diglossia revisited

20
Diglossia Revisited Peter Hawkins Queen Margaret College, Edinburgh ABSTRACT Ferguson's classic description and def'mition of diglossia was later extended by Gumperz and Fishman to other multi-lingual and multi-dialectal situations which are in some ways similar. However, all these accounts are based on uses of the two codes (the 'High' and 'Low' varieties) rather than on the linguistic relationships between the codes themselves. When the latter form the basis of the def'mition, "true' diglossia of the Greek or Arabic kind can be distinguished from multi-lingual, standard-nonstandard dialect, and possibly Creole, situations by certain critical differences. In true diglossia, the L is a natively-spoken, relatively independent variety for which descriptive grammars have been (or can be) written, while H is not a single form but varies indef'mitely and its description can best be derived from L by 'purification' rules (a 'bottom-up' description). This is the mirror-image of a dialect situation, where usually it is the H which is a natively-spoken, independent variety, and L exists in no 'pure' form but is indefinitely variable, its description proceeding from a basis of the standard (a 'top-down' description). This 'process' model of diglossia has advantages over the traditional 'static' model in accounting for the multiplicity of intermediate or 'mixed' varieties which arise, and which are trltimately more significant than 'pure' H or 'pure' L. In an investigation of Modem Greek diglossia, thirty subjects, 15 relatively 'educated' and 15 'uneducated', were tested by a version of Cloze procedure to assess competence in the use of H. All speakers, even the educated ones, were found to use H inconsistently, with frequent switching to L forms. INTRODUCTION The aim of this paper, which is based on research carried out by the writer in Salonica, Greece, is twofold: first, to examine differences between educated and uneducated speakers in knowledge of the High variety in a diglossic situation, and secondly, to investigate the relationship between the High (H) and Low (L) varieties from the native-speaker's point of view. As an outcome of this investigation, we must reconsider the nature and def'mition of diglossia. We shall also be testing,

Upload: peter-hawkins

Post on 17-Sep-2016

232 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Diglossia revisited

Diglossia Revisited

Peter Hawkins

Queen Margaret College, Edinburgh

A B S T R A C T

Ferguson's classic description and def'mition of diglossia was later extended by Gumperz and Fishman to other multi-lingual and multi-dialectal situations which are in some ways similar. However, all these accounts are based on uses of the two codes (the 'High' and 'Low' varieties) rather than on the linguistic relationships between the codes themselves. When the latter form the basis of the def'mition, "true' diglossia of the Greek or Arabic kind can be distinguished from multi-lingual, standard-nonstandard dialect, and possibly Creole, situations by certain critical differences. In true diglossia, the L is a natively-spoken, relatively independent variety for which descriptive grammars have been (or can be) written, while H is not a single form but varies indef'mitely and its description can best be derived from L by 'purification' rules (a 'bottom-up' description). This is the mirror-image of a dialect situation, where usually it is the H which is a natively-spoken, independent variety, and L exists in no 'pure' form but is indefinitely variable, its description proceeding from a basis of the standard (a 'top-down' description).

This 'process' model of diglossia has advantages over the traditional 'static' model in accounting for the multiplicity of intermediate or 'mixed' varieties which arise, and which are trltimately more significant than 'pure' H or 'pure' L.

In an investigation of Modem Greek diglossia, thirty subjects, 15 relatively 'educated' and 15 'uneducated', were tested by a version of Cloze procedure to assess competence in the use of H. All speakers, even the educated ones, were found to use H inconsistently, with frequent switching to L forms.

I N T R O D U C T I O N

The aim o f this paper , which is based on research carried out by the writer in

Salonica, Greece, is twofo ld : first, to examine differences be tween educa ted and

uneduca ted speakers in knowledge o f the High variety in a diglossic s i tuat ion, and

secondly , to investigate the relat ionship be tween the High (H) and Low (L) varieties

f rom the nat ive-speaker 's po in t o f view. As an o u t c o m e o f this invest igat ion, we

must reconsider the nature and def 'mit ion o f diglossia. We shall also be test ing,

Page 2: Diglossia revisited

2 Language Sciences, Volume 5, Number 1 (1983)

within a Greek context, two apparently irreconcilable hypotheses about the diglossic relationship which have hitherto been argued in the context of Arabic, but which could equally apply to Greek (and other cases of diglossia).

In a well-known paper, Ferguson (1959) outlined nine criteria for recognizing a diglossic situation. He illustrated the criteria by taking four cases which he regarded as canonic examples of diglossia: Arabic, Modern Greek, Swiss German and

Haitian Creole, in each of which a 'High' form of the language co-exists with a 'Low' form. Diglossia is defined as "a relatively stable language situation in which, in

addition to the primary dialects of the language. . , there is a very divergent, highly codified superposed variety [which is] the vehicle of a respected body of l i terature.. . is learned largely by formal education, and is used for most written and formal spoken purposes but n o t . . , for ordinary conversation."

The difference between a diglossic situation and merely 'stylistic' variation

within a language is a difference both of degree and of kind. The difference of degree is reflected in Ferguson's phrase very divergent: in all the canonic cases of

diglossia there are sharp differences between H and L, differences which cannot be due to the subject-matter, participants, setting, etc. but which seem to reside in the linguistic codes themselves, and which are all-pervasive, affecting the syntax, the

vocabulary, the inflectional morphology and the phonology and orthography. The

differences in Greek between Kathar~vousa, the 'High' variety, and Dhimotik[ (Low) can be illustrated by the parallel text below:

K D K D K D

o pat/r tu ~.fike fin p61in ke fit_hen is tas o pat6ras tu Lfise tin p61i ke tithe stin Athlnas, 6pu 6zisen os m6tikos pol~ 6ti, peripu Athina, 6pu 6zise san m6tikos pohl chr6nja, peripu triLkonda, m6chri tu than~tu tu tri~nda, m6chri to th~nat6 tu

'His father left the town and went to Athens, where he lived as a foreigner for many years, about thirty, until his death.'

The difference of kind between diglossia and stylistic variation lies in the fact

that, in diglossia, both varieties exhibit a range of styles, and furthermore the range

of each may overlap. H and L do specialize their functions to some extent but L is

not confined to spoken, informal or casual styles; it can be used for literary purposes,

for education and for official language. 2 Conversely, H has its own range of styles; it is used not only in strictly formal contexts (legal language, church services, etc.)

but is found, in relatively 'mild' versions, in journalism', higher education, etc.

However, it is true to say that the range of styles encompassed by H is more limited than the range for L and that the H range is concentrated at the 'formal' end of the spectrum and is almost entirely confined to the written medium.

Page 3: Diglossia revisited

Diglossia Revisited 3

The use of the term diglossia was extended by, for example~ Gumperz (1971) and Fishman (1971), who have pointed out that diglossic situations may involve two or more languages if, within the same speech community, one language takes the functions of H and the other the functioris of L. Rubin (1968) describes such a situation in Paraguay, where Spanish functions as H in complementary distribution with Guarani as L; Denison (1971) reports a trilingual situation of a similar kind from Northern Italy, and Abdulaziz-Mkilifi (1972)examines diglossic relationships between English, Swahili, and the local vernacular in Tanzania. Diglossia is, therefore, not a uniquely monolingual phenomenon but has affinities with bilingual and multilingual situations.

HYPOTHESES

In Greek diglossia Dhimotikl (henceforth D) must be seen as the dominant variety. It is the one used for everyday purposes and covers virtually the entire spectrum of styles as far as spoken language is concerned. It is the variety which the child acquires first and which he masters natively. Kathardvousa (henceforth K) is taught at school and is learnt via education, becoming the medium of instruction at some of the higher levels. It is also acquired via what one might call 'secondary' exposure, viz., from newspapers, news bulletins, official forms and documents, etc., this type of acquisition being essentially passive. We can suggest, in view of the teaching of K in schools, that educated members of the society are likely to acquire more knowledge of K, and greater ability to use it, than the less educated. Since K is still widely used for official and legal purposes, knowledge of K will, in itself, offer opportunities which are denied to those who do not possess it.

The present study seeks to confirm this hypothesis, and to measure the extent

of the difference between the two groups, which will in turn reflect the difference between being actively taught (and taught in) K, vs. passive acquisition via secondary sources .

The fact that L is acquired natively and H only at second-hand appears to contradict one of the most important of Ferguson's criteria. Discussing the question of standardization, Ferguson argues that a major difference between the two varieties

lies in the fact that H exists in a stable standardized form established by many years of tradition, while L is unstable and exhibits wide variation in grammar, vocabulary and phonology. H has the backing of "grammars, dictionaries, treatises on pronuncia-

tion, style and so o n . . . by contrast, descriptive and normative studies of L are either non-existent or relatively recent." This implies that H is the well established standard, the norm for the language as a whole, from which a variety of L's deviate in greater or lesser degree.

Such a view is challenged by Kaye (1972), who argues that, in fact, the exact

Page 4: Diglossia revisited

4 Language Sciences, Volume 5, Number 1 (1983)

opposite is the case. In Arabic, it is the various L's which are stable, while the H,

far from being stable, is indefinitely variable and can be characterized as ' i l l-defined'.

The phonology, morphology and syntax of H can, according to this view, be

correlated with (i.e., partly predicted by) the linguistic and sociolinguistic

background of the speaker. The speaker's L, his level of education, the given speech

situation, the at t i tudes of the speaker, and so on, all contribute to the particular

linguistic forms which will emerge in his use of H (1972: 40). We should, therefore,

regard L as the norm, and the phonology, morphology and syntax of H as derivable

from it.

We, thus, have two opposed views of the relationship between H and L in a

diglossic situation. It will clearly be of some interest to test these views with respect

to Greek, whose diglossia is parallel in many ways with that of Arabic. The opposing

hypotheses can be summarized as follows:

1. There is an estabhshed norm for the pronunciation, grammar, and vocabulary of K which allows variation only within certain Limits. D vanes widely in these respects. K is the norm from which D devmtes.

2. D is a well-defined system with norms of pronunciation, grammar, and vocabulary which are subject only to the same kinds of regional, social, and stylistic variation normally found in most languages. K is ill-def'med and variable. K forms can be predicted from D rather than vice-versa.

SAMPLE

The 'educated" group of subjects consisted of fifteen third-year students (six

male, nine female) from the Department of Psychology at Salonica University. The

"uneducated' sample consisted of fifteen subjects who had completed not more than

three years at high school, and who in most cases had had no formal education.

They were chosen so as to be roughly representative of different age groups, occupa-

tions and sex ( they included a policeman, two hotel porters, a salesman, a carpenter

and several housewives and pensioners).

The educated group had passed through high school during tbe years of the junta

(1967-74), which was the period when, for political reasons, the use of K was

actively promoted, particularly as the medium of education and in public life

generally. This factor should increase, if anything, the difference in knowledge o f

K between the two groups.

METHOD

Knowledge of K was tested by presenting the subjects with texts into which

blanks had been inserted; the subjects were asked to f'fll the blanks with an i tem they

Page 5: Diglossia revisited

Diglossia Revisited 5

considered appropriate. They were told that the investigator was interested in 'the language question' in general and in differences between K and D in particular. The texts were labeled simply as Texts 1, 2 and 3.

Of the three texts, which were taken from popular novels, one was written in K, one in D and the third in an intermediate variety which, for the time being, we can call 'simple K'. Knowledge of K was to be investigated via responses to the 'K"

text and the 'simple K' text; the D text was included partly to make a balanced variety of texts for presentation, and partly as a control in case subjects experienced difficulties due to the nature of the task itself. The texts were chosen so as to be

roughly comparable in subject-matter and, more importantly, in ease of comprehension, in order to minimize differences which might be caused by the topic

rather than the language. For these reasons, passages of conversational dialogue were chosen.

The K text, taken from a mid-nineteenth century novel, 3 is written in a fairly 'pure' or 'strict' variety. In addition to the usual markers of K, such as addition of

final -n in a number of inflectional endings, and the use of is and dhia (prepositions of location and purpose) for s and/a (e.g., is tin pblin (K) vs. stin pbli (D) ' to (at, in)

the city'), we f'md in this text such K features as the preposition en plus dative case

(corresponding to D me plus accusative), and K vocabulary items like tkift /(D t6tia); thus en ti~fti aplbn'n" vs. me t~tia aplbtita 'with such simplicity'. Features of K syntax include extensive use of participial clauses and genitive constructions. 4

The second text is written in straightforward Dhimotiki, marked particularly

by the absence of f'mal -n, the use of s for is and ja for dhia, ki as well as ke ('and'),

D forms of noun, verb and adjective suffixes, the omission of prothetic vowels (D

lighi vs. K olighi 'few'), and so on.

Text 3, from a story by the very popular author Papadiamantis, is the most interesting' linguistically, since its language is neither 'pure' D nor 'pure' K, but a

mixture. The syntax and most of the vocabulary are essentially D, while the

morphology and orthography are distinctively K. The choice of language is closely related to the text's function; a short story written for a popular audience. "The

author has aimed "for a style which, in terms of syntax and vocabulary, is close to the everyday spoken language of his intended readers, but which at the same time

exhibits a degree of 'purification' essential to a work of literary merit. The language must be impressive without becoming incomprehensible. Markers of K are present

in abundance: t'mal -n, is and dhia, iota subscripts in the orthography (in certain verb

suffixes), prothetic vowels (e.g., opiso for D plso 'behind'), the past-tense augment. Verb and noun suffixes appear in their K form, e.g., eparusi-6sthi for D parusi- [lstike ('he was present') and imdras for D mkres ('days', accus, pl .). Them are even K vocabulary items such as ikia (for D splti, 'house'), and 6fthalmos (D rr~ti, 'eye').

These markers give the language the outward appearance of K but the style

Page 6: Diglossia revisited

6 Language Sciences, Volume 5, N u m b ~ 1 (1983)

remains D at a more fundamental level. K syntactic constructions are avoided, and

the vocabulary, apart from a number of K words whose meaning would be widely known, is basically D. One interesting feature of the vocabulary is the inclusion of

several colloquial items which would not be allowed in a strictly K environment, words such as ghms~za ('witch'), ghurunop6dharin ( 'harpy'), and kism~ti ('fate').

The author often puts these words in inverted commas, as if to apologize for their disrespectable origins.

This text is then a good example of an intermediate variety. The language is

basically D but, through the judicious use of many superficial markers of 'purifica- tion', succeeds in giving the appearartce of K.

Blanks were inserted into these three texts at points where the subjects would, as far as possible, be forced to choose between a K response and a D one. In the main, these choices involved morphological elements, such as noun and verb suffixes, and a smaller number of lexieal items. We can illustrate this with reference to the

parallel text cited in the introduction above; o pat- tu would aim to force a choice of suffix between patlr (K) and patkras (D); 6fi- would require completion with -ke (K) or -se (D); m~chri t- thanat- would (hopefully) force a choice between m~chri plus the genitive (K) and mkchri plus the accusative (D); ~zise poldz- would aim to

elicit a lexical choice between ~ti (K) and chr6n/a (D). Our predictions can now be summarized as follows: ( t ) in the responses to the

D text, we expect few or no differences between the educated and uneducated

subjects; (2) in the responses to the K text, the educated group will use more K items, and fewer D, than the uneducated group, by virtue of their greater knowledge

of K; (3) in the responses to the 'mixed' text, both groups will use more D than they

used in the K text, i.e., both groups will shift towards D, since this text is less clearly 'K' in nature than the other.

CODING OF RESPONSES

The responses were classified with the help of two native-speaker informants.

Four categories of response were recognized: K, D, mixed, and neutral. Mixed responses, comprising about 4 percent of the total, contained a feature which was a

marker of D, as well as one which marks K. Thus m{numen 'we stay' has final -n

as a marker of K, but -u- marks a D form; the proper D form would be minume and the proper K form minomerL Likewise, apdndisen 'he answered' mixes final -n (K)

# .

with lack of augment (D). There were other types of mixture, such as the addition

of iota subscript (K) to a thoroughly D word like p~i ('go'), or the addition of the

K suffix -as to the stem turfst- ' tourist', a word which would only occur in D.

Neutral responses, accounting for a rather larger proportion (16% overall), were

items whose occurrence is common to both K and D, and which could, therefore,

Page 7: Diglossia revisited

Diglossia Revisited 7

not be counted as K or D responses. The test was designed to minimize the like-

lihood of neutral responses, by inserting blanks at points which would force a choice

between K and D wherever possible. In some cases, however, particularly where a

complete lexical item was deleted, subjects responded with an item which, although

appropriate to the context, did not match the target (or its equivalent in the other (K

or D) variety). If these non-target responses had K or D markers, they were classified accordingly; if not, they were classified 'neutral'.

RESULTS

None of the subjects experienced difficulty in completing any of the texts, as

indicated by the small percentage (2.8%) of the blanks which were left unfilled.

Table 1 gives the results for both groups of subjects combined, the responses being

classified according to the four categories outlined above; each text is presented

separately.

Table 1

No. of Responses: All Subjects Combined

K D Mix. Neut. Blank Total

(i) K text 374 294 33 203 26 930 % 40.2 31.6 3.6 21.8 2.8 100.0

(ii) D text 23 1326 7 - 24 1380 % 1.7 96.1 0.5 - 1.7 100.0

(iii) 'Simple K' text 472 644 61 151 22 1350 % 35.0 47.7 4.5 11.2 1.6 100.0

Table 1 shows that, as predicted, the responses to the D text are uniform and

consistent: 96.1% are D responses, showing that the subjects are in agreement in

assessing the language of the text as a whole and in their ability to respond appro-

priately. The incidence of K (1.7%) and 'mixed' responses (0.5%) is negligible.

By contrast, responses to the K text show substantial inconsistency. Only 40.2%

of the responses consist of K items, while 31.6% are D and 3.6% mixed. Adding the

D and mixed responses, we see that the non-K responses at 35.2% almost equal the

K responses at 40.2%.

The "simple K' text produced a divided set of responses fairly similar in

percentages to the K text. Here, however, the proportion of D items is higher, as

we anticipated. D responses at 47.7% now outweigh K responses at 35.0%, and the

number o f mixed responses is, at 4.5%, slightly higher than for the K text (3.6%).

This indicates that the subjects have switched towards D in response to a text whose

Page 8: Diglossia revisited

8 Language Sciences, Volume 5, Number 1 (1983)

language is less strongly K.

Differences between the educated and uneducated subjects can now be

examined. Table 2 presents the numbers and percentages of responses to the K text for each group.

Table 2

No. of Responses for Each Group Separately: K Text Only

(a) Student %

Co) Non-student %

K D

207 141 44.6 30.2

167 153 35.9 32.9

Mix. Neut. Blank Total

I0 94 13 465 2.2 20.2 2.8 100.0

23 109 13 465 5.0 23.4 2.8 100.0

It appears from Table 2 that the educated subjects achieved a higher proportion

of K responses m the K text, at 44.6%, than the uneducated at 35.9%. Conversely,

the former used fewer D responses (30.2%, against 32.9%) and fewer 'mtxed' responses (2.2%. against 5.0%).

The differences between the groups were subjected to a number of statistical

tests. The overall variation between the students and non-students was significant

by a X 2 test (X 2 = 10.99, p ~ . 0 5 ) , which takes into account the difference in use o f

D as well as in use of K, but the more reliable parametric text, the t-test, which

measures use of K vs. use of D, separately, and takes each individual subject's response

into account, showed no significant differences between the groups. We may

conclude that on the whole the educated subjects responded more sensitively to the

K text than the uneducated, but that their knowledge of K is not significantly greater

- a surprising result, in view of our prediction that the university students would

be more competent in K, by virtue of the formal instruction they had received.

There was not a single student who gave consistently K responses to the K text

throughout; the best performance had 23 K responses and only 2 D (out of 31

response items), and there were 4 students (out of the 15) who had fewer than six D

responses, but the worst student had only one K response and 24 D responses.

Among the non-students, the best performance achieved 19 K and only 3 D items,

while the worst had 3 K and 13 D items. The 'mixed' responses are a good measure

o f incompetence, since these were responses which included a feature of K and a

feature of D within the same lexical item, e.g., agribfsume ('we shall corrupt ') , the

proper K form being agri-bfsomen and the ordinary D form agri-~psume. The

students were better at avoiding 'mixed' items like this: they had only 2.2%,

compared with non-students 5.0%, and 8 students avoided mixed items altogether, compared with only 2 non-students (X 2 = 3.75, p = .06).

Page 9: Diglossia revisited

Digiossia Revisited 9

CODE SWITCHING

Table 2 shows that .for both groups o f subjects, D responses formed a considera-

ble proportion o f the responses to the K text. This implies the possibility of a great

deal of code-switching from D to K and K to D. In theory, the mixture of K and D

responses could or could not involve code-switching; there are three possible

strategies:

1. Subjects might begin to respond with items which come most easily to them (D items), but later, realizing that the text is in K, they switch to K and, then, maintain K responses.

2. "Subjects might begin to respond in K, in accord with the language of the text, but 'lapse' into D at some subsequent point.

3. Subjects might switch between K and D throughout.

An 'index of switching' was, theretore, devised to ascertain the type o f strategy

followed. The index is based on the number of times each subject switches from K

to D or from D to K. The results are as follows: during the course of the 31

responses made by each subject in the K text, the average number of switches was

9.98. The educated subjects proved rather less inconsistent, with an average o f 8.6

switches, than the uneducated with 11.3. The most consistent educated subject

switched only twice (this was exceptional); the most consistent o f the uneducated group switched seven times. Thus, in both groups, the level of switching is high and

is maintained throughout the text. Code-switching is found not only in the subjects' responses but in the language

o f Text 3 itself, where we find a number o f obviously D forms side by side with

corresponding K forms. For example, the K form embrbs ('in front of ') is followed

a few lines later by the D form brostd; K tis mitrbs appears alongside D tis rnit~ras

' o f his mother ' ; the locational preposition usually appears in its K form is but the D

form s- occurs in andikri sta 'opposite the'. In one remarkable example, the negative

particle appears as rni in one clause and as min (i.e., with final -n) in the following,

virtually identical, clause:

dhiatl na mi to 6chi sp~si, dhiatl na min to ~chi k~ipsi 'why hadn't she broken it, why hadn't she burnt it'.

These examples were all found within less than two pages of the text. Clearly,

even authors, who have time to consider their choice of words, are subject to

variability and cannot be fully consistent in their application of the 'purification'

devices.

Page 10: Diglossia revisited

10 Language Sciences, Volume 5, Number 1 (1983)

D I S C U S S I O N

The expected differences in knowledge of the use of H between the educated and uneducated groups of subjects were not strongly confirmed. While the student group as a whole was found to be more sensitive than the non-students in responses to the K text (X 2 = 10.99, p < .05), usage of K items on an individual basis did not" differ significantly. This is a surprising finding in view of the instruction both in K and through the medium of K, which the students had received as part of their education. One possible explanation, suggested by the fact that one student achieved only one K response out of 31 possible, is that D holds strong covert prestige (cf. Labov 1972: 249) and some students, therefore, felt antipathy towards K and deliberately responded with D forms. Such an explanation would be consistent with Hudson's concept of the individual using language to create "acts of identity in a multi-dimensional space" (1980: 13, 27). However, the explanation has to be rejected because the results show that all students used K to some extent; not one of them consistently gave D responses throughout. The most likely conclusion to be drawn from our results is that the formal teaching of K in high schools has little effect, s and that knowledge of K derives, for both groups, from 'secondary' exposure via reading (of literature and joumals), dealings with official forms and documents, and listening (e.g., to news bulletins and formal speeches). 6 Greek investigators of diglossia in education, such as Frangos (1972), have reached substantially the same conclusions.

The results also indicate that Ferguson's assumption of K as a stable, independent variety is untenable, if based on native-speaker knowledge. The present study provides many examples of the wide range of possible K variants for any particular lexical (or grammatical) item. It seems that the native-speaker has available a range of devices or strategies for 'purifying' a linguistic item which, in" comoination, produce a multiplicity of possible variants. To illustrate this we can give three examples (two verbs, one noun) from one of the texts. The D form is given first, followed by a set of increasingly 'pure' variants, with a description of the purificatory devices:

I.

2.

'he was present'

parusihstike (D form) eparusiastike (addition of augment) eparusihsthike (augment; t "---~0) eparusihsthi (augment; t "~0 ; K suffix)

'he was able'

mbo~se (D form) imbonJse (addition of augment)

Page 11: Diglossia revisited

Diglossia Revisited 11

imbbri (augment; K suffix) idhinato (augment; K suffix; K lexical item)

3. 'days' (Accusative, plural)

rd~res (D form) im~res (addition of prothetic vowel) imbras (prothetic vowel; K suffLx)

The examples show .that, alongside a relatively stable D form, there exists a range of intermediate variants, representing differing degrees or stages of purification. Similar evidence is given in detail by Householder (1962) and Browning (1969). As

with Kaye's account of Arabic diglossia, the H turns out to be not a unified, stable norm, but a range of variants. So the stability which Ferguson claims for it is derived only from the body of literature written in H, and the rules for writing H which are enshrined in grammars and manuals of style; H achieves only 'formal standardiza- tion' (Stewart 1968a). In the Greek case, even this stability is weakly-based, since grammars of K, being prescriptive (they cannot be an account of what native- speakers actually do, or used to do), tend to vary according to the predilections of the grammarian. 7

MEANS OF PURIFICATION

The verb and noun examples given above illustrate some typical means of achieving purification. Some devices are, from the native-speaker's point of view,

relatively simple and easily applied, such as the addition o f - n , augments, and

prothetic vowels. Others are more complex, involving either a change of lexical item

(e.g., dhiname for mborb in 2 above) or the use of K suffixes - more complex, because the speaker needs to know the correct noun and verb paradigms - or a K

syntactic construction. In general, we shall expect the simpler changes, those

involving orthography, phonology, f'mal -n, augments, etc., to be the most widely

employed in a K-demanding situation, and the more complex changes to be less well-

known and more rarely used. This was confirmed in our study; the average incidence

of K items in Text 1 is 40.2% but the incidence of K vocabulary items (where choices were possible) was only 6%. In Text 3, the proportions are 35.0% and 13%,

respectively. The same trend is observable in the language of Text 3 itself, where the writer, Papadiamantis, has used a K orthography and morphology but retains D forms at the more complex levels of lexis and syntax.S

TWO MODELS OF DIGLOSSIA

We can now outline two possible models for the relationship between H and L

Page 12: Diglossia revisited

12 Language Sciences, Volume 5, Number 1 (1983)

in a diglossic situation. The static model assumes the existence of two varieties of the language; K and D, in the Greek case. This is the model which most observers,

e.g., Mirambel, Ferguson, Householder, etc. have generally adopted hitherto. The process model takes L as a basis and assumes that an indefmite number of varieties of H can be created from it, by the application of rules for 'purification'. 9

Each model has its advantages and disadvantages. The static model gives the

benefit of positioning two clearly distinct 'poles' of reference (in this case, K and D). It would fit with Lyons' (1981) distinction between language behavior and language system, where behavior refers to "what speakers actually do' and system is a regularized or idealized account of behavior. In the Greek case, we would regard K and D as (idealized) systems, to which actual behavior (utterances) could be referred

for description. This model would account for the fact that there is more than one

definition of D, for, as Warburton ( 1980: 46) points out, D has been defined, on the one hand, "'on the basis of the people who speak i t . . . the language of the "common people'," on the other hand, on the basis of "certain internal grammatical

characterist ics. . , the language which exhibits a certain phonological structure and

certain morphological and syntactical features," and which has "remained static since

Triandafyllides' (1941 ) Grammar [of D] ." We could say that the first definition relates to actual behavior, the second to (idealized) system. For the process model

the lack of an agreed definition of D poses a problem, since the model requires a firm base as an anchor for the purification processes.

The disadvantages of this static model arise mainly from the description of intermediate varieties (see below).

The process model has the advantage of explaining why a lexical or grammatical

item can appear in a range of variant forms, all or most of which could be labelled K, rather than there being a single 'K' form (see the verb and noun examples above). But the main advantage of this model is m its handLing of intermediate varieties, to' Milch we now turn.

INTERMEDIATE VARIETIES

Intermediate varieties in Greek, i.e., styles which cannot be clearly labelled as

'D" or 'K', occur widely in current literature, in journalism, in scientific and technical language, and in pedagogical texts - in fact, in almost all registers of the current

written language. Wherever the language is not purely D, intermediate varieties have become the norm (cf. Browning 1969:111-2).

In the past, descriptions of these intermediate forms have been marred by a

proliferation of labels which attempt to classify the texts under various 'modifica-

tions' of K and D, but which only lead to confusion. Thus, Mirambel (1937)

describes five '~tats de langue' but his description of the differences between them

Page 13: Diglossia revisited

Diglossia Revisited 13

proves very difficult to apply to actual texts (or sentences) which are neither 'pure' K nor 'pure' D. The best example, however, of the proliferation of 'mixture' terminology is Householder (1962), who uses at least eleven different labels:

Extreme (archaic) K Ordinary K Mild K Moderate or Simple K Popularized (form of) K

Extreme (conversational) D Official D Journalistic D Moderate D Mildly purified D Modified D

Given this variety of labels, how does one distinguish, say, 'mildly purified D' from 'modified D', or, worse still, 'modified D' from 'Simple K'? How can one determine whether a text is 'basically' D, with modifications, or 'basically' K, with modifications? Usually, such a decision is either impossible or totally arbitrary. 1° Our Text 3 above is a good example. Assuming the static model, it could be described as basically K, or basically D, with modifications. Such labels are less than helpful.

If the process model is adopted, all texts are regarded as 'basically D'. What matters is the methods the writer uses to purify his text, i.e., the particular linguistic devices employed and the linguistic levels involved. The advantages of approaching the data descriptively in this way are, firstly, that the use of labels which are fundamentally meaningless becomes unnecessary and, secondly, attention can be focussed on the linguistic means by which the 'purification' has been achieved. This enables us at the same time to examine questions such as the consistency (or lack of it) of the writer's choices.

DIGLOSSIA VS. BILINGUAL AND BIDIALECTAL SITUATIONS

We can now consider the relationship between diglossia of the Arabic or Greek type, and other bilingual or bidialectal situations. Firstly, we must recognize that the nine criteria originally offered by Ferguson as characteristics of 'diglossia' apply equally well to many other types of situation involving two or more languages or dialects, including Creoles. Standard English, for example, holds much the same relationship to a non-standard dialect like Black English as does H to L in diglossia, in terms of Ferguson's criteria. It is a 'superposed' variety; functional specialization (criterion 1) is involved, i.e., H is used for certain purposes, L for others; Standard English has greater prestige (criterion 2) and a strong literary heritage (3); it is acquired secondarily through formal education (by Black speakers) (4), and is fully

standardized (5); the relationship is persistent through time (6) and the lexicon contains paired items (8); the phonological systems may interpenetrate (9). The only

Page 14: Diglossia revisited

14 Language Sciences, Volume 5, Number 1 (1983)

criterion which does not apply in this situation is part of (77; that is, the grammar of H is more complicated.

These criteria can be applied in a similar way to other situations involving two or more languages: to Creoles, as Ferguson himself reports for Haitian Creole vis-it-vis

standard French; to bilingual situations such as in Paraguay (Rubin 1968), and to multillingual situations (e.g., Denison 1971); for further examples, cf. also H. and R.

Kahane (1979). A number of writers such as Gumperz (1971) and Fishman (1971) have, therefore, taken the opportunity to extend the scope of the term diglossia to include situations of this type. We shall argue, however, that such an extension is

undesirable for a number of reasons, and that in fact a narrowing of its scope, to include only diglossia of the Greek/Arabic type, is preferable.

There are undoubtedly many similiarities between Greek or Arabic diglossia and other multilingual and dialect situations. Yet, there are also important differences,

and one further advantage of the process model proposed above is that it enables us to distinguish 'true' diglossia from the other situations. The critical differences

are to be found in the relationship between H and L. In the diglossic case, H varieties

are derived from L by the process of purification. The process has no def'mite terminal point; there is no "pure' H, only an indefinite process of approximations in

the direction of a more highly classicized representation. Consequently, while

descriptive grammars can be written for L, grammars of K must be based on prescriptive rules for purification - there is no natively spoken code. To describe the varieties of K linguistically we must work from the bottom up: we start from D, and then describe the purificatory devices.

In dialect (and possibly Creole - see note 11 below) situations, we find rather the reverse. Here, it is the High variety which has an independent existence, and the multiple varieties of L are described by reference to it. The most frequent

conclusion in descriptions of non-standard dialects is that there is no 'pure' L, bu't

only an indef'mite range of variants departing from, but describable in relation to,

the standard dialect or language. There is, for example, no pure form of Black

English; speakers may use rules for 'improving' their performance in the direction of

the standard, but observation of the results indicates that speakers cannot apply the

rules consistently: they produce utterances which are neither pure 'dialect' nor pure 'standard', but interlace features of both (Labor 1972:213 if; Trudgill 1974; De

Camp 1 9 7 1 : 2 6 ff).11 This is the mirror-image of diglossia, where speakers who are

competent in L fail to apply the rules for H consistently, producing 'mixed' forms, on the one hand, and a great deal of code-switching, on the other, as we observed in our subjects' switching behavior. 12

In short, the difference between diglossia and the dialect-situation can be seen,

under the process model, in terms of the language codes themselves. In diglossia, L

is a natively-spoken, fully specifiable code, while H is indefinitely variable. In

Page 15: Diglossia revisited

Diglossi,* Revisited 15

dialects (and in Creole, if one takes Bickerton's view), H is the natively-spoken, fully specifiable code, and L is variable. Thus, the difference lies ultimately in the nature and relationship o f the codes; the similarities are to be found in the uses o f the codes.

Ferguson's outline of diglossia is based on uses of the codes, rather than on the codes

themselves, and, therefore, fails to distinguish true diglossia o f the Greek or Arabic

type from the dialect-Creole situations which are so similar in many ways, yet so

different in others. Furthermore, we note that when diglossia is defined on the basis

of the codes, Ferguson's four 'canonic' cases are split down the middle, for while

Greek and Arabic are truly 'diglossic', Swiss German and Haitian Creole are variable

L types which can be described in relation to a fully-specifiable H, viz., Standard

German and Standard French, respectively.

One further difference between diglossia and dialect-Creole types involves the

range of styles and functions available to H and L, respectively. It is widely assumed

that the non-standard dialects of a language have a more limited functional range

than the standard. Thus, a statement like "the standard has the resources of coping

with the widest range of language func t ions . . , the stylistic variability of non-standard

dialects is generally restricted particularly at the more formal end of the spectrum"

(Pit Corder 1973) is a typical comment. In diglossia, however, it is not the L whose functional range is limited, but the

H. The L can be used in a full range of styles and functions in both written and

spoken media, whereas H, far from possessing the resources of a standard language,

is limited to certain formal, and above all written, purposes.

We can, for convenience, summarize the differences between diglossia and

dialect-Creole situations.

Diglossia

1. L is natively-spoken; H is not

2. L is relatively fully specit-mble, and independent

3. H is indefinitely variable; no 'pure' form

4. H described on basis of L ('bottom up')

5. L has widest range of styles and functions

Dialect-Creoles

H is natively-spoken

H is relative fuUy specifiable, and independent

L is indeffmitely variable; no `pure' form

L described on basis of H ('top down')

H has widest range of styles and functions

One Final point arises from the adoption of the 'process' model o f diglossia. I f

'purification' is an indefinite process, it should not be possible to say, as we did

originally (e.g., our Text 1, p. 5), that a particular text is 'in' K, or to provide parallel sentences, one of which is 'in' K and the other 'in' D (as in the parallel text in

Page 16: Diglossia revisited

16 Language Sciences, Volume 5, Number 1 (1983)

the Introduction). To speak of K texts or texts in K implies the existence of K as a system, and hence the static model of the relationship. If we accept the process

model, then strictly speaking we cannot use the ' text in K' terminology, since there are only 'more purified' and 'less purified' texts, or texts purified in one linguistic dimension but not in others, etc. In practice, however, it may be useful to say of a

text in which purification has penetrated to every level, including syntax, that the text is 'in' K. The difficulty of this terminology does not arise when the purification is thorough ('pure' K) or when there is none at all (pure D); it only emerges when we are faced with intermediate varieties, where we are forced to decide whether the

text is 'in' K or 'in' D, and hence to resort to unsatisfactory labels like 'simple K'

or 'modified D', both of which might in some cases be applied to the same text. Unfortunately for the static model, intermediate varieties are in abundance; pure K is a rare phenomenon. 13

Finally, what of the term diglossia itself?. It is clear that I would like to see it restricted to situations of the Greek or Arabic type, since these can be characterized by the critical language-code differences outlined above. Fishman's (1971) extension of the term to include a variety ofbi- and multi-lingual types is unsatisfactory in two

respects: it fails to account for the differences, and it weakens the term itself through over-generalization. As Gilbert (1975) remarks, the concept is "spread so thin that new specialized terms will be required to describe the H-L diglossia. . .

with which Ferguson was originally concerned."

A tentative re-definition of diglossia, taking account of the critical language-code differences outlined above, might be as follows:

Diglossia is the name given to a situation in which 'purified' varieties of a

language dependent upon, and describable in relation to, the everyday language, may arise particularly in the written medium through a process of classicization or

other artificial means; any or all linguistic levels may be affected, resulting in

indefinitely variable purified varieties, which are best described by means of a set of 'purification' rules, working 'upwards' from the natively-spoken L.

Diglossia in this newly-defined sense is not confined to Greek and Arabic; it will be found wherever a language undergoes a process of classicization or standardiza-

tion for the purposes of creating a more respectable, more prestigious variety for

literary or other reasons. Examples have been reported for Norwegian (Blom and

Gumperz 1972), Hindi in North India (das Gupta and Gumperz 1968), and Tamil (Ferguson 1959). While most of the Kahanes' (1979) instances of diglossia involve

bilingualism (and hence a fully specifiable and independent code as H), the use of

classical Latin in Western countries from the Renaissance onwards can be cited as a further example.

Page 17: Diglossia revisited

NOTES

Diglossia Revisited 17

1. A shorter version of this paper was given at the Sociolinguistics Symposium, Sheffield, March 1982. I would like to thank Professor Houssiadas and his colleagues in the Department of Psychology at Salonika University for making the research possible and for help with the analysis. I am also indebted to D. Gounelas for help with the choice of texts.

2. The overlap in stylistic ranges of H and L may lead to disagreement within the society as to which variety should be used for which purposes. In Greece, use of H vs. L in government, higher education, etc. is a political issue, forces of the Right favoring H, while more liberal opinion favors L. Kathar6vousa was actively promoted under the regime of the Colonels (1967-74); the policy of the present democracy is to encourage the use of Dhimotiki.

3. Text 1 (K) is from P. KaUigha's 'Thanos Vlekas', published in 'Pandora' 1855. Text 2 (D) is from 'Plusii ke Ftochi' by G. Xenopoulos (1907) and Text 3 ('simple K') is from a short story by A. Papadiamantis, 'Stringla Manna' (1905).

4. Householder (1962) provides a comprehensive description of the linguistic differences between K and D in Greek.

5. One further factor which is relevant to the students' performance is that the medium of instruction and examination at Salonika University is basically D, in contrast with Athens where K is required. Students at Salonika might, therefore, be less familiar with K than they would otherwise be.

6. The national radio broadcasts news in (slightly modified) D but the Right-inclined military radio station broadcasts news in K.

7. It must be emphasised that K has never been a spoken variety of Greek, i.e., it is not at attempt to restore one particular earlier stage of the language. The origins of K are a mainly nineteenth-century 'purification' of the modem spoken language by classicizing it. This accompanied the nationalistic movement: the language was to be purged of foreign elements, with the restoration of ancient

lexical and syntactic formations. Similar parallels elsewhere (e.g. Arabic, Hindi, and Norwegian) are noted in the literature. A good account of the origins and present position of K can be found in Browning (1969).

8. Elsewhere (Hawkins 1979), I have proposed that native-speakers, including literary writers, operate with sets of 'rules' for purification which behave like the variable rules found in other contexts (Bailey and Shuy 1973). Speakers' inconsistencies, and the resulting intermediate varieties of the language, can be explained (a) in terms of a hierarchy of difficulty ranging from orthography at one end to syntax at the other, and (b) in terms of a 'local determinacy' hypothesis in which the subject's choice of K or D form is strongly influenced by the immediately surrounding context rather than by the style of the text as

Page 18: Diglossia revisited

18 Language Sciences, Volume 5, Number 1 (1983)

a whole.

9. 'Purification' as a term for this process is unsatisfactory if it implies that D is in any sense 'impure'. The Greek name Kathar6vousa means literally 'purifying',

where the present participle implies the dynamic nature of the process. Another possible term for the process is classicization- see note 7.

10. Hence, Ferguson's bemused query to Householder's paper (1962: 130): "I would like to ask for a clarification o f . . . the terms simple K, moderate D, and spoken Greek." Householder's reply is a classic: "This will not take very long because we cannot define them precisely."

11. Bickerton (1975) applied this model to Creoles, arguing that there is no pure 'basilect'; Le Page, however, (p.c.) points out that the classic descriptions of

Creole grammars such as B.L. Bailey (1966) idealize the broad Creole and, then,

write rules for moving towards more standard forms. Bickerton's approach, thus, resembles what we are calling the 'process' model, while Bailey's is more akin to the static model which posits an idealized L as well as an idealized H.

12. Bilingual situations with an H and L relationship are different in some respects from both diglossic and dialect-Creole types. In a bilingual speech community

both the codes, H and L, may be fully specifiable, independent languages, natively-spoken, with descriptively-based grammars. This is true of, for example, Spanish and Guarani in Paraguay. The variability, then, arises not from the codes

themselves (as in the other cases) but from the individual speaker's knowledge of

the codes, which may be imperfect, and will result in 'mixed' forms and inconsistencies, as weU as code-switching (Gumperz and Hemandez 1971).

13. Warburton (1980: 47) suggests that pure D, at least among educated speakers,

is also a rare phonomenon. Discussing the emergence of a 'Common Modem

Greek', she points out that "areas where D was dominant in the past, such as casual speech, . . . show an increase in K d e m e n t s ; . . , the language as it. is

spoken t o d a y . . , contains a large number of words of K origin" (1980, 49). This is because speakers borrow lexical items from K, particularly items of a scientific, technical or learned nature, and in these items the K morphology and phonological patterns are often maintained, even in an otherwise D context.

REFERENCES

Abdulaziz, MkiUifi H. 1972 "Triglossia and Swahali-English Bilingualism in Tanzania," Language

in Society 1. 197-213. Bailey, Beryl L.

1966 Jamaican Creole Syntax, Cambridge: Cambridge University Press.

Page 19: Diglossia revisited

Diglossia Revisited 19

Bailey, Charles-James N. and R. Shuy (eds.) 1973 New Ways of Analyzing Variation in English, Washington, D.C.:

Georgetown University Press. Bickerton, Derek

1975 Dynamics of a Creole System, Cambridge: Cambridge University Press. Blorn, Jan-Petter and John J. Gumperz

1972 "Some Social Determinants of Verbal Behaviour," in Directions in Sociolinguistics, pp. 407-34 John Gumperz and Dell Hymes (eds.), New York: Holt, Rinehart and Winston.

Browning, Robert 1969 Medieval and Modem Greek, London: Hutchinson.

Corder, S. Pit. 1973 lntroducingApplied Linguistics, Harmondsworth: Penguin.

Das Gupta, Jyotirindra and John Gumperz 1968 "Language, Communication, and Control in N. India," in Language

Problems of Developing Nations, pp. 151-66, Joshua Fishman, et al. (eds.), New York: Wiley.

De Camp. David 1971 "The Study of Pidgin and Creole Languages," in Pidginization and

Creolization of Languages, pp. 13-39, Dell Hymes (ed.), Cambridge: Cambridge University Press.

Denison, Norman 1971 "Some Observations on Language Variety and Plurilingualism," in

Sociolinguistics pp. 65-77, John B. Pride and Janet Holmes (eds.), Harmondsworth: Penguin.

Ferguson, Charles 1959 "Diglossia," Word 15. 325-40. Reprinted in Language in Culture and

Society (1964), pp. 429-39, Dell Hymes (ed.), New York: Harper and Row.

Fishman, Joshua A. 1971 Advances in the Sociology of Language, Vol. 1., The Hague: Mouton

Row. Frangos, Christos P.

1972 I Epidrasi tis Glossikis Morris stin Anagnosi ke stin Katanoisi ('The Influence of Language Variety on Students' Reading and Comprehen- sion'), Monograph, University of Ioannina.

Gilbert, Glenn. 1975 "Review of Fishman's The Sociology of Language," Language 51. 234-

6. Gumperz, John J.

Page 20: Diglossia revisited

20 Llulgultlge Sciences, Volume 5, Number 1 (1983)

1971 Language in Social Groups, Essays selected and edited by Anwar S. Dil, Stanford: Stanford University Press.

Gumperz, John J. and Eduardo Hemandez 1971 "Cognitive Aspects of Bilingual Communication," in Language Use and

Social Change, pp. 111-25, W. H. Whiteley (ed.), London: Oxford University Press.

Hawkins, Peter R. 1979 "Greek Diglossia and Variation Theory," General Linguistics 19. 4.

168-87. Householder, Fred W.

1962 "Greek Diglossia," Georgetown Monographs 15. 109-32. Hudson, Richard A.

1980 Sociolinguistics, Cambridge: Cambridge University Press. Kahane, Henry and Renee

1979 "Decline and Survival of Western Prestige Languages," Language 55. 183-96.

Kaye, Alan 1972 "Remarks on Diglossia in Arabic," Linguistics 81.82-48.

Labov, William 1972 Sociolinguistic Patterns, Philadelphia: University of Pennsylvania Press.

Lyons, John 1981 Language and Linguistics, Cambridge: Cambridge University Press.

Mirambel, Andre 1937 "Les '6tats de langue' darts la Grece actueile," Conferences de l'Institut

de Linguistique de l'Universite de Paris, 5.19. Rubin, Joan

1968 "Bilingual Usage in Paraguay," in Readings in the Sociology of Language, pp. 512-30, Joshua Fishman (ed.), The Hague: Mouton.

Stewart, William A. 1968 "A Sociolinguistic Typology for Describing National Multilingualism,"

in Readings in the Sociology o f Language, pp. 531-45, Joshua Fishman (ed.), The Hague: Mouton.

Trudgill, Peter 1974 The Social Differentiation of English in Norwich, Cambridge: Cambridge

University Press. Warburton, Irene P.

1980 "Greek Diglossia and Some Aspects of the Phonology of Common Modem Greek," Journal of Linguistics 16.45-54.