CAPÍTULO VII 7.1 TÍTULO DE LA PROPUESTA.
7.6.1.6 SOBRE ANCHOS EN CURVAS DATOS:
7.6.3. ESTUDIO DE SEÑALIZACIÓN Y SEGURIDAD VIAL
Although this thesis makes a comprehensive contribution to the understanding of reaction time interference in the distractor search paradigm by considering both the probabilistic occurrence of attentional capture for the first selection and the subsequent target selection,
there are still open questions about the source of reaction time interference in the distractor search paradigm and its neural correlates and there is potential to increase predictability of attentional selection based on the probabilistic theory of salience.
The likelihood of attentional capture, as determined by the probabilistic theory of salience, is solely based on the bottom-up, environment-driven factor of salience. However, there is evidence that top-down control operates early in the attentional system (Müller & Krummenacher, 2006) and visual search theories such as Guided Search (Wolfe, 1994) assume selection to be influenced by top-down activation. Even authors of pure bottom-up salience models acknowledge the impact of top-down control (Itti & Koch, 2001; see also Fig. 1). Neurophysiological approaches suggest a ‘priority’ instead of a ‘salience’ map, which combines bottom-up salience and task relevance in order to determine the location for the attentional focus (Fecteau & Munoz, 2006). Also the brain area, which is suggested as site of the salience map combines bottom-up and top-down signals (Bisley & Goldberg, 2010) in a way that resembles the dynamics of a decision process (Gold & Shadlen, 2007). Therefore, a next step in developing the probabilistic theory of salience would be to include observer-guided control. This could be implemented in the model for example by a faster drift rate of the target accumulator, which would imply that the target’s stimulus salience is higher and consequently its selection probability is increased.
Besides the model adaptation for the first selection, the second selection needs to be investigated more extensively as well, since this was the very first study examining this selection (A. I. Koch, Goschy et al., 2013). The qualitative inspection of the distributions of reaction time interference suggests that it is only the first selection that varies as a function of relative salience, because increase of reaction time interference over the first percentiles is varyingly strong for the various salience differences, but the amount of decrease in later percentiles seems to be the same for all salience differences. However, this is only a qualitative inspection and it is possible that especially the disengagement process from the distractor is also salience-dependent to a certain degree.
In general, it would also be interesting to investigate the selection sequence in the distractor search paradigm with respect to neural correlates of the reaction time interference
pattern. EEG would be an appropriate method with a high temporal resolution. The EEG event-related potential N2pc, which is regarded as a marker of spatial orientation of attention (Eimer, 1996), could provide confirming evidence about proportionate attentional capture: the N2pc indicates attentional deployment by negative deflection of cortical activity contralateral to the attended location. In case of attentional capture by the distractor, this event-related potential should occur for the distractor and subsequently for the target location. If the distractor did not capture attention, only an N2pc for the target should be visible. Moreover, if the size of the proportion of capture is dependent on distractor relative to target salience, the amplitude of the distractor N2pc should vary as a function of this salience difference, because - analogously to reaction time interference – a high mean amplitude should result from many capture trials and a low amplitude from few capture trials. If slowed target processing also contributes to reaction time interference, as predicted by non-capture theories, there should also be a latency effect of the N2pc observable that is its onset should vary with relative salience.
Another method neural correlates of reaction time interference could be investigated with is functional magnetic resonance imaging (fMRI). This method has the advantage of high spatial resolution and would help to identify areas in the brain that process salience-based attentional selection. Manipulating global and local salience in Navon figures, Mevorach, Humphreys and Shalev (2006) found reaction time interference from the low salient distractor to increase when transcranial magnetic stimulation (TMS) was applied to the right posterior parietal cortex (PPC). On the other hand, interference from the higher salient distractor increased when TMS was applied to the left PPC. The authors concluded that the right PPC is involved when orienting towards a salient target stimulus and the left PPC being responsible for avoidance of salient stimuli. In an fMRI study using the same paradigm (Mevorach, Shalev, Allen, & Humphreys, 2008), the left intraparietal sulcus was significantly more activated when attending to a less salient target in the presence of a higher salient distractor compared to a higher salient target accompanied by a less salient distractor. Moreover, the blood-oxygen-level-dependent (BOLD) response difference between distractors more salient and less salient than the target correlated positively with reaction time interference. The advantage of the Navon Paradigm is that effects of salience on the PPC activation can be discerned from spatial influences, however salience of target and
distractor is correlated and a manipulation always affects both stimuli. This problem is circumvented in the distractor search paradigm, where target and distractor are spatially separated and can therefore be manipulated independently with respect to their salience. Based on the findings of Mevorach et al. (2006) and Mevorach et al. (2008), activation of the right PPC should increase with decreasing distractor salience in the distractor search paradigm, if the right PPC is responsible for orienting towards salient items. On the other hand, if the left PPC is responsible for avoiding salience, its activation should increase with increasing distractor salience. Therefore, activation in right and left PPC should be negatively correlated if they had dissociative and complementary roles in attentional salience-based selection. Additionally, as Mevorach et al. (2008) showed, the BOLD-response should correlate with reaction time interference.
Last but not least the theory of probabilistic salience needs to be tested on other paradigms than the distractor search paradigm to generalize its field of application. The ultimate objective should be to predict the selection sequence in natural images.