CC BY
30
ALGORITHM OF PERSUASIVE TALK IN PITCH AND DURATION
УДК 80, 311.3/.4
Tat'yana Shevchenko,
PhD in Philology, professor, Department of English Phonetics, Moscow State Linguistic University, RF
E-mail: tatashevchenko@mail.ru
Natal'ya Sadovnikova,
PhD in Economics, professor, Department of Theory of Statistics and Prognosis, Moscow State University of Economics, Statistics and Informatics, RF
E-mail: nsadovnikova@post.ru Lyudmila Sibileva,
Doctorate in Philology, associate professor, Marketing Department, Academy of National Economy, RF
E-mail: sibileva.ludmila@gmail.com
The research is concerned with prosodic rhythm coordinating the impact of pragmatic value and novelty factor on the prosodic properties of words. The corpus is unique in its authenticity and optimization for human perception through long-term practice. The words representing the cognitive scenario of persuasive monologue are found to belong to the foreground, a prominence category, within which gradience is established dependent on consecutive occurrences and pragmatic value of words. Prosodic prominence is further distributed in a patterned way for the initial, medial and final positions in the intonation phrase (ip). Thus the pragmatic and novelty forces appear to comply to the major type of prosodic rhythm of ip .measured in pitch and duration. The results of spoken language analysis might be relevant for speech coding, speech recognition and persuasion.
Keywords: spoken language, prosody, rhythm, language planning, novelty/givenness, pragmatic value, intonation phrase, pitch, duration.
Татьяна Ивановна Шевченко,
д.филол.н., проф., Кафедра фонетики английского языка, Московский государственный лингвистический университет Эл. почта: tatashevchenko@mail.ru
Наталья Алексеевна Садовникова,
д.э.н., проф., Кафедра теории Статистики
и Прогнозирования
Эл. почта: nsadovnikova@post.ru
Людмила Николаевна Сибилева,
к.филол.н., доц., Кафедра маркетинга, Российская академия народного хозяйства и государственной службы при Президенте РФ
Эл. почта: sibileva.ludmila@gmail.com
Алгоритм убеждающей речи в мелодике и длительности
Проведен статистический анализ высотных и темпоральных характеристик убеждающей речи американских бизнесменов. Установлены взаимосвязи между основными индикаторами мелодики и длительности американской речи относительно основных концептов и синтаксических позиций.
Ключевые слова: устная речь, просодия, ритм, языковое планирование, новизна/ данное, прагматическое значение, интонация фразы, высота, длительность.
1. Introduction
Our study stems from the necessity to check if rhythm of language planning, together with semantic and pragmatic factors, plays a role in prominence-lending information-bearing prosodic organization of spoken language.
A brief overview of major prosody research issues reveals that the rhythmic nature of information delivery in spoken language seems to be neglected.
Cognitive approach to spoken language following Chafe (1987) and Lambert (1994) has brought tangible results in correlating prosody with information structure [1; 2]. Within the framework of Autosegmental Metrical Phonology Pierrehumbert and Hirschberg (1990), Ladd (1996) have shown that different accent types mark the information status of individual discourse referents [3; 4].On elaborating fine-grained technique of speech analysis Baumann (2006, 2010) has applied it to spoken language corpus of German to prove that semantic-pragmatic categories are crucial for the interpretation of prosody in terms of information status and cognitive activation states [5; 6].
We assume that the rhythmic factor is a third regulating force which accounts for prosodic speech prominence of words in spoken language.
Speech rhythm is viewed as regularity, periodicity of commensurable speech events. Intonation phrase (ip) is taken to be the basic unit of prose rhythm, a higher-level unit in the hierarchy of phonological units, above the phonological word and the foot (Hirst and di Cristo 1998, Gussenhoven 2005) [7, 8].
We hold it that structurally ip is a construction consisting of three parts: the initial, the medial and the final (nuclear) ones. Pragmatically it is, as Halliday argued, an information unit [9] Cognitively it is a unit of speech planning, the length of which is determined by echoing memory, breathing and perception constraints.
The grouping of words into ip conforms to the speaker's global pragmatic aim of the talk, as well as her/his current intention, which together with the lexical choices, the construction schema and prosody results in bringing particular words into the foreground (focus) and thus taking the message across [10]. Prosodic means of prominence are accentuation and phrasing which are measured acoustically as Fomax, Fospan, and syllable duration of accented words and pause occurrences.
The goal of the present study is to explore how the impact of three factors: pragmatic value, degree of givenness and position in ip rhythmical structure, can be reconciled in the ongoing discourse.
2. Corpus
The unique features of the corpus 'Interviews with ten most successful networkers in the USA', (Upline 1986) (4hrs, 10 speakers, 3 women and 7 men)) is its authenticity and the fact that the talks have been previously repeated over the years and thus optimized for their effectiveness in communication aimed at persuasion. The talks are built around a stereotyped 'how to become a millionaire' scenario with the key concepts, or, rather, frames which we called 'success', 'business', 'leaders'.
The six subjects whose talks were selected for the narrow corpus (in search for gender differences we balanced the number of men and women to three in each gender group) are 'great communicators' sharing their experience in building and duplicating their businesses. The pragmatic aim is to persuade the listener, quite often a complete stranger, to join the company for one's own benefit, develop it through duplicating and find new leaders whose enthusiasm, energy and desire to succeed will guarantee further development.
The total for the narrow corpus is 18 min 30 sec, 413 ips containing 3131 words, with 315 words representing the 'success-business-leaders' scenario concepts and 74 words, adjectives and adverbs, which serve as intensifiers. The total amount of 389 tokens corresponds to 131 lexical entities, each being repeated approximately three times (2.97). Repetitions give us a chance to observe how the degree of 'givenness' affects prosodic prominence of the repeated words.
An intonation phrase which we assume to be the major unit of prose rhythm averages in the corpus at 2.3 sec (followed by an average pause of 0.46 sec) which is close to the echoing memory capacity of 3 sec. The length of ip in the present discourse is, therefore, optimal for human perception.
3. Method
Tapescripts of monologue parts from six interviews were prepared. Auditory analysis consists in the division into ips (from pause to pause), accents and tones marking.
Acoustic analysis (Speech Analyzer v. 2.5) parameters are: Fomax, Fomin, Foint-erval (span), mean syllable duration in 389 selected words representing the cognitive scenario and 'intensifiers'. The same measurements were taken for the rest of the text (minus the selected lexical items) in 413 ips.
The data was grouped:
• according to the novelty/givenness value in the 1st, 2nd,- 3rd and other occurrences of each lexical item in every speaker's performance;
• according to the pragmatic value of the words representing different frames (concepts) of the scenario: "success', 'business', 'leaders' and 'intensifiers';
• according to the position in ip construction: initial, medial or final. (One-word ip formed a separate group).
Statistical analysis (ANOVA) followed.
4. Results
4.1 Scenario words are in the foreground
The first observation based on auditory analysis consists in the following: with the exception of three words, the remaining 386 scenario words are accented. Acoustically there is considerable contrast in pitch range characteristics of the accented scenario words compared to the background of the discourse: the selected lexicon is realized within the pitch range of 7.5-10.6 st, while the rest of the text is realized within 1-7 st. Thus pitch prominence is reserved for the conceptual scenario words (together with intensifiers) and this distinction may be considered categorical: there is a clear-cut division into foreground, on the one hand, and background, on the other. The findings that followed evidenced gradience in the prominence category but one important point should be made here: the relative prominence does not mean that the scenario words could be pushed into the background.
4.2. Prominence is graded with given-ness
The next step was looking at pitch height, pitch range and duration variation in the three (or more) occurrences of each particular word in every speaker's talk. The degree of novelty/givenness, as we have stated, is assessed in the present study in a formal way, i.e. according to the first, second and third occurrences in discourse.
The general tendency transpires when
pooled and averaged over the three (or more) consecutive occurrences in all the talks (Table 1).
The relevant parameters indicating a drop in prominence are Fo max, Fo int (span). Syllable duration values fluctuate but a drop on the second word occurrence is constant.
We have, therefore, found gradience in novelty/givenness, or information status, operating on the prosodic level. Previous research results appear to be confirmed in a way which signals cognitive planning to be regular and, probably, automatically adjusting to each successive word mention.
4.3. Gradience is caused by pragmatic
value
The next finding consists in gradience within the prominence group determined by the relative pragmatic value of the three concepts/frames. In order of appearance in discourse they are graded prosodically: the 'success' words which symbolize the motive and the reward in the networkers activity (success, successful, opportunity, benefit, etc.) are more salient than the 'business'' words (business, network, marketing, organization, industry, distributor, duplicating, partnership, etc.) but the 'leaders' key frame words including the slot with leaders' distinctive characteristics are the most prominent ones (leaders, leadership, pick performers, passion, enthusi-
Table 1. Givenness in pitch and duration
Occurrences Parameters 1 2 3
Fomax Hz 183 175 170
Fomin Hz 112 108 112
Foint st 9.1 9.0 7.5
SyllD ms 232 210 222
Table 2. Pragmatic Value in Pitch Differences
Occrrences 1 2 3
Parameters Max Int st Max Int st Max Int st
Concepts Hz Hz Hz
success 179 9.3 160 10.7 150 7.7
business 173 9.8 169 10.6 156 8.1
leaders 191 8.0 187 7.0 195 7.9
intensifiers 188 9.1 178 7.0 179 6.4
Table 3. Regression model and statistics
Parameters Concepts Regression model F p SE R2
success Y=216.41+0.03 X 244.32 <0.001 60.88 0.907
business Y=216.18-0.10 X 916.25 <0.001 46.56 0.913
leaders Y=265.48-0.06 X 970.39 <0.001 78.69 0.831
intensifiers Y=329.54-0.58 X 517.95 <0.001 65.58 0.897
Table 4. t Stat Regression coefficient
Regression coefficient SE Coef t p
+0.028 18.11 2.778 <0.001
-0.101 3.56 -3.838 <0.001
-0.064 45.23 -3.411 <0.001
-0.575 31.25 -2.703 <0.001
Table 5. Initial position frequency of occurrence, pitch and duration
Parameters Frequency Fomax Syll Dur ms
Concepts % Hz
Success - - -
Business 4.5 166 223
Leaders 18.1 204 218
intensifiers 26.0 196 233
Table 6. Pitch and duration in ip (medial, final)
Parameters Fomax Fomin Foint SyllDur
Positions Hz Hz st ms
medial 170 113 7.1 201
final 179 102 10.6 250
asm, responsible, commitment, committed, noticeable development, personal growth, result, ability, mentality, communication skills, synergy, contribute, competitive, etc.). The 'in-tensifiers' group (powerful, specific, magic, burning, really, certain, very, great, always, important, right, extremely, enormously, incredible) comes second in this ranking (Table 2).
The data suggests that the established drop of prominence with each successive occurrence is generally preserved over the corpus. The exceptions are a few cases of one-word ips in the 'leaders' frame which form a crescendo sequence with growing pitch values.
Regression models suggest that the use of pitch in 'success' words positively correlates with syllable duration, while in the other three frames the correlation is negative (Tables 3, 4).
4.4. The regulating power of rhythm in ip structure
Concerning the location of content words relative to ip structure, it was found that the cognitive scenario words gravitate towards the medial and final positions in ip, whereas 'intensifies' are more common in the initial and medial positions. By considering prosodic properties of scenario words in the three positions the effect of ip power points on their relative prominence is revealed.
The initial position is much less common in the scenario words locations, as can be illustrated by the relative frequency of occurrence in the four groups. However, their pro-sodic values in the initial position (Fo max and mean syllable duration) are higher than those in the medial one (Cf. Table 5, and Tables 6, 10).
The regularity of maxima pitch and duration values concentrating in the final position proves that it is the most powerful position for both new information bearing and pragmatically relevant words (Tables 6, 11).
The regulating power of rhythm becomes evident when we discover that whatever the novelty/givenness or pragmatic value of the words might be, they comply to the power of the initial position increase in pitch, the medial position drop and the pitch and duration maxima location in the ip final position. The pro-sodic form of pitch contour in spoken American English is patterned that way. Pitch and duration parameters are automatically adjusted to the pattern. The wavy-like pulsation of thought is congruent with the wavy-like pitch maxima, pitch range and duration variance (Fig. 1, 3, 2).
5. Conclusions
The data presented above displays the work of speech planning mechanism governed by the pragmatic aim of persuasion. This lis-
Table 7. Descriptive Statistics: Occurrence 1
Concepts Parameters i
Fomax Fomin Foint SyllD
Succes Mean 179 115 9,3 217
SE 34,5 43,2 4,2 72,4
St.Deviation 42,5 48,2 4,9 81,8
Median 177 112 9,1 213
Business Mean 173 98 9,8 214
SE 46,6 28,8 4,5 48,9
St.Deviation 58,0 35,2 5,6 67,9
Median 175 96 9,9 210
Leaders Mean 191 119 8,0 260
SE 39,1 37,9 3,7 83,2
St.Deviation 51,5 42,9 4,1 93,5
Median 160 87 6,0 198
Intensifiers Mean 188 113 9,2 237
SE 43,8 35,4 4,8 70,8
St.Deviation 51,7 41,5 5,5 90,5
Median 192 111 7,0 215
MEAN 183 112 9,1 232
Table 8. Descriptive Statistics: Occurrence 2
Concepts Parameters 2
Fomax Fomin Foint SyllD
Succes Mean 161 90 10,7 199
SE 25,0 27,3 4,1 51,5
St.Deviation 29,7 34,4 4,5 52,5
Median 163 89 11,1 196
Business Mean 169 91 10,6 180
SE 43,9 29,9 4,7 40,0
St.Deviation 59,8 35,5 5,5 44,4
Median 170 93 10,9 182
Leaders Mean 188 137 7,0 240
SE 42,2 38,7 3,6 71,0
St.Deviation 53,8 43,5 4,9 96,2
Median 167 151 5,4 196
Intensi Mean 178 116 7,7 224
fiers SE 31,5 32,3 3,6 72,1
St.Deviation 38,5 36,9 4,4 88,4
Median 179 103 6,5 205
MEAN 175 108 9,0 211
tener-oriented talk needs a major type of rhythm which provides for the scenario words to be prosodically prominent and recurring at regular intervals in the most predictable salient positions in ip. The prominence category is gradient according to the order of occurrence (each degree of givenness is associated with decrease of prominence but a drop on second one is most noticeable); according to the pragmatic value of the words representing different concepts/frames of the scenario (the 'leaders' frame is decisively most prominent, next come 'success' and 'business'); according to the position in ip (the final position is most favorable, the initial one comes next, but it is very rare in the corpus as the scenario words gravitate towards cognitively more justifiable medial and final positions).
Rhythm proves to be an inherent property of ongoing discourse, not disturbing the information flow but, rather, regulating and rearranging it to facilitate comprehension.
References
1. Chafe, W., "Cognitive constraints on information flow," in R. Tomlin [Ed.] Coherence and grounding in discourse, Amsterdam, John Benjamins, 21-52, 1987.
2. Lambrecht, K., Information structure and sentence form, Cambridge, Cambridge University Press, 1994.
3. Pierrehumbert, J.B. and J. Hirschberg, "The meaning of intonational contours in the interpretation of discourse," in P.R. Cohen, J. Morgan, and M.E. Pollack, [Eds] Intentions in communication, Cambridge, MIT Press, 271311, 1990.
4. Ladd, D.R., Intonational phonology, Cambridge, Cambridge University Press, 1996.
5. Baumann, S., "Information structure and prosody: linguistic categories for spoken language annotation," in: S. Sudhoff et al. [Eds] Methods in empirical prosody research, Berlin, New York, Walter de Gruyter, 153-180, 2006.
6. Baumann, S. and Riester, A., "Annotating information status in spontaneous speech," in Speech Prosody 2010, Chicago, Online: SpeechProsody2010.Illinois.edu
7. Hirst, D. and Di Cristo, A., "A survey of intonation systems", in D. Hirst and A. Di Cristo [Eds.] Intonation systems, A survey of twenty languages, Cambridge, Cambridge University Press, 1-44, 1998.
8. Gussenhoven, C., The phonology of tone and intonation, Cambridge, Cambridge University Press, 2005.
9. Halliday, M.A.K., "Notes on transitivity and theme in English", Part 2, Journal of Linguistics 3: 199-244, 1967.
10. Valimaa-Blum, R., Cognitive phonology in construction grammar, Berlin, New York, Mouton de Gruyter, 2005.
Table 9. Descriptive Statistics: Occurrence 3
Concepts Parameters Э
Fomax Fomin Foint SyllD
Success Mean 150 96 7,7 253
SE 27,8 9,1 4,0 35,8
St.Deviation 31,8 10,8 4,2 38,2
Median 152 98 7,4 256
Business Mean 156 99 8,2 189
SE 33,4 25,2 3,5 45,2
St.Deviation 36,8 32,8 4,4 55,9
Median 154 101 8,3 184
Leaders Mean 195 129 7,9 250
SE 40,2 40,0 3,9 72,2
St.Deviation 55,9 44,6 4,8 88,7
Median 148 98 5,3 233
Intensi fiers Mean 179 127 6,4 195
SE 33,3 39,8 2,8 56,9
St.Deviation 43,9 45,1 3,2 71,9
Median 173 123 5,5 175
MEAN 171 113 7,5 222
Table 10. Descriptive Statistics: Occurrence medial
Concepts Parameters medial
Fomax Fomin Foint SyllD
Success Mean 159 101 8,5 221
SE 26,7 24,1 3,9 55,2
St.Deviation 31,6 31,8 4,2 65,5
Median 157 104 8,1 219
Business Mean 161 96 7,2 163
SE 42,0 25,6 3,6 33,9
St.Deviation 54,9 31,0 4,6 45,6
Median 163 99 7,7 166
Leaders Mean 184 128 6,5 218
SE 37,5 37,4 2,9 53,0
St.Deviation 46,1 42,0 3,6 67,5
Median 144 115 4,1 176
Intensi fiers Mean 179 126 6,2 200
SE 31,5 39,7 2,7 65,8
St.Deviation 40,9 37,9 3,3 83,2
Median 185 117 5,2 175
MEAN 171 113 7,1 201
Table 11. Descriptive Statistics: Occurrence final
Concepts Parameters final
Fomax Fomin Foint SyllD
Success Mean 180 111 10,1 221
SE 37,6 47,6 4,9 67,8
St.Deviation 64,5 50,2 5,3 76,8
Median 182 113 10,4 220
Business Mean 175 97 10,8 217
SE 47,7 31,5 4,9 42,0
St.Deviation 63,7 37,7 5,6 63,2
Median 172 94 10,2 212
Leaders Mean 190 109 10,2 306
SE 44,6 41,4 5,3 10,8
St.Deviation 62,0 47,9 6,9 99,4
Median 200 79 6,8 283
Intensi fiers Mean 171 90 11,2 257
SE 45,6 35,1 5,5 60,7
St.Deviation 50,0 40,8 6,2 60,1
Median 151 70 8,5 256
MEAN 179 102 10,6 250
