Eye movement data from visual world paradigm experiment on
anticipation and priming in L2 speakers and attrited L1 speakers of
Italian

Antonella Sorace, Thomas Bak, Frank Keller, Ian Finalyson
University of Edinburgh


OVERVIEW OF EXPERIMENT

The aim of this experiment was to compare the anticipation behavior of
two populations of English-Italian bilinguals: Advanced L2 learners of
Italian with English as native language, and attrited native speakers
of Italian with English as the L2. In a visual world experiment,
participants were eye-tracked while they heard prime sentences in
either English or Italian, followed by target sentences in Italian. We
manipulated whether the prime was in the same language as the target
(language switch or not), and whether the prime used the same
syntactic construction as the target (construction switch or not).

In the analysis, we related the degree of anticipation across
conditions to the type of bilingualism in the speaker (L2 or
attrited), and to their conflict score as measure by the Attention
Network Test (ANT), a test of executive function.


FILE STRUCTURE

README                  this file
analysis.R              R script for running the mixed model analysis
byItem.csv              data file with by-item means
bySubj.csv              data file with by-subject means
consent_form.pdf        consent form used for the experiment
information_sheet.pdf  instruction sheet used for the experiment
methods.pdf             brief description of methods, analysis and results

A detailed report on this experiment will appear at:

http://www.lel.ed.ac.uk/~antonell/


DEFINITION OF VARIABLES

Subj            Subject code
Item            Item code
group           Participant group (attriter: -1; learner: 1)
lang            Language switch (no-switch: -1; switch: 1)
cons            Construction switch (no-switch: -1; switch: 1)
CURRENT_BIN     Which 50ms bin does data come from
NP2.C           Number of trials where target item was fixated
N               Total number of trials per condition
elNP2           Empirical logit of proportion of looks to target
wtsNP2          Inverse of regression weight
t1              Linear component of time (as an orthogonal polynomial)
t2              Quadratic component of time (as an orthogonal polynomial)
conflict        Subject's standardized conflict score from ANT
                (higher scores = better performance)


DATA ANALYSIS

The 1600ms timecourse was divided into 50ms bins. For each bin in each
trial, it was recorded whether or not the target object was fixated.
Data was subsequently aggregated by subjects and by items (in
bySubj.csv and byItem.csv, respectively). For each condition, the
number of trials in which the target object was fixated was
counted. These counts were used to calculate empirical logits, and
appropriate weights for a regression analysis.