Linguistic Cues

Initial research on deception focused on nonverbal cues associated with deceptive speech, like fidgeting that could accompany anxiety. The leakage hypothesis stated that deception is arousing and engenders apprehension, guilt, and anxiety, which are then leaked through nonverbal behavior. However, research since the 1990s has started to focus on linguistic cues to deception and often uses computer applications to analyze texts for certain categories of words that are hypothesized to be related to deception. The underlying assumption is that senders engaging in deception change their language both strategically and nonstrategically compared with senders telling the truth.

Two theories that posit linguistic differences between truthful and false memories are criteria-based content analysis and reality monitoring theory. Criteria-based content analysis theorizes that there are 18 linguistic cues associated with truthfulness. These include a greater amount of logical coherency, more quotes, and more corrections. Reality monitoring theory also examines linguistic differences between truthful memories versus fabricated memories. Fabricated memories are hypothesized to contain less detail, especially sense-based, perceptual, and contextual detail because someone relating a truthful memory would have more access to this information.

Research using computer applications to detect linguistic differences has predominately used two software programs, Linguistic Inquiry Word Count (LIWC) and Coh-Metrix. These computer programs assign words to psychological (that is, negative-emotion words) or semantic (that is, first-person pronouns) categories. LIWC compares every word in the transcript to its internal dictionary, and it evaluates over 70 different word categories and identifies language indicative of emotional and psychological phenomena and parts of speech. Coh-Metrix also is a natural-language-processing algorithm. Although originally used to examine linguistic features related to the cohesion of a text, research has also applied Coh-Metrix to detect differences between truthful and deceptive text. Coh-Metrix has a larger number of linguistic indices than LIWC (over 700) and integrates more semantic analysis and parts of speech classifiers.

Some linguistic differences may reflect strategic choices deceivers make to reduce the likelihood of detecting their lie. One area in which deceivers may strategically change their language is through the amount of words. Some research has found that deceivers use more words than truth-tellers in order to craft a more believable lie or to control the interaction with the receiver. This has been termed the Pinocchio Effect. However, other research has found that deceivers use fewer words than truth-tellers, possibly to avoid contradicting themselves or to conceal information they do not want to disclose to the receiver.

One moderating variable may be the detectability of the lie. For lies where there is a verifiable fact that the sender could possibly compare with the lie the deceiver may strategically use fewer words to conceal and avoid contradiction with a knowable reality. However, when the receiver is not able to verify the lie (that is, the sender is lying about a nonverifiable opinion), the liar may strategically add more justifications and details to create a plausible story. In addition, some research has found that liars use more causation words (for example, “because”) for similar reasons of construction of a coherent and plausible lie. Other research has found that liars tend to use more words than truth-tellers in computer-mediated communication, where they have more time for strategic lie construction and editing than in face-to-face communication.

...