Network development.

Rapid
After only 2500 link selections (out of 150^2 possible links) both experimental networks had achieved a well-organised structure. This was in particular true for the second experiment, where the addition of the symmetry learning rule practically doubled the introduction of new links in the initial stages of network development.
Afterwards network development slowed down and settled in on the connections that had been established, due to:

limited network size: the reservoir of new meaningful connections was exhausted.
with increasing connection strength, the fixed rewards contributed relatively less and less to a specific connection strength.

Table 1 might illustrate how connections from the node 'MIND' are gradually built up and how the ordering of connections slowly settles in throughout the development of the network.

Clustering
"Hypertext Browsing is the ultimate accretion medium."[Jonassen, 1989]
After an initial phase of rapid development of new, meaningful links that superseded the random, initialised connections, nodes got clustered in medium-sized groups or "families" of associatively related nodes. The network's evolution seemed to have discovered a number of strongly related semantic "families". These clusters seemed to group the 150 nodes according to "higher" categorie-concepts, whose meaning was represented by the combined meaning of its constituent nodes.
A cluster analysis (using powers of the network's matrix to find strongly inter-connected groups of nodes) of the matrix of connections revealed a number of stable and separable clusters corresponding to highly general categories. The following 9 clusters of associated words, each denoted by an intuitive label for the underlying conceptual category, were found in the second experiment's final structure:

"Time": age, time, century, day, evening, moment, period, week, year
"Space": place, area, point, stage
"Movement":action, change, movement, road, car
"Control": authority, control, power, influence
"Cognition": knowledge, fact, idea, thought, interest, book, course, development, doubt, education, example, experience, language, mind, name, word, problem, question, reason, research, result, school, side, situation, story, theory, training, use, voice
"Intimacy": love, family, house, peace, father, friend, girl, hand, body, face, head, figure, heart, church, kind, mother, woman, music, bed, wife
"Vitality": boy, man, life, health
"Society": society, state, town, commonwealth
"Office": building, office, work, room

Although the learning algorithms only work on links and not on groups of nodes, it is remarkable how well the resulting clusters fit in with intuitive categories. With rare exceptions (e.g. "side" in the "Cognition" cluster), all of these words seem to be located in the right class.

Positive feedback and network development.
Our interface principle, ordering of available links according to their connection strength, combined with the Hebbian principle underlying our Frequency learning rule, introduced a positive feedback loop into the functioning of our Adaptive Web. Since the strongest connections always appeared on top of the list of available connections, they also had hightest probability of being selected by browsers. This in turn made these connections more elligeble for rewards administered by the Frequency learning rule. We expected this "Mathew"-effect to keep reinforcing the same existing connections and hamper the introduction of new, original connections into the system.
A temporal analysis of the development of the 20 strongest links in the network showed however that atleast 6 out of these 20 connection had been introduced by the Symmetry and Transitivity learning rule. This demonstrates that in spite of the positive feedback loop which reinforces only the already succesful connections, a large part of the connections still emerged without having been explicitly selected by human browsers. They had been independently proposed by the Adaptive Web system's Symmetry and Transitivity learning rules.
This could be explained as follows. Frequency only reinforces existing connections. It does not influence the introduction of new connections by the Symmetry and Transitivity learning rules. New connections however, if browsers think they are worthwhile and actually start using them in navigation, will due to positive feedback quickly be picked up by Frequency in a fast loop of self-reinforcing rewards, to finally reach their optimal connection strength among the other connections.
In this respect the positive feedback loop actually speedens network development and encourages the creation of new links.