Human association topology

In this chapter, the associations coming from various sources are compared based on ranked word lists. This does not reflect, however, the complex structure of human associations. These can be represented as weighted graphs with specific words in nodes and associations in the vertices. This graph can be then sub-divided into subnets by starting from a specific stimulus (word) and cutting off the net at a certain distance from this central stimulus. Those subnets can be treated as representative for a specific meaning of a word. The strongest associations are always correlated with the fact that they are bi-directional. But if we look at the each pair of connected words to find what the connection means, we see that connections can differ in meaning, e.g. home-mother says that a home is a place specific to a mother in contrast to home-roof which says that roof is a part of a building. Having analyzed all the pairs we may find that some of them connect the stimulus word in the same way, e.g. parents and family connect the home on the same principle as mother, and chimney and wall are parts of a building along with the roof This observation shows that the lexical meanings of the stimulus word are organizing subnets in an association network. We show two of them to illustrate the phenomenon. Figure 2.1 shows the sub-net for the meaning dom (“home”, as a place for family) and Figure 2.2 dom (“home”, as a building).

Figure 2.1. Human association subnet for dom (“home”, as a place for family)

The graph in Figure 2.1 shows the relations between the words: dom and rodzinny (family; adjective), stot (table), mama (mum), matka (mother), obiad (dinner), swigta (holidays), rodzice (parents) and rodzina (family).

Figure 2.2. Human association subnet for dom (“house”, as a building)

The graph in Figure 2.2 shows the relations between words: dom and komin (chimney), big (duzy), budynek (building), dach (roof), sciany (walls), chata (cottage), wies (village), ogrod (garden), osiedle (estate) and miasto (city).

Clearly both subnets were identified manually and it is hard to believe that it would be possible to extract those subnets automatically by a use of an algorithm operating solely on the network [GAT 16]. Then, we shall treat all the associations to a particular stimulus as a list disregarding what association means. Then, we may distinguish semantically valid associations comparing the Polish association list with the English associations obtained in the free word association experiment.