Object Recognition: Information Encapsulation & Evolution
18 April 2021
As a species, humans have developed many cognitive processes as a part of various mental systems, such as the visual system, that increase our chances of survival. In order to explain how human mental systems work, philosopher Jerry Fodor proposed the Fordorian model which states that mental modules are informationally encapsulated. The term informational encapsulation, or cognitive impenetrability, describes the condition that specific module systems cannot draw on the full range of beliefs, desires, expectations, or knowledge of an individual (Cartson, 1996). Fodor states that visual modules for object recognition are not able to access knowledge from the rest of the brain and are therefore encapsulated. However, there are also other arguments that disprove Fodor’s information encapsulation theory, specifically regarding object recognition. Despite the contrasting viewpoints, both encapsulated and unencapsulated processes are beneficial when perceiving objects. In this paper I will attest that both informational encapsulated and informational unencapsulated visual processes in object recognition are evolutionarily crucial for the survival of a species.
Fodor’s modularity thesis can be proven through the perception of visual illusions. An example of an illusion that validates Fodor’s visual module is the Müller-Lyer Illusion shown in Figure 1.
Figure 1. Adapted from Palmer, 1999. The Müller-Lyer Illusion.
In the figure, we can see that the straight line on the top seems shorter than the one on the bottom. In reality, the lines are the same length. However, even with knowing this fact, we still “see” the top line as shorter than the bottom line. Additional evidence to prove Fodor’s claim is the brain’s automatic ability to interpret two-dimensional drawings as three-dimensional objects. An example of an object that we perceive as such is shown in Figure 2.
Figure 2. Adapter from Palmer, 1999. A self-occluding cube.
The cube shown in the figure is an example of a self-occluded surface, where surfaces of an object are hidden from view by its own surfaces (Palmer, 1999). Although the sides that make it three-dimensional are hidden from view, we somehow “see” the hidden sides and interpret it as a three-dimensional cube. Even if you try to tell yourself that you are looking at a two-dimensional figure, your brain still perceives the drawing as three dimensional. These examples show that we cannot draw on our full range of knowledge when looking at objects, which defends Fodor’s stance that the visual mental module is informationally encapsulated.
Although the visual illusions above provide strong support for Fodor’s information encapsulated modules, there are also instances where the brain uses prior knowledge or expectation when identifying objects. Examples of these instances are ambiguous figures and the visual completion of objects.
Figure 3. Adapted from Palmer, 1999. An ambiguous figure of a vase or two faces.
An ambiguous figure is shown in Figure 3 with the full ambiguous figure on the left and the two possible objects you can see on the right. At first, people either see a vase or two faces. However, once someone else tells you to see the other object, your brain can identify it as well. After doing this, the image will become dynamic in that the two image possibilities alternate back and forth. Ambiguous figures provide evidence that visual object recognition relies on top-down mechanisms (Palmer, 1999). Top-down mechanisms or top-down processing is when your brain uses what it knows or outside information to anticipate what it sees. The low-level processing of identifying object features is affected by high-level processing of thoughts. This is a direct contrast to Fodor’s model, where the visual system cannot rely on outside information. Another tendency of our brains that undermines Fodor’s statement is the visual completion of objects. An example of this is shown in Figure 4.
Figure 4. Adapted from Palmer, 1999. Visual completion of shapes.
In the figure, we see a rectangle overlapped by a circle that is overlapped by a square. Although we do not see the whole rectangle or circle, our visual system “completes” the object due to prior knowledge of what a circle and rectangle are supposed to look like. This example illustrates that our brains actively construct a model of our environment due to our beliefs, desires, and expectations (Palmer, 1999). Ambiguous figures and the visual completion of objects provide evidence that object recognition is connected to outside informational processing.
While the two arguments presented above disagree with each other, it cannot be denied that both information encapsulated and information unencapsulated visual modules carry evolutionary significance. Visual illusions that supported Fodor’s encapsulated visual module, such as the illusion shown in Figure 1, imply that sometimes we do not have a very accurate perception of our environment. In other words, we do not have a veridical perception of the world. The objects we see are sometimes not how they are in real life. However, despite this, we can mostly rely on our visual system to identify objects accurately as optical illusions are not common in the environment. This ability to identify objects without having to rely on outside information is crucial for the survival of a species. For example, if we see a tiger at a grocery store, our visual system immediately identifies it as a tiger without consulting our expectations of there being a tiger in a grocery store. Recognizing objects immediately helps us avoid signs of danger and therefore increases our chances of survival. Additionally, the automatic process of interpreting a two-dimensional object as a three-dimensional object is valuable due to the fact that we live in a three-dimensional world. It is more useful to see the two-dimensional drawing in Figure 2 as a three-dimensional cube because there are no two-dimensional cubes in the real world. Perceiving two-dimensional objects as three-dimensional is a more accurate reflection of our environment and provides more comprehensive information to the brain (Palmer, 1999). Information unencapsulated processes are also crucial in evolution. As shown in Figure 3, our brains can easily adjust to the objects we see given additional outside information. This is particularly beneficial in the sense that we can quickly adapt to see other possible perceptions of objects, which results in a better understanding of the objects. Figure 4 illustrates how humans construct a model of their environment given limited visual stimuli. Constructing a model of the environment from our knowledge, beliefs, and expectations is evolutionarily crucial as it allows us to predict the future (Palmer, 1999). Due to prior knowledge and experience, we can often predict the amount of time and where a moving object will go next based on its speed. This is especially important when an object is coming towards us since we are able to know its trajectory and decide whether to avoid or ignore it. Furthermore, by knowing what objects are supposed to look like, we can avoid unexpected situations that arise because of an abnormality. These types of skills that aid in the awareness of our environment and our ability to make intelligent decisions are a result of both an encapsulated and an unencapsulated visual system.
In conclusion, even though there are disagreements regarding the informational encapsulation of object recognition, both encapsulated and encapsulated processes are crucial when looking through an evolutionary lens. While it is beneficial to understand how the visual system works as it relates to the degree of information encapsulation, it is more important to recognize what the visual system allows us to do and how the visual system’s capabilities are beneficial to a species. It is clear that without either of encapsulated or unencapsulated processes, perceiving organisms would be more challenged in survival.
References:
Carston, R. (1996). The Architecture of the Mind: Modularity and Modularization. Cognitive Science: An Introduction. Blackwell Publ.
Palmer, S. (1999), Vision Science, pp. 13-43