Biological Sciences 300/301, Smith College | Neurophysiology

Revised: April 20, 2009

TOP-DOWN COMPONENTS OF VISUAL PROCESSING

 
Experimental work on visual processing began with recordings from retinal ganglion cells, continued with Hubel and Wiesel's pioneering recordings from the lateral geniculate and primary visual cortex, and then moved into extrastriate cortex. The neurons in these areas seemed to be doing a "bottom-up" task, initially detecting light-dark borders in the retina, extracting edges in primary visual cortex, and continuing toward the detection and identification of objects in the inferotemporal cortex. The general consensus was that the perceived image was "built" by an ascending series of neurons with increasingly complex receptive fields.

In recent years, evidence has also emerged that in addition to this "feedforward" bottom-up progression, there is also significant "top-down" feedback from higher visual areas to lower ones. Even V1 neurons show signs of modulation by higher areas. But what is the function of top-down modulation? One proposal is that visual perception requires rapid bottom-up evaluation of an image's low spatial frequencies, followed by top-down modulation of information that is sent bottom-up to neurons in the temporal cortex, the neurons that are associated with visual perception of objects. In this theory, perception requires both feedforward and feedback components.

A completely different role for top-down feedback has been proposed: it could be to support visual attention, where the viewer focuses on a particular aspect of the visual world (for example, the location or the shape of an object). Attention appears to be generated in higher visual areas and modulates the responses of neurons at earlier stages in the visual pathway. Visual attention would be a top-down process that enhances responses of neurons coding for the attended object, while suppressing responses to other objects in the visual field.

The readings below are in two groups. The first group deals with top-down contributions to visual perception, while the second group is a brief glimpse of the extensive literature on visual attention. The assignment asks you to consider the evidence for top-down feedback in the visual pathway, and evaluate what its role might be.

The rules for this assignment:

We will have a class discussion on top-down elements of visual perception on Thursday, April 30, 2009. To prepare for the discussion, you are asked to read the articles listed below and then to write a thoughtful paper based on all of the readings. You may discuss the readings and the general topic freely with your classmates, but you must write your paper alone. Papers will be graded on the basis of their insight, thoughtfulness, and clarity. Since one purpose of the paper is to prepare the class for an informed discussion, it is important that both you and your paper be present in class. If you are absent or your paper is late, the paper will receive only a neutral (Pass/Fail) grade.

The readings: (* indicates color figures are available in the online version of the article)

First group: visual perception.

(1) Tovee MJ (1994) How fast is the speed of thought? Current Biology 4(12): 1125-1127.* JOURNAL LOCATOR

The oldest paper in this group discusses the timing of neural responses in the visual pathway, but favors a bottom-up process.

(2) Hupé, JM et al. (1998) Cortical feedback improves discrimination between figure and background by V1, V2 andV3 neurons. Nature 394: 784-787.

Blocking feedback by cooling regions of extrastriate cortex leads to changes in the activity of neurons in lower visual areas. NOTE: "Salience" is defined in this paper as the contrast between a stimulus bar and its background. "High salience" means high contrast (eg, white bar, very dark background); "low salience" means low contrast (eg, light gray bar, medium gray background).

(3) Bullier J (2001) Integrated model of visual processing. Brain Research Reviews 36(2-3): 96-107.

This often-cited review suggests that some "higher" visual areas respond earlier than areas that are considered below them in the hierarchy, and thus that it is possible for information to flow top-down as well as bottom-up during visual perception.

(4) O'Shea J, Walsh V (2006) Cognitive neuroscience: trickle-down theories of vision. Current Biology 16(6): R206-9. * JOURNAL LOCATOR

A synopsis of experiments conducted on human subjects. The results are consistent with a role for the orbitofrontal cortex in perceiving objects, but the authors also suggest another interpretation.

(5a) Kveraga K et al. (2007) Magnocellular projections as the trigger of top-down facilitation in recognition. J. Neurosci. 27(48): 13232-13240 (excerpt).

The first page from a paper by the same group as the next reading, explaining their theory of how low spatial frequency information organizes visual perception.

(5b) Bar M, et al. (2006) Top-down facilitation of visual recognition. Proceedings of the National Academy of Sciences [PNAS] 103(2): 449-454. * JOURNAL LOCATOR

This is the paper referred to in the commentaries (4 and 5a) above. Magnetoencephalography and FMRI measure the timing and location of activity associated with images that are recognized and images that are not. The results suggest that low spatial frequency components of an image arrive quickly at the orbitofrontal cortex, where they help to organize the image that is perceived in the temporal cortex. In reading the paper, some aspects will be technically out of reach for us, but the argument and much of the data should be accessible. (For some background on fMRI, see the first optional reading.)

Second group: visual attention.

(6) Peelen MV and Mruczek REB (2008) Sources of spatial and feature-based attention in the human brain. J. Neurosci. 28(38): 9328-9329.

A brief essay discussing two forms of visual attention.

(7) Mangun GR and Fannon SP (2007) Attention: control in the visual cortex. Current Biology 17: R170-172.

A commentary summarizing the issues in the following reading.

(8) Bestmann S et al. (2007) Spatial attention changes excitability of human visual cortex to direct stimulation. Current Biology 17: 134-139. * JOURNAL LOCATOR

In a task where cortical trans-cranial magnetic stimulation (TMS) of human V1 evokes phosphenes (illusory glowing disks), spatial attention modulates the threshold of V1 units.
 

Optional supplementary material (online only):

(9) Arthurs OJ, Boniface S (2002) How well do we understand the neural origins of the fMRI BOLD signal? Trends in Neurosciences 25(1): 27-31. * JOURNAL LOCATOR

An explanation of the neural activity that is most likely to produce the signal that is recorded in functional imaging.

(10) Paradiso MA (2002) Perceptual and neuronal correspondence in primary visual cortex. Current Opinion in Neurobiology 12(2):155-61. JOURNAL LOCATOR

Some of the details in this thoughtful paper are beyond our scope, but you may find the big ideas helpful in thinking about perceptual processing.

Chapters 25 and 28 in your textbook will also provide helpful background.

 

The assignment:

Write a 4 to 5 page paper analyzing the role of top-down activity in the visual system. Base your paper on all of the readings listed above, although you do not have to give equal weight to all of them. As you read descriptions of experiments, examine the figures carefully as potential evidence for your paper, and also be prepared to explain them in our class discussion. It is important to understand (as much as possible) the experimental basis for the authors' conclusions, even if you do not mention every experiment in your own paper.

As you read the articles, keep two questions in mind. First, what is the big picture? What is top-down processing? What are the problems with strictly bottom-up processing that some top-down information might solve? More specifically, is there evidence that the firing of neurons in V1 and other early levels of the visual pathway is influenced by higher visual areas? Second, what is visual attention? What is the evidence that feedback in the visual pathway is for attention rather than for visual perception? Which is more likely to be the role for feedback? Can the two possibilities be reconciled?


In your paper, be sure to include ideas from ALL eight of the articles (some are very short). The two optional articles are available online if you need them, but they do not need to be part of your paper. Since you cannot incorporate every detail in your own paper, ask yourself what is best regarded as background for this assignment, and therefore can be omitted from your paper. You may be tempted to write about the articles by creating a sequential synopsis of each one, as if you were typing up your notes, but this is NOT a good strategy. These papers address a general question from different angles, and you should try to step back and create a general structure to address the issues in your own voice.

In writing your paper, be careful to translate ideas from the readings into your own words. It is unlikely that you will need to quote sentences from the articles. Cite the source of major ideas, but in a way that doesn't litter your paper with citations. In citing the articles, either use the (author, year) style [e.g. (Tovee, 1994; Bullier, 2001)], or use the style of Science and Annual Reviews in which numbers in parentheses [(1,2)] refer to items in the bibliography. Since we all have the numbered bibliography of readings, you do not need to include a bibliography in your paper. The page on Tips for Writing in Neuroscience will also be helpful to you.