Skip to content
Publicly Available Published by De Gruyter Mouton March 17, 2015

Syllable structure and word stress effects in Peninsular Spanish nuclear accents

Nicholas Henriksen EMAIL logo
From the journal Laboratory Phonology

Abstract

In this study we analyzed temporal alignment between F0 turning points and acoustic landmarks in rising (L+¡H*) and falling (H+L*) nuclear pitch accents in Peninsular Spanish wh-questions. In the research design we devised two experimental factors based on nuclear syllable configuration: syllable structure (open vs. closed) and stress position (penultimate vs. ultimate). Regarding leading tone alignment, the L point of L+¡H* displayed close synchrony with the start of the nuclear syllable, whereas the H point of H+L* was more variable within the pretonic syllable. These findings provide only partial (i.e., accent-specific) confirmation for the syllable onset anchoring hypothesis. Regarding starred tone alignment, both accents showed the same result: syllable structure did not affect alignment in words with penultimate stress. Although these findings support the principle of segmental anchoring in nuclear position, we propose that anchoring landmarks for tonal targets may constitute entire segments themselves. All in all, this work contributes to the study of intonational phonology by documenting that patterns of temporal alignment are specific to individual pitch accent specifications.

1 Introduction

1.1 Tune-text coordination and syllable structure effects in intonational phonology

A key finding from the volume of research on tonal alignment is that the beginnings and ends of bitonal pitch accents coordinate with the segmental string in various ways. The collective findings of this research program have now led to a series of hypotheses on tune-text coordination in intonational phonology, based on acoustic evidence from multiple languages (e.g., Arvaniti et al. 1998 for Greek; Xu 1998 for Chinese; Ladd et al. 1999 for English; see also Atterer and Ladd 2004 for German; Schepman et al. 2006 for Dutch). However, such hypotheses are rarely tested on multiple accents within a particular language variety to assess their relevance within a single intonational grammar – this is the objective of the present analysis.

In keeping with the autosegmental-metrical (AM) framework to intonational phonology, it is assumed that utterance melodies are the realization of abstract sequences of H(igh) and L(ow) tonal targets (Pierrehumbert 1980; Beckman and Pierrehumbert 1986; Ladd 2008). One chief discovery within the AM framework is that melodic tunes in intonational languages may be decomposed into tones that associate with the segmental string in one of two ways: with metrically strong syllables (pitch accents), or with phrase edges (boundary tones or phrase accents). Pitch accents may constitute one tone (monotonal accents) or two tones (bitonal accents). In bitonal pitch accents, the starred (*) tone has a primary association with the metrically strong syllable (typically through alignment of the pitch accent with the segmental string); the non-starred tone precedes or follows the starred tone. [1] Thus, for accents like L+H* the L tone is considered the leading tone (it precedes the starred tone), whereas for accents like L*+H the H tone is considered the trailing tone (it follows the starred tone).

In acoustic terms, the most obvious manifestations of tonal targets are local F0 turning points or ‘elbows’ where F0 changes slope. Phonetically, tonal targets are defined by F0 minima or maxima quantified on two dimensions: F0 height and relative alignment to a segmental landmark. Early work on tonal alignment uncovered discrepancies between L valleys and H peaks in rising accents in a variety of languages: L valleys anchor to the onset of a stressed syllable, whereas H peaks are more variable (Prieto et al. 1995; Arvaniti et al. 2000). Work in this area also documented that F0 targets abide strictly by segmental anchoring, which is the notion that strict alignment effects on the L and H points of a single accent are not affected by changes in syllabic structure, segmental structure, or speech rate (Ladd et al. 1999; Arvaniti et al. 2000; Ladd et al. 2000). In his monograph on intonational phonology and autosegmental-metrical theory, Ladd (2008: 178) proposes that segmental anchoring is a cross-linguistic phonetic regularity: it can be relative to landmarks within the segmental string (e.g., “20 ms after a vowel onset”) or relative to a phonological domain (e.g., “onset of accentual syllable”). Of importance to the present analysis, phrase position can interact with alignment, causing F0 points on a nuclear stressed syllable to align earlier than comparable F0 points on a prenuclear stressed syllable. Multiple explanations have been offered to explain this effect. On the one hand, earlier alignment in nuclear accents may result from tonal crowding with the upcoming phrase or intonational boundary (Hualde 2002; Schepman et al. 2006; Ladd et al. 2009; but see Prieto 2009). On the other, earlier nuclear alignment may result from inherently different tonal compositions in prenuclear vs. nuclear accents (Face 2002).

For Spanish, the effect of syllable structure (i.e., open vs. closed syllables) on tonal alignment was studied in Prieto and Torreira (2007). Specifically, the authors tested the alignment of L and H turning points in Peninsular Spanish L+H* accents in non-final position. Their results showed that in open syllables the H peak was located near the end of the stressed vowel, whereas in closed syllables it was near the beginning of the coda consonant, far away from the syllable boundary. The authors claimed that since the H peak was not anchored consistently at a segmental landmark such as a vocalic or syllabic offset, this constituted evidence against a strict view of the segmental anchoring hypothesis (Arvaniti et al. 1998; Ladd et al. 1999). Similarly, D’Imperio (2000) found that H peaks were located closer to the vowel offset in closed syllables than in open syllables for Neapolitan Italian rising accents (see also Gili-Fivela and Savino 2003). Regarding speaking rate, studies show that tempo differences manifest instability effects on H peak alignment in particular (e.g., Xu 1998; Ishihara 2006). This accumulation of findings motivated some authors to put forth a weakened version of the segmental anchoring hypothesis, namely segmental anchorage (Welby and Lœvenbruck 2005, Welby and Lœvenbruck2006). Under the principle of segmental anchorage, it is feasible to posit a temporal range – as opposed to a fixed point in the segmental string – within which a turning point may coordinate.

Researchers in intonation have also tested F0 alignment with respect to other acoustic landmarks, namely syllable boundaries. Using acoustic data, Prieto (2009) studied the effect of syllable structure in Central Catalan H+L* nuclear falling accents. In open syllables the L aligned at the syllable offset, whereas in closed syllables it aligned just before the coda consonant. [2] A second finding of Prieto’s analysis was that the H peak of the fall movement was not affected by syllable structure and aligned at most 15–20 ms before the onset consonant. This confirmed that the syllable-initial boundary was the point at which greatest synchrony was achieved between acoustic landmarks and F0 elbows. Support for this position comes from the examination of bitonal accents consisting of a leading and a starred tone (i.e., L+H* or H+L*). In these accents, the leading tone is simultaneous with the stressed syllable onset, whereas the starred tone is more variable within (the last part of) the rhyme (e.g., Prieto and Torreira 2007). Of note, Prieto and Torreira (2007) and Prieto (2009) employ the terminology coupling mode hypothesis to describe the coordination between F0 turning points and syllable boundaries, framing their data in articulatory terms. For example, Prieto (2009: 874) interprets her findings such that “the coordination between tonal gestures and supraglottal gestures seems to be tighter (perhaps in-phase) in syllable-initial position and more variable in syllable-final position” (see also Prieto and Torreira 2007: 492). Since in the current study we use acoustic data only (i.e., we do not present kinematic results; cf. D’Imperio et al. 2007; Mücke et al. 2009; Niemann et al. 2011), we cannot make assessments with respect to articulatory-based coupling hypotheses. However, our experimental design does permit the assessment of whether there are co-occurrences between acoustic events (syllable edges) and tonal events (F0 elbows). This is the principle behind our syllable onset anchoring hypothesis, which we define in Section 1.3.

In this paper we study the effects of syllable structure and word stress on turning point alignment in H+L* and L+¡H* nuclear accents in Peninsular Spanish wh-questions. These configurations occur in most varieties of Peninsular Spanish, documented for northern, central, and southern varieties (Prieto 2004; Henriksen 2009; Estebas-Vilaplana and Prieto 2010; López‐Bobo and Cuevas‐Alonso 2010; Henriksen and García-Amaya 2012; Henriksen 2013). The phonetic properties of the H+L* and L+¡H* accents were examined in Henriksen (2014), where it was shown that the leading tones aligned within the last portion of the prestressed syllable and the starred tones aligned within the second half of the nuclear syllable. However, a recognized caveat was that only open syllables were included in the database of test sentences in Henriksen (2014). It remains to be seen whether the alignment of starred tones remains synchronous with the syllable edge when the nuclear rhyme incorporates a coda consonant. Given this background, the objective of this paper is to investigate the temporal organization of the F0 turning points in falling (H+L*) and rising (L+¡H*) nuclear pitch accents in Peninsular Spanish and to determine their temporal coordination with prosodic structure. Specifically, we focus on the interaction between syllable structure and word stress effects for these accents in wh-questions as a means of assessing two positions on temporal coordination effects in the AM framework: the segmental anchoring hypothesis and the syllable onset anchoring hypothesis. In the next section we proceed with a more complete description of these accents as they relate to the intonational grammar of Spanish.

1.2 H+L* and L+¡H* accents in Spanish

The Spanish ToBI (Sp_ToBI) proposal was initially put forth by Beckman et al. (2002) and has undergone subsequent revisions as new laboratory data are published on the intonational complexities of pan-Hispanic intonation (Face and Prieto 2006/2007; Estebas-Vilaplana and Prieto 2008; see also Aguilar et al. 2009). For Peninsular Spanish, there is a history of research on the intonational patterns of speakers from the northern and central areas of Spain (e.g., Face 2002, Face 2008, Face 2011; Prieto 2004; Estebas-Vilaplana and Prieto 2010), but the extent of intonational variation throughout the peninsula is not completely known. For Spanish generally, boundary falls are widely observed in wh-question intonation (Navarro Tomás 1944; Sosa 1999; Dorta Luis 2000; Sosa 2003; Prieto 2004; Henriksen 2009; Armstrong 2010; Willis 2010, among others), but detailed phonetic accounts with phonological implications are scarce. Prieto (2004) examined the speech data of two speakers from northern Spain and found that the most frequent contour contained a low-rise configuration (L*H%), although one speaker occasionally produced a contour comprised of a nuclear fall (H+L* L%). Henriksen (2009), Estebas-Vilaplana and Prieto (2010), and López‐Bobo and Cuevas‐Alonso (2010) likewise documented use of a nuclear falling contour in European Spanish wh-questions. One important drawback of these studies is that phonetic details on tune-text coordination were not offered since, for many of these authors, this was not the primary objective of their research.

A recent study on falling and rising accents in Peninsular Spanish examined the alignment behavior of H+L* L% and L+¡H* L% configurations with respect to the nuclear syllable under different contexts of tonal pressure (Henriksen 2014). [3] In Spanish, the H+L* accent is realized phonetically as a descending pitch movement over the stressed syllable, whereas the L+¡H* accent is realized as a rise to a peak located within the stressed syllable (see Figure 1). The acoustic results showed that the end of each F0 movement coordinated strongly with the nuclear syllable. Specifically, the valley of H+L* and the peak of L+¡H* were fixed within the nuclear vowel, motivating the starred notation for the L and H points, respectively. The fact that the nuclear peak in L+¡H* exhibited relatively high F0 (i.e., it was significantly higher than the utterance-initial peak) motivated an upstep analysis (¡H). [4] However, the focus of the paper was on the acoustic behavior of wh-question nuclear accents in open syllables only; on the whole, data on syllable structure effects in Spanish intonation are limited. Prieto and Torreira (2007) examined syllable structure effects for rising L+H* accents in non-final position, but it is unclear if the findings of their paper would apply to accents in nuclear (i.e., final) position. This is because one of the major points of debate in Spanish intonation is whether prenuclear and nuclear accents behave in an analogous fashion. Some researchers suggest that there is an essential difference between Spanish prenuclear and nuclear accents with respect to their phonological composition (Face 2002; Henriksen 2014; but see Hualde 2002), so it is not unreasonable to expect differences in the treatment of open vs. closed syllables in these positions. In sum, the findings of this paper stand to contribute new information to the prosodic grammar of Spanish, the description of its intonational melodies, and status of prenuclear vs. nuclear accents in the prosodic hierarchy. Specifically, we will explore the nature of syllable structure and word stress effects (and the interactions between them) for two different nuclear accents in Peninsular Spanish wh-questions: L+¡H* and H+L*.

Figure 1

Schematic of L+¡H* and H+L* accents in nuclear position in Peninsular Spanish wh-questions. Dotted line indicates F0 contour for L+¡H* accent, and solid line indicates F0 contour for H+L* accent. Shaded columns indicate stressed syllables for the hypothetical utterance ¿Quién ga el premio? ‘Who won the prize?’

1.3 Hypotheses

Given this background, we can put forth two hypotheses on temporal coordination patterns in Spanish wh-question intonation.

According to the segmental anchoring hypothesis, in the absence of tonal pressure, changes in the syllabic composition of stressed syllables should not modify the positioning of F0 turning points with respect to their anchoring landmark. In this paper we test this hypothesis for the L and H starred tones of the H+L* and L+¡H* accents, respectively. To support this hypothesis, we would expect the L valley of H+L* and the H peak of L+¡H* to remain synchronous with a fixed anchoring point and to demonstrate insensitivity to the following coda consonant (i.e., they will not deviate into the coda consonant in closed syllables). We would also expect syllable structure to interact with word stress, since nuclear syllables in oxytone words must realize the end of the F0 movement as well as the boundary fall within a restricted time domain. Tonal crowding effects are expected under the segmental anchoring hypothesis. On the other hand, given the findings of Prieto and Torreira (2007) and Prieto (2009), an alternative possibility is that starred tones will manifest later alignment (i.e., they may deviate into the coda consonant) in CVC syllables. This outcome would be problematic for the strict segmental anchoring hypothesis since it would confirm turning point instability under modifications in syllable structure.

According to the syllable onset anchoring hypothesis, F0 movements at stressed syllable onsets align more closely to the syllable boundary than F0 movements at stressed syllable offsets. To support this hypothesis, we might expect the F0 peak of the H+L* accent and the F0 valley of the L+¡H* accent to align more closely with the syllable-initial boundary (i.e., with the start of the onset consonant of the nuclear syllable), whereas the L valley of H+L* and the H peak of L+¡H* will display more variability with respect to the syllable-final boundary when coda consonants are present. These were the findings of Prieto (2009) for Central Catalan falling accents and of Prieto and Torreira (2007) for Peninsular Spanish rising accents. For example, H distance to the stressed syllable onset was typically 15 to 20 ms before the syllable boundary in Catalan H+L* accents. On the other hand, we might find more variability in leading tone alignment for both accents, namely it may deviate away from the syllable-initial boundary into the preaccentual vowel (leftward) or into the onset consonant or accentual vowel of the stressed syllable (rightward). These patterns are not predicted by the syllable onset anchoring hypothesis.

These hypotheses are motivated on various empirical grounds. First, a core finding from work on intonational languages is that there is a phonological basis for studying the temporal coordination of F0 turning points (e.g., Pierrehumbert and Steele 1989; D’Imperio and House 1997). Within the AM framework, it is assumed that F0 turning points are the phonetic manifestations of phonological tones (Bruce 1977; Ladd and Schepman 2003; Dilley et al. 2005). However, most research in this area focuses on rising accents; to our knowledge, only one experiment has been carried out on falling accents (Prieto 2009). Because of this, the present study could be regarded as an important addition to the body of work testing different theories on tune-text alignment. Second, we still lack empirical consensus on the validity of theoretical positions on segmental alignment – namely, the segmental anchoring hypothesis and the syllable onset anchoring hypothesis – and on their cross-linguistic applicability and their validity for different accent types (e.g., rises vs. falls) within a single language. Third, our knowledge on temporal coordination in Spanish is limited, especially for the effect of syllable structure on segmental alignment. The findings of Prieto and Torreira (2007) indicate later alignment in closed syllables in non-final position, but clearly there is a need for more detailed inquiry in this language, such as looking at different phrasal positions. Fourth, there are conflicting results regarding the effect of syllable structure on F0 alignment in nuclear position. For example, Schepman et al. (2006), Ladd et al. (2009), and Prieto (2009) all studied F0 alignment in phonetically short vs. long syllables in nuclear position. On the one hand, Schepman et al. (2006) and Ladd et al. (2009) did not find evidence that vowel length affects F0 alignment in syllables that contain short vs. long vowels in nuclear position. On the other, Prieto (2009) showed earlier F0 alignment in CV vs. CVC syllables in nuclear position. [5] At this juncture, it is not clear whether nuclear accents behave in a consistent fashion across languages, at least as far as syllable structure effects are concerned.

2 Procedure

2.1 Speech materials

For this experiment, speech data were produced in a sentence reading task designed to elicit a total of 80 wh-question test sentences via computer-based prompts. Each test sentence contained three or four lexically stressed words, and all began with the wh-word quién ‘who’ or cuándo ‘when’. The full database of test sentences is provided in Appendix A. Care was taken to maintain consistency in the total number of syllables prior to the utterance-final word such that in all cases eight syllables preceded the nuclear syllable. Also, in all cases a minimum three unstressed syllables preceded the nuclear syllable to avoid tonal clash effects with the previous prenuclear syllable (for discussion, see Prieto 2005). For clarity purposes, we use the terms ‘prenuclear’ and ‘nuclear’ in this paper to refer to stressed syllables with regard to their placement in an utterance. Prenuclear syllables are non-final stressed syllables, and nuclear syllables are utterance-final stressed syllables. [6]

The utterance-final word of each test sentence was created based on two manipulation parameters: nuclear syllable structure (open vs. closed) and word stress (penultimate vs. ultimate). [7] Intervocalic voiced consonants were chosen for utterance-final words because they permit the entire word to be voiced and allow for a continuous pitch track without F0 breaks. Example test sentences based on each combination of manipulation parameters are provided in (1) and (2) (stressed syllables are underlined, target syllables are in bold). There were five test sentences per condition, and each sentence was included four times, for a total of 80 test sentences in the experiment (4 conditions × 5 sentences × 4 repetitions = 80). No fillers were included in the speaking task.

(1)

Sample test sentences: open nuclear syllable

Penultimate stress: ¿Quién es bailando con laniña? ‘Who is dancing with the girl?’

Ultimate stress: ¿Quién es hablando con Ma? ‘Who is talking to Mom?’

(2)

Sample test sentences: closed nuclear syllable

Penultimate stress: ¿Quién es saliendo con Armando? ‘Who is going out with Armando?’

Ultimate stress: ¿Cuándo salisteis para Bailén? ‘When did you leave for Bailén?’

2.2 Speakers and recording procedures

Twelve speakers (seven men, five women) were recruited for participation in this study; their ages were between 18 and 24, and their average age was 20. Speakers were students at the Universidad de Castilla-La Mancha in Toledo, Spain, located approximately 70 km south of Madrid. All subjects reported that their mother and father were natives of Toledo. The speech data were recorded using a Marantz Pmd620 digital recorder and a Shure WH20 head-mounted microphone.

Each speaker read aloud the set of 80 wh-questions in a sentence reading task presented in Microsoft PowerPoint. Each slide contained one sentence, and sentences were sorted randomly and were presented without a previous discourse context. Speakers were instructed to read aloud the sentences keeping in mind a neutral context and to make their best effort to maintain the same intonational melody for all sentences. In most cases there was no need to monitor speakers, since the H+L* L% and L+¡H* L% configurations are used interchangeably as the wh-question melody in this variety of Spanish (see Henriksen 2013). The topic of intonational variation in Spanish has been addressed in previous research across dialects (Prieto and Roseano 2010) and also within dialects (Henriksen 2012, 2013). Since it is known that speakers can display variability in the type of utterance-final configuration that they use (due to stylistic or pragmatic factors), we needed to ensure that each speaker was fairly consistent in order to obtain a balanced data sample in his/her corpus. In a few cases speakers acknowledged that different configurations were possible at the right-periphery (e.g., H+L* L%, L+¡H* L%, or L* H%); thus they were asked to use the configuration with which they felt most comfortable.

Speakers were asked to repeat sentences containing any type of disfluency or unwanted phrasing. In total, five speakers (three women, two men) produced the H+L *L% configuration, six speakers (one woman, five men) produced the L+¡H* L% configuration, and one speaker (one woman) produced the L* H% configuration. The L* H% configuration is well-documented in Spanish question intonation, and the findings of experimental research (Henriksen 2013) suggest that this configuration is an artifact of read speech protocols, although there is possibly an interplay with pragmatic meaning (see also Savino 2012). Since the L* H% pattern was limited to one speaker, we discarded these data from the analysis and focused on the production of the remaining 11 speakers: five speakers produced the H+L* L% configuration, and six speakers produced the L+¡H* L% configuration.

2.3 Acoustic analysis

Syllable duration and F0 measurements were obtained for all target productions using the computer software Praat (Boersma and Weenink 2013). F0 contours were manually segmented and labeled by simultaneous inspection of waveforms and wide-band spectrograms following standard segmentation criteria (see Peterson and Lehiste 1960). F0 points presumed to represent tonal targets were also located manually.

Figures 2 through 5 show the acoustic output of four typical utterances, together with the labels for the putative intonational targets. In Figures 2 and 3 we show output for the H+L* accent produced in open and closed nuclear syllables, and in Figures 4 and 5 we show output for the L+¡H* accent in the same conditions. Following Arvaniti et al. (1998) and Schepman et al. (2006), segment labels represent the start of the segment in question (C = onset consonant; V = vowel; K = coda consonant), and syllable number labels represent syllable stress (0 = accented; 1 = postaccentual).

  1. C0: start of onset consonant of nuclear syllable

  2. V0: start of vowel of nuclear syllable

  3. K0: start of coda consonant of nuclear syllable

  4. C1: start of onset consonant of postnuclear syllable

  5. V1: start of vowel of postnuclear syllable

  6. H: turning point at start of fall (H+L* accent) or end of rise (L+¡H*)

  7. L: turning point at end of fall (H+L*) or start of rise (L+¡H*)

Figure 2

Waveform, spectrogram, and F0 for the H+L* accent in an open nuclear syllable in the test utterance ¿Quién se ha operado de la mano? ‘Who got their hand operated on?’

Figure 3

Waveform, spectrogram, and F0 for the H+L* accent in a closed nuclear syllable in the test utterance ¿Quién es saliendo con Armando? ‘Who is going out with Armando?’

Figure 4

Waveform, spectrogram, and F0 for the L+¡H* accent in an open nuclear syllable in the test utterance ¿Quién es viviendo con la niña? ‘Who is living with the girl?’

Figure 5

Waveform, spectrogram, and F0 for the L+¡H* accent in a closed nuclear syllable in the test utterance ¿Quién es viviendo con Raimundo? ‘Who is living with Raimundo?’

F0 points presumed to represent tonal targets were located manually. For all turning points, we ignored (as far as possible) any nasal-induced perturbations and obviously spurious values due to irregular voicing. For the H+L* accent, the H point was identified as the moment in the F0 contour when the previous high plateau turned into a clear fall; the L point was identified as the end of the steepest falling movement before the utterance-final low level stretch. For the L+¡H* accent, the L point was identified as the moment when the long stretch of high pitch after the initial rising accent turned into a clear rise. The H point was the peak of the nuclear rise, typically within the vowel of the nuclear stressed syllable. In those cases for which H peaks of the L+¡H* accent formed a plateau (i.e., no clear F0 value emerged as the highest point), we chose the point where the highest rate of F0 change occurred (cf. del Giudice et al. 2007). In those rare cases when microprosodic effects interrupted the F0 track (i.e., there was an abrupt lowering of F0), these were visually disregarded and not included as F0 valleys. Since the test consonants used to create the nuclear syllable were always sonorants, this was an uncommon problem. For H+L* utterances, 7 productions were discarded from the original corpus of 400, leaving 393 (98.2%) submitted to acoustic analysis. For L+¡H* utterances, 10 productions were discarded from the original corpus of 480, leaving 470 (97.9%) submitted to acoustic analysis. [8] Acoustic data were extracted using a series of Praat scripts.

As a check on the reliability of the measurement procedures, the H and L turning points of each accent were re-labeled for 100 productions in each accent, or approximately 20% of each database. Similar reliability criteria have been utilized elsewhere (e.g., Arvaniti et al. 2006). The absolute differences for each set of L and H measurements were calculated, and means and standard deviations were computed. The absolute mean differences for the H+L* accent were: 3.1 ms (SD = 8.9) for H; 3.0 ms (SD = 8.0) for L. The absolute differences for the L+¡H* accent were: 2.0 ms (SD = 7.9) for L; 2.9 ms (SD = 9.9) for H. At this point, there is no obvious reason to think that our conclusions are impaired by measurement error or bias.

2.4 Statistical analysis

In the statistical analysis we ran three linear mixed-effects models (LMEMs) with random effects of Speaker and Sentence within Speaker (thus allowing for two levels of correlations) and with fixed effects of Accent (H+L* or L+¡H*), SyllableType (open or closed), Stress (penultimate or ultimate), and the interaction combinations between them. Each LMEM corresponded to one of the three dependent variables in the analysis: leading tone distance to C0 onset; starred tone distance to V0 onset; and starred tone distance to V0 offset. To examine potential syllable structure differences in paroxytone vs. oxytone words between accents, post-hoc comparison tests with the Bonferroni correction were performed on the interaction combinations when these interactions returned significant results. In the following sections we report results for statistically significant effects only, as they related to the fixed factors and interaction combinations in each LMEM. For the complete report of results (including those factors which did yield significant effects on the dependent variables), see Appendix B.

All models were fitted using the MIXED procedure in SPSS (version 22), and degrees of freedom for approximate F-statistics for the fixed effects were computed using a Satterthwaite approximation (these degrees of freedom were used to determine p-values for F statistics). There is much debate over how degrees of freedom should be computed for approximate test statistics based on the fits of LMEMs, which do not follow exact F or t distributions. Given this debate and other methods that have been proposed for testing significance (e.g., Reubold et al. 2010), we fit our same models using the Kenward–Roger degrees of freedom approximation in SAS (see West et al. 2014) and did not find differences in terms of the overall significance of the approximate test statistics.

2.5 Syllable duration and speech rate

Prior to computing alignment data, we analyzed the effects of syllable structure and word stress on overall syllable duration. Since in this paper we are dealing with F0 timing based on syllable structure, we wanted to determine whether closed syllables were in fact longer than open syllables (see Maddieson 1985). We also wanted to determine whether vowel durations were longer in oxytone words due to phrase-final lengthening. Quantitative data are summarized in Table 1 based on nuclear accent type and manipulation condition. As a first observation, closed syllables are longer in duration than open syllables. A second observation is that ultimate stressed syllables are longer in duration than penultimate stressed syllables. We attribute this finding to final-lengthening in phrase-final position, a trend that is more robust in other languages like English, but is still prevalent in Spanish (Hualde 2005; Ortega-Llebaria and Prieto 2007).

Table 1

Nuclear syllable duration (in milliseconds) for each accent condition.

AccentConditionMean (ms)SD (ms)
H+L*Open penultimate15624
Open ultimate19038
Closed penultimate22130
Closed ultimate26447
L+¡H*Open penultimate14221
Open ultimate18633
Closed penultimate21434
Closed ultimate23738

We also wanted to determine comparability of results between speaker groups based on nuclear accent (H+L* and L+¡H*) and therefore calculated speech rate in syllables/second for each speaker group (individual data are reported in Appendix C). For H+L* speakers the average speech rate was 8.01 syll/sec (SD = 1.1 ms). For L+¡H* speakers the average speech rate was 7.95 syll/sec (SD = 0.9 ms). Given these similar data distributions, we chose to calculate alignment in this paper as a raw metric based on distance in milliseconds between an F0 turning point and a particular segmental landmark. However, we acknowledge that normalized alignment scores might more appropriately account for speech rate differences based on the other linguistic factors under study here (e.g., syllable structure and word stress).

3 Results

3.1 Alignment of leading tones

In this section we examine alignment of the start of each accent (i.e., the H point of H+L* and the L point of L+¡H*) with respect to the onset of the nuclear syllable. The data in Figures 6 and 7 present a distributional analysis of alignment scores relative to the start of the nuclear onset consonant for the four prosodic conditions that motivated the experimental design: open nuclear syllable, penultimate stress; open nuclear syllable, ultimate stress; closed nuclear syllable, penultimate stress; and closed nuclear syllable, ultimate stress. Negative scores represent alignment prior to the nuclear syllable, and positive scores represent alignment within the nuclear syllable. The data reveal a consistent trend: in almost all cases the start of the accent aligns within the pretonic syllable. It is very rare that speakers realize the H peak (of H+L*) or the L valley (of L+¡H*) within the onset consonant of the stressed syllable (although we acknowledge that error bars are large for a few speakers). As a second result, ranges of variation were greater for the H peak than for the L valley: leading H alignment was typically within the 30 ms to 80 ms interval before the nuclear consonant, whereas leading L alignment was more synchronous with the right edge of the pretonic syllable, within the 10 ms to 30 ms interval before the nuclear consonant. Recall that for Central Catalan H+L* accents, H alignment was typically between 15 and 20 ms of the syllable onset, retracting little into the pretonic syllable (Prieto 2009). In our data we show a comparable outcome for the L+¡H* accent only.

Figure 6

Distance (in milliseconds) from leading H point to nuclear onset consonant for H+L* accent based on manipulation condition. Error bars indicate 95% confidence interval.

Figure 7

Distance (in milliseconds) from leading L point to nuclear onset consonant for L+¡H* accent based on manipulation condition. Error bars indicate 95% confidence interval.

Regarding statistical results, the LMEM performed on the dependent variable coded for leading tone distance to C0 returned significant effects of Accent (F(1,9.044) = 23.845, p = 0.001), Stress (F(1,188.175) = 10.430, p = 0.001), and the interaction factor Accent × Stress (F(1,188.175) = 19.302, p ≤ 0.001). A Bonferroni adjustment used to determine the effects of factors in the Accent × Stress interaction returned a significant result for stress in the H+L* accent (t = 5.010, p ≤ 0.001), but not in the L+¡H* accent (t = 0.667 p = 0.390). The data in Table 2 reveal that the leading H was earlier in ultimate stress: it was 9 ms earlier in open ultimate syllables (compared to open penultimate syllables) and 20 ms earlier in closed ultimate syllables (compared to closed penultimate syllables). On the other hand, the leading L was phonetically invariant, and the greatest difference across conditions was 2 ms. Taken together, these data support the position that the L valley of L+¡H* displays stronger simultaneity with the onset of the nuclear syllable than the H peak of H+L*.

Table 2

Distance (in milliseconds) from leading tone to nuclear consonant onset. Negative scores indicate temporal alignment prior to nuclear syllable.

TargetConditionMean (ms)SD (ms)
H (H+L*)Open penultimate−3927
Open ultimate−4833
Closed penultimate−3331
Closed ultimate−5340
L (L+¡H*)Open penultimate−1115
Open ultimate−98
Closed penultimate−1119
Closed ultimate−1019

3.2 Alignment of starred tones

Next, we examine alignment of the end of both starred tones with respect to several acoustic markers. A decision was necessary regarding the acoustic landmark most appropriate for calculation of the starred tone. Schepman et al. (2006) argue that time alignment is most appropriately represented by quantifying the turning point in question relative to the nearest segmental landmark. Initial inspection of speaker data for both accents revealed that starred tones aligned at various points throughout the stressed vowel. Because of this, we computed alignment in two ways: relative to the beginning of the vowel, and relative to the end of the vowel.

Figures 8 and 9 are distributional analyses of the distance in ms from the start of the nuclear vowel to the starred tone for the H+L* and L+¡H* accents, respectively. Quantitative data are summarized in Table 3. First, alignment is relatively stable within the nuclear vowel when comparing open vs. closed syllables for each stress condition. This means that the end of each accent is near the syllable edge in the open condition, but considerably earlier with respect to the syllable edge in the closed condition. Second, alignment is generally earlier within the nuclear vowel in oxytone words, illustrated by the data dispersion between ultimate and penultimate conditions. [9] The LMEM returned a significant effect for the fixed factor Stress only: F(1,183.353) = 88.685, p ≤ 0.001.

Figure 8

Distance (in milliseconds) from starred L point to nuclear vowel onset for H+L* accent based on manipulation condition. Error bars indicate 95% confidence interval.

Figure 9

Distance (in milliseconds) from starred H point to nuclear vowel onset for L+¡H* accent based on manipulation condition. Error bars indicate 95% confidence interval.

Table 3

Distance (in milliseconds) from starred tone to left edge of nuclear vowel. Positive scores indicate alignment within the nuclear vowel.

TargetConditionMean (ms)SD (ms)
L (H+L*)Open penultimate4725
Open ultimate2619
Closed penultimate4530
Closed ultimate3130
H (L+¡H*)Open penultimate2916
Open ultimate1516
Closed penultimate3320
Closed ultimate1716

Next, we inspected starred tone alignment with respect to the end of the nuclear vowel. This facilitated our understanding of whether alignment was fixed at a specific landmark within the vowel or whether association was variable throughout the vowel itself. Data are presented in graphic format in Figures 10 and 11 for H+L* and L+¡H*, respectively, and quantitative results are summarized in Table 4. Regarding the effect of syllable type, starred tone alignment is within the nuclear vowel in both conditions. Alignment ranges were typically between the 40 and 100 ms interval prior to the vowel right edge. As for the effect of word stress, both accents show earlier F0 alignment in oxytone words than in paroxytone words. Earlier alignment in oxytone words is understood as a consequence of tonal clash, since it becomes necessary to realize the starred tone of the pitch accent in addition to the L% boundary tone in the cramped segmental space.

Figure 10

Distance (in milliseconds) from starred tone L point to nuclear vowel offset for H+L* accent based on manipulation condition. Error bars indicate 95% confidence interval.

Figure 11

Distance (in milliseconds) from starred tone H point to nuclear vowel offset for L+¡H* accent based on manipulation condition. Error bars indicate 95% confidence interval.

Table 4

Distance (in milliseconds) from starred tone to right edge of nuclear vowel. Negative scores indicate alignment within nuclear vowel.

TargetConditionMean (ms)SD (ms)
L (H+L*)Open penultimate−4126
Open ultimate−10233
Closed penultimate−4326
Closed ultimate−7239
H (L+¡H*)Open penultimate−5320
Open ultimate−11434
Closed penultimate−4323
Closed ultimate−7825

The LMEM showed significant effects for the fixed factors SyllableType (F(1,190.792) = 61.614, p ≤ 0.001), Stress (F(1,190.773) = 386.375, p ≤ 0.001), and the interaction SyllableType × Stress (F(1,190.776) = 39.433, p ≤ 0.001). A Bonferroni correction used to determine the effects of factors in the SyllableType × Stress interaction returned a significant result for SyllableType in ultimate syllables (t = 11.133, p ≤ 0.001), but not in penultimate syllables (t = 1.233 p = 0.255). We infer even earlier starred tone alignment in open ultimate syllables as a consequence of tonal crowding, namely speakers must realize the L% boundary tone in addition to the L valley (of H+L*) or the H peak (of L+¡H*) within the stressed vowel. In closed ultimate syllables, the coda consonant provides extra segmental space for the L boundary tone after the tonic vowel.

4 Discussion

In this study we examined the alignment of leading and starred tone turning points in nuclear H+L* and L+¡H* accents in Peninsular Spanish wh-question intonation. The effect of prosodic structure was tested for two experimental factors based on nuclear syllable configuration: syllable structure (open vs. closed) and stress position (penultimate vs. ultimate). Regarding the leading tones, these coordinated within the preaccentual syllable in almost all tokens, most likely so that the pretonic syllable is perceived as low or high relative to the accent on the tonic syllable. The H of H+L* retracted between 30 and 80 ms into the prestressed syllable; the L of L+¡H* displayed tighter synchrony with the syllable boundary, deviating no further than 30 ms into the pretonic vowel. Regarding starred tone alignment, this was not affected by nuclear syllable structure. Starred tones remained within the stressed vowel in all prosodic conditions, although they were not discernibly synchronous with either edge of this vowel. A major advance on the earlier findings of the H+L* and L+¡H* nuclear accents in Peninsular Spanish (Henriksen 2014) is that we establish that starred tones do not seek the nuclear syllable offset in their alignment; rather, they align with the nuclear vowel in CV as well as in CVC syllables. We also noted a uniform word stress effect such that, for both accents, starred tone alignment was earlier in oxytone words; these results are congruous to those of Schepman et al. (2006) and Ladd et al. (2009).

As for leading tone alignment, our findings on L+¡H* bear strong resemblance to previous findings on tune-text coordination. Recall that for Central Catalan H+L* (Prieto 2009) and Peninsular Spanish L+H* (Prieto and Torreira 2007), the leading tones displayed close synchrony with the nuclear onset consonant, whereas the starred tones displayed greater variability depending on the structure of the stressed syllable. Here, we show that the L leading tone of L+¡H* deviates little into the pretonic syllable and does not show time pressure effects with the upcoming phrase boundary, even in oxytone words. For this accent, we may infer tight coordination between F0 movements and acoustic landmarks at the syllable-initial position, confirming in this regard the syllable onset anchoring hypothesis. On the other hand, it seems that the H leading tone of H+L* behaves differently: in our data the H elbow is more variable and aligns substantially earlier within the preaccentual vowel than the L elbow of L+¡H*. In effect, this finding contradicts the predictions of the syllable onset anchoring hypothesis since the leading tone retracts well into the preaccentual syllable. As an interim summary, we find divergent outcomes on the syllable onset anchoring hypothesis: for the L+¡H* accent it is confirmed, but for the H+L* accent it is rejected. [10] We should also clarify that for both accents, there is support for the second portion of the syllable onset anchoring hypothesis (i.e., that alignment is more variable with respect to the syllable-final boundary) since starred tone alignment was not synchronous with the syllable offset in either case.

One seemingly recurrent result in the research on syllable structure effects is that starred tone alignment tends to be sensitive to syllable structure (e.g., D’Imperio 2000; Gili-Fivela and Savino 2003; Prieto and Torreira 2007; Prieto 2009). In our experiment, we tested starred tone coordination effects for two bitonal accents and showed that alignment was consistently stable within the nuclear vowel regardless of syllable structure. In fact, a follow-up check of alignment patterns reveals that only 3.2% (14/434) of F0 elbows produced in the closed syllable condition actually aligned past the start of the coda consonant (i.e., K0), an obvious minority (see histograms in Figures 12 and 13). Even though there was an interaction between syllable type and word stress in oxytone words, these tonal crowding effects are a clear prediction of the segmental anchoring hypothesis, proposed in Arvaniti et al. (1998) and Ladd et al. (1999). Taken together, our findings on H+L* and L+¡H* reinforce the notion that the location of tonal targets is governed by stability patterns with the segmental string. This is because in our data starred tone alignment is insensitive to syllable structure such that coda consonants do not provoke later alignment.

Figure 12

Histogram of L distance (in milliseconds) relative to start of coda consonant (K0) for all H+L* data. The vertical line at ‘0’ indicates start of coda consonant.

Figure 13

Histogram of H distance (in milliseconds) relative to start of coda consonant (K0) for all L+¡H* data. The vertical line at ‘0’ indicates start of coda consonant.

Notwithstanding this finding, we acknowledge that neither starred tone was fixed with a segmental edge; rather, our results indicate somewhat loose association within the stressed vowel (although we acknowledge the rare cases in which it deviated past the vowel itself). Welby and Lœvenbruck (2005, 2006) advance the notion of ‘anchorage’ rather than strict alignment to account for different anchor point locations depending on syllable structure in French. According to their proposal, there is a temporal range within which an intonational turning point can coordinate (as opposed to a strict segmental anchor), typically across a prosodic domain such as a syllable edge. For French this anchorage stretches from just before the end of the vowel of the last full syllable of the accentual phrase to the end of the phrase itself. Nonetheless, the authors are cautious that the concept of segmental anchorage may not account for alignment patterns in all languages. Given the findings of this paper, we feel that the principle of segmental anchorage does not properly characterize the overall stability of starred tone alignment in the L+¡H* and H+L* accents. This is because researchers typically invoke the notion of segmental anchorage when alignment is variable across a segmental boundary (e.g., Welby and Lœvenbruck 2005, Welby and Lœvenbruck 2006; Prieto and Torreira 2007). In our study, we found that F0 turning points were almost always synchronous with the stressed vowel itself. Our results therefore resonate more strongly with the original segmental anchoring hypothesis put forth in Arvaniti et al. (1998). [11] One possible adaptation to the segmental anchoring hypothesis, given our findings, could be that anchoring landmarks for tonal targets are either fixed points (e.g., edges of segments, or of constituents such as syllables or intonational phrases) or entire segments themselves (e.g., onset, nucleus, coda).

A combined analysis of syllable structure effects in Catalan and Spanish prompts the observations in Table 5. As an initial caveat, the information provided is not intended as exhaustive; we also acknowledge that straightforward comparisons are not always possible when taking into account a broad sample of findings for which completely uniform methodologies were not implemented (especially since phrase position varies across these studies). Nonetheless, these studies share one methodological commonality, which is that the structure of the prestressed syllable was controlled (it was open), and the structure of the stressed syllable was manipulated (it was open or closed). [12] Whereas Prieto and Torreira (2007) and Prieto (2009) obtained similar results for both hypotheses, a key discovery of our research is that we show a within-language discrepancy with respect to the syllable onset anchoring hypothesis: for the L+¡H* accent there was evidence in favor, but for the H+L* accent, there was evidence against. In this respect, our results underscore the notion that a single position on temporal coordination may not suffice to explain alignment patterns for all accents in one language variety.

Table 5

Summary of findings on the strict segmental anchoring hypothesis and the syllable onset anchoring hypothesis for Catalan and Spanish.

AccentLanguage varietySegmental anchoring hypothesisSyllable onset anchoring hypothesis
L+H*Peninsular Spanish (Prieto and Torreira 2007)
L+¡H*Peninsular Spanish (present study)
H+L*Central Catalan (Prieto 2009)
H+L*Peninsular Spanish (present study)

The body of research on syllable structure effects in Romance languages shows that closed syllables provoke later F0 alignment (e.g., D’Imperio 2000; Welby and Lœvenbruck 2005, Welby and Lœvenbruck 2006; Prieto and Torreira 2007; Prieto 2009). However, for the present data on Peninsular Spanish, we find no effect of syllable structure on starred tone alignment in nuclear position; this contrasts with findings from Prieto and Torreira’s (2007) earlier study on L+H* accents in prenuclear position. One possible explanation for this difference is that different dialect samples were used, similar to what Atterer and Ladd (2004) reported for Northern and Southern varieties of German. While the speakers in our study were from south-central Castile, the speakers in Prieto and Torreira were all from northern Castile or central Castile.

A second possibility is that differences arise due to the focus on prenuclear vs. nuclear accents as they relate to tonal crowding. This is because one core idea behind the segmental anchoring hypothesis is that when right-hand prosodic effects are excluded (i.e., when the tonal features under investigation are not in the vicinity of pitch accents or boundary tones), the alignment of F0 peak targets is consistently governed by segmental anchoring, regardless of changes in syllabic/segmental structure or speech rate. Schepman et al. (2006) and Ladd et al. (2009) show that for rising accents in Dutch and British English, respectively, nuclear peaks align earlier than prenuclear peaks. In both studies there was also an interaction between vowel length and phrase position such that prenuclear peaks aligned during the stressed vowel if the vowel was long and during the following consonant if the vowel was short. Crucially, this effect of vowel length on alignment did not apply in nuclear position. As indicated in Ladd et al., the failure to find an effect of vowel length on peak alignment in nuclear accents, in addition to the overall earlier alignment of nuclear peaks, suggests that greater priority is given to executing phrase-final F0 boundary movements than to respecting vowel length differences in the nuclear syllable. When taking into account the findings of Prieto and Torreira (2007) for prenuclear accents and those of the present study for nuclear accents, an analogous situation emerges for Peninsular Spanish: in prenuclear accents the F0 elbow of the starred tone aligns later in CVC syllables than in CV syllables (i.e., in phonetically longer syllables), whereas in nuclear accents there is no such difference. Interestingly, in the present study the outcome was identical for falling and rising accents alike, even though the amount of ‘pressure’ is distinct for each accent when followed by L%: the latter needs to compress three target pitch points within the utterance-final word (LHL), yet in the former there are two adjacent L targets at the right-periphery (HLL). This finding suggests that the ‘right context’ effect on the alignment of nuclear accents is not necessarily due to contrasting adjacent F0 movements. Instead, it seems to ensure that the boundary itself is executed, regardless of whether its phonetic implementation is a fall from a nuclear peak (as in LHL) or a low plateau from a nuclear fall (as in HLL).

Finally, although we have argued here that the observed alignment differences between prenuclear and nuclear accents in Spanish derive primarily from time pressure effects with the following boundary tone in nuclear accents, we acknowledge that alternative explanations are possible. For example, Face (2002) proposes that prenuclear and nuclear accents differ in their tonal configurations, assuming an additional layer of tonal structure between tonal units and pitch accents that helps to regulate F0 alignment. Although we did not design our experiment to test these two explanations explicitly, we feel that our results support an explanation based on time pressure effects for one important reason. Specifically, the failure to find an effect of syllable structure on starred tone alignment in nuclear position, in addition to generally earlier F0 alignment in nuclear position in Spanish (Henriksen 2012), suggests that any underlying effect of syllable structure can be overruled by the priority to realize phonetically the L% boundary. At the phonetic level, the result is that speakers must align the F0 elbow of the starred tone at an earlier point in nuclear accents than in prenuclear accents (see also Ladd et al. 2009: 158). At this juncture it is not clear how different tonal configurations in prenuclear and nuclear accents could motivate the observed results on syllable structure in Spanish, but obviously this point should be addressed more directly in future research.

5 Conclusion

In this paper we tested the coordination relations between F0 turning points and acoustic landmarks for two nuclear pitch accents common to Peninsular Spanish wh-question intonation: H+L* and L+¡H*. As for leading tone alignment, the L tone of the L+¡H* accent displayed close synchrony with the start of the nuclear syllable, whereas the H tone of the H+L* accent was more variable within the pretonic syllable. These findings confirmed the syllable onset anchoring hypothesis for the L+¡H* accent only. As for starred tone alignment, the effects were consistent for both accents: the open/closed condition did not affect starred tone alignment, and the stressed/unstressed condition provoked earlier alignment in oxytone words. Taken together with the results of Prieto and Torreira (2007), these findings yield a more complete picture of temporal coordination patterns in Spanish such that coda consonants provoke later alignment in non-final position only. All in all, one contribution of this paper is that we now establish that starred tone alignment in Peninsular Spanish nuclear accents is not synchronous with the stressed syllable offset; rather, it is within the nuclear vowel in CV and CVC syllables, due possibly to the tonal pressure from the upcoming boundary tone. A second contribution was our proposal that the segmental anchoring hypothesis may be adapted such that anchoring landmarks for tonal targets are either fixed points (e.g., edges of segments, or of constituents such as syllables or intonational phrases) or entire segments themselves (e.g., onset, nucleus, coda). A third contribution is that by testing the applicability of tune-text coordination positions for multiple nuclear accents in one language variety, we have shown that tune-text positions are accent-specific. Given this last point, we leave as a testable area of inquiry for future research the extent to which our claims hold for all prenuclear and nuclear accents in Spanish.

Acknowledgments

I am especially grateful to the Associate Editor Štefan Beňuš and two anonymous reviewers for their constructive feedback throughout the revision process. I further thank Andries Coetzee, Lorenzo García-Amaya, Lindsey Graham, Sarah Harper, Jelena Krivokapić, Bob Ladd, Rafèu Sichel-Bazin, Francisco Torreira, and Maria del Mar Vanrell for their comments on earlier versions of this paper, and Brady West (University of Michigan Center for Statistical Consultation and Research) for his help and advice on the statistical analysis. A portion of this paper was presented at the 10th International Seminar on Speech Production in Cologne, Germany, in May 2014. Of course, all shortcomings are my own.

Appendix A. Database of test sentences

Target nuclear syllables are in bold.

Open, penultimate stress

¿Quien está saliendo con lanena? Who is going out with the girl?

¿Quién está viviendo con Ramona? Who is living with Ramona?

¿Quién está bailando con laniña? Who is dancing with the girl?

¿Quién se ha operado de lamano? Who got their hand operated on?

¿Cuándo van a salir paraMula? When are they going to leave for Mula?

Open, ultimate stress

¿Cuándo se van para Alca? When do they leave for Alcalá?

¿Cuándo salimos para Ca? When do we leave for Calí?

¿Quién está jugando con Re? Who is playing with René?

¿Quién está hablando con Ma? Who is speaking with Mom?

¿Quien está saliendo con Da? Who is going out with Dalí?

Closed, penultimate stress

¿Quien está saliendo con Armando? Who is going out with Armando?

¿Quién está viviendo con Raimundo? Who is living with Raimundo?

¿Quién está bailando con Domingo? Who is dancing with Domingo?

¿Quién se ha operado de lanalga? Who got their buttock operated on?

¿Cuándo van a salir paraRonda? When are they going to leave for Ronda?

Closed, ultimate stress

¿Quien está saliendo con Ramón? Who is going out with Ramón?

¿Quién está hablando con Belén? Who is talking to Belén?

¿Quién está jugando con Fermín? Who is playing with Fermín?

¿Cuándo salimos para Hellín? When do we leave for Hellín?

¿Cuándo salisteis para Bailén? When did you leave for Bailén?

Appendix B. Statistical results of all LMEMs

Significant results are in bold.

Leading tone distance to C0 onsetStarred tone distance to V0 onsetStarred tone distance to V0 offset
AccentF(1,9.044) = 23.845, p = 0.001F(1,9.044) = 3.841, p = 0.082F(1,9.011) = 0.470, p = 0.510
SyllableTypeF(1,188.188) = 0.000, p = 0.986F(1,183.372) = 0.873, p = 0.351F(1,190.792) = 61.614, p 0.001
StressF(1,188.175) = 10.430, p = 0.001F(1,183.353) = 88.685, p 0.001F(1,190.773) = 386.375, p 0.001
Accent × SyllableTypeF(1,188.188) = 0.044, p = 0.834F(1,183.372) = 0.143, p = 0.706F(1,190.792) = 3.840, p = 0.052
Accent × StressF(1,188.175) = 19.302, p 0.001F(1,183.353) = 0.409, p = 0.523F(1,190.773) = 0.654, p = 0.420
SyllableType × StressF(1,188.183) = 1.873, p = 0.173F(1,183.358) = 0.644, p = 0.423F(1,190.776) = 39.433, p 0.001
Accent × SyllableType × stressF(1,188.183) = 1.926, p = 0.167F(1,183.358) = 1.926, p = 0.231F(1,190.776) = 0.735, p = 0.392

Appendix C. Mean rate of speech data

Figure 14

Mean rate of speech data (syllables/second) for each speaker in the H+L* speaker group.

Figure 15

Mean rate of speech data (syllables/second) for each speaker in the L+¡H* speaker group.

References

Aguilar, Lourdes, Carmende-la-Mota & PilarPrieto (eds.). 2009. Sp_ToBI training materials. http://prosodia.upf.edu/sp_tobi/ (accessed 18 December 2013).Search in Google Scholar

Armstrong, Meghan. 2010. Puerto Rican Spanish intonation. In PilarPrieto & PaoloRoseano (eds.), Transcription of intonation of the Spanish language, 155190. Munich: Lincom Europa.Search in Google Scholar

Arvaniti, Amalia, D. RobertLadd & InekeMennen. 1998. Stability of tonal alignment: The case of Greek prenuclear accents . Journal of Phonetics26. 325.10.1006/jpho.1997.0063Search in Google Scholar

Arvaniti, Amalia, D. RobertLadd & InekeMennen. 2000. What is a starred tone? In JanetPierrehumbert & MichaelBroe (eds.), Papers in laboratory phonology V, 119131. Cambridge: Cambridge University Press.Search in Google Scholar

Arvaniti, Amalia, D. RobertLadd & InekeMennen. 2006. Tonal association and tonal alignment: Evidence from Greek polar questions and contrastive statements . Language and Speech49(4). 421450.10.1177/00238309060490040101Search in Google Scholar

Atterer, Michaela & D.Robert Ladd. 2004. The phonetics and phonology of ‘segmental anchoring’ of F0: Evidence from German . Journal of Phonetics32. 177197.10.1016/S0095-4470(03)00039-1Search in Google Scholar

Beckman, Mary, ManuelDíaz-Campos, JuliaMcGory & Terrell A. Morgan. 2002. Intonation across Spanish in the tones and break indices framework . Probus14. 936.10.1515/prbs.2002.008Search in Google Scholar

Beckman, Mary E. & Janet B.Pierrehumbert. 1986. Intonational structure in Japanese and English . Phonology Yearbook3. 255309.10.1017/S095267570000066XSearch in Google Scholar

Boersma, Paul & DavidWeenink. 2013. Praat: Doing phonetics by computer (version 5.3.42). http://www.praat.org/ (accessed 18 March 2013).Search in Google Scholar

Bruce, Gösta. 1977. Swedish word accents in sentence perspective. Lund: Gleerup.Search in Google Scholar

del Giudice, Alex, Ryan K.Shosted, KatherineDavidson, MohammadSalihie & AmaliaArvaniti. 2007. Comparing methods for comparing “pitch elbows.” In Proceedings of ICPhS XVI, 1117–1120. Saarbrücken, Germany.Search in Google Scholar

Dilley, Laura, D. RobertLadd & AstridSchepman. 2005. Alignment of L and H in bitonal pitch accents: Testing two hypotheses . Journal of Phonetics33. 115119.10.1016/j.wocn.2004.02.003Search in Google Scholar

D’Imperio, Mariapaola. 2000. The role of perception in defining tonal targets and their alignment. Columbus, OH: The Ohio State University dissertation.Search in Google Scholar

D’Imperio, Mariapaola & DavidHouse. 1997. Perception of question and statements in Neapolitan Italian. In GeorgeKokkinakis, NikosFakotakis, & EvangelosDermatas (eds.), Proceedings of Eurospeech 1997, 251254. Greece: Rhodes.10.21437/Eurospeech.1997-90Search in Google Scholar

D’Imperio, Mariapaola, CaterinaPetrone & NoëlNguyen. 2007. Effects of tonal alignment on lexical identification in Italian. In CarlosGussenhoven & T. Riad (eds.), Tones and tunes, Vol. 2. 79106. Berlin: Mouton de Gruyter.10.1515/9783110207576.1.79Search in Google Scholar

Dorta Luis, Josefa. 2000. Entonación hispana: Interrogativas no pronominales vs. Pronominales . LingüÍstica española actual22(1). 5175.Search in Google Scholar

Estebas-Vilaplana, Eva & PilarPrieto. 2008. La notación prosódica del español: una revisión del sp_ToBI . Estudios De Fonética Experimental17. 263283.Search in Google Scholar

Estebas-Vilaplana, Eva & PilarPrieto. 2010. Castilian Spanish intonation. In PilarPrieto & PaoloRoseano (eds.), Transcription of intonation of the Spanish language, 1748. Munich: Lincom Europa.Search in Google Scholar

Face, Timothy L. 2002. Intonational marking of contrastive focus in Madrid Spanish. Munich: Lincom Europa.10.1515/prbs.2002.006Search in Google Scholar

Face, Timothy L. 2008. The intonation of Castilian Spanish declaratives and absolute interrogatives. Munich: Lincom Europa.Search in Google Scholar

Face, Timothy L. 2011. Perception of Castilian Spanish intonation: Implications for intonational phonology. Munich: Lincom Europa.Search in Google Scholar

Face, Timothy L. & PilarPrieto. 2006/2007. Rising accents in Castilian Spanish: A revision of sp_ToBI . Journal of Portuguese Linguistics6(/7). 117146.10.5334/jpl.147Search in Google Scholar

Fougeron, Cécile & PatriciaKeating. 1997. Articulatory strengthening at edges of prosodic domains . Journal of the Acoustical Society of America101. 37283740.10.1121/1.418332Search in Google Scholar

Gili-Fivela, Barbara & MichelinaSavino. 2003. Segments, syllables and tonal alignment: A study on two varieties of Italian. In Maria-JosepSolé, DanielRecasens, & JoqauínRomero (eds.), Proceedings of the XVth international congress of phonetic sciences, 29332936. Barcelona: Casual Productions.Search in Google Scholar

Henriksen, Nicholas. 2009. Wh-intonation in peninsular Spanish: Multiple contours and the effect of task type . Journal of Portuguese Linguistics8(1). 4774.10.5334/jpl.121Search in Google Scholar

Henriksen, Nicholas. 2012. The intonation and signaling of declarative questions in Manchego Peninsular Spanish . Language and Speech55(4). 543576.10.1177/0023830911434126Search in Google Scholar

Henriksen, Nicholas. 2013. Style, prosodic variation, and the social meaning of intonation . Journal of the International Phonetic Association43(2). 153193.Search in Google Scholar

Henriksen, Nicholas. 2014. Initial peaks and final falls in the intonation of Manchego Peninsular Spanish wh-questions . Probus26(1). 83134.10.1515/probus-2013-0003Search in Google Scholar

Henriksen, Nicholas & LorenzoGarcía-Amaya. 2012. Transcription of intonation of Jerezano Andalusian Spanish . Estudios De Fonética Experimental21. 109162.Search in Google Scholar

Hualde, José I. 2002. Intonation in Spanish and the other IBERO-romance languages: Overview and status Quaestionis. In CarolineR. Wiltshire & JoaquimCamps (eds.), Romance phonology and variation, 101115. Amsterdam: John Benjamins.10.1075/cilt.217.10huaSearch in Google Scholar

Hualde, José I. 2005. The sounds of Spanish. Cambridge: Cambridge University Press.Search in Google Scholar

Ishihara, Takeshi. 2006. Tonal alignment in Tokyo Japanese. Edinburgh, Scotland: University of Edinburgh dissertation.Search in Google Scholar

Ladd, D. Robert. 2008. Intonational phonology. Cambridge: Cambridge University Press.10.1017/CBO9780511808814Search in Google Scholar

Ladd, D. Robert, DanielFaulkner, HannekeFaulkner & AstridSchepman. 1999. Constant “segmental” anchoring of F0 movements under changes in speech rate . Journal of the Acoustical Society of America106. 15431554.10.1121/1.427151Search in Google Scholar

Ladd, D. Robert, InekeMennen & AstridSchepman. 2000. Phonological conditioning of peak alignment in rising pitch accents in Dutch . The Journal of the Acoustical Society of America107. 26852696.10.1121/1.428654Search in Google Scholar

Ladd, D. Robert & AstridSchepman. 2003. “Sagging transitions” between high pitch accents in English: Experimental evidence . Journal of Phonetics31. 81112.10.1016/S0095-4470(02)00073-6Search in Google Scholar

Ladd, D. Robert, AstridSchepman, LaurenceWhite, LouiseQuarmby & RebekahStackhouse. 2009. Structural and dialectal effects on pitch peak alignment in two varieties of British English . Journal of Phonetics37. 145161.10.1016/j.wocn.2008.11.001Search in Google Scholar

López-Bobo, María Jesus & MiguelCuevas-Alonso. 2010. Cantabrian Spanish intonation. In PilarPrieto & PaoloRoseano (eds.), Transcription of intonation of the Spanish language, 4985. Munich: Lincom Europa.Search in Google Scholar

Maddieson, Ian. 1985. Phonetic cues to syllabification. In VictoriaFromkin (ed.), Phonetic linguistics, 203221. San Diego, CA: Academic Press.Search in Google Scholar

Mücke, Doris, MartineGrice, J.Becker & AnneHermes. 2009. Sources of variation in tonal alignment: Evidence from acoustic and kinematic data . Journal of Phonetics37(3). 321338.10.1016/j.wocn.2009.03.005Search in Google Scholar

Mücke, Doris, HosungNam, AnneHermes & LouisGoldstein. 2012. Coupling of tone and constriction gestures in pitch accents. In PhilipHoole, LasseBombien, MariannePouplier, ChristineMooshammer, & BarbaraKühnert (eds.), Consonant clusters and structural complexity, 205230. Berlin: Mouton de Gruyter.10.1515/9781614510772.205Search in Google Scholar

Navarro, Tomás. 1944. Manual de entonación española. New York: Hispanic Institute in the United States.Search in Google Scholar

Niemann, Henrik, DorisMücke, HosungNam, LouisGoldstein & MartineGrice. 2011. Tones as gestures: The case of Italian and German. Proceedings of International Conference on Phonetic Sciences XVII, 1486–1489. Hong Kong.Search in Google Scholar

Ortega-Llebaria, Marta & PilarPrieto. 2007. Do complex pitch gestures induce syllable lengthening in Catalan and Spanish? In MarinaVigário, SóniaFrota, & M. João Freitas (eds.), Phonetics and phonology: Interactions and interrelations, 5170. Amsterdam/Philadelphia, PA: John Benjamins.Search in Google Scholar

Peterson, Gordon & IlseLehiste. 1960. Duration of syllable nuclei in English . Journal of the Acoustical Society of America32(6). 693703.10.1121/1.1908183Search in Google Scholar

Pierrehumbert, Janet. 1980. The phonetics and phonology of English intonation. Cambridge, MA: MIT dissertation.Search in Google Scholar

Pierrehumbert, Janet & ShirleySteele. 1989. Categories of tonal alignment in English . Phonetica46. 181196.10.1159/000261842Search in Google Scholar

Prieto, Pilar. 2004. Phonological targets in the tonal space. In Timothy L. Face (ed.), Laboratory approaches to Spanish phonology, 2959. Berlin: Mouton de Gruyter.Search in Google Scholar

Prieto, Pilar. 2005. Stability effects in tonal clash contexts in Catalan . Journal of Phonetics33. 215242.10.1016/j.wocn.2004.07.001Search in Google Scholar

Prieto, Pilar. 2009. Tonal alignment patterns in Catalan nuclear falls . Lingua119. 865880.10.1016/j.lingua.2007.11.014Search in Google Scholar

Prieto, Pilar, MariapaolaD’Imperio & BarbaraGili-Fivela. 2005. Pitch accent alignment in romance: Primary and secondary associations with metrical structure . Language and Speech48. 359396.10.1177/00238309050480040301Search in Google Scholar

Prieto, Pilar, EvaEstebas-Vilaplana & Mariadel Mar Vanrell. 2010. The relevance of prosodic structure in tonal articulation. Edge effects at the prosodic word level in Catalan and Spanish . Journal of Phonetics38(4). 688707.10.1016/j.wocn.2010.10.004Search in Google Scholar

Prieto, Pilar, & PaoloRoseano (eds.). 2010. Transcription of intonation of the Spanish language. Munich: Lincom Europa.Search in Google Scholar

Prieto, Pilar & FranciscoTorreira. 2007. The segmental anchoring hypothesis revisited: Syllable structure and speech rate effects on peak timing in Spanish . Journal of Phonetics35. 473500.10.1016/j.wocn.2007.01.001Search in Google Scholar

Prieto, Pilar, Janvan Santen & JuliaHirschberg. 1995. Tonal alignment patterns in Spanish . Journal of Phonetics23. 429451.10.1006/jpho.1995.0032Search in Google Scholar

Reubold, Ulrich, JonathanHarrington & FelicitasKleber. 2010. Vocal aging effects on F0 and the first formant: A longitudinal analysis in adult speakers . Speech Communication52. 638651.10.1016/j.specom.2010.02.012Search in Google Scholar

Savino, Michelina. 2012. The intonation of polar questions in Italian: Where is the rise? Journal of the International Phonetic Association42(1). 2348.10.1017/S002510031100048XSearch in Google Scholar

Schepman, Astrid, RobinLickley & D.Robert Ladd. 2006. Effects of vowel length and “right context” on the alignment of Dutch nuclear accents . Journal of Phonetics34. 128.10.1016/j.wocn.2005.01.004Search in Google Scholar

Sosa, Juan Manuel. 1999. La entonación del español: Su estructura fónica, variabilidad y dialectología. Madrid: Cátedra.Search in Google Scholar

Sosa, Juan Manuel. 2003. Wh-questions in Spanish: Meanings and configurational variability . Catalan Journal of Linguistics2. 229247.10.5565/rev/catjl.51Search in Google Scholar

Welby, Pauline & HélèneLœvenbruck. 2005. Segmental anchorage and the French late rise. Proceedings of Interspeech 2005. Lisbon, Portugal. 4–8 September.10.21437/Interspeech.2005-38Search in Google Scholar

Welby, Pauline & HélèneLœvenbruck. 2006. Anchored down in anchorage: Syllable structure and segmental anchoring in French . Italian Journal of Linguistics18(1). 73124.Search in Google Scholar

West, Brady, KathleenWelch & AndrzejGalecki. 2014. Linear mixed models: A practical guide using statistical software, 2nd ed.Boca Raton, FL: Chapman Hall/CRC Press.Search in Google Scholar

Willis, Erik. 2010. Dominican Spanish intonation. In PilarPrieto & PaoloRoseano (eds.), Transcription of intonation of the Spanish language, 123143. Munich: Lincom Europa.Search in Google Scholar

Xu, Yi. 1998. Consistency of tone-syllable alignment across different syllable structures and speaking rates . Phonetica55. 179203.10.1159/000028432Search in Google Scholar

Published Online: 2015-3-17
Published in Print: 2015-2-1

©2015 by De Gruyter Mouton

Downloaded on 4.12.2022 from frontend.live.degruyter.dgbricks.com/document/doi/10.1515/lp-2015-0001/html
Scroll Up Arrow