Reasoning with Pseudowords: How Properties of Novel Verbal Stimuli Influence Item Difficulty and Linguistic-Group Score Differences on Cognitive Ability Assessments.
Year: 2018
Type: dissertation
Abstract: Pseudowords (words that are not real but resemble real words in a language) have been used increasingly as a technique to reduce contamination due to construct-irrelevant variance in assessments of verbal fluid reasoning (Gf). However, despite pseudowords being researched heavily in other psychology sub-disciplines, they have received little attention in cognitive ability testing contexts. Thus, there has been an assumption that all pseudowords work equally and work equally well for all test-takers. The current research examined three objectives with the first being whether changes to the pseudoword properties of length and wordlikeness (how much a pseudoword resembles a typical or common word in English) led to changes in item difficulty on verbal Gf items. The second objective was whether boundary conditions existed such that changes to pseudoword properties would differentially impact two linguistic sub-groups of participants – those who have English as their dominant language and those who do not have English as their dominant language. The last objective was to index and explore performance on these verbal Gf items when pseudowords were replaced with real words. Hypotheses predicting how pseudoword properties influenced item difficulty, how stimulie type – pseudoword or real word, impacted performance across linguistic sub-groups, and how linguistic sub-group status interacted with pseudoword properties were tested. Four sets of pseudowords were developed – short and wordlike, long and wordlike, short and un-wordlike, and long and un-wordlike, as well as two sets of real words – short and wordlike, and long and un-wordlike. Sixteen verbal Gf items, adapted from the LSAT, were developed to accommodate the pseudowords or real words and explore these three objectives. While none of the hypotheses were statistically significant, the results did indicate further areas of exploration. Specifically, verbal Gf items were easier when they featured longer pseudowords and more difficult when they featured un-wordlike pseudowords. Additionally, while performance of English-non-dominant participants was fairly balanced across real and pseudoword sets, English-dominant participants performed better on items featuring real words. Similarly, linguistic status interacted with wordlikeness such that English-dominant participants featured a decrease in performance as pseudowords moved from wordlike to un-wordlike. A full discussion of the findings, their implications, limitations of the current study, and directions for future research are included.
Author: Paul Agnello
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W2787794184