Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Review
. 2019 Mar 25;92(1):127-137.
eCollection 2019 Mar.

Attention: The Messy Reality

John K Tsotsos  1
Affiliations
Review

Attention: The Messy Reality

John K Tsotsos. Yale J Biol Med. .

Abstract

The human capability to attend has been both considered as easy and as impossible to understand by philosophers and scientists through the centuries. Much has been written by brain, cognitive, and philosophical scientists trying to explain attention as it applies to sensory and reasoning processes, let alone consciousness. It has been only in the last few decades that computational scientists have entered the picture adding a new language with which to express attentional behavior and function. This new perspective has produced some progress to the centuries-old goal, but there is still far to go. Although a central belief in many scientific disciplines has been to seek a unifying explanatory principle for natural observations, it may be that we need to put this aside as it applies to attention and accept the fact that attention is really an integrated set of mechanisms, too messy to cleanly and parsimoniously express with a single principle. These mechanisms are claimed to be critical to enable functional generalization of brain processes and thus an integrative perspective is important. Here we present first steps towards a theoretical and algorithmic view on how the many different attentional mechanisms may be deployed, coordinated, synchronized, and effectively utilized. A hierarchy of dynamically defined closed-loop control processes is proposed, each with its own optimization objective, which is extensible to multiple layers. Although mostly speculative, simulation and experimental work support important components.

Keywords: Selective Tuning; attention; executive control; vision.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Classic control models. a) Open-loop control includes no method for monitoring or altering the system’s output. b) The standard closed-loop control model where the controller provides signals to the system that will bring it closer to a reference state. The system’s state is sensed continually and compared to the reference in order to determine these adjustments.
Figure 2
Figure 2
The breadth of problems inherent in pyramid representations. a) The Context Problem. A stimulus (black dot) within the receptive field of a top layer neuron, showing its spatial context defined by that receptive field. b). The Cross-Talk Problem. Two input stimuli activate feed-forward projections that overlap, showing the regions of overlap containing neurons that are affected by both. Those might exhibit unexpected responses with respect to their tuning profiles. c) The Routing Problem. Interacting top-down and bottom-up spatial search constraints are shown with the areas of overlap representing the viable search regions for best neural pathway. d) The Boundary Problem. The two units depicted in the second layer from the bottom illustrate how the extent of the magenta unit’s receptive field is entirely within the input layer while only half of the receptive field of the green unit is within the input layer. The bottom layer represents the input and higher layers the subsequent process representation. The boundary problem forces more and more of the periphery to not have a veridical representation in higher layers of the pyramid. (Adapted from [1]).
Figure 3
Figure 3
The closed-loop control of STAR-AX. The intent is the same as the standard closed-loop model except now there are several closed-loop controllers (horizontal loops) all tied to a higher level controller (vertical loop); it is hierarchical (with perhaps yet a higher level controller at the task level driven by a CP [14]). The color coding of the elements has similar meaning to that of Figure 1b. The reference can be dynamically set depending on task goals shown as an input to the central control loop (corresponding to the reference variable of Figure 1b, but now being dynamically set). The “orange” elements are the processes that are controlled. The “green” elements denote the measured or observed values for each of the control variables. The “blue” element corresponds to the computation of the sum of deviations from the goals; the system priorities are an input to this. The “red” element determines the values process control γ(i, t) based on system goals.
Figure 4
Figure 4
The cyclic control signals required for each of a number of different kinds of visual tasks. A. The set of tasks considered appear at the top, with the temporal extent of each shown in a different color. B. The set of control signals shown by the blues line to the right of the labels. These blue lines are intended to represent the timing of when the controller must generate an instruction appropriate to the indicated label. For example, in order to select the cFOA (central focus of attention), two actions are required at the end of the first feedforward pass through the visual hierarchy, namely, a selection of the strongest response followed by a matching function to compare what that strongest response represents to the goals of the task. Note how some signals (for example, “null blackboard”) have multiple and different colored pulses, one pulse for each layer of the visual hierarchy. C. ST has different temporally ordered stages of visual processing [33], beginning with top-down priming due to task instructions or knowledge and context, feedforward processing of visual signal, decision-making to confirm if a task is complete, recurrent processes to reduce network signal interference and permit identification of features or location, search of an image that would entail cycling through various parameterization of the previously listed tasks, and so on. Several experimental results support this conceptualization [1,34]. D. These oval circuits are intended to show the different kinds of attention cycles that this sequence of processing stages is capable of representing. Namely, a single attentive processing cycle may or may not include a priming stage, it would end at 150ms if the task was a simple categorization (no attention is needed), or it may require precise localization, etc. This degree of flexibility is a unique feature of STAR.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Tsotsos JK. A Computational Perspective on Visual Attention. Cambridge: MIT Press; 2011.
    1. Tsotsos JK. Complexity Level Analysis Revisited: What Can 30 Years of Hindsight Tell Us about How the Brain Might Represent Visual Information? Front Psychol. 2017;8:1216. - PMC - PubMed
    1. Kahneman D. Remarks on attention control. Acta Psychol (Amst). 1970;33:118–31.
    1. Allport A. Attention and control: Have we been asking the wrong questions? A critical review of twenty-five years. Attention and performance XIV: Synergies in experimental psychology, artificial intelligence, and cognitive neuroscience 1993:183-218.
    1. Egeth HE, Yantis S. Visual attention: control, representation, and time course. Annu Rev Psychol. 1997;48:269–97. - PubMed

Publication types

LinkOut - more resources