5. Políticas y programas públicos sobre el trabajo infantil
3.3. Explotación Infantil: un debate pendiente
3.3.2. Las condiciones laborales
4.3.1. Behavioural Data
Behavioural measures at study included the response time (RT) for making JOLs and response distribution across the 5-point JOL scale. At test, behavioural measures included overall recognition accuracy, recognition accuracy across JOL and RT for making old/new discriminations (in Experiment 1, overall cued recall accuracy and cued recall accuracy across JOL were also examined). These measures were taken primarily to confirm that participants behave consistently across experiments and in a way that is comparable to standard observations in the JOL literature. Analyses were carried out using repeated measures ANOVA with a significance criterion of 0.05. Post-hoc comparisons were carried out using t-tests with Bonferroni-corrections. Metamemory accuracy was assessed by calculating both the mean Gamma correlation coefficient and da (see Chapter 1). Specific details of the analyses will be outlined in the relevant data chapters.
4.3.2. ERP Data
The purpose of the ERP investigations reported in this thesis was to examine JOL related neural activity at both study and at test. At study, the rationale was to compare SM effects to any possible effects associated with JOLs. Contrasts were therefore made between i) items that were and were not subsequently remembered
(recalled in Experiment 1 and recognised in Experiments 2-4), and ii) items that were assigned low and high JOLs6. The explorative nature of the ERP research implied that no pre-experimental hypotheses were formulated regarding time windows that were submitted for analyses. Time windows were thus identified primarily on basis of visual inspection of the grand average waveforms (and varied across the four experiments).
At test, comparisons were made between old items correctly identified as old (through cued recall or recognition) and correctly rejected new items. Correctly identified old items were further subdivided into items that were assigned low and high JOLs at study. This division allowed the investigation of possible modulations of the well-characterised retrieval effects caused by JOLs. Choice of time windows submitted to analyses was primarily based on previous literature (Rugg & Curran, 2007) and corresponded well to the visual inspections of the grand average waveforms (the only exception being the picture version of Experiment 4). Time windows were as follows: 300-500 ms (mid-frontal familiarity effect), 500-800 ms (left-parietal recollection effect) and 800-1400 ms (right-frontal post-retrieval monitoring effect).
ERPs from study and test were first quantified by calculating, for each response condition, the mean activity during each latency period. The data were then submitted to repeated measures ANOVA. Typically (deviations are reported in the relevant data chapters) the initial analyses included five factors of location (frontal,
6 An alternative approach to comparing SM effects and JOL would be to divide remembered and not remembered items into high and low JOLs, however this strategy caused a significant loss of data
fronto-central, central, centro-parietal and parietal), two factors of hemisphere (left and right) and three factors of site (superior, medial and inferior, see Figure 4.2) in addition to a condition (response category) factor. Only main effects and interactions involving the factor of condition are reported. When interactions involving location were evident, the initial analyses were followed up by subsidiary analyses, examining each separate location (with two factors of hemisphere and three factors of site). The electrodes submitted for analyses were selected because they cover a large area of the scalp and, in most cases, seemed to capture the effects of interest (alternative electrodes were identified when effects exhibited foci on scalp locations that were not covered by the original set of electrodes). Using factors of location, hemisphere and site allows ERPs to be compared in terms of potential hemispheric and anterior-posterior differences and also give indications of the effects’ proximity to the midline.
Figure 4.2 Schematic illustration of the electrodes included in initial ERP analyses.
The front of the head is pointing upwards and left side is shown at left. Each circle represents an electrode and electrodes included in the analyses are marked with green. Electrodes from frontal, fronto-central, central, centro-parietal and parietal electrode rows provided five levels of a location factor, electrodes on left and right hemisphere provided two levels of a hemisphere factor and electrodes in each quadrant provided three levels of a site factor (superior electrode sites closest to the midline, medial electrode sites and inferior electrode sites).
The ANOVA model’s underlying assumption of sphericity (the requirement of homogeneity of co-variance for all factors) is usually violated in the case of ERP analyses. The consequence of this violation is an increased probability of a type 1 error and for that reason Greenhouse-Geisser corrections (Greenhouse & Geisser, 1959) are reported when necessary, to ensure a more conservative test of significance. As for the behavioural data, the significance criterion for all ERP analyses was set at 0.05.
To investigate potential qualitative differences between conditions or latency periods, topographic analyses were performed on difference waves (mean amplitudes of condition two subtracted from mean amplitudes of condition one) when robust ERP amplitude differences had been established. Prior to any topographic analyses, the data from all 62 active electrodes were normalised using the max/min method (McCarthy & Wood, 1985) described in chapter two. The analyses employed the same design as the ANOVA used to evaluate amplitude differences and only interactions involving factors of condition or latency period are reported.
4.4. Summary
The present chapter has provided an outline of the stimuli materials, experimental paradigms, EEG acquisition procedures and analyses that were employed in the research reported in the remaining chapters of this thesis. Although most experiments conform to the general methods, occasional exceptions exist and are highlighted in the relevant chapters.