Paper Four

Evolution, Categorization and Values



The aim of this paper is to present an evolutionary framework for categorization. Evolution needs an evaluation mechanism to work, and it is argued that primary values that the organism needs for its survival - such as food, mates for reproduction, and shelter - can drive the evolution of categories. Sensory stimulation is needed to build up the cognitive apparatus, but cannot in itself provide the evaluation mechanism for evolution. Categorization constrained by values will be dependent on the availability of sensory information, and its power as predictive of values. As perception and categorization are tied to the actions of the organism, it is argued that the unit of perception should be seen as larger than the usual single-dimension stimulus, and evidence is reviewed to support this claim. Covarying stimuli will also provide a much greater predictive power than single-dimension stimuli alone.



Download and order Go: Up Intro One Two Three Four Five Refs



Short note on the paper:

While the first three papers focus on the linguistic realization of meaning, Paper Four is a result of my curiosity about what meaning might be before it is lexicalized in language. The paper was written as a reaction against two research traditions that have exercized a great influence on my work - cognitive linguistics and constructivism - and the same critique applies to both these traditions: it is not enough to focus only on mental representations when studying meaning and cognitive processes. It is also necessary to consider the constraints imposed by the "real world." This is especially important in an evolutionary context. Evolution can hardly be imagined to be promoted by cognitive representations alone, but needs the physical constraints to build up selective pressure.

I am seriously in debt to Annika Wallin, Per Johansson and Christian Balkenius for many of the ideas in the paper, and much of the incentive for writing the paper came from the discussions at the conference "New Trends in Cognitive Science - Does Representation Need Reality?" in Vienna, May 1997. The paper has been submitted to Evolution and Cognition.

The focus in the paper is on how meaning is built up for living organisms, from things that are inherently meaningful for the organism, such as food, mates and shelter, to associations of these with previously meaningless things. The function of our senses is important in this connection. What is taken for granted by people who lack scientific schooling is the connection between our senses and what we perceive, but on closer inspection, we must separate the salient stimuli that the senses detect from what is behind the appearances, i.e. what is meaningful for us.

To give an example: our ears detect frequency variations in air pressure that we call sound. However, the sound is not useful for us in itself. Rather, the reason that we have developed ears is that the frequency variations point to distinctions in other domains that have proved useful in the course of evolution.

In everyday language we tend to mix the sensory impressions with the real-world consequences. For example, we use the same word "hurt" for both the sensation of pain ("it hurts") and the physiological process of injury ("he is hurt").

The physical entities that are useful for us I have called values in the paper. Together they constitute all our life processes - eating, mating, sleeping, etc. - in a value loop that we cannot escape from without dying. We must respect our meals!

In these "values" I found a firm foundation for meaning that I believe can be used for building up what is considered as meaning in language. A possible research program for a new semantic theory can be built around the question "What perspective can we take that includes both values and linguistic meaning?"

A first step in that direction is taken in the paper. I examine the phenomenon of covariation or property clustering. The gist of the argument is that real-world interaction is on the level of many properties at a time. When I eat a strawberry, there are several simultaneous processes going on: I see a red object, the strawberry is decomposed in physiological processes, I taste the sweetness and feel the texture and form with hand and mouth. Also, objects are subject to constraints such as physical coherence: if one part of the object moves, the rest usually moves with it. Thus, several properties form a cluster that can be used for cognitive processing, as the following quotation from Givón indicates:

Inference from clustering of categorial properties:

(a) "Individual members of a natural category do not share only a single criterial property. Rather, they most often share many properties, which are thus the definitional core of their categorial membership."

(b) "Therefore, if known members of a group exhibit properties A, B, C etc., and if a sample sub-group also exhibits property Z (to a statistically significant degree), then it is highly likely that the rest - untested - members also exhibit property Z." (Givón 1989:276)

As we argue in Paper One, what has been lexicalized in language as nouns builds upon the clustering of properties in several dimensions at a time (and adjectives, in contrast, on single dimensions).

There also seem to exist neurological processes in the visual cortex that react to complex properties, where the decomposition of the stimuli into simpler stimuli does not provoke the same response. (See Tanaka 1993 and section 3.4 in Paper Four.)

If we consider the inherently meaningful values as the foundation of cognition and action, it becomes natural to consider several interacting dimensions as a complex unit of perception. The whole sensory potential is present at the same time.

Science, however, as I have argued in the Introduction, tends to take one perspective at a time when examining the processes involved. There are very few tools developed for this kind of analysis where several dimensions are modeled simultaneously. However, computer simulation is rapidly changing the field, and I look forward to simulation models of meaning systems. (See Casti 1997; Harder 1991; Stewart & Cohen 1997; Paper Five.)

Download Paper Four. pdf 188 kB