Clinical Linguistics & Phonetics, February 2011; 25(2): 85–120 Perception of foreign accent syndrome speech and its relation to segmental characteristics Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. JANA DANKOVIČOVÁ1 & CLAIRE HUNT2 1 University College London, London, UK, and 2Adult Speech & Language Therapy Department, North West London Hospitals NHS Trust, London, UK (Received 1 April 2010; Accepted 31 July 2010) Abstract Foreign accent syndrome (FAS) is an acquired neurogenic disorder characterized by altered speech that sounds foreign-accented. This study presents a British subject perceived to speak with an Italian (or Greek) accent after a brainstem (pontine) stroke. Native English listeners rated the strength of foreign accent and impairment they perceived in speech of the FAS subject, alongside that of two native English speakers and Italian, Greek, and French L2 speakers acting as controls. The FAS subject was perceived to be as foreignsounding as the L2 control speakers, but was also perceived as mildly impaired. The FAS subject’s own perception of accents was also explored and it was found that his ability to distinguish presence and absence of accent does not seem to be affected. The relationship between listeners’ perceptions and features of the FAS speech is explored via correlational statistics and qualitative analysis. Impressionistic phonetic analysis, supplemented by acoustic analysis, confirmed a number of features consistent with a typical Italian (and also Greek) accent and the Italian and Greek L2 speakers. A pre-stroke and a post-stroke sample from the FAS subject were compared and the nature of post-stroke changes in segmental realizations is discussed. Keywords: Foreign accent syndrome, perception of foreign accent, perception of impairment, motor speech disorders, apraxia of speech Introduction Foreign accent syndrome (FAS) is the term used to describe a relatively rare disorder, in which acquired brain damage results in apparently foreign-sounding speech. Compared to patients with other acquired neurogenic speech disorders (e.g. apraxia, dysarthria), listeners find FAS speech intelligible and broadly fluent, but characterized by a non-native accent. Sometimes the accent resolves after weeks or months; sometimes it persists for years (Berthier, Rue, Massone, Starkstein, & Leiguarda, 1991; Haley, 2009). The syndrome emerges most frequently after stroke (Moen, 2000), but also after traumatic brain injury (Monrad-Krohn, 1947; Moonis et al., 1996; Carbary, Patterson, & Snyder, 2000; González-Alvarez, Parcet-Ibars, Avila, & Geffner-Sclarsky, 2003), and the progressive neurodegenerative diseases multiple sclerosis (Villaverde-González, Fernández-Villalba, Correspondence: Jana Dankovičová, Department of Speech, Hearing and Phonetic Sciences, University College London, Chandler House, 2 Wakefield Street, London WC1N 1PF, UK. Tel: þ44 20 7679 451. E-mail: j.dankovicova@ucl.ac.uk ISSN 0269-9206 print/ISSN 1464-5076 online © 2011 Informa UK Ltd. DOI: 10.3109/02699206.2010.513027 Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 86 J. Dankovičová & C. Hunt Moreno-Escribano, Alías-Linares, & García-Santos, 2003; Bakker, Apeldoorn, & Metz, 2004) and dementia (Luzzi et al., 2008). Altered accents are also reported in psychiatric disorders such as schizophrenia and conversion disorder (Reeves & Norton, 2001; Van Borsel, Janssens, & Santens, 2005; Verhoeven, Mariën, Engelborghs, D’Haenen, & De Deyn, 2005; Reeves, Burke, & Parker, 2007; Tsuruga, Kobayashi, Hirai, & Kato, 2008). However, some authors (e.g. Blumstein & Kurowski, 2006) argue that neurogenic FAS should be defined to exclude any psychogenic aetiology. Reversion to a previously spoken accent or dialect (Roth, Fink, Cherney, & Hall, 1997), and previous exposure to another language or accent (Seliger, Abrams, & Horton, 1992; Munson & Heilman, 2005) are also proposed exclusion criteria. The phonetic characteristics of FAS speech are complex and variable, affecting both segmental and prosodic features. Among the segmental characteristics, the vowel changes seem to be more dominant than consonantal changes. Both vowel quality and quantity may be affected, although the actual pattern of changes seems to differ widely across the reported FAS cases. Some FAS subjects display vowel tensing (e.g. Graff-Radford, Cooper, Colsher, & Damasio, 1986; Blumstein, Alexander, Ryalls, & Dworetzky, 1987), and/or lengthening (e.g. Graff-Radford et al., 1986; Miller, Lowit, & O’Sullivan, 2006), but some show the opposite tendency – vowel laxing, centralization and shortening (e.g. Ingram, McCormack, & Kennedy, 1992). Yet other FAS cases present with a mixed pattern; for example, with all vowels fronted, but some of these vowels more open and some closer (Dankovičová et al., 2001). A comprehensive survey of findings regarding the vowel quality/quantity can be found in Katz, Garst, & Levitt (2008). Among consonantal changes the most frequently reported are voicing errors and changes in manner of articulation. Again, the pattern of changes differs from case to case, with some studies reporting a pattern of Voice onset time (VOT) that is the reverse of what is expected for voiced and voiceless consonants (e.g. Laures-Gore, Henson, Weismer, & Rambow, 2006), some reporting devoicing of word-final plosives (e.g. Gurd, Bessell, Bladon, & Bamford, 1988; Ingram et al., 1992), regressive voicing assimilation (e.g. Blumstein et al., 1987) or reduced aspiration of voiceless plosives (e.g. Dankovičová et al., 2001). However, some FAS subjects seem to have VOT unaffected (e.g. Katz et al., 2008). In terms of prosodic characteristics, changes in speech rate, intonation and speech rhythm have frequently been reported. While a number of studies mention slower speech rate (e.g. Berthier et al., 1991; Kurowski, Blumstein, & Alexander, 1996; Coelho & Robb, 2001), some patients have been reported to have normal articulation rate or a rate similar to that before their stroke (e.g. Dankovičová et al., 2001). Unusual intonation patterns were observed in some FAS cases; for example, restricted F0 range and rising sentence-final pitch where falling pitch would be expected (e.g. Graff-Radford et al., 1986). The prosodic feature most frequently reported in FAS cases is a change in speech rhythm, specifically a tendency towards ‘syllabletiming’ in cases where ‘stress-timing’ would be expected (e.g. Blumstein et al., 1987; Gurd, Coleman, Costello, & Marshall, 2001; Scott, Clegg, Rudge, & Burgess, 2006). The perception of ‘syllable-timed’ rhythm can arise due to a number of characteristics; apart from unusually equal syllable durations, for example, lack of vowel reduction (e.g. Whitaker, 1982) and occurrence of epenthetic vowels (e.g. Ardila, Rosselli, & Ardila, 1988; Ingram et al., 1992). The 50 or so reported cases span over a century, but leave unanswered questions as to the underlying neuroanatomical and physiological bases of FAS. An extensive list of cerebral areas are implicated (Scott et al., 2006), with most cases resulting from an anterior left hemisphere lesion (Blumstein & Kurowski, 2006). However, right hemisphere aetiology is also reported (Dankovičová et al., 2001; Miller et al., 2006). Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 87 There is similarly diverse and inconclusive evidence for the underlying physiological mechanism(s) responsible, with a number of working theories suggested. First, a proposed ‘tense speech posture’ (Graff-Radford et al., 1986) has been adopted as a possible explanation for some reported cases (Kurowski et al., 1996; Moen, 2000; Ryalls & Whiteside, 2006). The second proposal that FAS may be a specific, mild form of apraxia of speech (Aronson, 1990; Duffy, 1995) is supported by the apraxic-type segmental and prosodic features and high variability in some FAS cases (Miller et al., 2006; Varley, Whiteside, Hammill, & Cooper, 2006). However, other authors note important qualitative differences between apraxic and FAS articulation and intonation (Kurowski et al., 1996; Blumstein et al., 1987; Carbary et al., 2000; Blumstein & Kurowski, 2006; Katz et al., 2008). A third theory that a prosodic disorder underlies the distortions produced in FAS (Blumstein & Kurowski, 2006) is supported by recent evidence from Katz et al. (2008). Some authors emphasize that these working theories about the underlying physiological mechanism should be viewed as complementary, not mutually exclusive (Moen, 2000; Dankovičová et al., 2001; Katz et al., 2008). The symptomatology of FAS presents as many challenges as the question of its underlying bases. Although its ‘most critical’ feature (i.e. the perceived accent) is also the ‘most obvious’ (Blumstein & Kurowski, 2006: 347), the detail beyond that is far from obvious. The accents attributed to FAS speakers are numerous, and the attribution of accents by naïve listeners is usually only reported anecdotally rather than explored experimentally (but cf. Christoph, de Freitas, Dos Santos, Lima, Araújo, & Carota, 2004; Di Dio, Schulz, & Gurd, 2006). Some reports attribute a single, specific accent to individual FAS speakers; for example, an apparently Polish-accented Czech speaker (Pick, 1919), German-accented Norwegian speaker (Monrad-Krohn, 1947) and French-accented English speaker (Gurd et al., 1988). More frequently, however, FAS speakers are attributed with a more generic accent (e.g. Eastern European, Slavic), and/or a range of accents (Blumstein et al., 1987; Ardila et al., 1988; Berthier et al., 1991; Ingram et al., 1992; Kurowski et al., 1996; González-Alvarez et al., 2003; Di Dio et al., 2006; Miller et al., 2006; Scott et al., 2006). This variability has led to different theories to explain how listeners attribute accents to FAS speakers, summarized and tested by Di Dio et al. (2006) in a study of listeners’ identification of accent in FAS and non-native speakers. Listeners showed some consistency in their identification of specific accents for two FAS speakers, and a ‘familiarity effect’, or a tendency to identify accents within their experience. Di Dio et al. suggest that their preliminary findings provide the strongest support for the ‘linguistic relative view’ of Miller et al. (2006), according to which listeners focus on salient elements of the speech that match accents they have themselves experienced. This is contrasted with two earlier, alternative views: (i) the ‘generic view’, stating that FAS emerges as a ‘generic’ rather than specific accent (Kurowski et al., 1996) and implying that there would be low accuracy and low agreement in accent judgements between listeners; and (ii) the ‘perceptual view’, suggesting that the perceived accent is entirely ‘in the ear of the beholder’, and not inherent in the FAS speaker’s actual speech (Carbary et al., 2000; Scott et al., 2006). A further question in FAS symptomatology is why this speech sounds foreign but not pathological. One explanation proposed is that, compared to apraxic speech, FAS distortions are all potential natural language sounds (Blumstein et al., 1987; Moen, 2000; Gurd et al., 2001; Blumstein & Kurowski, 2006), including ‘inter-language’ forms which match distortions commonly produced by non-native speakers (Katz et al., 2008). However, Miller et al. (2006) argue that this explanation relies on a reductionist linear approach which is methodologically weak (Weismer & Liss, 1991). Instead, they test this assumption experimentally by exploring which of the many variables within FAS speech are salient to judgements of foreignness. They found vowel and stress changes and vowel epenthesis to be strongly related to perceived foreign accent. They hypothesize that, whilst some potential natural language distortions were also noted in a Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 88 J. Dankovičová & C. Hunt control apraxic speaker, listeners did not judge this speaker to be foreign, perhaps focusing on other features more salient to a judgement of disorder. Miller et al. note, however, that the study design may have influenced judgements as listeners rated speakers on a scale of foreignness only and had no option to rate speakers on a scale of disorder. One unexplored aspect of FAS symptomatology is the patient’s own perception of speech and, specifically, accent (Gurd & Coleman, 2006). Amongst the reported cases, a mixed picture emerges as to whether FAS patients perceive the changes in their own speech. Some studies report patients being unaware of the changes to their accent until alerted by friends and family (e.g. Miller et al., 2006), while other patients noticed the changes themselves (Coughlan, Lawson, & O’Neill, 2004; Scott et al., 2006). In the majority of reported cases, the patient’s phonological input skills are not tested specifically. Models of ‘genuine’ foreign accent (see Major, 2001, for a review) emphasize the link between perception and production of second language (L2) speech. When perceiving L2 speech, non-natives ‘filter’ it through their first language (L1) ‘sieve’ (Major, 2001). In their own production of L2 speech, this results in L1 and hybrid inter-language sound substitutions. FAS literature has focused on speech production. We do not know whether a perceptual deficit, as in ‘genuine’ foreign-accented speakers, may exist in FAS symptomatology; or, if so, whether and how this interacts with production. As mentioned above, very few studies have tested perception of accent in FAS speech formally. Furthermore, no study, as far as we know, has tested formally the claim that FAS speech sounds foreign to listeners, but not disordered. In the present study we examine the perception of both accent and impairment in a case of a subject who was diagnosed with FAS after a stroke. We also were interested to know if FAS speech sounded to English listeners foreign to the same degree as speech of genuinely foreign (L2) speakers of English, and to what extent listeners agreed about the type of accent in the FAS speaker. The perception of both foreignness and speech impairment was also examined in the FAS subject himself, to gain some insight into the perceptual abilities of a particular FAS speaker, as compared to listeners without any speech or perceptual deficits. This is an area not explored in the literature so far. Further we investigated to what extent phonetic features in FAS speech are similar to phonetic features of genuine L2 speakers of English. In this paper we concentrate on the segmental features only, as the prosodic characteristics of our FAS speaker will be addressed in a separate paper (in progress). In the present paper, the investigation was based on an impressionistic and acoustic analysis of the FAS speaker, control English speakers, and several genuine L2 speakers. Finally we discuss the relationship between the phonetic features in our FAS speech sample and listeners’ perceptual judgements of foreignness and impairment. Case report In October 2003, our FAS subject – right-handed, male, aged 56, ex-smoker, with a history of diabetes, hypertension and ulcerative colitis – experienced a mild stroke characterized by sudden-onset slurred speech, right-sided weakness and facial asymmetry. He presented to hospital, was diagnosed clinically from the above symptoms with a stroke, and was discharged home the same day. Structural imaging was performed some months later during out-patient follow-up with his neurologist, due to an interest in the nature of his speech symptoms. A CT scan 6 months post-onset reported normal. A subsequent MRI scan 16 months post-onset revealed a small wedge-shaped area of porencephaly along the anterior margin of the mid pons immediately to the left of the midline, corresponding to the position of the descending Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 89 cortical fibres. Medical records note that his right-sided facial asymmetry and hemiparesis improved within a few days but, as his speech became less slurred, an altered accent emerged. In early 2005, our FAS subject, then 57, became aware of symptoms of increased fatigue, reduced left-sided lower limb mobility, reduced left-sided upper limb sensation and mild swallowing difficulties. He discussed these symptoms with his neurologist during a routine out-patient appointment in late 2005, and was diagnosed clinically with a suspected second mild stroke (no structural imaging was conducted at this time). An MRI scan in March 2008, conducted for the purpose of a television documentary about FAS (Tait, 2008), showed posterior parietal atrophy superior to the occipital sulcus. The FAS subject was born and raised in Essex, in the South East of England. His father was born and brought up in the East End of London. His mother was born in Egypt to Greek parents and grew up in Egypt and Greece, and then moved to London with his father. As an adult, the FAS subject lived in London and Essex. The FAS subject describes his accent before the stroke as ‘Cockney’ and identifies himself strongly in relation to his Essex upbringing. He had not spent significant time abroad or learnt any foreign languages, except social phrases picked up on holidays. His mother spoke English, not Greek, as he grew up, and, according to him, developed a Cockney accent herself. In short, the FAS subject had no significant exposure to any languages or accents other than British English. He lives with his wife, also from Essex, and has three adult sons. After the first stroke he returned to his job working with adults with learning disabilities. He has not worked since the second stroke. Our FAS subject and his wife report no significant changes to his speech since the emergence of the ‘accent’ a few days after the first stroke. The second stroke is not reported to have had any further impact on his speech. He had no formal or instrumental assessment of his speech following his first or second stroke. He participated in two speech therapy sessions in August 2005, reportedly focused primarily on oromotor exercises, but further input was not indicated as his speech was fully functional and intelligible, despite the ‘accent’. The FAS subject participated in a neuropsychological assessment at the University of Oxford in December 2004 (14 months after his first stroke). Information from the researchers involved reports that he performed normally on all except two tests (J. Coleman, personal communication). In the Boston Diagnostic Aphasia Examination (Goodglass, 2000) Oral Expression test, he scored poorly (8/12) on non-verbal oral agility tasks, with particular difficulty touching his upper and lower teeth with his tongue tip. This compared with a good score (14/14) on the verbal agility tasks. On the Apraxia test he scored 9/10 on transitive limb tasks, compared with 10/10 on bucco-facial, intransitive limb, and whole body tasks. Method Perceptual tests Pilot experiment. In order to choose control speakers with appropriate accents, we conducted an informal pilot experiment, in which 17 British English listeners (of whom two had a speech and language therapy background) were played four short clips of the FAS subject’s conversational speech (taken from a radio interview, see below), and asked simply to ‘describe the speaker’. The majority of listeners (14/17) included ‘Italian’ in their descriptions, while 4/17 noted ‘Greek’ in their descriptions. Interestingly, elements of ‘disorder’ were also mentioned in 8/17 descriptions (slurred speech, hesitancies, repetitions, stammer, slow speech). These were always mentioned alongside and secondary to the foreign accent references. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 90 J. Dankovičová & C. Hunt Control speakers. Five control speakers were selected for the study. All were male, aged 55–60 to match the FAS subject’s age (58 years) when the material used in this study was recorded, with no hearing, speech or language impairment, and no phonetics or pronunciation training. Three were genuine foreign speakers – an Italian, a Greek and a Frenchman (they are labelled henceforth as Italian, Greek and French second language [L2] speakers). Two control speakers were native English speakers, both from Essex, in order to match the FAS subject’s likely pre-stroke accent. The selection of an Italian and a Greek speaker was guided by the results of the pilot study, and also by the reports from the FAS subject himself that listeners usually attribute a foreign accent to him, generally Italian or Greek. (He also mentioned that native Italian listeners compared his accent to that of a speaker from southern Italy.) A French L2 speaker was included to widen the range of accents. Initially eight L2 speakers of English – two native Italian, three Greek, three French speakers – were recorded. One speaker per language background was then selected, taking into account difficulties reading English, influence of other languages and time spent in other countries, etc. The aim of the selection was to make as many factors as possible similar to the FAS subject. The speakers selected had lived only in their native countries and London. The Italian L2 speaker was from southern Italy – to match the specific accent attributed to the FAS subject by some listeners – and had lived in London for 34 years. The Greek L2 speaker was from Cyprus (2/4 references to ‘Greek’ in the pilot had specified ‘Greek Cypriot’) and had lived in London for 47 years. The French L2 speaker was from northeast France and had lived in London for 22 years. The two native English speakers were born in the same Essex town as the FAS subject, and had always lived in Essex. For the purposes of cross-speaker comparisons in the acoustic analysis, we also recorded a speaker with an RP accent (‘received pronunciation’), also described as Standard Southern British English pronunciation. The RP speaker was a 56-year-old male. Material. The FAS speech stimuli were taken from a radio interview with the FAS subject in 2005 (Robinson, 2005), 16 months after his first stroke and 3 months after his suspected second stroke. Twenty-one phrases and 20 single words were selected from the FAS subject’s conversational speech using a number of exclusion and inclusion criteria. Exclusion criteria ensured that listeners would not be influenced by autobiographical information (e.g. references to stroke, accents, placenames), or use of syntactic deviations, slang, or idioms. Inclusion criteria ensured that the stimuli included coverage of segmental features that are frequently mispronounced by Italian, Greek and French speakers of English (for Italians: Duguid, 2001; Pulcini, 2002; Atwell, Howarth, & Souter, 2003; Browning, 2004; Rogers & d’Arcangeli, 2004; for Greeks: Efstathiadis, 1974; Papaefthymiou-Lytra, 2001; Stathi, 2002; for French: Tranel, 1987; Dalby & Kewley-Port, 1999; Walter, 2001; Humbley, 2002). One phrase where the FAS subject impersonated a native Italian (‘What can I do for you?’) was included in order to explore phonetic features and listeners’ perceptions when he was intentionally adopting a foreign accent. Recordings were made of the same stimuli from the two native English (Essex) control speakers and from the three L2 speakers. Together with the FAS speaker, this yielded 246 stimuli in total. The single word stimuli were spoken within phrases (as originally spoken by the FAS subject) to preserve contextual influences. Speakers were encouraged to read the phrases naturally, as if they were part of conversational speech. They were not told the purpose of the study. Recordings were made using an Olympus WS-300M digital voice recorder with a Vivanco EM35 stereo microphone in a quiet room in speakers’ homes or workplaces. The stimuli were extracted from these recordings and from the FAS subject’s Perception of foreign accent syndrome speech 91 Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. radio interview using Adobe Audition computer software. Amplitudes were adjusted to minimize variation across the stimuli. The RP speaker recorded the same 21 phrases as the other speakers, though these were not part of the perceptual test and served only as the comparison baseline in the acoustic analysis. We were able to obtain from our FAS subject a sample of his pre-stroke speech. It was a very short clip (eight phrases) in which he was presenting a community cable TV show. Listeners. In the original study, 40 listeners were recruited among students at University College London: 10 native British English, 10 native French, 10 native Italian and 10 native Greek speakers. Each language group had an equal number of male and female listeners. They were all aged 20–35, with no hearing, speech or language impairment and no phonetics or pronunciation training. For the purposes of this study we report only results from the perceptual test by the native English listeners. These listeners had no more than very basic knowledge of French, Greek or Italian. Three of them were born and bred in Essex. Procedure. Listeners heard the stimuli via a Soundplayer program (Coleman, 2008) over high quality headphones, in a quiet room, either alone or alongside another listener also using headphones. The 246 test stimuli were randomized across speakers, words, and phrases, and divided into nine blocks of 24 stimuli, and a final tenth block of 30 stimuli, with a pause of 6 seconds between each stimulus. The listeners had 20-second breaks between each block and a 5-minute break between blocks 5 and 6. These timings were established by trialling a variety of between-stimuli and between-block durations with four listeners not used in the main experiment. Each stimulus was heard only once. Each target word or phrase was written on the response sheet and the listeners were asked to rate each stimulus on a scale of foreignness and also a scale of impediment (impairment). The scale for foreignness was 1 ¼ definitely English; 2 ¼ probably English; 3 ¼ probably foreign; and 4 ¼ definitely foreign; the scale for impediment was 1 ¼ no impediment; 2 ¼ slight impediment; 3 ¼ moderate impediment; and 4 ¼ severe impediment. When scoring 3 or 4 for foreignness, listeners could optionally state any accent recognized. The listeners were unpaid and blind to the purpose of the study, the details of the speakers and the total number of speakers. FAS subject’s perceptual test. The FAS subject’s perceptual task was a reduced version of the full test, and the procedure was modified to take into account likelihood of the test being more tiring for him than for other listeners. The test consisted of 125 stimuli (10 words and 15 phrases from five speakers: himself, one of the two Essex speakers, and the three L2 speakers). The selection was guided by the results of the main perceptual test in the following way. The Essex speaker chosen was the one rated most accurately by listeners in the main test as a nondisordered, native English speaker. The 10 words were chosen by discounting all one-syllable and some two-syllable words since listeners had found these difficult to recognize and rate (see below). The 15 phrases were chosen to represent a range of mean scores for both impairment and foreignness. Our FAS subject was tested in his home. He was given a large-print flashcard with the target word or phrase in advance of hearing the stimulus over headphones. He rated each stimulus on the same foreignness and impediment scales as the other listeners; however, he gave his responses verbally while the experimenter documented them. He could take as much time 92 J. Dankovičová & C. Hunt (within reason) as required and could ask for a stimulus to be repeated. Stimuli were presented in three blocks: (i) 50 phrases; (ii) 50 words; (iii) 25 phrases. Within each block, stimuli were randomized. He was allowed breaks flexibly throughout the testing. The FAS subject was aware that the test related to research into Foreign Accent Syndrome, but was not told the source of the stimuli, the number of speakers or any information about them or that his own speech was included. It was decided that if he recognized himself speaking and queried if it was him, confirmation would be given to him on an item-by-item basis. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Analysis The analysis consisted of three parts – an impressionistic phonetic analysis; an acoustic analysis; and examination of the relationship between classes of deviation from an RP accent and perceptual judgements of foreignness and impairment. The results reported here concern only phrases (see Appendix 1), as the single word stimuli turned out to be problematic for the perceptual test. They had to be extracted from a context (see above) and this distorted the word boundaries in some cases, making it harder to understand, especially if the word was a short one. The perceptual ratings were therefore more difficult to make and possibly thus less reliable. Results Perceptual test The response rate in the perceptual test was extremely high. For the foreignness ratings 95% of utterances were rated by all 10 English listeners and the maximum number of missing ratings per utterance was one (i.e. one out of 10 listeners not giving a response). For the impediment ratings 91% of utterances had a complete set of ratings across the listeners. One listener did not give any ratings to 5% of the total number of utterances (6/126), but this was exceptional, as the other listeners had no missing ratings or one or two at most. The maximum missing ratings per utterance was again one. Inter-listener variability. To examine the degree of agreement in ratings amongst listeners across all stimuli, we used Kendall’s coefficient of concordance W. This coefficient indicates the degree of association of ordinal assessments made by multiple appraisers when evaluating the same samples. Values of W range from 0–1; the higher the value of W, the stronger the association. Although the agreement between the listeners was significantly different from zero for both foreignness ratings and impediment ratings (p < 0.001), it was not particularly high (W ¼ 0.151 and W ¼ 0.261, respectively), thus revealing considerable variation among the listeners, especially for the foreignness ratings. Judgements of foreignness. Figure 1 shows the mean foreignness ratings across 10 English listeners for the individual speakers. The boxplots display the minimum and maximum (vertical lines), first and third quartiles (edges of boxes), median (horizontal line in the box) and outliers (asterisks and circles). The graph shows that while the two control Essex speakers were judged as definitely English, the FAS subject was rated as foreign to a similar extent as the genuine L2 speakers. The mean rating for FAS subject appears to be even higher than for the Greek L2 speaker (who had the widest distribution of ratings among the speakers). Perception of foreign accent syndrome speech 4.0 Mean foreignness rating Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 3.5 93 Foreign accent scale: 1 = definitely English 2 = probably English 3 = probably foreign 4 = definitely foreign 3.0 2.5 2.0 1.5 1.0 Essex1 Essex2 FAS French Speaker Greek Italian Figure 1. Mean foreignness ratings across all speakers. An analysis of variance (ANOVA) revealed a highly significant effect of speaker [F(5,120) ¼ 545.1, p < 0.001], but the post-hoc Tukey test (p < 0.05) indicated that not all speakers were significantly different from each other. The FAS subject was rated significantly more foreignsounding than the control speakers, and also than the Greek L2 speaker. However, he was significantly less foreign-sounding than the Italian L2 speaker. He was not significantly different from the French L2 speaker. The Greek L2 speaker was rated significantly more foreign than the control speakers but significantly less foreign than all the other speakers. The most foreign sounding speakers were the French and Italian L2 speakers (not significantly different from each other). As stated above, listeners had an option of stating a specific foreign accent when rating a stimulus as probably or definitely foreign. Of those who gave the accent information for the French L2 speaker, 59% identified his accent correctly; for the Italian L2 speaker it was 30% and for the Greek L2 speaker 9%. As mentioned above, anecdotal reports and the pilot survey suggested that the FAS subject’s ‘accent’ was consistently perceived as Italian or Greek in conversational speech. As stated in the Introduction, this apparent consistency is unusual since FAS speakers are usually attributed generic and/or variable accents by listeners. The results of our perceptual experiment are in line with the anecdotal report. They showed that the FAS subject was predominantly identified in utterances as having an Italian accent (36% of the total of 134 judgements). Three other accents were mentioned – ‘Indian’ and Spanish (both 16%) and ‘Mediterranean’ (13%). A number of other accents were also mentioned (e.g. French, Greek, German), but these impressions were more sporadic. Judgements of impairment. Figure 2 presents the mean ratings for impediment judgements. It indicates that the FAS subject’s impediment rating approaches the perception of a mild impediment, while the control speakers and the Italian and French L2 speakers were overall judged as having no impediment. Interestingly, the Greek L2 speaker, who had no reported 94 J. Dankovičová & C. Hunt 4.0 Mean impediment rating Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 3.5 Impediment scale: 1 = no impediment 2 = mild impediment 3 = moderate impediment 4 = severe impediment 3.0 2.5 2.0 1.5 1.0 Essex1 Essex2 FAS French Speaker Greek Italian Figure 2. Mean impediment ratings across all speakers. speech impairment, received the highest impediment rating of all speakers, including the FAS subject. The ANOVA revealed that the difference between individual speakers’ mean impediment ratings was significant [F(5,120) ¼ 33.4, p < 0.001]. Post-hoc tests (p < 0.05) showed that the FAS speaker sounded significantly more impaired than the control Essex speakers and the French and Italian L2 speakers (all of whom were not significantly different from each other). The Greek L2 speaker was rated significantly more impaired than the FAS speaker and all the other L2 speakers. The plots in Figures 1 and 2 show a higher number of outliers for impediment ratings than for foreign accent ratings. This might suggest that those features salient to an impression of impediment may be more sporadic (in these stimuli at least) than those features salient to an impression of foreign accent and/or that listeners are more used to listening to foreign accents than impaired speech, and thus their ability to spot signs of impairment is more varied (and weaker) among listeners. Perceptual results from FAS subject. As mentioned above, the FAS subject listened to a sub-set of the stimuli in the perceptual test. Tables I and II show his judgements of foreignness in phrases (15 phrases from five speakers, including himself; total 75 phrases) in comparison Table I. Mean foreign accent ratings (FAS and English listeners). Speaker FAS subject English listeners Essex2 French Greek Italian 1.1 1.0 3.0 3.7 2.2 3.3 3.5 3.8 Perception of foreign accent syndrome speech 95 Table II. Success rates identifying presence or absence of foreign accent (FAS and English listeners). Speaker Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. FAS subject English listeners Essex2 French Greek Italian 100% 100% 80% 91% 20% 78% 93% 99% with the English listeners. Table I shows the mean ratings for all the L2 speakers and the control (Essex2) speaker from the FAS subject, compared to the mean from the English listeners for the same set of phrases as the FAS subject was tested on (the results of the ratings on his own speech stimuli will be discussed separately). Table I shows that our FAS subject’s foreignness ratings were on average lower for all the L2 speakers than ratings from the English listeners, with the difference being most pronounced for the Greek L2 speaker. Table II looks at the same data but in terms of the success rate in placing speakers correctly on the foreign accent scale (‘correct’ means a rating 1 or 2 for the Essex English speaker; 3 or 4 for the non-native speakers). Our FAS subject’s success rate was lower than that of the other English listeners. Across all speakers, he scored 73% correct compared with the English listeners’ 92% correct. The discrepancy was largest for the Greek L2 speaker, which may be quite surprising, given that his mother was Greek. However, it would be in line with the FAS subject’s claim that his mother spoke English at home, and thus his exposure to Greek-accented English was minimal at best. English listeners also found the Greek accent harder to identify correctly than French or Italian. An ANOVA analysis revealed that the difference was significant, albeit relatively weakly [F(1,118) ¼ 5.87, p ¼ 0.017], and the degree of variance accounted for was very small (5%). Table III presents the results for the judgements of impediment for our FAS subject and the English listeners. The mean ratings are very similar – both the FAS subjects and the English listeners did not perceive impediments for the control speaker and the French and Italian L2 speakers, while perceiving the Greek L2 speaker on average to be mildly impaired. The similarity was confirmed by an ANOVA, which showed no significant difference between the FAS speaker and the English listeners. In terms of judging his own speech, our FAS subject recognized his own voice amongst the sample for 11/15 phrases (73%) stimuli. When he recognized himself, he said ‘it’s that Englishman with an impediment’ rather than giving numerical ratings on the two scales. On the remaining stimuli where he did not recognize himself (n ¼ 4), the mean rating of foreignness was 2.0 and the mean rating of impediment was 2.3. Thus, the FAS subject perceived himself alongside the Greek L2 speaker to be more foreign-accented than Essex2, but less foreign and more impaired than the French and Italian L2 speakers. This is illustrated in Figure 3. Table III. Mean impediment ratings (FAS and English listeners). Speaker FAS subject English listeners Essex2 French Greek Italian 1.3 1.4 1.3 1.3 2.6 2.1 1.1 1.2 96 J. Dankovičová & C. Hunt FAS subject's mean ratings 3.8 Foreignness rating Impediment ratings 3.3 2.8 2.3 1.8 1.3 Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 0.8 Essex2 FAS French Greek Italian Speaker Figure 3. FAS subject’s mean foreign accent and impediment ratings for each speaker. Finally, it should be noted that the phrase where the FAS subject was intentionally imitating an Italian accent (‘What can I do for you?’), see above, was rated second most foreign by the listeners. Impressionistic phonetic analysis All stimuli of all speakers were phonetically transcribed by the authors. Unlike Miller et al. (2006), who used their FAS subject’s pre-stroke accent (Tyneside) as a reference accent, for the reasons explained below, we used RP English as our reference, and any pronunciation different from RP was classed as a ‘deviation’. Whilst the pre-stroke accent is the most appropriate norm from which to classify deviations when analysing pre- to post-stroke speech changes, it may not be the best option when measuring the impact of speech features on listeners’ perception, for the following three reasons. First, this approach could conceal the impact of deviations which are consistent with the pre-stroke accent but may also, in the new context, contribute to the perceived foreign accent (e.g. h-dropping is present in Essex, Italian and French accents). Secondly, if stimuli from the Essex speakers were perceived to be impaired, it would be useful to see if differences due to this accent were linked with perceptions of impairment. However, it was also noted when a ‘deviation’ could be consistent with the pre-stroke Essex accent. Deviations were classified as segmental, prosodic and paralinguistic, with further sub-categories within these three main classes (e.g. shortening/lengthening of vowel, unreduced vowel, more close/open/front/back, change in consonant voicing and manner/place of articulation – see Appendix 2 for the full classification). In the following analysis, the FAS speaker’s sounds are therefore described with reference to RP phonemic norms. (See Appendix 3 for the transcription of all the phrases as produced by the FAS speaker, the Italian and Greek L2 speakers, with reference to a RP pronunciation. The phrases are ordered by the mean foreignness ratings for the FAS speaker.) Consonants. Impressionistic phonetic analysis showed some little or weak aspiration of the plosives /p/ (1/2 tokens) and /k/ (3/7). /t/ was always produced as [t] where an Essex accent might produce a glottal stop. /k/ was realized once as the fricative [x], in a syllable-final cluster with the fricative /s/. /d/ was once produced as dental [d̪ ]. Perception of foreign accent syndrome speech 97 Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. The fricative /z/ word-finally was devoiced (more than normal levels of word-final devoicing) to [s] (5/9) and omitted once. One word-medial /v/ was devoiced to [f]; one word-final / v/ was omitted. /ð/ was produced as alveolar [d] (1/8), dental [d̪ ] (2/8), or omitted (1/8). /h/ was omitted (2/3), which could be consistent with an Essex accent, and produced as pharyngeal [ћ] (1/3). The velar nasal /ŋ/ was omitted (1/4) and produced as [n] (2/4), possibly reflecting his original Essex accent. Velar [ŋ] was produced once for a target /n/ or /m/ (depending on assimilation processes in sandwich). The liquid /l/ was produced as clear [l] (4/4) where RP speakers would use dark [ł] and an Essex accent would vocalize to [o]. The liquid /r/ was omitted once but generally produced as labiodental [ʋ] (5/6), potentially consistent with the FAS subject’s original accent. Vowels. The lax monophthongs were produced more peripherally than would be usual for an RP speaker. /ɪ/ was fronted, usually to [i] (15/24), but also [ɛ] (1/24). /ɜ/ was also fronted to [ɛ] and [œ], once each. /ʌ/ was produced once as the more back [o]. /æ/ was realized once as the more open [a]. Our FAS subject often produced an unreduced full vowel where most British English speakers would reduce an unstressed vowel to schwa (13/20). A number of these cases concerned the preposition to, which in RP and most English accents, when unstressed, would have either /ʊ/ or /u/ in stranded positions (e.g. The university I wanted to go TO) or before a vowel (e.g. TO Oxford university), and schwa elsewhere. This rule was followed by the RP speaker and both Essex speakers, but the FAS subject and all L2 speakers produced /u/ in all contexts. This is likely to contribute significantly to sounding foreign. The tense monophthongs tended towards shortening, fronting and more open realizations. /ɑ/ was shortened (1/4) and fronted to [æ] (2/4) or [a] (1/4). /i/ was shortened (4/5). /e/ was produced more open as [ɛ] (2/10). /u/ was fronted to [ʉ] (2/10), possibly consistent with the original Essex accent, but also to [y] (1/10), and the more open [u̞] (1/10) and [o] (2/10). /ɔ/ was produced as the closer [o] (2/3). The diphthong /aɪ/ was realized as [æɪ] (5/15) and [æi] (1/15) or monophthongized to [a] (1/15) and [ə] (2/15). The diphthong /əʊ/ was realized as [oʊ] (6/8) or monophthongized to [o] (2/8). Other diphthong changes were less regular but some showed evidence of shifts consistent with those noted for component parts (e.g. /eɪ/ to [ɛɪ]). There were two instances of epenthetic vowels ([ə] and [ɜ], respectively), and two instances of omission of the indefinite article a. Prosodic features. There were no instances of altered lexical stress. Of the 21 phrases, six had potentially altered sentence stress: accenting additional, fewer or different words from those the listener would expect to be accented in normal continuous speech (e.g. and ˈthey ˈstart ˈmimicking me: additional accent they). There were no obviously atypical intonation patterns. The FAS subject’s speech rhythm was perceived as having a tendency towards syllabletiming. Apart from the impression that syllables were more equal in duration than would be usual in RP, this percept was probably reinforced by a number of features mentioned above, particularly, insertion of epenthetic vowels, lack of vowel reduction, shortening of long vowels and omissions of the indefinite article a. This would be consistent with the perceived ‘syllabletimed’ rhythm of, for example, Italian, Greek and French (e.g. Dauer, 1983). Table IV summarizes the most common segmental and prosodic features found in our FAS speaker. Similarities with typical Italian and Greek accents (i.e. accents that were mentioned most frequently by listeners in the pilot and in the perceptual test) are noted. 98 J. Dankovičová & C. Hunt Table IV. Summary of most common deviations from our FAS subject and comparison with typical features of selected accents. Accent feature is consistent with: Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Feature in FAS subject’s speech Consonants weak or no aspiration of voiceless plosives word-final /z/ devoiced to [s] /h/ omitted /ŋ/ produced as [n] dark /ɫ/ produced as clear [l] /r/ produced as [ʋ] Vowels /ɪ/ produced as [i] /i/ shortening /aɪ/ produced as [æɪ] /əʊ/ produced as [oʊ] monophthongization of diphthongs non-reduction of vowels Prosodic perceived ‘syllable-timed’ rhythm epenthetic vowels Italian Greek pffi pffi pffi pffi pffi pffi pffi pffi pffi pffi some pffi some pffi pffi pffi pffi pffi pffi Essex pffi pffi pffi pffi Paralinguistic features. Additionally, paralinguistic features were noted in our FAS speaker’s phrases. Three phrases contained perceived dysfluencies, and four phrases (minor) atypical pauses. Two of the three phrases with perceived articulatory struggle yielded two of the outliers with significantly higher mean impediment scores (Figure 2). A potential pattern was observed across the stimuli containing moments of articulatory struggle. All three struggles involved target alveolar consonants (/t/, and /ɹ/). These resulted in dysfluent, apparently unco-ordinated realizations. The impressionistic phonetic analysis of our Italian and Greek L2 speakers suggests that they are in most respects typical representatives of their accent group. They displayed all the features listed as typical for their accent in Table IV, apart from epenthetic vowels. In addition, they also frequently monophthongized diphthongs, similar to the FAS speaker. Acoustic analysis The acoustic analysis focused on the segmental features that the impressionistic phonetic analysis showed to be frequently affected in the FAS speech: vowel quality and aspiration. The analysis of vowel quality involved the following steps. All utterances were manually segmented and labelled for segments and syllables using Praat (Boersma, 2001). F1 and F2 formant values were extracted automatically, using a Praat script (Burg method) at mid-point for monophthongs and at 25% (‘peak’) and 75% (‘offglide’) of vowel duration for diphthongs. The formant values were checked and any values that seemed suspicious were re-measured manually. As we were comparing our FAS speaker with other speakers (three control speakers – RP and two Essex speakers – and three genuine L2 speakers), vowel normalization was deemed necessary, to avoid possible confounding effects of anatomical/physiological variation among the speakers. We used the Lobanov method (Lobanov, 1971), also known as z-score normalization, as it has been shown to be one of the best normalization methods in terms of eliminating anatomical/physiological variation, while preserving most sociolinguistic Perception of foreign accent syndrome speech 99 Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. variation (Adank, Smits, & van Hout, 2004). For comparison of the FAS subject’s pre-stroke and post-stroke speech, no vowel normalization was of course needed. For this part of the analysis we used the ERB scale (Glasberg & Moore, 1990) as we wanted to express relationships between vowels in a perceptually plausible way. VOT was measured only for voiceless plosives in stressed syllables, where aspiration would be expected in Southern British English, as little or no aspiration was noted in a number of cases for our FAS speaker in the impressionistic phonetic analysis. The measurements were done manually, using the visual examination of the waveforms and spectrograms, and then normalized across the speakers for possible differences in articulation rate, as it has been shown that VOT is affected by rate (Theodore, Miller, & DeSteno, 2009). Vowel quality. As the distinction between stressed and unstressed vowel sets might be realized differently in FAS speech and in genuine foreign accents, the two vowel sets were considered separately. For monophthongs, both stressed and unstressed vowels were analysed; for diphthongs, only stressed vowels were analysed, as the unstressed diphthongs were normally too short to measure both peak and off-peak values reliably. Five experienced phoneticians independently analysed (rhythmic) stress placements in all the utterances. The agreement between them was very high; in 91% of all syllables (n ¼ 925) the stressed/unstressed status was judged in the same way by at least 4/5 phoneticians. Only the vowels in syllables with this level of agreement on stress placement were included in the analysis. Vowel quality – comparison with RP speaker and L2 speakers. There were in total 11 different monophthongal vowels in stressed syllables (/i/, /ɪ/, /e/, /æ/, /ʌ/, /ɜ/, /ɑ/, /ɒ/, /ɔ/, /ʊ/, /u/), eight in unstressed syllables (/i/, /ɪ/, /e/, /ə/, /ʌ/, /ɒ/, /ʊ/, /u/) and four diphthongs in stressed syllables (/eɪ/ / aɪ/ /əʊ/ /aʊ/). Appendix 4 shows the number of tokens per vowel in each speaker. The results for vowel quality are presented for each speaker separately, in relation to the RP speaker. Stressed vowels (monophthongs). Figure 4 shows the results for the stressed monophthongal vowels. For each vowel (apart from /ʊ/, which had only one token per speaker) the mean normalized formant frequency for F1 is plotted against the mean normalized formant frequency for F2. The unfilled squares represent the values for the RP speaker, while the filled circles represent the values for each of the other speakers in turn. The lines show the approximate perimeter of the vowel space for each speaker (dashed line for the RP speaker). For the FAS speaker and all the L2 speakers the identity of vowels was assigned according to the lexical distribution of phonemes in RP, as in this way any ‘mispronunciations’/accentual deviations could be shown (e.g. if /ɪ/ would be expected in RP but the speaker pronounced this vowel more like /i/, /ɪ/ was still assigned for the purposes of the acoustic analysis). The examination of the vowel space for individual speakers in Figure 4 leads to the following observations. The FAS speaker seems to have a somewhat more compact vowel space than the RP speaker in terms of the F2 dimension (front vowels further back and back vowels further front). It also differs from the two Essex speakers. Essex1 has back vowels (particularly /ɒ/, /ɔ/ and /ɑ/) noticeably further back than the FAS speaker. The main difference between the FAS speaker and Essex2 is in the close (high) vowels (and also /ʊ/) being further front in Essex2. The vowels which turned out to be significantly different along this dimension for the FAS speaker compared with all the control speakers were /ɑ/ and /u/ (p < 0.01), with /ɑ/ more front and /u/ further back. The vowel space most similar to that of the FAS speaker appears to be that of the Italian L2 speaker (no vowels were significantly different between these two speakers). 100 J. Dankovičová & C. Hunt Essex 1 2.5 1.5 –2.5 Essex 2 1.5 2.5 F2 (norm) 0.5 –0.5 Greek F2 (norm) F2 (norm) 0.5 –0.5 –1.5 –2.5 FAS 1.5 F2 (norm) 0.5 –0.5 –1.5 –2.5 2.5 French 2.5 1.5 –2.5 F2 (norm) 0.5 –0.5 –1.5 –2.5 Italian 2.5 1.5 –1.5 –2.5 –0.5 –1.5 –2.5 F2 (norm) 0.5 –0.5 –1.5 –2.5 F1 (norm) –1.5 –0.5 0.5 1.5 2.5 –2.5 1.5 0.5 F1 (norm) –1.5 –0.5 0.5 1.5 2.5 –1.5 –0.5 0.5 F1 (norm) Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 2.5 1.5 2.5 Figure 4. Speakers’ vowel space – stressed vowels (comparison with RP speaker, represented by unfilled squares). When examining the individual vowels, it is apparent that some vowel contrasts (preserved in the control speakers) are neutralized by the FAS speaker, in particular /i/ and /ɪ/, /ɑ/ and /ʌ/, /ɒ/ and /ɔ/, and /u/ and /ʊ/ (none of the vowels within these pairings is significantly different from the other in a t-test). The lack of contrast between /i/ and /ɪ/ is also seen in all the L2 speakers, and could well contribute to the FAS speaker’s perceived foreign accent. A very similar pronunciation of /ɑ/ and /ʌ/ is also apparent in the Italian L2 speaker, although this would not be unusual in southern English accents either. Finally, with the exception of /ʊ/ in the Essex1 speaker, the vowels /u/ and /ʊ/ are fronted in the three control speakers, as expected 101 Perception of foreign accent syndrome speech in these accents nowadays (Torgersen & Kerswill, 2004). However, in the FAS speech both vowels are articulated clearly as back vowels. In this respect he is most similar to the Italian L2 speaker, though all L2 speakers seem to pronounce the /u/ and /ʊ/ further back than the native English speakers. A very back articulation of these two vowels could also make the speaker sound more foreign. Unstressed vowels (monophthongs). Figure 5 presents the normalized mean values for F1 and F2 in unstressed monophthongal vowels. Again, the vowel space is shown in contrast to the 2.5 –2.5 1.5 F2 (norm) 0.5 –0.5 –1.5 –2.5 Essex 2 F2 (norm) 2.5 0.5 1.5 –0.5 –1.5 –2.5 –1.5 –2.5 –1.5 –2.5 F1 (norm) –1.5 –0.5 0.5 1.5 2.5 FAS 2.5 –2.5 Greek F2 (norm) 1.5 0.5 –0.5 –1.5 –2.5 2.5 –2.5 2.5 1.5 F2 (norm) 0.5 –0.5 F1 (norm) –1.5 –0.5 0.5 1.5 2.5 Italian 2.5 –2.5 1.5 French F2 (norm) 0.5 –0.5 –1.5 1.5 F2 (norm) 0.5 –0.5 –1.5 –0.5 0.5 F1 (norm) Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Essex 1 1.5 2.5 Figure 5. Speakers’ vowel space – unstressed vowels (comparison with RP speaker, represented by unfilled symbols). Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 102 J. Dankovičová & C. Hunt RP speaker (unfilled symbols) and the lines (dashed for the RP speaker) indicate an approximate perimeter of the vowel space. With respect to the vowel space, the FAS speaker seems to have slightly closer articulation for some vowels when compared to the RP speaker and does not seem to be very similar to any of the L2 speakers. He also differs from the two Essex speakers, who themselves seem to have a different pattern from that of the RP speaker – Essex1 with a slightly more peripheral vowel space and Essex2 with a more compact space. The graphs indicate that in the FAS speaker’s unstressed vowels the distinction between /i/ and /ɪ/ and also /u/ and /ʊ/ is preserved, unlike in his stressed vowels (there were no tokens of /ɔ/ and /ɑ/ in unstressed positions to make comparisons with /ɒ/ and /ʌ/, respectively). In this respect the FAS speaker is more similar to the native English speakers than he is in his stressed vowels. The L2 speakers also have a clearer distinction within these pairs of tense-lax vowels in unstressed positions than in stressed positions. Finally, there is a similarity to the Italian and French L2 speakers in a relatively central articulation of /ɒ/. It should be noted that the extremely frontal position of /u/ in some speakers (particularly RP and Essex1) is likely to be influenced to some degree by coarticulation (all tokens appeared after /j/ in you and university). Diphthongs. In the analysis of stressed diphthongs (/aɪ/, /eɪ/, /əʊ/, /aʊ/) we were most interested in the extent of monophthongization, as monophthongization of some diphthongs was perceived in the FAS subject in the impressionistic phonetic analysis. Figure 6 shows the normalized mean F1 and F2 frequencies for all the speakers separately (though /aʊ/ was represented only by one token). The filled symbols indicate the values for the ‘peak’ and unfilled symbols for the ‘offglide’. The two parts of each diphthongs are connected by arrows. The vowel /aɪ/ is not included as the results are not very representative for this diphthong; one occurrence in the material was before dark [ł], which was vocalized in Essex speakers, and the second occurrence was in the word tired, which can involve smoothing of the triphthong /aɪə/ into [aə] (Gramley & Pätzold, 2004). The acoustic analysis largely confirmed the auditory impressions of monophthongization. Unlike in the control speakers (RP, Essex1 and Essex2) all the diphthongs (apart from /əʊ/ in Essex1) have the peak and the offglide considerably closer to each other in the FAS speech. This is likely to contribute to sounding foreign (at least in the context of the traditional accents of South-East England). All our L2 speakers also somewhat monophthongized the diphthong /eɪ/ – the peak and offglide are closer to each other than in the control speakers – but they did so to a lesser extent than the FAS speaker. Their other diphthongs did not differ in this respect from the control speakers. Monophthongization is largely related to dialectal differences: where RP has a diphthong, a different dialect may have a monophthong. However, monophthongization can also, to some extent, result from a fast articulation rate, where there may not be enough time to achieve the articulatory targets. We therefore checked the articulation rate across all speakers but found no significant differences, apart from the French L2 speaker, who was significantly faster than all the other speakers (p < 0.05). Therefore, speech rate explanation does not hold for our FAS speaker’s tendency towards monophthongization. Accent distances. In order to gain an overall picture of how similar/different our FAS speaker is from the three control speakers and the L2 speakers in terms of vowel quality (F1 and F2), we applied the ACCDIST metric of accent similarity (Huckvale, 2007). This technique has been shown to be superior to a number of other metrics in terms of accent recognition performance. Perception of foreign accent syndrome speech RP –3 FAS F2 (norm) 2 1 0 –1 F2 (norm) 2 –2 103 1 0 –1 –2 –1 –2 –1 –2 F1 (norm) –2 –1 1 2 Essex1 –3 Essex2 F2 (norm) 2 1 0 –1 F2 (norm) 2 –2 1 0 F1 (norm) –2 –1 1 2 3 Italian –3 Greek F2 (norm) 2 1 0 –1 F2 (norm) 2 –2 1 0 F1 (norm) –2 –1 1 2 3 French –3 F2 (norm) 2 1 0 –1 –2 –2 –1 1 F1 (norm) Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 3 2 3 Figure 6. Normalized mean formant frequencies for diphthongs in individual speakers (peak ¼ filled symbols; offglide ¼ unfilled symbols). 104 J. Dankovičová & C. Hunt Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. In the present study we used ACCDIST on normalized formant frequencies (Lobanov method as in the analysis above). Figures 7 and 8 show the visual representation of the accent distances, using multi-dimensional scaling (MDS) maps. Figure 7 presents the results for all the monophthongal vowels together and Figure 8 for stressed vowels and unstressed vowels separately. The graphs indicate that when all vowels are considered together, the FAS speaker clusters most closely with the Greek L2 speaker, less so with the Italian L2 speaker and least with the French L2 speaker. He is certainly at some distance from all three English control speakers (who are, reassuringly, very close together), suggesting that vowel quality indeed contributes to his sounding foreign (Italian and ‘Mediterranean’ were the accents mentioned in the perceptual test most frequently, with Greek mentioned as well). Figure 7. Accent distances for all vowels combined. Figure 8. Accent distances for stressed and unstressed vowels separately. 105 Perception of foreign accent syndrome speech Vowel quality – comparison pre-stroke and post-stroke FAS speech. The pre-stroke sample that was available to us contained only eight short phrases, as mentioned above, and so possibilities for comparison between pre-stroke and post-stroke speech in the FAS subject were limited. Nevertheless, we were able to make a number of interesting observations. The graphs in Figures 9 and 10 show the mean formant frequencies for F1 and F2, contrasting pre-stroke (unfilled symbols) and post-stroke speech (filled symbols). The analysis was again performed separately for the stressed vowels and unstressed vowels (Figure 9) and we also present the data on stressed diphthongs (Figure 10). The examination of the vowel space shows that after the stroke, in stressed positions, the FAS subject pronounced his front vowels and also /ɔ/ with a more front articulation than before his stroke (Figure 9, left). However, he lost the nowadays usual centralized articulation of /u/ and /ʊ/, and pronounced these vowels further back after stroke (both in stressed and unstressed positions). The slightly more front articulation of front vowels after stroke appears to be also true for his unstressed vowels. Importantly, the neutralization of vowel contrasts /i/ - /ɪ/, /ɒ/ - /ɔ/ and /u/ - /ʊ/ that we observed in the post-stroke speech clearly was not present before the stroke (no example of /ɑ/ was present in the pre-stroke speech to make any claims about the /ɑ/ - /ʌ/ contrast). The loss of contrast between tense and lax vowels in stressed positions in post-stroke speech may indicate a structural, phonological deficit (although this would seem to be at odds with the site of his lesion – sub-cortical, rather than cortical). The maintained distinction (at least in the case of /i/ - /ɪ/ between tense and lax vowels in unstressed positions remains to be explained (it should be noted that some relevant vowels were not represented in our material in the unstressed vowel set). The FAS speaker’s pre-stroke (stressed) vowel space is similar to that of our English control speakers (compare Figures 9 and 4). With regards to diphthongs, we noted a tendency in the FAS speaker to monophthongize diphthongs compared to the control speakers. When comparing his pre-stroke and post-stroke Stressed vowels F2 (ERB) 23 22 21 20 19 18 17 16 15 14 13 Unstressed vowels F2 (ERB) 23 22 21 20 19 18 17 16 15 14 13 7 8 9 10 11 12 F1 (ERB) Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. When examining stressed vowels separately (Figure 8), the FAS speaker stays closest to the Greek L2 speaker but is also closer to the Italian L2 speaker than he is to the control speakers. For unstressed vowels (Figure 8) the picture is somewhat different. The FAS speaker is closest to the Essex speakers but keeps some proximity to the Greek L2 speaker here too. This broadly confirms the observations of the analysis of stressed and unstressed vowels above. 13 14 15 Pre-stroke Post-stroke Pre-stroke Post-stroke Figure 9. FAS speaker’s formant frequencies (F1 and F2) for monophthongal vowels before and after stroke. 106 J. Dankovičová & C. Hunt 23 22 21 F2 (ERB) 20 19 18 17 16 15 8 9 11 12 F1 (ERB) 10 14 15 Post-stroke peak Post-stroke offglide Pre-stroke peak Pre-stroke offglide Figure 10. FAS speaker’s formant frequencies (F1 and F2) for diphthongs before and after stroke (unfilled symbols for pre-stroke sample, filled symbols for post-stroke). speech (Figure 10) we generally found a similar tendency in the post-stroke sample, despite the fact that post-stroke speech was significantly slower (although we have to bear in mind that a very small number of data points were available for post-stroke diphthongs – see Appendix 4). His diphthongs are also narrower in post-stroke speech as opposed to pre-stroke sample. VOT (aspiration). Rate-normalized duration of aspiration was examined for all the voiceless plosives in stressed syllables that occurred in the material (the number of tokens was small – two for /p/, two for /t/ and five for /k/ in each speaker). The mean values for VOT across the speakers are shown in Figure 11. When examining the means, the FAS subject showed no aspiration for /p/, which had VOT below 20 ms, and as such would be perceived as a voiceless unaspirated plosive (e.g. Laver, 1994: 349). The most similar to him in this respect is the Italian L2 speaker. However, the means for /t/ and /k/ in the FAS subject are within the range for aspirated plosives in English Duration of aspiration (ms) Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 13 110 100 90 80 70 60 50 40 30 20 10 0 /p/ /t/ /k/ RP Essex1 Essex2 FAS French Greek Italian Speaker Figure 11. Duration of aspiration across speakers (mean values; normalized for articulation rate). Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 107 (although two instances of /k/ had similar values to his /p/). This was not the case for the Italian L2 speaker, who produced unaspirated plosives also for /t/ and /k/. When compared to the control speakers, the FAS subject had much shorter aspiration than the RP speaker, who had the highest degree of aspiration of all speakers. However, apart from /p/, he was not dissimilar to the two Essex speakers. These differences may be partly due to the fact that VOT seems to be to some extent talker-specific (e.g. Theodore et al., 2009). However, when examining the individual tokens, the FAS subject turns out to have 2/5 instances of /k/ that fall within the unaspirated category, while the Essex speakers have all tokens consistently in the range typical for English aspirated plosives. The ANOVA on the rate-normalized duration of aspiration showed the factors of speaker and segment to be significant ([F(6,54) ¼ 11.8, p < 0.001] and [F(2,54) ¼ 3.2, p ¼ 0.05], respectively). We were able to compare the degree of aspiration in the FAS subject’s pre-stroke and poststroke speech only for /k/ (four and five tokens, respectively), as the pre-stroke sample did not contain any other voiceless plosives in stressed positions. The mean duration of aspiration for his pre-stroke /k/ was 57 ms (SD 14 ms) and for his post-stroke /k/ 43 ms (SD 14 ms). We compared the articulation rate pre-stroke and post-stroke, as VOT has been shown to decrease with increasing rate (Miller, Green, & Reeves, 1986; Theodore et al., 2009). We found the mean articulation rate to be 7 syllables/second in the pre-stroke sample and 4 syllables/second in the post-stroke sample. Thus, the duration of aspiration decreased after the stroke for our subject, despite considerably slower articulation rate. Association between deviations from the reference accent and listeners’ perceptual judgements Correlational (Pearson r) statistics were employed to identify those changes most strongly associated with perceived foreign accent and impairment in our FAS subject and the L2 speakers. Correlations were calculated between each deviation category (Appendix 2) and the perceptual ratings. The following correlations were significant (p < 0.05). In the case of foreignness ratings, vocalic deviations (in general) were significant for both the French and Italian L2 speakers. For the French L2 speaker, phrases containing more deviant vowels (i.e. vowels with at least one deviation) were perceived as more foreign-accented (r ¼ 0.449). For the Italian L2 speaker, phrases containing more vocalic deviations (whether on the same or different vowels) attracted higher foreignness ratings (r ¼ 0.469). For the FAS subject and the Greek L2 speaker, the only significant deviation category was a change in vowel quality (vowels produced with closer articulation); however this was a negative correlation: phrases containing more vowels produced closer than they would be in RP were perceived as significantly less foreign-accented (r ¼ 0.445, r ¼ 0.481, respectively). In the case of impediment ratings, no deviation categories were significant for the French or Italian L2 speaker. For the FAS subject and the Greek L2 speaker, however, there were a number of salient categories (p < 0.05, unless stated otherwise). Phrases containing more consonants with deviations and individual consonantal deviations (i.e. a single consonant could have deviations in both place and manner) were rated as more impaired (FAS: r ¼ 0.437, r ¼ 0.465; Greek: r ¼ 0.525, r ¼ 0.496). Changes in vowel quality were also significant (FAS: r ¼ 0.433; Greek: r ¼ 0.548). For the Greek L2 speaker articulation shifts that were more close (r ¼ 0.727, p < 0.001) and front (r ¼ 0.459) were significant. One vocalic deviation correlated negatively with impediment ratings for the FAS speaker: phrases containing more monophthongized diphthongs were rated as less impaired (rpb ¼ 0.649, p < 0.001). (The point-biserial correlation coefficient (rpb) is the standard Pearson correlation coefficient (r) as applied to data where one variable is dichotomous and the other is roughly continuous and more or less normally distributed (Howell, 2007).) Finally, some prosodic Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 108 J. Dankovičová & C. Hunt and paralinguistic deviations were also significant for the FAS subject: phrases which did not use ‘normal’ coarticulation processes (rpb3 ¼ 0.534), and phrases containing articulatory struggle (r ¼ 0.435) were rated as more impaired. It should be noted that the correlations have limited reliability. First, the sample was small and thus many error categories had restricted range and variability. The larger number of significant correlations found with impediment as compared with foreignness ratings may be in part an artefact of the more restricted range of ratings and/or presence of outliers on the impediment scale. Furthermore, the majority of correlations discussed above are relatively weakly significant (the variance accounted for varied from 19–42%). To supplement the correlational statistics, the transcriptions were inspected to identify those deviations found in the phrases rated most highly on both scales for the FAS subject. On the foreign accent scale, 5/5 of the highest vs 2/5 of the lowest rated phrases contain at least two consonantal deviations, including at least one deviation in voicing. Deviations in vowel quality are present in both low and high foreign-rated phrases; but 3/5 of the highest vs 0/5 of the lowest foreign-rated phrases contain unreduced vowel deviations. Articulatory struggle is present in 0/5 of the highest vs 3/5 of the lowest foreign-rated phrases. On the impediment scale, 4/5 of the highest vs 1/5 of the lowest rated phrases contain paralinguistic deviations; 3/5 of the highest vs 1/5 of the lowest rated phrases contain altered sentence stress. No patterns were identified for segmental deviations. Summary and discussion One of the main aims of this study was to examine impressions of foreign accent and speech impairment in an English FAS subject and to compare the results with native and non-native (L2) speakers. We also investigated the FAS subject’s own perception of foreignness and impairment, in an attempt to assess perceptual abilities of this kind in FAS. Furthermore, deviations from a standard English (RP) accent were examined and phonetic characteristics of the FAS speaker were compared to genuine L2 control speakers. Our FAS speaker was perceived as strongly foreign-accented by English native listeners, to a similar extent to the Italian and French L2 speakers, and significantly more so than the Greek L2 speaker (while the control English speakers were judged correctly as native English speakers). The most frequently mentioned accent was Italian, although ‘Indian’, Spanish, ‘Mediterranean’ and Greek accents were also mentioned by a number of listeners. The results of the perceptual test suggest that the general conception of FAS as a disorder in which the person sounds foreign but not impaired may not be entirely accurate. Our FAS subject was judged on average to be slightly impaired (but significantly more foreign-accented than impaired). However, the Greek L2 speaker, who did not have any speech impairment, was also judged as impaired (he was in fact rated more impaired than the FAS subject). This opens a wider question about how people perceive accents and speech impairments, which we will discuss in more detail below. The impressionistic phonetic analysis of the FAS subject’s post-stroke sample revealed a number of segmental and prosodic features that deviated from RP English, and also from a typical Essex accent (his pre-stroke accent). However, the hypothesized correlation between the total number of deviations and foreign accent ratings was not significant for the FAS subject. This contrasts with other studies of foreign accent ratings in FAS (Miller et al., 2006) and also of genuine foreign accent speakers (Major, 1986; Magen, 1998). Comparison with our French and Italian L2 speakers may provide a clue. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 109 While the FAS subject and the Greek L2 speaker show no deviations correlating positively with foreign accent ratings, but a number of deviations correlating with impediment ratings, the French and Italian L2 speakers show the opposite pattern. It appears, then, that the 2-fold rating scale in this study (impediment as well as foreign accent), compared with those of Major (1986), Magen (1998) and Miller et al. (2006), might explain the contrasting findings. Miller et al. hypothesized that deviations might fall into two distinct groups: markers of foreign accent vs markers of impairment. If so, for an individual speaker, a dominance of ‘foreign accent marker deviations’ might result in a correlation between total deviations and foreignness ratings (as in our Italian L2 speaker), and a dominance of ‘impairment marker deviations’ might result in a correlation between total deviations and impairment ratings (as in our Greek L2 speaker). Investigation of the correlations across all speakers does not, however, reveal a straight forward dichotomy of features acting intrinsically as either foreign accent or impairment markers. Several patterns were observed (although correlations were generally weak). For example, deviations in vowel quality correlated with perceived impairment for both the FAS subject and the Greek L2 speaker, but, for the French and Italian L2 speakers, vocalic deviations correlated with ratings of foreignness. This variability fits with the picture emerging from the growing body of research into perception of ‘genuine’ foreign accent. Currently there is no evidence that any specific segmental or prosodic feature is intrinsically significant to a general perception of foreign accent. Instead, a variety of features have been implicated across different studies with different accent groups (e.g. Van Els & de Bot, 1987; Carmichael, 2000; Riney, Takada, & Ota, 2000; Gut, 2007). According to this evidence, while no single feature stands out as a universal marker of foreign accent, individual features may be markers of accent for a particular speaker or accent group. There are likely to be a number of further reasons for the weak correlations in our study. First, our material consisted of a small dataset, which limited the number of possible deviations. Secondly, and more importantly, it may be unrealistic to expect a linear relationship between the number of deviations and the strength of perception of foreignness (and perhaps impairment too), as people’s perception of accents may not function in this way. There are a number of other factors to consider. It is possible that in the process of the perceptual test listeners became familiar with a speaker’s voice quality (despite the stimuli from all speakers having been randomized together), recognized the speaker, and based ratings on their ongoing impressions of the speaker rather than deviations in individual stimuli. As Miller et al. (2006: 403) point out, ‘. . . once a speaker has formed an impression of a person’s speech in their listening mind, their overall impression holds sway. Variables that do not match their interpretation are either sidelined or reinterpreted within their general impression’. Furthermore, any classification system applied to observed deviations may not capture the categories that actually differentiate listeners’ perceptions. Different categories, individual specific deviations, or particular loci and combinations of deviations may be most salient for listeners (Miller et al., 2006). Phonetic similarity/dissimilarity may also play a role. According to two key cross-language speech perception models (Best, 1995; Flege, 1995), phonetically dissimilar differences are more noticeable to non-native language learners than phonetically similar differences. Extending this theory to perception of foreign-accented speech, it may be hypothesized that deviations that are phonetically dissimilar to the native target are more noticeable to listeners than phonetically similar deviations. Our classification of deviations was not sensitive to phonetic similarity effects. Regarding loci of deviations, studies have found that, for native speakers, speech errors and dysfluencies such as stuttering (developmental and neurogenic) are most prevalent in wordonset position (Shattuck-Hufnagel, 1987; Van Borsel, Van Lierde, Van Cauwenberge, Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 110 J. Dankovičová & C. Hunt Guldemont, & Van Orshoven, 1998), whereas for non-native speakers of English errors are more prevalent in word-final position (e.g. Flege and Davidian, 1984; Hansen, 2001). We might hypothesize, therefore, that due to listeners’ experience of native (both pathological and non-pathological) and non-native speakers, the loci of errors might have a bearing on the perception of foreign accent vs impairment. The small sample in our study did not make it possible to control for the effect of these specific variables. The lack of a clear pattern of correlations between the types of deviations and perceptual ratings is likely to reflect to some extent the genuine underlying variability in listeners’ perceptions. In our study, as in Miller et al. (2006), inter-listener agreement, although significant, was quite low. Some of the variability of course relates to individual differences in listeners’ experience of accented and/or impaired speech (the ‘familiarity effect’ as described in Di Dio et al. (2006)). However, some variability may relate to how individual listeners apply the two categories ‘foreign accent’ and ‘impairment’, and whether these categories are conceptualized as related or unrelated. Miller et al. (2006) hypothesized a dichotomy of deviation types: (1) (2) Deviations consistent with the listener’s experience of non-native speakers’ pronunciation errors and which may be salient to a perception of foreign accent. Deviations consistent with the listener’s experience of impaired speech and which may be salient to a perception of impairment. We may add another type of deviation: (3) Deviations that are observed but not familiar may be perceived as ‘maybe foreign, maybe impairment’. Finally, as suggested by Flege (1988: 363), it is important to pay attention to ‘the partial approximation that is observed’ as well as the deviations. Thus, a fourth deviation type might be proposed: (4) Deviant features that are at least partially consistent with the listener’s experience of native speakers’ production and which may therefore be particularly salient to a perception of ‘no foreign accent’ or ‘nativeness’. Notwithstanding the importance of additional consideration of combinations and loci of deviations, as discussed above, how does this help explain why the FAS subject and the Greek L2 speaker were perceived as foreign and slightly impaired? The suggestion is as follows: listeners may have interpreted the majority of deviations in the French and Italian L2 speakers’ stimuli as type 1 from the list above, but deviations in the FAS subject’s and the Greek L2 speaker’s stimuli as types 2, 3, or 4. For the FAS subject, while some deviations match type 1 (typical foreign accent deviations – cf. Table IV), others (e.g. articulatory struggle) match type 2 (typical impairment deviations), or are not immediately attributable to a familiar foreign accent (e.g. /k/ to [x], type 3). The Greek L2 speaker’s higher impediment ratings may have occurred simply because listeners generally had less experience of a typical Greek (specifically Greek-Cypriot) accent, therefore interpreted more deviations as type 3. His lower foreign accent ratings may have occurred due to features in his speech resulting from assimilation with the accent of his local south London community (type 4 deviations). It should be noted that the perceptual ratings might also be partly influenced by the fact that nave listeners (especially our sample, recruited from multicultural London) are likely to be more familiar with foreign accents than with speech impairments and therefore more subtle manifestations of impairment might escape their attention. Also, when some deviations can occur both in foreign accents and in speech impairments, it is likely that listeners would categorize them under foreign Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 111 accent rather than under speech impairment (e.g. schwa insertion is claimed to be common in apraxia of speech, but it is also a marker of an Italian or Greek accent; Miller at al., 2006). We mentioned earlier that our FAS subject was, according to anecdotal reports and the pilot survey, consistently perceived as Italian or Greek in conversational speech (and in our material, an Italian accent dominated any other accents suggested by the listeners). This raises a question about the extent to which his deviations were similar to those found in our Italian (and Greek) L2 speaker. The impressionistic phonetic analysis revealed a number of features in the FAS subject that were consistent with typical Italian and Greek accents (Table IV). Both our Italian and Greek L2 speakers displayed nearly all these features as well, together with some others also shared by our FAS subject. In the acoustic analysis of vowel quality we compared our FAS speaker to three native English control speakers (RP and two Essex speakers) and also to our genuine L2 speakers. We found a number of features in which our FAS speaker was different from the native English speakers but similar to the L2 speakers. For instance, a lack of distinction between /i/ and /ɪ/, a tendency to pronounce /u/ and /ʊ/ as back vowels rather than more centrally, as would be common in standard English nowadays, and a tendency towards monophthongization of diphthongs. All these features may contribute to the perception of a foreign accent, although their salience for the listeners may differ, as discussed above. The acoustic analysis suggested that, in some aspects of vowel quality, our FAS subject was most similar to the Italian L2 speaker; in particular, in the overall shape and position of the vowel space relative to the reference RP speaker, but also in positioning of some vowels within this space. Furthermore, our FAS subject manifested a lack of distinction between tense and lax vowels /i/ - /ɪ/, /ɑ/ - /ʌ/ /ɒ/ - /ɔ/ and /u/ and /ʊ/. As already mentioned, the neutralization of / i/ - /ɪ/ contrast was shared by all the L2 speakers, but the neutralization of the /ɑ/ - /ʌ/ contrast was shared only by the Italian and Greek L2 speakers. To express the distances in vowel quality between the FAS subject and the control native and L2 speakers quantitatively, we used the ACCDIST metrics (Huckvale, 2007). The analysis showed somewhat different results for the stressed and unstressed vowel sets. In the case of stressed vowels, the FAS subject clustered most closely with the Italian and Greek L2 speakers (which is in line with the perceptual judgements), while in unstressed vowels, he clustered with the native English speakers. The difference between stressed and unstressed vowel sets was noted already in the acoustic analysis, but we cannot draw any strong conclusions about this difference, as it may simply reflect the fact that the vowels that made the FAS speaker most different from the control English speakers in the stressed vowel set happen not to be represented in the unstressed set. The use of ‘weak forms’ needs to be investigated further in FAS. The impressionistic phonetic analysis suggested a tendency towards syllable-timed rhythm in our FAS speaker. The impression of more equal syllable durations than would be expected in native British varieties of English would be only one contributing factor. The occurrence of epenthetic vowels, frequent lack of vowel reduction, and shortening of long vowels that was noticed in the impressionistic analysis are also likely contributors to the impression of altered speech rhythm. Italian is generally classified as a syllable-timed language (e.g. Ramus, Nespor, & Mehler, 1999; Dellwo, Fourcin, & Abberton, 2007), but Greek seems to be placed in a category between syllable-timed and stress-timed languages (e.g. Grabe & Low, 2002; Baltazani, 2007). Thus, the FAS subject’s speech rhythm may also add to the impression of sounding Italian (and to some, Greek). Further work is in progress on the prosodic characteristics of our FAS subject’s speech, in particular speech rhythm, but also speech rate and intonation. Only a very few studies have been able to examine differences between pre-stroke and poststroke speech in FAS subjects. We were able to make some comparisons for our FAS subject, Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 112 J. Dankovičová & C. Hunt though our pre-stroke sample was very short and did not contain instances of all the vowels present in the post-stroke sample, or the relevant consonants for the analysis of aspiration. We found that some of the prominent features in the post-stroke sample did not occur in the prestroke sample, suggesting that these changes arose as a consequence of the stroke. In particular, the vowels /u/ and /ʊ/ seem to be further back in the post-stroke sample, while they are centralized in his pre-stroke sample, just like in our control native English speakers. Other vowels seem to have more frontal articulation in the post-stroke than in the pre-stroke sample. A tendency towards monophthongization of diphthongs is apparent in the post-stroke, but not the prestroke sample. There is no obvious underlying physiological mechanism that could explain these changes. Interestingly, the distinction between tense and lax vowels is robust in the pre-stroke sample. This may suggest a structural, phonological deficit, although this claim remains speculative, unless more evidence can be found. (A purely phonological deficit might be surprising given the sub-cortical site of the lesion, although the neural basis and cortical–sub-cortical pathways for phonological processing are still not well understood.) Of course, physiological and phonological deficits do not have to be mutually exclusive. In our study we also investigated the FAS subject’s own perception of accent and impairment. The results showed that his ratings were generally in line with the ratings by the English listeners for both foreignness and impairment, providing no evidence that FAS has affected his ability to perceive accents. However, the perceptual task was not sufficiently sensitive to measure whether more specific skills (e.g. perception of voicing differences) were unimpaired. Therefore, further tasks exploring FAS subjects’ perception of individual features (word minimal pairs, intonation patterns, etc.) would be required before concluding that perception is wholly unaffected by FAS. Our study was based on a limited sample set, which made it difficult to obtain a more comprehensive picture of our FAS subject’s speech characteristics. A larger sample would cover a fuller set of phonological categories and more tokens per category. It would also facilitate the selection of stimuli with a more balanced range of deviation, and the control of psycholinguistic factors such as word frequency (Varley et al., 2006). Future studies of FAS might consider using a classification of phonetic deviations that picks out specific loci and/or combinations of features that are most salient to perceptions of foreign accent and/or impairment. It might also be useful to conduct the perceptual testing in three separate blocks (one rating foreign accent only; one rating impediment only; one rating both) to help control for the effects of correlation (in both directions) between the two scales. To conclude, there are four main questions concerning FAS: (i) Why is FAS speech generally perceived as sounding foreign rather than impaired?; (ii) Is there a particular neurological mechanism underlying FAS?; (iii) Are there common phonetic characteristics across different FAS cases?; and (iv) Can the phonetic aspects of FAS be explained by some underlying speech mechanism? The study reported here provides some answers mainly to the first question. We showed that when perception of FAS speech is specifically tested for speech impairment, FAS was, in fact, judged to involve slight impairment. The reasons for why FAS is normally perceived as foreignsounding and not impaired are to do with the fact that in FAS the deviations are normally of the kind occurring in at least some of the world’s languages (Blumstein et al., 1987), but they are also to do with the complex nature of people’s perception of foreignness and impairment and their experience in both, as we described above. In general, our study supports the ‘linguistic relative view’ (Miller et al., 2006), according to which that listeners focus on salient elements of the speech that match accents they have themselves experienced, rather than the ‘generic Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 113 (accent) view’ (Kurowski et al., 1996), or the ‘perceptual view’ that the perceived accent is entirely ‘in the ear of the beholder’ (e.g. Ardila et al., 1988; Carbary et al., 2000). With regard to the second question, our FAS case adds to the variety of lesion sites reported in FAS cases. His lesion was at the brainstem (pons), which is more unusual for FAS (but cf. Villaverde-Gonzalez et al., 2003), although more commonly implicated in cases of dysarthria (Kumran, Bayülkem, & Evyapan, 2002). We should, however, note that our FAS case may have a more complicated aetiology. As previously stated, some authors argue that neurogenic FAS should exclude any psychogenic aetiology (e.g. Blumstein & Kurowski, 2006), and previous exposure to another language or accent (Seliger et al., 1992; Munson & Heilman, 2005). In the case of our subject, a multi-factorial aetiology for FAS should be considered. A neurological aetiology in our FAS case was suggested by the emergence of the accent alongside other typical neurological symptoms, and the identification of a pontine lesion in the MRI scan. However, a recent case describes a FAS speaker presenting with a range of other neurological symptoms judged to be caused by psychological disturbance (conversion disorder) (Haley, Roth, HelmEstabrooks, & Thiessena, 2010). Also, in our subject’s case, it is noted that the lesion identified was small, of atypical localization (compared to other FAS cases) and, especially given the time lapse between the symptoms and the scan (14 months), might therefore be incidental to the symptoms. Therefore, potential psychogenic and/or behavioural factors might also be considered. First, although our FAS subject recalls that his mother spoke only English with him and herself developed a ‘Cockney’ accent, it is hard to exclude the possibility that she spoke Greek and/or English with a Greek accent with him in early childhood. Secondly, our subject reported how he fairly quickly adopted a new identity in line with the accent, for example attributing his emotional lability (a common stroke symptom unrelated to FAS) to an ‘Italian’ stereotype of being more ‘emotional’ than the British (cf. also the case described by Lippert-Gruener, Weinert, Greisbach, & Wedekind (2005), whose subject similarly found her accent helpful in ‘explaining’ other symptoms). This reaction cannot be excluded as a potential behavioural factor contributing to the persistence of the accent for years rather than the shorter periods seen in some other cases (see Haley (2009) for a summary of the recovery patterns seen in the literature). Finally, it is perhaps noteworthy that listeners rated the FAS subject’s intentional imitation of an Italian accent highly foreign-sounding, since most reported FAS speakers are unable to modify their ‘accent’ (Haley, 2009). Thus, our case supports Haley’s (2009) suggestion that further research is needed in the areas of differential (i.e. psychogenic/neurogenic) aetiologies and longitudinal recovery patterns of FAS. Finally, a few notes on the last two questions. Our FAS speaker demonstrates a number of phonetic features seen in some previous subjects, for example, monophthongization of diphthongs, non-reduction of unstressed vowels (cf. Whitaker, 1982; Blumstein et al., 1987; Berthier et al., 1991; Kurowski et al., 1996; Miller et al., 2006) and reduced aspiration of voiceless plosives (e.g. Dankovičová et al., 2001). However, the data suggest also neutralization of some phonological vocalic contrasts, which had not been reported in previous FAS cases, as far as we know. It is not obvious whether the overall pattern of deviations in our FAS speaker could be explained by a common underlying physiological mechanism. The pons (his lesion site) contains nuclei that control facial expression (and thus aspects of articulatory positioning), and nuclei that relay signals from the cerebrum to the cerebellum (which controls fine motor planning, coordination and timing) (Webb & Adler, 2007). Some of his deviations may be explained by impaired motor planning and timing, similar to an ataxic (cerebellar) dysarthria. Involvement of the cerebellum in FAS has recently been reported, for example, by Mariën, Verhoeven, Engelborghs, Rooker, Pickut, and De Deyn (2006), Marien and Verhoeven (2007) and Cohen, Kurowski, Steven, Blumstein, and Pascual-Leone (2009). 114 J. Dankovičová & C. Hunt However, we speculated that FAS may also (sometimes) involve a structural, phonological deficit. More research is needed to obtain a clearer picture about the underlying causes of the segmental and prosodic deviation patterns in FAS and their impact on perception of foreignness and impairment. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Acknowledgements We would like to thank the FAS subject for his participation, Jill House, Francis Nolan and Chris Lee for their comments on an earlier draft of this paper, John Coleman and Jennifer Gurd for sharing the FAS subject’s case history and neuropsychological assessment results, Sophie Scott for sharing MRI scan interpretation, Volker Dellwo, Mark Huckvale and Mike Coleman for technical support, and also Jyrki Tuomainen, Gordon Craig and Paul Iverson for their assistance with the statistical analysis. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. References Adank, P., Smits, R., & van Hout, R. (2004). A comparison of vowel normalization procedures for language variation research. Journal of the Acoustical Society of America, 116, 3099–3107. Ardila, A., Rosselli, M., & Ardila, O. (1988). Foreign accent: an apraxic epiphenomenon? Aphasiology, 2, 493–499. Aronson, A.E. (1990). Clinical voice disorders: An interdisciplinary approach. New York: Thieme. Atwell, E., Howarth, P., & Souter, C. (2003). The ISLE corpus: Italian and German spoken learners’ English. ICAME Journal, 27, 5–18. Bakker J.I., Apeldoorn S., & Metz L.M. (2004). Foreign accent syndrome in a patient with multiple sclerosis. Canadian Journal of Neurological Sciences, 31, 271–272. Baltazani, M. (2007). Prosodic rhythm and the status of vowel reduction in Greek. In Selected papers on theoretical and applied linguistics from the 17th International Symposium on Theoretical & Applied Linguistics, 1, 31–43. Department of Theoretical and Applied Linguistics, Salonica. Berthier, M.L., Ruiz, A., Massone, M.L., Starkstein, S.E., & Leiguarda, R.C. (1991). Foreign accent syndrome: Behavioural and anatomical findings in recovered and non-recovered patients. Aphasiology, 5, 129–147. Best, C.T. (1995). A direct realist view of cross-language speech perception. In W. Strange (Ed.), Speech perception and linguistic experience: Issues in cross-language research (pp. 171–204). Timonium, MD: York Press. Blumstein, S.E., & Kurowski, K.M. (2006). The foreign accent syndrome: A perspective. Journal of Neurolinguistics, 19, 346–355. Blumstein, S.E., Alexander, M.P., Ryalls, J.H., & Dworetzky, B. (1987). On the nature of the foreign accent syndrome: A case study. Brain Language, 31, 215–244. Boersma, P. (2001). Praat, a system for doing phonetics by computer. Glot International, 5, 341–345. Browning, S. (2004). Analysis of Italian children’s English pronunciation. Unpublished report contributed to the EU FP5 PF STAR Project. Accessed 1 July 2007, from: http://www.eee.bham.ac.uk/russellm/ItalianEnglishReport/ ItalianEnglish_report_v2.htm Carbary, T.J., Patterson, J.P., & Snyder, P.J. (2000). Foreign accent syndrome following a catastrophic second injury: MRI correlates, linguistic and voice pattern analyses. Brain Cognition, 43, 78–85. Carmichael, L. (2000). Acoustic variability and perceived foreign accent. Journal of the Acoustical Society of America, 108, 2651. Christoph, D.H., de Freitas, G.R., Dos Santos, D.P., Lima, M.A., Araújo, A.Q., & Carota, A. (2004). Different perceived foreign accents in one patient after prerolandic hematoma. European Neurology, 52, 198–201. Coelho, C.A., & Robb, M.P. (2001). Acoustic analysis of foreign accent syndrome: An examination of three explanatory models. Journal of Medical Speech-Language Pathology, 9, 227–242. Cohen, D.A., Kurowski, K., Steven, M.S., Blumstein, S.E., & Pascual-Leone, A. (2009). Paradoxical facilitation: The resolution of foreign accent syndrome after cerebellar stroke. Neurology, 73, 566–567. Coleman, M.D. (2008). Soundplayer. [Computer Program]. Division of Psychology and Language Sciences, University College London, UK. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 115 Coughlan, T., Lawson, S., & O’Neill, D. (2004). French without tears? Foreign accent syndrome. Journal of the Royal Society of Medicine, 97, 242–243. Dalby, J., & Kewley-Port, D. (1999). Explicit pronunciation training using automatic speech recognition technology. Special issue of the Journal of the Computer Assisted Language Learning Consortium (CALICO), 16, 425–445. Dankovičová, J., Gurd, J., Marshall, J., MacMahon, M., Stuart-Smith, J., Coleman, J., & Slater, A. (2001). Aspects of non-native pronunciation in a case of altered accent following stroke (foreign accent syndrome). Clinical Linguistics and Phonetics, 15, 195–218. Dauer, R.M. (1983). Stress-timing and syllable-timing reanalyzed. Journal of Phonetics, 11, 51–62. Dellwo, V., Fourcin, A., & Abberton, E. (2007). Rhythmical classification based on voice parameters. The 16th International Conference of Phonetic Sciences (1129–1132). Saarbrücken. Di Dio, C., Schulz, J., & Gurd, J.M. (2006). Foreign accent syndrome: In the ear of the beholder? Aphasiology, 20, 951–962. Duffy, J.R. (1995). Motor speech disorders. St. Louis: Mosby. Duguid, A. (2001). Italian speakers. In M. Swan & B. Smith (Eds.), Learner English (2nd ed. pp. 73–89). Cambridge: Cambridge University Press. Efstathiadis, S. (1974). Greek and English phonology: A comparative investigation. Greece: Thessalonike Aristoteleion Panepistemion Thessalonikes, Philosophike Schole. Flege, J.E. (1995). Second language speech learning: Theory, findings, and problems. In W. Strange (Ed.), Speech perception and linguistic experience: Issues in cross-language research (pp. 233–277). Timonium, MD: York Press. Flege, J.E. (1988). The productions and perception of foreign language speech sounds. In H. Winitz (Ed.), Human communication and its disorders. A review (pp. 224–401). Norwood, Nj: Ablex. Flege, J.E., & Davidian, R. (1984). Transfer and developmental process in adult FL Spanish productions. Applied Psycholinguistics, 5, 323–347. Glasberg, B.R., & Moore, B.C.J. (1990). Derivation of auditory filter shapes from notched noise data. Hearing Research, 47, 103–138. González-Alvarez, J., Parcet-Ibars, M.A., Avila, C., & Geffner-Sclarsky, D. (2003). A rare neurologically originated speech disorder: Foreign accent syndrome. Revista de Neurología, 36, 227–234. Goodglass, H. (2000). The Boston diagnostic aphasia examination (3rd ed.). Philadelphia: Lippincott Williams. Grabe, E., & Low, E.L. (2002). Durational variability in speech and the rhythm class hypothesis. Papers in Laboratory Phonology, 7, 515–546. Graff-Radford, N., Cooper, W., Colsher, P., & Damasio, A. (1986). An unlearned foreign‘accent’ in a patient with aphasia. Brain and Language, 28, 86–94. Gramley, S., & Pätzold, K.-M. (2004). A survey of modern English. London: Routledge. Gurd, J.M., & Coleman, J.S. (2006). Foreign accent syndrome: Best practice, theoretical issues and outstanding questions. Journal of Neurolinguistics, 19, 424–429. Gurd, J.M., Bessell, N.J., Bladon, R.A.W., & Bamford, J.M. (1988). A case of foreign accent syndrome, with clinical follow-up, neuropsychological and phonetic descriptions. Neuropsychologia, 26, 237–251. Gurd, J.M., Coleman, J.S., Costello, A., & Marshall, J.C. (2001). Organic or functional? A new case of foreign accent syndrome. Cortex, 37, 715–718. Gut, U. (2007). Foreign accent. In C. Müller (Ed.), Speaker classification 1 (pp. 75–87). Berlin: Springer. Haley, K.L. (2009). Dysprosody and the Foreign accent syndrome. Perspectives on Neurophysiology and Neurogenic Speech and Language Disorders, 19, 90–96. Haley, K.L., Roth, H.L., Helm-Estabrooks, N., & Thiessena, A. (2010). Foreign accent syndrome due to conversion disorder: Phonetic analyses and clinical course. Journal of Neurolinguistics, 23, 28–43. Hansen, J.G. (2001). Linguistic constraints on the acquisition of English syllable codas by native talkers of Mandarin Chinese. Applied Linguistics, 22, 338–365. Howell, D.C. (2007). Statistical Methods for Psychology (6th ed.). Belmont, CA: Thomas Higher Education. Huckvale, M. (2007). ACCDIST: An accent similarity metric for accent recognition and diagnosis. In C. Müller (Ed.), Speaker Classification II. Lecture notes in computer science series. Series edited by J. Carbonell & J. Siekmann (pp. 258–275). Berlin: Springer. Humbley, J. (2002). French. In M. Görlach (Ed.), English in Europe (pp. 108–127). New York: Oxford University Press. Ingram, J., McCormack, P., & Kennedy, M. (1992). Phonetic analysis of a case of foreign accent syndrome. Journal of Phonetics, 20, 457–474. Katz, W.F., Garst, D.M., & Levitt, J. (2008). The role of prosody in a case of foreign accent syndrome (FAS). Clinical Linguistics & Phonetics, 22, 537–566. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. 116 J. Dankovičová & C. Hunt Kumran, E., Bayülkem, G., & Evyapan, D. (2002). Clinical spectrum of pontine infarction. Journal of Neurology, 249, 1659–1670. Kurowski, K.M., Blumstein, S.E., & Alexander, M. (1996). The foreign accent syndrome: A reconsideration. Brain Language, 54, 1–25. Laures-Gore, J., Henson, J.C., Weismer, G., & Rambow, M. (2006). Two cases of foreign accent syndrome: An acoustic-phonetic description. Clinical Linguistics & Phonetics, 20, 781–790. Laver, J. (1994). Principles of Phonetics. Cambridge: Cambridge University Press. Lippert-Gruener, M., Weinert, U., Greisbach, T., & Wedekind, C. (2005). Foreign accent syndrome following traumatic brain injury. Brain Injury, 19, 955–968. Lobanov, B.M. (1971). Classification of Russian vowels spoken by different speakers. Journal of the Acoustical Society of America, 49, 606–608. Luzzi, S., Viticchi, G., Piccirilli, M., Fabi, K., Pesallaccia, M., Bartolini, M., . . ., Snowden, J. S. (2008). Foreign accent syndrome as the initial sign of primary progressive aphasia. Journal of Neurology, Neurosurgery & Psychiatry, 79, 79–81. Magen, H.S. (1998). The perception of foreign-accented speech. Journal of Phonetics, 26, 381–400. Major, R.C. (1986). Paragoge and degree of foreign accent in Brazilian English. Second Language Research, 2, 53–71. Major, R.C. (2001). Foreign accent: The ontogeny and phylogeny of second language phonology. Mahwah, N.J., London: Lawrence Erlbaum Associates. Mariën P., & Verhoeven J. (2007). Cerebellar involvement in motor speech planning: Some further evidence from foreign accent syndrome. Folia Phoniatrica et Logopaedica, 59, 210–217. Mariën, P., Verhoeven, J., Engelborghs, S., Rooker, S., Pickut, B.A., & De Deyn, P.P. (2006). A role for the cerebellum in motor speech planning: Evidence from foreign accent syndrome. Clinical Neurology and Neurosurgery, 108, 518–522. Miller, J.L., Green, K.P., & Reeves, A. (1986). Speaking rate and segments: A look at the relation between speech production and speech perception for the voicing contrast. Phonetica, 43, 106–115. Miller, N., Lowit, A., & O’Sullivan, H. (2006). What makes foreign accent syndrome foreign? Journal of Neurolinguistics, 19, 385–409. Moen, I. (2000). Foreign accent syndrome: A review of contemporary explanations. Aphasiology, 14, 5–15. Monrad-Krohn, G.H. (1947). Dysprosody or altered ‘melody of language’. Brain, 70, 405–415. Moonis, M., Svearer, J.M., Blumstein, S.E., Kurowski, K.M., Licho, L., Kramer, P., . . ., Drachmon, D. A. (1996). Foreign accent syndrome following a closed head injury. Perfusion deficit on SPECT with normal MRI. Neurology, 43, 381. Munson, P.D., & Heilman, B. (2005). Foreign accent syndrome: Anatomic, pathophysiologic and psychosocial considerations. South Dakota Journal of Medicine, 58, 187–189. Papaefthymiou-Lytra, S.C. (2001). Greek speakers. In M. Swan & B. Smith (Eds.), Learner English. (2nd ed., pp. 129–144). Cambridge: Cambridge University Press. Pick, A. (1919). Uber Anderungen des Sprachcharaketers als Begleiterscheinug aphasicher Storungen. Zeitscrift fur gemsamte Neurologie und Psychiatrie, 45, 230–241. Pulcini, V. (2002). Italy. In M. Görlach (Ed.), English in Europe (pp. 151–167). New York: Oxford University Press. Ramus, F., Nespor, M., & Mehler, J. (1999). Correlates of linguistic rhythm in the speech signal. Cognition, 73, 265–292. Reeves, R.R., & Norton, J.W. (2001). Foreign accent-like syndrome during psychotic exacerbations. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 14, 135–138. Reeves, R.R., Burke, R.S., & Parker, J.D. (2007). Characteristics of psychotic patients with foreign accent syndrome. Journal of Neuropsychiatry and Clinical Neurosciences, 19, 70–76. Riney, T.J., Takada, M., & Ota, M. (2000). Segmentals and global foreign accent: The Japanese flap in EFL. TESOL Quarterly, 34, 711–737. Robinson, T. (Presenter). (2005, April 9). Home Truths [Radio broadcast]. London: BBC. Edited version. Accessed 26 March 2007, from: http://www.bbc.co.uk/radio4/hometruths/20050411_italian_accent.shtml Rogers, D., & d’Arcangeli, L. (2004). Illustrations of the IPA. Journal of the International Phonetic Association, 34, 117–121. Roth, E., Fink, K., Cherney, L., & Hall K. (1997). Reversion to a previously learned foreign accent after stroke. Archives of Physical Medicine and Rehabilitation, 78, 550–552. Ryalls, J., & Whiteside, J. (2006). An atypical case of Foreign accent syndrome. Clinical Linguistics & Phonetics, 20, 157–162. Scott, S., Clegg, F., Rudge, P., & Burgess, P. (2006). Foreign accent syndrome, speech rhythm and the functional neuroanatomy of speech production. Journal of Neurolinguistics, 19, 370–384. Seliger, G.M., Abrams, G.M., & Horton, A. (1992). Irish brogue after stroke. Stroke, 23, 1655–1656. Shattuck-Hufnagel, S. (1987). The role of word-onset consonants in speech production planning. In E. Keller & M. Gopnik (Eds.), Motor and sensory processes of language (pp. 17–51). Hillsdale, NJ: Erlbaum. Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 117 Stathi, E. (2002). Modern Greek. In M. Görlach (Ed.), English in Europe (pp. 301–330). New York: Oxford University Press. Tait, O. (Executive Producer). (2008, August 7). My strange brain [Television broadcast]. London: Granada Factual London, for Five. Theodore, R.M., Miller, J.L., & DeSteno, D. (2009). Individual talker differences in voice-onset-time: Contextual influences. Journal of the Acoustical Society of America, 125, 3974–3982. Torgersen, E., & Kerswill, P. (2004). Internal and external motivation in phonetic change: Dialect levelling outcomes for an English vowel shift. Journal of Sociolinguistics, 8, 23–53. Tranel, B. (1987). The sounds of French: An introduction. Cambridge: Cambridge University Press. Tsuruga, K., Kobayashi, T., Hirai, N., & Kato, S. (2008). Foreign accent syndrome in a case of dissociative (conversion) disorder. Seishin Shinkeigaku Zasshi, 110, 79–87. Van Borsel, J., Janssens, L., & Santens, P. (2005). Foreign accent syndrome: An organic disorder? Journal of Communication Disorders, 38, 421–429. Van Borsel, J., Van Lierde, K., Van Cauwenberge, P., Guldemont, I., & Van Orshoven, M. (1998). Severe acquired stuttering following injury of the left supplementary motor region: A case report. Journal of Fluency Disorders, 23, 49–58. Van Els, T., & De Bot, K. (1987). The role of intonation in foreign accent. The Modern Language Journal, 71, 147–155. Varley, R., Whiteside, S., Hammill, C., & Cooper, K. (2006). Phases in speech encoding and foreign accent syndrome. Journal of Neurolinguistics, 19, 356–369. Verhoeven, J., Mariën, P., Engelborghs, S., D’Haenen, H., & De Deyn, P. (2005). A foreign speech accent in a case of conversion disorder. Behavioural Neurology, 16, 225–232. Villaverde-González, R., Fernández-Villalba, E., Moreno-Escribano, A., Alías-Linares, E., & García-Santos, J.M. (2003). Foreign language syndrome as a first sign of multiple sclerosis. Revista de Neurología, 36, 1035–1039. Walter, C. (2001). French speakers. In M. Swan, & B. Smith (Eds.), Learner English. (2nd ed., pp. 52–72). Cambridge: Cambridge University Press. Webb, W., & Adler, R.K. (2007). Neurology for the Speech-Language Pathologist. London: Mosby. Weismer, G., & Liss, J. (1991). Reductionism is a dead-end in speech research: Perspectives on a new direction. In C.A. Moore, K.M. Yorkston & D.R. Beukelman (Eds.), Dysarthria and apraxia of speech: Perspectives on management (pp. 15–27). Baltimore: Paul H. Brookes. Whitaker, H. (1982). Levels of impairment in disorders of speech. In R. Malatesha, & L. Hartlage (Eds.), Neuropsychology and Cognition (Volume 1). The Hague: Nijhoff. Appendix 1 Test phrases for all speakers: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) when I was a small child then I could not do it this is a country I want to go to but after a few days but that is what I am doing I feel a bit giddy I’m a happy-go-lucky person in the evening I come home joking about and they start mimicking me I went to Oxford University I’ve been many places but it will pass very quickly they asked me to do research with them and I go to order a sandwich and the man comes to me to serve me loads and loads of tests you become very tired 118 (19) (20) (21) J. Dankovičová & C. Hunt because I don’t want him to get upset what can I do for you and I went back to work Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Phrases from pre-stroke sample (1) (2) (3) (4) (5) (6) (7) (8) and it’s still holding up to its character behind me you see the cathedral come with me we’ll have a look which was severely bombed during the blitz of 1940 but visiting the inside of the structure today we visit the historical city of Coventry it still has a lot of nobility Appendix 2: Classification system for deviations Segmental Consonantal Vocalic: (a) Changes in vowel quantity Vocalic: (b) Changes in vowel quality Prosodic Stress Intonation Rhythm Paralinguistic Self-correction Articulatory struggle Atypical pause Omission of usual coarticulation process Omission/insertion Change in voicing (including deviations in aspiration and voice onset time) Change in place of articulation Change in manner of articulation Long vowel shortened Short vowel lengthened Monophthong diphthongized Diphthong monophthongized Unreduced vowel substituted for reduced More close More open More front More back More central Nasalization Altered word stress Altered sentence stress Altered intonation Epenthetic schwa Syllable omission 3.8 3.9 3.9 3.9 4 [bət ɑːftəɹə fjuː deɪz] [wɒt khən aɪ du fə ju] [ən(d) aɪ went bæk tə wɜːk] [dʒəʊkɪŋ əbaʊt] 3.7 3.7 3.6 3.6 [ən ðə mæŋ khʌmz tə miː tə sɜːv miː] [bɪkhəz aɪ dəʊnt wɒnt hɪm tə get ʌpset] [ðen aɪ khʊd nɒt du ɪt] [ðeɪ ɑːs(kt) mi tə du ɹɪsɜːtʃ wɪðəm] [ən ðeɪ stɑːt mɪmɪkɪŋ mi] 3.6 3.6 [ju biˈk ʌm veɾi ˎt aɪœɹd] [bʌt it {_}ɪ ̞ˈp¼as vɛɾi q`ˎkwɪkli] [aɪ ˈfiɫ ə ˈbit ˎgidi] [a ˈwe̞ntu ˈoksfodə juniˎvœsiti] [av ˈbin ˈmɛni ˎplɛsis] [en aɪ ˈgo tu ˈɔɹdœɾ ə ˎsɛŋwitʃ] [d̪is izøe ˈkhɑntɹi aɪ ˈwont̚ tu ˎgo tu] [bət ̪̚ ˈd̪æ̱ t iz ˈwot aɪ əm ˎduiŋ] [ˈlodz ən lodz ɒv ˎt¼ests] [ˈwen a ˈwɒz ə ˈsmoːɫ ˎtʃaɪld] [en ðə ˈmɛn ˈkhɑms tu mi tu ˎsœɹ{_} mi] [bik¼oz a ˈdõ{_} wont {_}im tu ˈgɛt əpˎ sɛt] [ˈd̪en a k¼ʊd̚ ˈnɒt̚ ˎdu it] [d̪ e ˈas mi ˈtuː ˈduː ɾiˎsœtʃ wɪd̪əm] [ʔɛn d̪e ˈstɑt ˎmĩmikiŋ mi] [ˈbʌt̚ ˈafteɾə fju ˎˏdeɪs] [ˈwo{_} khən aɪ ˎdu fə ju] [en ə wɛm ˈbɛk tu ˎwoɹk] [ˈdʒokiŋ əˎbaʊt] [aɪm {_} ˈћæpi go ˈlʌki ̮ p¼ɜːsən] [ɪn{_}i ˈifnĩ{_} ãɪ̃ khom ˏ{_} õʊ̃m] [ən ðə ˈmaŋ ˈkhʌm{_} to ˈmei tu ̮sœːv mi] biˈk¼əs æɪ ˈdõʊ̃{_} wɒnt ˈ{_}im tu ˈge̞t ʌpˎset̚ [ˈðen æɪ ˈkhʊd̚ nɒt̚ ˎdʉ ɛt] [d̪eɪ ˈas mi tu̞ ˈdu̞ ʋəʋiˎsɛtʃ wiðem] [ənd̚ ˈd̪eɪ ˈstæːt ˎmimɪkɪn mi] [bʌt ˈɑftə{_} ˈfjʉ ˏd̪eɪs] [ˈwɒtə k¼ən ə ˈdo fæ ˎjo] [aɪ ˈwɛntu bæk̚ tu ˎwɜːk] [ˈdʒokɪn ə ˎbæʊt̚ ] Greek L2 [ju bɪˈk am veɹi ˏthaɪə{_}] [ˈbʌʔ i{_} wil ˈphaz veɹi ˎkwɪkli] [a ˈfio ɛ ˈbit ˎgɪd̪ɪ] [a wen ˈtwoksfɔ{_} juːniˏvesɪti] h [bə{_} ˈɑːftə {_} ˈfjʉ ˎdeɪs] [ˈwoʔ ˈqhan a ˈdu fö ˎju] [ˈɛn{_} a ˈwenː{_} bak tə ˎwok] [ˎdʒokɪn əbɑ{_}] [ɛn ðe ˈmɛn ˈqhams tu mi n̩ ˎserv mɪ] [biˈkhɒs a ˈdõ{_} won{_} ˈhim tu ˈget apˏset] [ˈðenː æ ˈkhʊdenɒ{_} ˎdui{_}] [ðɛ ˈas mi tu ˈduɛ ɾɪˈsɛʃ ˎwɪðəm] [ˈɛn ðe ˈstɑːʔ ˎmĩmĩkĩ{_} mĩ] [a{_} bi{_} ˈmeni ˎpleiziz] [ɛn{_} a ˈgo ˈtwodɛ{_}e ˎsændɥiʃ] [ˈðɪs ɪs ə ˈkhʌntɹi ə wɒn{_}ə ˎgou {_}] [bəʔ ˈðaʔ is ˈwoʔ {_} am ˎduin] [ˈlods an ˈlods ov ˎthesː] [ə ˈwen ə wəs ə ˈzmo{_} ˎtʃao {_}] [am ə ˈhæ̱ pi go ˈlɑki ˎp¼œɹsn̩ ] [æm {_} ˈhapi go ˎlakɪ p¼œsə̃{_}] h [in ḏi ˈivəniŋ aɪ ˈk ɑm ˎhoʊm] [ˈin ðe ˈivənɪŋ a ˈqham ˎhoʊm] h [ˈjy pik ʌm ˈveʋi ̮t æɪəd] [ˈbət d d it wil ˈphæs veʋi ̖ kwɪkli] [æɪ ˈfil ə ˈbit ˈɜː ˎgɪdɪ] [aɪ ˈwent̚ tu ˈoxsfəd ju̞ːni ̮vɜːsəti] [a{_} bin ˈmɛni ˏplɛɪsɪz] [ænd æɪ ˈgoʊ t tu tu ˈodəʋə ˎsæŋwitʃ] [ˈðis is ə ˈk¼ʌntʋi aɪ ˈwɒntu ˎgoʊ tu] [bʌ̚t ˈdæt is ˈwɒt aɪ æm ˎduɪŋ] [ˈloʊdz ən ˈloʊdz əv ˎthes] [ˈwen ə wəs ə ˈsmol ˎtʃæild] h Italian L2 h FAS h Note: Deviations are marked in bold in transcriptions; omissions/indicated by {_}; phrases are ordered by mean foreign ratings for the FAS subject: scale from 1 ¼ definitely English to 4 ¼ definitely foreign. and the man comes to me to serve me because I don’t want him to get upset then I could not do it they asked me to do research with them and they start mimicking me but after a few days what can I do for you and I went back to work joking about [aɪm ə hæpi gəʊ lʌki I’m a happy-go-lucky person phɜːsən] in the evening I come home [ɪn ði iːvnɪŋ aɪ khʌm həʊm] 3.5 3.5 3.6 3.5 3.3 3.4 3.3 3.3 I feel a bit giddy [aɪ fiːɫ ə bɪt gɪdi] I went to Oxford University [aɪ went tu ɒksfəd juːnɪvɜːsɪti] I’ve been many places [aɪv biːn meni pleɪsɪz] and I go to order a [ən(d) aɪ gəʊ tə ɔːdəɹə sandwich sæmwɪtʃ] this is a country I want to go [ðɪs ɪz ə khʌntɹi aɪ wɒnt tə to gəʊ tu] but that is what I am doing [bət ðæt ɪz wɒt aɪ əm duɪŋ] loads and loads of tests [ləʊdz ən ləʊdz əv thes(ts)] when I was a small child [wen aɪ wəz ə smɔːɫ tʃaɪɫd] h FAS foreign. rating 2.9 3.1 h RP you become very tired [ju bɪk ʌm veɹi t aɪəd] but it will pass very quickly [bət ɪt wɪɫ phɑːs veɹi kwɪkli] Orthographic Appendix 3: Transcriptions of reference RP speaker, FAS subject, Italian, and Greek L2 speakers Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Perception of foreign accent syndrome speech 119 120 J. Dankovičová & C. Hunt Appendix 4: Number of tokens per vowel Clin Linguist Phon Downloaded from informahealthcare.com by Kainan University on 05/10/15 For personal use only. Speaker Vowel (stressed) i ɪ e æ ʌ ɜ ɑ ɒ ɔ ʊ u Vowel (unstressed) i ɪ e ə ʌ ɒ ʊ u Diphthongs (stressed) eɪ aɪ əʊ aʊ RP Essex1 Essex2 FAS post-stroke FAS pre-stroke Greek Italian French 4 5 9 5 4 5 4 3 3 1 4 3 6 6 5 4 5 4 3 1 1 5 2 4 6 4 4 4 3 4 3 1 5 3 6 8 4 3 5 3 4 2 1 6 4 8 — — 3 — — 4 1 1 1 4 6 8 6 6 5 4 3 3 1 6 3 5 7 5 4 5 4 4 2 0 4 4 5 7 4 4 5 4 5 2 1 5 11 20 1 32 2 2 3 3 9 18 1 29 1 1 3 2 8 19 3 28 2 2 3 2 11 18 2 29 2 2 3 2 3 19 — 17 1 — 1 1 12 18 2 24 1 3 2 2 10 19 3 28 1 2 3 1 10 19 1 29 2 1 3 3 2 3 7 1 2 2 7 1 2 2 7 1 2 2 7 1 4 3 2 1 2 2 7 1 2 2 7 1 2 2 6 1