Skip to main content
Social Sci LibreTexts

5.7: Horizontal Layers of Control

  • Page ID
    21234
  • \( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \) \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)\(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\) \(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\)\(\newcommand{\AA}{\unicode[.8,0]{x212B}}\)

    Classical cognitive science usually assumes that the primary purpose of cognition is planning (Anderson, 1983; Newell, 1990); this planning is used to mediate perception and action. As a result, classical theories take the form of the sense-think-act cycle (Pfeifer & Scheier, 1999). Furthermore, the “thinking” component of this cycle is emphasized far more than either the “sensing” or the “acting.” “One problem with psychology’s attempt at cognitive theory has been our persistence in thinking about cognition without bringing in perceptual and motor processes” (Newell, 1990, p. 15).

    Embodied cognitive science (Agre, 1997; Brooks, 1999, 2002; Chemero, 2009; Clancey, 1997; Clark, 1997, 2003, 2008; Pfeifer & Scheier, 1999; Robbins & Aydede, 2009; Shapiro, 2011; Varela, Thompson, & Rosch, 1991) recognizes the importance of sensing and acting, and reacts against central cognitive control. Its more radical proponents strive to completely replace the sense-think-act cycle with sense-act mechanisms.

    This reaction is consistent with several themes in the current chapter: the importance of the environment, degrees of embodiment, feedback between the world and the agent, and the integral relationship between an agent’s body and its umwelt. Given these themes, it becomes quite plausible to reject the proposal that cognition is used to plan, and to posit instead that the purpose of cognition is to guide action:

    The brain should not be seen as primarily a locus of inner descriptions of external states of affairs; rather, it should be seen as a locus of internal structures that act as operators upon the world via their role in determining actions. (Clark, 1997, 47)

    Importantly, these structures do not stand between sensing and acting, but instead provide direct links between them.

    The action-based reaction against classical cognitivism is typified by pioneering work in behavior-based robotics (Brooks, 1989, 1991, 1999, 2002; Brooks & Flynn, 1989). Roboticist Rodney Brooks construes the classical sandwich as a set of vertical processing layers that separate perception and action. His alternative is a hierarchical arrangement of horizontal processing layers that directly connect perception and action.

    Brooks’ action-based approach to behavior is called the subsumption architecture (Brooks, 1999). The subsumption architecture is a set of modules. However, these modules are somewhat different in nature than those that were discussed in Chapter 3 (see also Fodor, 1983). This is because each module in the subsumption architecture can be described as a sense-act mechanism. That is, every module can have access to sensed information, as well as to actuators. This means that modules in the subsumption architecture do not separate perception from action. Instead, each module is used to control some action on the basis of sensed information.

    The subsumption architecture arranges modules hierarchically. Lower-level modules provide basic, general-purpose, sense-act functions. Higher-level modules provide more complex and more specific sense-act functions that can exploit the operations of lower-level operations. For instance, in an autonomous robot the lowest-level module might simply activate motors to move a robot forward (e.g., Dawson, Dupuis. & Wilson, 2010, Chapter 7). The next level might activate a steering mechanism. This second level causes the robot to wander by taking advantage of the movement provided by the lower level. If the lower level were not operating, then wandering would not occur: because although the steering mechanism was operating, the vehicle would not be moving forward.

    Vertical sense-act modules, which are the foundation of the subsumption architecture, also appear to exist in the human brain (Goodale, 1988, 1990, 1995; Goodale & Humphrey, 1998; Goodale, Milner, Jakobson, & Carey, 1991; Jakobson et al., 1991).

    There is a long-established view that two distinct physiological pathways exist in the human visual system (Livingstone & Hubel, 1988; Maunsell & Newsome, 1987; Ungerleider & Mishkin, 1982): one, the ventral stream, for processing the appearance of objects; the other, the dorsal stream, for processing their locations. In short, in object perception the ventral stream delivers the “what,” while the dorsal stream delivers the “where.” This view is supported by double dissociation evidence observed in clinical patients: brain injuries can cause severe problems in seeing motion but leave form perception unaffected, or vice versa (Botez, 1975; Hess, Baker, & Zihl, 1989; Zihl, von Cramon, & Mai, 1983).

    There has been a more recent reconceptualization of this classic distinction: the duplex approach to vision (Goodale & Humphrey, 1998), which maintains the physiological distinction between the ventral and dorsal streams but reinterprets their functions. In the duplex theory, the ventral stream creates perceptual representations, while the dorsal stream mediates the visual control of action.

    The functional distinction is not between ‘what’ and ‘where,’ but between the way in which the visual information about a broad range of object parameters are transformed either for perceptual purposes or for the control of goal-directed actions. (Goodale & Humphrey, 1998, p. 187)

    The duplex theory can be seen as representational theory that is elaborated in such a way that fundamental characteristics of the subsumption architecture are present. These results can be used to argue that the human brain is not completely structured as a “classical sandwich.” On the one hand, in the duplex theory the purpose of the ventral stream is to create a representation of the perceived world (Goodale & Humphrey, 1998). On the other hand, in the duplex theory the purpose of the dorsal stream is the control of action, because it functions to convert visual information directly into motor commands. In the duplex theory, the ventral stream is strikingly similar to the vertical layers of the subsumption architecture.

    Double dissociation evidence from cognitive neuroscience has been used to support the duplex theory. The study of one brain-injured subject (Goodale et al., 1991) revealed normal basic sensation. However, the patient could not describe the orientation or shape of any visual contour, no matter what visual information was used to create it. While this information could not be consciously reported, it was available, and could control actions. The patient could grasp objects, or insert objects through oriented slots, in a fashion indistinguishable from control subjects, even to the fine details that are observed when such actions are initiated and then carried out. This pattern of evidence suggests that the patient’s ventral stream was damaged, but that the dorsal stream was unaffected and controlled visual actions. “At some level in normal brains the visual processing underlying ‘conscious’ perceptual judgments must operate separately from that underlying the ‘automatic’ visuomotor guidance of skilled actions of the hand and limb” (p. 155).

    Other kinds of brain injuries produce a very different pattern of abnormalities, establishing the double dissociation that supports the duplex theory. For instance, damage to the posterior parietal cortex—part of the dorsal stream—can cause optic ataxia, in which visual information cannot be used to control actions towards objects presented in the part of the visual field affected by the brain injury (Jakobson et al., 1991). Optic ataxia, however, does not impair the ability to perceive the orientation and shapes of visual contours.

    Healthy subjects can also provide support for the duplex theory. For instance, in one study subjects reached toward an object whose position changed during a saccadic eye movement (Pelisson et al., 1986). As a result, subjects were not conscious of the target’s change in location. Nevertheless, they compensated to the object’s new position when they reached towards it. “No perceptual change occurred, while the hand pointing response was shifted systematically, showing that different mechanisms were involved in visual perception and in the control of the motor response” (p. 309). This supports the existence of “horizontal” sense-act modules in the human brain.


    This page titled 5.7: Horizontal Layers of Control is shared under a CC BY-NC-ND license and was authored, remixed, and/or curated by Michael R. W. Dawson (Athabasca University Press) .

    • Was this article helpful?