Based on the studies reviewed in the preceding sections, we wanted to address the following questions, each in a separate experiment:
Does perceptual learning in the human visual system involve the earliest stages of visual cortex activity? This question was based, first, on animal liter-ature suggesting profound effects of perceptual learning on early activity in the visual system (Gilbert et al., 2001; Ito et al., 1998; Schoups et al., 2001); secondly, on be-havioural evidence in humans suggesting that retinotopically organised visual areas may be implicated in perceptual learning (Censor, Karni, & Sagi, 2006; Karni & Sagi, 1991;
Karni, Tanne, Rubenstein, Askenasy, & Sagi, 1994); and third, on previous fMRI studies
Introduction
(Furmanski et al., 2004; Schwartz et al., 2002) indicating changes in V1 activity follow-ing trainfollow-ing on the texture discrimination task developed by Karni and Sagi (1991). We expected to observe retinotopically specific changes in C1 amplitudes which, within the limits of precision afforded by this component (see Foxe & Simpson, 2002), would indi-cate modulation of the first sweep of activation passing through primary visual cortex as a result of prior task exposure. Note that although perceptual learning may appear as a very simple form of learning, it cannot be explained in terms of pure bottom-up processes but requires substantial top-down input (Herzog & Fahle, 1998; Li et al., 2004).
Can endogenous attention induce filtering effects involving the earliest stages of visual cortex activity? This question directly relates to a previous fMRI study (Schwartz et al., 2005) which assessed the effects of attentional load (Lavie, 1995, 2005) on the processing of peripheral, irrelevant distractors. The load theory suggests that attentional selection does not generally occur early (Treisman, 1969) or late (Deutsch
& Deutsch, 1963) during perception, but is adapted to the attentional demands of the current task. As previous studies had overwhelmingly reported no effects of endogenous attention on early visual cortex activity, we wanted to test this assertion using a task that would require attentional suppression rather than enhancement of stimuli. Again, we hypothesized that changes in attentional task demands would affect early visual cortex activity as indexed by the C1 component. Pilot studies for this second experiment suggested a third important question:
Do endogenous and exogenous attentional task demands interact at the earli-est stages of visual cortex activity? We observed that slight changes of stimulation parameters induced large differences in attentional effects on early visual cortex activity and decided to test this observation in a separate experiment. Again using an atten-tional load paradigm, we hypothesized that higher demands on exogenous attenatten-tional mechanisms would disrupt any effects of endogenous attention on the C1 component, in accordance with exogenous attention serving the role of a ‘circuit breaker’ (Beck &
Kastner, 2005; Connor, Egeth, & Yantis, 2004; Corbetta & Shulman, 2002; Serences et al., 2005). Previous studies had failed to demonstrate interaction effects this early in the visual hierarchy (Doherty, Rao, Mesulam, & Nobre, 2005; Hopfinger & West, 2006).
Stronger recruitment of exogenous attention was operationalized by simultaneous pre-sentation of target and distractor stimuli, as previous studies had indicated strong effects of synchrony on exogenous attention systems (Fournier, 1994; Kahneman, Treisman, &
Burkell, 1983; Burg, Olivers, Bronkhorst, & Theeuwes, 2008; Wilson & Singer, 1981), possibly linked to the Gestalt law of common fate (Blake & Lee, 2005).
Chapter 2
Methods
This chapter will give an overview of the experimental methods employed to study the questions and hypotheses detailed in Section 1.6. I will first describe the behavioral tasks used in each of the three experiments and then provide details on data acquisition and analysis procedures. All experiments were conducted in accordance with the Declaration of Helsinki and approved by the local ethics committee. Subjects provided informed consent and received a small payment for their participation. For details of the respective subject groups, please see the individual papers in the Appendix.
2.1 Experimental Stimuli and Tasks
Experiment 1 addressed the neural underpinnings of perceptual learning and employed an experimental task well-established in this field of research: the so-called texture discrimination task (TDT; Karni & Sagi, 1991). As shown in Figure 2.1, this is a dual task in which subjects are required to fixate a stream of rapidly presented letters while covertly monitoring peripherally presented stimuli containing a target texture, i.e. a small part of the overall stimulus that differs in its orientation from the background elements.
In our study, the letters presented were either an uppercase ‘L’ or an uppercase ‘T’
in varying orientations. Subjects indicated the identity of the letter with their right index and middle finger, respectively. Peripheral textures consisted of white horizontal line elements, with three diagonal elements hidden in the array. Here, participants indi-cated the overall orientation of the diagonal elements (i.e. whether they were arranged horizontally or vertically) with their right ring finger and pinky. Due to the high presen-tation rate and the fact that central letters and peripheral textures were simultaneously presented, this task is extremely difficult. In addition, a mask of randomly oriented
‘V’ shapes immediately followed this display in order to exclude any effects of retinal after-images.
Adequate performance on this task requires extensive training, which was provided during a first experimental session where target textures were selectively presented in one visual quadrant. It has been shown that initial training on most perceptual learning
Methods
Figure 2.1: The texture discrimination task used in Experiment 1. Participants had to detect the identity of a letter presented at fixation while covertly monitoring the periphery for three diagonal bars which could be presented either in a horizontal or vertical arrangement.
task requires sleep for consolidation of learning effects (Karni et al., 1994; Karni & Sagi, 1991; Mednick, Nakayama, & Stickgold, 2003). We therefore asked subjects to return the following day after a night of normal sleep and recorded EEG while stimuli were presented either in the trained quadrant or the opposite quadrant of the same visual hemifield. Two groups of subjects were trained and tested in either the upper or the lower visual hemifield. In order to exclude task difficulty as a confounding factor, stimuli in both quadrants were presented well above threshold during the EEG testing session. A comparison of visual evoked potentials between trained and untrained quadrants should thus yield an indication of the effects of perceptual learning on early visual processing.
Experiment 2 examined the effects of attentional load (Lavie, 1995; Lavie & Tsal, 1994; Lavie, 2005) on early visual processing. As shown in Figure 2.2, stimuli consisted of T-shapes rapidly presented at fixation, which differed in colour and orientation. In the periphery, textures similar to those used in Experiment 1 were shown with a variable stimulus onset asynchrony (SOA) with respect to the central task stimuli. However, in this experiment, peripheral textures were completely irrelevant and subjects were asked to ignore them as best as possible. Again, separate groups of subjects were tested in either the upper or the lower visual field.
On different experimental blocks, participants performed either an easy pop-out de-tection task, responding to the colour of the T-shapes independently of their orientation, or a difficult conjunction detection, where both the colour and the orientation of the T-shapes had to be monitored. According to the load theory of selective attention (Lavie, 2005), higher demands on attentional resources lead to increased filtering of irrelevant stimuli. Comparison of visual evoked potentials from experimental blocks with low vs.
high attentional load should thus allow for an assessment of the temporal characteristics of such increased attentional filtering.
For Experiment 3, we used essentially the same task and stimuli as for Experiment 2, with one important modification: central task stimuli and peripheral distractors were now presented simultaneously, based on psychophysical evidence that this would increase demands on exogenous attention (Fournier, 1994; Kahneman et al., 1983; Wilson &
Methods
Figure 2.2: Stimuli and task structure used in Experiment 2. T-shapes of different colour and orientation were presented at fixation, followed by task-irrelevant peripheral distractors. Subjects either monitored only the colour or both colour and orientation of T-shapes, yielding different degrees of attentional load which were expected to affect processing of the peripheral distractors.
Singer, 1981). As exogenous attention has previously been shown to affect early visual processing (Khoe et al., 2005), effects of attentional filtering under high load should be reduced if distractors are presented simultaneously.