Oliver Niebuhr 7th International Conference of Speech Prosody - PowerPoint PPT Presentation

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University “A little more ironic” – Voice quality and segmental reduction differences between sarcastic and neutral utterances Oliver Niebuhr 7th International Conference of Speech Prosody Dublin, Ireland Thursday, 22 May, 2014 22.05.2014 1 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Introduction  Ironic statements convey the opposite of what the speaker has put into words.   Since the number of ironic utterances is in principle infinite   Since the prosodic meanings have to be adjusted to clash with each of these verbal meanings  ...it is obvious that there cannot be THE prosody of irony  Phonetic analyses at least have to distinguish different subtypes of irony  The most frequently investigated subtype  clash between positive verbal and negative prosodic meanings = ‘sarcastic irony ’‚   subject of the present study 22.05.2014 2 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Introduction  Brief overview of (some) previous findings: Expression of sarcastic irony =  longer utterance durations (i.e. lower speaking rates)  F0 patterns are lowered and narrowed (in Western Germanic languages like German and English  but raised and widened in Cantonese and Roman languages)  Intensity range is lowered and flattened (i.e. less intensity dynamics), though this is less consistent across studies and languages  (1) But what about “the 4 th prosodic dimension“  Voice quality?  So far only anecdotal evidence (e.g., Muecke 1978): compared with sincere/neutral utterances, ironic ones have a softer, breathier voice  can we find experimental evidence for this claim?  = first aim of the present study 22.05.2014 3 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Introduction  (2) Given our increasing knowledge of the interrelations between seg- ments and prosodies, if/how does irony affect the pronunciation of words? eee ?aIdoUntnoU “I do not know ” = … but I will help you finding out = … and I don‘t care gUtenmoAgen gUmo: “ Guten morgen ” = just greeting, not talking please = hi, let‘s have a short chat • other examples: “ Entschuldigung ” (I am sorry)  Honest excuse vs. empty/routine phrase, and “ Danke ” (thank you) 22.05.2014 4 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Introduction  (2) Given our increasing knowledge of the interrelations between seg- ments and prosodies  Niebuhr & Kohler (2011) and Graupe et al. (2014): Degree of speech reduction is not just determined by competing demands of economy and comprehensibility (cf. H&H theory)   Degree of speech reduction also conveys speaker attitudes towards dialogue partner and/or content of message  Expressing (sarcastic) irony implies conveying: “I distance myself from the semantic content of my utterance”  Degree of speech reduction in ironic utterances higher than in sincere/neutral utterances?  = second aim of the present study 22.05.2014 5 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Method • Basis of the production study: • 20 Standard German target sentences, designed according to 4 criteria: • (1) Syntactically simply constructions of 4-8 frequent words •  facilitated a fluent reading with clear and consistent prosody • (2) Each target sentence included at last 2 phonetically weak forms •  by definition unstressed, prone to high degrees of speech reduction • for example: pronouns, aux. verbs, modal particles, articles, prep. etc. • (3) Each target sentence could be directed at an interlocutor • (4) Each target sentence made positive statements about the interlocutor or a third party •  basis for realizing sarcastic irony by inverting the positive propositions 22.05.2014 6 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Method • Basis of the production study: • 20 Standard German target sentences, designed according to 4 criteria: • Das ist aber mal interessant (That’s so interesting) • Das wird bestimmt spannend (This will surely be exciting) • Das läuft ja super (This works out perferctly) • Das ist ja lustig hier heute (It is indeed funny here today) • Er ist besonders wichtig (He is a very important person) • Das riecht richtig lecker (It smells really delicious) 22.05.2014 7 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Method • Target sentences produced by 10 Standard Northern German speakers • 5 males, 5 females • 23-29 years old • Recordings were conducted in individual sessions in the speakers’ homes • Interlocutor was a good friend / fellow student of all 10 speakers  Easier to elicit informal speech and the expressiveness required for • sarcastic irony • Speakers were randomly assigned to 2 subgroups •  first neutral, then ironic production of the 20 target sentences •  first ironic, then neutral production of the 20 target sentences • Sentence order was randomized after each round 22.05.2014 8 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Method • Prior to recording: • Speakers had time to familiarize themselves with the sentences • The concept of sarcastic irony was explained by means of comic strips  no audio examples You This expresses „Das ist ja mal interessant“ think? my personality (That‘s so interesting), 6x www.maescot.de 22.05.2014 9 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Analysis 1: Filtering out unclear cases • All 400 sentences were presented in differently randomized orders • … to 4 naïve Standard German listeners • Task: judge spontaneously for each sentence whether it was meant sincerely or ironically • Sentences were only kept in the final sample, if they were identified  75% correctly as being neutral or ironic neutral ironic 71 sentences 136 sentences 71 sentences 22.05.2014 10 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Analysis 2: Acoustic-prosodic measurements • F0 minimum of the sentence, measured in semitones relative to 100/200 Hz for male/female speakers • F0 maximum of the sentence, measured in semitones relative to 100/200 Hz for male/female speakers • F0 range in semitones, i.e. (b) subtracted by (a) • average F0 level, calculated in semitones relative to 100/200 Hz for male/female speakers • intensity minimum in dB within a non-silent sound section of the target sentence, • intensity maximum in dB, • intensity range in dB, i.e. (f) subtracted by (e), • average intensity level in dB, disregarding silent (sound) sections of the target sentence, • voice quality in terms of the amplitude difference in dB between the first and the second harmonic (H1-H2), measured in the midpoints of all vowels of a target sentence and then averaged across these vowels, • tempo, represented by the total sentence duration 22.05.2014 11 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Analysis 3: Degree of reduction • Analysis of the degree of speech reduction was done on an auditory basis •  2 trained phoneticians listened to the 2x71 sentences individually • … and counted the number of reductions (assimilations, elisions, lenitions) • … with reference to the canonical form specified in the DUDEN lexicon, published by the Institute for the German Language (IDS), Mannheim 22.05.2014 12 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Results 1: F0 measurements Compared with neutral/sincere sentences, ironic sentences had a lower and less melodic pitch contour • F0 max strongly lowered (about 4st) • F0 min weakly lowered (about 1.5 st) •  narrower F0 range •  overall lower F0 level 22.05.2014 13 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Results 2: Intensity measurements Compared with neutral/sincere sentences, ironic sentences had a reduced loudness level • Int. max lowered (about 7 dB) • Int. min lowered to the same extent •  lower intensity level •  constant intensity range 22.05.2014 14 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Results 3: Duration and voice-quality (VQ) measurements Compared with neutral/sincere sentences, ironic sentences had longer overall durations • = lower speaking rate (about 30%), also when measured in syll/sec • 7.1 (neutral)  5.5 (ironic) syll/sec Compared with neutral/sincere sentences, ironic utterances showed overall no systematic voice quality (VQ) differences, but had more variable voice qualities •  Levene test: clearly heterogeneous variances (p<0.001) for neutral vs. ironic 22.05.2014 15 Oliver Niebuhr

Analysis of Spoken Language, Dept. of General Linguistics, Kiel University Results 4: Degree of speech reduction • Compared with neutral/sincere sentences, ironic sentences showed a significantly higher degree of reduction (strong effect, but not consistent) 22.05.2014 16 Oliver Niebuhr