Phonetic/phonological aspects of typical and atypical speech development with emphasis on Greek

Title: Who drives cluster selection in child speech? A unifying model of parallel and multi-level cluster formation

Research in prosodic development in Greek and cross-linguistically examines cluster production mostly on the basis of the repair mechanisms which determine cluster formation as well as define the order of cluster development. It is evident that cluster simplification and positional faithfulness are the most prominent repair mechanisms. Especially simplification is driven by markedness/ sonority considerations. Additional studies have placed emphasis on the statistical power of these repair mechanisms and their durability throughout phonological development. Related claims are made for L1, L2 and SLI data (cf. Barlow, 1997; Ohala, 1998; Gierut, 1999; Steele, 2002; Gnanadesikan, 2004, for English; Fikkert, 1994, for Dutch; Lléo & Prinz, 1996, for German and Spanish; Lukaszewicz, 2000, for Polish). However, very few works delve into the factors that drive cluster formation in late phonological development. This is the aim of this paper.

In late phases of phonological development in Greek, clusters are either accurately produced or are substituted for other clusters. In the second case, unmarked clusters, i.e., clusters whose members are defined by bigger distance on the sonority scale, are substituted for marked clusters whose members are very close on the sonority scale (Tzakosta, 2017). Marked clusters are perceptually vague and, consequently, difficult to be produced (cf. Tzakosta, 2009). The production of marked clusters instead of unmarked ones seems like an oxymoron; however, such data emerge both in typical and atypical Greek (examples in (1) and (2), respectively) as well as other aspects of Greek, like the dialectal variants of the language (see examples in (3).

The pervasion of this phenomenon leads us argue that the Classical Sonority Scale (hereafter CSS) (cf. Selkirk, 1984, Steriade, 1982) and the bi-dimensional model proposed by Morelli (1999), in which cluster formation is evaluated on the basis of two distinct scales, i.e., the manner and place scales, are not adequate to account for cluster selection. We propose the Three Scales Model (hereafter TSM), according to which in addition to the scales of manner and place voicing needs to be established as a distinct scale, is a more adequate theoretical tool. Like in the CSS, in the TSM well-formedness is driven by the rightward satisfaction of the scales in combination with the distance holding among cluster members. The TSM suggests that different degrees of scale satisfaction and different distances existing between cluster members result in different degrees of cluster well-formedness and acceptability. Clusters are defined as perfect, acceptable and non-acceptable. Perfect clusters satisfy all scales, acceptable clusters satisfy at least one and maximally two scales. Voicing actually defines a cluster as acceptable, since its violation automatically marks it as non-acceptable. Therefore, the TSM broadens the pool of acceptable consonant clusters and the phonotactics of the language become more flexible.

To summarize, in the course of our presentation, we will show that superiority of the TSM to the CSS boils down to the fact that the TSM a) can explain the substitution of unmarked/ perfect clusters for marked/ acceptable one, b) allows for more flexibility in the phonotactic constraints of all aspects of a language, in our case Greek, c) provides information regarding the very subtle factors which affect intra-language and inter-language syllabification, d) accounts not only for two- but also for three-member clusters (cf. Tzakosta, 2009, 2013), e) makes predictions about the strength of phonotactic constraints at a diachronic but also a synchronic level, especially in the speech of native speakers and second language learners (cf. Tzakosta, 2019). As a result, we do not need different tools to account for exceptional and statistically sporadic data attested in various aspects of a language. On the contrary, we argue that such data may provide important information regarding the seemingly invisible principles which shape the phonological system of a language, and, consequently, provide new perspectives on language acquisition and learning.

Greek L1 data

(1a) /a.ftó/ → [a.ptó],[a.ftó] ‘this -DEM.PRO’                             (B: 1;11.27)

(1b) /sxo.lí.o/ → [θxo.lí.o] ‘school-NEUT’                                     (D: 2;07.06)

(1c) /pá.sxa/ → [pá.θka] ‘Easter’                                                      (B.M.: 2;09.25)

Greek SLI data

(2a) /a.'kri.δa/ → [a. 'stsil.δa] ‘grasshopper-FEM.NOM.SG.’        (S2)

(2b) /'pro.va.to/ → ['spo.δa.to] ‘sheep-NEUT.NOM.SG.       (S4)

(2c) /'kti.ri.a / → ['θti.ri.a] 'building-NEUT.NOM.PL.'                (S4)

(2d) /ka.'tsi.kes/ → [ka.'tsi.kses] ‘goat-FEM.NOM.SG.’          (S4)

Greek dialectal data

(3a) /tu.fé.ci/ → [tfé.ci] ‘rifle’                                   (Thessaly, Τzartzanos 1909)

(3b) /pi.θa.mí/ → [pθa.mí] ‘span’                            (Thessaly, Τzartzanos 1909)

(3c) /pu.ká.mi.so/ → [pká.msu] ‘shirt’                  (Meleniko, Andriotis 1989)

(3d) /ku.bá.ros/ → [kba.ré.ls] ‘bestman’             (Thassos, Tombaidis 1967)

(3e) /ku.vá.ri/ → [gvár] ‘κουβάρι’                           (Kozani, Margariti-Roga 1989)

(3f) /ku.δúni/ → [kδu.nél] ‘bell’                               (Thassos, Τombaidis 1967)

References

Tzakosta, M. & Stavgiannoudaki, A. (2013). Cluster production in Greek SLI children: A typological account of two-member consonant clusters. In S. Stavrakaki, M. Lalioti, & P. Konstantinopoulou (Eds.), Advances in language acquisition (pp. 454–463). Cambridge: Cambridge Scholars Publishing.

Tzakosta, M. (2012). Place and manner interactions in Greek cluster phonotactics. In P. Hoole, C. Mooshammer, M. Pouplier, & B. Kuhnert (Eds.), Consonant clusters and structural complexity. Vol. 26: Interface Explorations Series (pp. 93–118). Berlin/ Boston: Mouton de Gruyter.

Title: Emergence of marked structures as a result of conspiring reduction strategies targeting [OBSTRUENT + LIQUID] clusters in child Greek: A case study

This study investigates the conspiring reduction strategies in target [OBSTRUENT+LIQUID] (OL) clusters and the concomitant emergence of marked structures in the disordered Greek L1 grammar of a child (age 4;9). The data, elicited through a picture-naming task, are drawn from Stavgianoudaki (2010). In the child’s realizations, [OBSTRUENT1+LIQUID2] target clusters are simplified in (un)stressed positions (in 88.24% of productions). The child preserves the less sonorous [OBSTRUENT1] conforming to the cross-linguistically attested pattern (cf. Ingram, 1989, a. o.). Two conspiring reduction patterns apply in complementary distribution, according to the position of the target OL clusters in the word:

1. OL WORD-INITIALLY: the Manner of Articulation of the target [LIQUID2] is delinked, while its CORONAL Place (PoA) is faithfully preserved, licensed in adjunct position, thus realized as an extrasyllabic, cross-linguistically attested, sibilant, e.g. the target ['tɾno] ‘train’ is realized as ['ste.no], ['pɾo.va.to] ‘sheep’ as ['spo.ða.to], [fli.d^za.ni] ‘cup’ as [sfi.'d^za.ni].
2. OL WORD-MEDIALLY: (a) the target [LIQUID2] is systematically transferred to the preceding open syllable as a coda, e.g. the target [a.e.ɾo.'plno] ‘plain’ is realized as [a.e.ɾol.'pa.no], [a.θli.'tis] ‘athlete’ as [al.θi.'tis], [i.ɾa.'klis] ‘Hercules’ as [i.ɾal.'cis].

RESEARCH QUESTION: What drives the simplification of OL clusters and the concomitant emergence of marked syllabic structures in (i) and (ii)?

Our data indicate that, in the child’s grammar, adjuncts and medial codas are realized before complex OL onset clusters of rising sonority, i.e. the child has marked structures, such as closed CVC syllables and extrasyllabic consonants, but no branching onsets in his system.

We argue that the reduced realizations in onset position, resulting from the conspiring reduction strategies in (i) and (ii), satisfy the demands in the child's grammar, on the one hand, for an unmarked (simple) less sonorous onset, i.e. an obstruent; on the other, for an increased perceptual salience, within the realized optimal CV syllable, which occurs due to the maximal sonority slope between the (onset) obstruent and the syllable nucleus (Clements, 1990). Furthermore, we claim that the marked syllabic structures resulting from the patterns in (i) and (ii), i.e. licensing of an adjunct and the emergence of a coda respectively, result from the demand in the child’s grammar for faithful realization of the target segmental quantity, despite the (target) cluster simplification. Our analysis is couched in Optimality Theory (Prince & Smolensky, 1993) with the faithfulness constraints as defined in Correspondence Theory (McCarthy & Prince, 1995). At this point of the child’s phonological development, FAITH constraints are also highly ranked, in conflict with *COMPLEX, requiring preservation of the input (target) segmental quantity, thus satisfying the FAITH constraint MAX-IOSegm as well as FAITH of the PoA feature of the segment in question, i.e. by faithful realization of the CORONAL PoA of [LIQUID2] in the position of an extrasyllabic adjunct in (i) or by metathesis of [LIQUID2] in the preceding Coda position in (ii). Although adjuncts are considered to be marked structures, the constraint *APPENDIX that penalizes them is hierarchically lower ranked in this child’s grammar, resulting in realization of adjuncts prior to the realization of OL clusters. The above grammar reflects a stage where faithfulness to the segmental quantity is highly ranked, while the syllabic structure is yet to be expanded to branching onsets.

References

Ingram, D. (1989). First language acquisition. Method, description and explanation. Cambridge: Cambridge University Press.

Clements, G. N. (1990). The role of the sonority cycle in core syllabification. Papers in laboratory phonology, 1, 283–333.

McCarthy, J. J., & Prince, A. S. (1995). Faithfulness and reduplicative identity. Papers in Optimality Theory, 10. https://scholarworks.umass.edu/linguist_faculty_pubs/10

Smolensky, P., & Prince, A. (1993). Optimality theory: Constraint interaction in generative grammar. Optimality Theory in Phonology, 3.

Stavgiannoudaki, A. (2010). Realization of consonant clusters in the speech of children with language disorders [in Greek] [M.A. Thesis]. University of Crete.

Title: The Lombard effect in children with cochlear implants: Suprasegmental aspects

Cochlear implantation led to significant improvements in the speech of deaf children, yet, so far studies investigated their speech in quiet. In the real world, children with cochlear implants (CI) communicate in noisy environments, such as in classrooms, where multiple talker babble and also reverberation are present. As known, speakers compensate for the presence of noise by increasing their vocal effort, a phenomenon called the ‘Lombard effect’ (Lombard, 1911). Effects of Lombard speech in normally-hearing adults have been documented across several languages (e.g., Cooke & Lu, 2009; Lane & Tranel, 1971; Summers et al., 1988; Tartter et al., 1993), including Greek (Nicolaidis & Rispoli, 2005). Nicolaidis and Rispoli (2005) examined the effects of white noise on vowel productions of Greek-speaking adults and reported increases in vowel intensity and pitch, small increases in vowel durations and also spectral changes in F1 and F2 which led to more open and more neutralized vowel productions. Moreover, the Lombard effect was also documented for postlingually-deafened adults with CI (Lee et al., 2015, 2017). Okalidou et al. (2018) examined the Lombard effect in the speech of children with CI, focusing on the effects of speech-shaped noise (SSN) and speech-shaped noise with reverberation on stressed vowels. Changes in pitch and intensity, occasional changes in duration and also changes in F1 and F2 were noted; however spectral changes were highly variable.

The present study compared the vowel productions of children with CI with the ones produced by normally-hearing peers, using the experimental stimuli and conditions of Okalidou et al. (2018). Effects of noise on stressed vowel intensity and duration were investigated in the speech of 7 children with CI and 7 children with NH, aged 7-12 years.  Magnitude of change for intensity was also assessed in terms of the ratio of stressed to unstressed vowels. Results from planned comparisons of a general linear model revealed that a) in quiet condition, children with CI had similar stressed vowel intensity to their NH peers, b) in noise conditions, group differences emerged, i.e. stressed vowel intensity was greater in children with CI than NH controls, and c) in SSN condition, longer stressed vowels were produced by children with CI as compared to NH but the effect was shown for only /i/, /o/, /u/ when reverberation was added. The differences observed across groups have implications for speaking in classroom noise.

Title: Perception of voicing contrast in children with cochlear implants

Previous research compared the perception of voicing contrasts in children with cochlear implants and children with normal hearing and found lower performance for the former in different languages such as English (Peng et al., 2019; Giezen et al., 2010) and French (Bouton et al., 2012). An important factor associated with speech perception abilities in children with hearing loss is the age of implantation since previous research has shown that the earlier the device is implanted, the better the perception outcomes (e.g., Dowell, Blamey & Clark, 1997). The Greek stop consonant system consists of voiced (/b/, /d/, /g/) and voiceless stops (/p/, /t/, /k/) (Arvaniti, 1999). In this study, perception of voicing contrasts was examined in 24 Greek-speaking children with cochlear implants (CI) and 24 age- and gender- matched children with normal hearing (NH). Via the procedure of fast mapping (Dollanghan, 1987), each child was asked to choose a three-dimensional object that corresponds to a nonword with the target-sound in three repetitions (864 tokens). Participants were separated in two equal age sub-groups; the first consisted of children between 2;10 to 8;1 years old and the second between 8;2 to 13;4 years old. Speech stimuli were two syllable trochaic nonwords with the stops in initial position followed by the vowel /a/. Results showed lower performance for the young age subgroup with CI compared to the NH controls. Statistically significant differences in the perception of the stops /p/, /b/, and /g/ were found for this age subgroup (2;10-8;1). No significant differences in the perception of stops were found for the older children with CI and NH. Overall, the results of this study corroborate previous findings underscoring the importance of the extent of implant experience in speech perception outcomes.

References

Arvaniti, A. (1999). Standard Modern Greek. Journal of the International Phonetic Association 29, 167–172.

Bouton, S., Colé, P., & Serniclaes, W. (2012). The influence of lexical knowledge on phoneme discrimination in deaf children with cochlear implants. Speech Communication, 54, 189–198.

Dollaghan, C. (1987). Fast mapping in normal and language-impaired children. Journal of Speech and Hearing Disorders, 52, 218–222

Dowell, R., Blamey, P., & Clark, G. (1997). Factor affecting outcome in children with cochlear implants. In G. Clark (Ed.), Cochlear implants (pp. 297–303). Sydney: Monduzzi Editore.

Giezen, M., Escudero, P., & Baker, A. (2010). Use of acoustic cues by children with cochlear implants. Journal of Speech, Language and Hearing Research, 53(6), 1440–1457.

Peng, Z. E., Hess, C., Saffran, J. R., Edwards, J. R., & Litovsky, R. Y. (2019). Assessing fine grained speech discrimination in young children with bilateral cochlear implants. Otology & Neurotology, 40(3), 191–197.

Title: The fricative /s/ in Greek hearing-impaired and normal-hearing young adult speech: Acoustic characteristics and coarticulatory variability

Fricatives, and especially sibilants, constitute a late-appearing class of sounds in typical as well as disordered speech development, as they require both precise tongue placement and sufficient airflow to produce adequate turbulence (Mcleod & Crowe, 2018; Shriberg & Kwiatkowski, 1994). Accurate fricative production is even harder to achieve when perception at higher frequencies is impaired due to hearing loss. Both fricative discrimination and production studies report a disadvantage for children and adults with hearing impairment (HI) (Liker, Mildner & Šindija, 2007; Zeng & Turner, 1990). Previous work on obstruent and sonorant production, as well as V-to-V and C-to-V coarticulation by Greek speakers with HI has shown deviant articulatory and temporal patterns. With reference to fricatives, high variability and increased articulatory constriction across the palate have been reported (Nicolaidis, 2004, 2007; Sfakianaki et al., 2018).

The current study builds on this work by examining acoustic aspects of fricative production by Greek speakers with impaired (HI) vs normal hearing (NH). Although acoustic characteristics of Greek fricatives have been investigated in typical hearing adult speakers (Nirgianaki, 2014), there is no research on the acoustics of fricatives produced by Greek speakers with HI. In the present work, we focus on the production of the voiceless alveolar fricative/s/ which poses great supraglottal, laryngeal, and aerodynamic demands, requiring fine coordination of the lingual and laryngeal articulators and precise aerodynamic control. Five participants with NH (two male and three female), and nine participants with profound prelingual HI (four male and five female), aged 18-26, were recorded producing symmetrical disyllables of the form /pVsV/ with the three corner vowels /i, a, u/ and stress on the first or second syllable. Temporal, spectral and amplitude parameters of the fricative /s/ are examined, taking into account the effects of hearing status, gender and stress position. The results of the acoustic analyses are discussed in relation to hearing and intelligibility differences among speakers and in light of electropalatographic data previously obtained for a subgroup of the participants with HI.

References

Liker, M., Mildner, V., & Šindija, B. (2007). Acoustic analysis of the speech of children with cochlear implants: A longitudinal study. Clinical Linguistics & Phonetics, 21(1), 1–11. https://doi.org/10.1080/02699200400026991

Mcleod, S., & Crowe, K. (2018). Children's consonant acquisition in 27 languages: A cross-linguistic review. American Journal of Speech-language Pathology, 27(4), 1546–1571.

Nicolaidis, K. (2004). Articulatory variability during consonant production by Greek speakers with hearing impairment: An electropalatographic study. Clinical Linguistics & Phonetics, 18(6-8), 419–432. https://doi.org/10.1080/02699200410001703574

Nicolaidis, K. (2007). Aspects of spatio-temporal variability during consonant production by Greek speakers with hearing impairment. Clinical Linguistics & Phonetics, 21(6), 405–422. https://doi.org/10.1080/02699200701267377

Nirgianaki, E. (2014). Acoustic characteristics of Greek fricatives. The Journal of the Acoustical Society of America, 135(5), 2964–76. https://doi.org/10.1121/1.4870487

Sfakianaki, A., Nicolaidis, K., Okalidou, A., & Vlahavas, G. (2018). Coarticulatory dynamics in Greek disyllables produced by young adults with and without hearing loss. Clinical Linguistics & Phonetics, 32(12), 1162–1184. https://doi.org/10.1080/02699206.2018.1510987

Shriberg, L. D., & Kwiatkowski, J. (1994). Developmental phonological disorders I: A clinical profile. Journal of Speechand Hearing Research, 37, 1100–1126.

Zeng F. G., & Turner C. W. (1990). Recognition of voiceless fricatives by normal and hearing-impaired subjects. Journal of Speech and Hearing Research, 33(3), 440–9. https://doi.org/10.1044/jshr.3303.440.

Title: Developmental characteristics of Greek fricatives: An acoustic study

During speech production, articulatory gestures are performed within specific spatial and temporal extents (Maas & Mailend, 2017). Child speech develops gradually, as motor control skills develop, with some sounds such as fricatives, demanding more advanced skills, due their articulatory and aerodynamic characteristics (Ferguson, 1978). This is confirmed by the developmental pattern in fricative acquisition, with Greek speaking children acquiring fricative categories relatively late compared to other sounds, between the age of 3;7 and 4;0 (Mennen & Okalidou, 2007). Differences in the spatial, temporal and spectral characteristics of child fricative productions compared to adults have been observed (Fox & Nissen, 2005).

There are several studies on English and American fricatives produced by children and adults (e.g. Jongman Wayland, & Wong, 2000). For Greek, however, limited studies have examined fricative production by adults (e.g. Nirgianaki, 2014) and very few by children (Nirgianaki & Kampaki, 2021). Greek comprises ten fricative sounds, the labiodentals /f, v/, the dentals /θ, ð/, the alveolars /s, z/, the velars /x, ɣ/ and the palatals [ç, ʝ]. A number of linguistic factors such as fricative place of articulation, voicing, post-fricative environment  and syllable stress can affect the acoustic characteristics of  fricatives (e.g., Jongman et al., 2000), while non-linguistic factors such as age and gender can also affect fricative quality (Fox & Nissen, 2005).

The current study presents the acoustic characteristics of Greek fricatives in the productions of 20 children (10 boys and 10 girls), aged between 8 and 10 years old. A control group of adults was also recorded. Participants were recorded producing CVCV real words, in a carrier phrase, where C1 was one of the Greek fricatives and V1 one of the five Greek vowels (/i, e, a, o, u/), in all possible combinations including stressed and unstressed syllables.

Findings show slower speech rates in children productions and extensive temporal variability compared to adults. Also, children differentiate the first spectral moment of the alveolar category, with boys having lower and girls higher frequencies, indicating developmental or sociophonetic variation. Lower values in the second spectral moment of all fricatives were found for children, showing the production of a more compact spectrum.

References

Ferguson, C. A. (1978). Fricatives in child acquisition. In V. Honsa & M. N. Hardman-Bautista (Eds.), Papers on Linguistics and Child Language (pp. 93–116). The Hague: Mouton.

Fox, R. A., & Nissen, S. L. (2005). Sex-related acoustic changes in voiceless English fricatives. Journal of Speech, Language, and Hearing Research, 48(4), 753–765.

Jongman, A., Wayland, R., & Wong, S. (2000). Acoustic characteristics of English fricatives. The Journal of the Acoustical Society of America, 108(3), 1252–1263.

Maas, E., & Mailend, M. L. (2017). Fricative contrast and coarticulation in children with and without speech sound disorders. American journal of speech-language pathology, 26(2S), 649–663.

Mennen, I., & Okalidou, A. (2007). Greek speech acquisition. In S. McLeon (Ed.), The international guide to speech acquisition Part II. Clifton Park, NY: Thompson Delmar Learning.

Nirgianaki, E. (2014). Acoustic characteristics of Greek fricatives. The Journal of the Acoustical Society of America, 135(5), 2964–2976.

Nirgianaki, E., & Kampaki, V. (2021). The production of Greek voiceless fricatives by young children. ExLing 2021, 165.

Title: The development of speech production skills in typically developing Greek-speaking children

There is a practical interest and need to expand our knowledge on phonological development in typically developing Greek-speaking children. Greater knowledge of trajectories of speech production development in the early years could inform theories of phonological development as well as enable clinicians to deliver better informed assessment and intervention. In this paper a cross-sectional longitudinal approach was used to investigate the development of speech production skills in Greek-speaking preschool age children.

Two groups of typically developing children aged 3;0-3;5 years (group 1) and 4;6-5;0 years (group 2) were assessed at three assessment points six months apart. Assessment tasks included (a) a picture naming task, (b) a task of real word repetition and (c) a task of nonword repetition. Responses were scored on the accuracy of articulation a) on a whole word basis (WW) b) on Percentage of Consonants Correct in each word (PCC).

In both groups of children a strong time effect was found in the development of speech output processing abilities as assessed with tasks of naming, real word repetition and nonword repetition. Performance accuracy was higher on tasks including real words rather than nonwords; children performed more accurately on tasks requiring repetition rather than spontaneous naming of real words; repetition accuracy on 2-3 syllables items was significantly better than on 4-5 syllables items. Strong significant relationships were found between scores in different output production tasks. Both scoring methods, i.e. scoring for WW accuracy and for PCC, were sensitive in tracking change between consecutive time points in group 1. As regards group 2, scoring for WW accuracy was more sensitive than scoring for PCC in identifying significant changes in performance. Analysis of the data shows significant improvement in accuracy of speech production across time for both groups; yet, speech production accuracy is approaching ceiling for group 2.

Findings from the present study may be used as a starting point to diagnose children with speech production difficulties in Greece.