Kutatás / Research
Together with members of the Univeristy of Cologne, our research groups studies articulatory and acoustic features of vowels in several different speech modes and phonetic contexts. Our work focuses mainly of vocalic variation casued by factors within and outside the speech stream.
The collaboration is lead by Prof. Dr. Reinhold Greisbach (University of Cologne), and Andrea Deme (ELTE Eötvös Loránd University) (previously by Alexandra Markó). We received funding from the TKA (Tempus Közalapítvány, Hungary) and the DAAD (Deutscher Akademischer Austausch Dienst, Germany) several times.
Projects under review:
2025-2026 Audio-visual perception of vowels in Hungarian and German / Magyar és német magánhangzók audio-vizuális feldolgozása
Current project:
2023-2024 Phonetic aspects of fast speech in Hungarian and German / A gyors beszéd néhány fonetikai jellemzője a magyarban és a németben
Previous projects:
2020-2021 A cross-linguistic articulatory and acoustic analysis of the effect of vowel space density on the variability of vowels / A magánhangzó -készlet sűrűségének hatása a magánhangzók variabilitására: összevető artikulációs és az akusztikai elemzések
2014-2016 Articulatory and acoustic analysis of high-pitched sung vowels in Hungarian and German by means of electromagnetic articulography, ultra sound and inverse filtering of the acoustic signal. (lead by Alexandra Markó)
Previous and current members of the groups:
Márton Bartók
Andrea Deme
Julianna Jankovics
Kornélia Juhász
Alexandra Markó
Sarolta Murányi
Panna Soós
Zsuzsa Szánthó
Zsófia Weidl
Szabina Zsoldos
Fanny Burkhardt
Malte Dreist
Julia Eichholz
Solveigh Janzen
Reinhold Greisbach
Johanna Metz
Michelle Meier
Isa Samira Winter
Research Topics
For publications connected to the projects below, go to Publications page.
Vowel variation as a function of vowel desnity under vowel-to-vowel coarticulation
We analyze vowel variation induced by carryover V-to-V coarticulation under the effect of pitch-accent as a function of vowel quality (using a minimally constrained intervening consonant to maximize V-to-V effects). We tested if /i/ is more resistant to coarticulation than /u/, and if both vowels show increased coarticulatory resistance in pitch-accented syllables. Our approach is unprecedented in the sense that it involves the analysis of parallel acoustic (F2) and articulatory (x-axis dorsum position) data in a great number of speakers (9 speaker), and both real words of Hungarian, and non-words. To analyze the degree, direction, and magnitude of coarticulation, we calculated the differences between coarticulated and non-coarticulated vowels. We also gauged the dispersion of vowels withon the vowel space, andcalculated across context variability as relative standard deviation of midpoint F2 values and dorsum positions, again, in both domains.
Vowel and consonant durations and vowel contrasts in fast speech
Fast speech is the result of speech sounds produced shorter. However, it is expected that in terms of duration, not each segment may be reduced to the same extent, especially if duration serves linguistic functions, as in e.g., German, and Hungarian, where vowel length contrast is phonologically distinctive. In Hungarian, phonologically short and long high vowels, e.g., /i iː/ /u uː/, are traditionally assumed to be distinguished primarily by duration, while short and long low vowel pairs /ɒ aː/ and /ɛ eː/ differ also in their quality. In German, the situation is the other way round: (in accented syllables) we find no quality difference between the low vowels /a aː/, while there is a simultaneous durational and quality difference in high vowels, e.g., in /ɪ iː/, and /ʊ uː/. In vowels, temporal reduction in fast speech is also expected to be accompanied by some degree of spectral changes, due to target undershoot. As a result, increased speech rate is expected to endanger the vowel length contrast both in the temporal and the spectral domain. To this hypothesis, we find scarce, and to some extent, inconclusive evidence in Hungarian and German. With respect to duration, in Hungarian, there is some evidence that long vowels reduce to a higher degree than short vowels. Similarly, in German, long (tense) vowels were found to reduce more than short (lax) ones. In vowel quality, German short and long vowel spaces were found to be affected by fast speech in a similar fashion: they were both reduced, or (as opposed to expectations) increased (in different dialectal regions). In Hungarian, we find no systematic analysis of the spectral changes in fast speech. In our stiudies, we hypothesize that increased speech rate induces reduction of the vowel length contrast in Hungarian and German. Our aim is to explore if this reduction emerges differently in these two, typologically unrelated languages, where the phonological vowel length contrast is expressed using similar means, but in a different implementation.
Durational aspects of fast speech in German
Phonetic and phonological descriptions of the sound inventory and sound structure of a language are based on precise pronunciation by speakers of that language or (mechanically speaking) articulation at a slow speech rate. If the speaker speeds up, articulators have to move faster in order to produce more sounds in a shorter period of time. However, there are mechanical constraints on the speed and acceleration required to move an object, known as inertia. The inertia of the articulators limits the achievable speech speed. If the speaker tries to increase speed even further, it can only be achieved by reduction: lenition, and/or omission of some articulatiory movements that may also lead to missing sounds in the flow of speech (elision). Classical methods aiming to analyse these reduction phenomena focus on spontaneous speech, which is assumed to be faster than “normal” speech. For German it is known that spontaneous speech contains many sound elisions in unaccented syllables. In our studies, we analyse how some German vowels, and consonants in phrase accented syllable onset position are realized in fast speech, which we elicited using a controlled read aloud method.
Vowel and consonant durations in fast speech
In fast speech, speech sounds are produced shorter. However, according to previous studies, i) vowels are more compressible, and reduce more than consonants. In languages that show phonemic vowel length contrast, like Japanese, and Hungarian, vowels are also expected to vary in the extent of reduction as a function of their phonological length: in fast speech, ii) long vowels are expected to reduce more than short vowels, while iii) the vowel length contrast (as expressed in duration ratio) does not neutralize completely, as shown for Japanese. We analyze consonant and vowel durations produced and tests these three hypotheses.
Vowels in soprano singing
Sopranos are often required to sing at extremely high pitch. The articulatory maneuvers, however, singers use to increase fundamental frequency (f0) to these extremes are antagonistic to the production of certain vowel qualities. As a result, as singers raise the fundamental frequency of their voice, the vowels they produce become more and more distinct from their spoken counterpart — and merge into a unifrom /a/-like quality — both in production and perception. No wonder that we often experience serious problems understanding texts sung by these incredible performers…
We analyze articulation, acoustics, and percpetion of high-pitched sung vowels produced by professional (female) soprano singers in singing and speech, using all the nine qualitatively distinct vowels of Hungarian. In the articualtory studies we obtain EMA data, and compare vowels of German and Hungarian as well, which languages are different with respect to the density of their vowel spaces.
Vowels in motherese
According to Lindblom’s H & H theory, contrast is a crucial factor in speech production: in situations, in which the listener is assumed to be hindered in the speech perception task (due to e.g., adverse listening conditions or hearing loss), the speaker exaggerates contrast, hence (s)he makes the encoded information more accessible to the interlocutor. On this basis, linguistic contrast, e.g., the vowel length contrast is expected to be exaggerated in such speech styles as Lombard speech, i.e., speech produced in background noise, or in motherese, where the interlocutor is at an early stage of language acquisition.
In the Neurocognitive predictors of early language development [a korai nyelvfejlődés neuro-kognitív előjelzői] (NKFIH-K-115385) projekt we address the questions, if contrast exaggeration is observable in Hungarian primipara and multipara mothers’ infant-directed speech, if the contrast enhancement shows language-specific traits, and if contrast exaggeration is in any way correlated with the infant’s age and stage of language acquisition.
Voicing of the laryngeal fricative
The laryngeal fricative /h/ is claimed to be (breathy) voiced in some specific contexts generalizable over many languages. Yet, to date, little is known of the facilitating contexts of the voiced counterpart, not to mention the acoustic quality of this special voicing, and the articulatory and/or linguistic factors influencing it.
We investigate voicing characteristics of /h/ in Hungarian observing fine-grained phonetic variability, while we also try to take into account the articulatory mechanisms speakers use to produce vowels, voicing, and glottal frication.
Sociophonetics — identification of innovative language users
Investigation of language change is a key issue in sociolinguistics, but is also very challenging to tackle, as the process of change very often stretches over several lifespans. Nevertheless, sociolinguists since Labov try to identify and investigate ongoing change, mainly by identifying innovative and conservative speaker groups, and the analysis of their linguistic behavior. This way of research, however, can easily turn into being circular, if the variables of interest are also used as variables designating the innovative and conservative speaker groups in which the use of these specific linguistic variants is analyzed afterwards.
We contributed to this field of research by adapting a test from social psychology, the implicit association test (IAT), that is capable of detecting “more lenient”, and “less lenient” listeners with respect to the innovative and conservative linguistic forms they tolerate. The detection via LI-IAT is done independently of the linguistic variable(s) we set to examine from the perspective of change, and thus allows for the investigation of linguistic change more reliably.
Fur further topics, check out the site of the Department of Phonetics of Eötvös Loránd University, and the MTA-ELTE “Lendület” Lingual Articulation Research Group!