The Strategic Manipulation of Linguistic Complexity: A Test of Two Models of Lying

Abstract

To what degree is complex language driven by personal cognitive factors versus strategic self-presentation? Studies teasing apart these two influences on complexity are hard to design and evidence bearing on the question is not abundant. To fill this gap, the present studies explored two models relevant to a form of communication full of strategic implications: deception. The cognitive strain model suggests that because lies are cognitively draining, deception will generally reduce complexity, whereas the strategic model expects the liar to adjust complexity up or down depending on the perceived benefits. Three studies tested differential predictions from these models by scoring different forms of linguistic complexity (dialectical and elaborative) for deceptive communications in real-world and experimental contexts. Results from these studies support the value of a strategic model of the effect of lying on complex language, thus suggesting that people strategically manipulate the complexity of their language to accomplish specific goals.

Keywords

deception cognitive complexity integrative complexity interpersonal communication political communication

The complexity of language is widely predictive of very important phenomena, including violent outcomes in international crises (see, e.g., Abe, 2012), the economy (see, e.g., Abe, 2011), the behavior of terrorists (Conway, Dodds, Towgood, McClure, & Olson, 2011; Pennebaker, 2011), electoral success (see, e.g., Conway et al., 2012), and health outcomes (e.g., Davidson, Livingstone, McArthur, Dickson, & Gumley, 2007). However, in spite of its practical predictive importance, comparatively little is known about what causes complex language to begin with. At a broad level, one of the persistent questions yet to be fully answered is the degree that complex language is the result of personal cognitive factors versus strategic presentation/impression management concerns. Although personal cognitions and strategic self-presentation do not constitute mutually exclusive categories, nonetheless, some of the available evidence supports the idea that personal cognitive factors—such as the cognitive strain on the individual—affect complexity more than strategic concerns such as self-presentation (see Conway & Conway, 2011; Conway, Suedfeld, & Tetlock, 2001; Suedfeld, 1992; Tetlock, & Tyler, 1996). However, the evidence is not overwhelming and some research does suggest strategic concerns matter (e.g., Tetlock, Hannum, & Micheletti, 1984).

Indeed, as other researchers have pointed out, tests distinguishing strategic versus personal cognitive influences on complex language are hard to design and interpret (Tetlock & Manstead, 1985; also see Conway et al., 2001, for a summary), and it is possible that evidence for the strategic model has not been fully explored. In the present studies, we use a more recently developed linguistic analysis tool that allows for more nuance in measurement to pursue evidence from an emerging area that is at the intersection of this discussion: the relationship between lying and linguistic complexity.

Lying and Complexity

Although most research on cognitive complexity either assumes truthful statements (or does not distinguish between truthful and nontruthful statements), a small body of scientific research aimed at revealing the cognitive complexity behind lying has begun to take shape (e.g., Conway et al., 2008; Newman, Pennebaker, Berry, & Richards, 2003).¹ What is the relationship between lying and linguistic complexity? To date, a number of studies have revealed evidence for two diverging theories of what might be a reasonable expectation for cognitive complexity differences between lying and truth-telling. One of those theories focuses on personal cognitive processes (cognitive strain model), while the other focuses on presentational concerns (strategic model).

Two Models of the Effect of Lying on Complexity

Cognitive Strain Model

According to the cognitive strain/disruptive stress model (see, e.g., Suedfeld, 1992, 2010), people need cognitive resources available in order to think complexly. Thus, anything that adds too much strain would reduce complexity of all forms (see also, Richards & Gross, 1999, 2000).

Following from this, one theory of the effect of lying on complexity is that lying induces a degree of cognitive strain that truth-telling does not, and consequently, the resulting statements are less cognitively complex. This theory is somewhat ironic: Lying is a complex task, requiring managing multiple angles and stories, and the complexity of the task makes the actual output that people produce when they lie less complex. Consistent with this model, Newman et al. (2003) found that participants instructed to lie used a higher frequency of linguistic markers that are indicative of simplistic thinking (e.g., fewer qualifying words and more concrete verbs) compared with those instructed to tell the truth. Newman et al. (2003) suggest that lying inhibits one’s ability to produce complexity because the liar must expend cognitive resources on other cognitive tasks besides the lie itself, such as the stress that often coincides with lying.

While this framework for explaining the lying–complexity relationship is certainly useful, there is additional prior research that does not support this model. Specifically, there is other evidence that, while lying does decrease some forms of complexity, it actually increases other forms (Conway et al., 2008). This finding is inconsistent with the cognitive strain model, which predicts that all forms of complexity should be equally reduced by lying. Thus, it may be that the cognitive strain model does not fully capture the relationship between lying and all forms of complexity. An alternative, which we here call the strategic model, provides further insight into the complexity–lying relationship.

Strategic Model

The goal of lying is to deceive someone, and accomplishing that goal requires some degree of strategizing. The strategic model (see, e.g., Conway et al., 2008; Tetlock, 1985) suggests that people use complexity or simplicity depending on which avenue is believed to be most effective in deception. In particular, lies will be more complex when the liar perceives complexity as advantageous in deceiving someone, but simpler when the liar views simplicity as advantageous. These linguistic alterations can be uncovered via linguistic analyses.

There is some evidence that lends support to this theory. Specifically, Conway et al. (2008) found that participants who were instructed to lie about their true opinion on a particular topic showed less complexity in acknowledging alternative viewpoints about the topic (lower dialectical complexity, described below), yet more complexity in defense of the specific viewpoint they lied about (higher elaborative complexity). Consistent with the strategic model, this is because people likely assume the most effective strategy to deceive someone is to complexly elaborate on the lie—presumably to put forth a convincing defense of that lie—while not acknowledging the opposing position that might threaten belief in the lie.

While not couched within the same terms as the strategic model uses, Anolli, Balconi, and Ciceri (2003) also found evidence in support of this idea. They investigated the effect of deception on linguistic tendencies when an individual was communicating a lie to an acquiescent recipient or a suspicious recipient. In keeping with the strategic model, their findings indicate that participants employed differing strategies in response to communicating to the different types of recipient. When a participant was attempting to deceive a compliant recipient they were more likely to use ambiguous, less complex language, as opposed to when participants were attempting to deceive a skeptical recipient, to whom they tended to use more assertive and complex language.

Which Theory Is Right?

The purpose of the current article is to provide more direct tests of the cognitive strain and strategic models. While the cognitive strain model describes the process of lying as a taxing one, resulting in lies containing a decreased degree of complexity due to the fatigue the liar is under, the strategic model expects the liar to adjust their complexity up or down, depending on their perception of the benefit of complexity.

As we discuss in more detail below, complexity is not a unilateral construct—a person can be complex in different ways (Conway, Dodds, et al., 2011; Houck, Conway, & Gornick, 2014; Tetlock, Metz, Scott, & Suedfeld, 2014). One of the important distinctions between the strategic model and the cognitive strain model is that whereas cognitive strain ought to consistently reduce all forms of complexity equally, strategic concerns ought to affect only forms of complexity relevant to the immediate strategic goal. Thus, the strategic model suggests a more nuanced approach to the relationship between lying and complex language. In the present study, we utilize a validated scoring system (the multiple complexity model; Conway et al., 2008; Conway, Dodds, et al., 2011) for differentiating multiple forms of complexity.

Multiple Complexity Model: Dialectical and Elaborative Complexity

The present studies employ the multiple complexity model. This model derives its origins from integrative complexity (e.g., Suedfeld & Rank, 1976; Suedfeld, Tetlock, & Ramirez, 1977), a measure of cognitive complexity that identifies the degree that multiple ideas are differentiated and then (at higher levels) integrated. Integrative complexity assesses the complexity of spoken or written statements based on the underlying structure of the rhetoric, rather than on its content (see, e.g., Suedfeld, Conway, & Eichhorn, 2001; Suedfeld & Rank, 1976; Suedfeld, Tetlock, & Streufert, 1992). The integrative complexity score is determined by the level of differentiation and integration inherent in the statements being evaluated (see Baker-Brown et al., 1992). Differentiation occurs when different dimensions are present in the statement. Integration is present when connections are made between these differing dimensions.

In the multiple complexity model, passages can further be scored for two subtypes of integrative complexity that are the focus of this article: elaborative complexity and dialectical complexity (Houck et al., 2014). Elaborative complexity represents the development of a complex argument along a singular perspective. Dialectical complexity, on the other hand, involves the recognition of different perspectives that are in tension with one another.

The Present Studies: Distinguishing the Cognitive Strain and Strategic Models

The cognitive strain and strategic models are not mutually exclusive—and yet they do make different predictions with regards to dialectical and elaborative complexity that would help us determine when one or the other might be in operation. We focus on three such tests in the present set of studies.

Do Nuanced Strategic Effects Hold in Real-World Contexts?

One possible interpretation of previous work showing more nuanced strategic lying–complexity effects (eg., Conway et al., 2008) is that this work exclusively focuses on casual lies told in fairly artificial laboratory contexts. While this does not invalidate the value of the strategic model, it does open up the possibility that it is limited in its scope, and that the cognitive strain model is more likely to be in operation in more real-world contexts where people care more deeply about the lie. In Study 1, we attempt to replicate the nuanced, strategic pattern suggested by Conway et al. (2008) using a real-world lying context where it can be reasonably assumed that the importance level is much higher—a presidential debate.

Lying to a Split Versus Uniform Audience

The strategic model suggests that people will use complexity when they lie to accomplish a specific aim. As such, it leaves open the possibility that liars will manipulate the type of complexity that they use based on the strategic goal that type will attain. The cognitive strain model, in contrast, predicts no such subtle alteration of complexity type based on potential strategic goals—it only predicts a very straightforward downward trend of lying on complexity that is only qualified by the degree that lies might indeed be straining.

While prior work shows that people use more elaborative complexity and less dialectical complexity when they lie (Conway et al., 2008), that work does not offer any truly “strategic” reason for this pattern. There was no manipulation of possible factors influencing strategy. Furthermore, a strategic approach would predict that lies would differ on the degree that they contained elaborative versus dialectical complexity, based on the goal of the lie. Prior work offered no test of this hypothesis.

Thus, in Study 2, we manipulate one factor relevant to the goal of the lie on which the strategic and cognitive strain models make differing predictions: The type of audience. The strategic model suggests that the type of complexity liars use will depend on whether the audience they are talking to is split versus uniform. Specifically, if the goal of the liar is to maximize the amount of people in the audience that like her or him, then knowing the audience contained people with a high degree of consensus on a focal attitude, the liar ought to be especially likely to use a lot of elaborative complexity in line with the audience’s position—in order to increase the perception that they really do believe what the audience believes. However, if the audience is known to contain a lot of variability in attitudes on a focal topic, the strategic liar ought to be (relatively) more likely to use dialectical complexity in order to increase the possibility that a wider range of people will not take offense at what they said.

To the degree audience type influences the complexity type of persons when they lie, this suggests persons are being truly strategic in their manipulation of complexity.

Storytelling Versus Counterattitudinal Lying

Not all types of lies are the same. For example, the psychological properties of hiding one’s attitudes might be very different than the psychological properties of making up false stories. The strategic model predicts that different types of lies might show different strategic lying–complexity relationships. Strategic liars would be more likely than those affected by cognitive strain to alter their complexity levels based on the kind of lie being told. Study 3 directly tests the strategic model’s implication that the pattern of complexity would differ by the type of lie being told.

Study 1

Background

Study 1 evaluated the effect of lying on cognitive complexity using a real-world political event where evidence suggests persons were lying at one point, but not at another. The purpose of this study was to compare the results from a real-world scenario with the results from the Conway et al. (2008) study (which utilized experimentally manipulated conditions where participants told a very “casual” lie). The value of using a real-life example is that it addresses the potential issue of artificial/lab scenarios not producing the equivalent level of strain that individuals may feel when lying in real life.

Participants and Speech Selection

Speeches by two former presidential candidates, Richard Nixon and John F. Kennedy, Jr., were used for analysis. Specific speeches were chosen for analysis based on information revealed in Six Crises, a book written by Nixon. Due to knowledge that before the official debates of the 1960 presidential campaign began, Nixon had been urging the White House to take a stronger stance toward Castro and Cuba. During a conference with Castro in April of 1959, Nixon (1962) wrote a confidential memorandum stating that he was convinced that Castro was “either incredibly naïve about communism or under communist discipline” (p. 352) and that the United States needed to deal with him accordingly. Initially, Nixon’s tough position toward Castro left him in the minority; however, by early 1960, tensions had increased between the United States and Cuba, and Nixon’s position prevailed within the administration. At this time, the CIA launched a top secret initiative to provide arms, ammunition, and training to Cubans who had fled the Castro regime.

In the fall of 1960, Richard Nixon and John F. Kennedy were preparing for their fourth and final debate before Election Day. The debate was to cover foreign policy, and both candidates knew their words on Castro and Cuba would be critical in the outcome of the election. Privy to the details of the CIA operation, Nixon understood that advocating for very decisions that were already being brought to fruition secretly might jeopardize the operation. Consequently, Nixon prepared to alter his usual tough stance toward Cuba for the upcoming debate. Reports indicate that Kennedy was also briefed on the covert mission in advance of the debate, and that the information also altered Kennedy’s course of action for the fourth debate. While Nixon had long advocated for a harder line with Cuba, he came to the fourth debate ready to scale back and instead recommend an economic course of action. On the other hand, Kennedy had originally taken a softer line with Cuba, only to come to the fourth debate and recommend the U.S. aid rebel forces. In both cases, history suggests that the candidate acquired privileged information prior to the final debate, and that the information led to the presentation of a position that the candidate did not truthfully hold.

Speeches selected for inclusion in the present sample were the transcripts from four debates during the 1960 presidential election campaign. From those debates, sections were divided into “foreign policy” and “domestic policy” sections, and a random sample from each candidate was selected from each category; we also subdivided foreign policy up into Cuba- and non-Cuba-related statements (total paragraph N = 95). Subsequently, four trained coders scored the paragraphs for both dialectical and elaborative complexity (dialectical complexity α = .83, elaborative complexity α = .69).²

Results

Main Analysis

The main analysis of interest was a 2 (statement type: lying vs. truth) × 2 (type of complexity: dialectical vs. elaborative) mixed-model analysis of variance (ANOVA), with statement type as the between-subjects variable and type of complexity as the within-subjects variable. The statement type by complexity type interaction was significant, F(1, 93) = 6.25, p = .014, η_p² = .063. When the candidates were telling the truth, dialectical and elaborative complexity had very similar means (dialectical M = 1.51, elaborative M = 1.46). However, consistent with the strategic model and prior research in laboratory settings (Conway et al., 2008), when the candidates were lying, candidates showed much higher levels of elaborative complexity than dialectical complexity (dialectical M = 1.11, elaborative M = 1.72). See Figure 1.

Nixon Alone

Independent analyses were also conducted for each individual separately and, as Table 1 reveals, both candidates showed an identical pattern. First, a 2 (statement type: lying vs. truth) × 2 (complexity type: dialectical vs. elaborative) mixed-model ANOVA was conducted with data collected on Nixon, with statement type as the between-subject variable and complexity type as the within-subject variable. Results of the analysis revealed that for Nixon individually, the complexity type by statement type interaction was nearly significant, F(1, 55) = 3.91, p = .053, η_p² = .066. When telling the truth, dialectical and elaborative complexity levels were similar (dialectical M = 1.46, elaborative M = 1.42); however, when lying, dialectical and elaborative complexity levels were dissimilar (dialectical M = 1.10, elaborative M = 1.80).

Table 1.

Complexity Type by Statement Type Patterns for Nixon and Kennedy Alone.

	Nixon		Kennedy
	Truth	Lie	Truth	Lie
Dialectical	1.46	1.10	1.60	1.13
Elaborative	1.42	1.80	1.51	1.63

Kennedy Alone

Analysis on Kennedy also involved a 2 (statement type: lying vs. truth) × 2 (complexity type: dialectical vs. elaborative) mixed-model ANOVA. Results indicated that the complexity type by statement type interaction was nonsignificant, F(1, 36) = 2.32, p = .136, η_p² = .061. However, the mean pattern is consistent with the results for Nixon: When telling the truth, dialectical and elaborative complexity levels were somewhat similar (dialectical M = 1.60, elaborative M = 1.51); however, when lying, dialectical and elaborative complexity levels were dissimilar (dialectical M = 1.13, elaborative M = 1.63).

Analyses Considering Only Statements About Cuba

Data were also specifically analyzed for statements Nixon and Kennedy made regarding Cuba. For these analyses, all statements not involving Cuba from any debate were removed. Another 2 (statement type: lying vs. truth) × 2 (complexity type: dialectical vs. elaborative) mixed-model ANOVA, with statement type as the between-subjects variable and complexity type as the within-subjects variable was conducted. Results revealed that the complexity type by statement type interaction was significant F(1, 16) = 13.54, p = .002, η_p² = .458. When specifically discussing Cuba, the mean pattern of complexity levels was similar to the previously discussed pattern: When telling the truth, the candidates used more dialectical than elaborative complexity (dialectical M = 1.53, elaborative M = 1.17); however, when lying, they used more elaborative than dialectical complexity (dialectical M = 1.11, elaborative M = 1.72). This pattern for Cuba-only statements held when looking at Nixon and Kennedy separately. When telling the truth, Nixon used more dialectical than elaborative complexity (dialectical complexity M = 1.44; elaborative complexity M = 1.25), but when lying, he used more elaborative than dialectical complexity (dialectical complexity M = 1.10; elaborative complexity M = 1.80), F(1, 7) = 4.72, p = .066, η_p² = .403. Kennedy demonstrated the same pattern. When telling the truth, Kennedy used more dialectical than elaborative complexity (dialectical complexity M = 1.60; elaborative complexity M = 1.10), but when lying, he used more elaborative than dialectical complexity (dialectical complexity M = 1.13; elaborative complexity M = 1.63). F(1, 7) = 7.32, p = .03, η_p² = .511.

Discussion

These results are consistent with the findings from Conway et al. (2008), suggesting it is not a lack of sufficient strain felt by participants in a laboratory setting that results in a failure to find support for the cognitive strain model, but rather that engaging in counterattitudinal lying seems to reduce dialectical complexity while increasing elaborative complexity—consistent with what the strategic model predicts. Furthermore, this is consistent with the argument put forward by Levine and McCornack (2014) that despite the stakes of lie often being considered to be a critical element in research on lying, there is not consistent empirical research to support this view. Findings from Study 1 are consistent with this perspective. Despite the stakes of a lie told during a presidential debate being considerably higher than the stakes of a lie told during an artificial laboratory setting, they produced similar results when evaluating for the effect of lying on cognitive complexity.

Study 2

Background

Despite the evidence in favor of the strategic model across real-life and artificial lab scenarios, the question still remains of whether a reduction in dialectical complexity and an increase in elaborative complexity is truly strategic, or if perhaps there is another explanation for why counterattitudinal lying tends to follow this pattern. To gain further insight, Study 2 was developed to test whether this pattern would change when the complexity-related goals of the scenario changed. In other words, do people engage in strategic manipulation of complexity? To test this, the audience to which participants were ostensibly communicating with was manipulated. The goal of this manipulation was to determine if participants would strategically modify their use of complexity in response to whether their audience was all uniformly minded, or of diverse opinions on a given topic.

Participants

One hundred and thirty-three University of Montana undergraduates participated in this study during a mass testing session. In exchange for their participation, students received two research credits to be applied to course requirements for Psychology 100.

Materials and Assignment

Participants were first asked to identify their attitude toward one focal topic (either organized religion, death penalty, or easy access to birth control), to which they were randomly assigned. Because the nature of the study required participants to clearly agree or disagree with the focal topic presented to them, we asked participants the degree that they favored, opposed, or did not know their opinion on the focal topic. Participants who were unsure were excluded from analyses presented below, leaving a final N of 104 for analyses.

Participants were then told to imagine themselves as a politician about to give a statement on the assigned topic to a large crowd of people. Participants were randomly assigned to one of three conditions related to the hypothetical crowd they were to address. The first condition asked participants to prepare a statement which effectively asked them to lie to a crowd with a uniform opinion; “Imagine that the entire crowd of people [disagrees/agrees]³ with you on (insert topic) . . . lie to the crowd (when necessary) in order to win their vote. Please write a response that you think will make the largest number of people in the crowd want to vote for you.” The second condition asked participants to prepare a statement which effectively asked them to lie to a crowd whose opinion was split; “Imagine that half the crowd of people agrees with you on (insert topic), but the other half of the crowd disagrees with you . . . lie to the crowd (when necessary) in order to win their vote. Please write a response that you think will make the largest number of people in the crowd want to vote for you.”

These responses were scored by trained scorers for dialectical and elaborative complexity (α > .70) and their scores were averaged into two summary scores.

Results

A 2 (complexity type: dialectical vs. elaborative) × 2 (audience type: uniform opinion vs. split opinion) mixed-model ANOVA, with complexity type as the within-subject variable and audience type as the between-subjects variable, revealed a significant main effect of complexity type, F(1, 102) = 26.76, p < .001, η_p² = .21. However, this effect was qualified by the expected complexity type by audience type interaction, F(1, 102) = 4.95, p = .028, η_p² = .05. When participants were lying to a split audience, their levels of dialectical and elaborative complexity were modestly discrepant (dialectical M = 1.66, elaborative M = 1.95); however, when the participants were lying to a uniform audience, this difference became substantially larger, with participants showing substantially higher levels of elaborative complexity than dialectical complexity (dialectical M = 1.27, elaborative M = 2.00). No main effect emerged for audience type, p > .190.⁴ See Figure 2.

Discussion

Results from Study 2 provide evidence in support of the strategic model. When a person telling a lie must recognize opposing viewpoints in order to maximize approval, they are more likely to use dialectical complexity (and less likely to use elaborative complexity) than when lying to a group of uniformed–opinioned individuals. These findings confirm the implication of the strategic model that there will be different complexity outcomes according to a liar’s perception of the audience.

Study 3

Background

In Study 2, we saw that a similar type of lie could produce different outcomes on different types of complexity—if one alters the complexity required to achieve a strategic goal. In Study 3, we extend this result by examining the possibility that different complexity results emerge for different types of lies.

Not all forms of lying are the same. So far, all the lying we have discussed involves counterattitudinal lying: Trying to convince someone that your opinion is different than it is in order to accomplish some strategic goal (e.g., self-presentation). However, a very different form of lying—which we here call storytelling—involves trying to hide, not one’s attitude, but rather the actual facts of a presumed event. It is one thing to try to convince someone that one’s attitude is different than it is; it is another thing entirely to try to convince someone that an event happened that did not occur.

While these types of lies share similar properties, they also differ in many respects. Primary for our purpose is this: It is unclear that storytelling liars would benefit from increased complexity, regardless of whether that complexity is dialectical or elaborative. Concrete descriptions, no matter how detailed, are typically not complex (see, e.g., Newman et al., 2003; Suedfeld et al., 1992). Thus, even if a storyteller tried to increase the amount of information in their lie, it is unclear whether they would be motivated to increase the complexity of that information.

As a result, the strategic model would expect that, because the two types of lies differ in the functional relationship between lying and complexity, they might show different linguistic complexity patterns. However, the cognitive strain model predicts that lying reduces complexity across all forms of complexity for all types of lies because lying drains one’s cognitive resources, and thus would expect the same pattern for both kinds of lies.

To distinguish these two possibilities, Study 3 had some participants tell a counterattitudinal lie in a condition similar to prior research that showed lying produced more elaborative and less dialectical complexity (Conway et al., 2008), while other participants told a storytelling lie.

Participants

One hundred and eighteen University of Montana undergraduates participated in this study during a mass testing session. In exchange for their participation, students received two research credits to be applied to course requirements for Psychology 100.

Methods and Assignment

All participants were asked to write a letter to a friend regarding a hypothetical exit exam that the University of Montana was considering as a graduation requirement for seniors. Participants were assigned to one of two conditions. The first condition asked that participants lie to their friend by writing the opposite of their opinion in the letter (counterattitudinal lie condition). The second condition asked that participants “prank” their friend by writing a letter explaining that university administration had already decided to institute the exit exams (storytelling condition). In each condition, following drafting the letter to a friend, participants completed a survey of their actual opinion toward such an exam, questions regarding their feelings toward lying in general, and general demographic questions.

As before, all paragraphs were scored for dialectical and elaborative complexity by trained scorers (αs > .67) and averaged into summary scores for analyses.

Results

A 2 (complexity type: dialectical vs. elaborative) × 2 (lie type: general vs. storytelling)⁵ mixed-model ANOVA, with complexity type as the within-subject variable, and lie type as the between-subjects variable revealed a significant main effect of complexity type, F(1, 109) = 17.86, p < .001, η_p² = .141. Results also revealed a nonsignificant main effect of Lie Type, F(1, 109) = 2.50, p = .117. However, these effect were qualified by a significant complexity type by lie type interaction, F(1, 109) = 7.25, p = .008, η_p² = .062. When telling a counterattitudinal lie—consistent with prior research (Conway et al., 2008) and Studies 1 and 2 of the present package—levels of dialectical and elaborative complexity were dissimilar (dialectical M = 1.18, elaborative M = 1.65). However, when telling a storytelling lie, levels of dialectical and elaborative complexity were similar (dialectical M = 1.26, elaborative M = 1.36). See Figure 3.

Discussion

The primary purpose of Study 3 was to compare the effects of telling a counterattitudinal lie versus a storytelling lie on complexity levels. While findings for the effects of counterattitudinal lies on complexity levels were consistent with prior inquiries (and the strategic model) in which participants tended to use more elaborative complexity and less dialectical complexity when telling a counterattitudinal lie to an individual, storytelling lies resulted in a different pattern of complexity use. When participants told a storytelling lie, they tended to use similarly low levels of both dialectical and elaborative complexity. This is consistent with the idea that storytellers would have little to strategically gain by using either dialectical or elaborative complexity when they lie—producing generally low levels of both.

General Discussion

Utilizing tools only fairly recently available for scientific use (Conway et al., 2008; Conway, Dodds, et al., 2011) to examine the lying–complexity relationship in a more nuanced way, the present set of results overwhelmingly support the value of a strategic model of the effect of lying on complex language. Study 1 showed that, consistent with the strategic model, counterattitudinal lying results in decreased dialectical complexity and increased elaborative complexity regardless of whether the lie is told in a low-importance scenario (artificial lab setting) or high-importance scenario (presidential debate). Study 2 revealed two different patterns of elaborative and dialectical complexity use for lying to two different audience types, suggesting complexity is altered according to what strategy would make the most effective lie under the circumstances. Study 3 demonstrated that a completely separate distinction between storytelling lies and counterattitudinal lies similarly produced a nuanced pattern of elaborative and dialectical complexity that is consistent with the strategic model.

Although these results provide very different tests in very different contexts, they all support the basic premise of the strategic model that liars alter the complexity of their language to accomplish a specific strategic goal. These results provide more direct evidence—evidence that extends beyond prior work (e.g., Conway et al., 2008)—that impression management goals such as those that guide other areas of social cognition (see, e.g., Schaller & Conway, 1999, 2001) also play a key role in explaining how liars translate their goals into language.

Is the Cognitive Strain Theory Wrong?

Although the present results suggest strategic concerns affect complexity, this does not mean that cognitive strain plays no role in affecting the complexity of lies. Indeed, a long history of research and theory in psychology suggests that personal cognitive factors such as cognitive strain do affect the complexity of language (see, e.g., Conway et al., 2001; Conway, Gornick, et al., 2011; Suedfeld, 1992, 2010; Tetlock & Tyler, 1996; Thoemmes & Conway, 2007).⁶

In fact, although they sometimes make different predictions, the two theories are not mutually exclusive. They might be reconciled in multiple ways. First, it is possible that only certain types of lies produce extra cognitive strain—perhaps those that are both important and completely unrehearsed, or those that have to be maintained over a long period of time. This would be consistent with the argument made by Levine and McCornack (2014) that on average lies may be no more cognitively straining than truths, and at times, they may even be less cognitively straining. In other words, it may be the case that lying in general is not particularly cognitively draining—and that is why we do not find consistent evidence of a lies-reduce-complexity effect. But this also suggests that there might be a number of circumstances where lying would be draining, and in those contexts, it would reduce the complexity of the lie.

Second and on the other hand, evidence from other areas suggests that lying is associated with markers of heightened stress such as pupil dilation (e.g., Zuckerman, DePaulo, & Rosenthal, 1981). Thus, it is possible that the cognitive stress associated with lying does reduce linguistic complexity on average, but that sometimes strategic concerns can override that stress in certain circumstances. This integrative approach would suggest that, in the absence of clear strategic reasons for increasing any form of complexity, lying would generally reduce complexity. Thus, one interpretation of Study 3 would be that the low scores for both dialectical and elaborative complexity for storytelling liars represents the general effect of cognitive strain in the absence of a clear complexity-related strategic goal. This would be consistent with other work showing storytelling liars have lower complexity than truth-tellers (Newman et al., 2003).

Third and relatedly, it is also possible that cognitive strain operates as an additive effect above and beyond strategic concerns. After all, the existence of a nuanced interaction effect does not preclude the existence of a main effect (see, e.g., Conway, Houck, Gornick, & Repke, 2016). It is possible that there is a general overall reduction of cognitive strain that can be masked by studies evaluating strategic effects. If this were true, one would expect that, while there would be variability in the effect direction for overall summary complexity measurements, there would nonetheless be a general downward trend such that the cumulative relationship between lying and complexity is negative.⁷

It is important to note that the present studies were largely not set up to definitely argue that the cognitive strain model never operates on the lying–complexity relationship. Rather, they were set up to provide evidence that the strategic model is sometimes in operation in ways not easily predicted by simpler cognitive strain accounts. Whether any of the three above integrative approaches proves to be correct—or none of them—that does not undermine the primary value of our data. That value is to illustrate more definitively that truly strategic concerns affect the lying–complexity relationship across multiple contexts.

Concluding Thoughts: Personal Cognition Versus Strategic Self-Presentation

The dichotomy between personal cognition and strategic self-presentation is a fuzzy one. After all, strategic concerns are still the function of personal cognitions; and cognitive strain is often produced in a social context. However, the degree that statements are motivated by a desire to impress others versus some other more personal cognition is nonetheless a useful distinction—a distinction with many important practical implications. One of those implications involves the topic of the present article—the relationship between lying and complexity. If liars are using linguistic complexity to accomplish a specific strategic goal, then, their resulting language will look quite different than if something more private is going on behind the scenes. The present results, along with other work on strategically motivated use or interpretation of complexity (e.g., Anolli et al., 2003; Conway, Gornick, et al., 2016), suggest that scientists should more fully consider the strategic goals of language when evaluating what influences it.

Footnotes

Acknowledgements

We would like to thank the students who conducted the complexity coding of the statements under investigation here. We would also like to thank the anonymous reviewers and Howie Giles for their constructive comments on this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Meredith A. Repke is an experimental psychology PhD candidate at the University of Montana. She researches attitudes, behaviors, and communication relevant to the natural environment, well-being, and political orientation. Her research has been published in multiple peer-reviewed journals and featured in numerous The Huffington Post articles.

Lucian Gideon Conway III, received his PhD from the University of British Columbia in 2001 and is currently a full professor of psychology at the University of Montana. His primary research interests revolve around (a) how shared cultural beliefs emerge, persist, and have influence and (b) the causes of complex (as opposed to simple) thinking and the subsequent consequences on decision making in political and social arenas. He is the author of over 50 articles, commentaries, and book chapters in these areas, and his work has been featured in major media outlets such as The Washington Post, USA Today, and BBC Radio.

Shannon C. Houck is a faculty member at Syracuse University. She researches social and moral decision making, dynamics of peace and conflict, and the causes and consequences of complex (vs. simple) thinking.

References

Abe

J. A. A.

(2011). Changes in Alan Greenspan’s language use across the economic cycle: A text analysis of his testimonies and speeches. Journal of Language and Social Psychology, 30, 212-223.

Abe

J. A. A.

(2012). Cognitive–affective styles associated with position on war. Journal of Language and Social Psychology, 31, 212-222.

Anolli

Balconi

Ciceri

(2003). Linguistic styles in deceptive communication: Dubitative ambiguity and elliptic eluding in packaged lies. Social Behavior and Personality, 31, 687-710.

Baker-Brown

Ballard

E. J.

Bluck

DeVries

Suedfeld

Tetlock

P. E.

(1992). The integrative complexity scoring manual. In Smith

(Ed.), Motivation and personality: Handbook of thematic content analysis (pp. 605-611). New York, NY: Cambridge University Press.

Conway

L. G.

III Conway

K. R.

(2011). The terrorist rhetorical style and its consequences for understanding terrorist violence. Dynamics of Asymmetric Conflict, 4, 175-192.

Conway

L. G.

III Dodds

Towgood

K. H.

McClure

Olson

(2011). The biological roots of complex thinking: Are heritable attitudes more complex? Journal of Personality, 79, 101-134.

Conway

L. G.

III Gornick

L. J.

Burfiend

Mandella

Kuenzli

Houck

S. C.

Fullerton

D. T.

(2012). Does simple rhetoric win elections? An integrative complexity analysis of U.S. presidential campaigns. Political Psychology, 33, 599-618.

Conway

L. G.

III Gornick

L. J.

Houck

S. C.

Anderson

Stockert

Sessoms

McCue

(2016). Are conservatives really more simple-minded than liberals? The domain specificity of complex thinking. Political Psychology, 37, 777-798.

Conway

L. G.

III Gornick

L. J.

Houck

S. C.

Hands Towgood

Conway

K. R.

(2011). The hidden implications of radical group rhetoric: Integrative complexity and terrorism. Dynamics of Asymmetric Conflict, 4, 155-165.

10.

Conway

L. G.

III Houck

S. C.

Gornick

L. J.

Repke

M. A.

(2016). Ideologically motivated perceptions of complexity: Believing those who agree with you are more complex than they are. Journal of Language and Social Psychology, 35, 708-718.

11.

Conway

L. G.

III Suedfeld

Tetlock

P. E.

(2001). Integrative complexity and political decisions that lead to war or peace. In Christie

D. J.

(Ed.), Peace, conflict, and violence: Peace psychology for the 21st century (pp. 66-75). Upper Saddle River, NJ: Prentice Hall.

12.

Conway

L. G.

III Thoemmes

Allison

A. M.

Hands Towgood

Wagner

Davey

. . .Conway

K. R.

(2008). Two ways to be complex and why they matter: Implications for attitude strength and lying. Journal of Personality and Social Psychology, 95, 1029-1044.

13.

Davidson

Livingstone

McArthur

Dickson

Gumley

(2007). An integrative complexity analysis of cognitive behaviour therapy sessions for borderline personality disorder. Psychology and Psychotherapy: Theory, Research and Practice, 80, 513-523.

14.

Houck

S. C.

Conway

L. G.

III Gornick

L. J.

(2014). Automated integrative complexity: Current challenges and future directions. Political Psychology, 35, 647-659.

15.

Levine

T. R.

McCornack

S. A.

(2014). Theorizing about deception. Journal of Language and Social Psychology, 33, 431-440.

16.

McCornack

S. A.

Morrison

Paik

J. E.

Wisner

A. M.

Zhu

(2014). Information Manipulation Theory 2: A propositional theory of deceptive discourse production. Journal of Language and Social Psychology, 33, 348-377.

17.

Newman

Pennebaker

Berry

Richards

(2003). Lying words: Predicting deception from linguistic styles. Personality and Social Psychology Bulletin, 29, 665-675.

18.

Nixon

R. M.

(1962). Six crises. London, England: WH Allen.

19.

Pennebaker

J. W.

(2011). Using computer analyses to identify language style and aggressive intent: The secret life of function words. Dynamics of Asymmetric Conflict, 4, 92-102.

20.

Richards

J. M.

Gross

J. J.

(1999). Composure at any cost? The cognitive consequences of emotion suppression. Personality and Social Psychology Bulletin, 25, 1033-1044.

21.

Richards

J. M.

Gross

J. J.

(2000). Emotion regulation and memory: The cognitive costs of keeping one’s cool. Journal of Personality and Social Psychology, 79, 410-424.

22.

Schaller

Conway

L. G.

III . (1999). Influence of impression-management goals on the emerging contents of group stereotypes: Support for a social-evolutionary process. Personality and Social Psychology Bulletin, 25, 819-833.

23.

Schaller

Conway

L. G.

III . (2001). From cognition to culture: The origins of stereotypes that really matter. In Moscowitz

G. B.

(Ed.), Cognitive social psychology: The Princeton symposium on the legacy and future of social cognition (pp. 163-176). Mahwah, NJ: Lawrence Erlbaum.

24.

Suedfeld

(1992). Cognitive managers and their critics. Political Psychology, 13, 435-453.

25.

Suedfeld

(2010). The cognitive processing of politics and politicians: Archival studies of conceptual and integrative complexity. Journal of Personality, 78, 1669-1702.

26.

Suedfeld

Conway

L. G.

III Eichhorn

(2001). Studying Canadian leaders at a distance. In Feldman

Valenty

(Eds.), Political leadership for the new century: Lessons from the cross-cultural study of personality and behavior (pp. 3-19). Westport, CT: Greenwood.

27.

Suedfeld

Rank

D. A.

(1976). Revolutionary leaders: Long-term success as a function of changes in conceptual complexity. Journal of Personality and Social Psychology, 34, 169-178.

28.

Suedfeld

Tetlock

P. E.

Ramirez

(1977). War, peace, and integrative complexity: UN speeches on the Middle East problem, 1947-1976. Journal of Conflict Resolution, 21, 427-442.

29.

Suedfeld

Tetlock

P. E.

Streufert

(1992). Conceptual/integrative complexity. In Smith

C. P.

(Ed.), Motivation and personality: Handbook of thematic content analysis (pp. 393-400). New York, NY: Cambridge University Press.

30.

Tetlock

P. E.

(1985). Integrative complexity of American and Soviet foreign policy rhetoric: A time-series analysis. Journal of Personality and Social Psychology, 49, 1565-1585.

31.

Tetlock

P. E.

Hannum

K. A.

Micheletti

P. M.

(1984). Stability and change in the complexity of senatorial debate: Testing the cognitive versus rhetorical style hypotheses. Journal of Personality and Social Psychology, 46, 979-990.

32.

Tetlock

P. E.

Manstead

A. S. R.

(1985). Impression management versus intrapsychic explanations. Psychological Review, 92, 59-77.

33.

Tetlock

P. E.

Metz

S. E.

Scott

S. E.

Suedfeld

(2014). Integrative complexity coding raises integratively complex issues. Political Psychology, 35, 625-634.

34.

Tetlock

P. E.

Tyler

(1996). Churchill’s cognitive and rhetorical style: The debates over Nazi intentions and self-government for India. Political Psychology, 17, 149-170.

35.

Thoemmes

Conway

L. G.

III . (2007). Integrative complexity of 41 U.S. presidents. Political Psychology, 28, 193-226.

36.

Zuckerman

DePaulo

B. M.

Rosenthal

(1981). Verbal and nonverbal communication in deception. In Berkowitz

(Ed.), Advances in experimental social psychology (Vol. 14, pp. 1-59). New York, NY: Academic Press.