Tops Down

In Bottoms Up, we learned that "bottom-up" attention -- useful for detecting "pop-out" stimuli in the visual field -- is bottom-up (parietal neurons respond first, then prefrontal neurons), and that "top-down" attention required for effortful visual search is top-down (prefrontal, then parietal) [and that this merited publication in Science (Buschman & Miller, 2007).]


The timing of events in the lateral interparietal area (LIP), lateral prefrontal cortex (LPFC), and frontal eye fields (FEF) was as follows:

figure from Yantis (2003)

LIP, LPFC, and FEF neurons began finding the target 170 ms, 120 ms, and 35 ms before saccade, respectively' [pop-out] ...target location information reached significance in the FEF and LPFC at 50 and 40 ms before the saccade, respectively, followed by LIP at 32 ms after the saccade. [visual search]¹

It's nice to get the onset information all in one experiment, but did we need single-unit recording in awake behaving monkeys to tell us this? Unique opportunities to record intracranially in awake behaving humans occur clinically in the neurosurgical arena, to monitor for seizures in patients with intractable epilepsy (Dubeau & McLachlan, 2000). In a series of such experiments in the mid-90's, Halgren and colleagues recorded local field potentials from over two thousand cortical and subcortical sites while the patients performed an "oddball" task (reviewed in Halgren et al., 1998; see the three original papers for a full view of this tour de force).

In the oddball task, a series of simple auditory or visual or somatosensory stimuli are presented. Participants attend to rare target stimuli embedded in a train of standard stimuli, and press a button when a target is detected. A brain wave called the P300 (or P3b) can be recorded from the scalp at approximately 300 msec after target presentation. Rare, task-irrelevant distractor stimuli can also be presented, and these elicit the P3a component. Most germane here is the P3a, because it's often thought to be part of the automatic orienting response (Halgren et al., 1998):

The P3a is evoked by rare stimuli, regardless of whether they are targets or non-targets, overtly attended or unattended, auditory or visual. It is generated in a frontoparietocingulate system that has been associated with the orientation of attention... It is associated with an electrodermal response and represents the cortical component of the orienting response.

Employing a task with pure tone stimuli, Halgren et al. plotted the peak latencies of P3a at different recording sites, and demonstrated that frontal locations had earlier peaks than posterior locations. In essence, here we have evidence for "top-down" attentional mechanisms in a "bottom-up" situation.


from Halgren et al. (1998): The P3a has a significantly shorter latency in frontal sites (including anterior cingulate gyrus, aCg, and Brodman’s area 46 in the dorsolateral prefrontal cortex, a46), than in parietal sites (including posterior cingulate gyrus, pCg, and supramarginal gyrus, sMg), or temporal sites (including parahippocampal gyrus, pHg). At all sites, the depth P3a is earlier than the scalp P3.

Plus [for starters], there's some evidence for reverse hierarchies in the visual system (Hochstein & Ahissar, 2002)...

...we suggest a reversal of the way of understanding conscious perception and its relationship to cortical mechanisms. Based on results from feature search, vision at a glance and vision with scrutiny have been viewed as reflecting, respectively, low-level and high-level cortical representations. Thus, effortless simple feature detection has been seen as reflecting mechanisms operating at lower levels. Subsequent studies finding that the pop-out phenomenon also occurs for complex features challenged this view, while accumulating evidence for global precedence was viewed as an oddity.

We propose instead that vision at a glance reflects high-level mechanisms, while vision with scrutiny reflects a return to low-level representations. ... Thus, early spread attention reflects the large receptive fields found in high-level areas, and focused attention reflects localized low-level representations. High-level spread attention subserves our initial, crude global percept of the gist of the scene. Pop-out is but one aspect of this crude initial assessment. Associating early conscious perception with high cortical level mechanisms has implications for attentional phenomena as well.

...but that's a topic for another post!

References

Buschman TJ, Miller EK (2007). Top-Down Versus Bottom-Up Control of Attention in the Prefrontal and Posterior Parietal Cortices. Science 315: 1860-1862.

Dubeau F, McLachlan RS. (2000). Invasive electrographic recording techniques in temporal lobe epilepsy. Can J Neurol Sci 27 Suppl 1:S29-34.

Halgren E, Marinkovic K, Chauvel P. (1998). Generators of the late cognitive potentials in auditory and visual oddball tasks. Electroencephalogr Clin Neurophysiol 106:156-64.

Hochstein S, Ahissar M. (2002). View from the top: hierarchies and reverse hierarchies in the visual system. Neuron 36:791-804.

Yantis S (2003). To see is to attend. Science 299:54-56.


Footnote

¹ When the trials were time-locked to stimulus onset rather than response onset, the results were as follows:

For the pop-out condition, while the distribution was more variable due to the variability in reaction time, LIP showed selectivity for the target location approximately 50 ms after array onset, followed by LPFC and then FEF (after 120 and 220 ms, respectively). All these differences were also significant ... When aligning search trials on visual array onset, LPFC and FEF carried significant information at 250 ms after array onset, significantly preceding selectivity in LIP, which began at 320 ms after onset...