Dysarthric Bengali speech: A neurolinguistic studyN Chakraborty1, T Roy2, A Hazra3, A Biswas2, K Bhattacharya1
1 Department of Linguistics, University of Calcutta, Kolkata, India
2 Department of Neuromedicine, Bangur Institute of Neuroscience and Psychiatry, Kolkata, India
3 Department of Pharmacology, Institute of Postgraduate Medical Education and Research, Kolkata, India
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0022-3859.43510
Source of Support: None, Conflict of Interest: None
Background and Aims: Dysarthria affects linguistic domains such as respiration, phonation, articulation, resonance and prosody due to upper motor neuron, lower motor neuron, cerebellar or extrapyramidal tract lesions. Although Bengali is one of the major languages globally, dysarthric Bengali speech has not been subjected to neurolinguistic analysis. We attempted such an analysis with the goal of identifying the speech defects in native Bengali speakers in various types of dysarthria encountered in neurological disorders. Settings and Design: A cross-sectional observational study was conducted with 66 dysarthric subjects, predominantly middle-aged males, attending the Neuromedicine OPD of a tertiary care teaching hospital in Kolkata. Materials and Methods: After neurological examination, an instrument comprising commonly used Bengali words and a text block covering all Bengali vowels and consonants were used to carry out perceptual analysis of dysarthric speech. From recorded speech, 24 parameters pertaining to five linguistic domains were assessed. The Kruskal-Wallis analysis of variance, Chi-square test and Fisher's exact test were used for analysis. Results: The dysarthria types were spastic (15 subjects), flaccid (10), mixed (12), hypokinetic (12), hyperkinetic (9) and ataxic (8). Of the 24 parameters assessed, 15 were found to occur in one or more types with a prevalence of at least 25%. Imprecise consonant was the most frequently occurring defect in most dysarthrias. The spectrum of defects in each type was identified. Some parameters were capable of distinguishing between types. Conclusions: This perceptual analysis has defined linguistic defects likely to be encountered in dysarthric Bengali speech in neurological disorders. The speech distortion can be described and distinguished by a limited number of parameters. This may be of importance to the speech therapist and neurologist in planning rehabilitation and further management.
Keywords: Bengali, dysarthria, neurolinguistic study, speech defect
Dysarthria refers to a group of motor speech disorders affecting various domains related to language expression such as respiration, phonation, articulation, resonance and prosody,  that result from damage to the central or peripheral nervous system. It excludes stammering and speech disturbances due to anatomical deformities like cleft palate. Dysarthria has been grouped into six types depending on the nature of the speech defect - spastic, flaccid, mixed, hypokinetic, hyperkinetic and ataxic  . Depending on the extent of damage to the underlying neural mechanism, dysarthria severity can range from minimal speech dysfunction to complete absence of intelligible speech. One or more of these types are encountered as a feature or sequel to various neurological disorders such as stroke, Parkinsonism More Details, cerebellar disorders, and others.
Bengali is one of the major Indo-European languages. According to the 2001 census, it holds the second position in terms of the number of speakers among all the Schedule VIII languages of India. It is spoken by over 80 million people in India. Bengali is the mother tongue for most of the native population of the Indian states of West Bengal and Tripura. There are also large concentrations of Bengali speakers in other Indian states like Assam, Orissa, Bihar, and Uttar Pradesh. Internationally, Bengali is the national language of Bangladesh and there are many Bengali-speaking immigrants from the Indian subcontinent in countries like the US, UK, Canada, and Australia. In the world, Bengali holds the sixth rank in terms of the number of speakers which is estimated to be in excess of 185 million. 
Various studies related to dysarthric speech are available for other Indo-European languages. For instance, notable studies on dysarthric English speech are the Mayo Clinic dysarthria study,  the Frenchay Dysarthria assessment,  and Assessment of Intelligibility in Dysarthric Speakers by Yorkston and Beukelman.  Ziegler and colleagues have reported work with German-speaking dysarthrics. 
However, since speech sounds (phonetics) and patterns of stress and intonation in speech (prosody), appear to vary in the context of individual languages, findings from studies on dysarthric speech in other languages cannot be applied to Bengali speech. This paper deals with the perceptual analysis of dysarthric Bengali speech from the neurolinguistic point of view. To our knowledge, no study of this kind is available till date. We have attempted to identify and quantify the various parameters that are affected in dysarthric Bengali speech in neurological disorders, with respect to the five linguistic domains namely, respiration, phonation, articulation, resonance and prosody. On the basis of this analysis we have attempted to standardize the neurolinguistic assessment that can help in characterizing the speech defect, correlating it with the neural pathology and planning speech rehabilitation therapy.
We conducted a cross-sectional observational study on 66 dysarthric subjects (53 males), at least 12 years of age, attending the Neuromedicine Outpatient Department of a tertiary care teaching hospital in Kolkata. Data was collected during the six-month period of March to August 2004.
At presentation, all subjects were suffering from a primary neurological disorder and underwent clinical examination by a neurologist followed by investigations such as CT scan, MRI scan, or electrophysiological study as necessary. The neurological diagnosis was corroborated independently by a second neurologist. The time of onset of the neurological disorders was variable but all patients had symptoms of at least six-month duration. All subjects were right-handed and spoke Bengali as their first language. Each demonstrated distorted speech secondary to impairment of the central and/or peripheral nervous system. It is important to note however, that subjects with damage to primary cortical areas known to be associated with speech production were excluded and we studied speech execution defects (dysarthria) rather than speech motor programming abnormalities (aphasia). This distinction is important in the study of dysarthrias. 
Subjects exhibiting dysarthria of different neurological etiologies underwent the linguistic articulatory test to confirm the type of dysarthria. No formal sample size calculation was done. The 66 subjects selected represented a convenience sample of patients satisfying the selection criteria who were screened within the six-month study period.
Phonetic capabilities of the dysarthric subjects were assessed through perceptual linguistic analysis by one of the investigators who was blind as to the neurological diagnosis. Speech samples for the analysis were elicited from each subject using test material of two types - a basket of 184 words and a block of text consisting of nine sentences. The word basket included words selected on the basis of frequency of their usage in day-to-day Bengali prose or intuitive use by native Bengali speakers,  so that none of these words were likely to be unknown to the subjects, irrespective of their level of education. The nine sentences in the text block used included all Bengali vowel and consonant phonemes. The analysis on the word basket was the test of pronunciation of vowels and consonants in the initial, medial and final positions of the words by the dysarthric subjects. It also focused on how these subjects handled the consonant clusters in the initial, medial and final positions of the words. The nine-sentence text was read out by the dysarthric subjects at conversational rate. This was needed to identify the various defects within the domains of respiration, phonation, articulation, resonance and prosody.
Subjects' speech productions were recorded using a Sony TCM-150 cassette recorder with built-in microphone at a constant mouth-to-microphone distance of 5 cm. Speakers read each stimulus word and sentence following the test material. Orthographic representations of the stimulus material were provided in front of the speakers. If the subjects could not read the text spontaneously, the examiner modeled short phrases that were repeated by the subject. This recorded speech was used for the perceptual analysis. On-the-spot analysis was not done as finer defects require time and repeated hearing for assessment. For validation, two other observers apart from the researcher repeated the perceptual analysis.
The parameters that were assessed were modeled on the Mayo Clinic study  and were as follows:
These 24 parameters were grouped into the five domains of articulation (parameters 1 to 3), respiration (4 and 5), phonation (6 to 14), prosody (15 to 22) and resonance (23 and 24).
Data was entered into a Microsoft Excel worksheet and analyzed by Statistica version 6.0 (Statsoft Incorporation, Tulsa, Oklahoma, USA; 2001) software. Dysarthria scores were calculated for each subject on the basis of the number of parameters defective. Individual parameter score as well as total dysarthria score were compared between groups (dysarthria types) by Kruskal-Wallis analysis of variance, followed by pair-wise Mann-Whitney U test for post hoc testing if needed. Distribution of defective parameters between groups (dysarthria types) was also assessed by Chi-square test or Fisher's exact test, as appropriate. Parameters whose distribution differed significantly were included as identifying criteria for the different dysarthria types. A P value < 0.05 was taken to be statistically significant. All analyses were two-tailed.
Dysarthric subjects included in this study ranged in age from 12 to 73 years. Thirteen subjects (19.7%) were females. [Table 1] shows the age profile and distribution of neurological disorders with their anatomical substrates for the different types of dysarthria. Although the age range was wide, the majority of subjects were middle-aged adults.
The prevalence of different speech defects, under the five linguistic domains as mentioned above, has been summarized in [Table 2]. Parameters present in only a few subjects (less than 25% within a particular type of dysarthria) have been excluded. Imprecise consonants were an almost universal feature of all dysarthrias, with the highest prevalence among hyperkinetic (89%) and ataxic (87%) individuals.
Depending upon statistically significant differences in the frequency of various speech defects among different dysarthria types, it was found that individual parameters could help to distinguish certain dysarthria types from others. These results have been summarized in [Table 3].
Most dysarthric individuals tend to be shy of company and stay mute if they fail to establish effective communication. Society also tends to ignore or be intolerant of the dysarthric individual. However, rather than social isolation, recognition and evaluation of the problem will inspire remedial measures and improve the attitude of the community. The diversity of speech defects in dysarthric individuals, particularly in degenerative neurological disorders is daunting, but their management is crucial to improving the quality of life.  A study to evaluate dysarthric Bengali speech should be viewed in this context.
According to Darley et al., 1975, the authors of the now classic Mayo clinic study,  "Dysarthria is a collective name for a group of related speech disorders that are due to disturbances in muscular control of the speech mechanism resulting from impairment of any of the basic motor processes involved in the execution of speech." By impairment of the basic motor processes, they meant "coexisting motor disorders of respiration, phonation, articulation, resonance and prosody". For assessment of speech disorders they used 38 dimensions organized in seven groups - pitch, loudness, laryngeal, resonatory, respiration, prosody, articulation and overall. Although the speech defect parameters in this study are modeled on the 'dimensions' used in the Mayo clinic study, they do not match exactly. We selected 24 parameters grouped under the five domains of respiration, phonation, articulation, resonance and prosody, to enable perceptual analysis of dysarthric Bengali speech. The groups of pitch and loudness as suggested in the Mayo clinic study were included in prosody and phonation respectively. The dimensions included under laryngeal and resonatory dysfunction, were included under phonation and resonance respectively. No 'overall impression' parameter has been used.
Yorkston and Beukelman  proposed that "...individuals who are judging (intelligibility) must not have precise foreknowledge of what the speaker is saying". They circumvented this difficulty by employing a rigid procedure; speech samples recorded for assessment were always judged by individuals other than those who recorded the data. This was not the case in our study. In a validation exercise done with all subjects (data not shown), the intraclass correlation coefficient was above 0.9 for most of the 24 parameters in both intra-rater and inter-rater validation; the minimum value obtained being 0.74. This suggests that the intelligibility assessment was not biased by foreknowledge of the speech stimulus.
In spastic dysarthria, strained voice was displayed by 74% of the subjects, and monoloudness by 60%. Both these parameters pertain to the phonation domain. The articulation domain was also affected considerably - 67% of the subjects displayed imprecise consonants, although only 27% showed prolonged phoneme. Defects related to prosody were displayed through reduced stress (54%) and slow rate (47%). Hypernasality, a resonance domain parameter, occurred in 40%. The parameter of strained voice was present in much lesser degrees in other dysarthria types, with the exception of hyperkinetic dysarthria. However, strained voice along with monoloudness was a stronger indicator of spastic dysarthria since monoloudness was less common in the hyperkinetic type.
In flaccid dysarthria, hypernasality (80%) and imprecise consonants (70%) were the predominant defects. The coexistence of the prolonged phoneme defect (40%) indicated that the articulation domain was badly affected. Phonation defects manifested through monoloudness in 50% and breathy and strained voice quality in 40%. In addition, a prosodic defect showed through reduced stress in 40% subjects. Hypernasality and breathy voice, if present, could serve to distinguish flaccid dysarthria from other types.
Mixed dysarthria, as the name suggests, displayed a large number of defects representing other dysarthria types. However, in our sample, excess and equal stress and irregular articulatory breaks, that were among the more prevalent features of ataxic dysarthria, seemed to be absent in the mixed dysarthric subjects. Increase in overall rate was a feature noted only in mixed and hypokinetic types.
In hypokinetic dysarthria, phonation defects were pronounced, manifesting through presence of monoloudness and strained voice quality in 50% of the subjects. Reduced stress, a prosodic defect, also affected 50% along with slow rate and increase in overall rate, affecting at least 25%. The defects resembled those in spastic dysarthria, excepting hypernasality which did not occur in hypokinetic subjects and increase in overall rate which did not occur in spastic dysarthria.
Imprecise consonants affected 89% of hyperkinetic dysarthria subjects. Phonation defects were also frequent, including strained voice (78%), vocal tremor (56%) and monoloudness (34%). Prosodic defects included reduced stress (56%), slow rate (45%) and irregular articulatory breaks (34%). The resonatory feature of hypernasality occurred in 34%. Thus it appears that for both hyperkinetic and hypokinetic dysarthrias, in addition to the articulatory defect of imprecise consonants, the phonation and prosodic domains are likely to be considerably disturbed. However, vocal tremor is likely to be present in hyperkinetic rather than hypokinetic dysarthria and can serve as a distinguishing feature.
In ataxic dysarthria, the prosodic domain was largely affected - 63% had defects of both excess and equal stress and irregular articulatory break, while short rushes, variable rate, and reduced stress were found in at least 25%. Phonation components were also usually affected, voice quality being either hoarse or strained. Monoloudness and vocal tremor were also liable to be present. However, the high prevalence of excess and equal stress and irregular articulatory breaks can distinguish ataxic dysarthria from other types.
Our study has attempted to define the speech defects likely to be encountered in dysarthric Bengali speech. The relatively small convenience sample is a limitation, nevertheless we have found that out of the full range of parameters that may be assessed, a more limited number (15 out of the 24 studied) can suffice to describe the defect and distinguish between the various dysarthria types. From the linguistic point of view, this is important, as it will help not only in characterizing the defect but also in planning speech therapy or augmentative and alternative communication techniques for the dysarthric individuals. , From the neurological point of view, ascertaining the speech defects and dysarthria types correctly would be a pointer to the type of neurological lesions that may underlie the defect,  and alteration of speech defect patterns over time may indicate disease extension. Linguistic analysis may thus help in planning subsequent intervention and management.
The authors would like to thank Ms. Debjani Sarkar MSc, Research Scholar, Department of Pharmacology, Institute of Postgraduate Medical Education and Research, Kolkata for assisting in the analysis of speech data.
[Table 1], [Table 2], [Table 3]