General Methods
Introduction
This chapter outlines the methods common to all the EEG experiments reported in this thesis. Procedures specific to each experiment are described in the relevant methods sections. The studies reported employed the same criteria for participant inclusion in the first instance, and similar materials. The ERP recording parameters, as well as data processing and analysis, were identical in experiments 2, 3, 4, 5 and 6. In Experiment
1, the recording parameters and some elements of initial data processing were different, because the EEG was acquired on a different acquisition system.
Participants
They were all aged between 18 and 30 years of age, right handed, as defined by self report, and were native English speakers who reported normal, or corrected to normal, vision. Participants were screened for neurotropic medication, and any clinical
diagnosis of dyslexia. They were all students from Cardiff University and gave informed consent prior to commencing the experiments. They were paid £7.50/hour for their participation.
Materials
Stimulus lists consisted of words, ranging from four to nine letters in length. They were all low- frequency words (range 1-7 per million), and were drawn from the MRC psycholinguistic data base (http://www.psy.uwa.edu.au/MRCDataBase/uwa_mrc.htm).
110
The words were presented visually, in upper case white letters on a black background in central vision on a computer monitor. The images subtended a maximum of 5° of visual angle horizontally and 0.6° vertically.
All experiments included study and test lists. ERPs were acquired during the test phases only. Details regarding the construction of the study and test lists in order to control for item and order effects will be provided in the method sections in each of the experimental chapters. The experiments contained either two (Experiment 5) or four (Experiments 1, 2, 3,4 and 6) study-test cycles. For all experiments, study and test list combinations were constructed so that across participants all words were encountered as each class of item. These comprised new (unstudied) words, studied words, and some new words that were repeated. Within each study cycle, the order of presentation of study items for each participant in each experiment was determined randomly by the stimulus presentation software. The order of presentation of test items was such that the repeated new words were re-presented after randomised intervening lags of between 7-9 items. Across test lists within each experiment, words appeared both at study and at test (old items), or were presented only at test. In the latter case, they were presented, on different lists, either once (new items) or twice (repeated new items).
Experiment procedures
Participants were fitted with an electrode cap before commencing each experiment (see below). The procedure for all participants started with preparation of the skin with a cotton bud. This procedure helps to minimise the contribution of artefacts to the recorded EEG, by reducing the electrical impedance levels at the scalp. For each
participant, the inter-electrode impedance level at all sites was below 5 KQ at the outset of each experiment session. The participants were then seated in a sound- attenuated recording booth. They were situated approximately one metre in front of a computer monitor. Prior to each experiment, the participants were given a written instruction sheet to read, and the task requirements were also explained verbally.
Participants were asked to relax, keep still, and maintain fixation on the centre of the screen and blink as they would normally throughout the experiment. In all experiments there was a short practice session before the experiment began. In each experiment there was also a short break after each study-test cycle.
Electrophysiological recording procedures
For experiment 1 EEG was recorded from 34 silver/silver chloride electrodes, 32 of which were housed in an elastic cap. For experiments 2 - 6 , EEG was recorded from 27 silver/silver chloride electrodes, 25 of which were housed in an elastic cap. The sites were located at midline as well as left and right hemisphere locations (Jasper, 1958). The electrode locations are shown in Figure 4.1. For both recording procedures the remaining two electrodes were placed on the right and left mastoids.
For experiments 2 - 6 , EEG channels were referenced during acquisition to Fz, and were subsequently re-referenced off-line to the algebraic average of the two mastoids.
Data were sampled continuously at a rate of five ms per point for experiments 2, 3, 5 and 6, and 4 ms per point for Experiment 4. The durations of the epoched data were 1280 ms and 924 ms respectively, all including a 100 ms pre-stimulus baseline period relative to which all mean amplitude measures were taken. EEG and EOG were recorded with a bandwidth of 0.03-40 Hz (-3dB) in all except Experiment 1, where
112
EEG and EOG was acquired referenced to linked electrodes located midway between POz and P03/P04, respectively. EEG and EOG were recorded with a bandwidth range of DC-419Hz; initial sampling rate 2048 Hz. Data was high-pass filtered off-line (0.03Hz) and down-sampled to 200 Hz (5 ms/point). The duration of the recording epoch was 1280 ms, including a 100 ms pre-stimulus baseline period.
Figure 4.1. Selected sites from the international 10/20 system (Jasper, 1958). The sites employed in experiments 2, 3, 4, 5, and 6 reported in this thesis are shown in black. The sites used for Experiment 1 are those used for the remaining experiments with the addition of the 6 sites shown in red.
Data processing
Processing and analysis were completed off-line after the end of each experiment session. Trials containing large EOG artefact, and/or movement artefact, were
identified using set criteria and verified by visual inspection prior to rejection or inclusion. Trials containing A/D saturation, and/or baseline drift (difference between first and last data point) exceeding ±80pV were also rejected. Other EOG blink artefacts were corrected using a linear regression estimate (Semlitsch, Anderer, Schuster, & Presslich, 1986).
In all experiments, the averaged ERPs were subjected to a seven-point binomially weighted smoothing filter prior to analysis, and waveforms were formed for each of the critical response categories for each participant. Participants were excluded from subsequent statistical analyses, both for ERP and for behavioural data analysis, if they did not contribute a minimum of 16 artefact-free ERP trials to each of the critical response categories.
Data analyses Behavioural data
The accuracy and RT data were analysed by employing t-tests (corrected for multiple comparisons where appropriate) and repeated measures ANOVAs. The Geisser- Greenhouse correction procedure was implemented where necessary in all ANOVAs, for accuracy, RT and ERP data (Greenhouse & Geisser, 1959).
Two measures of discrimination (Pr: Snodgrass & Corwin, 1988) were calculated for all experiments. The first was computed to determine whether participants were able to distinguish between old words and new words (p(hit) - (FA)). The second was computed to determine whether participants were able to distinguish between old words and repeated new words (p(hit) - p(rFA)). Measures of response bias (Br:
114
Snodrgass & Corwin, 1988) were also calculated for all experiments, where Br = p(FA)/(l - Pr).
ERP data
Analyses of ERP old/new effects
Two sets of analyses were completed for all the experiments reported in this thesis.
Both set of analyses were performed over three time windows (300-500, 500-800 and 800-1100 ms). The first set of analyses (hereafter ‘global analyses’), included mean amplitude data from 18 electrode sites, split equally over the two hemispheres, with an equal number of sites at anterior, central, and posterior scalp locations (see Figure 4.1).
The analyses included the factors of condition (the conditions included will vary across experiments and will be identified for each experiment), anterior/posterior (AP) dimension (three levels: anterior: F3/F4/F5/F6/F7/F8, central: C3/C4/C5/C6/T7/T8, and posterior: P3/P4/P5/P6/P7/P8), hemisphere (two levels: left/right), and site (three levels: inferior: F7/F8/T7/T8/P7/P8, mid-lateral: F5/F6/C5/C6/P5/P6, and superior:
F3/F4/C3/C4/P3/P4).
The second set of analyses (hereafter ‘midline analyses’) included mean amplitude data from Fz, Pz and Cz. This set of analyses included the factors of condition, and site (three levels: anterior: Fz, central: Cz, posterior: Pz). The outcomes of the analyses at the midline will only be reported if there is an interaction involving condition as well as electrode site. These analyses were completed in order to provide a more sensitive assessment of changes of interest on the anterior/posterior dimension than would be obtained via the global analyses.
The main interest in the experiments reported in this thesis is on ERP differences according to condition. Therefore, only effects involving this factor will be reported.
Reliable two- and three-way interactions involving the critical factor will be followed up with appropriate directed analyses, as described in the relevant empirical chapters.
Directed analyses of mid-frontal ERP old/new effects
One set of directed analyses was carried out for the mid-frontal ERP old/new effect for all experiments. The analyses were carried out in the 300-500 ms time window for all experiments. This time window was selected as it has been shown that for anterior sites in this post-stimulus epoch, ERPs are sensitive to item familiarity (Curran, Tepe, & Piatt, 2006; Rugg et al., 1998). The analyses included data from F3, Fz, and F4 only, and included the factors of condition (as specified in each
experiment chapter) and site (three levels: F3/Fz/F4).
Directed analyses of left-parietal ERP old/new effects
The directed analyses for the left-parietal ERP old/new effect included factors of condition, hemisphere, and site (four levels: P3, P4, P5, P6), and were completed over the 500-800 ms time window. This time window was selected as it captures when and where the left-parietal ERP old/new effect is primarily evident (Friedman & Johnson, 2000; Rugg, Herron, & Morcom, 2002).
Presentation of Tables, Figures and Scalp Topographies
The tables for the critical behavioural data are show in the main body of text, with some additional material in the appendices where necessary. The tables for the accuracy data present the mean proportions of correct and incorrect responses,
116
separated according to condition. RT tables in the main body of the text display the mean RTs for correct and incorrect responses in all conditions.
All experimental chapters include tables showing the outcomes of the global and mid-line analyses on the ERP data. These are located at the end of each chapter. Only the highest order interactions in each contrast of interest are described in the main body of the text. The ERP figures and topographic maps are presented at the end of each chapter, with additional material in the appendices as necessary. All maps shown are computed from difference scores, obtained by subtracting the mean amplitudes associated with one critical condition from another.