Abstract
Aims and objectives:
Polysyllabic shortening is thought to contribute to the perception of stress-timed rhythm in some languages. Little is known about its use in the speech of children exposed to a language that incorporates it more frequently (e.g. English) and one that incorporates it less frequently (e.g. Spanish). The purpose of the current investigation was to explore polysyllabic shortening in bilingual children’s two languages compared to monolingual Spanish and English comparison groups.
Method/Design:
We performed a group-level, cross-sectional study comparing the magnitude of polysyllabic shortening for monolingual English- and Spanish-speaking children and Spanish-English bilingual children.
Data/Analysis:
Sixteen monolingual English speakers, 23 monolingual Spanish speakers, and 16 Spanish-English bilingual speakers produced two- and four-syllable words in English only, Spanish only, or both English and Spanish, respectively. Ages ranged from 4;5 to 7;7 (M = 5;10, SD = 7 months). English and Spanish words had the same syllable shapes and primary stress locations. Articulation rate was measured by syllables per second. A language history questionnaire and standardized vocabulary test were also administered. Comparisons were made both between and within groups.
Results:
Both monolingual English and Spanish speakers utilized polysyllabic shortening to similar degrees. Bilingual children produced polysyllabic shortening in English and Spanish to the same degree as their monolingual peers, but they produced it to a greater degree in their own Spanish than in their own English.
Conclusion:
Polysyllabic shortening might be a universal feature of speech that results from universal phonetic constraints. For the bilingual children, greater use of polysyllabic shortening in Spanish than English may be related to better Spanish than English articulatory control.
Polysyllabic shortening refers to a reduction in the duration of individual syllables as a word increases in the number of total syllables. It has been argued that polysyllabic shortening helps maintain the perception of rhythm in some languages (Lehiste, 1977). Although speech rhythm is critical for intelligibility (Ordin & Polyanskaya, 2014, 2015), little is known about its development in bilingual children. Instead, most research has focused on bilingual phonological development at the phonemic level. The current study attempted to contribute to our understanding of bilingual phonological development by exploring speech rhythm through the phenomenon of polysyllabic shortening.
Polysyllabic shortening and speech rhythm
Lehiste (1972) found that for adults, the speed per syllable in the word stickily was faster than in the word sticky, which was faster than in the word stick. The locus of this polysyllabic shortening was primarily on the stressed syllable (in this example, the stem stick), which became shortened, while the duration of unstressed syllables was affected to a lesser degree (Kim & Cole, 2005). Similar results were found by Port (1981), who measured speakers producing nonsense words (e.g. deeb, deeber, deeberly). Like Lehiste (1972), results showed that the stressed syllable became shorter as syllables were added to the stem. These results were interpreted to mean that polysyllabic shortening contributed to the perception of stress-timed rhythm. That is, speakers were thought to speed up and slow down their productions in order to maintain regular timing between stressed syllables.
Languages traditionally have been categorized into two rhythmic classes (Pike, 1945). Syllable-timed languages, like Spanish and Italian, are perceived to maintain a constant interval between syllables, and stress-timed languages, like English and Dutch, are perceived to maintain a constant interval between stresses. The strong view of this position has asserted that all languages fit into one of these categories (Abercrombie, 1967; see Kureta, Fushimi, & Tatsumi, 2006, regarding moraic-timed languages). These approaches necessarily assumed that speakers encoded and attempted to apply phonological representations for speech rhythm.
However, as researchers have failed to identify the phonetic correlates of these rhythmic classes (Bolinger, 1965; Borzone de Manrique, & Signorini, 1983), a weaker view emerged. For example, Roach (1982) compared the variation in syllable durations for three syllable-timed and three stress-timed languages. Because syllable-timed languages purportedly produce each syllable with relatively similar durations, one would predict less variation in their syllable durations than in the syllable durations of stressed-timed languages. This was not the case. Results showed greater variation in the syllable-timed than stress-timed languages. Dauer (1983) compared adult speakers of syllable-timed languages (Greek, Italian, and Spanish) with adult speakers of a stress-timed language (English). Contrary to predictions made by the strong view of rhythmic classes, measurements of the intervals between stresses were not more regular in English than in the other languages. Instead, the perception of stress-timed and syllable-timed languages appeared to be based on differences in the phonological structures of their syllables. Stress-timed languages had a wide variety of syllable shapes that influenced where lexical stress might fall (e.g. heavy syllables attract stress in English), while there was limited or no relationship between syllable shape and stress location in syllable-timed languages (which have a limited variety of syllable shapes; Dauer, 1983). In addition, stress-timed languages used vowel reduction, while syllable-timed languages typically did not (Dauer, 1983).
These results led Dauer (1983, 1987) and others (Roach, 1982) to propose that, rather than being categorized into one of the traditional dichotomous rhythmic classes, languages fell on a continuum between syllable-timed on one end and stress-timed on the other. Roach (1982) wrote that “there is no language which is totally syllable-timed or totally stress-timed—all languages display both sorts of timing; languages will, however, differ in which type of timing predominates” (p. 78). Where a language might fall on this continuum was determined by the phonological structure of syllables instead of a separate representation for speech rhythm. On this view, English fell toward the stressed-timed end of the continuum, and Spanish fell toward the syllable-timed end.
Development of language rhythm
Perception
A considerable body of literature has shown that even young infants can discriminate languages, and this likely is due to cues provided by speech rhythms. For example, four-day-old French infants in a study by Mehler et al. (1988) were presented with recordings of both French and Russian speakers. The infants’ sucking rates increased when French but not Russian was heard. The heightened sucking rates indicated that the infants recognized the familiar French but not the unfamiliar Russian. Four-day old infants who were neither from French nor Russian contexts responded similarly to French and Russian recordings, making no distinction between them. The authors interpreted this to mean that infants had become familiar with French while in the womb. Since they could not understand the words of French, the infants must have recognized that French speech rhythm was distinct from that of Russian.
The ability for infants to discriminate between languages has been identified in several language pairs, including but not limited to Spanish and English (Bahrick & Pickens, 1988, Moon, Cooper, & Fifer, 1993), Italian and English (Mehler et al., 1988), Japanese and Italian, and Japanese and English (Nazzi, Bertoncini, & Mehler, 1998). Although newborns can only discriminate languages from different rhythmic classes (e.g. French compared to Russian; Mehler et al., 1988), by age five months they can discriminate languages from within the same rhythmic class and even between dialects of the same language (Nazzi, Jusczyk, & Johnson, 2000). Although there are fewer studies investigating bilingual contexts, infants exposed to two languages appear to share the same developmental trajectory of language discrimination as infants exposed to a single language (Bosch & Sebastian-Galles, 2001; Byers-Heinlein, Burns, & Irker, 2010).
Production
Payne, Post, Astruc, Prieto, and Vanrell (2012) found that language-specific speech rhythms emerged as early as two years of age. This ability becomes increasingly fine-tuned with age. Grabe, Post, and Watson (1999) analyzed the audio recordings of three four-year-old monolingual English-speaking children and three monolingual French-speaking peers as they interacted with their mothers. The rhythmic speech patterns of the French children were similar to those of their mothers, which was interpreted to mean that the French rhythm (syllable-timed) was fully acquired by at least age four years. In addition, although the English-speaking children’s rhythm was different from that of their French peers, it was also different from that of their mothers. This suggested that the English rhythm (stress-timed) had not been fully acquired, which the authors attributed to the greater phonological complexity of English compared to French.
Similar claims were made by Allen and Hawkins (1980), who proposed that in the early stages of development, children from both syllable-timed and stress-timed languages use syllable-timed rhythms. This proposal was based on the consideration that the use of vowel reduction and the production of consonant clusters, both of which are key contributors to the perception of stress-timing (as noted by Dauer, 1983, 1987), require later-developing articulatory abilities. This view has found recent empirical support (Sirsa & Redford, 2011) and has been extended to posit that improved motor control is responsible for successfully producing the phonological components that contribute to stress-timing (Ordin & Polyanskaya, 2015).
Separation of speech rhythm in bilingual children’s languages
Few studies have investigated the acquisition of speech rhythm for bilingual children exposed to both stress-timed and syllable-timed languages. However, Lleó, Rakow, and Kehoe (2007) compared the speech rhythms of three groups of speakers, all of whom were three years old. Monolingual German-speaking and monolingual Spanish-speaking children used distinct speech rhythms. However, their German-Spanish simultaneous bilingual peers used similar rhythms for both of their languages, indicating that the speech rhythms for their two languages had not separated by age three. Mok (2011) similarly found this pattern when comparing Cantonese (syllable-timed) to English. The author compared the speech rhythms of six monolingual Cantonese-speaking, six monolingual English-speaking, and six bilingual Cantonese-English-speaking children at age three. The bilingual children’s English rhythm was significantly more syllable-timed than that of their monolingual English-speaking peers, again suggesting that the distinct speech rhythms of the two languages had not separated by age three.
Bunta and Ingram (2007) compared the speech rhythm of Spanish-English simultaneous bilingual speakers to that of monolingual peers in both English and Spanish. The participants included younger children (age 3;9–4;5) older children (age 4;6–5;2) and adults. Researchers elicited sentences in both English and Spanish by asking participants to describe pictures. Results showed that all three groups used different speech rhythms when speaking English compared to Spanish. However, older children’s speech rhythms were more starkly separated than the speech rhythms of the younger children, and neither the younger nor older groups produced speech rhythms undifferentiated from that of adults. The authors interpreted this to mean that, although significantly progressing by age four, the development of speech rhythm in bilingual speakers was still not complete by five years of age.
Possible polysyllabic shortening in Spanish
Our review of the literature has not found any studies that have directly tested polysyllabic shortening in monolingual Spanish speakers. However, the results of several studies suggest that its magnitude likely is weaker in Spanish than in English. For example, greater variation in syllable length would be consistent with polysyllabic shortening (i.e. if there were no variation between syllable lengths, then there could be no polysyllabic shortening). Studies examining the rhythm of Spanish and English have found less variation in the length of syllables in Spanish than in the length of syllables in English. For example, Hoequist (1983) found that the ratio of duration in stressed compared with unstressed syllables in Spanish was 1.24 to 1. Delattre (1966) found a similar result for Spanish, but found that the ratio in English was greater (1.6 to 1). This likely is related to the more recent findings that vowel length is similar for stressed compared to unstressed syllables in Spanish, but stressed syllables are considerably longer than unstressed syllables in English (Hammond, 2001; Ortega-Llebaria & Prieto, 2007). It must be underscored that these studies did not find that variation in syllable length was absent in Spanish. Instead, the data demonstrated that there was less variation in Spanish than in English.
Grabe and colleagues (Grabe, 2002; Grabe et al., 1999) developed objective phonetic measures to distinguish languages from different rhythmic classes. They measured the durations of vowels (vocalic duration) and the durations of the time between vowels (intervocalic duration). They reasoned that low variability between these measurements would be consistent with syllable-timed languages, and high variability would be consistent with stress-timed languages. Indeed, in their comparison of English and Spanish, Grabe and Low (2002) found that this prediction was borne out; English had greater variability in both vocalic and intervocalic duration than Spanish. Again, there was variation in Spanish, but to a significantly lesser degree than in English.
As mentioned above, Dauer (1983, 1987) proposed that stress-timed rhythm was not the result of speakers attempting to maintain isochrony between stressed syllables but instead emerged as a result of a language’s syllable structure as well as the use of vowel reduction. For example, syllables in Spanish are typically composed of V, CV, VC, or CVC shapes (and some consonant clusters), but English has a wide variety of shapes, up to and including syllables like in the word strengths, which is composed of a CCCVCCC syllable shape. Each syllable in a series of CV and CVC syllable shapes (as found in Spanish) should be perceived as similar in duration to every other syllable in the series. However, imagine a sentence composed of a CCCVCCC syllable followed by a CV syllable, followed by a CCVC syllable, and so on (as might be found in English). Each of the syllables in this series should be perceived as distinct in duration from its neighboring syllables. A language in which syllable shapes change from complex to simple, complex to simple, might be perceived rhythmically as stress-timed (i.e. it might sound as if the speaker is attempting to maintain a constant duration between each stressed syllable). Such complex/simple variational patterns were verified by Ramus, Nespor, and Mehler (1999) when they used instrumental measurements to calculate the percentage at which vowels and consonants made up utterances. In stress-timed languages like English, consonants made up a greater percentage of utterance duration than in syllable-timed languages like Spanish. Again, it should be noted that variation in syllable length was not absent in Spanish because of syllable structure. Instead, its variation was weak compared to that of English.
Another contributor to variation in syllable duration is the process of vowel reduction. Vowel reduction occurs when a full vowel like /ɑ/ is produced as a schwa. This can be observed by comparing the first vowel in the word botany, which is a full vowel, to the first vowel in the word botanical, which is reduced (Flege & Bohn, 1989). Reduced vowels are produced with shorter duration than their full vowel counterparts. English makes frequent use of vowel reduction, thus creating variation in syllable length, but recent studies by Byers and Yavas (2016, 2017) have demonstrated that this phenomenon does not occur in Spanish. This can be interpreted as yet another phenomenon that yields greater variation in syllable duration in English compared to Spanish.
Taken together, these studies support Roach’s (1982) assertion that languages incorporate both stress- and syllable-timed patterns. It is likely that polysyllabic shortening is present in the speech of monolingual Spanish speakers. However, Spanish likely engages in polysyllabic shortening to a lesser degree than English. Currently, we cannot answer this definitively since direct comparisons between the two languages have not been made.
Polysyllabic shortening in bilingual speakers
Although several studies like those reported above have explored the development of speech rhythm in bilingual speakers, only two studies have investigated bilingual polysyllabic shortening per se. In Gibson and Summers (2018), Spanish-English bilingual adults and their peers who were functionally monolingual in English repeated nonwords in both English and Spanish. The group that was functionally monolingual in English had received roughly 30% of their early language exposure in Spanish, but this had diminished to about 12% by adulthood. This corresponded to their self-ratings of English proficiency, which were twice as high as their Spanish proficiency. The authors predicted that because of their dominance in English, these participants would apply polysyllabic shortening to both their syllable-timed and stress-timed languages. The results did not support the prediction. Despite their limited Spanish exposure and proficiency, these participants felicitously applied their use of polysyllabic shortening, with more occurring in English than in Spanish. It appeared then that early exposure to English was key to learning and employing polysyllabic shortening appropriately into adulthood.
In the same study, Gibson and Summers (2018) reported that the fully bilingual participants had received about 40% of their childhood language input in English, and this increased to about 50% as adults; they rated themselves highly proficient in both English and Spanish. Like their functionally monolingual peers, these participants produced more polysyllabic shortening in English than in Spanish, which was predicted. However, the magnitude of their use of it in English was more than three times greater than that of their functionally monolingual peers. The authors interpreted this to mean that there might be a sociolinguistic motivation to use polysyllabic shortening to signal membership in the majority language.
In another study of bilingual adults, Krivokapić (2013) compared the use of polysyllabic shortening in two groups of English speakers. One group spoke Indian English (syllable-timed) as a second language (L2), and the other spoke American English. Although there were some qualitative differences in the implementation of polysyllabic shortening (i.e. differences in which syllables were affected), speakers of both dialects increased the number of syllables produced per second for longer compared to shorter words.
Research questions and predictions
Based on the above review of the literature, polysyllabic shortening should occur to a greater degree in English than in Spanish. With respect to Spanish-speaking children who are learning English in an English-majority context in the USA, it would be reasonable to predict that they would use Spanish rhythm when speaking Spanish, but also use that rhythm when speaking English. That is, they would speak Spanish and English with a syllable-timed rhythm, which in the current study is related to weaker polysyllabic shortening.
Because no direct comparison has been made of Spanish and English speakers’ use of polysyllabic shortening, we first sought to determine if indeed the literature was accurate in suggesting that there was more polysyllabic shortening in English than in Spanish. Secondly, because so little is known about polysyllabic shortening in bilingual children, we sought to determine if Spanish-English bilingual children produced polysyllabic shortening to the same degree as did their monolingual English- and Spanish-speaking peers. Finally, this study sought to determine if polysyllabic shortening was similar in Spanish-English bilingual children’s two languages. Therefore, we asked the following questions and made predictions based on the literature review.
Methods
Participants
This study included 55 participants. There were 16 Spanish-English bilingual children, 16 monolingual English-speaking children, and 23 monolingual Spanish-speaking children. The bilingual and monolingual English speakers lived in the Southeastern USA, and the monolingual Spanish speakers lived in Chile. To be included in the study, children had to have no hearing, language, speech, cognitive, or behavioral issues requiring intervention or considered problematic as reported by the classroom teachers (both in the USA and Chile). Ages ranged from 4;5 to 7;7 (M = 5;10, SD = 7 months). Two Spanish-English bilingual children attended preschool, and all remaining participants attended kindergarten. One child from the Chilean group was substantially older than the rest (7;7, while the next oldest was 6;10), but was included in the final analysis because there was no substantive difference in outcomes without this participant. There was no statistically significant difference in age between the three groups, F(2, 52) = 1.39, p = .16.
Participants in the USA were drawn from classrooms in a school with a low socioeconomic status (SES) background (96% of students at the school participated in the National Free Lunch Program). The Chilean children also were drawn from a school with a low SES background (100% of the children received free breakfast and lunch in a program similar to that of the US National Free Lunch Program). Standardized tests of vocabulary in English and Spanish indicated that the group of bilingual children were dominant in Spanish (with Spanish scores near the mean for the test). Providing bilingual children with the Spanish receptive vocabulary words in both Spanish and English (instead of only in Spanish) increased their standard scores by less than half a standard deviation, which is a further indication of Spanish language dominance (i.e. using English was not especially helpful). In addition, bilingual children in the study had on average only 17 months of cumulative English language exposure. The age of English exposure ranged from 0 to 76 months (M = 50.12; SD = 18.86; see Table 1 for demographic information).
Participant background information, Mean (SD).
BL: bilingual; ML: monolingual; PPVT: Peabody Picture Vocabulary Test; ROWPVT: Receptive One Word Picture Vocabulary Test; ROWPVT-S: ROWPVT Spanish-only score, based on Spanish-exclusive responses; AoE: age of regular English exposure.
Measures
Language history questionnaire
Caregivers and teachers completed the Bilingual Input-Output Survey, which is part of the Bilingual English Spanish Assessment (BESA; Peña, Gutiérrez-Clellen, Iglesias, Goldstein, & Bedore, 2014). The BESA itself was not administered. Caregivers provided an hour-by-hour report of children’s language input and output (Spanish only, English only, or both) for the average weekday and average weekend, and teachers reported hourly school day experience. We weighted home experience 66% and school experience 33% to calculate a total weighted current English experience percentage (its inverse being Spanish experience). Caregivers additionally reported whether children predominately heard Spanish, English, or both for each year of their children’s lives, which was used to determine the age of first regular English exposure (i.e. the year in which parents reported that children heard either only English or both English and Spanish was treated as the year of first regular English language exposure).
Vocabulary
Only US participants were administered vocabulary tests because of external constraints involved in testing the Chilean participants. We reasoned that the absence of vocabulary scores for this group should not negatively impact our ability to interpret results because the Chilean children were monolingual, from a SES background similar to that of the US group, and had to meet the criteria to participate in the study. However, we include this as a limitation in the Limitations section.
We administered the Peabody Picture Vocabulary Test—Fourth Edition (PPVT; Dunn & Dunn, 2007), a popular standardized English receptive vocabulary test (M = 100, SD = 15) in which participants heard a target word and pointed to its corresponding color picture from a field of four with a 2 × 2 orientation.
We also administered the Receptive One-Word Picture Vocabulary Test-4: Spanish-Bilingual Edition (ROWPVT; Martin, 2013), a standardized Spanish-English receptive vocabulary test (M = 100, SD = 15) in which participants heard a target word and pointed to its corresponding picture from a field of four consecutive color pictures. The tester provided a Spanish target word first. As per the examiner’s manual, if the child produced an incorrect or no response, the word was repeated in English, allowing the child an opportunity for success in English. Two standard scores were calculated. The first was based on correct responses in either language (as per the examiner’s manual). The second was based on Spanish-only correct responses (see Table 1 for vocabulary standard scores).
Stimuli
Previous studies have shown that when phonetic material is matched across languages (e.g. keeping the number of segments and syllables the same across languages), differences in speech rhythm can disappear (Os, 1988, as referenced by Nespor, Shukla, & Mehler, 2011). Therefore, the current study sought to minimize the role that cross-linguistic differences in syllable shapes in Spanish and English might play in polysyllabic shortening. For example, Spanish-dominant children might change their articulation rate on words containing unfamiliar syllable shapes (e.g. consonant clusters) or unfamiliar word stress (e.g. antepenultimate stress). Every attempt was made to find words that were both phonologically similar and phonologically familiar in both languages. We use phonological familiarity to refer to phonological patterns that are common in the language (e.g. CV syllable shapes and penultimate stress would be familiar phonological patterns in Spanish, while CCCVCC syllable shapes and antepenultimate stress would not). English is dominated by monosyllabic words (76.92% of English words are monosyllabic; Shriberg & Kent, 1982), while Spanish has very few (7.54%; Navarro, 1968). Therefore, identifying stimulus words was difficult (see the Appendix for the stimulus items).
Vowel duration can be influenced by neighboring consonants. For example, the voiced plosive /d/ in body should lengthen the duration of the preceding vowel, but the /t/ in bata should not lengthen the preceding vowel. In order to determine whether these sorts of variables were exerting a significant influence on articulation rate, a series of paired sample t-tests was performed. Even without a correction for multiple comparisons, paired sample t-tests found no statistically significant difference in bilingual children’s articulation rate (syllables per second) for any of these pairs of words (e.g. body compared to bata). Similarly, when we collapsed the words into two-syllable and four-syllable categories (i.e. Spanish two-syllable words compared to English two-syllable words), paired sample t-tests revealed that there was no statistically significant difference between languages at either syllable length for the bilingual children. This was interpreted as support for the adequacy of the stimuli.
Colored drawings accompanied the target words. Drawings were centered on a laminated 8.5 inch × 11 inch paper and bound by a key ring. Children heard a sentence with the target word, followed by a description of the target, followed by the question “What is this?” For example, children heard This is a
Articulation rate
Articulation rate and speech rate are similar but not identical. They are similar since both are measured by dividing the number of speech units by units of time (Tsao, Weismer, & Iqbal, 2006). However, they differ because speech rate includes pauses and hesitations, which indirectly measures cognitive processing (Segalowitz, 2016), but articulation rate measures only the actual movement of the articulators, which taps into speakers’ level of articulatory control (Crystal & House, 1990). The current study was concerned with articulation rate, since pauses and hesitation are unlikely to contribute to polysyllabic shortening.
We based our measurements on the procedures used by Gibson and Summers (2018). In that study, polysyllabic shortening was calculated by subtracting the syllables per second of two-syllable words from the syllables per second of four-syllable words. If each syllable of a word is produced at the same rate, there should be no discrepancy between syllables per second for two- and four-syllable words (i.e. there is no polysyllabic shortening). However, if there is a positive discrepancy, it would indicate that the articulation rate increased as the number of syllables increased (i.e. the presence of polysyllabic shortening).
To create the syllables per second variable, we first calculated the duration of each word. Measurements of word duration were performed using the acoustic software TF32 (Milenkovic, 2001), with 600 BW (Hz), frequency range to 5.06 kHz, dynamic range of 48 dB, and pre-emphasis of 6 dB per octave in higher frequencies. Listeners were fluent in the target language. For a word to be included in the measurement, it had to be a delayed repetition, conform to the targeted syllable shape, and possess the targeted stress location. Delayed repetition was used to minimize the possibility that polysyllabic shortening was due to mimicking the production heard in the stimulus. Productions were included if children substituted individual phonemes (e.g. seven produced as sefen) or metathesized phonemes (e.g. alameda produced as amaleda). Across all participants in both languages, only two tokens did not meet this standard and were not included in the analysis. Duration measurements began at the onset of the first phoneme to the final glottal pulse of the word. In Spanish, many children included the indefinite articles un or una before the word. Measurements included only the target word and not the articles. Two children frequently replaced final vowels with weak palatal fricatives; in those cases, the end of the frication of the final syllable was treated as the end of the measurement. Duration was converted into syllables per second.
Reliability
For the US participants, three pairs of research assistants performed the acoustic measurements (one pair for monolingual English, one for bilingual English, and one for bilingual Spanish). Research assistants were undergraduate and graduate students trained by the first author. For training, each pair of research assistants sat with the first author and performed measurements collaboratively on data that was not included in the current study. During the training, research assistants asked the first author questions and discussed the process. The research assistants then independently measured data that was not included in the current study. The pairs of research assistants met with the first author to measure inter-rater reliability on the training items and to discuss differences in measurements. This process was repeated on data not used in the current study until the pairs of listeners reached 100% consensus in their measurements to within 30 milliseconds. After training, all measurements were performed independently by the research assistants.
Three recordings were measured independently by each pair of research assistants, and Pearson’s product moment correlation was used to determine inter-rater reliability, which was high (r = .94 for bilingual English, r = .93 for bilingual Spanish, and r = .95 for monolingual English). The first author performed the measurements for the Chilean participants. Four recordings were randomly chosen and re-measured two weeks after the initial measurement to establish intra-rater reliability, which was high (r = .97).
Procedure
In the USA, all testing occurred in children’s schools by the first author and graduate students fluent in the language of testing. The first author trained the testers on test administration. Effort was made to test in quiet spaces with limited distractions. Vocabulary testing and polysyllabic shortening testing could occur on the same day, but this was based on time constraints. Language of testing for the polysyllabic shortening testing was counterbalanced, but there was no such counterbalancing for the vocabulary tests. However, participants received testing in only one language per day. For example, a child could receive a Spanish vocabulary test and Spanish polysyllabic shortening test on the same day, but never a Spanish vocabulary test and English polysyllabic shortening test on the same day (or any other mixing of Spanish and English testing on the same day).
Recording devices were positioned 12 cm from the participants’ mouths. If children shifted in their chairs, they were redirected to their original positions. Words were presented via over-the-ear headphones. A stereo splitter cable allowed the tester to hear the stimuli and advance the corresponding pictures appropriately. Testers and children were seated in chairs beside one another with the stimuli book in front of them on a table. Testers read instructions aloud in the language of testing. There were two two-syllable training stimuli for each language. If during testing a child made a significant error in pronunciation, the stimulus was played again. That is, if the child produced a distortion that made the word difficult for the tester to understand or produced an omission or addition that changed the syllable shape, the stimulus item was played again. In a few cases, the tester provided the child with a spoken model of the word and then played the stimulus recording again to guarantee a delayed repetition.
In Chile, all procedures were the same as those in the USA except that children did not hear pre-recorded stimuli through headphones. Instead, the tester, who was the second author, administered the test verbally in real time using the delayed repetition method.
Results
To compare monolingual English- and Spanish-speaking children’s use of polysyllabic shortening, a mixed model repeated measures analysis of variance (ANOVA) with Length (syllables per second for two- and four-syllable words) as the within-subjects variable and Language Group (monolingual English versus monolingual Spanish) as the between-subjects variable was performed. Monolingual English speakers were tested in English and monolingual Spanish speakers were tested in Spanish. There was a significant main effect for Length, F(1, 37) = 45.97, p < .001, partial eta squared = .55, with syllables per second faster for four-syllable words (M = 5.15, SE = .13) than for two-syllable words (M = 4.43, SE = .15). There was no significant main effect for Language Group, F(1, 37) = .387, p = .53, partial eta squared = .01, or for the interaction between Length and Language Group, F(1, 37) = .1.76, p = .19, partial eta squared = .04. This indicated that monolingual English and monolingual Spanish children produced polysyllabic shortening to a similar degree in their respective languages.
To compare Spanish-English bilingual children’s use of polysyllabic shortening to that of their monolingual English-speaking peers, a mixed model repeated measures ANOVA with Length (syllables per second for two- and four-syllable words) as the within-subjects variable and Language Group (monolingual English versus bilingual speakers tested in English) as the between-subjects variable was performed. There was a significant main effect for Length, F(1, 30) = 17.23, p < .001, partial eta squared = .365, with syllables per second faster for four-syllable words (M = 5.11, SE = .13) than for two-syllable words, (M = 4.6, SE = .17). There was no significant main effect for Language Group, F(1, 30) = 1.15, p = .291, partial eta squared = .03, or for the interaction between Length and Language Group, F(1, 30) = .36, p = .55, partial eta squared = .01. This indicated that bilingual children produced polysyllabic shortening in English to a degree similar to that of their monolingual English-speaking peers (see Figure 1).
Monolingual children’s average English and Spanish syllables per second by syllable length.
To compare Spanish-English bilingual children’s use of polysyllabic shortening to that of their monolingual Spanish-speaking peers, a mixed model repeated measures ANOVA with Length (syllables per second for two- and four-syllable words) as the within-subjects variable and Language Group (monolingual Spanish versus bilingual speakers tested in Spanish) as the between-subjects variable was performed. There was a significant main effect for Length, F(1, 37) = 149.08, p < .001, partial eta squared = .80, with syllables per second faster for four-syllable words (M = 5.45, SE = .11) than for two-syllable words (M = 4.45, SE = .11). There was no significant main effect for Language Group, F(1, 37) = .61, p = .44, partial eta squared = .02, or for the interaction between Length and Language Group, F(1, 37) = 2.62, p = .11, partial eta squared = .07. This indicated that bilingual children produced polysyllabic shortening in Spanish to a degree similar to that of their monolingual Spanish-speaking peers (see Figure 1).
To compare Spanish-English bilingual children’s use of polysyllabic shortening in English and Spanish, we performed a 2 × 2 repeated measures ANOVA, using only bilingual data, with Length (syllables per second for two- and four-syllable words) and Language (English versus Spanish) as within-subjects variables. Results showed no main effect for Language, F(1, 15) = .03, p = .84, partial eta squared < .01. However, there was a statistically significant main effect for Length, F(1, 15) = 13.03, p < .001, partial eta squared = .46, with four-syllable words produced at a higher articulation rate (M = 5.42, SE = .14) than the two-syllable words (M = 4.64, SE = .14). In addition to main effects, there was a statistically significant interaction between Length and Language, F(1, 15) = 13.61, p = .002, partial eta squared = .47 (see Figure 2). We used the Holm’s sequential Bonferroni procedure to control for multiple comparisons in the post-hoc analysis of six pairs of means. There were statistically significant differences between English two-syllable versus English four-syllable words, t(15) = 2.864, p = .012, d = .71, English two-syllable versus Spanish four-syllable words, t(15) = 3.82, p = .002, d = 1.02, English four-syllable versus Spanish two-syllable words, t(15) = 4.26, p = .001, d = 1.12, and Spanish two-syllable versus Spanish four-syllable words t(15) = 11.59, p < .001, d = 2.90.
Bilingual children’s average English and Spanish syllables per second by syllable length.
As a further post-hoc, we calculated the average difference between two- and four-syllable words (i.e. polysyllabic shortening) within English (M = .43, SD = .60) and Spanish (M = 1.12, SD = .38) and compared them using a paired sample t-test. There was a statistically significant difference, t(15), p = .002, d = 1.04, which indicated that polysyllabic shortening was greater in Spanish than in English (see Figure 3).
Histograms representing the effect sizes for the within-language difference between two- and four-syllable words (i.e. polysyllabic shortening).
Discussion
The current study attempts to contribute to our understanding of bilingual phonological development by exploring speech rhythm through the phenomenon of polysyllabic shortening. We asked three questions. Do monolingual English-speaking and monolingual Spanish-speaking children use polysyllabic shortening to the same degree? A direct comparison has never before been reported. Do Spanish-English bilingual children produce polysyllabic shortening in English and Spanish to the same degree as their monolingual English- and Spanish-speaking peers? And do Spanish-English bilingual children produce polysyllabic shortening to the same degree in their English and Spanish productions? These results are unexpected and complex and will require further study.
Comparison of monolingual speakers (Question 1): A universal phonetic constraint
To our knowledge, this is the first study to directly compare polysyllabic shortening in English and Spanish for any age group. We predicted that there should be more polysyllabic shortening in English, usually treated as a stress-timed language, than in Spanish, usually treated as a syllable-timed language. This was not the case. Indeed, in our study, monolingual English and Spanish speakers use polysyllabic shortening to the same degree. This result is unexpected since, as our literature review shows, many studies indicate that there should be limited or no polysyllabic shortening in Spanish but a significant amount of it in English.
At least in the current study, when syllable shapes are controlled for, there is no statistically significant difference in articulation rates for monolingual English- and Spanish-speaking children. This result appears consistent with the findings of Dauer (1983, 1987) and others that cross-linguistic differences in the use of polysyllabic shortening emerge from differences in languages’ syllabic structure rather than for the maintenance of speech rhythm.
However, these results raise the following question. Why is polysyllabic shortening occurring at all? To answer this question, we appeal to what Quené (2008) called a universal phonetic constraint. He analyzed the speech tempo from a corpus of interviews in Dutch and found that longer phrases were produced at a faster tempo than shorter phrases. He argued that speakers are constrained by universal phonetic factors. “In order to produce many syllables in one breath, speakers need to attain a fast tempo” (Quené, 2008; p. 1109). A similar result has been identified in an investigation of American English speakers who were reading text aloud (Bishop & Kim, 2018). Jacewicz, Fox, and Wei (2010) explain, however, that it is unlikely that the articulators are actually speeding up during the production of longer utterances but instead the syllables and segments within the utterance are being reduced. We propose that the same phenomenon is occurring in the current study, even though the lengths of utterances are quite short (single words). Polysyllabic shortening during single word utterances might occur because of the same universal phonetic constraints that apply to longer utterances.
Polysyllabic shortening in bilinguals compared to monolinguals (Question 2)
We predicted that bilingual participants’ use of polysyllabic shortening would be similar to that of their monolingual Spanish-speaking peers but less than that of their monolingual English-speaking peers. Indeed, bilingual children’s use of polysyllabic shortening in Spanish was similar to that of their monolingual Spanish-speaking peers. However, contrary to our prediction, their use of polysyllabic shortening in English was also similar to that of their monolingual English-speaking peers. This outcome, too, is consistent with the notion of a universal phonetic constraint.
What is salient in these results is that bilingual children produce the same degree of polysyllabic shortening in English as their monolingual peers despite having low English proficiency (as indicated by low English vocabulary scores and limited number of years of English exposure). This outcome suggests that the universal phonetic constraint responsible for polysyllabic shortening is not strongly related to language proficiency. This phenomenon might be observed in the results of Gibson and Summers (2018) in which adults with limited exposure to and knowledge of Spanish produced polysyllabic shortening when repeating Spanish nonwords. In fact, we speculate that, controlling for syllable shape, polysyllabic shortening would occur to similar degrees when comparing groups of speakers from any two languages. Future studies should explore this possibility.
Polysyllabic shortening in bilinguals’ two languages (Question 3)
We predicted that the degree of bilingual children’s polysyllabic shortening would be similar in both of their languages because of their limited experience with English. That is, they would behave like Spanish speakers whether speaking Spanish or English. This prediction failed (and as the results from Question 1 indicate, our assumptions about Spanish speakers were wrong). Instead, although bilingual children produce polysyllabic shortening in both of their languages, they produce more of it in Spanish than in English.
These results indicate that while the universal phonetic constraint might not be strongly associated with language proficiency, a relationship, in fact, does exist. Allen and Hawkins (1980) proposed, and subsequent research supported (Sirsa & Redford, 2011), that L2 learners begin with syllable-timed rhythms and move toward stress-timed rhythms, whether the first language (L1) is stress-timed or not (Ordin & Polyanskaya, 2015). This appears to be due to the growth of articulatory control as proficiency in the L2 improves (Ordin & Polyanskaya, 2014). Both the English and Spanish stimuli in the current study contained no consonant clusters and consisted of syllable shapes and stress patterns that occur commonly in both Spanish and English. Since frequency of use plays a substantial role in the development of speech patterns (Bybee, 2001, 2010; Pierrehumbert, 2003) and leads ultimately to their automatization, it may have been the case that these children had greater articulatory control in Spanish than English. We propose that the application of a universal phonetic constraint was more efficient in Spanish than in English and, therefore, these children simply were more successful at its implementation in Spanish than in English.
Limitations
The number of participants in each group was small so attempts to generalize from these results is speculative at this stage. In addition, there were only eight stimulus items in each language. Future studies should use more items or naturalistic data from connected speech to have a better understanding of this phenomenon as it is made manifest in bilinguals’ typical speech. The phonological makeup of the English and Spanish stimuli was common in both languages. This prevents us from determining that difference in syllable structure is indeed the cause of different perceived speech rhythms across languages. Future studies should include both English-typical and Spanish-typical syllable shapes to examine what degree syllable shape might play in polysyllabic shortening. In addition, data from the current study were taken from a single time point. Future studies should incorporate a longitudinal design to determine the role of experience in the use of polysyllabic shortening.
Because both the Spanish and English stimulus words were constructed of the same syllable shapes and the same lexical stress patterns, it may have been the case that differences at the segmental level influenced the outcome. As explained by Dauer (1983), vowel reduction is a phonological phenomenon that contributes to the impression of stress-timing. In the current study, stimulus words were not controlled for their likelihood to initiate vowel reduction. Future studies should construct nonword stimuli that control for vowel reduction to determine its role, if any, in the presence of polysyllabic shortening.
Finally, there were some differences in the procedures in Chile compared to the USA. For example, we were not able to administer vocabulary testing to the children in Chile. As outlined in the Methods section, we reasoned that the absence of vocabulary scores for this group should not negatively impact our ability to interpret results. However, it leaves open the possibility that the Chilean children, despite their similar SES background and absence of cognitive/linguistic impairments, were particularly advanced or delayed in their linguistic abilities compared to the US group. Future studies should include standardized language testing for all groups. In addition, children in Chile did not hear pre-recorded stimuli through headphones but instead had the task administered orally by an examiner. Results could potentially differ based on the different methods of presentation. Future studies should administer the stimuli to all participants using identical procedures.
Conclusion
It appears that polysyllabic shortening arises from the implementation of a universal phonetic constraint. That is, to produce many compared to fewer syllables on a single puff of air, speakers must increase their articulation rate. It is likely that the universal phonetic constraint responsible for polysyllabic shortening is only modestly related to language proficiency. For the bilingual children, greater use of polysyllabic shortening in Spanish than English may be related to better Spanish than English articulatory control. We speculate that, controlling for syllable shape, polysyllabic shortening would occur to similar degrees when comparing groups of speakers from any two languages.
Footnotes
Appendix
Stimulus items in English and Spanish (with English translation).
| English | Spanish |
|---|---|
| body | bata (robe) |
| lettuce | lejos (far) |
| hollow | jugo (juice) |
| seven | joven (young) |
| Alabama | alameda (park or boulevard) |
| avocado | abogado (lawyer) |
| Mississippi | mariposa (butterfly) |
| terracotta | calabaza (pumpkin) |
Acknowledgements
The authors wish to thank Paola Flores for her assistance in data collection. Special thanks to Lauren Aiello.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by a Humanities and Social Sciences Summer Research Fellowship from Louisiana State University and the Louisiana Board of Regents (Grant (LEQSF(2014-18)-RD-A-0).
