Metaphors are important. Once chosen, they determine what makes sense and what does not. If you say that travelling by train is like travelling by plane, a world of advanced bookings, seat reservations, and dynamic pricing opens up. If, on the other hand, you say that travelling by train is like taking the underground, things become simply a question of showing up, buying a ticket, and taking the next train that comes along. So, metaphors give rise to completely different and incompatible conceptions of the same phenomenon.
The field has cast attention as a gate. Consequently, it has become easy to imagine that attention acts as a bottleneck: some things pass through while most others do not. The gate has a certain capacity, and it is important to establish how large this capacity is and what factors influence its size. Observations that do not conform to the gate metaphor create problems and lead to complications and ad hoc solutions that have been pointed out in the target article.
The gate metaphor is an answer to the undeniable fact that there are limits to the bandwidth with which information can be transmitted in the visual system. But what if a gate is not the best metaphor for attention? What if rather than acting as a gate, attention acts as a modulator for the various parts of the visual field? Like an equalizer amplifying parts that are of interest and attenuating parts that are not. This also controls bandwidth, but in a much more gradual fashion.
In principle, the visual system covers the entire visual field (hence the latter’s name). By definition, that has to mean that information from the entire visual field enters the visual system. What differs, and where attention can play a certain role is how much information about a particular location enters the visual system.
Close to fixation receptive fields are small and a lot of detail is available. But the further into the periphery, the larger the receptive fields become, and the less detail can be extracted.
Saccades move the area with a lot of detail across a visual scene. Covert attention may modulate the gain for certain parts of the visual field, but only within the limits set by the physiology of the retina. Moreover, nothing ever gets completely excluded.
Together with Chris Olivers, I actually used the author’s idea about the computations of summary statistics in our model of visual search (Hulleman & Olivers, Reference Hulleman and Olivers2017). In easy search, summary statistics computed over a large area can yield enough information to determine whether a target is present. Hardly any saccades are needed, and reaction times will be fast. In difficult search, only the summary statistics computed over a small area can be used for target detection. This means a cumbersome and error prone search process where many saccades are needed to cover the search display widely enough to ensure that the target can be found. (Please note how this formulation of the visual search process, one of the main culprits in the target article, gets by without the use of the word attention.)
The target paper now dismisses this approach because it “essentially treats attention as spatial selection, gating the information available”. In my view, this is a consequence of the target paper getting trapped in its own gating metaphor. The gating metaphor encourages thinking in all-or-nothing outcomes. Information either comes through the gate, or it does not. This leads to theoretical problems when information that was supposed to have been left outside the gate turns out to influence behaviour after all. Thinking about attention as a modulator where all available information enters the visual system, but at varying spatial resolutions (as a consequence of the physiology of the retina and the limited actions of attention) allows a new and probably more appropriate perspective.
Especially when considered in combination with another valuable point made in the target article: What are we really asking our participants to do? Is it something that needs the high resolution of the fovea or can it be done with the lower resolution representations?
Here again, the gate metaphor of attention has a pernicious influence. If only a very small subset of the available information enters the visual system, it is important that its quality is as high as possible. For visual search, this leads to conceptualisations of the search process where high resolution representations of individual items are compared to a target template. But a lot of the time, the participant is merely asked to indicate whether a particular search display contains a target. When the target is different enough from the distractors, this is probably something that can be achieved based on low-resolution information from across the visual field.
When thinking about attention in terms of modulation of the entire visual field, the use of information from across the search display seems like a natural strategy. However, within the gate metaphor, this strategy is not even a possibility, since the low-resolution information from the periphery will never make it through the gate in the first place.
In conclusion, I would argue that the first step towards solving the problems described in the target article is to become aware of the power of metaphors. How picking any particular metaphor determines research questions and imposes implicit assumptions (e.g. if you say that attention is a spotlight, then one implicit assumption is that it can be actively moved around because this is how spotlights work). The next (and much more difficult) step is to choose a better metaphor for the fact that there is a limit to the rate with which information can be transmitted in the visual system. Here, I proposed an equalizer, but I make no claim that it is the best.
What I would say though is that, with the equalizer metaphor in mind, the use of summary statistics is a much bigger step towards the solution of the problem of attention than the author allows.
You have
Access
Metaphors are important. Once chosen, they determine what makes sense and what does not. If you say that travelling by train is like travelling by plane, a world of advanced bookings, seat reservations, and dynamic pricing opens up. If, on the other hand, you say that travelling by train is like taking the underground, things become simply a question of showing up, buying a ticket, and taking the next train that comes along. So, metaphors give rise to completely different and incompatible conceptions of the same phenomenon.
The field has cast attention as a gate. Consequently, it has become easy to imagine that attention acts as a bottleneck: some things pass through while most others do not. The gate has a certain capacity, and it is important to establish how large this capacity is and what factors influence its size. Observations that do not conform to the gate metaphor create problems and lead to complications and ad hoc solutions that have been pointed out in the target article.
The gate metaphor is an answer to the undeniable fact that there are limits to the bandwidth with which information can be transmitted in the visual system. But what if a gate is not the best metaphor for attention? What if rather than acting as a gate, attention acts as a modulator for the various parts of the visual field? Like an equalizer amplifying parts that are of interest and attenuating parts that are not. This also controls bandwidth, but in a much more gradual fashion.
In principle, the visual system covers the entire visual field (hence the latter’s name). By definition, that has to mean that information from the entire visual field enters the visual system. What differs, and where attention can play a certain role is how much information about a particular location enters the visual system.
Close to fixation receptive fields are small and a lot of detail is available. But the further into the periphery, the larger the receptive fields become, and the less detail can be extracted.
Saccades move the area with a lot of detail across a visual scene. Covert attention may modulate the gain for certain parts of the visual field, but only within the limits set by the physiology of the retina. Moreover, nothing ever gets completely excluded.
Together with Chris Olivers, I actually used the author’s idea about the computations of summary statistics in our model of visual search (Hulleman & Olivers, Reference Hulleman and Olivers2017). In easy search, summary statistics computed over a large area can yield enough information to determine whether a target is present. Hardly any saccades are needed, and reaction times will be fast. In difficult search, only the summary statistics computed over a small area can be used for target detection. This means a cumbersome and error prone search process where many saccades are needed to cover the search display widely enough to ensure that the target can be found. (Please note how this formulation of the visual search process, one of the main culprits in the target article, gets by without the use of the word attention.)
The target paper now dismisses this approach because it “essentially treats attention as spatial selection, gating the information available”. In my view, this is a consequence of the target paper getting trapped in its own gating metaphor. The gating metaphor encourages thinking in all-or-nothing outcomes. Information either comes through the gate, or it does not. This leads to theoretical problems when information that was supposed to have been left outside the gate turns out to influence behaviour after all. Thinking about attention as a modulator where all available information enters the visual system, but at varying spatial resolutions (as a consequence of the physiology of the retina and the limited actions of attention) allows a new and probably more appropriate perspective.
Especially when considered in combination with another valuable point made in the target article: What are we really asking our participants to do? Is it something that needs the high resolution of the fovea or can it be done with the lower resolution representations?
Here again, the gate metaphor of attention has a pernicious influence. If only a very small subset of the available information enters the visual system, it is important that its quality is as high as possible. For visual search, this leads to conceptualisations of the search process where high resolution representations of individual items are compared to a target template. But a lot of the time, the participant is merely asked to indicate whether a particular search display contains a target. When the target is different enough from the distractors, this is probably something that can be achieved based on low-resolution information from across the visual field.
When thinking about attention in terms of modulation of the entire visual field, the use of information from across the search display seems like a natural strategy. However, within the gate metaphor, this strategy is not even a possibility, since the low-resolution information from the periphery will never make it through the gate in the first place.
In conclusion, I would argue that the first step towards solving the problems described in the target article is to become aware of the power of metaphors. How picking any particular metaphor determines research questions and imposes implicit assumptions (e.g. if you say that attention is a spotlight, then one implicit assumption is that it can be actively moved around because this is how spotlights work). The next (and much more difficult) step is to choose a better metaphor for the fact that there is a limit to the rate with which information can be transmitted in the visual system. Here, I proposed an equalizer, but I make no claim that it is the best.
What I would say though is that, with the equalizer metaphor in mind, the use of summary statistics is a much bigger step towards the solution of the problem of attention than the author allows.
Financial support
The author was supported by UKRI grant ES/X000443/1.
Competing interests
None.