Ministry of higher and secondary special education faculty of foreign languages of bukhara state university department of english literature

2.2. Methods of analysis and results

Dialectological data are in most cases published on maps. Depending on the purpose of the study, the material is either transferred directly onto the map or the material is interpreted before it is integrated into the map. In the former case, the map covers the spatial distribution of the full variation concerning a particular linguistic item as collected by the fieldworker(s); in the latter case only those variables show up in the map which are interpreted as predominant within a particular area. The former type of map is called a display map and is the preferred form of language atlases, the latter type is called an interpretive map and is predominantly found in secondary studies which are geared towards more general conclusions and often take their data from display maps as their primary sources (Chambers and Trudgill 1998: 25).

In both types of maps, regions which differ with respect to a particular variable of interest can be separated by lines called isoglosses. Depending on political, historical or cultural factors, different patterns of isoglosses can emerge. The more isoglosses separate a given geographical area from another the more significant that dialect area will be. Thus, bundles of isoglosses are of particular interest in dialectology and – as will be shown below (cf. linguistic areas in section 3.3.3) – in typology as well.

Since maps are static and two-dimensional, while dialect variation is dynamic and multidimensional (apart from the geographical dimension there are social dimensions such as age, social class, education and gender), special techniques need to be developed for taking into account more complex dialect situations. If, for instance, the geographical diffusion of a certain variable depends on the age of the informants, a dialectologist may decide to keep the geographical aspect stable and to show the diffusion of the variable along the dimension of age by creating separate maps for each age group relevant in that particular study. The reader will then have to superimpose these maps to see the dynamics of the diffusion of that variable. Given new technological options, this situation may change in the future, but this stage does not seem to have been reached yet: “Drawing inferences from dialect maps gets more complex as the dialect situation adds dimensions. Eventually, we will have the wherewithal to create multidimensional displays involving stereoscopy or holography, but for the time being the best we can do is superimpose visual devices on flat planes.” (Chambers and Trudgill 1998: 119)

Discovering the geographical distribution patterns of variables crucially depends on quantification, i.e. on suitable statistical and mathematical methods. Two such methods will briefly be sketched in this section. One method is known under the term of dialectometry as introduced by Séguy (1973) in his sixth volume of the Atlas linguistique de la Gascogne (for more details, cf. Goebl 1982, 1993). Séguy’s method ultimately consisted in establishing a dissimilarity matrix. Together with his team, Séguy counted the number of differing variables for each of the contiguous sites in his atlas. For that purpose, he divided the items of his Gascony data (more than 400 items) into different types of variables (lexical, pronunciation, phonetic/phonological, morphological, syntactic). For each of these types he first calculated a separate percentage which he then integrated into an overall percentage of linguistic distance between each of the adjacent sites of his atlas. Finally, he plotted the results of his calculations of linguistic distance on his base map.

In the dialectometric method developed by Séguy, the geographical dimension is always the same irrespective of the degree of dissimilarity between the sites, i.e. each site is placed in the geographical space of the map of Gascony. The actual linguistic distance is only indicated by figures printed between the contiguous sites. In the more general approach of multidimensional scaling, the second statistical method to be discussed here, any type of quantified dissimilarity can be represented spatially by a straightforward procedure. The degree of dissimilarity between different sites, or between the results optained for different informants, to put it more generally, is graphically expressed by different proportions of spatial distance. Thus, if the dissimilarity between A and B is twice as much as between B and C, the spatial distance between A and B will be twice as large as between B and C as well. In the case of the dialectometric method as introduced above, the geographical dimension is represented directly, while other potential dimensions (social, cultural, etc.) must be represented indirectly. Multidimensional scaling allows to represent any dimension directly. The advantage of this procedure is very adequately characterized as follows by Chambers and Trudgill (1998):

One clear advantage of multidimensional analysis is that the investigator’s categorisations of the variables do not pre-judge the analysis or affect it in any other way. The statistical program searches the matrix for dissimilarities wherever they may be found and scales them independently for each dimension. Multivariate scaling often reveals correlations in the data that the investigator had not previously recognised. (Chambers and Trudgill 1998: 141)

The above interpretations of data concentrate either on the location of linguistic variability in the geographical space as such or on quantifying dissimilarity among different sites or individual informants be it with regard to geographical space or with regard to other domains. Another approach to variability concentrates on the covariation of structural variables as they are used by individual informants or within certain geographical or other domains. The following example from the variety of English spoken in Anniston (Alabama, USA) as described by Feagin (1979) is to illustrate this. In this variety, plural subjects often do not agree with the verb as in They was not at home (to use a constructed example; cf. also Chambers, this volume, on what he calls ‘default singulars’ in this and other non-standard varieties of English). As it turns out, the occurrence of nonagreement correlates with certain types of constituents in the subject position. The following table from Chambers and Trudgill (1998: 133) shows the lects attested in Anniston depending on whether the subject is they, a noun in the plural (NPpl), we, you (pl) or there. The (+) sign indicates that nonagreement is attested, the (ø) sign indicated that it is not.

Table 1. Occurrence of nonagreement in Anniston

	Nongreement with they NPpl		we	you (pl)	there
Lect 1	ø	ø	ø	ø	+
Lect 2	ø	ø	ø	+	+
Lect 3	ø	ø	+	+	+
Lect 4	ø	+	+	+	+
Lect 5 + + + + +

The above five lects attested in Anniston form an implicational hierarchy of the following type:

(1) they > NPpl > we > you (pl) > there

Thus, if a person from Anniston is heard using nonagreement with a certain constituent X one can conclude that she will also use it with a constituent further to the right. For instance, if a person uses nonagreement with the pronoun we in the subject position she will also use it with you (pl) and with there. As it will turn out in section 3.3.2, this approach parallels implicational universals and hierarchies.