Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


PhD dissertation presentation

No description

Nuria Gavalda

on 9 November 2015

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of PhD dissertation presentation

Index of Idiolectal Similitude for the phonological module of English applied to forensic speech comparison
Objectives, research questions and hypotheses
IIS Results
Núria Gavaldà Ferré

PhD Supervisors: Maria Teresa Turell (UPF)
Joan Carles Mora (UB)

Variationist Sociolinguistics
Forensic Linguistics
Theoretical Framework
Variation in the individual
Inter-speaker variation
Variation in the group
Intra-speaker variation
Idiolectal style
Proposed by Turell (2010) as more relevant to forensic linguistics.
It makes reference to:
a) how a speaker's linguistic variety, which is shared by all the members of their community, is used in a distinctive way.
b) the speaker/writer's production, which appears to be 'individual' and 'unique' (Following Coulthard 2004)
c)Halliday's proposal of 'options' and 'selections' from these options
General research question
Specific research questions and hypotheses
To create of an Index of Idiolectal similitude (IIS) that quantifies the idiolectal distance between two samples and determines whether they show inter- or intra-speaker differences.
Is it possible to distinguish between inter- and intra-speaker variation when looking at two linguistic samples?
RQ1: Is inter-speaker variation higher than intra-speaker variation?
H1. Inter-speaker variation will be higher than intra-speaker variation.
RQ2: To what extent is a speaker’s idiolectal style stable over time?
H2. A speaker’s idiolectal style will stay relatively stable over time.
RQ3: How does a long-term situation of language contact affect a speaker’s idiolectal style (intra-speaker variation)?
H3. LanCon speakers will show greater intra-speaker variation than InSit speakers.
RQ4: How does a long-term situation of language contact affect inter-speaker variation?
H4. Inter-speaker variation will be higher when comparing subjects from different subcorpora than when comparing subjects from the same subcorpus.
RQ5: How does gender affect inter-speaker variation?
H5. Inter-speaker variation will be higher when comparing subjects of different gender (a man vs. a woman) than when comparing men with men and women with women.
RQ6: How discriminatory will each variable be when considering them in isolation?

Perception test
Analysis and quantification
Related to the linguistic mechanism (vs. the physical mechanism) (Nolan 1997) --> Enclosed in the speaker's acquired patterns referring to their specific accent as well as individual choices that constitute their idiolectal style
Evidence that different individuals choose different patterns.
Discriminatory potential of phonetic and phonological characteristics that are undergoing change (Loakes 2006; Loakes & McDougall 2004, 2007, 2010; Moosmüller 1997; de Jong et al 2007a, 2007b).
Despite every individual in the community may eventually incorporate a sound change into their linguistic behaviour, if we look at the process at a synchronic point in time, we might see that speakers use the more conservative and the more innovative variants at different rates.

Speakers have a choice, and the total set of choices that they make constitute their idiolectal style.

Sociolinguistic definition of variable: a linguistic characteristic that varies as a result of a correlation with social, stylistic and linguistic factors.

Sameness of meaning: the variants can be used interchangeably in some contexts.
Sociolinguistic variables
1. High inter-speaker variability.
2. Low intra-speaker variability.
3. Resistant to attempted disguise or mimicry.
4. High frequency of occurrence.
5. Robust in transmission.
6. Easy to extract and measure.
7. As independent as possible.
Forensic variables
Phonological variables
Sociolinguistic analysis
1. Identify the variants of the variable: different realisations depending on different contexts.

2. Define the scope of variation: certain context which might have certain effects on the variants.

3. Devising a coding scheme to classify each variant accordingly. In the case of continuous variables that need to be turn into discrete, need to impose a classificatory criteria and break the continuum into identifiable parts, and make sure the same criteria is consistently maintained. (Chambers & Trudgill 1998; Gordon 2007).

4. Quantification of variables: principle of accountability (Labov 1972a, 1982): all the variants must be taken into account in the relevant environments that have been defined, even the Ø variant.
Forensic analysis
Auditory-acoustic analysis

Auditory: use of investigator's linguistic knowledge to interpret the samples.
Acoustic: use of software for the acoustic analysis of speech as well as the knowledge of physical and acoustic properties of the speech signal and segments.
To add further reliability to the coding of the variables dealing with processes most difficult to categorise:
Vowel alternation
Yod coalescence
Frication of plosives
Cohen's Kappa Coefficient (K): a measure used to examine the agreement between two raters on the assignment of categories of a categorical variable. An important measure in determining how well an implementation of some coding or measurement system works (Elliot and Woodward 2007).
Phi Coefficient
Measures the extent of relation between two sets of discrete variables which take only two values and it is adjusted to the sample size (Siegel & Castellan 1988).
Takes a value between 0 and 1.
Intra-speaker results
RQ2: To what extent is a speaker’s idiolectal style stable over time?
H2 is confirmed. It is pretty stable
RQ3: How does a long-term situation of language contact affect a speaker’s idiolectal style (intra-speaker variation)?
H3 is not confirmed. A long-term situation of LC does not cause higher intra-speaker variation.
Inter-speaker results
General results
RQ1: Is inter-speaker variation higher than intra-speaker variation?
H1 is confirmed: for both subcorpora, inter-speaker variation is generally higher (lower on the IIS continuum) than intra-speaker variation. But:
RQ4: How does a long-term situation of language contact affect inter-speaker variation?
H4 is not confirmed. ANOVA test shows no statistically significant difference between samples within the same subcorpus and samples across different subcorpora.
But: IIS values comparing speakers within the LanCon subcorpus in MT2 are generally higher (showing less variation). ANOVA tests show a significant difference between distributions of IIS values between the InSit and the LanCon subcorpora in MT2.
So, results show quite the opposite as expected: after a long-term situation of language contact, speakers show lower inter-speaker variation.
ANOVA tests show significant differences between the three groups. So:
Comparisons between speakers of different gender show more variation than comparisons between speakers of the same gender. H5 is confirmed.
Comparisons between women show less inter-speaker variation than comparisons between men.
IIS succeeds at discriminating better between samples produced by men.
56% of success rate for samples by women.
80% for samples by men.
RQ5: How does gender affect inter-speaker variation?
UVAL- ForensicLab
Institut Universitari de Lingüística Aplicada / Dept. Traducció i Ciències del Llenguatge Universitat Pompeu Fabra

Speaking tempo as an inter-speaker variation factor
Many of the phonological phenomena analysed may somehow be affected by the rate at which individuals speak. (e.g. allophonic variants of /t/).

Analysis to ascertain that the inter-speaker variation found is due to individual and idiosyncratic choices, and not differences in the subjects' speed of delivery.
Analysis of the discriminatory potential of the variables
Efficiency of IIS
100% of intra-speaker comparisons and 80% of inter-speaker comparisons.
The effectiveness is reduced with samples by women.
Some variables proved not to be discriminatory.
Scope of application is reduced due to:
Corpus of study
16 speakers have been analysed (13 in real time)--> 29 samples
Stratification by gender and language contact.
Nature of samples analysed
30 minutes long on average.
Good quality.
Variables reduced to a specific speech community.
Real-time study of 14 variables related to 6 phonological processes in SSBE.
Some of these processes are widely reported in sociolinguistic studies and dialectological descriptions of SSBE (e.g. t-glottalling, yod coalescence, linking /r/), but others not so much (e.g. frication of /t/ and /k/, vowel alternation, insertion of [t]).
Effect of gender, language contact and course of time in these variables.
The discriminatory potential of most of the processes had never been tested with forensic purposes.
Contribution to the Base Rate Knowledge with reference data.
IIS is proposed as a quantitative tool that can provide important information as regards the variation present in pairs of samples.
Future research
Formulation of other discriminatory variables for other speech communities, accents and languages.

Explore different independent variables such as style.

Study IIS with a wider corpus (different length / quality / channel).

Test IIS with real forensic data.
(Nolan 1983, Rose 2002)
Female speakers
Male speakers
To explore inter- and intra-speaker variation and determine whether it is possible to distinguish between these two types of variation.
Chi square test: compares the observed frequencies with the frequencies that would be expected to appear by chance.
RQ6: How discriminatory will each variable be when considering them in isolation?
Experimental design
Corpus of study
16 speakers of SSBE
Method for the calculation of the IIS
Full transcript