3.1 Introduction

The speech of the hearing impaired has several synonyms in literature: speech of the deaf, speech of the hard-of-hearing, dysenia, peripheral impressive disordered of language development, deaf-mutism, and audiogenic dyslalia.

There are several characteristics typical of the speech of the hearing impaired which serve to separate it from the speech of normal children. Numerous studies have been carried out to investigate the speech of the hearing impaired (Hudgins, 1934; Hudgins and Numbers, 1942; Carr, 1953; Calvert, 1962; John and Howarth, 1965; Fry, 1966; Markides, 1970; West and Weber, 1973; Oller and Kelly, 1974; Levitt, Smith and Stromberg, 1976; Smith, 1975; Monsen, 1974, 1976a,b,c, 1978; Sussman and Hernandez, 1979; Geffner, 1980; Levitt, Stromberg, et al., 1980). These studies have identified many recurring speech deviancies, thus indicating that the hearing impaired speech has a nature of its own.

Hudgins (1934), and Hudgins and Numbers (1942) were the pioneers presenting thorough investigations the characteristics of the speech of deaf children. Hudgins’ 1934 study revealed five particular characteristics of deaf speech: (i) extremely slow, laboured, and excessively breathy speech; (ii) prolonged production of vowels, resulting in either distortion or the creation of a new syllable; (iii) a tendency to devoice stops in all positions; (iv) excessive use of nasality with vowels and consonants; and (v) abnormal rhythm across utterance.

Further insights into the speech of the deaf were obtained from the study made by Hudgins and Numbers (1942). They collected the recordings of 1200 sentences from deaf pupils between 9 and 12 years. They found two general types of errors; errors involving both consonants and vowels, and errors of rhythm. The consonant errors were classified into seven general types as follows: (i) failure of distinguish between voiced and unvoiced consonants; (ii) consonant substitutions; (iii) excessive nasality; (iv) misarticulation of consonant clusters; and (v) omission of releasing consonants. The vowel errors are vowel substitutions; misarticulation of diphthongization of vowels; neutralization of vowels; and nasalization of vowels. Note that several of these are found in normal hearing children acquiring language. Several of these are found also in other speech disorders.

Nickerson (1975) has discussed the speech characteristics of the deaf under five topics: (i) timing and rhythm, (ii) pitch and intonation, (iii) velar control, (iv) articulation, and voice quality. Overall speech intelligibility, and supra-segmental errors of the deaf were discussed in a recent article by Gold (1980).

Nickerson (1975) cautioned that no two individuals whether they have hearing or not, produce speech that is exactly the same. Consequently, and description of the speech of a group is bound to fail in some respects to describe the speech of any given member in the group. This observation, of course, becomes true especially for the hearing impaired because a number of factors interact to influence the speech development in these children. Their speech development is dependent upon the age of onset of hearing loss, severity and type of hearing impairment, the age at which the hearing loss was detected and the age at which guidance was sought for from the speech therapist and audiologist, parent’s co-operation for following the therapist’s advice, the amount of speech and language stimulation given to the child, sibling status, whether the child attends a normal or deaf school, whether their parents are hearing impaired or normal, factor of lip-reading, audio-visual speech reception for speech development, self-monitoring and the type of speech and language training implemented. In addition, if the child has multiple handicaps such as blindness, other physical deformities, middle ear pathology, sensorineural problems, or mental retardation, the speech development is further retarded. But most of the studies have shown that on the average the speech of the individuals with profound hearing impairment that dates from birth or shortly thereafter tends to differ in many ways from the speech of the normal children (Nickerson, 1975).

In this chapter an attempt is made to review and discuss the speech of the hearing impaired:

- Common approaches used to obtain speech samples of the hearing impaired      and their analysis of the speech samples;
- Segmental errors;
- Voice aspects;
- Supra-segmental errors; and
- Overall speech intelligibility in the hearing impaired.

3.2. Methods of Collection and Analysis of Speech Samples of the Hearing Impaired

The present investigators find that almost all the studies analyzing the speech samples of the hearing impaired have used a cross-sectional approach. Longitudinal studies of the articulatory, or the phonological development of a single hearing-impaired child or of a group are not found. The spontaneous speech samples have been collected through picture naming tasks. Imitation of the spoken word has been tested by standardized articulation tests, screening tests of articulation and photo articulation tests. Tests meant for normal children have been modified to suit the conditions of the hearing impaired. For instance, Geffner (1980) has adapted the Goldman-Fristoe test of articulation by taking out the words not occurring in the known vocabulary of 6-year old deaf children and supplanting it with other words known to exist in the child’s vocabulary. While most of the tests consist of individual vocabulary items, some tests have employed sentences as well. For instance, Smith (1975) has specially constructed 20 sentences for the analysis of deaf children’s speech. These sentences consist only of the vocabulary expected within the vocabulary of the deaf child of 8 years. These sentences are of different lengths and cover a range of syntactic constructions. Places of articulation employed are analyzed and sentences are so constructed that one transition from each place of articulation is at least ensured. There is also at least one transition from each of the vowels to each place of articulation. All consonants are represented in all the positions initial, medial and final positions, in consonance with such admissibilities in the language. No sound of interest is allowed to appear in an unstressed syllable and no sound of interest is allowed to appear in the initial or final word of sentence.

The analysis of the samples has been done through phonetic and phonemic transcription of the samples. In addition to perceptual judgements, voice onset time aspects, formant frequency analysis, analysis of oral air flow characteristics during the production of consonants, and the analysis of articulatory movements during speech production have also been made. The perceptual and acoustic analysis of prosodic errors has also been made. Thus, both perceptual judgements and objective analysis have been used to evaluate the speech of the hearing impaired.

Most of the investigators have attempted to find out the significance of the errors to the overall speech intelligibility in the hearing impaired.

In our opinion, the concept of “error” is, indeed, a tricky one. An item is conceived as an error in terms of the normal language. It may also be considered as an error if there is deviancy from the linguistic behaviour generally observed in the hearing impaired population. It may also be considered as an error if the hearing impaired subject uses one item regularly, but fails to do so on occasions. In this last case the errors may be systematic or non-systematic.

With sparse language, the hearing impaired’s utterances have been subjected mainly to an analysis of segmental sounds. Sequential occurrence of the segmental sounds has not been given much attention. Lexical choice, sequential occurrence of lexical items in a sentence, use of grammatical elements, etc., have not received much attention. There is a vast area of research waiting for enterprising linguists to explore. And yet research in this area will not be fruitful unless the concept of error is changed in consonance with the language of the hearing impaired – in consonance with retrieval modalities and methods of the hearing impaired. Omission, addition, substitution and distortion are not errors in the language of the hearing impaired if a consonant pattern is maintained and if there is a consistent association of objects, events, etc., with the above four categories – in other words, the sparse and special nature of the language of the hearing impaired demands a different approach from the ones we generally adopt in the identification of errors in the language of the normals. Elison and/or holding on to one cue exclusively is an integral part of the communicative mode of the hearing impaired and this is so because of inherent constraints. Such items form regular paterms. And hence there is a need to identify, first of all, the patterns of linguistic behaviour of the hearing impaired with reference to his/her own utterances. Secondly, there is a need to arrive at a general picture of the linguistic behaviour of a group of hearing impaired individuals of similar age, linguistic background, type of hearing loss, etc. Thirdly, one should aim at a description of the deviancies in the individual hearing impaired in terms of his/her own linguistic behaviour and in terms of the linguistic behaviour of the group of which he/she is a member. Only then an insightful picture and analysis of the “errors” of the hearing impaired will become available.

3.3. Errors of Segmental Sounds

The errors in segmental sounds produced by the hearing impaired are discussed here under the consonantal errors, vowel errors, and errors involving diphthogs and consonant-vowel and vowel-consonant transitions. In General, many studies (such as Geffner, 1980) have found that vowels are produced correctly more often than the consonants. Many explanations have been offered for the higher frequency of errors reported for the consonants. One explanation is that since more consonants than vowels are generally produced in speech, there is greater likelihood of there being more errors of consonants than of vowels (Gold, 1980). However, note that the generalization that more consonants than vowels are produced may not be found valid for all contexts of the language of the hearing impaired. Another explanation given is that, since vowels carry more energy, they are easier to hear and are, therefore, produced correctly more often. A third explanation given is that vowels require less difficult tongue adjustments. A fourth explanation given revolves around the finding that the normal ear accepts a wide range of vowel distortion and that, as a consequence the experimenters are less critical in determining the degree of vowel accuracy. A fifth explanation is an stated by Monsen (1976) that the vowels, in general, convey to the listener more consonantal cues than the vocalic cues conveyed by the consonants. Therefore, he said that it is reasonable that a reduction in effects of transition from one speech sound to the next observed in the hearing impaired children results in fewer consonants being correctly produced than vowels. In other words, no hearing impaired subject demonstrates the use of all the phonemes – vowel and consonants – found in the inventory of the language of the normal hearing. We find in our study, however, that of the available consonants and vowels, the hearing impaired subjects did produce more consonants correctly than the vowels. The incorrect production of vowels was seen to be the result of excessive nasality which coloured the quality of vowels produced in many subjects. It was also seen to be the result of confusion between the heights of vowels. We found that while the part of the tongue and lip rounding were correctly handled, the height of the tongue for the vowels could not be correctly handled by the hearing impaired subjects. In our own investigation, we find that, only in the case of the partial hearing, consonants are produced more in number than the vowels. In the case of the profoundly impaired subjects, predominance of the consonants in terms of number cannot be established. In all the cases, the number of consonants was equal to that of the vowels, or in one case it was even less. Secondly, a more notable feature is that neither in the profoundly impaired nor in the partially hearing was the list of consonants or the vowels complete.

Scholars have classified the articulatory errors into addition, substitution, omission, and distortion. Many scholars have suggested that omission in the most frequent type of error in severely hearing impaired children. Several studies (such as those of Geffner, 1980; Markides, 1970; and Morely, 1972) suggest that omission errors are more frequent in the deaf and that substitution errors are more among the partially hearing. Markides (1970) reported that the omission errors are more numerous followed by substitution and distortion. Amongst the partially hearing, substitution errors are more numerous, followed by omissions and, finally, distortions. It may be noted here that omissions also form a major part in the language acquisition of normal children. In normal language acquisition while all the four types of errors indicate in some manner and degree the types of deviancies, the process of omission reflects more insufficient acquisition of language, whereas the other three indicate more the efforts of corrections towards approximations of normal language. The fact that errors of omission are found more in the language of the profoundly hearing impaired only shows the insufficient inventory of items in the concerned language. It is also likely that an inadequate/insufficient inventory will be the chief feature of the language of the profoundly hearing impaired, whereas problems of proper use of the inventory would dominate the language of the partially hearing impaired. This fact was borne out clearly in language data of our subjects of investigation. For example, all the three profoundly hearing impaired subjects, irrespective of their varying age groups, showed an utter insufficiency in terms of the number of phonemes, both vowels and consonants. In addition to this paucity of number of phonemes, they also showed a failure to acquire phonemes, both vowels and consonants, under the various categories of point and manner of articulation. In other words, a subject not only failed to master consonants of a particular category such as stop but he failed also to master some whole categories, such as the series of fricatives in Kannada. While this was the picture presented by the profoundly hearing impaired, the partially hearing impaired acquired most of the phonemes of the inventory of phonemes of the normal hearing. Also there was a fair distribution of these phonemes in that almost all the points of articulation and manner of articulation were found in their inventories of phonemes. However, use of these phonemes in proper positions and in acceptable combinations was at variance with the language of the normal hearing subjects and this variance is described as falling within the errors of addition, substitution, distortion, or even omission.

Several reasons have been suggested in the literature to explain why more errors of omission than other errors occur. The place of articulation of the consonant intended and the position of the consonant in the utterance seem to influence the occurrence of omission errors in the hearing impaired children. Levitt, Stromberg, Smith and Gold (1980) suggest that the consonants produced near the front of the mouth (labial and labiodental) are substantially less prone to omission errors than the consonants produced near the center or back of the mouth. Geffner (1980) has also reported that the velar consonants k, g and nׂ, which are not visible, are omitted in greater proportion than the more visible front consonants, namely, bilabials and labiodentals. There are conflicting findings as regards the positions in which consonants are deleted. Some studies suggest that it is the initial consonant that is deleted more frequently and some other studies suggest that it is the final consonant that is deleted more frequently. Hearing impaired subjects are seen to reduce the multisyllable words by deleting unstressed initial syllables in languages which have stress as an important meaning conveying phenomenon. Frequent deletion of the unstressed syllables and weak phonemes has been noticed. Among the nasals, omission of final nׂ is found more frequent in English. The deletion of final voiced consonants is more frequent than the final voiceless consonants; velar consonants seem more susceptible to final consonant deletion. In our investigation, which was done with Kannada subjects, we found that the subjects were not motivated in several cases by the position of syllables – the deletions were done at random. In fact, in many cases, the initial syllables were retained in the partially hearing impaired. While in the profoundly hearing impaired subjects, the inventory of phonemes available to the children dictated and regulated the retention process, in the case of the partially hearing children, it is the case of production that appears to influence the retention and deletion of sounds. And yet it is, indeed, not possible for us to specify clearly what really contributes to this ease of pronunciation, since this ‘ease’ appears to vary from the subject to another. Of the omissions attested in our data, we could state that the initial syllables have a greater potential for retention if the sounds involved in the syllable are found in the inventory of the hearing impaired, and if these syllables are with the (non-affricate) stops, or with the bilabial nasal, or with an alveolar lateral, in that order.

The correlation of omission errors to intelligibility has been obtained by Markides (1970) and smith (1975). Smith has reported a correlation of (-0.65) and Markides has found a high correlation of (-0.87) between the omission errors in final position and intelligibility. Gold (1980) cautions against hasty interpretations of correlations. Suppose that, of the two errors A and B, error A causes a substantial reduction in intelligibility and error B has little effect on intelligibility. But error B occurs about every time error A occurs. There will thus be a large negative correlation between error B and intelligibility. Despite the high correlation between error B and intelligibility, correction of error B will have little effect on intelligibility. The present investigators suggest that some of the following variables should be considered while correction between errors of omission and intelligibility are sought: intelligibility to whom, what is the familiarity between the hearing impaired subject and the normal hearing, what is the content of communication, is the word familiar and frequent, what is the functional load of the content word, what is the functional load of the sound in the particular context, can not the word be retrieved through other means – by retaining some other sound cues or by signs, or by assuming/understanding it through context, and so on, what is the length of the word, what is the length of the utterance, native/borrowed status of the word and the sequences of sounds.

As already reported, the substitution errors are said to be the frequent type of error in the partially hearing children. The substitution process noticed are voicing-devoicing errors, using stops in place of corresponding fricatives, using glides in place of liquids, denasalization of nasals, fronting of consonants, and substitution between aspirated and unaspirated sounds. These findings are applicable to our data also. Nasalization of vowels, exchange of alveolar lateral for the retroflex lateral, exchange of the alveolar nasal for the retroflex nasal, likewise the exchange of a dental stop for the retroflex stop are also noticed in our data. In other words, there were substitutions of points of articulation as well attested in our data. Here also the partially hearing subjects, who alone attested these processes as a pattern, appeared to be engaged in the simplification of the production process. This raises the question whether the subjects did perceive the difference between the sounds at the acoustic levels, and whether the problem, if they have perceived the sounds correctly at the acoustic level, is related only to the difficulties with regard to production. It may be pointed out that there are similarities in the manner of production between the pairs of sounds found to be substituted in a regular fashion. This is reflected also in the acoustic level wherein one finds that the pairs fall within the same range of frequency, intensity, etc., more or less. Thus there is a range of similarity brings about a regular pattern in the substitution process. The reasons for random substitutions are to be sought in the specific production mechanisms of the sounds involved vis-a-vis the individual abilities for such a production in the individual subjects.

Some have observed the devoicing of stops to be the frequent error while others have suggested that the voiced stops are frequently correct in the deaf children. Unvoiced plosive consonants have been seen substituted for their correspondences, substitution towards the voiced member of the pair is more common than the substitution towards the voiceless correspondence. The hearing impaired children were seen also avoiding voiced sounds (Oller and Kelly, 1974). Several reasons have been offered for devoicing found in the hearing impaired children. Smith (1975) suggested that the mistiming of laryngeal activity resulted in inappropriate devoicing in the deaf children. Gold (1980) offered hypothetical explanations for the predominance of voicing errors in deaf children. Deviant duration of the preceding vowel, he said, may affect the listeners’ perception of voicing in the consonant, and that the problem of continuous phonation may result in predominance of voiced sounds. In any case, the present investigators find the evidence offered to be inadequate and controversial. In our own investigations, we find that the profoundly hearing impaired and partially hearing subjects did not show the devoicing of voiced consonants as a pattern. It is found that the subjects did not have some of the voiced consonants, but this did not amount to a regular pattern of devoicing. It may be noted that voicing is phonemic in Kannada. Secondly, there was another pattern of giving a voiced, unaspirated consonant for the voiceless aspirated consonant. This was found to be having some patterning in some of our subjects.

Fricativization of stops in the hearing impaired has been noticed by several investigators. This also is conspicuious by its absence in our data. Fricatives in genral are absent in the language of our subjects. Kannada has sibilants and glottal fricative as fricative phonemes in the normal hearing. Some investigators noticed that the final position fricatives are often substituted for by the stop consonants in English. Replacement of the liquid consonants by glide consonants in prevocalic position and by mid and back vowels in post-vacalic position in a hard of hearing subject has been reported by Oller, Jensen and Lafayette (1978). Studies report substitution of b for m and d for n. Subjects have been not only denasalizing but also devoicing the nasals. In our study we find that, whereas nasalization, which is not phonemic in Kannada, is more frequently used on the vowels, the nasals are infrequently used. There are only two nasals out of the 4 or 5 nasal phonemes, namely, the bilabial and the alveo-dental, are used in the hearing impaired language. Elsewhere it is reported that substitutions are carried out between homorganic sounds. The substitution of the frontal consonants for the posterior ones has been reported. Velar and palatal fronting, dentalization and blading, and labiodentalization of dental consonants have been reported.

Consonant cluster reduction process is another phenomenon that is observed clearly both in the literature and in our study. The hearing impaired subjects reduce the fricative-stop clusters to a stop consonant, the nasal-stop clusters to a stop, and the liquid consonant to a glide. We note, however, that our subjects did not reduce the non-identical cluster with a stop as the first member and an alveolar trill as the second member, although the trill was sometimes extremely distorted. The other processes reported are assimilation, substitution of unvoiced unaspirated consonants for initial aspirated stops, and substitution of affricates by one of its components occasionally with a voiced-voiceless error (Levitt, Smith and Stromberg, 1976; Levitt, Stromberg, Smith and Gold, 1980). The same authors have suggested that the deaf children place their articulators in roughly the right place, but the consonant substitutions appear to involve errors in timing and control, such as improper velum control, and inappropriate voicing or lack of voicing. We find in our data that our subjects were influenced more by the manner of articulation in that they retain the manner but change the point of articulation.

As already stated, only a low proportion of distortion has been reported in the deaf children. Distortion is generally a sort of co-articulation process indulged in by the hearing impaired. Distortion in our study was found more frequently as regards the trill sound, and the sibilants in the case of the partially hearing. Distortion to some extent was also noticed with regard to the affricates and retroflex sounds.

Almost of studies indicate that, for both partially hearing and deaf subjects, errors involving the final consonants are more numerous than the errors involving initial consonants: Omissions are more frequent in the final position. The deviations are also more often in the final than in the initial and medial positions. These were borne out in our study also. However, in the initial positions, the occurrence of omission was found more frequent if the sounds involved happen to be fricatives including sibilants and retroflex sounds. Also omission in the final positions have to be linked with the ability of the hearing impaired to produce multisyllabic words and longer sentences. We found that, when a word had only a single syllable or is of a disyllabic structure, the final sound is not usually omitted; it could, however, be substituted or distorted. Geffner (1980) reported that the proportion of correct production of consonants in the initial, medial and final positions were 0.30, 0.22 and 0.19 respectively. Her analysis indicated that the errors of omission are greater for final consonants (0.77) than for the initial consonants (0.61). Errors of substitutions which constitute a minimal percentage are greater for initial position (0.08) than for the final position (0.02).

West and weber (1973) have found no relation ship between errors involving place of articulation to hearing abilities or speech intelligibility. The poorest groups in speech intelligibility made about the same proportion of place errors as the best. A multiple regression analysis made by Levitt, Stromberg, Smith and Gold (1980) on the relation between percentage of intelligibility and different error types of different word positions indicates that the errors involving phonemes in the initial or medial positions carry considerably more weight than the errors involving phonemes in word final positions. Note that all these are relevant to and found in the use of language by the normals. For instance, the initial and medial consonants/vowels would have already set the cues for a particular word, since the succession of sounds in a word carries in it the cues for the recognition of a word. Even before a word is fully pronounced, we are able to recognize what the speaker has attempted to produce. In other words, the final sounds have, in most cases, less value for the recognition of the word attempted. The present authors find that, in arriving at the weightage, adequate attention has not been given to minimal pairs. Where minimal pairs occur in an utterance, the weightage for the position in which the pair of contrasting phonemes occurs will be different from the normal weightage found for that particular position. This is valid equally for the language of the hearing impaired.

Several studies have discussed the order of emergence of the place and manner of articulation of consonants and the acquisition of correct consonantal production. Irwin (1947 and 1948) reported that the back consonants appear first in the normal infant’s speech. Carr (1953) studied the spontaneous speech sounds of 5-year old deaf born children. The vowel and consonant sounds in these children did not continue much beyond the level of hearing infants from 12 to 13 months. The data revealed that the front consonants were used by these children more often than the back consonants. According to Carr, this was in agreement with the early rehabilitators of the deaf, who had suggested that the front consonants are the first to be taught. West and Weber (1973) found that, in their four-year old hard of hearing subject, frontal consonants were better established than the posterior consonants. They also found that voiced consonants requiring back lingual placement were seldom used. Smith (1975) reported that the consonants with a bilabial place of articulation headed the list of correct productions, followed by all glides and i, v and n. Stops in alveolar and velar region, h and the linguadental fricatives succeeded. Greater frequency of errors were reported for velar nasal, and alveolar and palatal fricatives in that order. Affricates had greatest proportion of errors in Geffner’s study. She has also given the proportion of correct production of consonants classified on the basis of manner and place of articulation. Geffner’s conclusion is to be noted: the visibility of phonemes proves to be an important factor in the spontaneous ability to produce speech.

Our study was not a longitudinal one; it was conceived to be a cross-sectional study focusing on the age and type of hearing loss as the major variables. We originally assumed that a cross-sectional study of the language behaviour of the hearing impaired may also bring out possible hierarchies of language elements in the language acquisition process, as found valid in the study of normal subjects. However, it soon became apparent that the behaviour of the hearing impaired was so diverse from one another that cross-sectional study will not yield any hierarchy. In a way, this finding leads us to doubt whether, in view of the diversity in individual language behaviour of the hearing impaired subjects, we could work out detailed hierarchies of the emergence of linguistic units in the hearing impaired population. It is also to be checked whether this position is not valid for all language disorders in general. That is, while it is possible to identify the overall types and general characteristics which define a disorder, it may be difficult to work out strict hierarchies of emergence of language units, as done in the case of normal language acquisition. We find in our observations that, while it is possible to conclude that in the case of the deaf subjects the earliest consonants or vowels to emerge are those whose production is more clearly visible (frontal ones), the position that this is in contrast to the earliest emergence of sounds in the normal learning children cannot be agreed to in full. This is so because there are complicating evidences in the normal language acquisition. For example, Jakobson (1971) proposes the priority for bilabial sounds. This partially supports the statistical universals proposed by Greenberg (1963). Moreover, in most studies, a strict hierarchy between the front and back sounds cannot be maintained.

In general, the studies reveal that the labials and labiodentals are the more correctly produced consonants, and the alveolars, palatals and velars are prone to be produced more incorrectly. Though the glottals are not visible, they are more correctly produced than the other back sounds. When the consonants are classified accordingly to manner of production, the laterals, glides, and stops head the list of correct production, and the fricatives, nasals and affricates follow them.

Vowel Errors

As already reported, investigations have shown that vowels are produced correctly more often than the consonants. However, even vowels undergo certain changes in the language of the hearing impaired. These errors include substitutions, diphthongization of vowels, nasalization of vowels, and durational distortions of vowels. The findings in the area of vowel errors committed by the hearing impaired are listed below: (most of the findings are on English speaking hearing impaired subjects).

1)	Vowel substitutions in the eharing impaired children are said to be typically towards a more central vowel (Angelocci, 1964; Levitt, 1972; Levitt, Smith and Stromberg, 1976; Smith, 1975; Levitt, Stromberg, Smith and Gold 1980)
2)	All Vwels are substituted by the vowel ∂ fairly often (Smith, 1975); Levitt, 1972; Levitt, Stromberg, Smith and Gold 1980) and by Λ slightly less frequently (Smith, 1975).
3)	Greater the closeness between the heights of the vowels, greater is the mutual substitutability between them (our own investigation).
4)	Hudgins and Numbers (1942) found that vowel neutralization contributed to 19% of the total errors.
5)	Diphthongization is the least common mistake made by the deaf and partially hearing subjects (Markides, 1970; Levitt, Smith and Stromberg, 1974). It seldom exceeded 1% of the time (Levitt, Smith and Stromberg, 1976).
6)	Excessive nasalization has been noticed in many hearing impaired subjects.
7)	Both the absolute and relative duration of vowels is distorted in hearing impaired children.
8)	The deaf achieve vowel differentiation by making excessive laryngeal variations with only minimal articulatory movements (Angelocci, Kopp and Holbrook, 1964).
9)	The errors in placement, when they occurred, tended towards the more neutral position. This is related to the point made under item (1).
10)	Levitt, Smith and Stromber 1976) interpreted from their deaf subjects that the placement of articulators by their subjects is in roughly the right place. But the aim is not perfect.
11)	The front vowels are used more often than the back vowels (Carr, 1953; and West and Waber, 1973). The low vowels are more correctly produced than the higher vowels (Geffner, 1980).
12)	Substitutions from a diphthong to one of its components is common in the deaf subjects.
13)	The deaf drop more commonly the second component of the diphthong while at the same time prolonging the first component (Markides, 1970).
14)	Creation of two syllables and reduction of diphthong to a single vowel are the two ways in which the deaf subjects manifest the error (Hudgins and Numbers, 1942).
15)	We found in our investigations that substitution among the short vowel and its counterpart is found more often; the long vowels are generally substituted by their short vowels.
16)	u and i are said to have greater stability in the organic formation, which implies that they are more consistently recognized by the listener (Geffner, 1980). In Geffner’s study, the deaf showed some indication that these “stable vowels” are produced correctly more often.
17)	Diphthong errors constitute for 55% of the total errors (Hudgins and Numbers, 1942).
18)	Auditory discrimination of the vowels by the normal listeners is different from that produced by the hearing impaired. This is attributed to the domination of the visual cues in the phonemic systems of the deaf.
19)	Vowel neutralization seems to be greatest for back and open vowels.
20)	While the profoundly hearing impaired subjects do not master the entire inventory of vowels in the language of the normal hearing, the partial hearing subjects are found to have more difficulty in mastering the patterns of usage of vowels, our data revealed.
21)	As we pointed out in the earlier chapter, out data indicates that both the profoundly hearing impaired and the partially hearing subjects do not ever substitute a consoant for a vowel or vice-versa.
22)	Both the profoundly hearing impaired and the partially hearing subjects do not exhibit mastery of the allophnoic patterns of the vowels found in the normal hearing populations. (This was based on our observation of some Tamil speaking hearing-impaired subjects.)

3.3 Consonant-Vowel and Vowel-Consonant Transitions in the Hearing Impaired

In speech utterances, sounds are not produced as isolated entities. They follow or precede other sounds and, in their occurrence, the sounds influence one another in points of articulation and/or manner of articulation. Moreover, phonetic realization of phonemes is one area in which several peculiarities may be noticed in a language. Apart from the mutual influence based on points and manner of articulation, which may be progressive or regressive, coalescence of two sounds may also take place, leading either to the deletion of one sound and retention of the other, or substitution of a sound not related in any manner. There are certain universal constraints. From the universal constraints and universal sets, individual languages draw their own peculiar combinations. Not all these aspects have been studied in the language of the hearing impaired. Most of the studies undertaken so far concentrate on the consonant-vowel and vowel-consonant transitions; in other words, neither the allophonic distribution not an in-depth study of the assimilation /dissimilation/ coalescence processes have been undertaken. Because of inherent constraints in eliciting exhaustive data, and because of the better efficacy of physical analysis, the investigators have often resorted to spectographic and electromyographic investigations to study the transitional aspects.

Investigators find that deaf speakers exhibit difficulty in coordinating articulation with phonation. Rathman (1976) finds that normally occurring co-articulation effects are minimized in the deaf. Rothman finds that the transitions from one sound to another in the speech of deaf speakers have a restricted range and a slower rate of movement than in the speech of the normal speakers. Further, the deaf group manifests articulatory stereotyping – that is, they tend to start each articulatory sequence in much the same way as they do any other, regardless of the context. Monsen (1976) finds reduced second formant transitions both in time and frequency in the hearing impaired. Monsen concludes that since the formant transitions are important acoustic cues for the adjacent consonants, reduced second formant may be an important factor in the low intelligibility of the speech of the deaf. If we assume that vowels, in general, convey to the listener more consonantal cues than what vocalic cues the consonants do give, then it is reasonable that a reduction in the effects of transition may result in fewer consonants being correctly produced than the vowels.

Rathman’s (1977) study finds that the deaf group, in contrast to the normal groups, (i) extends the duration of the intervocalic stop closures and the duration of the individual phonemes, (ii) exhibits difficulty in coordinating articulation with phonation, and (iii) exhibits greater inter-speaker variability. Smith (1975) finds that the transitional process generally reveals a process of insertion. This happens because, according to Smith, the tongue is moving toward or away from the correct position but too slowly. Smith finds that the sequence of vowel and fricatives are produced as vowel, stop and fricative, and the fricative vowel sequence is produced as fricative, stop and vowel or as the fricative, glide and the vowel. Thus, Smith’s study reveals the insertion of another consonant during the consonant-vowel and vowel-consonant transitions.

The present investigators suspect that, while the process itself may be commonly found among subjects of diverse language groups, the choice of the particular items for insertion may differ from language to language and from individual to individual. The closeness or otherwise between the sounds occurring in succession in terms of points and manner of articulation, the length of the utterance, the number and variety of sound units, complexity of phonological structure, the functional value of the sounds involved vis-a-vis the communicative content appear to influence the choice of elements for insertion. For example, in our investigation, there were at least two types of insertions noticed. The first type of insertion was found both in the normal hearing and hearing impaired subjects – vocalic release at the end of a word and between clusters. The second type of insertion was peculiar to the hearing impaired subjects. In the profoundly hearing impaired subjects, pauses between the consonants that constitute a cluster were dominant. Also, these subjects inserted a vocalic release, nasalization and slurring, apart from some consonant or the other between the consonant clusters. The consonant thus inserted cannot be accounted for by any phonological conditioning.

Rothman (1976) suggests a change in focus. A primary emphasis, Rothman argues, on teaching the deaf speakers the correct articulation of individual sounds without accounting for the effects of context on phonemes, syllables and words would result in faulty speech production. Rothman emphasizes that the rules describing (allophonic) phoneme variation due to co-articulation and context should be established and incorporated into early oral speech training programmes for the deaf. The rules so established should always relate to the speech process as a dynamic synergistic event determined by inter-relationship of phonemes in the context. Rothman’s suggestion is linguistically motivated. The goals of therapy, however, cannot be only based on linguistic description of normal language. The type of hearing loss is an important factor in deciding on whether allophonic distribution should be part of therapy.

3.4. Suprasegmetnal Errors

Studies on normally hearing children have shown that early communication is largely suprasegmental, as discussed in Chapter-2. In contrast, the hearing impaired children stop or regress in their vocal behaviour in the age at which suprasegmental production gets differentiated in a normal child. Generally speaking, suprasegmental production precedes segmental production in a normal child. But, the hearing impaired may start imitating both the aspects at the same time during the speech training sessions. As Ling (1976) suggests, if the residual hearing in low frequencies is good, then the sequence of development of segmental and suprasegmental features may be normal.

Relatively less attention has been given to the analysis of suprasegmental errors in the hearing impaired children. The studies on suprasegmental errors are discussed here under intonation, stress and quantity.

Intonation

Since intonation contours carry meaning, they must be morphemes according to Bloomfield (1933). Since the intonation contours are determined by various pitches, he said, they must also be phonemes. In our investigation, we find that intonation is characteristically absent or imperfect in the deaf. And this is clearly in consonance with the picture of segmental phonemes in the language of the hearing impaired. Restricted range of pitch variation ahs been reported by Stark and Levitt (1974). Levitt, Smith and Stromberg (1976) found that the most common suprasegmental errors, as reported by the judges, are inappropriately monotonous rate, insufficient or excessive variability of intonation, inappropriate stress, and spasmodic control of phonation. Stark and Levitt (1974) examined the inter-relationship between reception and production of stress, intonation for Yes-No questions, and location of pausal juncture. They hypothesized that significant information is contained in lower frequencies which may be within the range of residual hearing of deaf children. However, the study revealed poor performance in both reception and production of prosodic features, the highest error rates occurring for the intonation feature. In other words, it is not the availability of frequency range supposedly involved in the perception of intonation that determines the perception of intonations. It appears that something more than the frequency range is required. Perhaps, something related to cognition is involved here.

Susman and Hernandez (1979) tested the intonation control in ten hearing impaired subjects. The subjects read three sentence pairs, each having a declarative and interrogative form. Terminal drops in means fundamental frequency were found for both sentence types. Phillips, Remillard, Bass and Provonost (1968) had earlier suggested that a terminal pitch rise may be more difficult for a deaf child to produce than the terminal fall.

Levitt and others (1980) reported that the deaf use durational cues in place of intonation. In many Indian languages including Kannada, durational cues get merged with intonation cues. For example, the interrogative intonation is formed by adding –a: to the sentence final word. In such cases also, the profoundly hearing impaired had difficulty in maintaining the length of the vowel, whereas some neutralization was noticed in the case of the partially hearing subjects. This is in keeping with our earlier statement that the hearing impaired subjects had difficulty with maintaining vowel length.

We found in our investigation that the hearing impaired children appeared to use idiosyncratic durational cues contrastively between questions and statements, in the absence of consistency in maintaining vowel length. Note that normal hearing children also go through similar stages at earlier ages. However, the normal hearing children to ultimately acquire consistency in maintaining the length for negative interrogation, whereas the hearing impaired do not.

Nickerson (1975) had earlier suggested that the deaf speakers who tend to produce each syllable of equal duration may also generate a similar pitch on each syllable.

Recently Indira (1981) analyzed and compared the intonation patterns of 4 normal hearing and 4 hearing impaired subjects. The subjects read two simple stories in Kannada, each of which depicted the primary emotions – fear and anger, and joy and sorrow respectively. The equipments for the acoustical analysis included the Beat frequency oscillator, pitch analyzer, measuring amplifier, and the level recorder. It was found that the hearing impaired subjects did not have the same rise and fall patterns as normal hearing subjects. The pitch variations in the hearing impaired was also relatively restricted when compared to normal subjects. An additional finding was that the duration of the speech segment was more in the case of the hearing impaired subjects. (Note this is in conflict with our statement made earlier.) It was suggested that this factor would have been the reason for the little changes in the intonation patterns observed in the hearing impaired subjects. In contrast, the normal hearing subjects showed sharp changes in intonation patterns.

In short, it has been reported by various researchers that the intonation control in the hearing impaired subjects is not adequate. Restricted or excessive pitch variation, gradual pitch variation, and the use of durational cues in the place of intonation feature are some of the deviancies described so far.

Several explanations have been reported for the deviant intonation patterns. It is well known that pitch is primarily an auditory domain. Phillips, Remillard, Boss and Provonost (1968) said that the lack of intuitive grasp of the concept of pitch by the deaf may result in their attempts to raise their pitch by increasing the intensity. But Ling (1966) has suggested that the residual hearing need not be extensive for this purpose. Still others have claimed that the attempts to increase the amount of proprioceptive feedback by the hearing impaired speakers result in excessive pitch variations. In contrast, Nickerson (1975) has hypothesized that the excessive pitch variations in the deaf is the result of inappropriate muscles or muscle groups in controlling vowel articulation. These inappropriate muscle contractions may result in inadvertent tensing or slackening of the vocal cords, resulting in greater than average pitch variations. In our own investigations we find that performance of intonation is related to severity of loss. A monotonous utterance is the general rule, with variations depending upon the severity of the hearing loss. Our tool covered information on intonation in its last section where a free narration of a story was elicited from the subjects. Out of the total of 11 subjects, only 4 responded to this. The narrations were faulty and did not contain clear-cut distinctions between various intonations – only a monotonous note was discernible. However, the intonational content could be appropriately conveyed through facial and hand gestures.

Stress

In general, the stressed syllables have a higher fundamental frequency, grater intensity and a longer duration than the unstressed syllables. Nickerson, Stevens, Boothroyd and Rollins (1974) measured the duration of syllables in four short utterances read by 25 deaf and 25 normal hearing children. The results revealed that the deaf children failed to produce the durational differences between the stressed and unstressed syllables. The deaf subjects produced the unstressed syllables with increased duration.

Susman and Hernandez (1979) selected ten hearing impaired subjects, to determine the subjects’ ability to control stress. Each subject read consecutive CV syllables and he was instructed to place stress on one syllable within each trail. In addition, five noun-verb pairs were embedded in carrier sentences to test the subject’s ability to place correct stress in continuous speech. Group means indicated a slightly higher fundamental frequency for stressed syllables in insulation, but the difference was small and statistically not significant. In the noun-verb pairs spoken in sentence format, in contrast to the performance in the production of nonsense syllable, the subjects did not consistently produce higher intensity or frequency levels for the stressed syllables. Four subjects showed the reverse tendency. The intensity of the unstressed syllables exceeded that of the stressed syllables.

Mc Garr (1976) reported that the hearing impaired subjects place equal stress on all words. None of the researches cited above, studied all the three aspects of stress in the hearing impaired. Failure to produce the difference in duration between the stressed and unstressed syllables, greater than normal duration of the stressed vowel when a derivational suffix is added to the base word, omission of the unstressed syllables, insignificant or absence of increase in frequency or intensity while producing stressed vowels are some of the deviancies reported. Therefore, as far as duration is concerned, it is prolonged for both stressed and unstressed syllables. This lack of differentiation between length of stressed and unstressed syllables contributes to the listener’s perception of improper account or stress in the speech of the hearing impaired (Gold, 1980). Placement of equal stress on all words, while reading, may add to the monotonous speech production problem in the hearing impaired.

In most of the Indian languages, stress is not phonemic. It is found to occur in sentence initial position, perhaps with the function of separating one sentence from the other. In many languages stress is placed on words that require to be focused. This generally happens in languages wherein the word order is not rigid. Our problem, then, should be to investigate how the hearing impaired separate one sentence from another and how they focus on significant words. While the speech of the partially hearing has fluency, the speech of the profoundly hearing impaired is generally of a faltering type. In both the cases, however, there are more pauses between words noticed than in the language of the normally hearing. There are also more number of elliptical sentences than found in the language of the normally hearing. One sentence is seen separated from another based on the familiarity the hearer has with the context. Another important factor in this regard is the lengthened pause between sentences in some subjects. Whether the increase in duration of pause is a significant factor to isolate one sentence from another in a stretch of utterance in the profoundly hearing was tested impressionistically by the project investigators by asking two normally hearing subjects. These two felt that the pauses, indeed were longer between two apparently different sentences in a stretch of utterance. However, this judgement could also be motivated by an imposition of their implicit knowledge of sentence structure in Kannada on the utterances of the hearing impaired. In any case, the duration and placement of pauses in the utterances of the hearing impaired were seen to be different from those of the normally hearing. Another important feature was the slurring phenomenon. both the profoundly hearing impaired and the partially hearing subjects tended to slur the syllables usually after the second syllable in a word and almost the entire word in the final position in an utterance. This was in addition to simplification and reduction of syllables in the words uttered. This was, however, not the case with those adults who became deaf or partially hearing in later life. Slurring of syllables and the retention of syllables of a word are two different processes. Whatever be the syllables of a word retained, the slurring does take place as per the rule stated above.

Length

The duration of sounds may be conditioned by the following factors: point and manner of articulation of the segment itself, preceding and following segmental sounds, suprasegmental factors (especially stress), and position of the sound within a higher level of phonological unit. Duration may function as an independent variable at word level (quantity) and sentence level. A thorough investigation of all these factors has not been done in the hearing impaired subjects.

Lehiste (1970) has said that the physiological mechanism that is ultimately responsible for the quantity phenomenon is the process involved in the timing of articulatory movements.

Calvert (1962) found that the deaf speakers typically distort absolute and relative vowel duration. Monsen (1974) studied two aspects of vowel duration: the inherent durational differences between the two vowels i and I and the modifying influence of the following consonant. In the deaf subjects, the two vowels occupied much more to create durational ranges. The subjects tended to creatre mutually exclusive durational classes for the two vowels. The duration of one vowel could not approximate that of the other even when they occurred in the presence of different consonants. The absolute durations of the two vowels in normals overlapped if the accompanying consonants differed.

Studies by Whitehead and Jones (1977) revealed that the severity of hearing loss is an important factor that determines the ability of the hearing impaired children to control vowel duration. The oscillographic analysis revealed that for normal hearing and moderately hearing impaired subjects, a significant difference in vowel duration occurred between voiced-voiceless, and between plosive, fricative consonant environments. But, for the deaf subjects, however, there was no significant difference in vowel duration for any of the consonant environments. The study by Susman and Hernandez (1979) revealed that the performance of the hearing impaired subjects conformed to the normal group by producing longer vowels before the voiced stops than before the voiceless stops.

Reily (1979) investigated the absolute and relative vowel durations with respect to type tense-lax feature, lexical stress, pre-pausal pattern and sentence final lengthening. The absolute duration of vowels produced by the deaf subjects was two to three times longer than that produced by the hearing group. But the relative vowel duration with respect to the vowel type, tense-lax feature, lexical stress, and sentence pattern conformed to the normal patterns. It was found that the relative duration of vowels was closely related to intelligibility but the absolute duration of vowels was not.

Thus, most of the studies indicate that the absolute vowel duration is distorted in most hearing impaired subjects. But controversial result have been reported with respect to relative vowel duration in varied consonant environments. Whitehead and Jones (1977) reports indicate that, the more profound the hearing impairment, greater is the chance of relative vowel duration being distorted. Further, the relative vowel duration, rather than absolute vowel duration, is related to intelligibility.

Prolonged duration has also been reported for consonants. Whitehead and Jones (1978) suggested that the hearing impaired make use of the residual hearing to develop consonantal durational patterns similar to those of the normal-hearing population. In contrast, the deaf do not learn to combine phonemes and view them as dynamic articulatory events; rather, they view speech in terms of distinct individual phoneme units. Therefore, both the absolute and relative consonant duration are distorted in the profoundly hearing impaired population. In Kannada there are long vowels and long consonants (taking all identical clusters as single units). The tool we administrated took into account all the long vowels and their consonants both with their corresponding short vowels and other long vowels. We found that there is a qualitative difference between the vowels of the hearing impaired and the normal hearing in terms of number, and manner of production. First of all, not all the vowels are acquired by the congenitally hearing impaired subjects. Secondly, there was a qualitative difference in the height of the vowels produced in words. For example, the i was always either longer than the normal hearing sound or more tense. Thirdly, the long counter parts of the short vowels, rather the pair of short-long vowels, was not found for all the vowels acquired, in the case of the congenitally hearing impaired children, whether profoundly impaired or partially hearing. A chief characteristic was that the corresponding short vowel found in the normal hearing and the long vowels of the normal hearing were not produced appropriately. Thus performance with vowels was found to be deficient both in terms of their number and quality of production. Another important point noticed was that frequently the vowels are nasalized was opposed to non-nasalization in the language of the normal hearing.

3.5. Speech Characteristics Identified in Instrumental Analysis

Voce onset time (VOT) characteristics of stop consonants (Monsen, 1976), oral air flow characteristics during the production of stop consonants and fricatives (Whitehead and Jones, 1978; and Wirz, Subtelny and Whitehead, 1980) and electromyographic investigation of articulatory movements have been investigated in the hearing impaired children. These studies are reviewed in brief here.

The start of vocal fold vibration in relation to the noise burst or some other articulatory event is referred to as voice onset time or VOT (Fry, 1973). It serves as an important cues for the distinction between voiced and voiceless phonemes. Monsen (1976) stated that because the deaf child constructs a linguistic code by means of cues which are acoustically different from those available to the normal hearing child, his linguistic system may also be different from the normal in some proportionate degree. The results of his study revealed that:

(1) Some deaf subjects produced the distinction between voiceless aspirated stops and (voiced) unaspirated stops in the same way as do the normally hearing subjects.

(2) The deaf subjects who failed to produce the voiced-voiceless distinction have a tendency to produce what appears to be a single phoneme for both stops, and this single phoneme tends strongly to be the unaspirated member at the labial and alveolar places of articulation, and to be either the unaspirated or the aspirated member for the velars.

(3) Both the normal subjects and those deaf subjects who produced the voiced-voiceless distinctions appropriately, did produce the stops whose point of articulation is further back in the mouth to have the longest interval between the release and the onset of voicing. However, it was interesting to note that this same tendency was also found for the deaf subjects who failed to produce a normal distinction during the production of voiced and voiceless stops. It was also found that the exact place of articulation of the stop may be different from the position used in the language of the normal hearing. The tendency for a deaf child to vary the place of articulation is greatest where there is physiologically the most space for variation: least variation is to be expected for the labials and the most for the velars.

In our investigation, we find that the subjects did not make any distinction between the aspirated and unaspirated sounds. In some stray cases, some sounds were produced with aspiration. Such an aspiration was found usually with the velar voiceless stop. Aspiration is not normally used in the everyday speech of Kannada. However, for stylistic and sociolinguistic reasons, aspiration is used even in normal everyday Kannada, even as it is phonemic in learned/deliberately acquired Kannada. Thus, since the occasions in which aspiration is phonemic are not generally accessible to the hearing impaired, it is not surprising that the speech of the profoundly hearing impaired and the partially hearing does not have aspiration as a pattern. However, voicing is phonemic and voicing is utilized in the production of Kannada consonants in the language of the hearing impaired. And yet there is no neatness to pattern maintained in the language of both the profoundly hearing and the congenitally partial hearing subjects. That is, not all the voiceless consonants, which have their voiced counterparts in the normal hearing language are found to have their voiced consonants used in the sparse language emitted by the profoundly deaf and the partially hearing populations. That is, in addition to the fact that the language of the congenitally hearing impaired children does not have all the consonants found in the language of the normal hearing, the subjects exhibit a lack of pattern among the sounds available to them.

Whitehead and Jones (1978) measured the peak rate of oral airflow during the production of plosive consonants in 5 normally hearing, 5 moderately to severely hearing impaired, 5 profoundly deaf adult males. The airflow was measured during the production of CV and VCV syllables. All the subjects were able to intelligibly produce the syllables. The results indicated that for the normally hearing and hearing impaired subjects, in both syllable environments, the voiceless plosives were characterized by significantly greater oral airflow than their voiced counterparts. However, the results for the deaf subjects were not consistent with those for the normally hearing and hearing impaired.

Wirz, Subtelny and Whitehead (1981) indicated that for all the groups of subjects (normal) hearing, profoundly hearing impaired and partially hearing impaired, in both syllable environments, the voiceless sounds (fricatives) were characterized by greater air flow than their voiced counterparts. In deaf subjects, however, the average individual flow rates for each of the four fricatives were significantly greater in the CV context and significantly less in the VCV context, when compared with normal hearing and hearing impaired subjects. In short, the profoundly hearing impaired subjects did not conform to the normal hearing group during the production of plosive consonants, even though they were capable of producing them intelligibly. During the production of fricative consonants, the syllable environments seem to be a factor, which is not so in the normal hearing group. The moderately to severely hearing impaired subjects, on the other hand, conform to the normal group in oral airflow characteristics.

Deviant articulatory movements during the production of consonants in the hearing impaired have also been sought and identified. Lever and Sievers (1976) studied four normal and 11 deaf subjects. The test consisted of 10 single words (CVC type) and three sentences with typical intonation patterns. Investigation using the X-rays, Kinemascope and picture synchronized tone recording revealed three types of articulatory deviance: a deficient deviations in tongue movements, very slight movement in the pharynx area, and adoption of a tongue position peculiar to each subject.

Huntington, Harris and Sholes (1968) compared the articulatory movement in the normals and deaf speakers. The electrodes of the ectromyogram were placed on some diagnostic location on the facial and tongue musculature, and the patterns of contractions were measured for eleven consonants spoken in a disyllabic frame. The results showed that the patterns of facial muscle contractions of the deaf speaker are in general exaggerated. On the other hand, the tongue muscle patterns of the deaf speakers were stereotyped, but were frequently wrong, and there was no consistent pattern in direction of errors. Although they found that the deaf were more likely to produce a consonants correctly if they had a visual model to follow, they suggested that visibility itself was not the crucial factor determining why the bilabial sounds were more often correct than other consonants. They proposed that tongue movements were actually harder than lip movements, and therefore, lingua-alveolars, lingua-dentals and lingua-velars would be hard to produce.

The interpretation made by Huntington, Harris and Sholes is consistent with Geffner’s (1980) findings. Geffner’s study revealed that the linguadental, palatal, alveolar and velar had the least proportion of correct production. The study also revealed that the glottals are produced correctly more often that the lingual consonants. Note that Geffner’s (1980) hypothesis that the visibility of consonants was an important factor for correct production of consonants does not explain the glottals being produced correctly more often than the lingual consonants.

Though Huntington, Harris and Sholes (1968) have concluded that tongue movements are harder in deaf children, it may not be the ultimate cause for the poor utterance of lingual consonants. Inappropriate auditory and speech perception of these consonants may still be the major factors. It is also probable that the glottals are perceived with less difficulty through the tactile modality, and are hence produced correctly in a relatively greater frequency.

Fundamental Frequency Characteristics

It is well known that the fundamental frequency lowers as one moves from childhood through adolescence and then to adulthood. This is said to require some acoustic monitoring (Boone, 1966). Boone stated that the voice of deaf children of 7 to 8 years is not higher than the normal group. But, as the deaf children grow older, they do not necessarily develop the lower voices of the normal post-adolescent. The deaf child or adult, who lacks auditory feedback, appears to need some external guidance in using the acceptable pitch levels of this age group.

Some investigators have reported higher fundamental frequency in their deaf subjects. Engelberg (1962) reports that the falsetto voice is said to be the result of most vibration of the most anterior segments of the vocal cords. Anagelocci (1964) also found higher fundamental frequency in deaf subjects in their spectrograms.

Gilbert and Campbell (1980) reported the following: As the hearing impaired chilren become older, they may be exposed to more auditory training and they may use auditory cues to lower the fundamental frequency. Secondly, the growth of the laryngeal mechanism in both males and females may over-ride the auditory feedback deficit and reduce the fundamental frequency. Thirdly, the emphasis on speech and auditory training in oral communication schools may contribute to the speaking fundamental frequencies which are chosen by the normally hearing population than reliance on the total communication process. In Collins’ (1979) study the subject was fitted with an experimental hearing aid which transposed the sound upward in frequency to the range she hears best. Analysis of the vowels i, a and u in sentences in the aided and unaided condition revealed a systematic downward shift in the fundamental frequency in the aided condition. The study reveals the importance of auditory monitoring for pitch control.

Loudness

Not much has been reported on the loudness dimension of voice in the hearing impaired children. It is well known that the loudness of a voice is determined by the pitch of the voice. This may be the reason why the researchers have concentrated their attention to evaluate the pitch rather than the loudness dimension in the hearing impaired.

Voice Quality

The most frequent quality deviation reported in the hearing impaired is excessive nasality (Hudgins and Numbers, 1942; Sherman, 1952).

Some investigators have reported that, in general, the deaf speakers are perceived to be more nasal than the normal speakers. Further, the deaf speakers are perceived to be more nasal when speaking at a reduced tempo than when speaking at a normal tempo. They concluded that much of the nasality perceived is the natural consequence of the reduced speaking tempo rather than due to the velopharyngeal dysfunction. The tendency of the speakers to break the velopharyngeal seal when their rate of speech is reduced was an explanation offered for the perception of nasality in the speech of the hearing impaired.

Stevens, Nickerson, Boothroyd and Rollins (1976) have investigated nasality in deaf children using an accelerometer attached to the nose. The accelerometer was in turn connected to a visual display. They found that 76% of the subjects had excessive nasalization in at least half the vowels in monosyllabic words and 36% of the subjects nasalized at least 8 of the 10 vowels tested. Fletcher and Daly (1976) concluded that, in contrast to Colton and Cookers’ findings, the rate of speaking was not significantly related to the nasalance scores in the hearing impaired subjects. Boone (1966) suggested that the nasality in the deaf subjects is due to cul-de-sac resonance. The cul-de-sac resonance in this population is the result of retraction of the tongue towards the pharyngeal well instead of placing it in a normal, more forward and higher position.

According to Nickerson (1975), learning velar control is difficult for the deaf children for two reasons: (1) raising and lowering the velum is not a visible gesture and is, therefore, not detectable by lip-reading, and (2) the activity of the velum produces very little proprioceptive feedback. Further, the movement of the velum must be timed accurately when producing words with abutting nasal and stop consonants if the appropriate sounds are to be produced and if the resulting speech is to be produced and if the resulting speech is to be fluent. Tense or harsh voice is another deviant voice quality reported in the deaf (Whitehead and Bancroft, 1979; Wirz, Subtelny and Whitehead, 1981). Thus, hypernasality, harshness, and breathiness are the deviant voice qualities reported in the hearing impaired children.

A study by Seaver III, Andrews and Grants (1980) has indicated that the hearing impaired subjects do not have a deficiency in the anatomical structure needed for the velophrayngeal closure.

In India, so far the assessment of the speech characteristics of the hearing impaired has not included instrumental analysis. In fact instrumental analysis and experimentation of the speech of the hearing impaired using Indian languages has begun in the National Institute for the Hearing Handicapped at Bombay and the All India Institute of Speech and Hearing in Mysore. Neither the deaf schools nor the individual privately practicing speech therapists show much interest in these aspects. The Departments of Linguistics have their own priorities which do not include practical applications and descriptions of the present sort.

3.7. Rate of Speech Production

The rate of speech production in the hearing impaired is said to be slower. The normal subjects produce 3.3 syllables per second (Pickett, 1968) but the deaf produce 2.0 syllables per second (Nickerson, Stevens, Boothroyd and Rollins, 1974). Thus, according to the above authors, the overall speech rate is termed slow in the hearing impaired subjects. The reduced speech rate could be due to the factors discussed here: excessive absolute duration of vowels and consonants, the distorted relative durations of vowels and consonants and hence of the difficulty in making smooth transitions, insertion of more pauses, pauses inappropriate places and prolonged pauses. In addition, long duration of stressed syllables may also contribute to the slow speech rate in the hearing impaired subjects. The review suggests that problems in phonatory-respiratory control resulting in a production of short duration of breathy voice, and deviant timing of articulatory movements may be contributing to the errors. Emphasis of the teachers on isolated phoneme production by the deaf may also be a contributing factor for the reduction in speech rate in the hearing impaired speakers.

3.8. Other Features

Investigations have generally ignored several features of speech and focused on items we have dealt with above. Some of these omitted features are length of sentences, use of juncture, features that mark one sentence from another, and features of discourse. Our investigations partly focused on these elements. The results of our findings are as follows:

(i) Length of Sentences

The profoundly hearing impaired subjects produced only sparse language data. The data consisted only of ‘broken’ sentences – mostly words, generally not connected by any grammatical devices. Secondly, the data consisted also of some phrases with grammatical linkages between the words in such phrases. The sentences were, indeed, very few. The few sentences that were uttered in a complete form were mainly short utterances of the N + V verbal sentences type. The length of the sentences did not generally exceed two or three words. A chief characteristic of the sentences was the excessive nasality, lengthening of the vowels, monotonous delivery and some breathy voice. The connections between sentences were not marked since all the words in a stretch of utterance were marked by pauses in between them almost in an equal measure.

The congenitally partially hearing subjects had more well-formed complete sentences than the elliptical ones as found in the speech of the profoundly hearing impaired. However, there was nasality, lengthening of vowels and breathiness in voice in most of the subjects. There was a blend of single word utterances, elliptical utterances and sentences of various types. However, the length of the sentences did not exceed five words in a sentence. In other words, although the phrases, and other utterances did indicate a variety of linguistic structures and discourse connectors, the language of the partially hearing continued to be deficient in length. We found that N + N sentence type and N + Adj. + N, N + V, and N + Adv. + V sentence types dominated the utterances. Adjectival and adverbial markers tended to be slurred or dropped. In fact the speech of the partially hearing suffered from a tendency to telescope the morphemes of a long word. This telescoping was generally achieved by slurring, but partly by dropping of affixes, and faltering individual production of affixes with pauses in between affixes.

The above characterization does not hold good for persons who acquired hearing impairment later in life.

In all the cases, there was a tendency to grope for words, a tendency caused perhaps by an inability for self-monitoring and reception of feedback from others.

(ii) Juncture

Juncture plays an important role at the speech level in signifying the structural complexity in language. Our subjects and great difficulty in exploiting and making proper use of this features of speech for comprehension of the structural complexity at the language level. This was mainly because of the indiscriminate use of pauses that the hearing impaired exhibited between all words in a stretch of utterance. The routine occurrence of pauses that are lengthier than the ones found in the language of the normal hearing did not give room for use of such pauses to distinguish between the underlying grammatical structures as done in the language of the normal hearing.

(iii) Separation of One Sentence from Another

In normal language, separation of one sentence from another in speech is marked by intonation patterns at the speech level. It is also marked by sentential pauses. In some languages, it could be signified by euphonic devices in the form of occurrence of certain vowels/ consonants at the end of a sentence, or even at the beginning of a sentence. It could also be signified by suitable interjections. That an utterance is not a single sentence and that it is a combination of sentences is marked by a change in the intonation at the end of the “sentences” preceding the main sentence. At the segmental level, such a combination could also be accompanied by morphophonemic changes. Certain structural characteristics could also contribute to the separation of one sentence from another – such structural characteristics could be the location of some selected grammatical category at the “end” of a sentence; it could also be certain inflections to the word occurring at the “end” of the sentence; it could be a combination of affixes in a particular order. Occurrence of such particles may also have certain specific phonetic characteristics.

The type of hearing loss appears to be the dominant factor in the hearing impaired’s ability to utter sentences separate from one another. In most case, the pauses, both interword and intersentential, are indistinguishable from the general appearance of faltering in the speech of the hearing impaired. As already pointed out, the profoundly hearing impaired tend to have pauses between all words in a stretch of utterance. Both the congenitally profoundly hearing impaired and the partially hearing impaired do not have appropriate sentence intonations. Relatively speaking, the partially hearing subjects in our investigation had presented several structures that resemble sentence intonation, but, then, the excessive use of pauses and the monotonous presentation of utterances make it difficult to recognize them as proper intonations, making one sentence from another. The language of the normal hearing (Kannada) uses the euphonic device of a word/sentence final –u to mark the end of a word/sentence along with the intonation. This euphonic device is not clearly used in the language of the profoundly hearing impaired subjects, whereas several of the partially hearing subjects did use at random this device. There was slurring over individual sounds and syllables, where there was breathiness and nasality over the portions of an utterance, a sentence-like utterance. Thus, the occurrence of slurring and breathiness of voice did not allow much room for the use of the euphonic device in the speech of the hearing impaired. Note that what is most easily exhibited almost as an involuntary release becomes a hard to perform task, thus revealing the value of the particular euphonic device of the language of the normal hearing.

The interjections found in our subjects appeared to have a mechanical function in that the interjections were not related to thought processes, deliberate physiological preparations for speech, or stylistic differences in speech delivery.

The morphophonemic changes are not attested except the occasional u / v and i / y alternations and the velarisation of the alveolar nasal in word final positions when followed by a word beginning with a velar stop. The few morphophonemic changes that we notice in our hearing impaired subjects were, thus, the rules of external sandhi. It appears to us that the language of the hearing impaired is marked by the absence of rules of internal sandhi of various types noticed in the normal hearing Kannada language.

(iv) speech Elements of the Discourse

The speech characteristics of the discourse in normal language include the features that we listed under the sub-section Separation of One Sentence from Another. It also includes the futures of continuity, appropriate speed of delivery, accommodation of the listener and other paralinguistic features. Features of continuity include appropriate uses of intonations to mark that a sentence is complete or is still in progress, appropriate use of pronouns of reference, cohesion between lexical items chosen and used, and coherence of communication of ideas, etc. The profoundly haring impaired in our investigation were seen to be marked out by their difficulties with or non-use of sentence intonations, as already pointed out. That a sentence is still in progress is the feeling that a normal hearing person always gets when an utterance is uttered by a hearing impaired persons even in contexts in which the sentences might have been uttered in full form. This is valid even for the congenitally partially hearing impaired children and for those who acquired the impairment later in life. We have already discussed the speed of delivery of speech in the hearing impaired. One of the significant marks of the speech characteristics of the language of the hearing impaired is the absence of relative pronouns in the profoundly hearing impaired subjects of our study. Even in the case of the partially hearing subjects, we find that the subjects engage themselves in strategies that avoid production of sentences that would involve back references through use of pronouns. That is, the utterance are so framed that back references are to be supplied by the listener only. Lexical strategies are exploited for the description of the person referred to through labeling and using the names, rather than using pronouns of back reference.

3.9. To Conclude

There is more to hearing impairment than hearing loss or inadequacy of hearing. The hearing impairment could affect the cognitive processes in several ways. Reference to this has already been made by us in earlier sections. There are several indicators that the hearing impairment does affect cognitive processes. Our subjects had difficulty with adjectives of colour and quantity – both the profoundly hearing impaired and the partially hearing exhibited these difficulties. The sparse data that we could obtain from the profoundly hearing impaired did not show the process of abstraction adopted by them; however, the partially hearing subjects adopted strategies to indicate possession and belongingness in ways different from the normal hearing. While this is surely an indication that the hearing impaired have their own strategies of lexical voice and word order for communication, we also found that these strategies were not clearly extended as a general pattern, even in the subjects that had more than sparse utterances. We shall see the difficulties with grammatical structures the hearing impaired has in another monograph planned.

An important distinction between the normal hearing and the partially hearing young children was the drastic reduction in the loud egocentric speech in the hearing impaired young children. The relationship between and essentiality of the egocentric and inner speech together for later day thought processes have been generally agreed upon. The hearing impaired young children were seen to have difficulties with the progress of egocentric and inner speech. Kyle, et al. (1978) propose that reading difficulties with language materials the hearing impaired have are to be related to the representational inadequacy of the inner speech that these children were seen to have. It is seen by investigators (for example, Darbyshire, 1977) that while evolution of play follows Piaget’s three-fold development pattern in young hearing impaired children, the play of the hearing impaired children slows down in relation to hearing children as they grow older. Darbyshire also reports that games involving rules were the most difficult for hearing impaired children.

Thinking with restricted language is a good description of the thought processes of the congenitally profoundly hearing impaired and the partial hearing young children. The implications of this restricted language on this thinking are reflected in the lack of adjectives, qualifying adverbs and the lack of elaborate processes of abstraction in the language of the congenitally hearing impaired.

The partially hearing impaired make use of several alternate strategies in their language communication. We already referred to changes in labeling possession and belongings adopted by our subjects. In the absence of elaborate labeling tools in the form of words, most of the prelingually hearing impaired subjects tend to categorize people and objects around them into some broad categories such as good or bad, black or white, long or short, etc. That is, while the basic elements of categorization of people and objects around them is seen in the language behaviour of the prelingually deaf, they are left with only the minimum, and are not able to go beyond for an intricate network of classification and categorization.

There is yet another tendency seen – the prelingually hearing impaired children show a tendency towards easier recognition of accurate nouns. That is, even within a single category of linguistic items they are not in a position to recognize all the sub-components of the particular category.This latter tendency is matched by the deficient use of the physiological speech production mechanism in the prelingually hearing impaired children. While they produce some sounds of a category, they fail to produce even a full category of sounds such as the aspirated group in Kannada. They are not able to use certain sounds such as the sibilants effectively in our investigation. Among the physiological mechanisms for speech production, the prelingually hearing impaired children are seen to have greater difficulty with the use of language. In other words, while the children are able to use some articulators, they are unable to use other articulators efficiently.

In many areas of language acquisition, it is shown that while there is not much difference between the rate and quality of language acquisition in the early stages (Paskowitz and Bond, 1982-1983) soon the hearing impaired children show difficulties in maintaining the rate and quality of acquisition exhibited by normal hearing children (Oller, et al., 1978). That is, the prelingually hearing impaired acquire whatever language they also do not acquire fully all the essentials of the language of the normal hearing.

One of the problems that prelingually hearing impaired children exhibit is their inability for anticipatory coarticulation which is a very important part of language communication involving more than one person. This is to be also related to the inappropriate balance between the speech production and speech reception skills in the prelingually hearing impaired. Novelli-Olmstead and Ling (1984) found that training in speech production led to highly significant gains both in speech production and in auditory discrimination of speech, whereas training in listening only showed slight gains in speech production and no gains in auditory discrimination. Robb, et al. (1985) find that hearing impaired speakers’ speech motor skills vary in relation to auditory perception. They find that for each level of hearing loss (moderately severe, severe, profound) there are speech-timing coordination deficits commensurate with the level. Experiments have indicated that the normal hearing speakers are aware of the articulatory movements associated with the speech sounds (Ruscello, et al., 1980). However, in the case of the hearing impaired, the results indicate instability in the production of speech sounds, with variability both in physiological and acoustic measurements. That is, there is no consistency noticed in the use of articulators. This has led to the conclusion that, with reference to consonant articulation, the hearing impaired may place their articulators accurately but fail to coordinate the movement of several articulators, showing deficiency in motor movements. In the case of vowel production, the hearing impaired may or may not move their articulators only within a restricted range. Also the hearing impaired are seen to be unable to control suprasegmental features of speech (McGarr and Harris, 1980). The hearing impaired subjects showed more variable tongue shape and position than normal subjects (Tye and others, 1983). Also note that tasks such as sustained vowel and syllable repetitions yield differential respiratory behaviours between the hearing impaired and normal hearing subjects (Itoh, et al., 1982). Incorrect production of intonation contours is seen as common phonatory problem among the hearing impaired (Friedman, 1985). In other words, there are differences between the normal hearing and the prelingually hearing impaired children with respect to the developmental aspects of language, articulatory patterns, suprasegmental patterns, production patterns, control mechanisms and speech intelligibility.

A distinction is generally made between surface phonetic appearances of a form and its underlying phonological representation. For instance, the Tamil form Maraṅgaļ ‘tree’ includes the underlying phonological form maram ‘tree’. This latter form undergoes several changes when the plural suffix –kaļ is added to it. The normal hearing child relates the two forms and masters both underlying phonological representation as well as the surface forms. This is only a small instance. There are areas in grammar in which construction of phrases depends entirely on an understanding or grasp of the nature of the formal characteristics of the underlying form inflections for tense, number, case, gender, among others, are decided by the nature of the formal characteristics of underlying phonological representation. Even ambiguity in the meaning conveyed by a phrase or sentence is at times resolved by taking an unconscious recourse to the characteristics of the underlying phonological representations. Investigators have attempted to identify the nature of understanding of the rules of phonological representations in the language of the hearing impaired. Prelingually, profoundly hearing impaired individuals would not have been exposed to develop the rules of relationship between surface forms and underlying phonological representations. That is, the hearing impaired children would not have been expected to have the same access to phonological information as hearing children do. This should result in difficulties in using phonological structure to relate different morphological forms of words. This feature of the language of the prelingually hearing impaired also contributes to the peculiarity of the language.

Literature clearly indicates that the channels of reinforcement are affected in a vicious cycle in the language learning contexts of the hearing impaired children. The hearing impaired children show difficulties with regard to a proper interrelationship between the three essential dimensions of speech production and output, namely, the acoustic signal, the articulatory shapes generating it and the muscle contractions controlling the movements of articulators. These are the three essential dimensions in any speech production and output (Harris, 1976). The problems with the inter-relationship with these dimensions on that part of the hearing impaired children lead the caregivers to modify their own linguistic behaviour with the intention of helping the hearing impaired comprehend what the caregivers utter. This has been very well attested in the studies of maternal speech addressed to the hearing impaired children. Complexity of speech features, use of selected utterances and sentence types, repetitions, expansions, utterance acceptability and fluency (Nienhuys, et al., 1984). In several ways, the hearing impaired condition of children comes to regulate the entire communicative acts of the caregiver, and this, in its turn, does not help maintain and widen the channels of reinforcement of oral language. It is not as if the caregivers alone are doing this. Even the peers are known to modify their communicative acts in several ways when they interact with the hearing impaired children. For example, Arnold and Tremblay (1979) report on how hearing children interacted with hearing impaired children and modified their communication skills depending upon the extent of hearing impairment. They found that hearing children interacted more frequently with other hearing children on several behavioural categories. There was a tendency for the hearing impaired children to approach the other hearing impaired children more frequently than they approached the hearing children and yet there was no effort on the hearing impaired children to seek for everything the hearing impaired. That is, while hearing impaired children showed no peer preference, they were least preferred by normal peers. Thus, the problems of hearing impairment go from acoustic and articulatory bases to a much wider social-psychological contexts of the hearing impaired.