"Audio-Lingua": pronunciation improvement
through sound perception training
ETE: DE NOUVEAUX STAGES
INTENSIFS CHAQUE SEMAINE
SEPTEMBRE: LES STAGES
PROGRESSIFS DEMARRENT
|
A partir de
400 €
POUR 30 HEURES
|
When learning a new language, various components have to be considered: grammar (morphology
and syntax), lexicon and phonetics. Whereas in grammar and lexicon considerable progress can be
made in a cognitive way (e.g. through the study of texts and memorisation), this is not the case with
sound discrimination ability and, consequently, pronunciation and intonation. Here adult learners
feel they face an insuperable barrier.
The importance of listening in language learning and its role in helping the learner overcome this
barrier has long been recognised. Indeed, there is hardly an institution without a language laboratory
or tape-recorders. But we also know from our own experience that there are cases when students
simply do not succeed in producing words or sentences correctly even after hours in the language
laboratory. Is it because they just fail to reproduce the correct sounds or is it possible that they do
not hear them correctly? Voice production is closely connected to auditory comprehension. This
means that pronunciation problems are not always articulation problems and, therefore,
pronunciation exercises can be of little help if, from the very beginning, the auditory input is not
being decoded correctly. Consequently, how can traditional pronunciation training help if the
student keeps hearing sounds incorrectly? Pronunciation exercises should, therefore start with a
battery of exercises aiming at sound perception training!
Audio-Lingua" is the name of a research project which tested the efficiency of the SPT method
("Sound Perception Training"). This method proved to have clearly positive effects on oral
production (pronunciation and intonation) of the target language. The project was supported by the
European Union through the Lingua Programme, which is now called the Socrates Programme, and
the support lasted for the maximum three years (1993, 1994 and 1995). The project had two goals:
1. to test the efficiency of the SPT method in terms of pronunciation improvement through
acoustical stimulation;
2. to develop didactic pronunciation material (book with tapes/CD) for the target languages
German, Italian, Dutch and Spanish.
Apart from the co-ordinating institution, the University of Bologna (School for Interpreters and
Translators in Forlì), a further five universities and two non-university institutions in Italy,
Germany, Belgium, the Netherlands and Spain collaborated.1
1. Listening and language learning: the principles of the SPT method ("Sound Perception Training")
The SPT method aims at making people more perceptive to the sound system of the target language.
Since comprehension is always ahead of production, aural comprehension is ahead of oral
production. The results of our project have confirmed that students who undertake this training have
a far better pronunciation than those who work with pronunciation exercises only.
Alfred Tomatis, the French specialist in ear, nose and throat diseases, carried out research on sound
perception mechanisms in the 1950s. He compared voice spectra (spectrograms) with listeningcurves (audiograms) and found a close correspondence between the distribution of Hz-frequencies2
for hearing and production. Those frequencies which could not be perceived by the ear were also
missing in the voice spectrum3. For a non-defective ear this means that there is an automatic
selection in the registration of auditory input, which is revealed in the quality of voice production4.
The auditory selection can be voluntary, for example when we succeed in understanding someone
in a very noisy environment (where, if one measured it, the volume of the noise might be higher
than the voice we are trying to understand)5. The auditory selection can also be habitual, for
example, a mother might sleep through all kinds of noises, but when her baby merely whimpers, she
is immediately wide awake. Someone else might not even be aware of this sound because they do
not attach any meaning to it.
One encounters the same phenomenon of selective hearing when learning a foreign language. Our
auditory system for languages
is strictly connected to the specific language(s) we are brought up with: namely, our mother tongue
or, in the case of bilinguals, our "parent" or "caretaker" tongue(s)6. Tomatis calculated "envelope
curves" for different languages, which show the characteristic mean frequency spectrum of a
language7. Other scientists, such as Delattre, made similar calculations and compared differences in
vowel formants in various languages8. Certain foreign language sounds which do not exist in one's
mother tongue are heard and interpreted in the closest way they come to these sounds in the listener'
own native language sound inventory. This substitution strategy is one of the main reasons why the
learners of a foreign language speak the language with a "foreign accent". Often, they do not even
realise this: their mental decoding mechanism does not react to certain unfamiliar foreign language
sound systems.
Young children do much better than adults in terms of phonological development. When acquiring
more than one language at a time, they have no "foreign accent" in any of the languages. Adult
learners (learners in the so-called post-critical phase, from puberty onwards) have great difficulties
with new sound systems. In fact, in other areas, such as grammar or vocabulary, it has been shown
that older learners do better because of their use of cognitive skills9. In pronunciation, cognitive
help is useful only up to a certain degree (e.g. through visualisation of the speech organs, melody
curves, verbal description of the production of certain sounds).
Correct pronunciation can only rarely be achieved automatically or by mere imitation. Adult
learners already have fossilised listening habits and not everything which is heard can be
reproduced correctly because the sounds of the foreign language pass through a filter which consists
of the sound inventory of the mother tongue10. The language laboratory is of limited use if students
are not able to hear correctly, i.e. if they are not able to discern unfamiliar sound patterns. New
listening and hearing habits have to be developed in order to improve productive competence.
These new listening models can then control and correct one's sound production and train the
speech apparatus.
As a result of the above-mentioned differences in language perception, a speaker of one language
will not necessarily perceive and decode all the phonological input from another language, and,
consequently, will not be able to encode it into oral production either. For example, the German
phoneme /ö/ (as in "böse") does not exist in many languages and is often substituted by the sound
which comes closest to it in the learner's own language through the generalisation of sound patterns:
either /e/ ("bese") or /o/ ("bose"). Indeed, differentiation tests11 have shown that a large number of
students who make errors in this kind of speech production cannot easily differentiate the sound
when presented acoustically. The same is true for other sounds or sound combinations which the
learner is not familiar with. Another example is the English fricative [—] as in "father". Germans
frequently have problems with this sound and substitute it with a sound from their own sound
inventory - [z] or [t] - which comes close to it.
Tomatis developed an electronical instrument (which he called "electronic ear") that helps to
stimulate sound perception capability and thereby increase the amount of perceived information.
Improved sound perception leads to an improvement in listening ability, to a better comprehension
of the language in general and better language production (e.g. pronunciation), i.e. language
learning is facilitated. The electronic ear educates the automatic auditory selection mechanism and
trains the ear to react in order to perceive unfamiliar sound patterns. In this way, it modifies one's
way of hearing and hence one's way of speaking. This instrument, is called "Sound Perception
Trainer" or "SPT" and completed with sound-sources (microphone, tape player)12 and special
headphones.
2. Modifications of sound by the SPT instrument
The training, as developed by Tomatis, re-educates the ear's organs so that the innate listening
abilities are stimulated. This occurs by means of the electronic modification of the sound and sound
transmission. The SPT can be programmed either individually or for homogeneous groups in
accordance with the following four modifications:
2.1 Filtering
For the pure recognition of vowels, only the main formants (the lower formants) are important13.
The differences in the timbre of voice, on the other hand, lie mainly in the higher zones, especially
in the secondary formants. They do not carry phonemic information and tend to be overshadowed
by the strong main formants so that fine differences are not perceived. Through filtering from low
to high frequencies (from 500 to 9,000 Hz), the learners are taught to become perceptive in the area
of the overtones and to learn to distinguish sounds better.
2.2 Transmission of sound through two channels
The instrument uses two channels for the transmission of sound14. The sound structure in channel 1
(C1) enhances the frequency profile of the learner's way of speaking and hearing: lower frequencies
are strengthened, higher ones weakened. In channel 2 (C2) the opposite occurs: the lower tones are
weakened and the higher tones strengthened. By means of sudden "jumps" between the two
channels, the ear is forced to adapt to the unfamiliar sound structure of the target language and the
automatic selection mechanism described above is eliminated.
2.3 Bone conduction of sound
The headphones of the SPT have an additional transmitter for the bone conduction of sound in order
to intensify sound perception by directly leading the sound vibrations to the inner ear through the
skull. In this way there are three sources of sound transmission: the right ear, the left ear and the
bone transmitter on top of the skull. Sound not only travels through the air, it also travels almost ten
times faster through dense material like bone16. In the hearing/listening process, sounds are always
received through bone and air conduction17.
2.4 Intensity differentials in the output
The volume with which the manipulated sound message reaches the students' ears differs from one
transmitter to another. It is especially reduced in the left ear. This modification is made because the
two ears do not receive sounds in the same way. According to Tomatis, the right ear has a dominant
function: it controls and guides the aspects concerning intensity, timbre, intonation, melody and
semantics18. One reason for the right ear's dominance is that the length of the nervus recurrens, the part ofhe nervus vagus which controls the tympanum, pharynx and larynx, is 40 to 50 cm longer on
the left side than on the right. This means a remarkable delay of the nervous signal19.
Fig. 2: Abdominal part of the nervus vagus: the two recurrens nerves
Tomatis gives another reason for the preference for the right ear during the training with the
apparatus: the areas of the brain responsible for the analysis and recognition of speech, and the
processing of phonetic information when acoustically presented are in the left hemisphere. Signals
perceived by the left ear would first arrive in the right hemisphere and then be sent to the left for
decoding. This means a loss of quality if the speaking process is controlled by the left ear20.
3. Set-up of the SPT training programme
The sound perception training programme precedes or accompanies conventional language study. It
is not meant to be a competitive or substitute approach - it is a complementary or remedial one
aimed at overcoming the barrier of acquiring the correct pronunciation. After a certain number of
listening sessions, the training is complete and should not lose its effect over time. The SPT should
be seen as an instrument for the correction of rigid listening habits, educating the automatic auditory
selection mechanism which modifies hearing and hence speaking. The training, therefore, provides
a "shortcut for learners" (McLaughlin, 1987, p. 48).
The training which first stimulates the listening ability and is then integrated into pronunciation
exercises needs to follow a regular schedule. Each session, which takes place on a daily basis, lasts
a minimum of 60 and a maximum of 120 minutes (120 minutes turned out to be too long for
students who have to find spare time in their university schedule). The total time involvement is
around two months, but this can be extended if necessary.
Our training programme uses pronunciation exercises and texts which were developed within the
"Audio-Lingua" project21. It is divided into two stages. The students start with a receptive stage and
then proceed to a productive stage after a pause of two to three weeks. During the receptive stage,
the learners listen to tapes with texts in the target language. The SPT performs the modifications
mentioned above. The learners are not always consciously aware of these modifications, but the
SPT forces their ear to "open up" to unfamiliar sounds.
During the productive stage, the students listen to recordings in the target language (words,
phrases, sentences and short texts), repeat them and also read aloud themselves. While they are
repeating or reading, the SPT instrument is constantly adapting their ears – and thus their voices –
to the typical frequency spectrum of the target language so that the training has a double effect: It
affects both listening and speaking and gives the students the possibility of self-monitoring. The
students learn to experience how a native speaker of the language they are studying hears the
language, and this gives them an advantage over someone who tackles it with only his/her own
ethnic sound inventory to rely on. During and after the training, students are more likely to speak
with the correct pronunciation and intonation of the foreign language.
4. Testing of the SPT method within the "Audio-Lingua"
project
The study was carried out in five universities with three testing groups, all students of German:
Forlì, Antwerp, Saragossa, Milan and Brescia. In all universities the same material was used
(German as a target language22) since German was the first pronunciation course to be developed
within the project and in this way the results are comparable. Potentially, the results can be
transferred to any target language and any source language. The three testing groups had the
following profile:
1. PT group: The students worked with the SPT instruments using the pronunciation
material developed within the project (German as a target language). A teacher was
only occasionally present but did not intervene or help.
2. Control group: The students worked in a conventional language laboratory with the same
material (German as a target language, as developed within the project), but without any
sound manipulation. Here, a teacher was always present, and controlled and corrected the
students' performance. – Saragossa did not have a control group.
3. Zero group: The students of this group did not undergo any specific pronunciation
training, they merely carried on with their everyday university courses in German. –
Forlì did not have a zero group.
All students of all three groups underwent an initial and final test; the experimental group took an
additional interim test. Initially, there were 203 people; by the end, the number was reduced to 158,
128 of which were regarded as statistically significant23. All tests consisted of a listening test (sound
perception threshold, selectivity and laterality), a "receptive test" (phoneme differentiation, multiple
choice and word stress tests) and a "productive test" (reading aloud, repetition and free speech)24.
| listening test · |
· sound perception threshold
· selectivity
· laterality |
"receptive "test
(in the language lab) |
· phoneme differentiation
· multiple choice (sound discrimination)
· word stress |
"productive" test
(individual test, tape
recordings for segmental
and suprasegmental
evaluation) |
· reading aloud (dialogue and prose)
· repetition
· free speech |
| questionnairequestionnaire |
· student's own impression |
Fig. 4: Tests which all participants had to undergo
5. First results
The Dutch testing institute CITO helped with the set-up of the tests and is evaluating the test results
using the one-parameter Rasch model25 which is a model from the Item Response Theory (IRT).
The Rasch model specifies a relationship between observable test performance and an unobservable
trait or ability assumed to underlie this performance26.
Each of the tests was assigned 248 items27 for statistical evaluation. Each single item was given an
attribution which encodes the result of the students' performance on the single item (the total data
set is over 1,600 item-person encounters). By applying the IRT model, the relationship between the
observable result of a test and the ability of the person undergoing the test can be calculated28. The
difference in this variable between the ability of a subject and the difficulty of an item determines
the probability of a correct response. Thus the elaborated test results can be described in ability
scores, which are closely connected to the linguistic competence of the participant. Even though
CITO is still working on the preparation of the final report, in which single aspects of the overall results
will be looked at, compared and discussed, they have so far produced figures which visualise
the overall results (all the tests together) on an ability scale, which shows the students' ability to
learn or make progress in the foreign language on the basis of their test results. The following table
shows the evaluation of the linguistic ability of the participants in T1 (initial test), T2 (interim test,
only in Forlì) and T3 (final test)29:
Fig. 5: Evaluation of the linguistic ability of the participants
Results show that in all the testing groups at all partner universities, the SPT training led to an
increase in phonetic ability, with the SPT groups showing a significant advantage in comparison to
the control groups and the zero groups.
Let us take a closer look at the growth in the ability score of the single groups in Forlì, Antwerp and
Saragossa:
Forlì growth SPT group: 62.35 %, control group: 28.90 %
Antwerp growth SPT group: 64.07 %, control group: 45.12 %
Saragossa growth SPT group: 114.42 %, zero group: 55.20 %
The high numbers in Saragossa probably result from the fact that the level of German of both
groups was very low to start with. This led to a faster improvement than in Forlì or Antwerp (both
schools for interpreters and translators).
When comparing the overall results of the SPT groups and the control groups at T 1 (initial test)
and T 3 (final test, about three months later), the results provide an impressive picture: in general,
the ability of the SPT groups increased by 70.95 % whereas the control groups had a mean increase
of 36.52 %:
Fig. 6: Ability increase between T 1 and T 3 of all SPT and control groups
These results show clearly that the method might mean a remarkable innovation in foreign language
learning and teaching. If transferred into time units, a student working with the SPT method would
be able to achieve the same results in almost 50 % less time as someone from the same course who
worked without any specific sound perception training.
6. Conclusion
The test results show that so far students who undergo the SPT training have a significantly better
pronunciation than those who work using only pronunciation exercises. Sound perception training
turned out to be a valuable instrument to support and facilitate language courses and to accelerate
oral production ability. It may be of considerable help in a multilingual comprehension approach
because the training literally "opens up" the ear and the learner acquires greater sensitivity towards
auditory input in general.
Four groups of learners in particular may be able to profit from the SPT method:
1. Adult learners, who are known to have more difficulties than children in learning foreign
languages in terms of pronunciation and intonation;
2. Advanced learners who already have a high level of language proficiency and see
themselves at a level from which they cannot progress without additional help;
3. Isolated learners who have to learn a language autonomously, without the presence of a
(native) teacher or the possibility to spend time in the country of their target language.
4. Young learners with special hearing difficulties (remedial teaching).
It is also believed that average learners, at whatever stage of proficiency, can greatly profit from
systematic or occasional training with the materials developed for the SPT30.
Footnotes
1 Our partner universities were:
1. Katholieke Vlaamse Hogeschool (Antwerp, Belgium)
2. Gerhard-Mercator-Universität-Duisburg (Duisburg, Germany)
3. Università La Sapienza (Rome, Italy)
4. Università Cattolica del Sacro Cuore di Milano (Milan and Brescia, Italy)
5. Universidad de Zaragoza (Saragossa, Spain)
The partner universities tasks were:
· to test the SPT method;
· to develop pronunciation material for the target languages German, Italian, Dutch and
Spanish.
The names and tasks of the two non-university institutions were:
1. Diapason Ltd., Milan, Italy, which provided us with the electronic instruments and the
technical know-how;
2. CITO (National Institute for Educational Measurements), Arnem, Netherlands: consultation
in the setup of the tests and the evaluation of the data and test results.
2 Frequency means the number of sound waves per second and is measured in Hertz. The higher the
frequency, the higher the sound. A human ear can perceive from 16 up to 16,000 Hertz (sometimes
20,000 Hertz: cf. v. Essen 1979, p. 152)
3 See the research by M. Pierre and P. Grassé , presented by R. Husson , 1957.
4 To give some figures: according to J. Wendler and W. Seidner (1987, p. 86) total auditory
registration is ca. 100,000 bit/sec, whereas the auditory perception is 50 bit/sec, which is only a
2,000th of the total information.
5 Psychologists call this phenomenon of auditory selection "cocktail-party effect"
6 Studies show that after three months a baby starts to form a listening pattern which is fully
developed by the age of 16 (Caneau 1992, pp. 19, 39).
7 Cf. Tomatis, 1991.
8 Cf. Delattre, 1965.
9 As Krashen says, the adult learners' poorer language acquisition ability results in a high affective
filter, which works as a barrier against acquisition. Adults use a cognitive controlling instrument,
which he calls "monitor". (Krashen, 1985)
10 In the 1950s, Nicolay Trubetzkoy talked about "phonological grids" which function as filters:
unknown sound characteristics of the foreign sound system would pass through the grid (which
means that the decoding mechanism fails) whereas those familiar to the learner would be "held up".
(Trubetzkoy, N.S. Grundzüge der Phonologie. Göttingen: Vandenhoek & Ruprecht, 1958)
11 The kind of differentiation tests which were taken by the students at the University of Bologna
were, for example, rows of words which differed slightly. The student had to indicate the one which
sound different from the others, e.g. "Miller - Müller - Miller - Miller", where the right answer
would have been a mark at the second position.
12 Sound sources can be tape recordings as well as the learner's own voice (which immediately
reaches the learner's ears through the headphones via a microphone and an automatic circuit).
13 In the spectra of vowels, the main formants of the vowels are located in the following areas: /i/:
200 - 400 and 3,000 - 3,500 Hz; /e/: 400 - 600 and 2,200 - 2,600 Hz; /a/: 800 - 1,200 Hz; /o/: 400 -
600 Hz; /u/: 200 - 400 Hz. (Cf. Trendelenburg in Habermann 1986, p. 77).
14 Two channels do not mean that one channel is used for the right ear and the other is used for the
left ear. The sound is being modified constantly but not synchronized on both sides through the two
channels.
15 I indicates the intensity of sound, which is measured in decibels (dB), F stands for frequency,
which is measured in Hertz (Hz).
16 D.R. Lide compares numbers for the speed of sound transmission in meters per second: (Cf. Lide,
1993)
dry air, 0°C - 353 m/secsea water, 25°C - 1,531 m/sec bone - 3,380 m/sec
We can easily test bone conduction in our own body when we put our fingers in our ears and speak
aloud. Everything we hear is mainly transmitted by the bones of our skull and not through the air.
We are always used to hearing ourselves in a combination of bone transmission and air
transmission. That is why tape recordings of our own voice sound so strange to us. In the latter case
the bone conduction is mainly eliminated.
18 For lateralisation of the hearing process see Tomatis, 1986, pp. 64-85.
19 Cf. Kahle et al., 1991, pp. 108-09.
20 Cf. A. Tomatis, 1977, p. 86 and Wirth 1994, pp. 76-88.
21 Cf. Kaunzner, 1994, pp. 72-73 and Kaunzner/Gianni, 1997.
22 The textbook (with 5 audio cassettes) for Italian native speakers is now available: U.A. Kaunzner,
1997. I Suoni del Tedesco. Deutsche Aussprache für italienischsprachige Lerner. Bologna,
CLUEB. A general version on CD (for any source langage) is in print by Julius Groos Verlag,
Heidelberg
23 There were several reasons for excluding some of the participants and for not considering Milan
and Brescia at all:
1. organizational problems;
2. technical problems with the instruments (which turned out to need further improvement in
order to be used with large groups);
3. the fact that some participants could not follow the whole set of sessions or missed the final
tests.
24 For a more detailed description of the testing material see Kaunzner/Gianni, 1997.
25 Cf. Rasch, 1980.
26 If the applied model holds true for all items in a particular set, all items measure the same ability,
and a subject of higher ability will have a higher probability of a correct response for all items in the
test than a subject of lower ability. Similarly, the positive differences in mean ability between
groups of subjects will result in positive differences in probability for these groups to succeed in
each and every item in the test. The computer programme OPLM (Verhelst et al., 1995) is being
used to estimate the item difficulty parameter, given the data set. By using the conditional
maximum likelihood (CML) estimation procedure, it was possible to estimate the item parameters
without any constraint on the ability distributions within each group or across groups. (John De
Jong, Director of the Language Testing Unit in CITO, final report in preparation).
27 "Item" means a single aspect in a test which is evaluated (for example, whether the sentence
stress in one of the repetition exercises is being repeated with the stress on the right word).
28 John De Jong, Director of the Language Testing Unit in CITO, final report in preparation.
29 By applying Item Response Theory, differences in scores between the groups at T1 become
irrelevant.
30 A detailed report on the test results will be published by CITO by the end of 1997.
References
Cauneau, I. 1992. Hören - Brummen - Sprechen. München, Klett.
Delattre, P. 1965. Comparing the phonetic features of English, German, Spanish and French.
Heidelberg, Groos.
Essen, O. v. 1979. Allgemeine und angewandte Phonetik. Berlin, Akademie Verlag. Habermann, G.
1986. Stimme und Sprache. Stuttgart, Thieme.
Fromkin, V., Rodmann, R. 1993. An introduction of language. Fort Worth, Harcourt Brace.
Habermann, G. 1986. Stimme und Sprache. Stuttgart, Thieme.
Husson, R. 1957. Etude expérimentale des modifications éventuelles de la fourniture vocalique sous
l'influence de fournitures auditives stimulatrices concomitantes. (Resumé of the experimental work
by M. Pierre and P. Grassé for the Académie des Sciences.).
Kahle, W. et al. 1991. Taschenatlas der Anatomie: Nervensystem und Sinnesorgane, Vol. 2.
Stuttgart, Thieme.
Kaunzner, U. 1994. Neue Wege zur Verbesserung des Hörverstehens und der Aussprache beim
Fremdsprachenerwerb. Zielsprache Deutsch, 25, 2, 68-74.
Kaunzner, U. 1995. Fast forward: Electronic Acceleration and Improving of Comprehension and
Speaking in Foreign Language Learning. In R. Dirven, J. Vanparys (eds.) Current approaches to
the lexicon. Frankfurt/M., Lang, 503-17.
Kaunzner, U. 1996. Elektronisches Hörprogramm zur Verbesserung des Hörvermögens und der
Aussprache im DaF-Unterricht. Materialien Deutsch als Fremdsprache 42. Regensburg, 199-212.
Kaunzner, U., Gianni, F. 1997. Il progetto Audio Lingua: miglioramento della comprensione
auditiva e della espressione orale di una lingua straniera. Quaderni del Centro Linguistico
dell'Università di Verona. Verona, UP. (in print)
Lide, D.R. ed. 1993. Handbook of chemistry and physics. Boca Raton, CRC Press.
Luchsinger, R., Arnold, G.E. 1965. Voice-speech-language. Belmont, Wadsworth.
McLaughlin, B. 1987. Theories of second-language learning. London, Arnold.
Rasch, G. 1980. Probalislistic models for some intelligence and attainment tests. Chicago, UP.
Tomatis, A. 1977. L'oreille et la vie. Paris, Laffont.
Tomatis, A. 1978. L'oreille et le langage. Paris, Éditions du Seuil.
Tomatis, A. 1986. Vers l'écoute humaine, Vol. 1. Paris, Les Editions ESF.
Tomatis, A. 1987. L'oreille et la voix. Paris, Laffont.
Tomatis, A. 1991. Nous sommes tous nés polyglottes. Paris, Fixot.
Verhelst, N.D. et al. 1995. One-Parameter Logistic Model OPLM. Arnhem, Cito.
Wendler, J., Seidner, W. 1987. Lehrbuch der Phoniatrie. Leipzig, Thieme.
Wirth, G. 1994. Sprachstörungen - Sprechstörungen - Kindliche Hörstörungen. Köln, Deutscher
Ärzte-Verlag.
