Suppression in visual cortex and parietal top-down control

In this vein, experiments have been conducted to examine the brain network responsible specifically for the suppression of unwanted visual information (though not in the context of action). In two studies, participants were required to respond to the angle of one of two visual stimuli, while being scanned with fMRI (Sylvester, Jack, Corbetta, &

Shulman, 2008; Sylvester, Shulman, Jack, & Corbetta, 2007). Which stimulus they were required to attend to was cued with one of two auditory beeps 6-10s prior to the

presentation of the visual stimuli. This pre-cue and subsequent delay allowed the

researchers to measure the neural signature of the allocation of attention to both the cued and uncued target locations during this preparatory period. They found that the allocation of attention led to a correlated difference in the visual cortical representations of the attended (activity enhanced) and non-attended (activity suppressed) locations (Sylvester et al., 2007), and that the suppression of activity at non-attended locations was larger when the expected discrimination was going to be more difficult (lower contrast, Sylvester et al., 2008). These results fit well within a competitive framework where attention toward the cued location is accompanied by a suppression of activity at the non- cued location, especially when the difficulty demands even greater attentional resources. Additional findings from these studies also suggest that the modulations in preparatory activity observed in early visual cortex were generated by top-down attentional control signals from a frontoparietal network (see also Bressler, Tang, Sylvester, Shulman, & Corbetta, 2008; Silver, Ress, & Heeger, 2007). A similar result was found in another experiment (where the likelihood of the appearance of distractors was cued) which demonstrated that the amount of preparatory activity observed in visual cortex was contingent upon whether or not participants expected distractors to appear together with the targets (Serences, Yantis, Culberson, & Awh, 2004).

One neurophysiological study, where recordings were made from the IPS (specifically the lateral intraparietal (LIP) area) of rhesus monkeys during a visual search task (Ipata, Gee, Gottlieb, Bisley, & Goldberg, 2006), also provides neural evidence for the role of the parietal cortex in distractor suppression. In this experiment, search displays always contained a target that was difficult to detect amongst several distractors one of which

was highly-salient and ‗popped out‘. Initially, neuron responses to the pop-out distractor were higher than for non-pop-out distractors. However, with training, the responses of the neuron for the pop-out distractor became slower and smaller than the responses for the non-pop-out distractors. This was taken as evidence that with sufficient knowledge (in this case training) the representation of salient but irrelevant information in the IPS could be suppressed.

Finally, a recent set of experiments has used fMRI and transcranial magnetic stimulation (TMS) to confirm the role of the parietal cortex when human participants responded to one set of visual information while ignoring another (Mevorach, Hodsoll, Allen, Shalev, & Humphreys, 2010; Mevorach, Humphreys, & Shalev, 2005, 2006, 2009; Mevorach, Shalev, Allen, & Humphreys, 2009). Specifically, these researchers used compound letter stimuli where a larger letter is comprised of smaller elements of a different letter (e.g. a large letter ‗H‘ comprised of small letter ‗D‘s, see Figure 1.4). In these compound stimuli there are two orthogonal dimensions, the global, large letter and the smaller, local letters, whose salience can be varied (i.e. blurring the local elements makes the large letter more salient, while making local elements different colours makes them the more salient dimension, see Figure 1.4). Importantly, one can then contrast behaviour and neural activity on trials when participants are responding to the high-salience or low- salience dimension independent of whether that dimension is the global or local target. Brain areas responding to this contrast would then specifically be implicated in the selection (or suppression) of high- or low-salience information. Using repetitive TMS (10 min of stimulation over one site, which leads to a disruption of that area for several minutes) these researchers showed that disrupting the right PPC interfered with

participants‘ ability to respond to the high salient dimension (again, irrespective of whether it was global or local) while disrupting the left PPC interfered with participants‘ ability to respond to the low salient dimension (Mevorach et al., 2006). The most

relevant conclusion to the current thesis, that the left PPC biases selection away from the salient dimension (i.e. enabling one to ignore distracting information), was replicated and extended in two follow up studies. First, using transient TMS (where a brief burst of TMS is given at different points in a trial), Mevorach et al. (2009) showed that the deficit in responding to low-salience information after left PPC disruption was maximal with

TMS prior to stimulus onset, indicating that the suppressive response was preparatory in nature. Second, using fMRI, these researchers (Mevorach, Shalev et al., 2009) further localized the preparatory left PPC activity to the IPS.

Figure 1.4. Compound letter stimuli used by Mevorach and colleagues (from Mevorach et al., 2010). A large global letter (in this case H) is made up of smaller local-element letters that are either congruent (in this case H) or incongruent (in this case S or D) with the larger letter. By providing a high resolution stimulus (top row) with different coloured local elements, the local dimension can be made more salient. By blurring the stimulus (bottom row) the global dimension can be made more salient. A contrast can therefore be made between high and low salient information that is independent of the global / local level of the information.

One recent study (Mevorach et al., 2010) brings together the work showing a suppression of visual cortex activity at the location of an anticipated distractor (Sylvester et al., 2008; Sylvester et al., 2007) and the increase in left PPC activity when suppression is required (Mevorach, Shalev et al., 2009). In this study, Mevorach and colleagues (2010)

combined TMS and fMRI to specify a causal relationship between left IPS activation and visual cortex suppression. First, in a pre-TMS fMRI scan, they confirmed that the left IPS was activated more on trials where participants were required to respond to low salient, as compared to high salient information. From this data, they also demonstrated that the increased left-IPS activation was correlated with a decrease in activity in visual cortex. Finally, using repetitive TMS followed immediately by an fMRI scan (such that the disruptive TMS effects were still evident) they showed that left-IPS disruption led to an increase in the blood oxygenation level dependent (BOLD) fMRI response in visual cortex, but only when participants were required to respond to the low-salience

dimension of a compound stimulus. That is, in cases where the left IPS should have been generating the suppressive response required to ignore the irrelevant high salient

information, its disruption led to an increased visual response, corresponding, they

argued, to a lack of suppression. Taken together, these TMS and neuroimaging studies of the suppression of distracting visual information suggest that the detection of irrelevant information is generated by preparatory activity in the PPC (specifically the left IPS) which leads to the suppression of the corresponding retinotopic location in early visual cortex.

It should be noted that there are many other tasks which likely rely on suppressive mechanisms which are similar to those described in this section, including Go/No-Go (or cancellation, e.g. Curtis, Cole, Rao, & D'Esposito, 2005) tasks (where the participant is required to countermand a planned response, e.g. Rubia et al., 2001) anti-saccade tasks (where the participant is required to make an eye movement to a location opposite a visual cue, e.g. Connolly, Goodale, Desouza, Menon, & Vilis, 2000; Connolly, Goodale, Menon, & Munoz, 2002) and, perhaps most relevantly, anti-pointing tasks (e.g. Connolly et al., 2000). Interestingly, as was the case for the distractor suppression studies

described above, a frontoparietal network (in which the PPC plays a crucial role) has been implicated in all of these related tasks. Key differences between these tasks and

obstacle avoidance, however, make a further discussion of them beyond the scope of this thesis. In anti-pointing and anti-saccade tasks, while a movement to the visual cue must be suppressed, there is also considerable remapping required to generate a movement to the new anti-location, which is not required in an obstacle task where no target

remapping is needed. For Go/No-Go tasks neural activity resulting from the complete cancellation and corresponding lack of response is likely to be very different from obstacle tasks where an action is always performed, and only a single object, and not an entire action, must be suppressed.