VOLUMEN 1 - RESUMEN NO TÉCNICO
1. INTRODUCCIÓN Y OBJETIVOS DEL PROYECTO
This chapter summarised the common methods used in the following three chapters. The differences specific to each study were documented separately in the Method sub-section in the respective chapter.
3.1 Participants
Table 3-1 presented the participant demographic data. Study 1 (Chapter 4) excluded one participant from the analysis because of chance-level responses.
Study 2 (Chapter 5) conducted three experiments. The second experiment in Study 2 excluded two participants because one participant performed inadvertently the varied cue condition twice and another dropped out without attending the second visit. Thus, the second experiment in Study 2 analysed 19 valid participants. The third experiment recruited 26 volunteers. Six of them dropped out before the second visit and one completed, relative to other participants, the experiment with less per-condition trials (70). Thus, the third experiment analysed also 19 valid participants. All volunteers took part in exchange of course credits or cash, reported normal or corrected-to-normal vision and signed a consent form before carrying out in the study. The procedure was reviewed and granted permission to proceed by the Ethics Review Committee at the University of Birmingham.
Table 3-1. Participant demographic data. Exp: experiment; P: participant; S: study; F/M:
female to male ratio; L/R: left hander to right hander ratio.
P number Age (years; Mean ± SE) F/M L/R
S1 40 18-22 (18.9 ± 1.01) 33/7 5/35
S2 Exp 1 20 18-31 (23.27 ± 3.85) 18/2 5/15
Exp 2 22 18-32 (20.18 ± 3.33) 13/9 5/17
Exp 3 26 18-27 (20.42 ± 2.50) 14/12 0/26
S3 10 19-20 (19.4 ± 0.52) 8/2 0/10
3.1.1 Summary
A clear association was observed between the non-decision time and the shift parameter when the other distributional parameters were kept constant. As for the other two distributional parameters, the simulation study do not indicate a clear distinction for the scale and shape parameters with the decision-making process. Specifically, an Increase in the scale parameter raised the boundary separation and slowed down the drift rate, suggesting a negative correlation between the scale parameter and RTs. In contrast to the scale parameter, an increase in the shape parameter suggests it correlates positively with RTs, because of the decrease in the boundary separation and the increase in the drift rate, although the increase in the shape parameter appears also to increase the non-decision time.
Overall, the current evidence supports a clear link between the shift and the non-decision time but seems to suggest that an overarching influences of the scale and shape parameters on the decision-making process. Even though the detailed examination of the latter two parameters indicates that an increase in the
scale parameter affects more the decision components and less the non-decision component and that an increase in the shape parameter affects both, it is far from clear-cut to assert the scale or the shape parameter reflects a distinctive cognitive process.
3.2 Apparatus and Stimuli
In Study 1 and Study 2, the order and timing of the paradigms were controlled by PsyToolkit (Stoet, 2010), which is a GNU C library designed to implement cognitive paradigms. The paradigms were carried out on a Linux PC, using a kernel specifically tweaked to a hard real-time system (Linux kernel 2.6.31-11-rt). A hard real time system will treat a designated computer programme as system critical, thereby responding immediately when the programme is called. I designated cognitive paradigms as critical programmes when they were running. In contrast to general operating systems, such as Windows 7 and Windows 8, the user is not allowed to tweak their kernel. This type of operating systems prioritises system-related programmes, but not the programmes running cognitive paradigms. In some situations, the programming running cognitive paradigms may be interrupted by, for example, system updates.
This is not good news, because the programmes executing system updates mostly own a higher priority than those running cognitive paradigms. Further the system-related programmes are controlled by IT staffs (maybe remotely by the staffs in software companies), suggesting that the time the programmes interrupt experiments is not completely random.
The graphic card, NVidia GeForce 8500 GT, controlled visual displays in Study 1 and Study 2. Study 3 used E-Prime 2.0 on a Windows 7 personal
computer, equipped with an NVidia GeForce GT430 graphic card to control the timing of stimulus presentation. All participant responses were made using, a Cedrus RB-830 response pad. All experiments presented stimuli on a Sony CPD-G420 CRT monitor set at the resolution of 1152 × 864 pixels and 100 Hz refresh rate, except the second experiment in Study 2, which presented stimuli on a Dell P991 CRT monitor at the 1024 × 768 resolution and 85-Hz refresh rate.
Participants sat about 60 cm in front of the monitor in a well-lit cubicle and were asked to make speeded responses without compromising their accuracy.
In Study 1 and Study 2, the visual stimuli were presented on a 2.526°
invisible circle in black and grey (white in Study 2) colours onto a grey background (RGB, 190, 190, 190). The visible area contained the entire screen, but the relevant stimuli were all drawn within the viewing area of 7.59° × 7.59°. Study 1 used the visual stimuli similar to the benchmark paradigm (Wolfe et al., 2010).
Study 2 used 13 English uppercase letters, A, B, D, E, F, G, H, J, M, N, Q, R, &
T, sized 0.63° × 0.63°. In Study 3, the search items (A, B, D, E, F, G, H, M, N, R, & T) were scaled to 0.32° × 0.46° in black colour, presented on E-prime’s default grey colour background. The search items were randomly allocated to 10 possible locations on an invisible circle (see Figure 4-1 for an illustration). While viewing an imperative stimulus, participants indicated whether the target was present or absent (Study 1) or whether the target was on the left or the right side of the invisible circle (Study 2 and Study 3).
3.3 Design
To minimise one of the experimenter biases related to the analysis of null hypothesis significance testing (NHST; Kruschke, 2010), the studies set a fixed
target sample size (20 participants in Study 1 and Study 2; 10 participants in Study 3) before collecting data. The target sample size was determined based on commonly used sample sizes (approximately 5 to 20 participants) in visual search literature. The data from participants who withdrew and completed only part of the tasks were not analysed; these participants were replaced with other individuals.
3.4 Decision-making Models
Three decision-making models will be applied in the thesis in separate chapters. Chapter 4 applied the simplified decision-making model, EZ2 to estimate the decision parameters in the benchmark search parameter. The EZ2 model differs slightly from the other two decision-making models, applying in Chapter 5 (DDM and LBA) and Chapter 6 (the fast-dm version of DDM). EZ2, because of the mathematical simplification, estimated the drift rate, non-decision time and boundary separation. Specifically, the boundary separation merges the decision threshold and the initial bias (as estimated separately in DDM and LBA) as one parameter. Study 2 (Chapter 5) fited data with LBA and DDM and used the model selection procedure described in Donkin, Brown and Heathcote (2011) to balance the good fits and parsimonious factors to fit model. Study 3 (Chapter 6) allowed all experimental factors to depend on most DDM parameters, thereby fitting data with a saturation model.