This paper examines the differences in the division of intonation phrases and in the tonal structure of the nuclear configuration (i.e., the last pitch accent and the following boundary tone) in imitated and in authentic English-accented Spanish. The same Spanish text was read by four native speakers of American English, who produced the text with a real English foreign accent in Spanish, and six native speakers of Spanish, who read the text twice: in L1 Spanish and in fake English-accented Spanish. An auditory analysis of the data was carried out along with an inspection of the
Forensic Phonetics is the branch of Legal Linguistics that examines the characteristics of human speech for legal purposes. It covers a variety of issues, such as the identification of the speaker’s phonetic profile (e.g., sex, social class, dialectal variety, etc.) or the comparison between known and unknown speech samples so as to help identify a particular speaker.
However, the identification of a speaker through his voice tends to be a difficult task and a great challenge for Forensic Phonetics since it is common amongst criminals to voluntarily disguise their accent. According to Masthoff (
There are many ways to modify one’s voice, both electronically and physiologically (Rodman,
This study deals with this last type of voice alteration, that is, the imitation of a foreign accent. In particular, it examines the production of imitated English-accented Spanish produced by L1 Spanish speakers as opposed to real English-accented Spanish produced by L1 English speakers.
Even though most studies on fake foreign accent agree in that there are certain cues in the imitated speech that might help identify the fake productions (Markham,
There are not many studies on the imitation of a foreign accent in Spanish. From a perceptual perspective, Schoonmaker-Gates (
Gibson, Blecua, and Cicres (
From a production perspective, Cicres and Fernández Trinidad (
As far as suprasegmental features are concerned, Estebas-Vilaplana (
Estebas-Vilaplana (
The present paper analyzes two other prosodic features in imitated and in authentic English-accented Spanish, namely, the division of a text into intonation phrases and the intonational patterns at the end of those phrases. The aim of this study is twofold: 1) to examine the location of prosodic breaks and the tonal structure of the nuclear configurations in imitated English-accented Spanish (IEAS), and 2) to compare the results with both L1 Spanish and authentic English-accented Spanish (AEAS).
The materials used in this study consist in a phonetically balanced text in Spanish that was read by 4 native speakers of American English and 6 native speakers of Spanish. The text was obtained from Bruyninckx, Harmegnies, Llisterri, and Poch-Olivé (
The informants were given the text before the recording, so that they could prepare it in advance for about 5 minutes. The Spanish native speakers had to read the text twice. For the first recording, they were instructed to read it imitating a typical English accent in Spanish. In the second recording, they read it in their normal L1 Spanish accent. The English speakers read the text once, with their real English-accented Spanish.
Thus, the corpus comprised 4 recordings in authentic English-accented Spanish (AEAS), 6 recordings in L1 Spanish and 6 recordings in fake or imitated English-accented Spanish (IEAS).
The recordings of the native Spanish speakers were carried out in a soundproof booth at the Phonetics Laboratory at the University of Girona, with an Audio-Technica AT2050 microphone and a Focusrite Scarlett 2i2 digitizer card. The American English informants were recorded in the Speech Laboratory of the University of Navarra, under the same conditions.
The informants were 3 male and 3 female native speakers of Peninsular Spanish, without a particularly regional accent, and 2 male and 2 female native speakers of North American English, all aged between 20 and 40 and with higher education.
The Spanish-speaking informants had an intermediate level of English, whereas the American English speakers had an intermediate level of Spanish.
The Spanish speakers are identified as SS1–6 and the English speakers as ES1–4.
Auditory and acoustic analyses of the data were performed. For these analyses, the text was divided into 15 phrases, which allowed us to carry out a detailed prosodic inspection of the
(1)
01. | El joyero Federico Vanero│ |
02. | ha sido condenado por la Audiencia de Santander│ |
03. | a ocho meses de arresto mayor│ |
04. | y cincuenta mil pesetas de multa│ |
05. | por un delito de compra de objetos robados.│ |
06. | La vista oral se celebró el miércoles pasado│ |
07. | y, durante ella,│ |
08. | uno de los fiscales, Carlos Valcárcel,│ |
09. | pidió para el joyero tres años de prisión menor│ |
10. | y una multa de cincuenta mil pesetas.│ |
11. | Gracias a las revelaciones de Vanero│ |
12. | de hace dos años y medio│ |
13. | se llegó a descubrir la existencia de una sospechosa mafia policial en España,│ |
14. | parte de la cual se vio envuelta en el llamado│ |
15. | “caso el Nani”. |
Each phrase was stored on a separate audio file. Overall 180 phrases were analyzed for the 6 Spanish speakers (90 produced in L1 Spanish and 90 in IEAS) and 60 phrases for the 4 English speakers (all of them produced in AEAS). The prosodic analysis of the corpus was done by means of
For this study, no distinction was made between levels of prosodic phrasing, that is, the presence of a prosodic break was equally transcribed irrespective of being a major break (or intonation phrase) or a minor break (or intermediate phrase). In the acoustic analysis, all breaks were marked as ‘PB’ (prosodic break). The main cue to identify a prosodic break was the presence of a boundary tone. In some cases, this tone was accompanied by a pause. False breaks, such as stammering or disfluencies, were excluded from the analysis.
The prosodic annotation of the final configuration of the pitch contours followed the Sp_ToBI system conventions, firstly proposed by Beckman, Díaz-Campos, McGory, and Morgan (
In this paper, only the final configuration of the
The Sp_ToBI conventions were used to examine not only the productions of the Spanish speakers (both in L1 Spanish and in IEAS) but also those of the English speakers in AEAS. Even though there might be differences in the intonation patterns produced by these two groups of speakers, it was considered necessary to annotate the contours with the same labelling system so as to facilitate the comparison between the two languages. Since all the data were produced in Spanish, the Sp_ToBI system was considered the most suitable labelling tool for the prosodic transcriptions. Furthermore, the Sp_ToBI annotation conventions used in this study could account for the majority of the final pitch movements found in the data of both groups of speakers.
Sp_ToBI notation conventions for different final pitch patterns.
Sp_ToBI | Pitch pattern |
---|---|
L* L% | Low-fall |
(L)+H* L% | High-fall |
(L)+H* H% | Rise |
(L)+H* !H% | Fall-to-mid |
(L)+H* LH% | Fall-rise |
* =% | Sustained |
The annotation process was performed by two trained transcribers. Each transcriber generated a time-aligned display of the speech signal (waveform and
Example of the acoustic analysis (waveform,
Finally, a statistical analysis of the data was carried out to compare the proportion (or frequency of occurrence) of the final tonal configurations in the three samples of speech (L1 Spanish, IEAS, and AEAS). The statistical analysis was performed using a Chi-square test and the results of contingency tables. The adjusted residual values were calculated, which have a mean of 0 and a standard deviation of 1. The values that are higher than 1.96 and lower than -1.96 indicate that there is a significant lack of proportion in favor of one variant or the other. IBM SPSS Statistics 21 was used for this purpose.
First, the results of the Spanish speakers both in L1 Spanish and in imitated English-accented Spanish (IEAS) are presented. Then, the productions of the English speakers in authentic English-accented Spanish (AEAS) are examined. For each group of speakers, the results are divided into two sections: 1) phrasing and 2) the tonal analysis of the nuclear configuration (last pitch accent and boundary tone).
The presence of a prosodic break, signaled by a boundary tone and an optional pause, was analyzed for the two productions of the six Spanish speakers, namely, in L1 Spanish and in IEAS. The text in (2) shows the places where a prosodic break was produced in L1 Spanish and the percentage of occurrence.
(2) L1 Spanish
01. | El joyero (33.3 %)│Federico Vanero (100 %)│ |
02. | ha sido condenado (33.3 %)│por la Audiencia de Santander (100 %)│ |
03. | a ocho meses de arresto mayor (100 %)│ |
04. | y cincuenta mil pesetas de multa (100 %)│ |
05. | por un delito de compra de objetos robados. (100 %)│ |
06. | La vista oral (83.3 %)│se celebró el miércoles pasado (100 %)│ |
07. | y, durante ella, (100 %)│ |
08. | uno de los fiscales, (100 %)│Carlos Valcárcel, (100 %)│ |
09. | pidió para el joyero (66.6 %)│tres años de prisión menor (100 %)│ |
10. | y una multa de cincuenta mil pesetas. (100 %)│ |
11. | Gracias a las revelaciones de Vanero (100 %)│ |
12. | de hace dos años (66.6 %)│y medio (100 %)│ |
13. | se llegó a descubrir la existencia de una sospechosa mafia policial en España, (100 %)│ |
14. | parte de la cual (83.3 %)│se vio envuelta en el llamado (100 %)│ |
15. | “caso (66.6 %)│el Nani”. (100 %)│ |
The text in (3) shows the places where a prosodic break was produced in IEAS and the percentage of occurrence. The phrases marked in bold indicate that there was no break in this position when the text was read in L1 Spanish, whereas those marked in bold and italics show that there has been an increase in the number of prosodic breaks in this position. There are no cases with a break in L1 Spanish and no break or less occurrences of it in IEAS.
(3) Imitated English-accented Spanish (IEAS)
01. | |
02. | |
03. | |
04. | |
05. | |
06. | |
07. | y, durante ella, (100 %)│ |
08. | uno de los fiscales, (100 %)│Carlos Valcárcel, (100 %)│ |
09. | |
10. | |
11. | Gracias a las revelaciones de Vanero (100 %)│ |
12. | |
13. | |
14. | parte de la cual (83.3 %)│ |
15. | “caso (66.6 %)│el Nani”. (100 %) |
The analysis of phrasing in the two texts shows that there are more instances of prosodic breaks in the IEAS reading than in the L1 Spanish reading.
Example of the utterance
Example of the utterance
The higher number of intonation breaks in IEAS is confirmed with the results provided in
Individual and total number and mean value of prosodic breaks produced by the Spanish speakers (SS) in the two readings of the text: in L1 Spanish and in imitated English- accented Spanish (IEAS).
Speakers | Number of prosodic breaks | |
---|---|---|
L1 Spanish | IEAS | |
SS1 | 23 | 33 |
SS2 | 24 | 32 |
SS3 | 23 | 33 |
SS4 | 21 | 29 |
SS5 | 21 | 23 |
SS6 | 20 | 30 |
Total | 132 | 180 |
Mean | 22 | 30 |
Percentage of (dis)agreement in the presence of a prosodic break (PB) in L1 Spanish and in imitated English-accented Spanish (IEAS).
L1 Spanish | IEAS | |
---|---|---|
% of agreement in the location of PB | 69.5 | 45 |
% of disagreement in the location of PB | 30.5 | 55 |
In order to analyze the final pitch movements, utterances have been divided into two kinds: 1) those that convey unfinished information and 2) those that indicate that the message is complete. This distinction has been used since Navarro Tomás (
The tone inventory proposed by Navarro Tomás (
Number of occurrences of the final tonal configurations produced by the six Spanish speakers in L1 Spanish in the
Sp | Number of occurrences of the final tonal configurations in L1 Spanish | ||||||
---|---|---|---|---|---|---|---|
(L)+H* H% | (L)+H* !H% | (L)+H* LH% | * =% | L* L% | L* L% | (L)+H* L% | |
SS1 | 17 | - | 1 | - | 1 | 4 | - |
SS2 | 19 | - | - | - | 1 | 4 | - |
SS3 | 13 | 5 | - | - | 1 | 4 | - |
SS4 | 15 | - | - | 2 | - | 4 | - |
SS5 | 16 | 1 | - | - | - | 4 | - |
SS6 | 13 | 1 | 1 | - | 1 | 4 | - |
- |
The text includes four phrases in the
(4) Phrases in the
05. | (…) por un delito de compra de objetos robados.│ |
06. | (…) se celebró el miércoles pasado.│ |
10. | (…) y una multa de cincuenta mil pesetas.│ |
15. | (…) “caso el Nani”.│ |
All speakers produced the four phrases in the
The most recurrent tonal configuration in the
Example of a fall-to-mid tone (L+H* !H%) and a rising tone (L+H* H%) for the utterance
A minority of cases presented a final fall-rise (L+H* LH%) or a sustained tone (* =%).
Example of final fall-rise (L+H* LH%) and a rise (L+H*H%) at the end of the two phrases of the utterance
Example of final sustained pitch (L+H* =%) for the phrase
The results also show four cases of L* L% in the
Number of occurrences of the final tonal configurations produced by the six Spanish speakers in imitated English-accented Spanish (IEAS) in the
Sp | Number of occurrences of the final tonal configurations in IEAS | ||||||
---|---|---|---|---|---|---|---|
(L)+H* H% | (L)+H* !H% | (L)+H* LH% | * =% | L* L% | L* L% | (L)+H* L% | |
SS1 | 29 | - | - | - | - | 4 | - |
SS2 | 26 | 1 | - | 1 | - | 4 | - |
SS3 | 27 | 2 | - | - | - | 4 | - |
SS4 | 15 | 2 | 8 | - | - | 4 | - |
SS5 | 16 | 1 | - | - | 2 | 4 | - |
SS6 | 16 | 1 | 8 | - | 1 | 4 | - |
As with L1 Spanish, all speakers used L* L% in the
Example of the utterance
The fall-rise pattern (L+H* LH%) was found in some of the productions of two speakers (SS4 and SS6).
Example of the utterance
All speakers, except for SS1, showed a few cases of a final fall-to-mid tone ((L)+H* !H%).
Example of the utterance
Finally, two speakers (SS5 and SS6) used the low-fall (L* L%) typical of the
Percentage of occurrence of the tonal configurations in
Tonal configurations | Percentage of occurrence of tones in |
|
---|---|---|
L1 Spanish | IEAS | |
(L)+H* H% | 86.1 | 82.7 |
(L)+H* !H% | 6.6 | 4.5 |
(L)+H* LH% | 1.8 | 10.3 |
* =% | 1.8 | 0.6 |
L* L% | 3.7 | 1.9 |
Percentage of occurrence of the tonal configurations in
Tonal configurations | Percentage of occurrence of tones in |
|
---|---|---|
L1 Spanish | IEAS | |
L* L% | 100 | 100 |
L+H* L% | 0 | 0 |
In order to verify whether the different reading conditions, namely, L1 Spanish versus IEAS, had any statistical effect on the frequency of occurrence of the different tonal structures, an analysis of contingency tables was carried out for the
According to these results, there is a clear usage of L1 Spanish intonation in the production of IEAS. The tones of the
The text in (5) shows the places where a prosodic break was produced in authentic English-accented Spanish (AEAS) and the percentage of occurrence. The phrases marked in bold indicate that neither in L1 Spanish nor in IEAS there was a break in this position. Thus, only some English speakers produced a prosodic break after the phrases in bold.
(5) Authentic English-accented Spanish (AEAS)
01. | El joyero (100 %)│ |
02. | ha sido condenado (100 %)│por la Audiencia de Santander (100 %)│ |
03. | a ocho meses (25 %)│de arresto mayor (100 %)│ |
04. | |
05. | por un delito (25 %)│de compra de objetos robados. (100 %)│ |
06. | La vista oral (100 %)│se celebró (75 %)│el miércoles pasado (100 %)│ |
07. | |
08. | uno de los fiscales, (75 %)│Carlos Valcárcel, (100 %)│ |
09. | pidió para el joyero (66.6 %)│tres años de prisión (25 %)│menor (100 %)│ |
10. | y una multa de cincuenta mil pesetas. (100 %)│ |
11. | Gracias a las revelaciones de Vanero (100 %)│ |
12. | de hace dos años (75 %)│y medio (100 %) |
13. | |
14. | parte de la cual (100 %)│se vio envuelta (25 %)│en el llamado (100 %)│ |
15. | “caso (25 %)│el Nani”. (100 %)│ |
The results in (5) show that in 98 % of cases English speakers agree with Spanish speakers in the location of prosodic breaks. There are only four instances, corresponding to 2 % of the total number of breaks, in which an intonation unit was produced in AEAS but not in L1 Spanish or IEAS. Most of these breaks, for instance,
Individual and total number and mean value of prosodic breaks produced by the English speakers in authentic English-accented Spanish (AEAS).
Speakers | Number of prosodic breaks in AEAS |
---|---|
ES1 | 26 |
ES2 | 27 |
ES3 | 28 |
ES4 | 27 |
Total | 108 |
Mean | 27 |
Percentage of (dis)agreement in the presence of a prosodic break (PB) in authentic English-accented Spanish (AEAS).
AEAS | |
---|---|
% of agreement in the location of a prosodic break | 55.9 |
% of disagreement in the location of a prosodic break | 44.1 |
These results are more similar to the behavior observed in IEAS (with 45 % of agreement and 55 % of disagreement in the location of prosodic breaks) than that detected in the L1 Spanish productions (with 69.5 % of agreement and 30.5 % of disagreement; see
Number of occurrences of the final tonal configurations produced by the four English speakers in authentic English-accented Spanish (AEAS).
Sp | Number of occurrences of the final tonal configurations in AEAS | ||||||
---|---|---|---|---|---|---|---|
(L)+H* H% | (L)+H* !H% | (L)+H* LH% | * =% | (L)+H* L% | L* L% | (L)+H* L% | |
ES1 | 6 | 3 | 8 | 2 | 3 | 3 | 1 |
ES2 | 9 | 1 | 12 | - | 1 | 4 | - |
ES3 | 15 | 1 | 8 | - | - | 3 | 1 |
ES4 | 8 | 1 | 12 | - | 2 | 3 | 1 |
As far as the tonal patterns for the
Examples of low fall (L* L%) at the end of the second phrase and a rise (L+H* H%) at the end of the first phrase for the utterance
However, whereas all Spanish speakers, both in L1 Spanish and in IEAS, only produced the L* L% configuration in the
Example of two high-fall patterns (L+H* L%) in
Even though the productions with an L+H* L% pattern in AEAS in the
Percentage of occurrence of the tonal configurations in
Tonal configurations | Percentage of occurrence of tones in |
---|---|
(L)+H* H% | 41.3 |
(L)+H* !H% | 6.5 |
(L)+H* LH% | 43.5 |
* =% | 2.2 |
(L)+H* L% | 6.5 |
Percentage of occurrence of the tonal configurations in
Tonal configurations | Percentage of occurrence of tones in |
---|---|
L* L% | 81.25 |
(L)+H* L% | 18.75 |
The variability in the final falling movement found in AEAS is not present in L1 Spanish or in IEAS, which consistently show a final L* L% contour. The low-fall or
Thus, although at this point no clear conclusions can be reached from our data, given the reduced number of productions, the presence of both a high-fall (L+H* L%) and a low-fall (L* L%) in the declarative sentences of AEAS suggests that this might be an issue to study further, since the variability in the production of the final accent in statements can be a helpful cue to distinguish real English-accented Spanish from a fake accent.
In the
Example of two fall-rise patterns (L+H* LH%) in the
Finally, the fall-to-mid accent (L+H* !H%) and the sustained tone (* =%) are used in a minority of cases in AEAS, namely, 6.5 % and 2.2 % respectively, as shown in
Example of a fall-to-mid tonal configurations (L+H* !H%) in
The low occurrence of the fall-to-mid and the sustained patterns in AEAS is very similar to behavior observed in L1 Spanish and in IEAS. As exhibited in
Finally, the results of the statistical analysis comparing the frequency of occurrence of the tonal configurations found in IEAS and in AEAS in the
This study has compared the realization of some prosodic features in a corpus of read declarative sentences produced in L1 Spanish, in imitated English-accented Spanish (IEAS), and in authentic English-accented Spanish (AEAS). In particular, it has examined the main differences and similarities in the location of prosodic breaks and in the tonal categorization of the final or nuclear configuration of an intonation phrase, namely, the last pitch accent and the following boundary tone.
As far as phrasing is concerned, the results show a higher number of prosodic breaks in IEAS than in L1 Spanish. This seems to indicate that in imitated speech speakers may exaggerate some of their productions and overdo the number of prosodic breaks. The exaggeration in the production of phonetic features in imitated speech has been attested by some authors, such as Zetterholm (1997), who studied the imitation of the voice of Swedish politicians by other L1 Swedish speakers. She concluded that for a good impersonation it is important to exaggerate several typical features of the target speaker. Estebas-Vilaplana (
The comparison in the location of prosodic breaks in IEAS and in AEAS shows similar results, with a 90 % of coincidence in the number of intonation units in the real and in the fake accents. The main reason for this high number of prosodic breaks in IEAS and in AEAS might be due to a different cause. Whereas in IEAS the increase in the number of intonation units might be related to a tendency towards exaggerating an imitated accent, in AEAS the high number of breaks may respond to the hesitation effects typical of a foreign accent (Reitbrecht Hirschfeld, 2015). In any case, the similar number of prosodic breaks found in IEAS and AEAS indicates that phrasing is not a reliable cue to differentiate real English-accented Spanish from fake English-accented Spanish since it works very similarly in the two conditions.
The analysis of the tonal structure in the final configuration of phrases shows that Spanish speakers clearly adopt the Spanish intonation patterns of declarative sentences in their productions of IEAS. The typical falling pattern (L* L%) found in the
The tonal configurations found in the
Thus, our results suggest that the occurrence of a particular tonal configuration may not be a sufficient factor to detect the production of fake English-accented Spanish since the same tonal patterns have been observed in the productions of the two groups of speakers (in IEAS and AEAS). However, a more reliable cue seems to be found in the frequency of occurrence of a particular tonal structure, as in the case of the fall-rise pattern (L+H* LH%). Even though both English and Spanish speakers use this contour at the end of the
Percentage of occurrence of the tonal configurations in the
Whereas the productions in L1 Spanish and in IEAS show a predominant use of the rising configuration (L+H* H%), those of AEAS indicate that both a rise pattern and a fall-rise pattern (L+H* LH%) are equally used in the
The other pattern in the
This paper has examined the phrasing and the tonal structure of the final configurations of declarative sentences in a read corpus produced by a group of Spanish speakers in L1 Spanish and in imitated English-accented Spanish (IEAS), and by a group of English speakers in authentic English-accented Spanish (AEAS). The results have shown that, as far as phrasing is concerned, there are differences in the presence of prosodic breaks between L1 Spanish and imitated English-accented Spanish. However, a comparable number of intonation phrases were produced in imitated and in real English-accented Spanish, indicating that phrasing does not seem to be a reliable cue to detect a fake accent. The type of final tonal configuration is not a definite cue either since both languages show similar
Authors acknowledge funding from the Spanish Ministerio de Economía, Industria y Competitividad: EMULANDO (FFI2014-59848-C2-1-P).
The jeweler Federico Vanero│has been sentenced by the Court of Santander│to eight months in prison│and a fine of fifty thousand pesetas│for the crime of buying stolen goods.│The oral hearing was held last Wednesday│and, during the session,│one of the prosecutors, Carlos Valcárcel,│requested for the jeweler three years of simple imprisonment│and a fine of fifty thousand pesetas.│Due to Vanero’s report two and a half years ago,│the existence of a suspicious police mafia in Spain was disclosed,│ part of which was involved in the so-called│ “caso el Nani”.