C H A P T E R 7

Network Structure and Linguistic Change JAMES MILROY AND LESLEY MILROY

[T]he Belfast research design here depends on the idea of n o r m maintenancej which we have operationalized i n terms of social nenuork, and w i t h i n this model we have distinguished between relatively weak and strong network ünks. I n any real community individuáis and groups w i l l vary in the relative intensity of ties, and this is what makes it possible to compare them in these terms. But behind this there lies an idealization which predicts that i n a communi ty bound by maximally dense and multiplex network ties linguistic change would not take place at all . N o such community can actually exist, but the idealization is impor-tant, because it also implies that to the extent that relatively weak ties exist i n communities (as i n fact they do), the conditions w i l l be present for linguistic change to take place. This perception was partly borne out even in the inner-city research. We noted that very few individuáis had markedly low network strength scores, and furthermore that these individuáis tended to use language much less cióse to the core Belfast vernacular, w i t h a much lower use of the 'close-tie' variants (such as [A] in words of the {pulí) class). T h e idea that relative strength of netiuork tie is a powerful predictor of language use is thus implicit i n the interpretative model we have used throughout; it predicts, amongst other things, that to the extent that ties are strong, linguistic change w i l l be prevented or impeded, whereas to the extent that they are weak, they w i l l be more open to external inñuences, and so linguistic change w i l l be facihtated.

Weak ties are much more d i f f i cu l t to investígate empirical ly than strong ones, and the inst inct of the network-based ethnographer is usually to study relatively self-contained small communit ies that are internally bound by strong

Source: ' N e t w o r k Structure and Linguis t ic Change' in M i l r o y , J (1992) Linguistic Variation and Change: On the Historical Sociolinguistics of English, (Oxford: Blackwell Publishing) pp. 176-91.

92

t^etwork Structure and Linguistic Change 93

links and relatively insulated f r o m outside inñuences. T h e ethnographic work reported in Cohén (1982), for example, focuses on peripheral áreas of the British Isles that have a strong sense of local ' community ' . A l t h o u g h we may surmise that urban situations (such as Belfast) are likely to exhibir lower den-sity and mul t ip lex i ty i n personal ties than remote rura l ones (and are by the same token also likely to be more open to outside inñuences), many stud-ies, both urban and r u r a l , have shown that a close-knit network structure functions as a conservative forcé, resisting pressures for change or ig inat ing from outside the network; conversely, those speakers whose ties to the local-ized network are weakest approximate least closely to localized vernacular norms, and are most exposed to external pressures for change (J. JVlilroy and L . M i l r o y 1985). T h i s second observation suggests that since strong network structure seems to be implicated i n a rather negative way i n l inguistic change, a closer examination of zueak network ties might be profitable.

The d i f f i cu l ty i n studying weak ties empirically means that the quantitative variable of network (which can be readily applied to close-knit communities) cannot be easily operationalized i n situations where the populat ion is socially and/or geographically mobile. T h e networks of mobile persons tend to be loose-knit; such persons f o r m (relatively weak) ties w i t h very large numbers of others, and these are often open-ended, seldom f o r m i n g into close-tie clus-ters. I t is therefore di f f icul t , i n studying loose-knit situations, to produce direct empirical (quantitative) evidence of the k i n d usually used to support sociolin-guistic theories, and indeed (as we noted above) the speaker-innovator cannot easily be directly observed and located. However, we have argued (J. JVlilroy and L . M i l r o y 1985) that the speaker-innovator is a necessary theoretical con-struct i f we are to clarify what is involved i n solving the actuation problem. Therefore, as we are again dealing w i t h an idealization here, we use a mode of argumentation that differs f r o m the usual inductive mode favoured by quan­titative l inguists, and to support the argument we adduce evidence f r o m var i -ous sources.

T h i s evidence is of several different kinds. As is the case so often in network analysis, we find that anthropological and sociological studies of small-scale communities (as in Cohén 1982) are i l l u m i n a t i n g . O n the basis of evidence f r o m a number of such studies, M e w e t t (1982) has observed that class di f -ferences i n small communities begin to emerge over t ime as the propor t ion of multiplex relationships declines (mult iplexi ty being an impor tant charac-teristic of a close-knit type of network structure). Th i s observation, i n addi-t ion to associating social class stratif ication w i t h the dechne of close-knit networks, suggests a f ramework for l i n k i n g network studies w i t h large-scale class-based studies i n f o r m u l a t i n g a more coherent mult i - level sociolinguis-tic theory than we have at present. But we have also derived insights f r o m important work by Granovetter (1973, 1982), who has argued that 'weak' and

94 PART I: LANGUAGE VARIAVON

uniplex interpersonal ties, although they may be subjectively perceived as unimportant, are in fact important channels through which innovation and influence flow from one close-knit group to ancther, linking such groups the wider society. This rather larger-scale aspect of the social function of weak ties has a number of important imphcations for a socially accountable theory of linguistic change and diffusion, some of which we shall briefly outline.

Granovetter's working definition of 'weak' and 'strong' ties is as foUows: 'The strength of a tie is a (probably linear) combination of the amount of time, the emotional intensity, the intimacy (mutual confiding) and the recip-rocal Services which characterise a tie' (1973: 1361). This is probably suffi-cient to satisfy most people's feeling of what might be meant by a 'weak' or 'strong' interpersonal tie, and it fits in fairly well with our indicators for meas-uring network strength in the Belfast inner-city communities. I t also fits in with the principies followed in comparing inner-city with outer-city Belfast (see Milroy, J. 1992, chapter 4): broadly speaking, the former is character-ized by stronger and the latter by weaker ties. Thus, although strength.ofjie is a continuous variable, for the purpose of exposition Granovetter treats it as i f it were discrete, and we need always to bear in mind that we are speaking in relative terms: a tiejs 'weak' i f it is less strong than the other ties against which it is measured. Granovetter's basic point is that weak ties between groups regularly provide bridges through which Information and influence are diffused. Furthermore, these bridges between groups cannot consist of strong ties: the ties must be weak (that is, relatively weak when measured against internal ties). Thus, weak ties may or may not function as bridges, but no strong tie can. This is shown in Figure 7.1.

Strong ties, however, are observad as concentrated within groups. Thus, they give rise to a local cohesión of the kind that we explored in inner-city Belfast; yet, at the same time, they lead paradoxically to overall fragmentation. Clearly, this perception is potentially very illuminating in accounting for dif-ferent language states at different times and places at many levéis of generality, ranging from the interpersonal situations, through dialect-divergent, bilingual and code-switching communities to the very broadest of language situations, and it throws light on the question of convergence and divergence. The model of strong and weak ties presented graphically in Figure 7.2 can be thought of as an idealized representation of (for example) an urban community which consists of clumps connected by predominantly strong ties, which in turn are connected to other clumps by predominantly weak ties, but it can of course represent other kinds of language situation that we might conceive of

The important point (from our perspective) that foUows from all this is that weak inter-group ties are likely to be critical in transmitting innova-tions from one group to another, despite the common-sense assumption that strong ties fulfil this role. For example, Downes (1984: 155) suggests that the

¡\¡etwork Strucwre and Linguistic Cliange

— Strong tie - - - Weak tie

Figure 7.1 Weak ties as bridges

Figure 7.2 Idealized representation of an urban community in which weaker ties are more numerous in middie social

groupings and between the groups

96 PART I: LANGUAGE VARIflJION

nerwork concept is important in developing a theory of l inguist ic di f fusion, but assumes that it is strong ties that w i l l be cr i t ical . T h i s assumption seems to be shared by many linguists who have considered the matter; indeed, as we have noticed, Labov (1980: 261) presents a model of the innovator as an indiv idual w i t h strong ties both inside and outside a local group. Clearly, this conflicts w i t h the arguments presented here, which predict that to the extent that ties are strong, l inguistic change w i l l be impeded, not facilitated.

Granovetter's principie seems at first sight to go against ' common sense', and for this reason we need to expound i t a l i t t le further . First of all , it is likely that weak ties are much more numerous than strong ties, simply because the t ime and energy invested i n the maintenance of strong ties must place an upper h m i t on hovv many i t is possible to have, whereas weak ties require l i t t le effort. Second, many more individuáis can be reached through weak ties than through strong ties; consider for example the bridges set up by participants at academic conferences, which l i n k cohesive groups associated w i t h each ins t i -t u t i o n and thr ou gh which new ideas and Informat ion pass. Conversely, in for -mat ion relayed throu gh strong ties tends not to be innovatory, since persons l inked by strong ties tend to share contacts (that is, to belong to overlapping networks). So they may, for example, hear the same r u m o u r several times. Th is general principie entails that mobile individuáis who have contracted many weak ties, but who as a consequence of their m o b i i i t y occupy a posit ion marginal to some cohesive group, are i n a part icularly strong posit ion to carry in format ion across social boundaries and to diffuse innovations of all kinds.

I n view of the norm-enforc ing capacities of groups bui l t up mainly of strong ties, i t is easy to see why innovators are hkely to be persons weakly l inked to the group. Susceptibility to outside influence is likely to increase i n inverse p r o -por t ion to strength of tie w i t h the group. Where groups are l o o s e - k n i t - that is, l inked mainly by weak ties - they are likely to be generally more susceptible to innovation. We might note that this contention is consistent w i t h the principie enunciated by Labov and K r o c h that innovat ing groups are located centrally i n the social hierarchy, characterized as upper w o r k i n g or lower-middle class ( K r o c h 1978; Labov 1980: 254). For i t is l ikely that i n B n t i s h (and probably also N o r t h American) society the most close-knit networks are located at the highest and lowest strata, w i t h a majori ty of socially and geographically mobile speakers (whose networks are relatively loose-knit) fa l l ing between these two points.

One apparent d i f f i cu l ty w i t h the proposal that innovators are only mar-ginally l inked to the group is i n explaining how these peripheral people can successfully diffuse innovations to central members of that group, who are of course resistant to innovation. One part of the answer here is that cen­tral members often do not accept the innovat ion: henee, for example, the

Metwork Structure and Linguistic Cíiange 97

persistence of regional varieties and m i n o r i t y languages i n strong-tie situa-tions (compare here Andersen's (1986) idea of endocentric dialect c o m m u n i -ties). But to the extent that they do accept innovations, two related points are relevant. First , since resistance to innovat ion is likely to be strong i n a n o r m -conforming group, a large number of persons w i l l have to be exposed to i t and adopt it i n the early stages for it to spread successfully. N o w , i n a mobile society, weak ties are likely to be very much more numerous than strong ties (especially i n urban communit ies) , and some of them are likely to f u n c t i o n as bridges through which innovations flow. Thus , an innovation like the L o n d o n merger between /d, 0/ (as i n brother, thin) and /v, f/ reported i n N o r w i c h teen-age speech ( T r u d g i l l 1986: 54ff.) is likely to be transmitted t h r o u g h a great many weak links between Londoners and N o r w i c h speakers, and T r u d g i l l suggests tourists and football supporters as individuáis who m i g h t contract such l inks. Quite simply, before it stands any chance of acceptance by central members of a group, the links through which it is original ly t ransmit ted need to be numerous (compare Granovetter 1973: 1367). T h u s , the existence of numerous weak ties is a necessary condi t ion for innovations to spread: i t is the quantity as well as the qual i ty of l inks between people that is crucial here.

The second point we need to make i n explaining the success of marginal members of a group as innovators relates more directly to Labov's view of the innovating personality type. As Granovetter suggests, persons central to a close-knit, norm-enforc ing group are likely to find innovat ion a risky activity (indeed i t is probably more i n their interests to main ta in and enforce norms than to innóvate); but adopting an innovat ion that is already widespread on the fringes of the group is very much less risky. There is of course a t ime dimensión involved, and i n this dimensión a point may be reached at which central members begin to accept that it is i n their own interests to adopt the innovation. I n f o r m a l observation of cu l tura l and polit ical innovation suggests that this is generally true. As an example we may cite the final adoption of a margina] cul t (Christ ianity) i n ancient Rome: i t took centuries for this innov­ation to penétrate to the centre. Central members o f a group d i m i n i s h the risk of potential ly deviant activity by adopting (after a lapse of time) an innovation f rom persons who are already non-peripheral members of the group, rather than by direct i m p o r t a t i o n f r o m margináis, who tend to be perceived as devi­ant. T h u s , we can i n this way understand how acceptance - under certain conditions ~ can be a rational strategy on the part of central members of the group.

W i t h i n the network model , therefore, the existence of numerous weak ties is a necessary condi t ion for innovat ion to be adopted. But there must be addi -tional conditions, and at least one of these is psychosocial: this is that speakers f r o m the receptor c o m m u n i t y want to ident i fy for some reason w i t h speakers

98 PART I: LANGUAGE VARIATÍON

f r o m the donor community . Thus , the N o r w i c h speakers cited by T r u d g i l l i n some sense view L o n d o n vernacular speakers as persons w i t h w h o m , i n Andersen's (1986) terms, they wish to express solidarity. Ul t imate ly j for an innovat ion to be adopted, it seems that the adopters must believe that some benefit W themselves and/or their groups w i l l come about through the adop-t i o n of the innovation. T h e cost of adopting the innovat ion i n terms of effort w i l l thus be perceived by the adopters as less than the benefit received f r o m adopting i t . I t also seems that an explanation based on the idea of group identi ty and solidarity is more satisfactory than one that relies on prestige i n a social class dimensión (see J. M i l r o y 1992, chapter 7).

Bearing all these points i n m i n d , it is appropriate now to r e t u r n to Labov's account of the innovator and compare i t w i t h our o w n . T h e most general difference is that Labov's account is about a type of person, whereas ours is abstract and structural , focusing on the nature of interpersonal l inks: i t is based on relationships rather than on persons. We m i g h t describe Labov's innovator as a person who is sociable and outgoing, and who has many friends both inside and outside the local group. Intui t ively , i t seems very likely that Informat ion of all kinds ( inc luding linguistic innovation) can be diffused by such persons, for the reason that they have many contacts. But according to our account, such individuáis could not be near the centre of a close-knit group and at the same t ime have many strong outside ties. M o r e probably, they w o u l d have relatively few multiplex l inks w i t h others, and many of their l inks w o u l d be open-ended and henee low on density; they w o u l d have a pre-dominance of weak l inks, inc luding many that constitute bridges between groups. I n class terms such persons w o u l d probably be mobile, and their profile would therefore fit i n w i t h Labov's view that socially mobile sectors (upper w ork i ng to lower-middle class) are the ones i n w h i c h l inguist ic innov­ation and change are carried. I t seems, however, that this profile is not that of the innovator at al l , but that of an early adopter, and we shall consider this point ful ly i n the next section.

W h a t we have presented here is an abstract model , supported by the insights of Granovetter, which i n effect implies that a c o m m u n i t y characterized by maximal ly strong network ties (and henee m a x i m a l norm-enforcement) w i l l not p e r m i t change to take place w i t h i n i t . Real communit ies , however, con-ta in varying degrees of internal cohesions and varying degrees of openness to outside influence through weak ties. T h e speaker-innovator w i t h i n this model is not a close-tie person, but one who is marginal to more than one (relatively) close-knit group and who therefore forms a bridge between groups across which innovations pass.

Empir i ca l support for our model l ing of the speaker-innovator is provided by Rogers and Shoemaker's (1971) studies of about 1500 cases o f innovat ion

Network Structure and Linguistic Change 9')

in many áreas of Ufe, inc luding , for example, innovations i n agricultura I, educational and technological methods. I n the present discussion, the mosi important principie emerging f r o m this work is the dis t inct ion between the innovator and the early adopter. As the innovator has weak l inks to more thaii one group and forms a bridge between groups, he or she is, i n relation lo the close-tie groups, a marginal indiv idual . Rogers and Shoemaker's s t iu l -ies c o n f i r m the marginal i ty of innovators and fur ther suggest that innova-tors are often perceived as underconforming to the point of deviance. I f this is correct, the innovator does not resemble Labov's (1980) characterizarimi (an indiv idual who has 'prestige' both inside and outside the local group), but actually seems to have more in common w i t h the famous 'lames' of tlie Har lem study (Labov 1972). Conversely, Labov's ' innovator ' resembles wliat Rogers and Shoemaker cali the 'early adopter'.

Early adopters are relatively central to the group and relatively confor iu -ing to the group norms. Once the innovation reaches them, i t diffuses to i l ic group as a whole, and at this stage i t moves into the middle part of the S-curvc structure that is associated w i t h the di f fusion of innovations generally. ThuH, although l inguist ic processes are much more complex than many of the ot l icr processes that have been studied f r o m this point of view, they share this |"iai-tern of d i f fus ion w i t h other kinds of innovation. Later, once the new formn are estabhshed i n the group, they may diffuse f r o m the centre outwards. A t the macro-level, therefore, i t is tempt ing to see these patterns i n broad svvccps of cul tura l and l inguist ic history (the history of Chr is t iani ty comes agaiii lo m i n d ) , but we must leave this speculation aside and r e t u r n to the mattcr in hand, because there seems to be no easy way for empirical studies of chany^e in progress to identi fy i n the data the crucial dis t inct ion between innovators and early adopters.

However, we should again recall that we are not at tempting to dcscii l ic the characteristics of personality types, but of relations between groups and individuáis, and these may vary considerably according to different social and cu l tura l conditions. T h a t is to say that we are not t h i n k i n g of identify i ii)'. some i n d i v i d u a l who lurks around the margins of a group and labelling i i i n i or her 'the innovator' . Ñor is it a case of 'once an innovator, always an inno vator', and i t is obviously true that people who are innovative in some ways may not be innovative i n others. We are t h i n k i n g i n s tructural terms, and so we are concerned w i t h the kinds of relationship between persons that determine the conditions i n which l inguistic innovations can be accepted or rejectcd. Thus , the whole question is relative, just as the def ini t ion of the weakiu-ss of a tie is relative. W h a t is clear, however, is that i f innovations are t ransmii -ted across relatively tenuous and marginal l inks i n fleeting encounters that are perceived as u n i m p o r t a n t , we are unlikely to observe the actuation of a

100 PARÍ ;; LANGUAGE VARIATION

change. However sophisticated our methods may be, we are m u c h more likely to observe the take-up and di f fus ion of the innovation by the more socially salient early adopters.

Bearing these diff icult ies i n m i n d , we now t u r n to some detailed examples i n order to demónstrate how the model developed here affects the interpreta-t ion of hnguist ic variat ion i n speech communit ies . I t is usual to suppose that the di f fus ion of hnguistic change is encouraged by relatively open channels of communicat ion and discouraged by boundaries or weaknesses i n hnes of communicat ion . I n Belfast, howeverj there are many patterns that are dif -ficult to explain i n this apparently common-sense way, and we shall consider two of t h e m here. T h e y are: (1) the social conf igurat ion of the spread of /a/-backing f r o m the Protestant east of the city into the Clonard , a West Belfast Catholic c o m m u n i t y ; and (2) the c i ty-wide younger generation consensus on the evaluation of the (pulí) variable, as against conf l ic t ing patterns in the older generation. T h e backing of /a/ is led by East Belfast males. Figure 7.3 shows this , and i t also shows that the movement of back /a/ into West Belfast is not led by Protestant males in the Hammer , as might be expected, but by the younger female group i n the Catholic Clonard área. T h i s is the group that exhibits the crossover pattern and reverses the generally expected 'sta-ble n o r m ' patterns. I n this group the c i ty-wide female movement away f r o m /a/-backing is reversed: the incidence of /a/-backing i n the group is higher than i n the other older and younger female groups, higher than amongst older females i n the same área, and - surprisingly - also higher than amongst their young male counterparts i n the Clonard . W h e n measured against other groups, these young women are deviant.

F i g u r e 7.3 B a c k i n g of /a/ in B a l l y m a c a r r e t t , t h e C l o n a r d and t h e H a m m e r

Metwork Siructure and Linguistic Change 101

The social barriers that inh ib i t contacts between working-class communities have been well described for many locations throughout the w o r l d (examples are cited by L . M i l r o y 1987), and they were evident i n our inner-city fieldwork even inside sectarian boundaries. Inter-ethnic conflict i n Belfast, however, has had the effect of strengthening the barriers that are present i n all such c o m m u ­nities (Boal 1978). I n fact, the major sectarian boundary i n West Belfast is now marked physically by a brick and barbed-wire structure, which is described by the mi l i ta ry authorities, apparently without intentional irony, as the 'Peace Line'. The puzzle is that an East Belfast pattern can be carried across these boundaries, evidently by a group of young women whose movements and face-to-face contacts have been constrained f r o m a very early age. [TJhere is a long-term shift i n the vowel system towards back /a/, and this dif fusion pattern from east to west is a continuation of i t . T h a t this shift is cont inuing across the i ron barriers (both physical and psychological) that sepárate the Protestant east and Catholic west, is a fact for which we are obliged to seek a principled explanation.

The most accessible, and possibly the only, explanation is one that takes account of weak ties and the dis t inct ion between the marginal innovators and the early adopters. I t seems that the Clonard young women are central members of the group, and so they resemble early adopters rather than i n n o ­vators. T h i s is quite clear f r o m their N e t w o r k Strength score (as reported by L . M i l r o y 1987: 204): they all score extremely h igh on this - much higher than the young Clonard males. T h e i r average score is 4.75 out of a possible máximum of 5.00.

Fur ther personal In format ion about this group points rather clearly to innovation through múltiple weak ties. These young women, unl ike their male counterparts, were i n f u l l employment: they all had regular jobs out-side the Clonard c o m m u n i t y at a rather poor city-centre store. Here they were very likely to be i n weak-tie contact w i t h large numbers of people f r o m all over the city, bo th Catholic and Protestant. Thus , they would be well placed to adopt innovations transmit ted by persons peripheral to their core networks, and as a result exposure to innovatory forms would be frequent. Given the large nuraber of service encounters i n the store, it becomes possible for the weak-tie encounters w i t h back [a] users to exceed greatly the n u m -ber of strong-tie encounters w i t h non-back [a] users. Henee the capacity of innovation-bearing weak ties to compete w i t h , and i n this case overeóme, the innovation-resisting strong ties.

I f we have a theoretical perspective such as the one developed here, which explicit ly predicts that an innovation w i l l be transmitted through (frequent and numerous) weak ties, we have a solution to the problem of explaining how back [a] can diffuse i n this way, and we can present a plausible account of how the innovat ion can appear to j u m p across a barrier of brick and barbed

102 PART I: LANGUAGE VARIATION

wire. If , however, we make the usual assumption that innovations are diffused through strong ties, the pattern is very diff icul t to explain. Yet, i t is only i f we make this strong-tie assumption i n the first place that (a] diffusion appears to be a puzzle at all .

Whereas back [a] diffusion is mainly a change in a phonetic segment, the change of pattern in the (pulí) variable is a change of evaluation (or of agree-ment on norms). This variable is quantified on the basis of a small phono-lexical set consisting of items such as pidl, push, took, shook, foot, which exhibit vowel alternation between [A] and [u]. Al though the [A] variant is recessive, it has very strong affective valúes and is a very salient marker of casual speech between cióse acquaintances. But here we vvish to point out only one th ing - the change i n Consensus on norms over the generations. Whereas the (bag) variable shows consensus across the different groups - oíd and young, male and female - the

.(pulí) variable (shown i n Figure 7.4) shows consensus only i n the younger gen-eration, where it has become a marker of gender-differentiation. The question is: how can this normative consensus come about i n this divided city?

The pattern here is one in which the older groups do not agree on the gender mark ing in use of the ' in-group ' variant [A ] . I n Ballymacarrett , males favour [A ] , but i n the two West Belfast communities, gender preference is reversed: the [A] variant is favoured by the females. The younger groups.

% ( A ]

75 r

50

25 • Clonard

Ballymacarrett

' Hammer

Men 40-55

Women 40-55

Men 18-25

Women 18-25

F igure 7.4 D i s t r i b u t i o n of t h e (pulí) va r i ab le ( p e r c e n t a g e of [A ] var iants a r e s h o w n ) by age, sex and á r e a in inner-c i ty Be l fas t

Network Structure and Unguistic Change 103

however, show the same pattern i n all three communit ies: i n all cases [A] preference is stronger among males and weaker among females.

Again the puzzle is to explain how young people l iv ing i n closed commu­nities, whose outside links are quite tenuous, could reach cross-community consensus on the social va lué to be attached to the two variants of (pulí). I n their parents' youth there was greater freedom of movement, and people fre-quently formed friendships across regional and sectarian divisions; however, since the beginning of the c ivi l disorder in 1969, people have been much less able to form strong ties outside their communities. Yet, despite this, the absorption of the (pulí) variable into the regular socio-linguistic structure of Belfast vernacular has continued unhindered. Again , it is only i f we accept that weak ties are the normal channel for the diffusion of innovations that the apparent paradox is resolved.

I n these examples, we have selected instances based on extensive quanti­fied informat ion which is very fine-grained and which is ful ly accountable to the data, but the general pattern here had already become evident f rom observation of other cases, and the (pulí) variable can be regarded as test-ing out a hypothesis that had already been formed. There are many other examples involving different dimensions o f variation ( including, for example, phonological mergers) that can be observed fairly easily and that appear to show this general pattern of consensus in the inner-ci ty younger generation. Indeed, once you are 'clued i n ' to the possibilities (especially w i t h regard to gender-differentiation), it is remarkable how readily you can observe the trends in everyday encounters. Perhaps the most dramatic of these trends is the progressive loss of localized lexical items and reduction of phono-lexical variable sets of the (pulí) type (J. M i l r o y 1981). Another example is the three-way merger (or apparent merger) of words of the type furifirifair, which are very cióse to being ful ly merged amongst younger speakers. The few elderly speakers that we studied (around 70 years oíd in 1975), however, exhibit a three-way differentiation, and middle-aged speakers often have a two-way differentiation. As indications of this greater consensus amongst younger people had already been observed before we started our quantitative analysis, we spoke i n terms o f ' t he rise of an urban vernacular': the first research appli-cation in 1975 proposed the hypothesis that we were witnessing in Belfast a fairly early stage i n the development of a focused urban vernacular, in which there is a generally observable trend towards greater consensus on norms.

However, we have also emphasized i n this chapter the psychosocial barrier of the sectarian difference i n Belfast, which we might expect to inh ib i t the trend towards consensus, and from the beginning of our research we natu-rally wished to discover whether the ethnic difference was consistently and reliably reflected i n language. I n our pi lot research, therefore, we looked at two East Belfast communit ies, one Catholic and one Protestant (Catholics

104 PART I: LANGUAGE VARIAVON

being a small minority in this part of the city). In our analysis of the tapes, however, we could find no appreciable differences between the two groups: the Catholics spoke with an East Belfast accent (including back varieties of /a/) just as the Protestants did, and their speech was more similar to East Belfast Protestants than it was to West Belfast Catholics. Subsequently, after comparing different communities very fully in our inner-city study, we were able to State, rather cautiously, that 'there is as yet no persuasive evidence to show that the two ethnic groups in Belfast (and Ulster) can be clearly identi-fied by differences in accent' (J. Milroy 1981: 44). Indeed, it seems that those features of differentiation that in the past could have been seized upon as eth­nic markers, have been abandoned in favour of greater inner-city agreement on norms of age, sex and contextual style. I n general, the Catholic immigrants arrived in the city later than the Protestants and brought from mid and west Ulster a number of features (such as palatalization of initial [k]) that coiild have been used to reinforce differences, but this does not seem co have hap-pened. Both groups seem to be moving in the same direction in the younger generation even though there may be divergent movements in small details; similarly, both groups appear to evalúate variants in much the same way (and this evaluation is often quite divergent from 'standard' evaluations).

REFERENCES

Andersen, H. 1986. Center and periphery: Adoption, diffusion and spread. Paper delivered to the Conference on Histórica! Dialectology. Poznán, Poland.

Boal, F. W. 1978. Territoriality on the ShankilJ-Falls divide, Belfast: The perspective from 1976. In D. A. Lanegran and R. Palm (eds.) An Invitation w Geography. 2nd edition. New York: McGraw Hül. 58-77.

Cohén, A. (ed.). \9S2. Belonging. Manchester: Manchester University Press. Downes,W. 1984. Language and Society. Bungay, Suffolk: Fontana. Granovetter, M . 1973. The strength of vveak ties. American Journal of Sociology 78:

1360-1380. Granovener, M . 1982. The strength of vveak ties: A network theory revisited. In P. V.

Marsden and N. Lin (eds.) Social Structure and Neiimrk Analysis. London: Sage. Kroch, A. 1978. Toward a theory of social dialect variation. Language in Society 7:

17-36. Labov, W. 1972. Language in ihe Inner City. Philadelphia, Pennsylvania: Pennsylvania

University Press. Labov, W. (ed.). 1980. Locating Language in Time and Space. New York: Academic

Press. Mewett, P. 1982. Associationai categories and che social location of relationships in a

lewis crofting community. In A. Cohén (ed.) Belonging. Manchester: Manchester University Press. 101-130.

Mñroy,J. 1981. Regional Accents of English: Belfast. Belfast: Blackstaff.

Network Structure and Linguistic Change 105

Milroy, J. 1992. Linguistic Vanation and Change: On the Historical Sociolinguistics of English. Oxford: Blackwell.

Milroy, J. and L. Milroy 19S5. Linguistic change, social network and speaker innovation. Jowrna/o/Lin^uúricj 21: 339-384.

Milroy, L. 1987. Language and Social Networks. 2nd edition. Oxford: Basil Blackwell. Rogers, E. M. and F. F. Shoemaker. 1971. Communication of Innovaiions 2nd edition. j

New York: Free Press. Trudgill, R 1986.The apparent time paradigm: Norwich revisited. Paper presented at

the 6th Sociolinguistics Symposium, University of Newcasüe uponTyne.