** AhISee **   PSYCHOLOGY > Top-Down, Bottom-Up?  |<|Up|?|Menu|Search/Map|Translate|>| 

Top-Down or Bottom-Up? Can attention be understood at a cellular level?

What is attention? Can attention be understood at a cellular level? Do you care? You should. 'Attention' is a sort of code-word for consciousness. Biological psychology has made many strides towards an understanding of brain processes over recent years (well, it thinks it has). But has it got anywhere with understanding where consciousness (coded as 'attention') arises?

In this essay I argue that the answer to that question is 'no'.

This essay has been reformatted for on-line presentation.


Essay Outline




This essay reviews evidence for a cellular basis for attention. We will quickly explore some ideas on - and a theory of - attention, then look at recent evidence as to neurological factors. At the end of the essay, we will come back to what the question really implies. But first, what is attention? Every psychologist knows what it is, -don't they? Here are three definitions:

  1. 'A general term referring to the selective aspects of perception which function so that at any instant an organism focuses on certain features of the environment to the (relative) exclusion of other features'

    (Reber 1995: Penguin Dictionary of Psychology)

  2. 'Hypothetical process that either allows a selective awareness of a part or aspect of the sensory environment or allows selective responsiveness to one class of stimuli'

    (Kolb & Wishaw 1996)

  3. 'Attention is the experimental psychologist's code name for consciousness'

    (Allport 1980)

Perhaps it is not so easy to define after all. If we follow Allport (1980), then perhaps attention implies consciousness. Does it require self-awareness? What cognitive capacity is necessary? Perhaps an example would help; read the words in bold in the next sentence:

Whilst It was walking so down wet the this hill morning that when day, I my went happy out mind I was got far soaked away.

What exactly did the un-emboldened text say? Is cognition involved here, or is it 'the wiring'? In seeking a cellular basis for attention, we will have to leave Allport on one side for now, because it is difficult to infer conscious acts at a cellular level - though we will return to this point later. Perhaps it is more useful to ask, 'What might be the function of attention?' An inseparable attribute of attention is that of selectiveness; it implies discrimination - the selection of particular features (or perhaps combinations of features) of the environment to attend to, to monitor. We can infer that the capacity to selectively attend to specific features of the environment is adaptive for the organism, and that 'higher' organisms will have more selective attentional discrimination. Indeed, as the evidence is that 'higher' organisms perceive more of their environments (see, for example, Lettvin et al (1959) and Hubel & Wiesel 1968, 1974, 1977), a better attentional or selective mechanism is presumably necessary to enable the organism to extract significant information from its expanded sensory gamut.

So, in looking for a cellular basis for attention we are looking for mechanisms that may help us account for these phenomena of attention - selection, discrimination, monitoring. These specific characteristics are such that we will be looking for cells that are differentially triggered by the same stimulus dependant on the coexistence of some other stimulus at the same time; that is, whether the cell responds or not depends on the attentional level not the stimuli.


Theories of Attention

Treisman et al

In order to seek attentional mechanisms, it is necessary to have a working theory of what attention is. Treisman et al (1986) produced a fruitful (though disputed) model, the feature integration theory of visual attention: this is described in Table 1 and diagrammed in Figure 1   Open in popup window . One advantage of this model is that it mapped quite well onto what was then known of various features of the striate ('striped') occipital cortex; cells which behave as 'feature detectors' are found there in systematically incrementing rows and columns (originally described by Hubel & Wiesel in their (1977) 'ice cube' model). Indeed, groups of cells with such extreme particularization of feature extraction have been found as to give rise to such half-jocularities as 'grandmother cells' (Lettvin 1972)  [NOTE 1] . A map of the cortical pathways perhaps corresponding to Treisman's model is in Figure 2.   Open in popup window

No. Stage Name Characteristics


Pre-attentive Radical

or primitive features of the stimulus are detected in parallel by what might be called feature detectors



Features such as edges, orientation, color, saturation or brightness - amongst others - are then integrated by the attentional spotlight

Table 1: Treisman's (1985) Feature Recognition model


Top-Down? Bottom-up?

Another issue enters here; whether attention operates from the top-down or bottom-up (Table 2).

Name Characteristics

Endogenous: volitional control selects the attentional field Examples: the 'cocktail party' effect whereby we are capable of selecting a single voice from the whole enormous chatter


Exogenous: specific features of a stimulus 'pop up' into attention Examples: see Figure 3.   Open in popup window


Table 2 : Top-down -vs- Bottom-up: based on Theeuwes (1993)

An example of the bottom-up process is given in Figure 3   Open in popup window; the large red X is self-evident. In Treisman's model, there would be parallel detection of red objects, green objects and round objects, with a serial 'integration' picking out the big red X. In a review, Theeuwes (1993) commented that Treisman's pre-attentive stage has three qualities or properties: see Table 3.

It is unlimited in capacity. There was no significant change in the reaction time of subjects detecting targets distinguished by primitive features no matter how large the field

It is spatially parallel across the visual field - objects with a single primitive difference show the pop-out effect - see Figures 2 and 3.

It is bottom-up. Theeuwes found no evidence of top-down control.

Table 3: Theeuwes (1993) analysis of characteristics of the pre-attentive stage of attention

There is some evidence of top-down control in spatial terms; Eriksen & James (1986) found evidence of a capacity to selectively focus attention to small parts of the visual field. 



Various metaphors have been coined, the best known being Posner's (1980) 'spotlight of attention' metaphor, whereby attention is assumed capable of being focused - like a spotlight. A metaphor preferred by Eriksen & James (1986) is the 'zoom lens' -- in this idea, the 'beam' of attention can be widened or narrowed.

All of these descriptions primarily make use of visual metaphors, but it is worth remembering that these are metaphors. The human 'visual' cortex does not deal with pictures or images, it deals with action potentials. A 'spotlight' of attention is not likely to 'illuminate' an area of the cortex. Equally, the visual nature of these metaphors is too restrictive for a general model of attention. Is there aural attention? Certainly, there is - the 'cocktail party' effect is an exemplar of selective aural attention. Is there a 'spotlight' of aural attention? Or tactile attention?

Kolb & Wishaw (1996) comment that Treisman's model requires a "...sharp focus of attention..." and then argue that this "...seems unlikely..." but that it "...seems more reasonable..." that there is a gradual degradation of attention around the spot [NOTE 2] . This 'seeming' is however again based on a 'visual' idea of attention - the cocktail party selective attention effect is not a 'gradual' but a sharp focusing, at least in the view (and experience) of this author, as is the focusing of visual attention [NOTE 3] .


Neuropsychological evidence


Moran & Desimone (1985) used macaque monkeys to demonstrate the type of discrimination needed for an attentional mechanism; Figure 4   Open in popup window is a map of the macaque brain areas, whilst the experiment itself is detailed in Table 4. The experimenters monitored a neuron whilst the animal carried out a matching task. The important result was that the monitored cell, when presented with the same stimulus, only responded when the animal's attention was in the 'right' place. When the stimulus appeared elsewhere within the cell's receptive field the cell did not respond, thus meeting the top-down attentional criterion established above. Later, Spitzer, Desimone & Moran (1988) in a hard-and-easy delayed matching task found that orientation-discriminating cells in the visual area 4 (V4) of the monkey responded more strongly and selectively in the harder task, which they infer is due to the extra attention required.

Experimental Sequence

Animals were trained on a matching task

Animals were required to hold a bar while attending to a fixation point on a screen Stimuli were presented at various points in the receptive field (RF) of a particular neuron in the V4 area which was being monitored.

Case 1 

Two stimuli were given about 500ms apart. These could be identical or different; on an identical pair the animal was given a food reward on the immediate release of the bar.

Case 2

the animal was trained to ignore the stimulus when it was in a different part of the RF

Results The cell monitored responded differentially to the stimuli

One stimulus was a cross-hatched symbol, and this triggered the cell strongly. The other was an open symbol, to which the cell did not respond. In this way, the exact same stimulus could be required to be attended to or not, depending on its location. The interesting result was that the cell, when presented with the same (cross-hatched) stimulus, only responded when the stimulus was in the 'right' place . The animals' attention was focused at that point; when the stimulus (to which the cell responded normally) appeared elsewhere (but still within the cell's RF) the cell did not respond, thus meeting the attentional criterion established above


Table 4: modus operandi of Moran & Desimone's 1985 study

Does this work support the attentive stage in Treisman's (1986) theory? In a review by Maunsell & Ferrera (1995) Undergleider & Mishkin (1982) are cited as stating that the visual cortex appears to show two parallel streams of processing, these being the parietal and temporal pathways. They speculate that the temporal or ventral pathway might be concerned with deciding what objects are, whereas the parietal or dorsal pathway might be concerned in determining where objects are. However, Milner and Goodale (1993) recently proposed ventral as recognizing an object - what - and dorsal as how to. In 1993 Chelazzi et al investigated the pathways. They recorded from infero-temporal (IT) cells whilst the monkey altered fixations. Here there was evidence of the cell's response being suppressed by attentional factors (see Table 5).

Experimental Sequence


For 700 ms fix gaze on fixation point 


After 300 ms cue is given 


3000 ms delay 


Monkey chooses between 'Good' cue picture (for which the cell is normally responsive) and a 'Poor' non-target picture (for which it is not) (The monkey had to make an eye movement to the cue target).


At cue onset the cell's response to the 'Good' cue is strong, to 'Poor' cue is weak. If it is the 'Good' cue the cell continues to respond during the time delay But after the choice has been made, and for about 100ms before the eye moves: For 'Good' cue cell activity remains high: For 'Poor' cue cell activity is very low as though the RF was suppressed by the attentional factor

Table 5: Study by Chelazzi et al(1993)


Differential responses

Maunsell & Ferrera (1995), again working with the macaque and a stimulus-matching paradigm, identified a neuron in area V4 (halfway up the visual 'hierarchy') that did not have a feature-dependent response. They nonetheless managed to elicit a differential response from it (with differing stimuli) and suspect that they have found a memory effect; the animal needs to store a representation of the stimulus it is to match, but attentional factors were dismissed. Desimone and Duncan (1995) recently reviewed progress towards understanding neural mechanisms of attention. They emphasize the three phenomena that they regard as defining the problem of visual attention (see Table 6).

Factor Significance
the limited capacity of attention:

only a small part of available information can be processed

the selectivity of attention:

behavioural imperatives force choices

the integration of responses:

these being from multiple properties of the stimulus


Table 6: after Desimone and Duncan (1995)

They view attention not as a 'spotlight' in the Treisman model nor as a 'zoom lens' but as "an emergent property of many neural systems working to resolve competition for visual processing and control of behavior" [NOTE 4] . They go on to say: "At some point (or several points) between input and response, objects in the visual input compete for representation, analysis, or control. The competition is biased, however, towards information that is currently relevant to behavior. Attended stimuli make demands on processing capacity, while unattended ones often do not." (Desimone and Duncan 1995)

Recent work by Colby et al (1996) has concerned the lateral intraparietal area or LIP. It is neurally connected both with the supposed pattern recognition system in the inferior temporal cortex and also with the superior colliculus, which is involved in saccades of the eyes. According to Corby et al , most neurons in LIP respond to visual stimuli. They noted that traditionally a flashing light was used as a stimulus in studies of the LIP responses, but that Jonides & Yantis (1988) had pointed out that a flashing light has (fairly obviously) an attentional component.


Gating of cells?

Colby et al therefore carried out an experiment to distinguish between the attentional and visual components of the response. They used a stable array task, whereby the monkey was trained to make a visual saccade that brought the fixed stimulus into the RF of the monitored neuron. In this way, the stimulus did not stimulate the cell by its appearance, but by its entry in the RF. The response in these circumstances was significantly less than if the stimulus flashed on in the RF. By manipulation of the conditions, though, stable stimuli could be made to trigger a strong response. This could be done either by novelty, when the animal was trained to make a saccade immediately on perceiving the stimulus to bring the stimulus into the RF of the cell, or by training the animal to 'anticipate' the stimulus and then saccade to it. In this way the criteria established earlier of a differential response of a particular cell to the same stimulus are fulfilled. Colby et al hold that these results show that the LIP neurons are gated by attention; the author has however only been able to examine an abstract of this rather complex work.



We have seen how certain cells in area V4 of the temporal lobe respond differentially, and at a significant level, to the same stimulus depending on other criteria; that being another stimulus. Other V4 cells may remember criteria for comparisons. We have also seen that there is evidence that many (perhaps most) parietal cells are gated by attentional factors. Does this give us an understanding of attention at a cellular level?

This author's response is, ' Yes - and No '. Yes, various discriminatory responses have been identified by ingenious work. Yes, the studies show that (or taking into account the minimal statistical evidence cited, suggest that) cells exhibit the type of differential response under the same stimulus that is a prerequisite for there to be a cellular basis for attention. The bottom-up process it would seem presents no great problem. All this work is painstaking and clever, but we do already know that humans and other animals can 'pay attention' - thus, the existence of a mechanism to support this is not exactly surprising.

Are we looking at this the right way round? Clearly, a principal function of the sensory systems is to filter and reduce the amount of information picked up by the organism; see for example Ornstein (1983) or Desimone & Duncan (1995). This quantity is potentially so huge, with such a number of possible stimuli, that perhaps the interesting question is not, 'how is attention elicited at a cellular level' but 'how is the normal suppression of the eliciting of attention overridden?'

The fascination of cognitive neuroscientists with computer models strikes the student from all sides. 'Top-down' and 'bottom-up' started out as descriptions of two ways of writing computer programs. 'Feature recognition' with 'stroke recognition' is exactly how text- and handwriting-recognition programs are implemented. This is all fine and useful as long as it is remembered to be metaphorical and is not reified into a reality. Computers may well work a little like brains; it is after all our brains that design them, based on how we conceive problem-solving [NOTE 5] . Going the other way, to assume that brains are like computers is perhaps a little simple-minded; we have so often been wrong about how biological systems work. Is it time for the concept of GIGO [NOTE 6]  to be revived? Evolutionary competitive models will, no doubt, be the next great fashion.

Overall, there is still a great difficulty here. In the monkey who is subjected to this process, it is its volitional or endogenous attention that is being manipulated (by a food reward which it desires) in such a way that it attends - or does not attend - to a stimulus. If a monkey is capable of 'choosing', then it chooses to pay attention. In a human, the top-down process of cutting out the noise in the background to attend to one particular person (say) is similarly volitional, and here we have no problem with the word 'choose'.

But this volitional act is affecting a cellular mechanism - that is, there is still an inherent dualism in this type of approach to the problem. The ghost is still in the machine, until a cellular mechanism for consciousness is demonstrated. The conviction that 'consciousness' is inherent in the actions of neurones is not equivalent to a proof [NOTE 7] . So, finally, we arrive at 'No'. At the moment we can only say that a cellular result of attention is perhaps demonstrable - not that the cellular basis for attention has been demonstrated.



Allport, D.A. (1980) Patterns and actions: Cognitive mechanisms are content specific. (1996) IN G. Claxton (Ed.), Cognitive psychology: new directions . London: Routledge, Kegan Paul.
Chelazzi, L., Miller, E.K., Duncan, J., & Desimone, R. (1993). A neural basis for visual search in inferior temporal cortex. Nature, 363 , 345-347.
Colby, C., Duhamel J-R.& Goldberg, M. (1996) Visual, presaccadic and cognitive activation of single neurons in monkey lateral parietal area. Journal of Neurophysiology, 76 . 2841-52
Desimone, R. & Duncan, J. (1995) Neural mechanisms of selective visual attention. Annual Review of Neuroscience, 18 . 193-222.
Eriksen, C. & James, J. (1986). `Visual attention within and around the field of focal attention: A zoom lens model', Perception and Psychophysics 40, 225-240.
Gazzaniga, M. (Ed.) (1995) The Cognitive Neurosciences . Bradford: London.
Hubel, D., & Wiesel, T (1974) Sequence regularity and geometry of orientation columns in the monkey striate cortex. Journal of Comparitive Neurology 158 . 267-93
Hubel, D., & Wiesel, T (1977) Functional architecture of macaque monkey visual cortex . Proceedings of the Royal society of London 198 . 1-59.
Hubel, D., & Wiesel, T. (1968). Receptive fields and functional architecture of monkey striate cortex. Journal of Physiology (London) 195 . 215-43.
Jonides, J., & Yantis, S. (1988). Uniqueness of abrupt visual onset in capturing attention. Perception and Psychophysics, 43 , 346-354.
Kolb, B. & Wishaw, I. (1996) Fundamentals of Human Neuropsychology . 4th Edition. New York: Freeman & Co.
Lettvin, J.,Maturana, H., McCulloch, W., & Pitts, W. (1959) What the frog's eye tells the frog's brain. Proceedings of the Instyiute of radio Engineers 47 . 1940-51.
Maunsell, J. & Ferrera, V. (1995) Attentional Mechanisms in Visual Cortex IN M. S. Gazzaniga (Ed.) The Cognitive Neurosciences . Bradford: London. 451-61
Milner, A., & Goodale, M (1995) The visual brain in action . Oxford: Oxford University Press
Moran, J. & Desimone, R. (1994) Selective attention gates visual processing in the extrastriate cortex. Science 229. 782-4.
Ornstein, R. (1983) The Psychology of Consciousness . London: Penguin.
Posner M. (1980) Orienting of attention. Quarterly Journal of Experimental Psychology 32 :1:3-25
Reber, A. (Ed.) (1995) The Penguin Dictionary of Psychology. Second Edition. London: Penguin
Spitzer, H., Desimone, R.J. and Moran, J. Increased Attention Enhances Both Behavioral and Neuronal Performance. Science, Vol. 40 338-340, 1988.
Theeuwes, J. (1993). `Visual selective attention: A theoretical analysis', Acta Psychologica 83, 93-154.
Treisman, A. (1986) Features and objects in visual processing. Scientific American, 254 . 114-124
Ungerleider, L. & Mishkin, M. (1982) Two cortical visual systems. IN D. Ingle, R. Mansfield & M. Goodale (Eds.) The analysis of visual behaviour. Massachusetts: MIT Press. 549-586.


 [Note 1] Harris (1980) mocked the idea of extreme particularity in feature extraction with the proposal of 'yellow Volkswagen cells' - the point being the idea can be carried too far.

 [Note 2] Is it surprising to see such imprecise and unsupported language in a neuropsychology textbook? It does demonstrate honesty, and how little definite knowledge there is in this area.

 [Note 3] The focussing of attention to discriminate a particular musical instrument from ensemble is (experientially, speaking as an (ex?) musician and producer) sharp and not at all fuzzy.

 [Note 4] We can note that evolutionary competition has, rather ß la mode and free-market driven, arrived at last in the nervous system. Gosh.

 [Note 5] Please note that the author is not in any way opposed to making computer models, having spent years making computers do more-or-less what was required. However, the map is not the territory. We know how computer models work, because we design them - we don't know if the brain does things in the same way as we conceive it doing them.

 [Note 6] Garbage In - Garbage Out.

 [Note 7] Gazzaniga (1995) says in Churchillian mode in his preface: "...The future of the field is working toward a science that truly relates brain and cognition in a mechanistic way. That task is not easy...Yet that is the objective" (my emphasis). Why? Why is 'that' the objective? Is this science or politics? Is the aim to find out the 'truth' - or to prove a presupposition?

|<|Up|?|^|Menu|Search/Map|Translate|> GOTO 


ahisee?? AhISee!

Valid HTML 4.01!

Entire Site Copyright ę 2001 J. Roberts and AhISee.com - All Rights Reserved.
Problems, questions or comments about the content? Go to the contacts page
Problems, questions or comments about this web site? Send mail to our webmaster