Analysing prosody and meaning in Spanish : implications for the teaching of English as a foreign language

The aim of this work is to identify the difficulties that Spanish learners of English may have when confronted with the L2 intonation. The different tone choices made by native speakers of Spanish in North Patagonia, Argentina, are analysed in the light of the Autosegmental-Metrical framework (LADD, 1996) and the Relevance Theory (SPERBER and WILSON, 1995, 2004). The findings are compared with the tone choices produced by native speakers of English. The main results indicate that the same procedural instruction is manifested in both languages by means of L* L% and H+L* L%. However, Spanish favours L* whereas English prefers H+L*.


Introduction
This work emerges from a research project on prosody and relevance in English and Spanish that is being carried out at Facultad de Lenguas, Universidad Nacional del Comahue in Argentina.

The overall aim of this research paper is to
LGHQWLI\ WKH SRWHQWLDO GLI¿FXOWLHV WKDW 6SDQLVK OHDUQHUV of English may have when being confronted with the English intonation system.The tone choices 1 made by native speakers of Spanish at Comahue region, North Patagonia, Argentina, will be examined with the purpose of paving the way for the improvement in teaching and learning intonation.The present study intends to analyse the prosody of declaratives from a pragmatic and a phonological point of view.A theoretical background will be outlined, particularly SPERBER and WILSON's Relevance Theory (SPERBER andWILSON, 1995, 2004), and its application to the study of prosody (HOUSE, 1990;WILSON and WHARTON, 2006).The gathered data will be analysed in the light of the Autosegmental-0HWULFDO IUDPHZRUN /$'' DQG WKH ¿QGLQJV RQ Spanish as an L1 will be compared with the tone choices made by native speakers of English from the South East of England.

Relevance Theory
Pragmatic theory is concerned with the interpretation of the speaker's meaning expressed through utterances.In Gricean pragmatics (cf.GRICE, 1991), it is considered an inferential process of hypothesis formation and evaluation.The possible hypotheses are evaluated in the light of certain general principles of communication which speakers are expected to obey. 1

Within Relevance
Theory the term choice is understood as an element of procedural encoding.A speaker 'chooses' certain intonational contours unconsciously to cause an effect on the listener and guide him/her in the interpretation of utterances.
Relevance Theory, a neo-Gricean cognitive pragmatic theory, proposes that our cognitive life and consequently our linguistic utterances are guided by the search for relevance.This means that the interpretation of utterances is governed by two general principles: a communicative and a cognitive one.
Communication, according to this theory, FRPELQHV WKH WUDGLWLRQDO SURFHVV RI FRGL¿FDWLRQ ZLWK processes of inference.This combination suggests the existence of a speaker that wants his utterances to be understood, and a listener that provides a context for the interpretation of those utterances.The so called ostensive-inferential process holds that the listener infers the speaker's intention in order to process the LQIRUPDWLRQ LQ DQ HI¿FLHQW ZD\ 7KLV LV DFKLHYHG LI WKH speaker directs the listener's attention to what the former considers relevant.
The general cognitive principle states that we pay attention to what is relevant.Hearers are equipped with a very general cognitive criterion to evaluate interpretations, which is based on the fundamental assumption mentioned above.Every utterance has a variety of possible interpretations, but not all of them occur to the hearer simultaneously: some demand more HIIRUW WR ZRUN RXW 7KHUHIRUH UHOHYDQFH LV GH¿QHG LQ WHUPV of contextual effects and processing effort.Contextual effects are achieved when newly-presented information interacts with a context of existing assumptions in one of three ways: by strengthening existing assumptions; by contradicting and eliminating existing assumptions; or by combining with existing assumptions to yield a contextual implication, a synthesis of old and new information.
An utterance which provides great cognitive effects requires little effort on the part of the listener.The relevance results from the interaction between the new information that the utterance conveys and the one that is known or derived from the context: the lower the effort to process the message, the greater the relevance.On the contrary, an utterance that requires greater effort to be interpreted produces fewer cognitive effects, thus, the less relevant it will seem.

Autosegmental-Metrical Model
The autosegmental-metrical model (AM) is based on PIERREHUMBERT's (1980) doctoral thesis.It was originally meant to describe the English language; however, it has been revised and employed in other languages, including Spanish.
In the AM model, intonation is conceived of as an independent level from the segments, and contours are broken down into contrastive components.It basically distinguishes two types of prosodic domains: the intonational phrase and the intermediate phrase.
There are two phonological units: pitch accents, which are anchored with prominent syllables, and boundary tones, which are associated with the edges of prosodic domains.This approach proposes two tone levels: H (High) and L (Low), which are linked to pitch accents and edge tones.The (*) symbol indicates the association of the prosodic event with the prominent syllable.The (-) symbol corresponds to the intermediate phrase, and the (%) relates the tone to the edge of the intonational phrase.A third level, M (Mid), for the boundary tone has been described in the analysis for the Spanish language.
Based on the AM model prosody has been analysed by means of the Tone and Break Indices (ToBI) transcription system (BECKMAN, 2002).This notational tool examines strings of phonological events and consists of different tiers to annotate tones, words, syllables, break indexes and any other extra linguistic aspect.ToBI has been implemented for the study of several languages, including Spanish (Sp-ToBI) (BECKMAN ET AL., 2002;HUALDE, 2003;SOSA, 2003;ESTEBAS VILAPLANA and PRIETO, 2008).The Sp-ToBI transcription system has resulted from agreement reached by specialists working in the autosegmental-metrical framework (BECKMAN DQG (/$0 7KH ODWHVW PRGL¿FDWLRQV WR 6S7R%, have been introduced by ESTEBAS and PRIETO ( 2008) and the Sp-ToBI conventions for Argentinian Spanish have been put forward by GABRIEL et al. (2010).

Methodology Corpus
The data collection was organised in two stages.,Q WKH ¿UVW SODFH D QDUUDWLYH SLHFH ZDV FKRVHQ WR EH UHDG aloud by the participants as this genre is characterised by containing mainly declaratives.The short story selected was "A Imagen y Semejanza" by BENEDETTI (1968).This text was shortened as a fragment of the tale would provide D VXI¿FLHQW VDPSOH RI WKH ODQJXDJH IRU LWV ODWHU DQDO\VLV Then, the resulting version was translated into English.Attention was placed on declaratives which coincided with full stops in the short story.There is a clear correlate between punctuation and intonation (CRUTTENDEN, 1997).One of the functions of punctuation is to delimit the beginning and end of a tone unit, not necessarily by means of a pause.Accordingly, the readers in this work segmented their speech in a similar fashion due to the LQÀXHQFH RI SXQFWXDWLRQ +RZHYHU WKLV GLYLVLRQ RI RUDO language does not imply the same choice of tones.This is why instances of similar segmentation but different tone choices were encountered.
The second stage consisted of gathering the Spanish and English data.As regards Spanish, the subjects of study were two male and two female university students aged 21 who had just started the teacher training and translation courses with some basic previous knowledge of English.All of them were native speakers of the variety of Spanish spoken in the Comahue region, North Patagonia; they were all born and raised in this area and their parents had lived there for at least 20 years.As to the English data, the subjects of study were four male native speakers from the South East of England in their late 20s.All the participants were asked to read the fragment out loud obtaining one recording per subject.Once the interviews were carried out, the linguistic data were transcribed.

Analysis and transcription tools
One of the tools used for analysing speech and transcribing data was the computer programme Praat 5.2.25 (BOERSMA and WEENINK, 2010).This software shows the waveform together with the F0 in parallel and coordinated windows and allows researchers to follow the conventions for ToBI, i.e. the notation may be divided into different tiers such as words, tones, break indexes, syllables and the like.ToBI was not developed for interlinguistic analysis and this means that English ToBI and Sp-ToBI are different.In order to be able to compare the preliminary observations obtained from both languages only one transcription system was necessary.In keeping with the aims of the research project, Sp-ToBI was favoured.
The other tool used for analysing speech was the auditory perception of the researchers.It is ZRUWK PHQWLRQLQJ WKDW WKH ) QRW DOZD\V UHÀHFWV WRQH movements exactly.In the case of low-pitched voices and voiceless sounds, the F0 track is broken and, as a consequence, errors in the representation of it may occur.This is why the perceptual analysis of the researchers is combined with computational tools to optimise the analysis of the data.

Phonetic and phonological analysis
During the process of analysis, the first step consisted of exploring the readings by means of auditory perception.Each of the recordings was divided into intonational phrases and then prominent syllables and nuclear accents were located.One aspect taken into consideration when selecting the intonational phrases to be analysed was the location of the nuclear accent.In order to make a clear and separate analysis of the nuclear pitch accents and boundary tones the focus was placed on intonational phrases that did not bear the nuclear accent on the last syllable.The next step comprised the segmentation of the audio files so as to proceed with the acoustic analysis.The preliminary auditory testing was checked by means of the software Praat: nuclear accents and boundary tones were identified.Then, the performance of the four Spanish participants was contrasted with one another.The same procedure was followed for the English subjects.Finally, the information about both languages was compared.

Nuclear accents
The nuclear tones used by the Spanish subjects were of three types: L*, H+L* and L+H*+L.In the case RI (QJOLVK WZR QXFOHDU WRQH W\SHV ZHUH LGHQWL¿HG / and H+L*.A schematic representation, together with a brief description of each pitch accent, is shown in table 1 below.The most frequent nuclear tone choice made by the Spanish speakers was L*, standing for a 59 per cent of the corpus, followed by H+L* as a second choice with a 35 per cent of occurrence.In opposition, the English subjects preferred H+L* as the nuclear pitch accent in a noticeable 71 per cent, leaving a low 29 per cent for L*.The tritonal nuclear accent L+H*+L was only found in the Spanish data in a six per cent.This nuclear tone is typically used in Argentinian Spanish intonation in utterances with a contrastive or emphatic reading (GABRIEL ET AL., 2010).The following pie charts show the frequency of occurrence of each nuclear tone in both languages.

Boundary tones
Concerning boundary tones, the analysis revealed that both languages make use of L%.The Spanish corpus also presented the occurrence of M%.Table 2 below summarises these edge tones.In both languages the tendency was to use the low level tone target: L%.In Spanish, this edge tone was found in 82 per cent of all instances, whereas in English, L% appeared in 100 per cent of the analysed GHFODUDWLYHV 7KH 0 ZDV RQO\ LGHQWL¿HG LQ WKH 6SDQLVK data, representing an 18 per cent of the corpus.Figures 3 and 4 show the frequency of occurrence of each boundary tone in both languages.

Spanish -Boundary tones English -Boundary tones
With regard to the percentages of nuclear FRQ¿JXUDWLRQV SHU FHQW RI WKH GHFODUDWLYHV LQ 6SDQLVK were produced with a low accent L* followed by a low boundary tone, L%.This percentage can be directly compared with the one obtained in the English data, as only 29 per cent of all the cases were uttered with WKLV QXFOHDU FRQ¿JXUDWLRQ /LNHZLVH PRVW GHFODUDWLYHV 71 per cent, were produced with a H+L* nuclear pitch DFFHQW DQG D / SKUDVH ERXQGDU\ 7KLV FRQ¿JXUDWLRQ displays a falling movement of the F0 track down to the pitch baseline () and it appeared in the Spanish data as well, but in a lower percentage, 26 per cent.The Spanish corpus also revealed that 12 percent of the prosodic units combined L* with a mid boundary tone, M%, and 9 per cent of the cases showed a H+L* pitch accent followed E\ D PLG HGJH WRQH 0 7KLV FRQ¿JXUDWLRQ GLVSOD\V DQ ascent of the F0 in the post tonic material up to a mid level point ( ).According to the data, the tritonal tone L+H*+L followed by a low boundary tone, L%, was the least used by the Spanish speakers as just 6 per cent of WKH GHFODUDWLYHV ZHUH SURGXFHG ZLWK WKLV FRQ¿JXUDWLRQ )LJXUHV DQG UHSUHVHQW DOO WKH WRQDO FRQ¿JXUDWLRQV detailed above:

Pragmatic analysis
Intonation works as a procedural indicator of the relative relevance in utterances.By means of different YRLFH LQÀHFWLRQV WKH VSHDNHU JXLGHV WKH OLVWHQHU WRZDUGV the intended message, so that the hearer makes as little effort as possible when processing oral input.A previous study on Spanish spoken in the Comahue region, North Patagonia, (ARANA ET AL., 2010) showed WKDW LQWRQDWLRQ DQG PRVW VSHFL¿FDOO\ ERXQGDU\ WRQHV HQFRGHV VSHFL¿F LQVWUXFWLRQV DV WR KRZ HDFK XWWHUDQFH should be interpreted: 6XVSHQGHG WRQDO FRQ¿JXUDWLRQV J) such as H+L* M%, L+H* M% or L* M%, indicate continuity: they tell the hearer that this intonational unit is part of a bigger structure.In spite of being complete, grammatically and semantically speaking, the WRQDO FRQ¿JXUDWLRQ VKRZV LQFRPSOHWHQHVV RU QRQ¿QDOLW\ DQG WKXV WKH KHDUHU OHDUQV WKDW WKH relevance of the utterance is to be found in what follows.
)DOOLQJ WRQDO FRQ¿JXUDWLRQV ) of the type L* L% DQG +/ / VLJQDO ¿QDOLW\ WKH\ VSHFLI\ WKDW WKH sense of incompletion achieved by means of the previous utterances is about to end.The relevance of this utterance depends on what is achieved by the phrase itself; it is part of the directly relevant information.
7KHVH ¿QGLQJV ZHUH WKH VWDUWLQJ SRLQW RI WKLV work and the scope of analysis was broadened by also examining English data.Attention was directed towards falling contours in declaratives at the end of intonational phrases in both languages, so as to observe possible differences between Spanish and English.
The 21% of cases of declaratives at the end of LQWRQDWLRQDO SKUDVHV ZLWK D ULVLQJ QXFOHDU FRQ¿JXUDWLRQ found in the Spanish data, L* M% or H+L* M%, could be interpreted as uncertainty statements (GABRIEL ET AL., 2010).However, they will be not examined pragmatically in this work because they deserve a detailed analysis in a further study.
(2) K_O_1 makes up the context for processing information, as in the Spanish data.Accordingly, it invites the hearer to postpone evaluating the relevance of the utterance until a falling contour follows.The last intonational phrase "and couldn't keep up with his mates" shows the falling tonal FRQ¿JXUDWLRQ +/ / ).By means of this contour, the speaker indicates that the information of the present tone unit and the previous one should be processed.A falling contour shows conclusiveness: the speaker has offered the directly relevant information.
It can be observed that the boundary tone is the same as the one in the Spanish example above, but the nuclear accent is different: H+L*.It seems that the procedural instruction encoded in the falling contour is the same in both languages, irrespective of the nuclear tone choice.The procedural instructions for both languages could be summarised as follows: a) in the presence of a suspended contour (J ), the listener should postpone the evaluation of the global relevance of this utterance until a L% tone is reached; b) when there is a falling contour () the listener should evaluate the global relevance of the previous tone unit(s) in the light of the content of the this particular intonational phrase.Through the tone choices the speaker makes, the KHDUHU LV KHOSHG WR ¿QG WKH PRVW UHOHYDQW LQWHUSUHWDWLRQ of the utterances.
The data analysed in the light of Relevance Theory give evidence of procedural encoding.The tone XQLWV WKDW VKRZ D VXVSHQGHG WRQDO FRQ¿JXUDWLRQ FUHDWH great cognitive effects.This optimises the interpreting process and when the listener reaches the tone unit with a falling contour, little effort is necessary: the lower the effort to process the message, the greater the relevance.This interpretation procedure would require some extra effort on the part of the listener if each tone unit was processed separately.

Discussion
Having analysed and contrasted both languages, differences have been found as to the preferred nuclear tone choices, but not as to the selection of the boundary tone.It can be clearly observed that while Spanish prefers L* L% when reaching a full stop, English, on the other hand, chooses H+L* L%, in the examined corpus.This difference does not seem to have any pragmatic implications, although further study is necessary.
The difference in the choice of nuclear tones might be related to a phonetic phenomenon.The abrupt fall in the F0 produced by the bitonal H+L* pitch accent followed by L% is preferred by English speakers over the less notorious fall manifested in a L* L% combination in Spanish.It could be argued that the analysed Spanish speakers did not produce an abrupt fall of the F0 given that Spanish in North Patagonia seems to display a narrower pitch range, which coincides with the descriptions of other varieties of Spanish.This pitch range difference might be associated with the rhythmical structure of each language.Even though, more empirical evidence is needed.
The above-mentioned remarks pave the way IRU SUHGLFWLQJ RXU VWXGHQWV ¶ GLI¿FXOWLHV LQ OHDUQLQJ the English intonation system.The phonetic characterisations of both languages might suggest that these students need to broaden their pitch range when performing English tunes.
A future step in the analysis would consist of asking the students to read the tale in English and see LI 6SDQLVK LQWRQDWLRQ LQÀXHQFHV WKH WDUJHW ODQJXDJH If so, it would be interesting to examine which aspects of intonation are the ones which learners apply when reading English so as to design a course of action to help solve these problems.

Conclusion
Intonation plays a crucial role when speakers instruct the hearers how to process the information.Relevance Theory seems to give a suitable framework for DFFRXQWLQJ IRU WKH PHDQLQJ RI GLIIHUHQW YRLFH LQÀHFWLRQV 7KH IDOOLQJ QXFOHDU FRQ¿JXUDWLRQ LQVWUXFWV WKH KHDUHU WR process the information presented as directly relevant.
In this work, the pragmatic information conveyed in Spanish and English was substantially the same, i.e. the hearer was instructed to process the utterances as foreground information at the moment the pitch movement was occurring.However, the differences lied on the phonetic and phonological manifestation of the analysed data.Spanish speakers told the hearer to process their message by means of L* L%, H+L* L%, and L+H*+L L%, whereas English speakers preferred H+L* L% and L* L% in order to give the same procedural instruction.
This study on pragmatic values and phonology has thrown new light on cross-linguistic similarities and GLIIHUHQFHV 7KHVH SUHOLPLQDU\ ¿QGLQJV SURYLGH XV ZLWK an insight into what should be done in order to help our students learn the English intonation system.

Table 1 :
Inventory of nuclear tones found in declarative sentences in the studied varieties of Spanish and English.

Table 2 :
Inventory of boundary tones found in declarative sentences in the studied varieties of Spanish and English.