Morphemes, Pseudo-morphemes, and Relational Structure: Effects on production and comprehension of English words

点击翻译按钮获取中文摘要

Many theories of word processing assume the activation of constituent representations, but little is known about how these representations are integrated. By exploring factors that are involved in this process, and by separately identifying the role played by morphemic structure and relational structure, this project will greatly contribute to our understanding of how multi-morphemic words are comprehended and produced. Ultimately, this project will provide deeper insight into lexical productivity and semantic compositionality, which are the cornerstones of the human language and cognitive system. The proposed research has three streams. The first examines whether the language system attempts to construct a morphemic structure whenever morphemes are available, regardless of whether it is appropriate. Schreuder and Baayen (1995) found that response latencies and subjective frequency ratings of a monomorphemic word are influenced by the number of morphologically related forms for that word, which suggests that a morpheme actives morphologically related words. However, what is not yet fully known is whether similar effects occur when a morpheme is used non-productively. To illustrate, the word ‘carpet’ has no morphemic structure yet contains two morphemes (‘car’ and ‘pet’). I will conduct a series of experiments that will test whether properties of these morphemes (such as word frequency and positional family size) affect ease of processing as measured by lexical decision latencies. Semantically opaque and transparent compounds will serve as comparison conditions. The second stream tests whether relational information is used only after a morphemic structure has been established. I will use the possible relations task (Gagné & Spalding, in press) to create a distribution of possible relational interpretations for each item. The distributions will be used to derive two measures that reflect relational information: diversity and relational entropy. Diversity indicates the extent to which the possible interpretations are concentrated on a small set of relations. Relational entropy measures the extent to which the relational distribution for a specific item differs from the total relational distribution. Recent evidence shows that diversity and relative relational entropy affect the ease of processing both semantically transparent and opaque compounds, but in different ways (Spalding & Gagné, in press). I will explore the reasons underlying this difference and will also examine whether these variables influence the processing of pseudocompounds (e.g., ‘kitten’). If meaning construction is attempted for all items that contain embedded morphemes, then the processing of pseudocompounds will be affected by relational information. The third stream uses a typing task to examine the role (if any) played by relational information during the production of compound words. Consistent with the claim that morphemes might act as planning units during the production of morphemically complex words (Roelofs & Baayen 2002), Libben (2011) found elevated typing times at the morpheme boundary. My goal is to identify factors that affect this elevation and, in particular, to test whether relational information influences production. I will begin by examining whether diversity and relational entropy differentially affect typing times for semantically transparent and opaque compounds. I will also examine whether a context that supports a novel meaning (e.g., footrace = race that is one foot long) increases the size of the boundary delay and whether context interacts with diversity and relational entropy.