed on the thyroid cartilage as normal. On phonation, the Laryngograph supplied the PC with a measure of the larynx periods (Tx) during voicing and the strobe was automatically triggered to flash at 1 or 2 Hz less than this. Eight images of the vocal folds were sequentially captured at equal and progressive intervals through the vibratory cycle. A simultaneous recording of Lx was made and the points on the waveform that corresponded to each image were marked. The eight images could then be viewed as a continuous "movie" loop. The fundamental frequency of vocal fold vibration for a particular voice sample was also displayed. For an unobstructed view of the larynx, subjects were asked to lean forward and tilt their head back slightly. The tongue was pulled forward by the ENT consultant with a piece of dental gauze, to prevent the base of the tongue impeding the view of the larynx and to stop subjects attempting to swallow. Subjects were asked to phonate on [i], which causes the base of the tongue to move forward in the mouth. The examination was repeated until an image fidelity score of 90% or above was achieved. No anaesthetic spray was used as all subjects were able to tolerate the rigid endoscope.
2.5 Test materials for the acoustic analyses
The vowels [i] and were elicited from each subject. These vowels were chosen because their production encourages the tongue to take up an extreme position [i] and the most neutral position in the oral cavity. The rationale was that changes brought about by altered hydration status might be most evident when the vocal apparatus was taxed during phonation. For the same reason the speakers were asked to produce the vowels at a high as well as a comfortable fundamental frequency. Each token was held for approximately five seconds to allow for analysis of the mid-portion of the vowel and was repeated to allow a comparison across different productions of the same vowel.
2.5.2 Reading passage
A reading passage was also recorded (The Natural World - Barry et al., 1990). This text was chosen as it contains all the sounds of Standard Southern British English in a number of phonetic contexts. The passage takes approximately two minutes to read which, according to Horii (1975), makes it more likely to give reliable estimates of voice fundamental frequency than a shorter text.
Conversation was included in the test materials as measures of voice fundamental frequency are known to be different in an informal task as opposed to a more formal task such as reading (e.g. Snidecor, 1943). Subjects responded to standard questions (about sports or holiday plans) for up to two minutes so that a sample of informal speech was obtained for analysis.
2.6 Experimental procedure
Each subject was led through the following procedure which took approximately two hours:
Between recordings subjects were shown to a waiting room where excessive talking was discouraged to avoid vocal fatigue, which can lead to naturally more irregular phonation (Scherer et al., 1987).
2.7 Data analysis
2.7.1 Speech and Laryngographic data
The speech and Laryngograph recordings were acquired on to a PC using the PCLX software. The glottal periods (Tx) are supplied to the PCLX interface card by the Laryngograph Processor. The software allocates each Tx value to one of 128 frequency "bins" (covering 30-1000Hz on a log scale) to build a simple histogram of the data (Dx1). Alternatively, only consecutive Tx values that fall into the same frequency bin can be included in a modified histogram (Dx2). This excludes irregular phonation from the analysis, giving a measure of the frequency range involved in regular phonation. A third analysis produces a cross-plot (Cx) which is constructed by treating consecutive Tx values as the x and y co-ordinates on a scatter plot. Regular phonation will therefore fall along the x-y diagonal or very close to it. The Cx plot can provide an indication of the amount of irregular phonation in the speech sample.
From these three analyses, measures of central tendency and dispersion of Tx can be obtained. The measures chosen for this study were modal Fx and Fx range (90%) (both taken from Dx2) and an estimate of the percentage of irregular vocal fold vibration in the sample (percent irregularity taken from Cx). Modal Fx was chosen as the measure of central tendency as it is most directly related to what, perceptually, is often called the "habitual pitch" of the speaker (Baken, 1987). The 90% Fx range was chosen as it excludes the outlying 5% of the distribution at both ends of the frequency range. The estimate of irregularity calculated the percentage of the sample that fell more than one position away from the diagonal on the Cx plot. All the Laryngograph analyses were performed on both reading and conversation.
Sustained vowels were used for calculating frequency (period) perturbation measures, that is, the cycle by cycle variations in Fx more commonly known as jitter. During speech, many voluntary changes in Fx occur to achieve appropriate stress and intonation patterns. Measures of jitter attempt to quantify the frequency variability in Fx not accounted for by these voluntary changes. As Baken (1987) states "..even during sustained vowels, (fundamental frequency) may undergo slow changes that are not of interest but that may inflate the measure of jitter." A measure implemented in the latest version of the Laryngograph analysis software is relative average perturbation or RAP (Koike, 1973). RAP minimises the effects of these relatively slow, volitional changes in Fx by averaging across a triplet of successive period values to provide an estimate of the value of the middle period had there been no perturbation present (a "corrected" period value). The jitter for a given cycle is then taken to be the difference between the corrected period and the actual period. RAP measures were made on the mid-portion (3 seconds) of each vowel so that extra irregularity associated with the onset and offset of phonation was not included. A measure was taken for each token of a vowel and the measures were averaged across repeated tokens.
2.7.2 Laryngostrobe data
The primary purpose of including the Laryngostrobe analysis was to screen out subjects with laryngeal pathology. However, one of the experimenters also looked informally at the data for differences in the appearance and vibratory pattern of the vocal folds with de- and rehydration. The tentative findings are presented here.
3.1 Laryngostrobe data
On examination, seven of eight initial subjects, were classed as having a normal larynx. Subject 2 was found to have a unilateral sulcus vocalis. It is unclear whether this condition is congenital or arises as a result of vocal misuse (Hirano and Bless, 1993) and, as there is an accompanying hoarse voice quality, the data obtained from this subject will not be considered further here.
The informal visual examination of each subject when dehydrated and then rehydrated did not highlight any systematic differences. It appeared as though there were some changes in the consistency and amount of mucus present and in the propagation of the mucosal wave, but no systematic trend was observed. More importantly, comments made by an informed observer are less likely to be objective.
The Laryngostrobe data was also used to ascertain whether there was a visible difference in the vocal folds of smokers versus non-smokers, as has been reported in previous studies of these two groups (e.g. Gilbert and Weismer, 1974, found 87% of their subjects had visible laryngeal pathology linked to smoking). However, in this study, systematic differences in the mucus, the obvious presence of blood vessels, the amplitude of vibration and the closure patterns of the folds were not observed. Subject 3, a smoker, did show considerably more mucus than any other subject but, as this feature was not apparent in the other two smokers, it may have been an idiosyncracy. In spite of the lack of visible signs of smoking (perhaps due to the very small number of cigarettes reportedly consumed) the unequivocal findings of earlier studies led us to treat the data from the smokers separately from that of the non-smokers in our statistical analyses.
3.2 Fundamental frequency data
|Subject 1 (S)|
|Subject 3 (S)|
|Subject 4 (S)|
Table 1: PCLX data for reading and conversation tasks
The fundamental frequency data for all subjects in all conditions reading a text and in informal conversation is presented in Table 1.
A General Linear Model (GLM) repeated measures ANOVA was computed for measures of fundamental frequency mode, range, irregularity and jitter, to establish whether hydration status had a significant effect on these measures. Smoking was incorporated as an additional factor to determine whether any difference in the data due to smoking was at all significant.
For 6 of the 7 subjects the mode increased, but only very slightly, while for 1 subject it decreased (Figure 1(a)). The GLM ANOVA revealed that hydration status for measures of modal frequency in the reading task was not significant (F(1,5)=1.181, p=0.327). There was also no significant within-subject interaction between hydration and smoking (F(1,5)=0.002, p=0.967). Between-subject variability due to smoking was not significant (F(1,5)=2.657, p=0.164).
Figure 1(a) : Changes in modal frequency with rehydration in a reading task
In conversation, as in reading, 6 of the 7 subjects showed an increased modal frequency with rehydration, while one subject showed no change. This change was significant (F(1,5)=7.960, p=0.037) (Figure 1(b)). There was again no significant interaction between hydration and smoking within subjects for conversation (F(1,5)=0.143, p=0.720). Between-subject variability due to smoking was again non-significant (F(1,5)=1.723, p=0.246).
Figure 1(b): Changes in modal frequency with rehydration in conversation
Our measurements (Table 1) showed that there was no consistent change in frequency range for either reading or conversation.
No consistent effects in irregularity were seen. The GLM repeated measures ANOVA revealed that hydration status had no significant effect on irregularity in the reading task (F(1,5)=1.453, p=0.282). However, as seen in Figure 2(a), there was some interaction between hydration and smoking, although not a statistically significant one (F(1,5)=4.54, p=0.086). It appears that the degree of irregularity varied considerably between subjects in the non-smoking group when dehydrated, but that this variation was less after rehydration. Showing the opposite trend, the smokers had slightly more variation between subjects in the rehydrated condition. The possible reasons for this are discussed below.
Figure 2(a): Changes in percent irregularity with rehydration in a reading task
In conversation, neither hydration (F(1,5)=0.047, p=0.838) nor the interaction between hydration and smoking (F(1,5)=3.679, p=0.110) reached significance. Figure 2(b) shows that, once again, the degree of irregularity varied considerably between subjects in the non-smoking group when dehydrated, but that this variation was rather less after rehydration. Again, showing the opposite trend, the smokers had more variation between subjects in the rehydrated condition. Finally, between subjects, smoking was again a non-significant factor (F(1,5)=2.335, p=0.187).
Figure 2(b) : Changes in percent irregularity with rehydration in conversation
3.6 Jitter (Relative Average Perturbation - RAP)
The same statistical test was applied to the RAP measures, taken from the mid-portion of the sustained vowels and [i] at comfortable and high Fx. RAP was expected to be higher in dehydrated subjects but, without exception, no significant hydration effect was found (for example, F=0.000, p=0.991 for at a high fundamental frequency - Figure 3).
Figure 3: Changes in Relative Average Perturbation (RAP) with rehydration for at a high fundamental frequency
Similarly, no significant interaction was found between hydration and smoking within subjects and there was no significant variability between subjects due to smoking. This was true for all tokens. The possible reasons for this are discussed below.
3.7 Summary of results
- fundamental frequency range in reading and conversation|
- vocal fold irregularity in reading and conversation
- jitter (RAP) in sustained vowels
4. Discussion and conclusions
The results obtained from this study are inconclusive but they indicate that hydration status, as it was manipulated in this experiment, does not have a particularly marked influence on fundamental frequency mode, range and regularity. Possible explanations for this are suggested below, but first it may be useful to consider how these results may have been complicated by an important issue.
The inclusion of three smokers in the subject group was considered to be preferable to reducing the sample size to just four people. Indeed, we found no significant main effects or interactions with smoking. There were, however, non-significant trends showing non-smokers' irregularity decreasing with rehydration, while smokers' irregularity increased. The combination of smoking and altered hydration status may create the opposite effect from altered hydration alone. This is an interesting issue, but the aim of this experiment was to examine the effects of dehydration on the voice when all other variables had been controlled for. It would be tempting to make hypotheses about why the reverse effect was found in smokers, but the sample size is really too small to justify this and not enough background information about smoking was gathered, as it fell outside the central issue of the study. This, then, requires further study.
The fact that hydration status only had a significant effect in increasing modal frequency in conversation requires careful consideration. One important issue that has some bearing on this is the choice of materials used to collect the speech samples. Clearly, this has to be considered, as different effects were found for the mode in reading and in conversation. The issue revolves around whether reading or conversation provides data that is more representative of other speaking situations. This has long been the subject of debate (for a review, see Baken, 1987) and it is generally argued that in a reading task, people adopt a more formal style that can produce a more restricted frequency range. Conversation, on the other hand, is thought to be more representative of the speaker's normal frequency range and is generally more informal, but the content of the sample is usually unrepeatable. It may be the case here that the formal reading task, in which subjects seem to perform more, masked any change in the modal frequency, but that this showed up in conversation because subjects talked in a more relaxed and informal way. Therefore, any changes which may affect the vocal apparatus, such as altered hydration, show up less in a reading task, where subjects make more of an effort to perform well and overcome any changes in their voice. Small changes brought about by dehydration may be more likely to show up in conversation, as subjects may also become distracted by the topic being discussed and feel more at ease. Consequently, they may make less of an effort to perform and a more accurate representation of their voice status is obtained. This idea could be tested out by using other test materials, such as picture description, which would encourage the speaker to adopt a more formal style. The subject would then be repeatable also.
The lack of effect in the jitter analyses is not altogether surprising. RAP and other jitter measures routinely take measurements from the mid-portion of a sustained vowel (see Baken, 1987), which eliminates any irregularity associated with the onset and offset of phonation. Our measures were therefore taken at a point when vocal fold vibration was likely to be at its most regular. Different vowels were produced on a comfortable fundamental frequency and then on a high fundamental frequency, with the reasoning that the effects of dehydration would be more evident when the larynx is forced to work harder. However, as jitter measures require that readings are taken from the most reliably produced part of the vowel, any interesting differences are probably ignored. This is a problem of the measure itself and not just the experiment.
In normal voice production, the vocal folds generally vibrate in a smooth and regular fashion. This is achieved because the vocal apparatus is kept well-lubricated and because the vocal folds are covered with mucus (Fourcin, 1981). A small degree of irregularity is often found in normal voice production, particularly at lower frequencies, associated with the offset of phonation when airflow through the larynx is reduced (often called creaky voice). When the body's hydration status is lowered, the mucosal covering might be expected to become dry and sticky, increasing viscosity and causing the vocal folds to vibrate in a less regular manner. Acoustic analysis of a poorly hydrated voice was expected, therefore, to detect an increase in irregularity, that is, a greater count of consecutive glottal periods of different lengths. Rehydration was expected to cause viscosity to be reduced again, so that the vocal folds resume regular function causing irregularity to fall. That more dramatic changes in irregularity were not seen is surprising and the following explanation is offered for this.
During the course of the recordings it became obvious that, if subjects were experiencing any difficulty at all, it was in initiating phonation. Analysis of the onsets and offsets of phonation may well be more revealing about the effects of dehydration on the voice, which would correlate well with the findings of Verdolini-Marston et al. (1990), in which greater subglottal pressure was required for phonation when the vocal folds were dry. Alterations in hydration status may affect the first few cycles of vibration, but it is possible that once phonation has been initiated and the vocal folds are set in motion, the vibratory pattern is relatively normal. If this is the case, no large changes are likely to occur in irregularity as currently measured. It is therefore possible that significant changes in the voice did occur with de- and rehydration, but that other measures such as analysis of the Lx waveform would be necessary to detect them. It might be interesting to see if listeners can detect changes in voice quality arising from changes in vocal parameters not studied here. A perceptual analysis of the data may therefore reveal which parameters should be monitored in future experiments of this type. It may also be useful to systematically collect the speaker's own impressions of what features of their voice have changed under the two experimental conditions.
Secondly, it may be possible that the subjects were not in fact very dehydrated, although every effort was made to establish this quantitatively. It was expected that subjects' TBW would decrease by one or two litres during the eighteen hour period of fasting. However, this reduction in TBW does not necessarily mean that body fluids have fallen below the level at which a subject is well-hydrated. They may have had less body water in the dehydrated condition, but they may still have been operating with enough ECF and TBW to maintain healthy cells and tissues. There would be no reason to expect changes in the voice, therefore. Furthermore, a period of eighteen hours of fasting may be significant in altering hydration status for one subject, if he habitually drinks a lot, but not for another, whose fluid intake may generally be low anyway. This point is not so crucial to this study, as each subject acted as their own control, but it does emphasise how unrealistic it is to suggest that all human beings function at the same level of hydration and that an attempt to change hydration status will have the same effect on all people. It was certainly true that some of the subjects were not used to drinking a lot and did not appear to feel worse for having abstained from food and water for 18 hours. However, as there was a significant change in the modal frequency, it may be more likely that subjects were dehydrated to some degree, but that greater dehydration is required to affect the mode more significantly and to affect other vocal parameters.
A third possibility is that subjects were mildly dehydrated, but that systemic dehydration does not significantly affect voice fundamental frequency and regularity. It may well be the case that when TBW is reduced, the larynx is maintained well-hydrated because it is situated at a critical point in the airway. In this sense, its healthy function is essential to life. If this is true, then mechanisms must be in place to ensure that the larynx and vocal fold mucosa continue to be well-lubricated during mild dehydration, possibly at the expense of other body systems. In this scenario, no changes would occur in the viscosity of the vocal folds, vibration would be regular and smooth and propagation of the mucosal wave would be as normal. Topical drying of the vocal folds may well have produced different results. As this is often how many people present in voice clinic, that is, with mild systemic dehydration resulting from low fluid intake, this explanation has important implications for the advice given by many clinicians. If this method and degree of dehydration has little or no effect on the voice, then encouraging people to improve their hydration level may not be justified in terms of improving voice quality, although it may be justified in terms of general health.
In view of the fact that the human body functions in a very complex way and that little information exists on exactly which changes occur with dehydration and how the voice may be affected, all of the above explanations seem plausible. Their validity can only be determined experimentally. This study, therefore, has certainly acted as an interesting precursor to future work.
The authors would like to thank Julian McGlashan for his time and expertise with the Laryngostrobe system, and for his help in interpreting the data gathered with the equipment. Thanks are also due to Laryngograph Ltd for the loan of the Laryngostrobe and to Andrew Faulkner for his helpful comments on a draft of this paper.
Abitbol, J., de Brux, J., Millot, G., Masson, M-F., Mimoun, O.L., Pau, H., Abitbol, B. (1989) Does a hormonal vocal cord cycle exist in women? Study of vocal premenstrual syndrome in voice performers by videostroboscopy-glottography and cytology on 38 women. Journal of Voice 3, 157-162.
Baken, R.J. (1987) Clinical Measurement of speech and Voice. London: Taylor and Francis.
Ball, V., Faulkner, A., Fourcin, A.J. (1990) The effects of two different speech coding strategies on voice fundamental frequency control in deafened adults. British Journal of Audiology 24, 393-409.
Barry, W.J., Goldsmith, M., Fourcin, A.J., Fuller, H. (1990) Stability of laryngeal measures in speech. Alvey Project MMI/132 Speech Technology Assessment.
Colton, R.H., Casper, J.K. (1996) Understanding Voice Problems - A Physiological Perspective For Diagnosis and Treatment. Williams and Wilkins.
Costill, D.L., Sparks, K.E. (1973) Rapid fluid replacement following thermal dehydration. Journal of Applied Physiology 34, 299-303.
Engell, D., Hirsch, E. (1991) Environmental and sensory modulation of fluid intake in humans. In Ramsey, D. J., Booth, D. A. (eds.) Thirst: Physiological and Psychological Aspects. ILSI Human Nutrition Reviews. London: Springer-Verlag.
Finkelhor, B.K., Titze, I.R., Durham, P.L. (1988) The effect of viscosity changes in the vocal folds on the range of oscillation. Journal of Voice 1, 320-325.
Fourcin, A.J., Abberton, E. (1971) First applications of a new laryngograph. Medical and Biological Illustration 21, 172-182.
Fourcin, A.J., Abberton, E., Miller, D., Howells D. (1995) Laryngograph: speech pattern element tools for therapy, training and assessment. European Journal of Disorders of Communication 30, 101-115.
Gilbert, H.R., Weismer, G.G. (1974) The effects of smoking on the speaking fundamental frequency of adult women. Journal of Psycholinguistic Research 3, 225-231.
Gonzales-Alonso , J., Heaps, C.L., Coyle, E.F. (1992) Rehydration after exercise with common beverages and water. International Journal of Sports Medicine 13, 399-406.
Greene, M.C.L., Mathieson, L. (1989) The Voice and Its Disorders. 5th edition. Whurr.
Heyward, V.H., Stolarczyk, L.M. (1996) Applied Body Composition Assessment. Human Kinetics.
Hirano, M., Bless, D.M. (1993) Videostroboscopic Examination of the Larynx. San Diego; Singular Publishing Group.
Hiroto, I. (1981) Introductory Remarks. In Stevens, K.N., Hirano, M. (eds.) Vocal fold Physiology. Tokyo: University of Tokyo Press.
Hirson, A., Roe, S. (1993) Stability of voice and periodic fluctuations in voice quality through the menstrual cycle. Voice 2, 78-88.
Horii, Y. (1975) Some statistical characteristics of voice fundamental frequency. JSHR 18, 192-201.
Koike, Y. (1973) Application of some acoustic measures for the evaluation of laryngeal dysfunction. Studia Phonologica 7, 17-23.
Lawrence, V.L. (1981) Handy household hints: to sing or not to sing. NATS Bulletin 37, 23-25.
Montague, J.C., Jr., Hollien, H., Hollien, P.A., Wold, D.C. (1978) Perceived pitch and fundamental frequency comparisons of institutionalised Down's syndrome children. Folia Phoniatrica 30, 245-256.
Nose, H., Mack, G.W., Shi, X., Nadel, E.R. (1988) Role of osmolality and plasma volume during rehydration in humans. Journal of Applied Physiology 65, 325-331.
Ramig, L.A., Ringel, R.L. (1983) Effects of physiological aging on selected acoustic characteristics of voice. JSHR 9, 353-360.
Sataloff, R.T. (1987) The professional voice: part I. Anatomy, function and general health. Journal of Voice 1, 92-104.
Sataloff, R.T. (1987) The professional voice: part III. Common diagnoses and treatments. Journal of Voice 1, 283-292
Scherer, R.C., Titze, I.R., Raphael, B.N., Wood, R.P., Ramig, L.A., Blager, R.F. (1987) Vocal fatigue in a trained and untrained voice user. In Baer, T., Sasaki, C., Harris, K.S. (eds.) Laryngeal Function in Phonation and Respiration. College-Hill: Boston/Toronto/San Diego.
Snidecor, J.C., (1943) A comparative study of the pitch and duration characteristics of impromptu speaking and oral reading Speech Monographs 10, 50-57
Stoicheff, M.L. (1981) Speaking fundamental frequency characteristics of non-smoking female adults. JSHR 24, 437-441.
Sulter, A.M., Schutte, H.K., Miller, D.G. (1995) Differences in phonetogram features between male and female subjects with and without vocal training. Journal of Voice 9, 363-377.
Taber's Cyclopedic Medical Dictionary (1981). 14th edition. Thomas, C.L. (ed.). Davis.
Tortora, G.J., Grabowski, S.R. (1996) Principles of Anatomy and Physiology. 8th edition. Harper Collins.
Verdolini, K. (1988) Practice good health and prevent those voice disorders. Choristers Guild LETTERS, XL, 40-44.
Verdolini-Marston, K., Titze, I.R., Druker, D.G. (1990) Changes in phonation threshold pressure with induced conditions of hydration. Journal of Voice 4, 142-151.
Verdolini-Marston, K., Sandage, M., Titze, I.R. (1994) Effect of hydration treatments on laryngeal nodules and polyps and related voice measures. Journal of Voice 8, 30-47.
© Julia Selby and Ginny Wilson.