Entangled Associative Structures and Context

Scientists are engaged in mapping the universe, the land, the genome, and word knowledge. In this paper we describe results of a 30-year effort to map and understand how preexisting word knowledge affects memory for a recently experienced word. We show how some of these findings are inconsistent with widely held views in psychological science and support the incorporation of the quantum formalism in our attempts to understand how prior knowledge interacts with recent experience and context. We relate our work to the state context property (SCOP) quantum formalism. Introduction There exists a large body of evidence showing that seeing or hearing a word activates words related to it through prior language experience. For example, seeing PLANET activates the associates earth, moon, and so on, because planet-earth, planet-moon, moon-space and other associations have been acquired in the past. This activation aids comprehension, is implicit, and provides rapid, probably synchronous access to what is associated to a word. Understanding how such activation affects memory requires a map of links among known words, and nearly 30 years ago we set about this task by using free association to construct an associative map of word knowledge. In free association, we present individual words to samples of 150 participants asked to produce the first associated word to come to mind. We started using this task to index the strengths of preexisting links between pairs of words, with strength computed by dividing the production frequency of a response word by the sample size (e.g., the probabilities that the word planet produces earth and mars are .61 and .10, respectively). We soon discovered that some words had relatively small sets of associates whereas others had mu ch larger sets. Although no theory at the time predicted that a word’s preexisting associative set size would affect memory, we set about exploring its effects (e.g., Nelson and McEvoy 1979; Nelson, Schreiber, and McEvoy 1992). By the late 1980s, we realized that link strength and set size were capturing important dimensions of associative structure but were ignoring links between the associates (e.g., moon-space) and between the associates and the initiating stimulus (earth-planet). We then set about collecting free association norms for each word’s associates. Norms were collected through the 1990s until over 5,000 words were normed using 6,000 participants (Nelson, McEvoy, and Schreiber 2004). We discovered that the associative structures of individual words differ in both size and connectivity. Each word can be represented in an nxn matrix that describes the number and strength of three types of links: target-toassociate, associate-to-associate, and associate-to-target (see Nelson et al. 2004, for a database of 4,000 examples). Figure 1 shows the word PLANET in a network format to illustrate the three types of links. Figure 1. Planet’s associative structure Planet produces a relatively small set of 9 target-toassociate links (e.g., earth, moon) having many associateto-associate links (e.g., mars-to-earth) and many associateto-target links (e.g., earth-to-planet). Other words have relatively small sets and few associate-to-associate and associate-to-target links. All combinations of set size and connectivity are represented in the database. Because of the size of the database and because these indices are not highly correlated (Nelson and Zhang 2000), we were able to select words that systematically varied on these links in order to investigate how they affect recall and recognition. Initially we were interested in determining whether the nature of a word’s preexisting associative structure had any PL AN ET