514 JOURNAL OF LAW AND POLICY
writers compared is applied). In other words, for two samples to
be attributed to the same speaker or writer, the IIS values must
all be near 1 in all modules; for two samples to be attributed to
different speakers or writers, IIS values must all be between 0.7
and 0.5 (same speech variety) or between 0.5 and 0.3 (different
speech varieties).
The disparity of results obtained in some of the IIS values
has had a direct effect on the design of further experiments and
on future data collection. Future research will focus on
increasing the number of languages as objects of analysis (e.g.,
Arabic), the sample size (i.e., more subjects for each language),
and also on the stratification of the corpora by genre in order to
confirm hypothesis 3. Additionally, other indicators such as
gender, age, or educational level will be examined to contribute
to the Base Rate Knowledge of population distribution.
In conclusion, the IIS measure can provide reliability to the
concept of idiolectal similitude or distance, and once the
protocol for its calculation is consolidated, the IIS measure may
be successfully complemented with other approaches to forensic
speech and text comparison to be used in real forensic cases. In
addition to this, research towards the establishment of the IIS
measure can also provide forensic linguistics with a Base Rate
Knowledge of population distribution as regards several
linguistic variables for the three modules and the three languages
under study, which is a fundamental issue in current forensic
linguistic work.