Català | Español


Four different methods have been tested for the establishment of the IIS:

Method 1

Estimate and comparison of the percentage of occurrences of each variable and variants, considered for each pair of analysed speakers/writers. Application value variant. % of variant realization /each variant and calculate difference between the 2 speakers. Sum up all the % differences and calculate the average. Divide this figure by 100. Substract 1 to the final figure.

Method 2

Calculation of the Adjusted Residual Value. Cross-tabulation running with SPSS. Calculation of the Adjusted Residual Values (ARV) for each variable.  Assignment of a value to each A range:

The IIS  is obtained by the calculation of the following formula:

Metòde 2

Method 3

Calculation of the Phi coefficient. Calculation of χ2 and obtention of the Phi coefficient, which is located between 0 and 1, indicating the relation between variables. The square of the Phi coefficient (Phi2) represents the percentage of overall variance. The IIS is obtained through the calculation of the average of Phi:

Metòde 3


Corpus for the English modules

Corpus for the Catalan modules

Corpora for the Spanish modules


For the phonological modules, we examine oral samples by means of the auditory-acoustic approach which combines the use of specially developed techniques ―normally involving specialized computer software aimed at the acoustic analysis of speech― together with the phonetician’s knowledge on physics and the acoustic properties of the speech signal, especially those of the properties of the sounds of the language under analysis.

For the analysis of the morpho-syntactic and discourse-pragmatic modules, we analyse each written sample by identifying and coding the variables under analysis.

Preliminary results and conclusions

  1. Method 3, based on the Phi Coefficient has turned out to be the most reliable method and has triggered the most robust results for both intra and inter speaker/writer variation, and in particular in the phonological modules, with some exceptions.
  2. For all modules and languages, hypothesis 1 is confirmed in that there is more variation and thus more idiolectal distance between speakers and writers’ samples than between two samples of the same speaker or writer, which show a quite steady idiolectal similitude throughout time.
  3. For all modules and languages, hypothesis 2 is confirmed in that an individual’s idiolectal style (spoken or written) doesn’t seem to vary much throughout time.
  4. Inter-speaker/writer IIS values may seem to be too high, but it has to be borne in mind that, except for the phonological module of Catalan, all the speakers or writers considered are from the same language variety.
Last updated: 8-02-2013