LG405 Language Variation and English Phonology The Environments for /L/ Vocalisation in Romford, Colchester and Ipswich: London Diffusion or Independent Development? By Chloe Gathern Abstract This project investigates the specific feature of /l/ vocalisation in the towns of Romford and Colchester, both in Essex, and Ipswich in Suffolk, and the common assumption that the feature is present in these localities due to a process of diffusion moving northward from London. Despite evidence from Kjederqvist (1903) describing the first ‘recent’ recorded incidence of /l/ vocalisation in Pewsey (Wiltshire), of late it has been customary to imply that many of the developing linguistic features, including the vocalisation of dark /l/, originate in London, and have spread into surrounding dialects. However, there is some discussion as to whether /l/ vocalisation would not have occurred in these localities as a natural, independent progression of the dialect, without the alleged diffusion from London. My findings, on the whole, showed a rapid progression of the feature through the region since the SED data was recorded in the 1950s, but also that Colchester was a higher vocalising dialect than Romford, a town geographically closer to London, and closer to the heartland of cockney speech. This would suggest that Colchester developed /l/ vocalisation through the natural development of the dialect, and whilst the influence of the London dialect is undoubtedly present, it is not the sole determining factor. The similarities of some aspects of the data between Colchester and Ipswich also suggest that Ipswich, which has been developing vocalisation since before 1960, is influenced by Colchester in a way that is completely independent from the features influenced by London. 1. Introduction At a time when living in one town and working in another, potentially an hour or more away is commonplace, the commuter culture has greatly increased both the economic, and the linguistic mobility of many communities. With this has come greater levels of dialect contact, and the accommodation of new features. With the current public fascination with ‘Estuary English’, it has become easy to assume that the influence of London is the be-all and end-all of language change in the southeast, ignoring the possibility that the reasons for changes in the dialect may lie elsewhere. Johnson and Britain (2003) believe that /l/ vocalisation, in particular, is a natural phenomenon, and occurs in line with a general trend in language change toward the unmarked. In investigating the different levels of vocalisation in the three communities, I will also examine the different environments which appear to favour vocalisation in each of the communities, to discover whether the development of /l/ vocalisation in that area can be attributed to the diffusion of the feature from London. I will examine data taken from recordings of 12 conversational interviews made with people from the three localities; Romford, Colchester and Ipswich, restricting my investigations to female speakers from differing age groups, in order to gain a perspective of the change across an age difference, not gender. In addition to examining existing studies discussing the linguistic restraints upon the levels of /l/ vocalisation, I will also look at the Sociolinguistic factors that can have a great influence over the distribution of a feature, and current Geolinguistic theory relating to the diffusion of dialect features. 2. Data Collection Methodology In order to gain a perspective of the change over time, I looked at data from both the SED (Survey of English Dialects, 1971), and from a corpus of recordings from each of the speech localities I’ve studied as part of this project (Romford, Colchester and Ipswich), shown in fig. 1 below. 2.1 Corpus Recordings All recordings that I used form part of a larger archive held at the University of Essex, including the Colchester Archive of Spoken English. All the recordings I used consisted of taped conversations, all of which were seemingly informal. I decided to focus on female speakers because of their tendency to promote linguistic innovations, and in order that the results could indicate reliable patterns chose two informants from each age group of ‘Old Speakers’ (aged 30+) and ‘Young Speakers’ (less than 30 years old). The background information for the Colchester archives indicated that the informants had lived and worked in the Colchester area all their lives, and whilst no background information data is held in the archives for the Romford and Ipswich recordings, the details given by the informants during their interviews suggest they too have lived in their respective localities for most of their lives. Fig. 1 – Map showing the locations of Romford, Colchester and Ipswich, as well as the SED localities Kersey, West Bergholt and Doddinghurst. (http://maps.google.co.uk) Despite a marked variation in the quality of sound in the recordings, I focussed on highlighting the environments where dark /l/ is found, and taking the same amount of tokens from each speaker, established if vocalisation of dark /l/ took place, then looked at the wider environment of the token to see if the surrounding element could be an indicating factor in the decision to vocalise or not. 2.2 SED data The Survey of English Dialects is the report of a questionnaire of speakers in rural areas around the UK published in 1971. In order to gain a historical perspective on how far the feature has spread since that time, I looked at the closest localities in the SED to those I studied from the recordings corpus; Doddinghurst (nr. Romford), West Bergholt (nr. Colchester) and Kersey (nr. Ipswich) all of which are shown in (fig. 1) above. Whilst it has been suggested, by Peter Trudgill, that the recent spread of /l/ vocalisation follows the Urban Hierarchy model of linguistic diffusion, this approach seemed the fairest way to search the SED, for at that time dialect surveys focussed on smaller, rural communities, with no alternative localities being of a similar population or urban status to the three towns I am studying the corpus data of. In this way the structure of the SED also undermines the female focus of this project, through the informants in all these areas all being older male participants. I took token words from the SED which exemplified the environments in which dark /l/ exists, and which, according to the literature, promote vocalisation; specifically non-prevocalic /l/ and syllabic /l/ environments. I elicited token words with a variety of preceding vowels and consonants in order to establish if a preferential environment possibly exists. 3. The /l/ Variable 3.1 Brief Description /l/ is described by the literature as a coronal lateral approximant, implying a lateral release of air across the sides of the tongue, as well as contact being made at the alveolar ridge. /l/ can appear in syllable onset positions as ‘clear’ (also referred to as ‘plain’ or ‘light’) but in many dialects can also be realised in syllable coda positions as ‘dark’ /l/ through a process of velarisation due to the retraction of the tongue to the velum. Giles and Moll (1975) found that for dark /l/ in English, coronal contact is not even achieved for all speakers. In dialects where the clear/dark /l/ dichotomy exists, this loss of coronal contact (tongue tip at the alveolar ridge) induces a vocoid realisation of dark /l/, occasionally with increased lip rounding. This vocalisation is exclusive to dialects which accept both clear and dark realisations of /l/, and even where this dichotomy is a recent development, such as in East Anglia (Trudgill 1999), there is a resistance to the vocoid form. The transcription of the vocoid allophone is subject to differing opinions, but is transcribed by Tollfree (1999) as []. 3.2 /l/ Vocalisation: Past and Present The vocalisation of dark /l/ is by no means a new phenomenon. Words such as ‘calf’ and ‘talk’ where the /l/ preceded a labial or velar sound underwent vocalisation in the 16 th century. However, it would be fair to describe the current wave of vocalisation currently affecting the UK a recent event, with earliest references dating back just over 100 years. Whilst many linguists argue that London is the main source of major linguistic innovations, Ellis, writing in 18891, makes no mention of a clear/dark /l/ distinction, let alone of vocalisation. The lack of vocalisation references is to be expected when we remember that a dialect must contain dark /l/ first, before any vocalisation can take place, and dark /l/ did not begin to appear in many dialects until the mid 19th century. The first reference to the current spread of vocalisation originates not from London, but Pewsey in Wiltshire, when Kjederqvist (1903:10) states: “the treatment of l is a most characteristic feature of the Pewsey dialect…I often found the children did not sound the l in football …and other words where l was final” To illustrate the ferocity with which the feature spread, it is noted in some studies of Cockney (as in Matthews 1938), less than 50 years after Kjederqvist’s first reference, that the vocalisation of /l/ was already a marked characteristic of their dialect. Fig. 2 – The realisation of /l/ in the nucleus position in mid-20th century England. (derived from Orton et al (1978)’s Linguistic Atlas of England)2 1 2 Ellis 1889, as cited in Johnson and Britain (2004): LG405 Course Notes, Uni of Essex. Image scanned from Johnson & Britain (2003) Fig. 2 above shows the distribution of /l/ allophones within England in the late 1970s, showing the pockets of vocalised /l/ in the Oxfordshire area, as well as the ‘belt’ like arrangement spreading from the south coast up into south Essex. 3.3 Constraints Upon Vocalising Environments Whilst Kjederqvist’s original 1903 analysis of the Pewsey dialect showed vocalised /l/ appearing in a post-vocalic environment, Linguists now observe several environments in which the vocoid realisation of /l/ is possible. Horvath and Horvath (2001) note that the “vocalisation process should be understood in terms of 3 primary places within the syllable; post-vocalic in a coda position, post-vocalic in a cluster, and syllabic /l/”. However, /l/ vocalisation has been shown to arise in varying environments according to the dialect, and Tollfree (1999) also notes examples of vocalisation in word-final intervocalic positions primarily amongst young speakers in her survey of Southeast London dialects. In a post-vocalic environment, the length of the preceding vowel can affect the levels of vocalisation, with various studies suggesting that a preceding long vowel or diphthong creates a very conducive environment for promoting vocalisation. Spero’s 1996 Colchester data3 shows a large difference in the varbrul weighting of long vowels (with a 0.6 weighting) when compared to short vowels (with a 0.35 weighting). Horvath and Horvath (2001) during their investigations into Australian and New Zealand also not a marked difference in the distribution of vocalisation after long and short vowels, both studies thus supporting the theory that preceding long vowels are a more popular environment. This discernable difference in vocalisation is contradicted by Meuter’s 2002 Colchester study, where Colchester 6 year olds were actually shown to prefer vocalising after short rather than long vowels. This appears however to be solely a trend amongst young speakers, who by the age of 10 had apparently acquired the dominant adult patterning of preferring long vowels. Wells (1982) notices a trend towards a shortening of the preceding long vowels in London, resulting in words such as ‘fill’ and ‘feel’ to become completely homophonous. Bauer’s 1994 study of the high-vocalising New Zealand English also described this development, leading us to believe that it is part of a continuing development in dialects where /l/ vocalisation is already firmly established. As vowel shortening was not reported in the Fens, a relatively newly vocalising dialect, by Britain’s 2003 data, this would seem to corroborate the theory. Syllabic /l/ remains an ever-popular environment for vocalisation and, as with preceding vowel length, different dialects vocalise this segment to varying extents. Again, the preceding sound proves to be an 3 Statistics as shown in Johnson & Britain (2004): LG405 Course Handouts. Colchester, University of Essex. important consideration, with most dialects following a different preceding consonant type hierarchy. Horvath and Horvath’s New Zealand study in 2001 showed that 3 New Zealand cities preferred a preceding labial consonant, and favoured a dorsal over a coronal example. This is the same hierarchy found in Britain’s 2003 Fenland data, and appears to closely resemble the same hierarchy that existed during the 16th century vocalisations. However, the hierarchy described by Horvath and Horvath for Australia (2001:44) differs greatly in that a preceding labial consonant is considered the least likely environment, and a dorsal the most likely. This is the same pattern found by Meuter (2002) amongst Colchester children, and although the dorsal is placed first, the ranking of the labial and coronal can be interchanged. Spero’s Colchester and Southend data unexpectedly places dorsal sounds as the least likely candidate for vocalisation, in so far as we would expect a preceding dorsal consonant to promote vocalisation due to the shared place of articulation with the dark /l/ and its vocalic alternatives, but Spero’s data concurs with the New Zealand data in placing labials as the most conducive environment. It is interesting to note the absence of coronal sounds from the highest ranking in any of the hierarchies, but Walsh Dickey (1997) comments that, in Jamaican English, when a coronal segment such as [t] or [d] appears before a syllabic /l/, that the dorsal feature is spread and assimilates the preceding consonant, so that they become realised as a dorsal sound. This dorsal spread assimilation has caused words such as ‘little’ to be realised as [lkl]. If the coronal sound were to remain as coronal before a syllabic /l/, we would expect that its place of articulation would inhibit the production of a dark /l/ and in this way its vocalised counterpart. According to Tollfree (1999), vocalisation in word-initial and word-internal intervocalic environments is blocked regardless of morphology. However she found pre-vocalic vocalisation to be very widespread amongst younger speakers in Southeast London. Also amongst young speakers, Spero (1996) found 33% vocalisation before a vowel in Southend, and 16% in Colchester, indicating perhaps a developing trend amongst younger speakers. It is not only a development found in the UK, as Horvath and Horvath (2001) find a 20% vocalisation in this environment when examining Australian and New Zealand speakers. It is perhaps another development that is restricted to highly vocalising dialects, as Britain’s Fens data found only 2 tokens in the entire corpus but instead show that speakers in that area prefer to insert a linking /l/ instead when the following vowel occurs across a word boundary. This pattern, according to Johnson and Britain (2003:28) indicates a “constraint against vocalic onsets”, Uffmann (2003) claims that high vocalising dialects choose to retain the vocalised /l/, but insert a glide [w] to avoid a vocoid onset segment. Tollfree (1999) also finds evidence of word-final intervocalic vocalisation, but only amongst her younger speakers, in examples such as ‘legal info’ [ligwnf], although the frequency of vocalisations is noticeably less than in other environments. Gick (1999) notices that the same environment in the speech of South-central Pennsylvania, but says that a linking /l/ is always inserted. Wells (1982) maintained that this environment was extremely prohibitive to vocalisation, leading to presumption that this development may be a change in its early stages. 3.4 Why Does Vocalisation Occur? The manner in which dark /l/ is produced greatly facilitates its vocalisation. Phonetically speaking, both allophones of /l/ are produced by a combination of coronal (consonantal) and dorsal (vocalic) gestures, the difference between them lying in the order in which the gestures occur. The label of dark /l/ as ‘velarised’ belies the fact that both allophones have a dorsal gesture, but the production of clear /l/ necessitates an initial coronal gesture, thus rendering the dorsal element weaker. This process is reversed when producing a dark /l/, placing the vocalic dorsal gesture before the coronal one. It is this weaker rendering of the coronal gesture which results in the potential loss of tongue tip contact, producing a vocoid realisation. Gess (2001) claims that the vocalisation of dark /l/ is a result of the CAE (conserve articulatory effort) constraint, and that by limiting the movement of the tongue, we are more prone to missing the coronal target when producing a dark /l/. Sproat and Fujimura (1993) go further to explain that the positioning of dark /l/ at the syllable rhyme makes the coronal gesture inherently weaker, because coronal gestures tend to be stronger in a syllable initial position. Therefore, this principle makes the dorsal gestures in a dark /l/ stronger in the syllable final position. If a dorsal gesture is a better rhyme segment than a coronal, logically, given its vocalic nature, we can assume that a vowel sound would be even better suited to the environment. This vocalic nature explains further why a preceding long vowel is considered such a conducive environment for vocalisation, with the initial dorsal gesture in the dark /l/ strengthened during the realisations of the vowel sound, thereby rendering the coronal gesture much more difficult to obtain. 3.5 Is Vocalisation A Natural Progression? English is not the only language to undergo the process of /l/ vocalisation. Walsh Dickey (1997) shows that evidence exists from languages as diverse as Catalan where vocalisation occurs before a bilabial stop [b] (‘aube’ [aub] was originally [alb]), and Mehri, which resists the vocalisation of word-medial intervocalic /l/ (‘Third’ can be realised as [o:l] if masculine, or [we:t] if feminine). We can reasonably assume, because of the cross-linguistic material, that there must be some underlying reason behind such similar changes in many languages. Many linguists now suggest that the patterning behind many instances of language change seems to be moving towards what is considered ‘unmarked’ or more ‘natural’. Some linguists, like Gnanadesikan (2004) point to child language as an indicator of what is ‘natural’, because in their early speech, children are only able to produce natural forms, thus conforming to the markedness constraints that then become superceded by the faithfulness constraints that we develop later as we harness an adult phonology. Where children have been observed acquiring a dialect that has the clear/dark /l/ dichotomy in adulthood, evidence shows that the /l/ present in the syllable rhymes is often vocalised; Amahl Smith, for example, renders ‘table’ as [bebu] at 2 years (Smith 19724). Furthermore, it is often noted that liquids are acquired at a later stage than stops or glides, leading to many children producing a [w] where an adult of relevant English dialects would use a dark /l/. Clear /l/, conversely, is acquired early in languages such as French (which has no clear/dark /l/ dichotomy) thus implying it to be more ‘natural’ and ‘unmarked’ than its dark /l/ allophone. Statistics in the literature showing a higher percentage of /l/ vocalisation amongst younger speakers indicate a language change moving in the direction of the apparently unmarked forms, suggesting that /l/ vocalisation should be considered a natural evolution of the language, and not a change instigated through laziness as some non-linguists presume. 4. Sociolinguistic considerations 4.1 Dialect Contact and Gender Labov (1990) maintains that, “in the majority of linguistic changes, it is females who are instrumental in the diffusion of prestige forms” and whilst males are more likely to promote more vernacular forms, many linguists maintain that women are more linguistically innovative. However, in their 2001 study of /l/ vocalisation in Australian and New Zealand English, Horvath and Horvath cast doubt on Labov’s claim, noting that in terms of /l/ vocalisation, “sex and social class are only weakly associated with the variability” (2001:40) suggesting instead that the age of speakers and the locality form more important considerations. 4.2 Dialect change and Age Considered an important factor by Horvath and Horvath (2001:40), the age of speakers seems to influence their readiness to use innovations. Young speakers are traditionally seen as more innovative, 4 Smith (1972), as described in Johnson, W & D. Britain (2004): LG405 Course handouts. Colchester, Essex University. being less likely to cling to existing traditional forms, perhaps because the age group is likely to be considerably more mobile, and has more contact with other dialect speakers. Britain’s 2003 Fens data, for example, shows a marked difference between levels of /l/ vocalisations between the younger and older age groups surveyed, with younger speakers consistently vocalising more. Patterns like this lead to the assumption that the difference between the speech of the younger and older age groups can often be indicative of a language change in progress. 4.3 Influence of locality All three localities investigated are within a commutable distance of London, often cited as the most influential urban centre in the southeast. Hernández-Campoy describes the linguistic influence of London as constituting the “principal spatial diffusing nucleus, from which innovation normally radiates outward” (1999:18). Linguistic influence through spatial diffusion necessitates contact and interaction between speakers of the existing and innovative dialect forms. This interaction can occur in many forms, e.g. travelling to school or work and can lead to an accommodation of features, which may be adopted and spread if the interaction results in a ‘target’ form amongst younger speakers. Trudgill 1986 notes that “it (/l/ vocalisation) reached Cambridge and Colchester some decades ago and is beginning to affect Ipswich”. Ipswich is located just within the Suffolk borders, and the Ipswich dialect contains many features in contrast with the two Essex localities. 4.4 Expansion Diffusion Britain describes expansion diffusion as involving “the spread of linguistic forms from a core locality to neighbouring ones through everyday interactions between core and neighbouring localities.” (Unpublished5:11) There are varying models to explain expansion diffusion patterns. ‘Wave model’ diffusion, suggests that innovations radiate slowly from the central point purely on a distance basis, rather than taking the size of the settlement into account. However, /l/ vocalisation is more commonly associated with the ‘Urban Hierarchy Model’ which suggests that in addition to a wave-like spread from a central point, the innovations affect localities according to a hierarchy of settlement size, from large city down through to village. Radford et al. (1999) observe this patterning in the Fens area, where levels are much higher in urban areas than in the rural settlements that surround them. Britain (unpublished6 :14) comments that “the theory is that interaction between urban centres is greater, and facilitates the transmission of innovations better then urban to rural.” 5 6 http://homepages.tesco.net/~david.britain/7.pdf http://homepages.tesco.net/~david.britain/7.pdf Trudgill (1974b) goes further to suggest that ‘gravity models’ can be used to predict the course of an innovation’s expansion, by calculating the interaction using the populations of the place of origin and target destination, as well as the distance between them. This method has been used successfully by Trudgill (1986) to calculate the levels of h-drop diffusion in East Anglia, notably Norwich, Lowestoft and King’s Lynn. However, this equation assumes that distance and population are the only determining factors, and ignores that the varying social factors can affect the way the distance is experienced. A community is not 100% intrinsically mobile, whether large or not, and more economically mobile groups have wider ranging contacts than socially static groups. However, the ease in which a feature is diffused between dialects largely depends how salient the variant is considered to be, according to Kerswill (2001), who claims that forms which are neither overly stigmatised nor too far below the consciousness of most speakers are the most effectively transmitted. Vocalisation of dark /l/ would appear, then, to be a prime candidate for such diffusion, as a feature which is classed by Horvath and Horvath (2001:39) as lying “below the level of consciousness” 5. Results and Discussion 5.1 SED findings Having learnt of the ferocity with which /l/ vocalisation spread within the Cockney dialect, I was expecting to find very low levels of vocalisation, if it existed at all, in the two Essex based localities (Doddinghurst and West Bergholt), and was not expecting to see any evidence of vocalisations in the Suffolk locality, Kersey (near Ipswich). I was surprised to find no evidence at all of /l/ vocalisation in West Bergholt, considering its close proximity to Colchester which is now considered a high vocalising community. Doddinghurst only vocalised in three of the total 19 token words I examined, which was a lot less than I had expected because it was geographically the closest locality to Romford, and therefore to London. Only the absence of vocalisations in Kersey appeared to concur with my hypothesis. S p e e c h l o c a l i t y S y l l a b i c / l / % v o c a l P r e . s h o r t V % v o c a l P r e . l o n g V % v o c a l D o d d i n g h u r s t 0 1 / 0 81 2 . 5 0 % 0 1 / 0 4 2 5 % 0 1 / 0 71 4 . 2 0 % W e s t B e r g h o l t 0 0 / 0 8 0 % 0 0 / 0 4 0 % 0 0 / 0 7 0 % K e r s e y 0 0 / 0 8 0 % 0 0 / 0 4 0 % 0 0 / 0 7 0 % Fig. 3 – Percentage vocalisations from SED data. Looking at the tabulated percentages in fig. 3 above, it would be impossible to glean any kind of preferential environment within the Doddinghurst data, with equal amounts of token being vocalised in each of the three most common environments for vocalisation today. There were no instances at all of word internal intervocalic vocalisation, but this is expected, as the literature has suggested that any type of intervocalic examples will only develop in dialects where vocalisation is already extremely common. It is impossible to devise whether vocalisation would take place in a word final intervocalic position across a word boundary, as the SED does not record the surrounding environment to the token words. My conclusions from these localities would be to assume that /l/ vocalisation had not yet reached these communities. Certainly, this would appear to be true for Kersey, Suffolk. However, in examining the entries for other localities surveyed within the Essex county borders, there were several other communities mentioned which frequently employed vocalisation, but were geographically further from the Essex localities I have investigated (Romford and Colchester). In particular, three communities stood out with high levels of vocalisations; Little Bentley (to the east of Colchester), Little Baddow (to the west of Maldon, Essex) and Canewdon (close to Southend, Essex). All three communities consistently vocalised syllabic /l/ after velar plosive [k], as well as after the long [ü:] in ‘school’, and the short [ε] in ‘twelve’. There does not appear to a way to account for these anomalous results, as even an urban hierarchy diffusion model cannot be applied here, as the three high vocalising communities are apparently small communities. 5.2 Corpus Results 5.2.1 General vocalising levels across localities Locality No. of vocalised tokens % Vocalisation Romford 123/140 88% Colchester 126/140 90% Ipswich 90/140 64% Fig. 4 - % Vocalisation according to locality Both Romford and Colchester, in accordance with my prior suppositions, have high levels of vocalisations. Colchester, with 90% token vocalisation appears to vocalise more than Romford, which would cast doubt on the theory that the feature has spread in a ‘wave diffusion’ manner outwards from London. The figure for Ipswich, whilst significantly lower than the others, is still fairly high. But as Claxton (1960:4) maintained of the Suffolk dialect; “…the consonant l is nearly always turned into a ‘w’ when preceding a ‘d’ or a ‘t’…” showing that /l/ vocalisation was found in Suffolk a lot earlier than is perhaps realised, thereby explaining the elevated levels. 5.2.2 Vocalisation across Age Groups It is apparent from fig. 5 below, that age is indeed an important consideration when investigating levels of vocalisation. Once again, the levels of vocalisation in Romford and Colchester are very similar, hovering around the 90% mark. Vocalisation amongst the Colchester under-30 age group is considerably higher than the other localities, and we can observe an interesting equidistance between the levels of Colchester and Ipswich vocalisations due to the difference in vocalisations between the two age groups in both localities being very similar (9% in Ipswich and 8% in Colchester). This suggests regularity in the increase of vocalisations in the two communities. The Romford statistics present an unexpected result, showing a lower level of vocalisations amongst its young speakers than its older speakers. However, we must also note there is no great difference in the amount of vocalisation either, with the younger speakers vocalising just 4% less than the older ones. 100 90 80 70 60 50 40 30 20 10 0 15-30 yrs 30-65 yrs Romford Colchester Ipswich Fig. 5 – Vocalisation according to age. If we were to assume that a higher level of vocalisation among the younger age group means a change is underway, then the relative equality of the older and younger speakers could signify the development has ground to a halt, with no further changes taking place. However, Both Romford and Colchester have similar levels of vocalisation overall, so we should wonder why, if the development should stop at that level of vocalisation, that this hasn’t occurred in Colchester, where the younger speakers still have a marked 8% increase on their older speakers. The 4% lower vocalisation levels experienced by the younger Romford informants is unexpected, but may well be an anomalous result, perhaps affected by external elements during the interview process; the style of the conversation, how well at ease they felt with the interviewer, etc. It could also be that one of the informants used much less vocalisation overall, thereby inadvertently affecting the percentages. This last hypothesis is supported by Przedlacka (2001:7) as she notes during her observations of speakers of estuary English; “In the present sample, the extent of /l/ vocalisation seems to be an idiosyncratic characteristic, with one speaker using twice as many vocalised tokens as the other.” 5.2.3 Impact of preceding vowel length In general, the results seem to concur with previous studies, in that a preceding long vowel seems a more likely site for vocalisation of /l/. Whilst Meuter’s 2002 results showed roughly equal chances of vocalisation after long or short vowels amongst Colchester 6 year olds, both the Colchester and Ipswich young speaker groups show a distinct preference for a short vowel over a long vowel as the most promoting environment for vocalisation. This would appear to contradict Meuter’s findings that by the age of 10 Colchester children had been converted to the adult system. The preference for short vowels amongst these age groups could be due to an increased amount of parallel vowel shortening taking place in these communities, as noticed in London by Wells (1982). However, it is also acknowledged by linguists that vowel shortening only takes place in very stable consistently high vocalising dialects, and whilst Ipswich can no longer be considered as a low vocalising community, particularly amongst its younger speakers, it is perhaps a pattern we would expect more of Romford, where instead the converse has occurred, and it’s younger speakers are the only young age group across all three localities to have a vocalising percentage lower than their older counterparts (85% among Romford young, compared to 58% for Ipswich old and 74% for Colchester old speakers.) Only the older informants from Ipswich and Colchester vocalise short vowels less often. 100 90 % Vocalisation 80 70 60 50 40 30 20 10 0 ROM OF ROM YF COL OF COL YF IPS OF IPS YF Locality and Age Group Prec Short Prec Long Fig. 6 – percentages of vocalisation with preceding long and short vowels. It is also interesting to note that, looking over the variation in general across the entire sample, whilst the levels of preceding short vowel vocalisations vary widely across both ages and localities, that the levels of preceding long vowels are more or less identical in each locality, with an extremely low percentage difference between age groups. This would suggest that the environment of preceding long vowel vocalisation is a much more stable environment in general. The general preference for preceding longer vowels concurs with the theory held by Sproat and Fujimura (1993) whose tracings show that a long vowel incites a longer dorsal gesture, rendering it more prominent which makes the secondary coronal gesture much more likely to fail, producing vocalisation. 5.2.4 Syllabic /l/ and preceding consonant type In an attempt to be objective in my data collection, I only took examples of syllabic /l/ from within the first 35 tokens of possible dark /l/ environments from each speaker. As a result there are several consonant categories (notably dorsal examples) where there are virtually no tokens from which I can draw conclusions and predict a hierarchy. From the table below (fig. 7) we can, however, see that of all four consonant categories, the glottal examples are the only one to incite 100% vocalisation across the majority of speakers in all three localities. V o c a lis e d T o k e n s :A c tu a l/P o te n tia l/% V o c a lis e d P h o n e tic E n v ir o n R O M R O M C O L C O L I P S I P S P r e c S y lla b ic /l/ O F Y F O F Y F O F Y F L A B I A L 2 /3 /6 7 % 5 /8 /6 3 % 3 /5 /6 0 % 5 /5 /1 0 0 % 2 /3 /6 7 % 4 /5 /8 0 % C O R O N A L 1 /1 /1 0 0 % 4 /4 /1 0 0 % 4 /4 /1 0 0 % n /a 0 /2 /0 %n /a D O R S A L n /a 1 /2 /5 0 % 2 /2 /1 0 0 % n /a n /a n /a G L O T T A L 2 /2 /1 0 0 % 2 /2 /1 0 0 % 8 /8 /1 0 0 % 9 /9 /1 0 0 % 3 /3 /1 0 0 % 3 /4 /7 5 % *n /a s ig n if ie s la c k o fr e le v a n tto k e n s ,n o tv o c a lis a tio n s Fig. 7 – tabulated results of syllabic /l/ vocalisation. The older speakers from Colchester are the highest vocalising overall, with 3 of the 4 categories being vocalised 100%. Amongst Colchester and Romford speakers, preceding coronals and glottals appear to be more popular sites for vocalisation with 100% of tokens having vocalic realisation. In Ipswich, amongst the older speakers, a preceding coronal would appear to be the least favourite environment, with no vocalisation in either of the 2 tokens that presented themselves. A preceding labial appears to be a moderately favoured preceding environment across the board, with it providing the highest level of vocalisation amongst younger Ipswich speakers. If I were to theorise a hierarchy, the data from Romford and Colchester would appear to prefer a preceding glottal/coronal, with a preceding labial being in turn preferred over a preceding dorsal sound. This tallies with Spero’s 1996 data from Colchester, in placing a dorsal as the least likely candidate, but also contrasts entirely with the same study by inverting the order of the two most favoured environments from the Colchester hierarchy. The Ipswich data would also appear to prefer preceding glottal sounds, followed by labial sounds, but in contrast to the data from the two Essex localities, would rank coronal sounds very low. Without any examples of a syllabic /l/ with a preceding dorsal in the entirety of the Ipswich data I collected, it is impossible to say where a dorsal sound would be placed in this order of preference. It has been shown in other studies of Colchester and in the Fenland area of England that, as occurs with the Colchester and Romford results in my investigation, that a dorsal sound has an almost inhibiting effect on the vocalisation process, whilst it would otherwise be thought to be a more promoting environment due to the phonetic production of a dark /l/ involving the realisation of a dorsal gesture before a coronal. Johnson and Britain’s 2003 Fens data shows that consonants that have some connection to the articulation of /l/ (i.e. coronal and dorsal consonants) have a somewhat inhibiting effect on the potential vocalisation, but that those consonants that share no place of articulation with /l/ at all have a more promoting effect. This supports the hierarchy suggested from the Ipswich speakers data, and Johnson and Britain (2004) go further to suggest that the dislike for a preceding dorsal may be a result of an OCP effect, where a preceding dorsal stop for example necessitates a lateral release of the dorsal closure, and this happens prior to the vocalic gesture of the dark /l/. Ladefoged (1993:55) also endeavours to explain the dislike for preceding coronals by claiming that “the air pressure built up during the stop can be released by lowering the sides of the tongue”, which, as a result of lateral plosion, causes a coronal gesture to precede the dorsal, rendering vocalisation very unlikely. 5.2.5 Pre-vocalic dark /l/ vocalisations Colchester has the highest levels of word final pre-vocalic vocalisation across both age groups, with the older speakers vocalising 100% of possible tokens where the dark /l/ occurred word-final pre-vocalically across a word boundary. The Colchester older speakers are shown to vocalise syllabic /l/ when it occurs pre-vocalically with a preceding labial or glottal consonant, whereas a word final syllabic /l/ seems to be prohibiting environment for the younger Romford speakers, where a preceding voiceless labial [p] appears to prevent vocalisation, despite a voiced labial [b] occurring in the same environment and permitting vocalisation. Two examples of vocalisation before the glide [j] in ‘tell you’ were found amongst Romford speakers, and whilst this has been noticed in Colchester speakers independently of this investigation, there were no examples within my data sets for me to examine. However, the fact that there are any examples of vocalised /l/ before [ju] is a contradiction of Gimson (2001:203), who claims that vocalisation, in this environment, is impossible. Interestingly, in both Romford and Colchester, the younger speakers consistently produce less prevocalic vocalisations than their older counterparts, this is especially marked in Romford, where younger speakers produce 57% vocalisations compared to the 88% of older speakers. This could perhaps indicate a trend that pre-vocalic vocalisations will become less popular in these areas over time. My data of vocalisation levels before vowels therefore contrast greatly with Spero’s results from 1996 where she suggests that vocalisation in this environment is on the increase (Spero 1996:65), whereas my data from the Essex localities shows the feature to be in decline in this area, especially so in Romford. However, this phenomenon is seemingly reversed in Ipswich, where the young speakers produce 50% vocalisations of prevocalic /l/, more than double the older speaker rate of 23%. 2 of the 3 vocalised tokens from the older Ipswich speakers were word final intervocalic, and the remaining being a word final syllabic /l/ with a preceding glottal. Pre-vocalic Vocalisation levels 100 % pre-vocalic vocalisations 90 80 70 60 50 40 30 20 10 0 RomOF RomYF Col OF Col YF IpsOF IpsYF Locality and Age Group Fig. 8 – Total percentage of vocalised pre-vocalic tokens. Interestingly the way in which the hiatus between the dark /l/, whether vocalised or not, is realised seems to again contradict other linguist’s theories. Johnson and Britain (2004) discover from their Fenland data that when a vocalisation of a word-final prevocalic /l/ doesn’t occur, they insert a linking [l] to create a hiatus and provide a lower sonority element, showing that they retain an underlying /l/ in their phonology. They also state that those who do vocalise, however, no longer have a linking /l/ available to them, filling the hiatus instead with a [w]. From this they have the assumption that those who do vocalise, in losing their underlying /l/ in this position means they never produce an /l/ in a syllable rhyme. However, my data shows that speakers in all three localities, of both younger and older age groups, have produced both vocalised and non-vocalised /l/s prevocalically, and whilst their use of a clear [l] hiatus filler after a non vocalised /l/, and the [w] only appearing in environments where vocalisation does take place confirms Johnson and Britain’s findings, both forms being used by the same speakers surely disproves their assumption that /l/ never appears in a syllable rhyme of those who vocalise. 5.2.6 Vocalisation in clustered /l/ Clustered consonants seem a very popular site for vocalisation, with the levels of /l/ vocalisation across the three localities appearing to follow a similar distribution pattern as my results found looking at vocalisation before vowels. This environment is again more popular again with older Romford and Colchester speakers, with both groups vocalising 100% of their possible tokens. This pattern is again reversed for the Ipswich data, where the younger speakers vocalise 30% more clustered /l/s than their older speakers. The popularity of clustered /l/ vocalisation can be explained by looking at by observing that other forms of language change often move to simplify the language, and by simplifying consonant clusters, we conserve articulatory effort. This is further qualified by markedness theory which suggests that changes move toward unmarked elements, and in child language, often pointed to as an indicator of unmarked features, the dislike for consonant clusters suggests that they are considered marked features and that any trend would move away from such clusters towards simplification. Locality Young Speakers Older Speakers Romford 96% 100% Colchester 86% 100% Ipswich 94% 64% Fig. 9 – percentage vocalisation of dark /l/ in a consonant cluster. 6. Concluding Comments The region under investigation forms part of what is considered the ‘heartland’ of Estuary English, and with a historical link between the area and London, especially since the slum clearances of the 1960s; it is surprising that dialects from London do not exert a greater influence. However, this is by no means an exhaustive investigation, but the topic remains an interesting one, meriting further study. An additional angle would be to introduce the notion of gender difference into the study, following on from the supposition by Milroy and Milroy (1985) that men are in fact more inclined to use and spread vernacular forms. Also a wider number of participants, as well as a larger amount of tokens taken from each informant, would help make the results more credible, thus potentially avoiding the anomalous results that have surfaced during my analysis. Having the wider ranging data that comes from a larger pool of results would also help in the creation of a more reliable ‘preceding consonant type before syllabic /l/’ hierarchy than is detailed here. Horvath and Horvath (2001:41) state that, from their data, clustered /l/s appear to be the least likely to vocalise, with coda /l/ and syllabic /l/ preferred in Australia and New Zealand respectively. They also acknowledge that the higher vocalising localities prefer syllabic /l/ is preferred over both coda and cluster /l/. This illustrates how the ratios of the three vocalisation types as the most promoting vocalisation environment changes as the general vocalisation levels increase. The three localities studied as the main focus of this project are all high vocalising areas, and certainly in Colchester, the vocalisation of a syllabic /l/ is virtually 100% with all consonant types, proving Horvath and Horvath’s assumption to be well founded. The transition between the absence of vocalised tokens in the SED data taken during the 1950s, and the high levels of vocalisation seen across all three communities is a testament to the speed at which the feature has spread throughout the region. However, my results have shown that it would be wrong to suggest the acquisition of the feature took place purely through the influence of the areas proximity to London. The data from Colchester, in terms of levels of vocalisations, often exceeds that of Romford, where we would possibly have expected a diffused feature to be more frequently closer to the source dialect. Younger speakers from Colchester are also the highest vocalising group overall (see fig. 4) but their preference for a preceding short vowel is in contrast with the general trends discussed in the literature. There are several other results (namely the dislike for syllabic /l/ vocalisation with a preceding coronal in Ipswich, and the lower vocalising levels of younger Romford speakers compared to their older speakers) which appear to be anomalous, but in actual fact serve to illustrate that the integration of a feature into a dialect is a result of a combination of two forms, the incoming innovation, and the traditional form. If the original traditional dialects are different, then the accommodation of that feature will differ from place to place within a region, or as Britain (2002:618) says; “in each case, there will be local outcomes determined by local circumstances.” 7. Index of Figures Fig. 1 (p. 3) Map showing locations of investigated speech localities. (from http://maps.google.com) Fig. 2 (p. 6) The realisation of /l/ in the nucleus position in mid-20th century England. (derived from Orton et al (1978)’s Linguistic Atlas of England) Fig. 3 (p.14) Vocalisation levels from SED data. Fig. 4 (p.15) Vocalisation according to locality. Fig. 5 (p.16) Vocalisation levels according to age group. Fig. 6 (p.18) Percentages of vocalisations with preceding vowel length. Fig. 7 (p.19) Tabulated results of Syllabic /l/ vocalisation. Fig. 8 (p.20) Total percentage of vocalised pre-vocalic tokens. Fig. 9 (p.23) Percentage vocalisation of clustered /l/. 8. References Britain, D: Geolinguistics and Linguistic Diffusion (Unpublished) http://homepages.tesco.net/~david.britain/7.pdf Britain, D (2002): Space and Spatial diffusion. In JK Chambers et al (eds): Handbook of Language Variation and Change. Oxford, Blackwell. (Pages 603-37) Caxton, A (1960): The Suffolk dialect of the 20th Century (2ed). Ipswich, Norman Adlard and Co. Ltd. Gess, R. (2001): On re-ranking and explanatory adequacy in a constraint based theory on phonological change. Ms University of Utah. Gick, B (1999): A gesture based account of intrusive consonants in English. In Phonology 16.1 Giles, S.B & K.L. Moll (1975): Cinefluorographic Study of selected allophones of English /l/. In Phonetica 31 (206-227) Gnanadesikan, A (2004) as cited in Johnson and Britain (2004): LG405 Course Materials, University of Essex. Horvath B. & R. Horvath (2001): A Multilocality study of a sound change in progress: the case of /l/ vocalisation in New Zealand and Australian English. In Language Change and Variation 13 Hernández-Campoy, J-M (1999): Geolingüística: modelos de interpretación geográfica para lingüistas. Murcia, Universidad de Murcia. Johnson, W & D. Britain (2003): L vocalisation as a natural phenomenon. In ERRL 44. Johnson, W & D. Britain (2004): LG405 Course Notes, Colchester, University of Essex. Kerswill, P (2001): Koineisation and Accommodation. In JK Chambers et al (eds): Handbook of Language Variation and Change. Oxford, Blackwell. (Pages 669-702) Kjederqvist, J (1903): The Dialect of Pewsey (Wiltshire). London, The Philological Society. Ladefoged, P (1993): A Course in Phonetics. Fort Worth, Harcourt Brace. Labov, W (1990): The intersection of sex and social class in the course of linguistic change. In Language Variation and Change 2: 202-254 Matthews, W. (1938). Cockney past and present. London, Routledge and Kegan Paul. Meuter, A. (2002): L Vocalisation and labiodental variants of /r/ in the speech of Colchester Primary School Children – The Acquisition of a sound change? MA Dissertation, Colchester, Essex University. Orton, H & E. Dieth (Eds.) (1971): Survey of English Dialects, Volume 3: East Anglia. Leeds, E.J. Arnold. Radford, A et al. (1999) Linguistics: An Introduction. Cambridge, CUP. Spero (1996) L-Vocalisation in Southeastern Britain: MA Dissertation. Colchester, University of Essex. Sproat R. & O. Fujimura (1993): Allophonic variation in English /l/ and its implications for phonetic implementation. In Journal of Phonetics 21:291-311 Tollfree, L. (1999): South east London English: Discrete versus continuous modelling of consonantal reduction. In Foulkes and Docherty (eds): Urban Voices. London, Arnold. 163-184 Trudgill, P (1974b): Linguistic Change and Diffusion: Description and Explanation in Sociolinguistic Dialect Geography. In Language and Society 3: 215-246 Trudgill, P (1986): Dialects in Contact. Oxford, Blackwell. Trudgill, P. (1999): Norwich: endogenous and exogenous linguistic change. In Foulkes and Docherty (eds): Urban Voices. London, Arnold. (124-140) Uffmann, C (2003): Intrusive [r] and optimal epenthetic consonants. Unpublished ms, University of Marburg. Walsh Dickey, L (1997): The Phonology of Liquids. PhD Dissertation UMASS. Distributed by the GLSA. Wells, J (1982): Accents of English. Cambridge, CUP.