Stylistic Analysis As Evidence And Its Application To Deception Detection In Corporate Fraud Cases

In recent years, unveiled high-profile corporate frauds have greatly shattered investors' confidence in corporate governance and information sharing system, and eventually professional managers' ethics. Not only does corporate fraud erode hard-earned investment from individuals, but also it is disappointedly formidable to convict those "white collars" who mastermind the malfeasance due to the complexity of such cases and the suspects' attempts to mislead the trials. There are voices calling for effective means out of legislative efforts to decode deceptive behaviors in order to bring punishment to the guilty, and linguistic analysis is considered a new approach to collecting forensic evidence. To this end, this thesis discusses the possibility of detecting deception with stylistic research methods and related stylistic indicators.The first three chapters embark on stylistic analysis in forensic linguistics and its common approaches. By listing the qualitative and quantitative markers used in practice, research methods of locating specific stylistic indicators are manifested. Chapter Four introduces different schools of deception detecting techniques and achievements, which shows a course of deception detection development. Much emphasis has been put on Chapter Five, which presents a research designed and conducted based on four most influential studies to test the feasibility as well as predictability of some deception detecting criteria from stylistic point of view. After analyzing up to 20 sample hearing transcripts from two highlighted "white-collar" cases, i.e. Enron and WorldCom, eight stylistic markers derived from previous studies were found to occur in witness/suspect's narration, among which five are related to deception at a statistically significant level, with some criteria showing high tendency of co-occurrence. A further study revealed that the combined predictability of these five criteria is much higher than any single indicator used alone. The concluding chapter points out some limitations about the research results and hopefully portends the future study direction in this field.