Phonological delay of segmental sequences in a Greek child’s speech

ABSTRACT This paper investigates persistent elements of protracted phonological development (PPD) at ages 5;10 and 6;3 in a monolingual Greek girl’s speech following earlier articulation intervention (3;6) and post-intervention assessment (4;3). The re-assessment data examined here, five months apart, were elicited using the Phonological Assessment for Greek (PAel). Results reveal interesting idiosyncratic patterns in the production of segmental sequences (VV, CC) in complex syllables and longer words, with a striking imbalance between singletons and sequences, which holds especially true for inconsistency in the acquisition of /ɾ/ across CV, CC, and C.C contexts. Phonological delay surfaces as chronological mismatches, idiosyncratic forms and, most notably, disparity between segmental and structural development.


Introduction
In this paper we examine the protracted (delayed) phonological system of a monolingual Greek girl at ages 5;10 and 6;3, utilizing the Phonological Assessment for Greek (PAel) (Babatsouli,

Prosodic structure
Greek is a syllable-timed language with a basic trochaic (Sw) foot (Holton et al. (2002). The few monosyllables are function/loan words. Longer words comprise up to five syllables, while compounds may exceed five syllables. Onsets, while optional, can have up to three consonants, but CV is most common. Codas are also optional and more restricted in inventory (see below) (Setatos, 1974). Greek syllable distributions are comprehensively outlined in Setatos (1974) and discussed in Babatsouli (2019). Greek stress falls on either of the last three syllables, being primary, secondary or enclitic. Primary stress is contrastive, e.g., adult surface /miˈʎa/ 'speech' vs. /ˈmiʎa/ 'miles'. Enclitic stress is morphosyntactically determined, as in the+noun+possessive pronoun, e.g., to /ˈpɾosoˌpo tis/ 'her face' (Holton et al., 2002).

Greek phonological development
Greek phonological development is addressed in cross-sectional and single child case studies (Babatsouli, 2017;PAL: Panhellenic Association of Logopedics, 1995). Though no exact criteria are discussed, monophthongs are reported acquired by 4;0 (PAL: Panhellenic Association of Logopedics, 1995). See also Babatsouli (2020) for information on hiatus and diphthong acquisition.

The case
This study uses PAel (Babatsouli, 2019) to examine phonological delay in a monolingual Greek girl at 5;10 (T1) and 6;3 (T2), including interactions of singletons and sequences. Phonological delay is indicated by non-age-level phonological simplifications and inconsistency of productions across contexts.

Participant
"Dorothea" is a monolingual standard Greek speaker living in Greece. She is healthy and enjoys socializing and school. Parental written consent was obtained for the study following the Declaration of Helsinki ethical procedures. She was assessed by the second author at the Clinic of the Department of Speech and Language Therapy of the University of Patras.

Procedures
Re-assessment took place at 5;10 (T1) and 6;3 (T2) using PAel (Babatsouli, 2019, http://www. phonodevelopment.sites.olt.ubc.ca) which comprises two word lists, a Screener (50 words) and Extended A&B (51 + 51 = 102 words) lists, plus a 1,001-word narrative (N). Word-list data were elicited using the PAel culture-relevant colour images (one per word) that are freely available in slideshow software from: http://www.phonodevelopment.sites.olt.ubc.ca. Sentence cues (cloze technique), and/or phonemic, syllabic cues, included in the slide show for clinical use to motivate children during the test, were minimally utilized as needed. This battery for nonlinear analysis was preferred for reassessment because of its comprehensive coverage of Greek phonology in terms of phones, consonant clusters, word shapes, stress patterns, and syllable types per word position; for instance, there is a parallel between Setatos (1974) computations of segmental (including allophones) distributions in Greek and those in PAel. For a detailed discussion, see Babatsouli (2019). Dorothea's running speech was also assessed at T2, as she retold the story in her own words following her mother's narrative (N) readings. Digital recordings on a laptop placed 50 cm away with a built-in Lenovo Audio Device were configured to 2 channel, 16 bit, 48000 Hz; the data were stored in compressed M4A format. Transcriptions in IPA were performed independently by the authors, both native speakers and trained phoneticians. Minimal inter-rater disagreement (6%) was resolved through consensus in follow-up listening sessions. Supplemental File 1 provides the word list and transcriptions of productions. For examples in the text, we refer to Supplemental File 1 with number-codes as follows: Screener (S), Extended lists A, B, e.g., A3 refers to the third word in Extended A. Supplemental File 2 summarizes the data in the nonlinear analysis scan. Quantitative analysis was computed in spreadsheets.

Case profile
Global measures are presented first to show Dorothea's strengths that contrast with very specific needs regarding consonant clusters, presented in detail for the bulk of the profile. The following measures are included for T1 probes regarding clusters: cluster reductions, Percent Consonants Correct (PCC), Percent Clusters Correct (PClC), Measure for Cluster Proximity (MCP) (there was less than 10% change at T2).

Global measures for T1 screener
Global measures for T1 Screener data are presented first because the Screener enables a preliminary account of the child's phonological skills. Screener words comprise 44% (125/281) vowels and 56% (156/281) consonants. Production of monophthongs and diphthongs shows no errors in the Screener.

Global measures at T1 and T2
The section presents global measures comparing the child's performance at T1 and T2. Starting with vowels, Percentage Vowels Correct (PVC) is 98% at both assessment times, as shown in  Table 2 also compares consonant performance at T1 and T2 in terms of PCC, PClC, and WWM values. It is seen that there is, by and large, consistency in the child's production of consonants at the two assessment periods, though some progress is indicated at T2.

Word structure: general
Mismatches affecting word length and shape occurred in vowel hiatus, V 1 .V 2 , of multisyllabic words. The productions in the examples below indicate syllable reduction through deletion of a vowel in the hiatus (examples 1, 3, 4), consonant epenthesis/metathesis

Consonant sequences
The major constraint in Dorothea's system concerned consonant sequences. We first present global measures and then discuss the various clusters in detail.

Cluster measures at T1
There were 39 WI clusters and 22 WM clusters in all three probes: Screener, Extended A and B (See Tables 2 and 3.) Though not balanced for phoneme type and frequency, a comparison of PCC in singletons and sequences is presented. While the child's singleton PCC was 92%, her  PCC in clusters (individual consonants in clusters) was 64% (67% WI, 59% WM) and her heterosyllabic clusters PCC was 55%, preliminary evidence of difficulty in consonant sequences. On similar grounds, all four measures indicate better performance word-initially than word-medially (Table 4), though the numbers are too small to make conclusive generalizations. This holds for WI OBSTRUENT+OBSTRUENT (/sC/, STOP +FRICATIVE, STOP+STOP, and FRICATIVE+FRICATIVE /ðʝ/, /fç/) and WI clusters involving SONORANT (NASAL+NASAL, STOP+LATERAL). An exception is WM FRICATIVE+NASAL /zm/. Among all, NASAL+NASAL, /s/+STOP, and FRICATIVE+FRICATIVE do well irrespective of position. Reductions to the obstruent occur overwhelmingly in OBSTRUENT+SONORANT clusters: all /ɾ/ and /n/ tokens, but also /l/ in FRICATIVE+/l/, and once word-medially in STOP +/l/. Less consistent reductions occur for OBSTRUENT+OBSTRUENT in single tokens of /pt/, /tç/, /vð/, /γð/.

Production variability at T1, T2 and T2N
Comparing the child's productions across assessment times, we see that they were not variable, overall. Nevertheless, some inconsistency across different repetitions at T1, T2 and T2 (N) is noted. Nine (9) of the words targeted in PAel across probes are produced differently on one occasion only, such as at T1: /ˈku.klɐ/ [ˈku.kɐ] 'doll' (S13), /kɐ.

Discussion
PAel (Babatsouli, 2019) was used to facilitate fine-grained assessment of a monolingual Greek girl's phonology at 5;10 and 6;3, who had received articulation therapy between 3;6-4;1 but continues to show delay. The non-linear analysis of the nature and degree of her persisting phonological difficulties draws attention to needs not observed during previous initial and post-intervention assessments.
More variation in cluster productions is reported by Stemberger and Bernhardt (2018) for children with PPD than TD children. Dorothea's OBSTRUENT+SONORANT mismatches reveal such variability. Her /Cl/ and /Cn/ are special in that, while cluster members are mastered as singletons, reductions show universal preference for C1-OBSTRUENT, /xn/→[x, s, ç], /γl/→[γ], but also occasionally retain C2-SONORANT, /xn/→[n], /γl/→[l]. Additionally, while acquired as singleton, /x/ is mostly substituted in /xn/. By comparison, because nonacquired singleton /ɾ/ is always substituted by [l], her /Cɾ/ reduces to C1, as expected, but /ɾ/ is never substituted by [l] in /Cɾ/, as reported for a younger TD child (Babatsouli, 2021), and despite the fact that Dorothea produces STOP+LATERAL and /fl/. This is idiosyncratic and indicates a misalignment between realizations across singleton and cluster contexts. Thus, her phonological delay is influenced by structural constraints imposed by segmental sequences on the syllable and word levels, negatively impacting timely acquisition. This finding is significant because, though clinical approaches to intervention commonly focus on speech sounds (Baker et al., 2018), a non-linear approach to intervention accounting for higher prosodic units (Baker & Bernhardt, 2004) is not only justified, but also mandated.

Proposed intervention plan
A non-linear constraints-based approach is advisable for continuing intervention that targets structures, segments and features (see Supplemental File 2 -Scan; page 1 goals). Baker and Bernhardt (2004) analyzed the data of a child with a phonological impairment who had already attended intervention and report on goals achieved by adopting a non-linear approach to intervention. System intrinsic support seems to be efficacious to treatment. The literature also indicates that speech therapy intervention is most effective when it is intensive and is provided for at least a few weeks (Allen, 2013), including the case of intervention based on the principles of nonlinear phonology (Lundeborg Hammarström et al., 2019). On this basis, it is suggested that the client attends speech therapy sessions lasting at least 30 minutes with a frequency of two sessions per week for a period of eight weeks. It is expected that if a re-assessment is conducted upon such an intervention, significant progress will be observed.
Treatment needs (Table 5) (not ordered goals) for Dorothea are:
A phased approach to intervention would include the following goals: • Phase 1: singleton /ɾ/, and other singletons in challenging word structures. • Phase 2: support for segmental sequences, problematic syllable shapes (CVC), and multisyllabic words. • Phase 3: mostly liquid clusters and WM consonant sequences.

Conclusion
This paper has investigated the delayed phonology of a monolingual Greek-speaking girl revealing persisting needs even after intervention. The study brings to light a discrepancy in vowel and consonant production between singletons and segmental sequences in complex contents, like long words, onsets, and CVCs. Interestingly, though C+LATERAL is acquired, /ɾ/ in C+RHOTIC is deleted rather than substituted by [l], as for /ɾV/. Although a single child case study, the results complement the scarce literature on Greek phonological delay, recording under-reported patterns that provide information about the acquisition of phonological complexity in Greek, and lay the groundwork for future cross- [+lateral] to Liquid [-lateral], /l/ to /ɾ/ sectional studies. PAel is effective for non-linear phonological analysis, showing constraints that hamper the child's skill in structural complexity. The present study has illuminated the interplay of underlying hierarchical representations and provides a useful example in Greek for intervention protocols that target higher level representations, beyond simply articulatory skill.