Cues and Cue Theories
Much thinking and research on the topics of deception and deception detection involves the idea of “cues.”
I am a vocal critic of cues. Truth-default theory rejects cues as providing a viable path to deception detection and provides an alternative to cue theories. I think of cues as the stuff of folk myth. Scientifically, I think cues are red herrings.
On this page, I explain my thinking and provide links to my most relevant peer-reviewed published work.
What are cues? Cues (as I think of them) are specific observable behaviors that are used to differentiate between honest and deceptive communication. They are “tells,” signals, or indicators. Cues can be verbal or nonverbal. Examples of cues include averting eye contact, changes in vocal pitch, fidgeting, the number of unique details in a statement, or the number of third-person pronouns (just to name a few).
Cue theories refer to a category of influential social scientific theories, accounts, and arguments that share a common logical structure explaining why deceptive and honest communication should produce specific behavioral differences (cues). For example, maybe liars won’t look you in the eye because they feel shame. Maybe increased fidgeting and increased vocal pitch signal lying because lying is more arousing than being honest in high stakes situations. Maybe liars use fewer first-person pronouns to linguistically distance themselves from their lies. Maybe liars provide fewer verifiable details because they want to strategically avoid detection. Maybe honest people have fewer speech errors because honesty is less cognitively demanding than deception.
By the way, there are no honest cues and deception cues per se. Difference in one direction or the other are in comparison. Compared to honesty, lies involve more (or fewer) of a given cue or vice versa.
I have 4 big objections to cues and cues theories in current theory, research, and practice.
First up, my reading of decades of scientific evidence is that cue findings simply do not replicate. While there is plenty of evidence for this cue or that cue in individual studies, cue findings do not hold up over time and across labs. As I show in Duped, the more often a cue is studied, the weaker the evidence. If findings that hold up are the hallmark of solid science and if claims need to be falsifiable to be scientific, then acceptance of science and beliefs in cues seem incompatible. If you have heard of the replication crisis in social science, cue research is a poster child for the problem. Getting technical for a moment: there is lots of p-hacking and overfit modeling and relatively little cross-validation.
My second objection is practical. The main value of cue approaches is in their promise for application. Cues and cue theories promise to help us be better lie detectors. This is false advertising for reasons beyond just a lack of scientific evidence that replicate. Let me explain.
To date, scientifically speaking, probably the best documented cue effect is for the number of details. On average, and all else being equal, liars provide fewer details than honest communicators. This finding is highly statistically significant meaning that is not just chance, and we can scientifically dismiss the idea that details are exactly the same for honest and deceptive statements. The size of the difference is small to moderate on social scientific standards. Honest and deceptive statements differ in the number of details (on average and under controlled conditions) by one-fifth to one-third of a standard deviation. So, as example, for scores on a typical test in one of the classes I teach with an average of 75% and an SD of 10%, this is like the difference between a student who scores 73% and one who scores 76%. I would not have much confidence that the higher score student knows more. Or, more relevantly, in deception detection experiments where the average accuracy is 54% with a SD of 6%, this like the difference between 53% and 55%. The difference is real, but it is not big.
Here is the real problem for real life deception detection. The difference is an average across many truths and lies that are otherwise experimentally equivalent (random assignment, same topic, homogeneous population etc.). The effect is statistically useful if we are judging large numbers of statements and our goal is to probabilistically be better than chance under controlled conditions. If this were a gambling situation, using details is a winning strategy in the long run. You play long enough, you come out ahead. It is not helpful, however, for placing any single bet (or assessing the honesty of specific message).
Cues are completely unhelpful in assessing a particular message. If you want to know if a statement is honest or not, you need a cut-point or decision rule. How few details do there need to before we conclude that it is a lie or how many details are exculpatory? There are many other things besides honesty-deception that influence how detailed a statement is. How long ago was it? Was the person paying attention? How memorable was the event? How good is the person’s memory? I, for example, lack visual memory. If I can’t report a visual detail, it does not mean that I am lying. If you can’t provide as detailed a description of Bond and DePaulo’s (2006) findings as I can, it does not mean that you are dishonest. Maybe I have just read that article more times than you or I have more experience reading social science articles.
As armchair academic speculation, cues and cue theories are interesting ideas worth investigating, that (at least so far) seem to have not panned out. When cues research findings are sold to practitioners, funding agencies, and the public as applied science, however, I call shenanigans.
My third point involves a distinction I see between cues and demeanor. If cues are specific stars in the sky, demeanor is a constellation. Usually when we look up at the nighttime sky, we don’t focus on one star. Instead, there is a mosaic of stars. Communication works like this. On the sender side, cues do not occur in isolation but instead co-occur with many other specific behaviors which are all inter-correlated. Cues are given off as parts of larger behavioral displays which stream over time. And, this is how they are perceived. Most cue research artificially isolates cues (violating statistical best practice) and providing an artificial and misleading portrayal of how communication works.
Finally, all the focus on cues distracts from communication content, evidence, and persuading honesty which I believe provide more promising approaches to improved lie detection.
Here is an important caveat. I am not saying that cues don’t exist. They do. They are easy to spot if you look for them. Details exist and can be measured, quantified, and counted. People fidget, they use pronouns, they change their gaze, etc. My argument is that cues do not reliably differentiate between honest and deceptive in ways that hold up over time, across people, and across situations. Cues are too ephemeral to be practically useful and cues are affected by too many things other than honesty-deception. Thus, cue- and demeanor-based lie detection is inherently error prone in predictable ways.
Want to read more? Looking for peer-reviewed work on the topic?
My account of the empirical evidence for cues and cue theories:
Levine, T. R. (2018b). Scientific evidence and cue theories in deception research: Reconciling findings from meta-analyses and primary experiments. International Journal of Communication, 12, 2461-2479. https://ijoc.org/index.php/ijoc/article/view/7838
A short article distinguishing between cues, demeanor and content in the context of interrogation to detect deception:
Levine, T. R. (2022a). Content, context, cues, and demeanor in deception detection. Frontiers in Psychology, 13, https://doi.org/10.3389/fpsyg.2022.988040
My demeanor studies which brought Malcolm Gladwell to my work:
Levine, T. R., Serota, K. B. Shulman, H., Clare, D. D., Park, H. S., Shaw, A. S., Shim, J. C., & Lee, J. H. (2011). Sender demeanor: Individual differences in sender believability have a powerful impact on deception detection judgments. Human Communication Research, 37, 377-403. https://doi.org/10.1111/j.1468-2958.2011.01407.x
Finally, and essay on alternatives to cues for improving deception detection:
Levine, T. R. (2015). New and improved accuracy findings in deception detection research. Current Opinion in Psychology, 6, 1-5. https://doi.org/10.1016/j.copsyc.2015.03.003
And of course, there is Duped. Chapter 2 summarizes cue findings. Chapter 4 reviews cue theories. Chapter 5 critiques cue theories. Chapter 13 explains how cues and demeanor push accuracy down toward chance. Chapter 14 talks about alternatives for cue-based lie detection.