fala : um estudo piloto Methods of analysis speech rate : a pilot study

Purpose: To describe the performance of fluent adults in different measures of speech rate. Methods: The study included 24 fluent adults, of both genders, speakers of Brazilian Portuguese, who were born and still living in the metropolitan region of Belo Horizonte, state of Minas Gerais, aged between 18 and 59 years. Participants were grouped by age: G1 (18–29 years), G2 (30–39 years), G3 (40–49 years), and G4 (50–59 years). The speech samples were obtained following the methodology of the Speech Fluency Assessment Protocol. In addition to the measures of speech rate proposed by the protocol (speech rate in words and syllables per minute), the rate of speech into phonemes per second and the articulation rate with and without the disfluencies were calculated. We used the nonparametric Friedman test and the Wilcoxon test for multiple comparisons. Groups were compared using the nonparametric Kruskal Wallis. The significance level was of 5%. Results: There were significant differences between measures of speech rate involving syllables. The multiple comparisons showed that all the three measures were different. There was no effect of age for the studied measures. These findings corroborate previous studies. Conclusion: The inclusion of temporal acoustic measures such as speech rate in phonemes per second and articulation rates with and without disfluencies can be a complementary approach in the evaluation of speech rate. DOI: 10.1590/2317-1782/20162015039 42 Costa LMO, Martins-Reis VO, Celeste LC CoDAS 2016;28(1):41-5 INTRODUÇÃO A velocidade de fala é uma importante medida da fluência de fala, sendo essa inversamente proporcional à gravidade da gagueira. Além disso, é um parâmetro que permite verificar o processamento motor da fala. Desta forma, torna-se importante caracterizar a velocidade de fala tanto de indivíduos gagos quanto de fluentes. A fluência pode ser definida como o fluxo contínuo e suave de produção da fala e é comumente analisada de acordo com os seguintes parâmetros: tipologia das rupturas (disfluências comuns e gagas), frequência das rupturas (porcentagem de descontinuidade de fala e porcentagem de disfluências gagas) e velocidade de fala, em palavras e sílabas por minuto. Para as autoras, a velocidade de fala em palavras por minuto mede a taxa de produção da informação, e em sílabas por minuto, a taxa de velocidade articulatória. Entretanto, na literatura encontramos propostas diferentes para análise da velocidade de fala, tanto no que se refere à unidade de medida (palavras por minuto, sílabas por minuto ou fonemas por minuto), quanto à exclusão ou não das rupturas da fala, o que muitas vezes pode levar o clínico a ter dificuldades em escolher uma metodologia. Em outra perspectiva teórica (sistemas dinâmicos), existe uma unidade temporal de análise, que não a sílaba, chamada unidade vogal-vogal (VV). Nessa perspectiva, as medidas derivadas da unidade VV também refletem a organização temporal do discurso. É importante lembrar que trabalhos sobre a fluência da fala, desenvolvidos por fonoaudiólogos, já utilizaram essa abordagem. Neste estudo, optou-se por não acrescentá-la, uma vez que envolve outra abordagem teórica divergente. A velocidade de fala pode ser medida manualmente por cronômetro ou por medidas acústicas temporais em programas computadorizados, em diferentes tarefas de fala. Situações estruturadas, onde seja possível controlar a ocorrência de pausas — como na repetição de sentenças — minimizam a influência da formulação da linguagem na velocidade e refletem melhor a performance dos mecanismos motores de produção de fala. Essa variedade de tarefas pode interferir no diagnóstico de gagueira, tornando-se fundamental verificar a melhor metodologia de análise e estabelecer valores de referências para tais. Alguns estudos propõem uma série de medidas relacionadas à organização temporal do discurso. Três medidas de duração simples são propostas como ponto de partida: tempo de elocução (duração total de um determinado enunciado), tempo de pausas (duração total dos espaços de silêncio no discurso) e tempo total de articulação (que é a duração resultante da subtração do tempo de pausas do tempo total de elocução). Com base nas três medidas de duração anteriormente expostas, pesquisadores propuseram variáveis temporais considerando o número de sílabas expressas e as medidas do tempo total de elocução e tempo total de articulação. Os autores propuseram que ao se dividir o número total de sílabas pelo tempo total de elocução, chega-se à taxa de elocução. Essa fornece ao ouvinte uma sensação global da velocidade de fala. Os mesmos colocaram, ainda, que para se obter a taxa de articulação, basta dividir o número total de sílabas pelo tempo total de articulação. Essas medidas e variáveis de tempo foram utilizadas em estudos posteriores na leitura de adultos, na leitura de crianças, na leitura de crianças com dislexia, na fala no comando militar, na fala de gagos e fluentes, dentre outros. Ao tomar medida de um enunciado ou suas pausas, por exemplo, acredita-se que tal exame seja relacionado ao nível físico, uma vez que se limita a averiguar qual o tempo gasto em um determinado espaço (representado pelo sinal de fala). No entanto, a partir do momento em que as medidas se relacionam à organização temporal do discurso, passamos para o nível fonético de análise. Isso quer dizer que os parâmetros acústicos retirados do sinal e reinterpretados dentro de uma perspectiva linguística nos levam ao nível fonético de análise. Dentro do nível de análise encontramos, então, as variáveis temporais. Ressalta-se, portanto a importância de se estudar a melhor metodologia para avaliação da velocidade de fala, uma vez que a mesma pode ser utilizada não só para avaliar os distúrbios de fluência como a efetividade comunicativa de forma mais ampla. Na literatura encontram-se estudos que se utilizam de metodologias diferentes, mas não foi localizado nenhum estudo que comparasse as diferentes metodologias. O objetivo deste estudo é descrever o desempenho de adultos fluentes em diferentes medidas de velocidade de fala.


INTRODUCTION
Speech rate is an important measure of fluency of speech, being inversely proportional to the severity of stuttering (1)(2)(3) .Moreover, it is a parameter that allows assessment of the motor speech processing (2) .Thus, it is important to characterize the speech rate both in stutterers and in fluent individuals.
Fluency can be defined as the continuous and smooth flow of speech production (2) and is commonly analyzed according to the following parameters: type of breaks (typical disfluencies and stuttering), frequency of breaks (speech discontinuity percentage and percentage of stuttering), and speech rate in words and syllables per minute (1,2,4) .According to the authors of previous studies (1,2,4) , the speech rate in words per minute measures the information production rate and, in syllables per minute, the articulatory rate.
However, we found different proposals for the analysis of speech rate in the literature, both with regard to the measurement unit (words per minute, syllables per minute, or phonemes per minute) (1,2) and the exclusion or not of speech disruptions (5) , which can often lead the clinician to difficulty in choosing a methodology.
In another theoretical perspective (dynamic systems), there is a temporal unit for analysis other than the syllable, called vowel-vowel (VV) unit.From this perspective, the measures derived from the VV unit also reflect the temporal organization of speech (6) .It is worth highlighting that studies on the fluency of speech, developed by Speech-Language Pathologists, have used this approach before (7) .In this study, we chose to not add it, as it involves another divergent theoretical approach.
The speech rate can be manually measured by stopwatch or by temporal acoustical measures in computerized programs during different speech tasks (8)(9)(10) .Structured situations, in which it is possible to control the occurrence of breaks -similar to the repetition of sentences -minimize the influence of the formulation of language in speed and better reflect the performance of the motor mechanisms of speech production (11) .This variety of tasks may interfere with the diagnosis of stuttering, being important to check the best analysis methodology and establish reference values for such.Some studies (12,13) propose a series of measures related to the temporal organization of speech.Three simple duration measures are proposed as a starting point: time of utterance (total duration of a given statement), time of breaks (total length of silent spaces in the speech), and total articulation time (which is the resulting length of subtracting the time of breaks from the total time of utterance).
On the basis of the three aforementioned duration measures, researchers (13) have proposed temporal variables considering the number of expressed syllables and measures of the total utterance time and total articulation time.The authors proposed that, by dividing the total number of syllables by the total utterance time, one can calculate the speech rate.This gives the listener a global sense of the speech rate.They also proposed that, to obtain the articulation rate, one has to simply divide the total number of syllables by the total articulation time.
These temporal measures and variables were used in further studies of the reading performance of adults (14) , of the reading performance of children (15) , of the reading performance of children with dyslexia (16) , in the speech performance of military commanders (17) , in the speech performance of stutterers and fluent individuals (5,18) , among others.
When taking measurements of an utterance and its breaks, for example, it is believed that this examination is related to the physical level, as it is limited to assess the time spent in a given space (represented by the speech signal).However, from the moment that the measures relate to the temporal organization of speech, we move to the phonetic level of the analysis.This means that the acoustic parameters taken from the signal and reinterpreted within a linguistic perspective lead us to the phonetic level of the analysis.Inside the level of analysis we find, then, the temporal variables (5) .
It should be noted, therefore, the importance of studying the best methodology for assessing speech rate, as it can be used not only to evaluate fluency disorders but also to evaluate communicative effectiveness more broadly.In the literature, there are studies that use different methodologies; but, no study was found comparing the different methodologies.The objective of this study is to describe the performance of fluent adults in different speech rate measures.
This study was approved by the Research Ethics Committee of Universidade Federal de Minas Gerais (CAAE: 01460612.4.0000.5149).All participants signed an informed consent, which explained the research with its risks and benefits, guaranteeing freedom of participation, refusal or withdrawal, and the confidentiality of personal data.
Participants were asked to participate voluntarily in public places such as parks, schools, universities, among others, located in the metropolitan region of Belo Horizonte, Minas Gerais.
Inclusion criteria were being in the age range determined by the research and having signed the informed consent.Exclusion criteria were adults with personal and/or family complaints of stuttering and other communication disorders; presence of neurological changes and/or psychiatric illness; and who have undergone previous Speech-Language and/or psychiatric therapy.All participants must have been born and lived in Minas Gerais for the past 10 years.
The patient's history was surveyed for the inclusion and exclusion criteria.Participants underwent speech sample collection from a picture as a visual stimulus.They were given the following order: "Please look at this picture and tell me whatever you want about it" (1) .The speech was only interrupted by Speech rate analysis CoDAS 2016;28(1):41-5 questions and/or comments, where there was need to encourage the production of discourse to obtain 200 uttered syllables (fluent), required for sample analysis (1) .
Speech samples were recorded on a digital mono voice recorder, Sony ® ICD-PX333, and transcribed literally, seeking to survey speech disruptions (hesitation, interjection, revision, unfinished word, word repetition, segment repetition, phrase repetition, syllable repetition, sound repetition, prolongation, block, pause, and sound or segment intrusion) and obtaining the speech rate in words and syllables per minute, according to the methodology proposed by the Speech Fluency Protocol (1) .
To obtain the speech rate and speech disruption measures, an acoustic analysis was performed using the software Praat, version 5.0.03.To this end, the speech samples recorded on the digital recorder were transferred to a notebook for acoustic analysis.
In the acoustic analysis, the following actions took place: total utterance time (TTEe), break time and unfilled hesitation (TPHnp), pause time (Tp), and the total time of disfluencies (TTdisf).
To calculate the speech rate in words (TxE w/m) and speech rate into syllables per minute (TxE s/m), the methodology proposed by the Speech Fluency Assessment Protocol was used (1) , wherein the total number of syllables and words is divided by the total time of utterance and multiplied by 60.
The speech rate in phonemes per second (TxE p/s) refers to the total phonemes of the message uttered, divided by the total utterance time (TTEe).To check the number of phonemes present in the utterance, the phonetic transcription was performed for their numerical count in the 200 syllables uttered.
TTEe refers to the total time spent in the utterance of the sentence (200 syllables uttered).For the calculation, the silence time (pauses and hesitations unfilled) and the time spent in the production of speech disruptions were not accounted for.This measure was performed by acoustic analysis.
The total time of articulation (TTArt) refers to the TTEe minus the time of break and hesitations unfilled (TPHnp) in order to ascertain the exact time used only with articulation.The duration of TPHnp was verified by the acoustic analysis of the utterance.
The articulation rate (TxArt) was calculated with and without the disfluencies.To calculate the articulation rate with disfluencies (TxArtCdisf), the previously selected uttered messages were reexamined and disfluencies produced were computed on the syllable count.Then, all produced syllables (with or without disfluencies) were multiplied by 60, and this result was divided by the time of articulation with disfluency (TArtCdisf).The TArtCdisf refers to TTEe minus the Tp, obtained by the acoustic analysis.
To calculate the articulation rate without dysfluency (TxArtSdisf), the total number of syllables without dysfluency (200 syllables uttered) was multiplied by 60 and divided by the total time of articulation (TTArt).The TArtSdisf refers to the TTEe minus the total time of disfluencies (TTdisf).
Descriptive statistics measures and inferential tests were performed.To compare the speech rate measures involving the syllable count (speech rate in syllables per minute and articulation rates), we used the nonparametric Friedman test and the Wilcoxon test for multiple comparisons.The age groups were compared using the nonparametric Kruskal Wallis test.The significance level was 5%.

RESULTS
Table 1 presents the descriptive statistics of the speech rate measures considered in this study.The Friedman test showed significant difference between the speech rate measures involving syllables (χ 2 =45.52; p<0.001) and the multiple comparisons (Wilcoxon test) showed that the three measures are different (Table 2).
It can be seen that there was no effect of age on any of the measures studied (Table 3).

DISCUSSION
This study sought to describe the performance of fluent adults in different speech rate measures.Studies show that the analysis of speech rate has proven valuable in evaluating individuals with typical and altered speech development (1,2,15,16,(19)(20)(21)(22)(23) .However, there is no consensus on the best methodology for assessing speech rate, which should be associated with the clinician's objective.
In Brazil, the most traditional protocol for assessing speech rate (1) takes into account the utterance rate (words and syllables per minute).This protocol has been used extensively with fluent speakers and individuals with speech and language disorders (1,2,(19)(20)(21)(22)(23) .The question raised is that the measurement into syllables per minute intends to study the speech rate but does not discount the break time.When the occurrence of breaks is controlled, it minimizes the influence of language formulation in speed, better reflecting the performance of the motor speech-production mechanisms (11) .
The average utterance rate found was of 90.25 words per minute and 170.04 syllables per minute.These data corroborate with other studies for Brazilian Portuguese (18,19) .The average TxArtSdisf is higher compared with the TxE s/m and with the TxArtCdisf, which is expected, as the disfluencies interfere in this measure.
When the time of dysfluencies is subtracted from the articulation time, the articulation rate aims, in this case, to determine the average duration of each syllable produced and only the effectively produced syllables.This proposal is interesting, because it allows the comparison of the average duration of each syllable of the speech of nonstuttering individuals and the fluent speech of stuttering individuals (5) , reflecting the speech rate without influence or with less influence of linguistic processing (11) .
The methodology of the inclusion of the length of disfluencies in the articulation time is coherent, as disfluencies are articulatory attempts to produce a particular phoneme.As its name implies, the articulation time includes the entire duration of the moments in which the segments were articulated (5) .
The development of motor speech processes extends past the age of 16 years and may be established at around 21 years (23) .In this study, in which the sample consisted exclusively of adults (18-59 years), no effects of age were observed in speech rate measures.These findings corroborate with those of previous studies (19,24) , whose results showed that speech rate varies along the stages of life, which may indicate a process of acquisition, development, stabilization, and degeneration.However, it is noteworthy that the previously conducted research (19) observed a decrease in the age group 48-59 years, which differed from the two neighboring groups (38-47 years and 60-69 years), which was not observed in this study.Children speak more slowly compared with adolescence and adulthood.The profile is very close between children and the elderly, initially suggesting maturation of the central nervous system (19) .
The findings for speech rate measurements analyzed in relation to age reinforces the findings in the literature on the variability of flow between individuals (5,18,19) , even when all are considered fluent, as shown by the high standard deviation values.
This study is a pilot project.By definition, the pilot study is a test on a small scale of procedures, materials, and methods proposed for a given type of research (25) .The importance of conducting a pilot study is in the possibility to test, evaluate, revise, and improve the instruments and research procedures (25,26) .
It is believed that this pilot study with fluent speakers has contributed with relevant information.However, it is worth mentioning the importance of further studies to investigate speakers with speech fluency disorders.Only then, it would be possible to establish the best methodology for assessing speech rate.
It would also be interesting that the methodology was applied in other Brazilian regions, allowing the analysis of the influence of "regionality" and "influence of the source language" variables on the speech rate measures, and in a study involving other age groups.
The inclusion of acoustic temporal measures for assessing speech rate proved to be an effective methodology.However, analyzing the cost benefit of these measures in the clinical practice in Speech-Language Pathology and Audiology is of paramount importance, because the analysis proposed by other author (1) is faster than the analyses in this study.Nevertheless, the use of acoustic analysis is suggested to survey the length of the speech samples, as the use of the stopwatch depends on the examiners reaction time and can interfere with the accuracy of the calculation of the speech rate.

CONCLUSION
The results of this study suggest that the inclusion of acoustic temporal measures such as speech rate in phonemes per second and articulation rates with and without disfluencies can be a complementary methodology to evaluate the speech rate.

Table 1 .
Description of the speech rate measures

Table 3 .
Effects of age on the speech rate measures