«A. Kochetov and L. Colantoni Proceedings of Meetings on Acoustics Volume 19, 2013 ICA 2013 Montreal Montreal, Canada 2 ...»
A. Kochetov and L. Colantoni
Proceedings of Meetings on Acoustics
Volume 19, 2013 http://acousticalsociety.org/
ICA 2013 Montreal
2 - 7 June 2013
Session 5aSCb: Production and Perception II: The Speech Segment (Poster Session)
5aSCb29. An electropalatography (EPG) study of nasal-trill/lateral sequences in
Alexei Kochetov* and Laura Colantoni *Corresponding author's address: The Department of Linguistics, The University of Toronto, Toronto, M5S 3G3, Ontario, Canada, firstname.lastname@example.org Trills and laterals require relatively precise articulatory and aerodynamic settings that are at least partly incompatible with setting necessary to produce nasal stops. Historically, this incompatibility has often been resolved through assimilation, deletion, or epenthesis in within-word [n+r] and [n+l] clusters (e.g. in Romance). It is expected that similar, yet gradient effects will be observed in across-word or hetero-morphemic sequences of nasals and liquids. This study examines the production of Spanish nasal-liquid sequences using electropalatography (EPG).
Linguopalatal contact data were collected from 9 native speakers of Spanish (representing 3 dialects) producing various utterances with nasals before /r/ and /l/ (as well as before /t/). The analysis of C1 and C2 using standard indices of constriction location and degree showed that nasals had a more retracted and partly deocclusivized constriction before /r/, and a lowered tongue dorsum before both /r/ and /l/. These differences, indicative of substantial anticipatory coarticulatory effects in constriction location and degree, were similar across speakers, regardless of their dialect background and the default realization of final nasal (alveolar or velar). The results thus confirm the articulatory source of historical developments of combinations of nasals and liquids.
Published by the Acoustical Society of America through the American Institute of Physics © 2013 Acoustical Society of America [DOI: 10.1121/1.4800067] Received 22 Jan 2013; published 2 Jun 2013 Proceedings of Meetings on Acoustics, Vol. 19, 060273 (2013) Page 1 A. Kochetov and L. Colantoni
INTRODUCTIONApical trills are known to require very precise articulatory and aerodynamic settings: in order for trilling to occur, the tongue tip has to be positioned sufficiently close to the alveolar ridge/upper teeth andthe airflow has to be strong enough to initiate and sustain the tip’s vibration (Ladefoged & Maddieson, 1996). The optimal positioning of the tongue tip for an apical trill involves active lowering of the tongue dorsum, which contributes to the trills ‘resistance’ to coarticulation to preceding or following vowels and consonants (Recasens & Pallarès, 1999a, 2001).
Some adjacent articulations, however, can interfere with trilling, and there is a particularly strong incompatibility between apical trills and nasal consonants or vowels. This is because the lowering of the velum for nasals crucially reduces the intra-oral pressure necessary to produce vibration (Solé, 2002, 2007ab). Unlike trills, lateral approximants are not incompatible with the velum lowering and nasal airflow of adjacent nasal consonants.
However, the production of laterals requires lowering of the tongue sides and maintaining substantial oral airflow, which can interfere with the oral occlusion and the nasal-only airflow for nasal stops. Laterals are also known to be relatively resistant to coarticulatory influences from adjacent vowels and consonants (Recasens & Pallarès, 2001).
Altogether, these factors make sequences of nasal + trill and nasal + lateral sub-optimal from the point of view of speech production. Not surprisingly, there is ample evidence from historical linguistics that [n+r] and [n+l] clusters undergo various changes involving, assimilation, metathesis, deletion, or epenthesis. Such changes, for example, have occurred in word clusters in the history of Romance languages: nr/nl rr/ll (*inreal irreal ‘unreal’), nr rn (GENERU yerno ‘son in law’), nr ndr (poner he pondré ‘I will put’) (Lloyd, 1993; Ariza, 2012). Despite these developments, however, clusters of nasals with laterals and rhotics still marginally occur in modern Romance languages, mainly across morpheme boundaries or in sequences of words (e.g. en+roscar ‘to screw’; en+listar ‘to enroll’; hacen ruido ‘they make noise’; hacen lío ‘they make a mess’ in Spanish).
The goal of this paper is to determine how the incompatibility between trills/laterals and nasality is resolved synchronically at the level of articulation. Specifically, we are using electropalatography (EPG) to examine acrossand within-word clusters /n/ + /r/ and /l/ in Spanish, produced by speakers of three different dialects. EPG tracks the contact between the tongue (the tip, the blade, and the front part of the dorsum) and the roof of the mouth using an artificial palate with built-in electrodes. Previous EPG work on the closely related Catalan language (Recasens & Pallarès, 1999b, 2001) showed that the relatively unconstrained alveolar and palatal nasals /n/ and /ɲ/ assimilated to the following highly constrained apical trill /r/ in constriction location (becoming more retracted). The alveolar lateral approximant /l/ was also found to show some resistance to coarticulation, particularly to the raising of the tongue back. Given this, we would expect to find similar effects in our data, with nasals assuming a constriction and a tongue shape configuration similar to /r/ and /l/ respectively. It may be possible that some aspects of the trill and lateral articulation are also affected by the preceding nasal, resulting in a shift in the constriction location or degree, or in the loss of trilling. Such a change in the liquid, especially a tighter constriction after a nasal, would be consistent with other phonological processes observed in the language, specifically the voiced stop (after a pause or a nasal) vs. approximant (other contexts) alternations. It was also of interest to examine other potential influences, such as stress, word/morpheme boundary, and vowel quality. Given the previous finding of the liquids (and particularly the trill) being resistant to coarticulation or prosodic effects (Recasens & Pallarès, 1999ab, 2001), these were expected to play a minor role.
Participants The participants were nine native speakers of Spanish: five from Buenos Aires, Argentina (hence referred to as A1-A5), three from Havana, Cuba (C1-C3), and one from Madrid, Spain (P1). All were female except one male (A5), and were 23 to 49 (on average 39) years old. The speakers had university education and at the time of the experiment were residing in Toronto, Canada. All the participants reported to use Spanish extensively on a daily basis and reported no history of hearing or speech difficulties. For the experiment, the participants wore custommade artificial EPG palates with 62 electrodes.
Proceedings of Meetings on Acoustics, Vol. 19, 060273 (2013) Page 2 A. Kochetov and L. Colantoni Materials The stimuli used in this study are presented in Table 1. They involved target items with nasal + /r/ and /l/ clusters in several phonetic contexts: across and within (hetero- and mono-morphemic) words, preceded by unstressed and stressed vowels, and by vowels of different quality (front or back, low or mid). The control items included single intervocalic /r/ and /l/, and the nasal before /t/ in some of the vocalic and stress contexts (previously used in Kochetov & Colantoni, 2012ab). The stimuli were randomized and produced with other utterances. The participants were instructed to read the sentences at a normal, casual speaking rate.
Stimuli with the low vowel contexts (a_##_á and á _##_á) were used for the primary sets of comparisons – to examine the effect of the following /r/ or /l/ on the nasal, or the effect of the preceding nasal on /r/ and /l/. They were also used to examine the effect of stress on the realization of both the nasal and the liquids. The other items were used to explore possible contributions of word and morpheme boundaries and vowel quality.
Our previous EPG work (Kochetov & Colantoni, 2012a) showed that, despite the differences in dialect background, rhotics and laterals (/r/, /ɾ/, /l/) were produced overall similarly across speakers (as apical anterior/posterior alveolars), with the exception of a somewhat more posterior realization of /r/ by Cuban speakers, consistent with frequent pre-aspirated realizations (see Quilis, 1993). The choice of the /nt/ cluster as a nasal control was also motivated by the relative uniformity in its production (as laminal denti-alveolar) across dialects; see Kochetov & Colantoni, 2012b). Note that in the Argentine and Peninsular (Madrid) dialects a syllable-final nasal occurring utterance-finally or prevocalically is realized as alveolar [n], while in the Cuban dialects it is commonly velar [ŋ]. In all dialects, however, the nasal assimilates to a following consonant, becoming similar to it in place and stricture (Navarro Tomás, 1918; Hualde, 2005). Figure 1 presents averaged linguopalatal contact profiles for the nasal in [nt] clusters in two stress conditions (a) and single intervocalic /r/ and /l/ (b). It can be seen that the nasal before [t] has a constriction in the first 2-3 rows of the palate (laminal denti-alveolar), regardless of the stress condition. This realization is almost identical to that of the denti-alveolar stop /t/ (see Kochetov & Colantoni, 2012b). In contrast, the trill involves a narrow constriction in rows 2 or 3 (apical alveolar), with an often incomplete closure; it also has a lower side contact further back, indicative of the tongue dorsum lowering. The location of the lateral constriction ranges from rows 1-2 to 3-4 across the speakers, yet it is a consistently tighter constriction than for /r/ and has considerably less side contact (in order to produce the lateral airflow) than for either /r/ or /t/. Given these differences, we would expect the nasal before /r/ and /l/ to have contact patterns similar to these consonants and significantly different from the nasal before /t/.
FIGURE 1. Linguopalatal EPG contact profiles of the nasal before /t/ (a; from Kochetov and Colantoni, 2012b) and single liquids /l/ and /r/ (b; from Kochetov and Colantoni, 2012a) produced by 9 Spanish speakers (the same as in the current study); the means are based on 6 tokens from the first recording session; black = contact 100% of the time, white = contact 0% of the time, with shades of grey indicating intermediate values).
Instrumentation and Analysis A WinEPG system by Articulate Instruments (Wrench et al., 2002) was used to collect simultaneous articulatory and acoustic data, sampled at 100 Hz and 22,050 Hz respectively. The recordings were done in the Linguistics Phonetic Lab at the University of Toronto. They were performed in two separate sessions, with 6 repetitions of each sentence elicited per session, with the exception of speakers C2 and C3, for whom only one session (6 repetitions) was performed. This gave us in total 1,344 tokens for analysis (14 stimuli x 12 repetitions x 7 participants + 14 stimuli x 6 repetitions x 2 participants).
The Articulate Assistant software (Wrench et al., 2002) was used for data collection, segmentation, annotation, and analysis. Measurements of the tongue-palate contact were made both at the midpoint and at the point of maximum contact (PMC) of both C1 (nasal) and C2 (/r/ an /l/). The results in this paper are based on PMC only, which was defined as the frame showing the highest number of activated electrodes within the interval of C1 or C2.
Extracted PMC data were converted to several articulatory indices described below (following Fontdevila et al.
• Contact Anteriority in the alveolar region (CAa): the frontmost position of the constriction in the first 5 rows, with higher values corresponding to a more anterior constriction.
• Contact Posteriority for the alveolar region (CPa): the backmost position of the constriction in the first 5 rows, with higher values corresponding to a more posterior constriction.
• Contact Centrality for the alveolar region (CCa): the degree of central occlusion in the 5 central columns of the first 5 rows, with higher values corresponding to a greater central occlusion.
• Quotient of activation for the palatal region (Qp): the amount of contact in the last 3 rows, with higher values corresponding to sounds with greater palatal constriction.
Statistical analyses included Repeated Measures ANOVAs based on individual means for the indices CAa, CPa, CCa, and Qp of C1 and C2 of the nine participants per each condition. Bonferroni posthoc pairwise comparisons were performed to determine differences between levels within factors. Given the variation in the realization of the trill, the articulatory analysis was supplemented by an informal acoustic inspection, using Praat (Boersma & Weenink, 2012).
3-4 (except for P1 who has a very front constriction in rows 1-2). The constriction is occasionally incomplete and covers a narrow part of the alveolar ridge. It is thus quite different from the nasal before /t/ and similar to the single /r/ in Figure 1. The same can be said about the nasal before /l/, which is also characterized by markedly less side contact in the posterior half of the palate. Individual differences in the degree of retraction and the (in-)completeness of the constriction seem parallel to those observed in the realization of the single liquids (Figure 1).
FIGURE 2. Linguopalatal EPG contact profiles of the nasal before /r/ and /l/ separately for each speaker; the means are based on 6 tokens from the first recording session.