Polychronia Christodoulidou, Katerina Nicolaidis, Dimitrios Stamovlasis
Aristotle University of Thessaloniki
polychri@enl.auth.gr, knicol@enl.auth.gr, stadi@edlit.auth.gr

Vowel reduction is a phenomenon that is directly influenced by duration, with vowels reduced in duration exhibiting formant undershoot (Moon & Lindblom, 1994). Research on vowel reduction in Greek adult speech has shown that unstressed vowels, which are reduced in duration, do not reach the ideal acoustic targets observed in stressed vowels, resulting in smaller vowel space areas for the unstressed condition than the stressed one (Fourakis et al., 1999). Despite the extensive study of vowel reduction in adults, research on children’s speech is very limited both in Greek and cross-linguistically. Moreover, the focus is on children’s temporal patterns (Redford, 2015), while spatial data would be also valuable for clinical purposes and speech technology.

To investigate vowel reduction developmentally, this study examined 72 typically developing Greek speakers across 9 age groups (3-, 5-, 7-, 9-, 11-, 13-, 15-, 17-year-olds, and adults), with an equal gender distribution in each group. Participants took part in a delayed repetition task, producing words of the form [pVpV] with stress on the first or second syllable (V = [i, ε, ɐ, o, u]). The words were symmetrical (e.g., [ˈpipi, piˈpi, ˈpεpε, pεˈpε], etc.) and were repeated 5 times within the carrier phrase [ˈlεο to __ pɐˈdu] ‘I say __ everywhere’. Both vowels of each word were studied, totaling 7,200 vowels being examined (72 participants × 2 stress conditions (stressed, unstressed) × 2 foot types (trochee, iamb) × 5 vowels × 5 repetitions).

Measurements included absolute and relative vowel duration, as well as (non-)normalized vowel space areas (72 participants × 2 stress conditions × 2 foot types × 5 repetitions = 1,440 (non-)normalized vowel space areas), which were analyzed using linear mixed model ANOVAs. An exploratory statistical procedure followed, with age, gender, stress, foot, and vowel as independent variables (vowel was not included in the analysis of vowel areas) and with the speaker as the random-effects factor, aiming to retain only statistically significant variables in the models. The results showed a significant decrease in absolute/relative duration and non-normalized vowel areas with age, with females exhibiting larger non-normalized vowel areas than males. Stressed vowels had longer absolute/relative duration than unstressed ones across all ages. Moreover, unstressed vowels were located within the vowel areas of stressed ones across all ages. However, the degree of vowel reduction in children was not always the same as in adults (less temporal and spatial degree for young children and greater temporal degree for preadolescents were found). This finding, in combination with a weaker positive Pearson’s correlation between relative duration and normalized vowel areas for younger age groups, suggests variation in spatio-temporal organization across ages. Additionally, although stressed vowels had longer absolute/relative duration and larger (non-)normalized areas in trochees compared to iambs, Ballard et al.’s (2012) suggestion that children demonstrate no difficulties marking stress in the trochaic foot was not confirmed.

References

Ballard, K. J., Djaja, D., Arciuli, J., James, D. G. H. & van Doorn, J. (2012). Developmental trajectory for production of prosody: Lexical stress contrastivity in children ages 3 to 7 years and in adults. Journal of Speech, Language, and Hearing Research, Vol 55, 1822-1835.

Fourakis, M., Botinis, A. & Katsaiti, M. (1999). Acoustic characteristics of Greek vowels. Phonetica, Vol 56, 28-43.

Moon, S. J. & Lindblom, B. (1994). Interaction between duration, context, and speaking style in English stressed vowels. The Journal of Acoustical Society of America, Vol 96 (No 1), 40-55.

Redford, M. A. (2015). The acquisition of temporal patterns. In M. A. Redford (Ed.), The Handbook of Speech Production (pp. 379-403). Malden: Wiley.

Vicky Papachristou
CITY College, University of York Europe Campus
vpapachristou@york.citycollege.eu

Previous research on child speech development suggests that children produce formant frequencies and temporal values that are higher than those of the adults, revealing great intra- and inter-speaker variability (Christodoulidou et al., 2023; Kelmali, 2020; Lee et al., 1999; Romeo et al. 2013). Also, it has been established that female speakers exhibit higher formant values and have a wider vowel space compared to male speakers (Hillenbrand et al., 1995; Periot & Arnold, 2021; Sfakianaki, 2002; Simpson, 2009). This study aims to report on the differences in the production of L1 Standard Modern Greek vowels between young children and adolescents. To that end, recordings of twenty (10 nine-year-old and 10 fifteen-year-old) students were obtained, which consisted of a list of short sentences containing words with the Greek vowels set in a carrier phrase. The words were of the form /'pVCV/ or ⁄'bVCV⁄ with vowels = ⁄i, e, ɐ, o, u⁄ and consonant = /t/ (appearing after the vowels tested) (πίτα ⁄'pita⁄, πέτα  ⁄'peta⁄, πάτα ⁄'pata⁄, μπότα  ⁄'bota⁄, μπούτια  ⁄'butça⁄) set in the carrier phrase ‘Λέγε _ πάλι’ (⁄'lege _    'pali/). All the data were analysed firstly acoustically by measuring the first two formants (F1 and F2) and the duration of the vowels from wideband spectrograms, and secondly, statistically to identify any significant differences between the two age groups. The results showed that Greek vowels produced by young children were placed more “downwards and to the left” compared to those produced by teenagers, as the former exhibited higher formants than the latter.  Also, female adolescents’ vowel space was found to be the broadest, while male adolescents’ vowel space was shown to be the narrowest. Regarding temporal values, Greek high vowel /i/ was the shortest, while low vowel /ɐ/ and high-mid vowel /o/ were the longest for both age groups and genders. The findings of the present study constitute an original contribution to the existing literature, as the collection of data by young children is not extensive, hence offering useful insights into young speakers’ speech.

Keywords: Standard Modern Greek; vowels; production; formants; children; adolescents

References

Christodoulidou, P., Nicolaidis, K. & Stamovlasis, D. (2023, 20-24 August). Vowel reduction by Greek-speaking children: The effect of stress and word length. INTERSPEECH, Dublin, Ireland.

Hillenbrand, J., Getty, L. A., Clark, M. J., & Wheeler, K. (1995). Acoustic characteristics of American English vowels. The Journal of the Acoustical Society of America, 97(5), 3099-3111.

Kelmali, E. (2020). An acoustic analysis of Greek fricatives and vowels produced by adults and children. (Doctoral dissertation). Aristotle University of Thessaloniki, School of English.

Lee, S., Potamianos, A., & Narayanan, S. (1999). Acoustics of children’s speech: Developmental changes of temporal and spectral parameters. The Journal of the Acoustical Society of America, 105(3), 1455- 1468.

Periot, E. & Arnold, A. (2021). Cross-Gender Differences in English/French Bilingual   Speakers: A Multiparametric Study. Perceptual and Motor Skills, 128(1), 153-177. DOI: 10.1177/0031512520973514

Romeo, R., Hazan, V., & Pettinato, M. (2013). Developmental and gender-related trends of intra-talker variability in consonant production. The Journal of the Acoustical Society of America, 134(5), 3781- 3792.

Sfakianaki, A. (2002). Acoustic characteristics of Greek vowels produced by adults and children. In Selected Papers on Theoretical and Applied Linguistics from the 14th International Symposium 2000 (pp. 383– 394). Thessaloniki: Department of Theoretical and Applied Linguistics, School of English, Aristotle University of Thessaloniki.

Simpson, A. P. (2009). Phonetic differences between male and female speech. Language and linguistics compass, 3(2), 621-640.

George Georgiou, Elena Theodorou
University of Nicosia, Cyprus University of Technology
georgiou.georg@unic.ac.cy, eleni.theodorou@cut.ac.cy

Children with developmental language disorder (DLD) find it challenging to perceive phonological contrasts (Georgiou & Theodorou, 2023). This study examines how children with DLD and children with typical development (TD) discriminate consonants of their native language and their processing speed in this discrimination. It also investigates how the perception of phonological structures correlates with the perception of grammatical and semantic structures. The participants were 14 Cypriot Greek children with DLD (nfemales = 7) and 14 peers with TD (nfemales = 5) with an age range of 7;10 – 10;4; the two groups matched for age and nonverbal intelligence. The speech material included five fricative/stop voiced consonants and their voiceless counterparts embedded in trisyllabic /CCV.CV.CV/ nonsense words; the target consonants were the first phonemes of the first syllable. The children completed an AX discrimination test on a computer-based Praat script in which they listened to two words through headphones. Then, they were instructed to select whether the words were acoustically the same or different. Similarly, the grammaticality and semantic judgment tests were completed on Praat scripts and asked children to judge whether different sentences were syntactically or semantically correct. The Bayesian regression models demonstrated strong evidence that the DLD group exhibited lower performance compared to the TD group, while no such evidence was found for the processing speed across the two populations (see Figure 1). In addition, the correlation tests did not show significant correlations between any of the tests. These findings highlight the role of voicing discrimination as a potential diagnostic marker of DLD as opposed to reaction times. They also denote that each cognitive linguistic domain operates independently being in line with modularity of mind theories.

Figure 1: Performance in the phonology test (left) and mean reaction time (right)

References

Georgiou, G. P. & Theodorou, E. (2023). Abilities of children with developmental language disorders in perceiving phonological, grammatical, and semantic structures. Journal of Autism and Developmental Disorders, 53, 4483-4487.

Georgiou, G. P
University of Nicosia
georgiou.georg@unic.ac.cy

The field of nonnative sound perception in language learning has witnessed the development of several influential models, such as the Speech Learning Model and its revised version (Flege & Bohn, 2021), the Perceptual Assimilation Model/L2 (Best & Tyler, 2007), and the Second Language Linguistic Perception model (van Leussen & Escudero, 2015). A recent addition to this landscape is the Universal Perceptual Model (UPM) by Georgiou (2021), designed to offer comparable insights into learners’ abilities to perceive nonnative sounds. UPM asserts that speech sounds are fundamentally perceptual, shaping the recognition of phonetic categories extracted from the speech signal. Grounded in the outcomes of prior phonetic training studies (e.g., Georgiou, 2022), UPM proposes a universal human capacity to acquire nonnative sounds throughout life. The model posits that the discrimination of sound contrasts is contingent on the degree of overlap between the two contrastive sounds. This overlap hinges on the perceptual association of the nonnative sounds with each other, based on their classification in terms of one or more first language sounds. According to UPM, sound contrasts can be classified as completely overlapping (CO), wherein they share the same above-chance responses; partially overlapping (PO), indicating at least one shared above-chance response; or nonoverlapping (NO), signifying no shared responses. This work presents the outcomes of empirical studies evaluating the ability of the UPM model to predict the discrimination accuracy of nonnative sound contrasts based on perceptual similarity. Employing diverse statistical methods, ranging from mixed-effects models to Bayesian regression models, and various tests like AXB and AX, these studies encompass a range of target languages, both familiar and unfamiliar. Collectively, the findings underscore that NO contrasts demonstrate the most accurate discrimination, followed by PO and CO contrasts, aligning with UPM’s assumptions. Moreover, these studies emphasize the crucial role of acoustic and perceptual cues in predicting patterns of nonnative sound discrimination. Future investigations could explore additional contrasts, including consonantal ones, and consider learners at various developmental stages in second or foreign language acquisition.

References

Best, C. T., & Tyler, M. (2007). Non-native and second-language speech perception: Commonalities and complementarities. In O-S. Bohn & M. J. Munro (Eds.), Second language speech learning: In honor of James Emil Flege (pp. 13-34). John Benjamins.

Flege, J. E., & Bohn, O. S. (2021). The revised speech learning model (SLM-r). Second language speech learning: Theoretical and empirical progress, 3-83.

Georgiou, G. P. (2021). Toward a new model for speech perception: The Universal Perceptual Model (UPM) of Second Language. Cognitive Processing, 22(2), 277-289.

Georgiou, G. P. (2022). The impact of auditory perceptual training on the perception and production of English vowels by Cypriot Greek children and adults. Language Learning and Development, 18(4), 379–392.

Van Leussen, J. W., & Escudero, P. (2015). Learning to perceive and recognize a second language: the L2LP model revised. Frontiers in psychology, 6, 1000.

Anastasia Deri
Aristotle University of Thessaloniki
anasnter@webmail.auth.gr

Introduction: This study investigates lingual coarticulation in Greek children using the technique of ultrasound tongue imaging.  It is the first study to present kinematic data to address the maturation of intrasyllabic coarticulation, together with the development of speech motor control and speech production. There are still crucial gaps in our knowledge about unit organization in children, with converging evidence showing segment-specific patterns (Rubertus and Noiray, 2020, Noiray et al., 2018, Zharkova, 2018). The present ultrasound study is expected to provide important insights into changes in the extent of coarticulation across different phonemes.

Method: We recorded tongue movements for ten children 7-8 years, ten adolescents 14-15 years and ten adults. Vowel-to-consonant effects are investigated in the first CV syllable of disyllabic real and nonsense words, where C=/p, t, k, f, s, x/ and V=/i, a/. Vocalic effects are measured at the midpoint of the consonant and differences in the magnitude are compared across age cohorts and consonantal contexts.

Results: In all age groups, the magnitude of coarticulation is sensitive to the articulatory properties of the consonants under investigation. Overall, anticipation of the following vocalic gesture is more extensive for labial consonants compared to lingual ones, whose patterns differ according to their articulatory characteristics. The results show significant across-and-within-speaker variability, as well as a significant decrease from childhood to adolescence and adulthood.

Discussion: The patterns of coarticulation are segment-specific, in line with the theory of the Degree of Articulatory Constraint model (Recasens et al, 1997). Previous studies in other languages have provided evidence that the articulatory properties of the consonants determine the magnitude of coarticulation and relate to developmental patterns observed in child speech (Zharkova et al, 2018, Noiray et al., 2018). Significantly greater variability in children compared to adolescents and adults reflect an immature speech motor control system and a developing phonological system. Implications for theories of speech development in relation to the Greek language are discussed.

References

Noiray, A., Abakarova, D., Rubertus, E., Krüger, S., & Tiede, M. (2018). How do children organize their speech in the first years of life? Insight from ultrasound imaging. Journal of Speech, Language, and Hearing Research, 61(6), 1355-1368.

Recasens, D., Pallare`s, M. D., and Fontdevila, J. (1997). A model of lingual coarticulation based on articulatory constraints. The Journal of the Acoustical Society of America, 102, 544–561.

Rubertus, E., & Noiray, A. (2020). Vocalic activation width decreases across childhood: Evidence from carryover coarticulation.

Zharkova, N. (2018). An ultrasound study of the development of lingual coarticulation during childhood. Phonetica, 75(3), 245-271.

Ioanna Kappa
University of Crete
kappa@uoc.gr

In Standard Modern Greek (SMG) clusters of [-voice] Stops are of equal sonority according to the language-specific sonority scale: Stops < Fricatives_[-voi] < Fricatives_[+voi] < s < z < Nasals < Liquids (< : less sonorous than) (cf. Author, 1995). For the study, we examine naturalistic speech data from eight typically developing children, acquiring Greek as L1 (ages 2;04.16–2;11.25, 4 boys (B) and 4 girls (G)). We note that all children are in the intermediate developmental phase, and they exhibit the following common characteristics regarding the realization of target Obstruent clusters:

1. The falling sonority clusters [Fricative₁ Stop₂] /ft/, /xt/, /st/, /sp/, sk/ and [s₁Fricative₂] /sf/, /sx/ are reduced (100%) to the less sonorous Stop₂ and Fricative₂, respectively.
2. The flat sonority clusters consisting of [+cont], voiced [Fric₁ Fric₂], i.e. [vð, ɣð, vɣ],  are reduced (100%) to the segment with the more marked PoA (Dorsal or Labial)  according to the relative markedness hierarchy ranking of Place features in children’s phonological grammar, i.e. Dorsal > Labial > Coronal (Author, to appear).

The data in (1, 2) show that all children can not accommodate flat sonority clusters of [-cont], unvoiced Stops, both in a perceptually strong position (word-initial or stressed syllable), and word-internally, therefore reduction occurs (100%).

Sonority plays no role in the reduction of the target clusters in (1, 2) because C₁ and C₂ are of equal sonority. Positional Faithfulness to a word-initial position or a stressed syllable also seems irrelevant for preserving either  C₁ or  C₂. At first glance, it seems that Contiguity, guides the reduction in (1) and (2), as in the case of [Stop₁Stop₂] clusters in child Hebrew, e.g. /k₁t₂a'na/®[t₂a'na], ‘small’ [p₁k₂ak] ®[k₂ak] ‘cork’ (Bloch 2011:50), but this is not our case. In this study, we claim that in the children’s phonological grammar, the reduction of flat sonority Obstruent clusters is driven by the preservation of unmarked or marked, dependent on the target cluster makeup. To be more specific, the preservation of an unmarked or marked for PoA segment is related to the (un)marked MoA of the target cluster members, i.e., in the unmarked for MoA [-cont] [Stop₁Stop₂] clusters, it is preserved the segment with the more harmonic PoA,  i.e., the unmarked for PoA Coronal segment (1, 2). In contrast, in the marked for MoA [+cont] [Fric₁ Fric₂] clusters, it is preserved a marked for PoA segment (see (ii) above). By the latter analysis, we offer a unified account of all children’s outputs regarding the reduction of flat sonority voiceless or voiced Obstruent clusters. The formal analysis is couched in the framework of Optimality Theory (Prince & Smolensky, 1993/2004).

Eleftheria Geronikou
Department of Speech and Language Therapy, University of Patras
egeronikou@upatras.gr

There is a growing body of literature suggesting that impaired phonological processing affects the development not only of phonological knowledge but also of semantics, morphology and syntax. In this account processing of auditory input is essential for the mapping process between those different components through which the words and sentence structure are established. Phonological skills, including speech discrimination and integrity of phonological representations, may be at the root of speech and language difficulties. Studying speech input processing skills development of typically developing Greek-speaking children holds theoretical and clinical implications.  Greater knowledge of trajectories of speech input processing 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 input processing skills in Greek-speaking preschool age children. The aim of the paper is twofold: firstly, to evaluate whether the experimental tasks are developmentally sensitive; secondly, to investigate the developmental progression of speech processing skills in typically developing Greek speaking 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 (T1, T2, T3). Assessment tasks included (a) an auditory discrimination with picture choice task, (b) a task of real word auditory discrimination and (c) a task of nonword auditory discrimination with matched stimuli used across tasks. In order to assess the speech sound processing abilities of the children, a number of phonological minimal pairs were chosen; the phonological properties of the words such as voicing, manner and place of articulation and the phonotactic structure such as consonant clusters or close syllables were taken into consideration, in order to ensure broad representation of the Greek phonological system.

Results showed that for Group 1 performance fluctuated below the level of chance in auditory discrimination tasks. Nevertheless statistically significant differences were observed as children got older. When real word stimuli were used significant difference was observed between T1 and T2, whilst when nonword stimuli were used a significant difference was only observed between T2 and T3. Data indicate that a higher performance was achieved in tasks involving real words, as compared to the corresponding tasks involving nonwords.

Each of the input processing tasks was developmentally sensitive for the age range tested. Having taken account of task difficulty issues, these tasks appear to capture the development of input processing skills in typically developing children. Different tasks may be more representative of the maturity of the speech processing skills at different time points.

Theophano Christou¹, Maria Mastropavlou¹, Anna Sfakianaki¹, Marika Lekakou¹, Alexandra Prentza¹ & Theodore Marinis^2
1University of Ioannina, ² University of Konstanz
theophano_christou@hotmail.com, mmastrop@uoi.gr, asfakianaki@uoi.gr, mlekakou@uoi.gr, aprentza@uoi.gr, t.marinis@uni-konstanz.de

Recent research on the processing and perception of rhetorical questions (RQs) has indicated that speakers mainly depend on prosodic cues and linguistic context, rather than lexical or syntactic cues, when interpreting questions (Neitsch et al. 2018, Neitsch 2019, Kharaman et al. 2019). Additionally, research on Greek (Arvaniti et al. 2020, Baltazani et al. 2020) has shown that questions that are not used for seeking information differ from information-seeking ones in the way they are prosodically realized, with the former being associated with a flat-ending tune (LH* L-L%) and the latter with a high-ending tune (L*H L-H%). Nonetheless, neither the prosodic realization nor the perception and processing of RQs in Greek have been previously investigated.

Attempting to fill this literature gap, the present research focuses on the perception of RQs by adult speakers of Greek. Our aim is to explore the effects of prosodic, lexical, and contextual cues on the way speakers interpret a question as rhetorical (RQ) or information-seeking (ISQ). Therefore, the main research question addressed involves whether and how (a) lexical cues, (b) prosodic cues (specifically the boundary tone of the question), and (c) the linguistic context facilitate or determine the discrimination between RQs and ISQs. We further aim to look into the way these factors interact, so that we can establish a hierarchy of factors listeners depend on when interpreting questions.

Two experiments were conducted for the purposes of the study; a preliminary production experiment was carried out with 19 adult native speakers of Greek, aiming to look for prominent prosodic characteristics of RQs as opposed to ISQs, since no adequate background information specifically on RQs was available in the literature. The production data revealed a difference in boundary tone between RQs and ISQs, with a L% (flat) tune associated with RQs and a H% (rising) tune with ISQs, in line with previous studies on the prosodic realization of Greek non-ISQs vs ISQs (Baltazani et al. 2020). Based on these findings, a subsequent, perception experiment was designed and run on a separate group of 50 adult speakers through the SoSciSurvey platform (https://www.soscisurvey.de/admin/index.php). In the perception experiment, speakers were presented with a set of questions, while the explored factors, namely prosody, lexical cues and linguistic context were manipulated to enable the investigation of their impact on interpretation. Specifically, sentences were presented in three versions with respect to linguistic context: no context, with RQ-associated and with ISQ-associated linguistic context; in two versions with respect to prosodic realization: with a RQ-associated (L%) and an ISQ-associated (H%) boundary tone; and in three versions with respect to lexical cues: with no cue, with a RQ-associated cue (i.e. sambos), and with an ISQ-associated cue (i.e. araje). Participants were asked to mark their preferred interpretation of each question as RQ or ISQ on a 5-point Likert scale, while additional acceptability judgments of all questions were also requested to establish the effects of conflicting cues (e.g. RQ-associated prosody with an ISQ-associated lexical cue).

The findings obtained from the two experiments will be presented and discussed with respect to the effects of the investigated factors on the interpretation of questions and will constitute the basis for the subsequent investigation of the perception and comprehension of rhetorical questions during language acquisition.

References

Arvaniti, A., Baltazani, M., & Gryllia, S. (2020). The contribution of pitch accents and boundary tones to intonation meaning. Proc. Speech Prosody 2020, 356-360.

Baltazani, M., Gryllia, S., & Arvaniti, A. (2020). The intonation and pragmatics of Greek wh-questions. Language and Speech, 63(1), 56-94.

Kharaman, M., Xu, M., Eulitz, C., & Braun, B. (2019). The processing of prosodic cues to rhetorical question interpretation: Psycholinguistic and neurolinguistics evidence. Interspeech, 1218–1222.

Neitsch, J., Braun, B., & Dehé, N. (2018). The role of prosody for the interpretation of rhetorical questions in German. In 9th International Conference on Speech Prosody 2018, 192-196.

Neitsch, J. (2019). Who cares about context and attitude? Prosodic variation in the production and perception of rhetorical questions in German. PhD Thesis, University of Konstanz.

Eirini Ploumidi
University of Crete
eirini.ploumidi94@gmail.com

Consonant deletion is a marginal strategy that occurs in several child languages (e.g. Costa & Freitas, Karni 2012). This study explores the application of onset consonant deletion by investigating longitudinal data, obtained from spontaneous speech and a picture-naming task, of two typically developing Greek-acquiring children (#67G, #61Β, G: girl, B: boy, ages: 1;10-2;08, intermediate phase of acquisition). Emphasis is placed on the conditions that drive onset deletions.

The data show that onset consonant deletion results in onsetless syllables (ratio: 4% of the data, see data in (1). This strategy is conditioned by grammatical restrictions on the realization of onset consonants. Specifically, obstruents that are marked with respect to MoA and/or voicing, i.e. voiced stops and voiceless/voiced fricatives, and sonorants, i.e. nasals and liquids, are deleted in onsets of (un)stressed syllables. The data further show that onset deletion may apply in every onset of the target word allowing the production of di-, tri- and multisyllabic consonant-free words (1a-b). These data confirm the presence of consonant-free words in typical child speech, contrary to Ben-David’s (2001) suggestion that this pattern does not occur there. Also, the data show that voiceless stops are the only onset consonants that resist deletion (1c).

(1)	Child	Target	Output	Gloss	Age	Target	Output	Gloss	Age
a.	#61Β	ˈði.o	ˈi.o	‘two’	2;03.16	ˈba.la	ˈa	‘ball’	2;04.02
b.	#67G	le.o.fo.ˈri.o	e.o.ˈi.o	‘bus’	2;00.06	ma.ˈri.a	a.ˈi.a	‘Mary’	2;03.15
c.	#67G	ˈpi.so	ˈpi.o	‘back’	2;00.06	fa.ɣi.ˈto	a.ˈto	‘food’	2;03.22

The explanation of the patterns in (1) suggests that markedness and sonority factors influence the children’s forms. In this acquisition path, the emergence or deletion of onsets is determined according to a sonority scale (e.g. Selkirk 1984). Specifically, onsets that are marked with respect to MoA and/or voicing and are more sonorous than voiceless stops are deleted. Hence, onset realization entails the formation of ‘optimal’ syllables with respect to sonority (see e.g. Clements 1990 for ‘optimal’ syllables), namely the presence of a sharp sonority rise from the onset to the vowel nucleus is a requirement. If this condition is not met, deletion occurs.

The data in (1) seem to reflect a ‘paradox’ since target CV syllables that are crosslinguistically unmarked with respect to syllable structure (e.g. Clements 1990) and favored in child speech (e.g. Jakobson 1941; Ben-David 2001) tend to surface as V ones, which are marked and less preferred in the (early) child speech. Nevertheless, it is argued that these productions do not reflect a ‘genuine’ preference for the V shape. Instead, they are driven by restrictions on the segments that are permitted in onsets and by a preference for onsets of the lowest sonority that are also unmarked with respect to MoA and voicing. Hence, the realization of V syllables turns out to be an epiphenomenon of the children’s grammars restrictions on onset consonants.

Concluding, the patterns of deletion in marked onsets and preservation of voiceless stop onsets that result in V and ‘optimal’ CV syllables, respectively, reflect grammatical preferences that prevail in the early child speech. The fact that these patterns are still produced by the children’s intermediate-state grammars indicate that they reflect cases of ‘relics’ of the initial phase that survive in later stages of acquisition. Hence, these patterns constitute a possible non-dominant acquisition path of the intermediate phase in child Greek.

Selected references

Ben-David, A. (2002). Language acquisition and phonological theory: Universal and variable processes across children and across languages. PhD dissertation, Tel-Aviv University.

Clements, G. N. (1990). The role of the sonority cycle in core syllabification. In J. Kingston & M. Beckman (Eds.), Papers in laboratory phonology I: Between the grammar and the physics of speech (pp. 283–333). Cambridge: Cambridge University Press.

Costa, J., & Freitas, M. J. (1998). V and CV as unmarked syllables: Evidence from the acquisition of Portuguese. Paper presented at the Conference on Syllable Typology and Theory. Universidade de Lisboa.

Jakobson, R. (1941/1968). Child language, aphasia, and phonological universals. The Hague: Mouton.

Karni, N. (2012). Minimizing faithfulness violation in the acquisition of Hebrew onset. Brill's Journal of Afroasiatic Languages and Linguistics, 4(1), 80–103.

Selkirk, E. (1984). Phonology and syntax: The relation between sound and structure. Cambridge, Massachusetts/London, England: The MIT Press.