Introduction A topic of growing interest during the last decade in Guatemala has been the state of language maintenance among the Mayan languages. In response to this interest, the Summer Institute of Linguistics has augmented its ongoing language survey efforts by investigating not only dialect differences but also language attitudes and use (McArthur and McArthur 1987; Powell 1989). In 1986, SIL began a two-year study in the sociology of language in the K'iche'- speaking area (described in Lewis 1987) which was primarily focused on language maintenance and which looked specifically at not only language use but the ethnolinguistic vitality of the target communities as well. Only some very partial results of that study have been previously reported on (Crossley 1989; Lewis 1991; Lewis 1993; Lewis 1995; Lewis, PelSiz and Willems 1989) and a more thorough but still not exhaustive analysis of the data was the topic of my doctoral dissertation (Lewis 1994). It is my goal in this paper to offer a summary presentation of the language use data which were collected along with my analysis and interpretation of the implications for K'iche' language maintenance.

Methodology

Language use data were collected beginning in August, 1987 in many K'iche' communities. The resulting data were sufficient for statistical analysis in only seven of these, Chichicastenango, Cunen, Joyabaj, Santa Cruz del Quiche, Sacapulas, San Andres Sajcabaja, and Totonicapan. In Joyabaj only a partial sample was collected so the resulting analysis is incomplete and will be included here only insofar as the data permits. These communities represent more than 25% of the K'iche'-speaking population and are also communities which represent major dialect centers. Two, Cunen and Sacapulas, are also linguistically and ethnically diverse communities where more than one linguistic variety is spoken. In both cases, we did not differentiate in the analysis between the various Quichean varieties being spoken, making a somewhat ad hoc decision that all of the Mayan varieties would be on a more or less equal footing viz a viz Spanish. This assumption can be questioned, and both communities merit a much more thorough study of the dynamics of their interethnic relations.

We set as our goal the collection of authentic language use data, rather than (equally valuable) data about attitudes towards the languages available for use. For this reason, and because of the sensitivity of the region after many years of the civil and political strife, we chose not to use direct methodologies which employ interviews or questionnaires. This primary goal also influenced us to choose an unobtrusive methodology which would minimize the observer's effect on the communication situation. The methodology which resulted was an observational one wherein our observers, both expatriates and community residents, kept rather simple mental tallies of observed speech behavior and later, away from the site, recorded these observations on tally sheets. For each observed speech interaction, we recorded the location/situation category, the race, sex and age of the participants and the language used by each participant with each interlocutor. There was also space for other comments which the observer might feel were important to provide an understanding of the situation. We collected a convenience sample but we attempted to make observations in the most significant domains in each community and during the normal course of our observer's daily routine. We had originally set a goal of obtaining l,000 observations in each community but were unable to achieve this goal. We obtained 4,920 observations in the seven communities which were made up of a total of 11,220 interchanges between participants.

Data Analysis

The resulting data were analyzed in several different ways. One method, reported on briefly in Lewis (1995) was to use categorical models maximum likelihood analysis to determine the significant independent variables and interactions of those variables which affect the choice of language in each speech interaction. For this analysis, four independent variables were identified, race of the speaker (RACE), race of the interlocutor (IRACE), sex of the speaker, (SEX) and sex of the interlocutor (ISEX). Table 1 shows the significant variable interactions for each of the communities.

TABLE 1: SIGNIFICANT VARIABLES & VARIABLE INTERACTIONS

VARIABLES CHI CUN JOY SAC SAJ SCQ TOT

RACE X

RACE*IRACE X X X

RACE*ISEX X

SEX*IRACE X

SEX*ISEX X

SEX*RACE*IRACE X

SEX"RACE*IRACE*ISEX X X X

Although RACE is significant by itself in only one of the communities (Sacapulas), it is the single most influential variable in determining language choice. As might be expected, Indians speak K'iche' more often than Ladinos. Furthermore, Indians speaking to Indians (RACE*IRACE) speak K'iche' more often than Indians speaking to Ladinos. Another characteristic of the sociology of language in these communities which is made clear by the data is the differential in conversational interaction patterns between the two races. While Ladino speakers converse with other Ladinos as well as with Indian interlocutors, most Indian speech transactions are with Indian interlocutors. This pattern is iconic of the general nature of relations between Indians and Ladinos with the more-powerful Ladino dominating conversational interactions while the Indian is generally seen and heard very little in such situations.

The analysis also showed an interesting differential in some of the communities in language use based on gender. Chichicastenango, Cunen, San Andres, Santa Cruz del Quiche, and Totonicapan are characterized by a generally greater use of K'iche' between members of the same sex. Neither Sacapulas nor Joyabaj (the latter perhaps because of incomplete data) show this pattern. The data also demonstrate the generally distinct patterns of Spanish acquisition which exist between men and women in the communities. Generally Indian men are the ones who have greatest contact with the outsider Spanish-speaking world, and thus have the greatest opportunities to acquire and use Spanish. In most of the communities Spanish use by Indian male participants is greater than Indian women's use of Spanish though Santa Cruz del Quiche is an exception to this with Indian women tending to use Spanish with greatest frequency than Indian men.

Although this analysis provides some insight into the social dynamics which affect language choice, a much more revealing analysis in terms of language maintenance can be had by looking at language use by age groups and according to domains of use. For the purposes of this analysis, a Language Maintenance Index was calculated. This is a number between 0 and 2 which reflects the proportion of K'iche' use in the speech transactions which were observed. The Language Maintenance Index was arrived at by assigning a weighting factor to each of the language varieties used (K'iche' = 2, Code-Mixed = 1, and Spanish = 0). The frequency count for each of these varieties was multiplied by the weighting factor and that total was divided by the total number of observations. This technique was used for each age group and for each domain of use. In addition, a global language maintenance index number was calculated for each community using all of the observed speech transactions.

As a further means of interpreting the data, the language maintenance index scores were classified as either strong, moderate or weak. This categorization was arrived at by calculating the average (mean) index score and the standard deviation for all of the observations. All of the scores within one-half standard deviation of the mean were considered to represent moderate language maintenance. All those above the moderate range were considered to represent strong language maintenance, and all those below the moderate range were categorized as representing weak language maintenance. Table 2 shows the ranges of index scores which fall into each category.

TABLE 2 - LANGUAGE MAINTENANCE INDEX LEVELS

1.66 - 2.00 Strong language maintenance

1.28 - 1.65 Moderate language maintenance

0.00 - 1.27 Weak language maintenance

X= 1.45, s = .33

Using the language maintenance index scores the data can be examined in order to look for evidence of what Fasold (1984:54) has called "leaking diglossia". If a number of domains are identified in which it is unclear which language is appropriate, or in which the language used is not the one traditionally expected in such a domain, we can interpret that as evidence of a "leaking" diglossia. Ferguson (1959) identified stability as one of the defining characteristics of diglossia and stability along with compartmentalization of functions are the two characteristics of diglossia which Fishman (1967) retained in his redefinition of the term. Leakage in any of the domains of use, can be taken then, as a destabilization of a stable diglossic state, and if accompanied by increasing bilingualism is quite likely to be indicative of language shift.

Since language shift occurs through the lack of transmission of a language from one generation to the next, lower levels of language maintenance in the younger age groups can be taken as evidence of incipient language shift. In addition the levels of language maintenance evidenced by young adults who are the producers and caretakers of the next generation may also provide indications of the prospects for language maintenance in the communities. Table 3 shows the Language Maintenance Index scores for each community by age group.

TABLE 3: SUMMARY OF LANGUAGE USE BY AGE GROUPS

TOWN 1-12 13-24 25-34 35-44 45-54 55+

Chichicastenango 1.54 1.37 1.35 1.31 1.57 1.30

Cunen 1.93 1.59 1.62 1.59 1.38 1.74

Joyabaj 1.18 1.46 1.50

Sacapulas 1.74 1.54 1.66 1.79 1.50 1.07

San Andres

Sajcabaja 1.75 1.49 1.48 1.66 1.68 1.88

Sta. Cruz del Quiche 0.98 1.16 1.58 1.77 1.61 1.71

Totonicapan 1.40 0.93 1.34 1.45 1.58 1.65

The community with the lowest levels of language maintenance in the younger age groups is Santa Cruz del Quiche where the two youngest age groups have language maintenance index scores that show only weak maintenance. While the young adult and adult age groups (25-34 and 35-44) show moderate and even strong language maintenance respectively, it must be remembered that Santa Cruz is also the community in which women have taken the lead in acquiring and using Spanish. Totonicapan is the community with the next weakest profile of language maintenance in the younger age groups. While the youngest age group has an index score in the moderate range (1.40), the next oldest age group has the lowest index score of any age group in any of the communities (0.93) and none of the older age groups have a maintenance index that represents anything more than moderate maintenance. In the categorical models analysis, it was also found that the men of Totonicapan are taking a quite strong lead in the acquisition and use of Spanish.

With data from only three age groups, it is difficult to draw any firm conclusions regarding Joyabaj, though it is noteworthy that the 13-24 year old age group score is only slightly higher than the score for the same age group in Santa Cruz del Quiche which is the second lowest in the study. Both of the other age groups for which we have data from Joyabaj (the 25-34 and 35-44 year old age groups) show moderate language maintenance.

At the other extreme, Cunen is the community with the strongest language maintenance profile in the youngest age groups, though still only the youngest and oldest age groups demonstrate strong language maintenance. All of the other age groups in Cunen are demonstrating moderate K'iche' maintenance. Sacapulas and San Andres Sajcabaja don't demonstrate as high a language maintenance index in the youngest age group but do have an overall pattern that is quite similar to that of Cunen. San Andres Sajcabaja shows stronger maintenance in the older age groups but slightly weaker maintenance in the younger age groups.

Chichicastenango falls between these two groups. Though not as low as Santa Cruz del Quiche and Totonicapan in its younger age groups, it is not nearly as strong as Cunen or Sacapulas. All of the age groups in Chichicastenango are demonstrating only moderate language maintenance which seems to corroborate Langan's (1990) characterization of language attitudes and use in Chichicastenango as "ambivalent".

What is clear from these data is that though the language maintenance profiles of the communities based on age groups differ in their details, there is a general tendency for relatively higher levels of maintenance to be seen in the very youngest and in the older age groups. Lower maintenance levels are generally seen in the more economically active and socially mobile middle age groups. This indicates that in most of the communities K'iche' is being transmitted to the youngest age group in the home domain. When the children begin school, however, and then move on to young adulthood they are exposed to both the opportunities and the pressures to learn and use Spanish. It is difficult, with these data, to measure precisely the impact of schooling on language use, but the data are highly suggestive of a shift towards Spanish use in this age group as a result of the increasing attendance and retention levels of K'iche' children in the elementary and secondary schools. While current intergenerational language transmission seems to be relatively intact, the trend towards greater Spanish use by the young adult groups ought to sound a warning.

A similar sort of analysis can be done using the frequency counts and calculated Language Maintenance index scores according to domains of use. The topic/location categories originally assigned by our observers have been consolidated and recategorized into ten domains of use which correspond in general terms to the domains identified by Fishman (1991) who pointed out that in a situation characterized by broad diglossia (to use Fasold's (1984) term), it is to be expected that the more intimate, more solidary domains would be the most likely to show higher levels of maintenance of the L language. If broad diglossia obtains in the K'iche' communities, we should expect to find higher levels of K'iche' maintenance in the more intimate domains and lower levels of K'iche' maintenance in the more formal domains. Of particular interest is the Home domain which is the most crucial domain for the intergenerational transmission of K'iche'. Table 4 shows a summary of the language maintenance index scores for each community according to domain categories.

This analysis shows a more robust and intact diglossia than does the analysis by age group. In all of the communities, the most intimate, informal and solidary domains demonstrate relatively strong levels of language maintenance. Again, Santa Cruz del Quiche has the weakest overall profiler but in the Home domain it is no weaker than San Andres Sajcabaja and both communities show moderate language maintenance. Cunen is also the community with the strongest overall profile and demonstrates very strong language maintenance in the Home domain. The other communities have language maintenance profiles that are somewhere between these two extremes. It should be noted that only Cunen, Sacapulas and Totonicapan show strong language maintenance in the Home domain. All of the other communities are characterized by moderate language maintenance in that domain which may be attributable to the fact that it is only the older speakers who are using K'iche' in the Home domain. On the other hand, all of the communities except Chichicastenango, Santa Cruz del Quiche and Totonicapan demonstrate strong language maintenance in the Street domain which includes encounters between individuals in public on the streets. These three exceptions demonstrate only moderate maintenance levels in this second most intimate domain.

TOWN HOME STREET PLAY MARKET WORK RELIGION STORES MEDIA SCHOOL GOVT

Chichicastenango 1.53 1.54 1.21 1.55 1.76 1.13 1.21 1.51 1.13 1.06

Cunen 1.99 1.70 1.89 1.80 1.88 1.43 1.77 0.97 1.80 1.07

Joyabaj ---- ---- ---- 1.32 1.89 1.05 0.70

Sacapulas 1.81 1.82 1.50 1.72 1.55 1.21 1.60 1.41 1.71 1.64

San Andres Sajcabaja 1.46 1.73 1.80 1.78 1.77 1.12 1.46 1.79 0.79 1.29

Sta. Cruz del Quiche 1.46 1.65 0.62 1.70 1.54 0.87 1.32 1.76 0.36 1.53

Totonicapan 1.89 1.55 0.66 1.75 1.33 0.91 0.96 0.91 1.08 1.23

<-------MORE INTIMATE----------------------------------------------------------------------LESS INTIMATE------>

Again, Joyabaj presents a somewhat anomalous picture and unfortunately we have no data for the Home domain. The strongest domain for K'iche' maintenance for which we have data from Joyabaj is the Work domain. The Market in Joyabaj shows only moderate K'iche' maintenance and has a Language Mainenance index score (1.32) which is much lower than any of the other communities.

The domains in each community which show the lowest levels of K'iche' maintenance are, as might be expected, the domains which are most formal and public. In all of the communities (except possibly Joyabaj) there is a clear break between those domains which are more intimate, and thus most likely to be associated with K'iche' use and those which are more public and associated more generally with Spanish. While Home, Street, Play, Market and Work are clearly K'iche' domains, Religion, Stores, Media, School and Government Offices are all domains which show markedly lower levels of K'iche' use. There are some exceptions, however, which represent to some degree, a resistance to language shift either through the retention of K'iche' in a domain where it has been traditionally used or through the introduction of K'iche' into a domain where it had not previously been considered appropriate.

An example of the former occurs in the Store domain in Cunen and Sacapulas and to a somewhat lesser degree in San Andres where Ladino merchants were observed to accommodate to the code of their K'iche'- speaking clients. In the other communities, Spanish was more frequently used in Stores. An example of the second kind of language maintenance, the introduction of K'iche' into a domain where it had not previously been used or considered appropriate, is that of the school domain where in both Cunen and Sacapulas there was considerable use of K'iche' in the school setting.

Both communities were observed to have active and enthusiastic bilingual school teachers working under PRONEBI, the government bilingual education program, who were promoting the use of K'iche' in their classrooms and around the school in general. In spite of these efforts, however, formal public schooling is not a generally accepted K'iche' domain and the domain of education (formerly an informal, intimate domain administered by parents and siblings) is a domain which is leaking.

The weakest of the more formal domains is Religion where only in Cunen is language maintenance seen to be in the moderate range. In all of the other communities language maintenance in the domain of Religion is in the weak range. Our sample was cleanly biased towards Protestant and Catholic religious observances, and generally towards the more formal and public religious services. Although we did have some observations of Mayan traditional religious practitioners who obviously use K'iche' predominantly, these were not enough to present a completely representative picture of language use in the religious domain. Clearly, however, in spite of efforts by both Protestants and Catholics to promote vernacular language liturgy and written materials, a large part of the Christian segment of these Mayan communities do not deem it appropriate to use K'iche' in formal public worship. Religion, too, could be categorized as a leaking domain.

Another domain which shows only weak levels of K'iche' use overall is Government Offices. Here again, there are exceptions. In Sacapulas, and surprisingly, in Santa Cruz del Quiche the level of K'iche' use in the town hall and other government offices was in the moderate maintenance range. In all of the other communities K'iche' maintenance is weak, K'iche' has not traditionally been the language associated with the functions of Ladino-dominated government and in most Mayan communities monolingual speakers of a Mayan language are tolerated and accommodated to out of necessity rather than out of any value placed on the use of the vernacular.

Summary and Conclusions

Each of the communities included in this study can be seen to be characterized by a unique language maintenance profile when examined in terms of the interaction of the independent variables of RACE, IRACE, SEX and ISEX as well as in terms of age groups and domain categories. These language maintenance profiles correspond roughly to other socioeconomic characteristics of the communities which have not been reported on here (but see Lewis 1993; 1994 and 1995).

The three communities which have been reported to have the weakest overall language maintenance profiles, Santa Cruz del Quiche, Totonicapan and Chichicastenango, are also the most urbanized and economically developed of the communities. Two are departmental capitals and have considerable ease of communication with the outside world. All three, but especially Chichicastenango and Totonicapan are moving away from subsistence farming and adopting a cash-based economy based on non-traditional crops and goods and services which they are marketing not only regionally but nationally and internationally. These communities have schools available not only in the town centers but in the rural areas as well.

Secondary schooling is also available in all three communities. These communities have undergone an identity shift in which they have adopted a world view and a set of values that coincides in many respects with that of their Ladino neighbors. This identity shift is resulting in language shift. In contrast, Joyabaj, which seems to exhibit many of the same weaknesses in language maintenance as these three communities, is not characterized by the same socioeconomic vitality although the socioeconomic disruption of the 1976 earthquake and the subsequent massive influx of modern goods and services has had a significant effect on the interethnic dynamics of that community.

The communities with the stronger language maintenance profiles are those which are more remote, more difficult to get to, and have less access to schooling and other services identified with a modern, technological identity. I have characterized these communities as having maintained not only their language but, more importantly, their identity and world view. It is this identity maintenance which undergirds the maintenance of K'iche'.

Another way to categorize the communities is in terms of Fishman's (1991) Graded Intergenerational Disruption Scale (GIDS) which provides a means of measuring the level of disruption which is produced by the processes of language shift. Because the GIDS is focused primarily on the reintroduction of a language where language shift has already occurred it may not be the optimal metric of language maintenance. It may well be that the process of loss of domains differs in character from that which Fishman proposes as appropriate for the planned reintroduction of language use into new or previously abandoned domains. Indeed, it seems to me that the GIDS is not a precise enough measure for use in situations such as that described here, particularly where the levels of disruption are greatest. Nevertheless it does provide a metric in a field where such tools are hard to find.

In a somewhat subjective manner, I have used the age and domain data along with ethnographic data to classify each community on the GIDS and have found the communities cluster right around and on both sides of GIDS level 6 which Fishman describes as the threshold level for language maintenance. Fishman (1991:92) observes that the lion's share of the world's intergenerationally continuous languages are at this very stage and they continue to survive and, in most cases, even to thrive, without going on to subsequent ('higher') stages "(emphasis in the original), I have classified Chichicastenango and Totonicapan as being at GIDS Stage 6.

Fishman (1991:92) states that at this stage the "shifting" language "... is the normal language of informal interaction between and within all three generations of the family ... "While there are families in these communities where this intergenerational use of K'iche' is not the case, the general pattern of K'iche' use in the home and between all three generations is still relatively intact. In addition, the endangered language at Stage 6 must be the language of interfamily interactions. This is clearly the case in these two communities.

I have ranked Santa Cruz del Quiche (and probably Joyabaj as well) at GIDS Stage 7. At this stage, Fishman (1991:90) notes that though the intergenerational transmission of the language has been disrupted, the older generation which still speaks the endangered language is present in the home domain and is still integrated into the family. At this stage, however, the childbearing generation is not passing on the language to their children. Our data from Santa Cruz del Quiche includes many examples of individuals who do not themselves normally speak K'iche' and who are not passing K'iche' on to their children, but who have K'iche' speaking parents and grandparents residing in their homes. Although there is evidence of K'iche' being used in the Home domain in Santa Cruz as described above, it is likely that it is not the youngest age group which is using it.

The communities with the stronger profiles, Cunen, Sacapulas, and San Andres Sajcabajar have been placed at GIDS Stage 4. This categorization represents the higher levels of language maintenance within these communities in both the core (intimate) domains and younger age groups as well as in a greater number of domains and age groups overall. Fishman's focus in the GIDS at this level is more on the expansion of language use into the less intimate domains especially into lower level education and literacy. Although none of these communities shows a particularly strong tendency towards the use of K'iche' for literacy and education, they do show a general retention of K'iche in more domains and both Cunen and Sacapulas have active bilingual teachers who are promoting the expanded use of K'iche' in the educational and other settings.

The conclusion that I have drawn from this study is that K'iche' language maintenance is being eroded in the K'iche' communities for which we have data though not at the same rate and to the same degree in all of them. The communities are at different levels of language maintenance and this can be related to the level of identity maintenance (or shift) which each community has experienced as its residents have changed their subjective vitality assessments in response to a changing demographic, political, economic, social and cultural environment.

References Cited

Crossley, Charissa. 1989. El uso del idioma K'iche' en los hogares de Chuixchimal, Totonicapan, Guatemala. WINAR: Boletin Intercultural. 5.1.3-33.

Fasold, Ralph. 1984. The sociolinguistics of society. (Language in Society). Oxford: Blackwell.

Ferguson, Charles A. 1959. Diglossia. WORD. 15.325-40.

Fishman, Joshua A. 1967. Bilingualism with and without diglossia; diglossia with and without bilingualism. Journal of Social Issues. XXIII.2

Fishman, Joshua A. 1991. Reversing language shift. Philadelphia: Multilingual Matters.

Langan, Katherine A. 1990. Language proficiency use and attitudes in Santo Tomas Chichicastenango: A study of language competition. Unpublished doctoral dissertation, Georgetown University.

Lewis, M. Paul. 1987. Un estudio de la sociologia del lenguaje del idioma quiche. WINAK: Boletin Intercultural. 2.4.249-255.

_____________1991. Language maintenance, standardization and national trends: lessons learned from the Quiche survey. Proceedings of the Summer Institute of Linguistics International Language Assessment Conference, ed. by Gloria Kindell, 131-139. Dallas, TX: Summer Institute of Linguistics.

_____________1993. Real men don't speak Quiche: Quiche ethnicity, Ki'che' ethnic movement, K'iche' ethnic nationalism. Language Problems and Language Planning. 17.1.37-54.

_____________1994. Social change, identity shift and language shift in K'iche' of Guatemala. Unpublished doctoral dissertation, Georgetown University

_____________1995. Measuring K'iche' (Mayan) language maintenance: a comprehensive methodology. 1995 Annual Meeting, Linguistic Society of America in New Orleans, LA.

Manuel Peliz Pacajoj and Reginald Willems. 1989 El uso de K'iche' en las radiodifusiones, WINAK: Boletin Intercultural. 5.2.10 8-128.

McArthur, Richard and Carol McArthur. 1987. San Luis Jilotepeque: Un pueblo bilingue en camino a ser monolinge. Unpublished manuscript, Guatemala: Summer Institute of Linguistics.

Powell, Patricia M. 1989. The use of Cakchiquel in two communities: Santa Ana Chimaltenango and San Martin Jilotepeque. Unpublished manuscript, Guatemala: Summer Institute of Linguistics.