May 16th, 2013
Anéla Conference: Radboud University Nijmegen

Keynote Prof. dr. M. Coene

Clinical assessment of pitch perception and its relevance for the acquisition of oral grammar in hearing impaired children
Martine Coene & Paul Govaerts, Vrije Universiteit Amsterdam & The Eargroup, Antwerp

In oral communication, there are a number of linguistic means by which speakers may signal important parts of information.
Germanic languages such as Dutch have been argued to make use of prosodic prominence rather than using syntactic operations
to mark contextually important information (compare e.g. Du. een mooi MEISJE - een MOOI meisje to It. una bella ragazza – una ragazza
bella). Pitch-marking allows the speaker to make particular elements stand out in running speech. In the oral language input to
children, it helps to draw the child’s attention to new referents in the discourse (e.g. Kijk eens wat een lekker KOEKJE !).

In previous work, we have shown that particular morphemes, such as the indefinite article (een ‘a’), occur more frequently in the
presence of a pitch-marked noun. This may explain why children acquire indefinite articles before their definite counterparts (de/het ‘the’),
in spite of the overall higher frequency of the latter morpheme in the input. Surprisingly, the grammar of cochlear-implanted children
does not show the same early emergence of indefinite articles. The hypothesis put forward in this study is that the observed
difference in natural order of emergence of both types of articles is the result from the deficient perception of voice pitch found in
cochlear implant users.

In order to verify this hypothesis, different populations of deaf and hearing subjects have been tested on their perception of pitch
with respect to intonation patterns that occur in natural language. To this end, a clinical pitch assessment procedure
was used to test 19 deaf adults, 19 deaf children (age range 5-15 years) wearing different types of hearing devices (CI, classical
Hearing Aid and bilateral CI+HA) and 30 hearing adults and 19 hearing children (age range 6,5-12 years). In a number of intonation
perception tasks pairs of mono- or bi-syllabic words were to be discriminated on the basis of differences in pitch accent in a same/different paradigm
(16 pairs of natural speech words and 16 pars of words low-pass filtered at 500 Hz) mimicking intonation patterns associated with either definite or indefinite contexts.

Performance outcomes in discrimination were examined in terms of hearing, the type of hearing device and the type of stimulation. The
results confirm that CI-users perform significantly worse on intonation tasks than hearing controls and that the simultaneous use of
a classical hearing aid in the opposite ear significantly improves the deaf children’s performance.

Based on these results we take CI-children to have difficulties in hearing the subtle changes in fundamental frequency necessary to
discriminate particular intonation patterns which occur in natural language. This makes pitch-marked nouns to be less easily perceived
in incoming speech, resulting in a delay in the acquisition of indefinite articles. The optimal use of residual low-frequency hearing
(classical hearing aid in the non-implanted ear or electro-acoustic stimulation in the implanted ear) is expected to improve low-frequency hearing
and as such to have a beneficial effect on the acquisition of intonation-related morphology.

Lecture Anneke Van der Lee, MA
Watch the lecture here!

Are speech perception errors determined by linguistic factors?
Searching for phonetic and lexical explanations from initial consonant confusions in speech audiometry
Anneke van der Lee 1,2, Paul Govaerts 1,2, Bart Vaerenberg 2, Martine Coene 1,2
1) Vrije Universiteit Amsterdam & 2) Otoconsult/Eargroup Antwerpen

Speechaudiometry is common clinical examination to analyze hearing deficits. In a conventional test setting the patient is asked to repeat words that are orally presented to him or her. The audiologist will then register which phonemes are not repeated correctly and calculates the percentage of correct speech utterances.

The central question of this study is how patients fill in a gap when they are not hearing (part of) the words of a speechaudiometry test well enough. After all, understanding speech is a complicated process in which both auditive (threshold of audibility, ability to distinguish) and non-auditive factors (linguistic skills, cognitive capacities) play a part. In this study we want to analyze if the phoneme confusions could be declared by linguistic grounds.

Methods and materials .
We compare stimuli that are used in speech audiometry tests in the Dutch speakers area with the responses and test the following three hypotheses.
(i) The alternative hypothesis has to be formed purely by the availability of phonemes. (“The phonemic hypothesis”).
(ii) The listener chooses an existing alternative in the native language. (“The lexical hypothesis”)
(iii) The choice for an existing alternative word occurs on the basis of lexical frequency. (“The frequency hypothesis”).

The analyzed data set consists of 35.040 stimulus-response pairs. The results belong to a pilot where the stimulus word 'bel' has been analyzed. The word has been presented 236 times and this led to 47 alternatives from which the initial consonant differed from the original /b/. The analysis consists of a comparison of the alternatives by matrixes with the expected and observed values. This will be done for each of the previous mentioned hypotheses. It is assumed that the hypothesis with the smallest deviation between both values has the biggest probability to be true.

Results & conclusions.
When the phoneme confusions are compared for the three hypotheses, it is found that the lexical hypothesis seems to be
the best explanation for the set of alternatives (0,812). On the second and third place respectively follows the phoneme route (0,938) and the frequency route (1,359).

On the basis of this pilot it seems that people opt for a meaningful alternative when they do not hear the initial consonant well. This could have important implications for the current speech audiometry tests. This pilot is just a small part of the study and in the weeks to come, it will be expanded to other initial consonants.