کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
326322 | 542230 | 2016 | 15 صفحه PDF | دانلود رایگان |
• We show how the form of a multinomial processing tree can be inferred from data.
• A tree is inferred from data in the literature on recall from a phonological network.
• A comparability graph can combine results from different experiments to infer a tree.
• Uniqueness of the tree depends on partitive sets.
• We describe the form of partitive sets in multinomial processing trees.
In a multinomial processing tree, processes are represented by vertices in an arborescence, i. e., a rooted tree with arcs directed away from the root. Processing begins at the root. When a process is completed one of its possible outcomes is produced. This outcome may result in another process starting, or in a response at a terminal vertex. The form of a multinomial processing tree can sometimes be inferred from data. Suppose an experimental factor, e.g., an item’s serial position, changes probabilities of outcomes of a single vertex in the tree, all else invariant. The factor is said to selectively influence the vertex. Suppose each of two factors in an experiment selectively influences a different vertex in a multinomial processing tree. The tree is equivalent to one of two relatively simple trees and the data will indicate which applies. As an example we construct a multinomial processing tree from data of Vitevitch et al. (2012) on immediate serial recall. One factor was the clustering coefficient in a phonological network of a to-be-remembered word. The other was the word’s serial position. The clustering coefficient selectively influences a vertex that precedes a vertex selectively influenced by serial position. The clustering coefficient has two opposing effects on recall. Part two of the paper is about a technical question. Suppose we learned for every pair of processes in a task whether they are executed in order or not. A comparability graph can represent the information. If there is an underlying arborescence it can be constructed with a procedure in the literature, the Transitive Orientation Algorithm. More than one arborescence may be possible. Different possibilities arise from subsets of vertices called partitive sets. We show that a partitive set in an arborescence has a simple form. Vertices in commonly used Multinomial Processing Trees can be ordered in such a tree in only one way.
Journal: Journal of Mathematical Psychology - Volume 71, April 2016, Pages 7–21