Proceedings Fonetik 2009 - Institutionen för lingvistik

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm Universitywith a few cluster simplification patterns, suchas [tva] for “tavla” (picture). For the third ofthese children (a boy aged 4;4), and for all ofthe other children included in the study, nospeech production deviations was noted orcould be detected in the recordings. This mightsuggest a correlation between deviant speechproduction and difficulties of recognizing therecorded voice as one’s own. However, a contradictoryexample was also found that had tobe excluded from the study. Dentalisation (i.e.systematic substitution of [t], [d] and [n] for /k/,/g/ and /ng/, respectively) was noted for onegirl who could not participate for a second test,and who was therefore excluded from thisstudy. Interestingly, this girl scored 23 of 24 onthe first test. These single cases do certainly notpresent a uniform picture of the relation betweendeviant speech production and the abilityto recognize the recorded voice as one’s own,but rather illustrate the need for further investigationof this relation.The results in this study give support to theuse of recordings in a clinical setting, e.g. whenpromoting awareness in the child of deviationsin his/her speech production. An example of aneffort in this direction is presented in Shuster(1998), where children were presented withoriginal and “corrected” versions of their ownspeech production. The great variation betweenchildren in their ability to recognize their recordedvoice as their own requires further exploration.ConclusionsThe findings in this study indicate that childrenin the ages of 4-5 and 7-8 years can indeed recognizetheir own recorded voice as their own;average performance results are well abovechance. However, there is a large variabilityamong the children, with a few children performingat chance level or worse, and manychildren performing with more than 90% accuracy.No significant difference was found betweenthe younger and the older children’s performance,suggesting that self-voice perceptiondoes not improve between these ages. Furthermore,a time span of 1-2 weeks between recordingand identification seems to make theidentification task more difficult for the olderchildren, whereas the same time span does notaffect the younger children’s results. The findingshere support the use of recordings in clinicalsettings.AcknowledgementsThis work was funded by The Swedish GraduateSchool of Language Technology (GSLT).ReferencesBartholomeus, B. (1973) Voice identificationby nursery school children, Canadian Journalof Psychology/Revue canadienne depsychologie 27, 464-472.Mann, V. A., Diamond, R. and Carey, S. (1979)Development of voice recognition: Parallelswith face recognition, Journal of ExperimentalChild Psychology 27, 153-165.Maurer, D. and Landis, T. (2005) Role of boneconduction in the self-perception of speech,Folia Phoniatrica 42, 226-229.Rosa, C., Lassonde, M., Pinard, C., Keenan, J.P. and Belin, P. (2008) Investigations ofhemispheric specialization of self-voice recognition,Brain and Cognition 68, 204-214.Shuster, L. I. (1998) The perception of correctlyand incorrectly produced /r/, Journalof Speech, Language, and Hearing Research41, 941-950.Spence, M. J., Rollins, P. R. and Jerger, S.(2002) Children’s Recognition of CartoonVoices, Journal of Speech, Language, andHearing Research 45, 214-222.AppendixOrthography Transcription In English1) k /ko/ (the letter k)2) kaka /kka/ cake3) kam /kam/ comb4) karta /ka/ map5) katt /kat/ cat6) kavel /kvl/ rolling pin7) ko /ku/ cow8) kopp /kp/ cup9) korg /korj/ basket10) kula /kla/ marble11) kulle /kl/ hill12) kung /k/ king13) tåg /to/ train14) tak /tk/ roof15) tant /tant/ lady16) tavla /tvla/ picture17) tidning /tin/ newspaper18) tiger /tir/ tiger19) tomte /tmt/ Santa Claus20) topp /tp/ top21) tub /tb/ tube22) tumme /tm/ thumb23) tunga /ta/ tongue24) tupp /tp/ rooster139