Several recent studies have explored the applicability of the preferential attachment principle to account for vocabulary growth. According to this principle, network growth can be described by a process in which existing nodes recruit new nodes with a probability that is an increasing function of their connectivity within the existing network. The current study combined subjective estimates of the age of acquisition (AoA) and associations among words in a large corpus to estimate the organization of semantic knowledge at multiple points in vocabulary growth. Consistent with previous studies, the number of connections or relations among words followed a power law distribution in which relatively few words were highly connected with other words and most words were connected to relatively few words. In addition, the growth in the number of connections of a word was a linear function of its initial number of connections, and the ratio of connections to any two words was relatively constant over time. Finally, number of connections to known words was a reliable predictor of a word's AoA. All of these findings can be shown to be consistent with the preferential attachment principle.