The most elementary kind of localization is dead reckoning (sometimes also called path integration or odometry), which determines the creature’s position by monitoring its motion through space.
– Cognitive Maps and the Language of Thought
Thus, human similarity judgments do not appear to be simple, context independent distance computations but depend on the relative position of the objects to compare.
– Concepts and Compositionality: In Search of the Brain’s Language of Thought
To lie at someone’s door is an idiom. It is typically used to attribute responsibility for something which has gone wrong to someone. In a sentence, it might be used the following way: The delay cannot be laid at their door. This idiom is useful to help us appreciate the underlying meaning through its use of imagery.
If something lies outside your door, you would have to pick it up or in some way deal with it. It is after all your door. The idiom is often used in disputes or negotiations both traditionally considered the exclusive province of logic. It is a reasonable assumption that to make clear one’s position and to build an argument, a series of statements would have to be made, “each of which must be true if the statement before is true” (definition of logic, Collin’s Dictionary).
Logic is linear, sequential and follows certain structural rules. When engaging in an argument therefore, one would scrutinise the syntactic structure of the sentences which are uttered in sequence. Attention would be had to a special class of connectives known as “adversarial conjunctions” such as “but, still, and yet, nevertheless” (Semantics of Paragraphs, Zadrozny & Jensen, 2003). None of these words lend themselves readily to a pictorial representation. Accordingly, it is easier to conceptualise participants in a battle of words more like programmers working with strings than artists working with a brush.
It is therefore interesting, that to be effective in making a logical point, recourse is had to imagery such as the threshold of someone’s door by professionals whose bread and butter is logic. Why are images or to be more precise situating concepts in an image, effective in conveying logic?
The answer may have something to do with the way we think. The language of thought hypothesis (LOTH) suggests that thoughts have a language like structure. This means to form thoughts we combine a limited number of concepts in unlimited ways just like we form sentences. This applies also to non-linguistic thought (Frankland & Greene, 2020) – this is thought which is not quite sharply defined in sentences in our mind. In the LOTH, thought is amodal; it does not constitute sentences or pictures but is formed the way sentences are constructed.
Some researchers and philosophers suggested that all thoughts are formed only in sentences. Others suggested that we think in pictures only and then convert these mental images to language. There were strong arguments on both sides.
Rescorla (2009) found the “logicist”- sentence structure model wanting because as others have pointed out, babies or animals and those who for some reason do not have developed linguistic capabilities still find their way around the world. Many researchers did not agree that we think only in images. She therefore attempted to offer a model which showed how the brain operates with images but combines such images by way of rules such as those used in sentence construction.
The model she preferred is known as cognitive mapping. In essence what she suggests is that in different situations, appropriate conceptual spaces are created in the brain, which look like maps. In these context specific mental spaces, each concept is positioned in relation to others just like how in a map any element only has meaning in relation to some other element.
The map analogy is apt here because the process of thinking is likened to navigation. It deals with the following three functions; “localization, mapping and path planning” (Rescorla, 2019). According to her, localization asks the question where am I in relation to the entities around me. Mapping entails asking where different entities are located. Path planning relates to finding a way to reach some goal in a novel terrain.
Localization is done through what is known as dead reckoning and piloting. Dead reckoning begins with first fixing a point of origin. Then, taking speed, time transpired and direction of travel, current location is estimated. This is done even without awareness of the salient features of an environment. This method of navigation is used mainly only by necessity if for some reason the visual mode of perception is not available. Rescorla (2009) gives the example of the “desert ant” which “can return directly to the nest with remarkable accuracy, even lacking relevant external cues.” This mode is prone to “errors” which are “cumulative” over time because, it does not account for directional drift.
To correct the errors of dead reckoning, the traveller must take into account the position of external objects in the environment in relation to her own position. This is known as piloting. This is also not without difficulties because, the traveller must have an accurate map of existing elements in an environment from which to take reference.
The difficulty of mapping an unknown environment while at the same time locating a traveller’s position in it is known as Simultaneous Location and Mapping (SLAM) – (Titov & Motorin, 2019). The solution to this problem would lead to widespread adoption of autonomous mobility. Rescorla explains that a Bayesian probabilistic approach could be a solution – where a hypothesis about location or terrain is formed and confirmed or rejected based on evidence gathering.
Frankland and Green (2020) have suggested that thinking is structured both like maps and like sentences and that both types of thinking work together. Here, the map analogy is appropriate because of the notion of conceptual distance of one entity in relation to another.
To make their case, they referred to neuroimaging studies which documented which parts of the brain were active during different types of thinking.
Brain regions they refer to in their paper include the Default Mode Network (DMN) which refers to “an interconnected group of brain structures” (neuroscientificallychallenged.com) such as the angular gyrus (responsible for reading, writing and interpretation of language) and the medial prefrontal cortex (mPFC) which has a “fundamental role in a wide range of social cognitive abilities such as self-reflection, person perception, and theory of mind/mentalizing” and has been shown to be used for “learning and predicting the likely outcomes of actions” (Frontiers in Human Neuroscience).
They also referred to regions beyond the DMN such as the left-mid superior temporal cortex (lmSTC) which has an important role in “sentence comprehension”. When readers were shown two phrases with actor and patient (subject and object) reversed – the baby kicked the grandfather vs the grandfather kicked the baby, the strength of response in the amygdala (responsible for fight or flight responses) differed based on how activity in the ImSTC varied.
The DMN is able to generate context specific conceptual maps. These conceptual maps have been shown to have a grid like structure with directionality and distance as key factors which determine placement of concepts within such structures. For example, “a fax machine is in between a photocopier and a phone or that a chair is closer to a loveseat than a lamp” (Frankland & Green, 2020). To illustrate directionality, they give the example of “the president is like a child” and “a child is like the president” – the comparison works only one way. They suggest that the fax machine could be invented because the phone and the photocopier were for some reason mapped in the same conceptual space.
According to them, the grid like maps of the DMN are activated when thinking about “continuous variables” which can be situated in a “semantic space”. Here proximity in the representational space (e.g. adjacency on a map) corresponds
to “proximity in the represented domain” and the “graded relational information is implicitly coded”. So, a Vespa can be situated in a semantic space with vehicles in general and some might locate it closer to bicycles than to cars while others may locate the bicycle further away because it has no motor. It will be very easy to acquire and contextualise new information, say for example when someone looks at an ATV for the first time because the DMN would be able to activate the relevant semantic map. This concept of an ATV would be implicitly understood because of the DMN’s role.
Consider the sentences; A vespa stung her and She stung a vespa. Vespa is not a scooter; it is a kind of hornet or yellow jacket. Here a different type of semantic map has to be activated because the context has changed. Then, the regions beyond the DMN would have to be activated to accept the truth value of A Vespa stung her and reject She stung a vespa. We would reject She stung a Vespa because, we would assume She is a human and know that it is highly improbable for a human to be the actor or agent in that sentence and conclude that she could only be the patient (the one the action was done to). This “assignment of values” to “discrete, non-continuous variables” – [there being no similarity between humans and hornets (physically)] happens in an “algebraic way”; the way sentences are formed – we choose a word based on the word preceding it.
In other words, though default maps exist in the brain and are very useful, we need to take note of two things. First, we need to ensure the right map has been activated. Second, we need to rely on something beyond maps or default settings to make value judgements.
It is biological and natural to map novel elements or new people into existing maps – this entity/person resembles/is different from an existing entity/person; therefore, it/she/he must be close/far in space to such entity/person. We assign values to what we encounter based on the similarity or lack thereof to what has been previously encountered.
However, there is no map to El Dorado; the man or the place.
The Brain Dojo
