Frankly, We Do Give a Damn

Abstract

There are two conflicting perspectives regarding the relationship between profanity and dishonesty. These two forms of norm-violating behavior share common causes and are often considered to be positively related. On the other hand, however, profanity is often used to express one’s genuine feelings and could therefore be negatively related to dishonesty. In three studies, we explored the relationship between profanity and honesty. We examined profanity and honesty first with profanity behavior and lying on a scale in the lab (Study 1; N = 276), then with a linguistic analysis of real-life social interactions on Facebook (Study 2; N = 73,789), and finally with profanity and integrity indexes for the aggregate level of U.S. states (Study 3; N = 50 states). We found a consistent positive relationship between profanity and honesty; profanity was associated with less lying and deception at the individual level and with higher integrity at the society level.

Frankly my dear, I don’t give a damn.

Gone with the Wind (1939)

Profane as it is, this memorable line by the character Rhett Butler in the film Gone with the Wind profoundly conveys Butler’s honest thoughts and feelings. However, it was the use of this profane word that led to a US$5,000 fine against the film’s production for violating the Motion Picture Production Code. This example reveals the conflicting attitudes that most societies hold toward profanity, reflected in a heated debate taking place in online forums and media in recent years—with passionate views on both sides. For example, the website debate.org, which conducts online polls and elicits general public opinions on popular online debates, has many comments on the issue, with a 50–50 tie between the two views (Are people who swear more honest?, 2015). This public debate reflects an interesting question and mirrors the academic discussion regarding the nature of profanity. On the one hand, profane individuals are widely perceived as violating moral and social codes and thus deemed untrustworthy and potentially antisocial and dishonest (Jay, 2009). On the other hand, profane language is considered as more authentic and unfiltered, thus making its users appear more honest and genuine (Jay, 2000). These opposing views on profanity raise the question of whether profane individuals tend to be more or less dishonest.

Profanity

Profanity refers to obscene language including taboo and swear words, which in regular social settings are considered inappropriate and in some situations unacceptable. It often includes sexual references, blasphemy, objects eliciting disgust, ethnic–racial–gender slurs, vulgar terms, or offensive slang (Mabry, 1974). The interest in understanding the psychological roots of the use of profanity dates back to as far as the early 20th century (Patrick, 1901), yet the literature in this domain is scattered across different scientific fields with only recent attempts to connect the findings into a unified framework (Jay, 2009).

The reasons for using profanity depend on the person and the situation, yet profanity is commonly related to the expression of emotions such as anger, frustration, or surprise (Jay & Janschewitz, 2008). The spontaneous use of profanity is usually the unfiltered genuine expression of emotions, with the most extreme type being the bursts of profanity (i.e., coprolalia) accompanying the Tourette syndrome (Cavanna & Rickards, 2013). The more controllable use of profanity often helps to convey world views or internal states or is used to insult an object, a view, or a person (Jay, 2009). Speech involving profane words has a stronger impact on people than regular speech and has been shown to be processed on a deeper level in people’s minds (Jay, Caldwell-Harris, & King, 2008).

The context is important for understanding profanity. Profanity can sometimes be interpreted as antisocial, harmful, and abusive—if, for example, intended to harm or convey aggression and hostile emotions (Stone, McMillan, & Hazelton, 2015). It also violates the moral foundations of purity (Sylwester & Purver, 2015) and the common norm for speech, suggestive of the potential to engage in other antisocial behaviors that violate norms and morality. However, profanity may also be seen as a positive if it does not inflict harm but acts as a reliever of stress or pain in a cathartic effect (Robbins et al., 2011; Vingerhoets, Bylsma, & de Vlam, 2013). Profane language can serve as a substitute for potentially more harmful forms of violence (Jay, 2009) and can alert others to one’s own emotional state or the issues that one cares about deeply (Jay, 2009). Profanity is also used to entertain, attract (Kaye & Sapolsky, 2009), and influence audiences (Scherer & Sagarin, 2006) as illustrated by the frequent use of profane language in comedy, mass media, and advertising (Sapolsky & Kaye, 2005). Profanity has even been used by presidential candidates in American elections (Slatcher, Chung, Pennebaker, & Stone, 2007) as recently illustrated by Donald Trump, who has been both hailed for authenticity and criticized for moral bankruptcy (Sopan, 2015).

Dishonesty

In its most basic form, dishonesty involves the conscious attempt by a person to convince others of a false reality (Abe, 2011). In this work, we operationalize dishonesty as a generalized personal inclination to obscure the truth in natural, everyday life situations. The most common type of such dishonesty is represented by “white lies” or “social lies” that people tell themselves or others in order to appear more desirable or positive (DePaulo, Kashy, Kirkendol, Wyer, & Epstein, 1996; Granhag & Vrij, 2005). While most people claim to be honest most of the time (Aquino & Reed, 2002; Halevy, Shalvi, & Verschuere, 2013), research suggests that minor cases of dishonesty are quite common (DePaulo & Kashy, 1998; Hofmann, Wisneski, Brandt, & Skitka, 2014; Serota, Levine, & Boster, 2010), especially when people believe that dishonesty is harmless or justifiable (Fang & Casadevall, 2013) or that they can avoid any penalties (Gino, Ayal, & Ariely, 2009). In other words, people tend to rationalize their own dishonesty (Ayal & Gino, 2012) and perceive it as less severe (Peer, Acquisti, & Shalvi, 2014) or nonexistent (Mazar, Amir, & Ariely, 2008).

The Relationship Between Profanity and Dishonesty

There are two opposing perspectives on the relationship between profanity and dishonesty. As dishonesty and profanity are both considered deviant (Bennett & Robinson, 2000) and immoral (Buchtel et al., 2015), they are generally perceived as a reflection of a disregard for societal normative expectations (Kaplan, 1975), low moral standards, lack of self-control, or negative emotions (Jay, 1992, 2000). In this regard, profanity appears to be positively related to dishonesty, explaining why people who swear are perceived as untrustworthy (Jay, 1992) and why swear words are often associated with deceit (Rassin & Van Der Heijden, 2005). Previous work has also linked the use of swear words to the dark triad personality traits—namely, narcissism, Machiavellianism, and psychopathy—all indicative of social deviance and a higher propensity for dishonesty (Holtzman, Vazire, & Mehl, 2010; Sumner, Byers, Boochever, & Park, 2012). Swearing has also been shown to hold a negative relationship with the personality traits of conscientiousness and agreeableness, which are considered the more socially aware and moral aspects of personality (Kalshoven, Den Hartog, & De Hoogh, 2011; Mehl, Gosling, & Pennebaker, 2006; Walumbwa & Schaubroeck, 2009).

On the other hand, profanity can be positively associated with honesty. It is often used to express one’s unfiltered feelings (e.g., anger, frustration) and sincerity. Innocent suspects, for example, are more likely to use swear words than guilty suspects when denying accusations (Inbau, Reid, Buckley, & Jayne, 2011). Accordingly, people perceive testimonies containing swear words as more credible (Rassin & Van Der Heijden, 2005).

The Present Investigation

This work explores the relationship between profanity and honesty to address the paradoxical perspectives in the existing literature. Study 1 examined the relationship between profanity use and honesty on a lie scale. Study 2 examined behavior in real-life naturalistic setting by analyzing behavior on Facebook: looking at the relationship between users’ profanity rate and honesty in their online status updates, as indicated by a linguistic detection of deception. Study 3 extended to society level by exploring the relationship between state-level profanity rates and state-level integrity. The Online Supplemental Materials include power analyses, procedures, and stimuli used in the three studies, and data and code were made available on the Open Science Framework (https://osf.io/z9jbm/).

Study 1—Honesty on a Lie Scale

We began our investigation with a test for the relationship between profanity and honesty, captured by a widely used lie scale.

Method

Participants and Procedure

A total of 307 participants were recruited online using Amazon Mechanical Turk. Of the sample, 31 participants failed attention checks (10%) and were excluded from the analysis, leaving a sample of 276 (M _age = 40.71, SD _age = 12.75; 171 females). The exclusion of participants had no significant impact on the reported effect sizes or p values below. Participants self-reported profanity use in everyday life: given the opportunity to use profanity, rated reasons for the use of profanity, and answered a lie scale.

Measures

Profanity use behavioral measure

In 2 items, participants were asked to list their most commonly used and favorite profanity words: “Please list the curse words you [1 – use; 2 – like] the most (feel free, don’t hold back).” By giving participants an opportunity to curse freely, we expected that the daily usage and enjoyment of profanity would be reflected in the total number of curse words written. Participants’ written profanity was counted and coded by the first author and a coder unrelated to the project, who was unaware of the study hypotheses and data structure. The interrater reliability was .91 (95% confidence interval [CI] [.87, .94]) for most commonly used curse words and .93 (95% CI [.91, .97]) for favorite curse words, indicating a very high level of agreement.

Profanity self-reported use

To supplement the behavioral measures, we also added self-reported use of profanity. Participants self-reported their everyday use of profanity (Rassin & Muris, 2005) using 3 items: “How often do you curse (swear/use bad language)” (1) “verbally in person (face to face),” (2) “in private (no one around),” and (3) “in writing (e.g., texting/messaging/posting online/emailing”; 1 = never, 2 = once a year or less, 3 = several times a year, 4 = once a month, 5 = 2–3 times a month, 6 = once a week, 7 = 2–3 times a week, 8 = 4–6 times a week, 9 = daily, 10 = a few times a day; α = .84).

Reasons for profanity use

Following Rassin and Muris (2005), we also asked participants to rate reasons for their use of profanity (0 = never a reason for me to swear; 5 = very often a reason for me to swear) and asked questions regarding the general perceived reasons for using profanity (0 = not at all; 5 = to a very large extent; see Online Supplemental Materials).

Honesty

Honesty was measured using the Lie subscale of the Eysenck Personality Questionnaire Revised short scale (Eysenck, Eysenck, & Barrett, 1985). The Lie subscale is one of the most common measures for assessing individual differences in lying for socially desirable responding (Paulhus, 1991). The Lie scale includes 12 items, such as “If you say you will do something, do you always keep your promise no matter how inconvenient it might be?” and “Are all your habits good and desirable ones?” (dichotomous Yes/No scale). In these examples, positive answers are considered unrealistic and therefore most likely a lie (α = .79). The Lie scale was reversed for the honesty measure.

Results

The means, standard deviations, and correlations for the honesty and profanity measures are detailed in Table 1. Honesty was positively correlated with all profanity measures, meaning that participants lied less on the Lie scale if they wrote down a higher number of frequently used (r = .20, p = .001; CI [.08, .31]) and liked curse words (r = .13, p = .032; CI [.01, .24]) or self-reported higher profanity use in their everyday lives (r = .34, p < .001; CI [.23, .44]), even when controlling for age and gender (Behavioral 1: partial r = .20, p = .001; CI [.08, .31]; Behavior 2: partial r = .12, p = .049; CI [.001, .24]; self-report: partial r = .32, p < .001; CI [.21, .42]).

Table 1.

Study 1: Means, Standard Deviations, and Correlations for Variables.

Variables	Mean	SD	1	2	3	4	5
1. Honesty	7.63	3.00	(.79)
2. Profanity self-report	6.51	2.56	.34***	(.84)
3. Profanity behavioral 1	4.09	2.61	.20**	.46***	(—)
4. Profanity behavioral 2	1.60	1.62	.13*	.41***	.45***	(—)
5. Age	40.71	12.75	−.13*	−.34***	−.05	−.08	(—)
6. Gender	1.62	0.49	−.06	−.03	−.07	−.04	.08

Note. N = 276. Gender coding: 1 = male, 2 = female. Scale α coefficients are on the diagonal. Profanity behavioral 1 = number of most frequently used curse words written; profanity behavioral 2 = number of most liked curse words written.

*p < .05. **p < .01. ***p < .001.

We asked participants to rate their reasons for use of profanity. The reasons that received the highest ratings were the expression of negative emotions (M = 4.09, SD = 1.33), habit (M = 3.08, SD = 1.82), and an expression of true self (M = 2.17, SD = 1.73). Participants also indicated that in their personal experience, profanity was used for being more honest about their feelings (M = 2.69, SD = 1.72) and dealing with their negative emotions (M = 2.57, SD = 1.64). Profanity received a lower rating as a tool for insulting others (M = 1.41, SD = 1.53) as well as for being perceived as intimidating or insulting (M = 1.12, SD = 1.36). This supports the view that people regard profanity more as a tool for the expression of their genuine emotions rather than being antisocial and harmful.

Study 2—Naturalistic Deceptive Behavior on Facebook

Study 1 provided initial support for a positive relationship between profanity use and honesty, with the limitations of lab settings. Study 2 was constructed to extend Study 1 to a naturalistic setting—using a larger sample, more accurate measures of real-life use of profanity, and a different honesty measure.

With a stellar growth, Facebook has become the world’s most dominant social network and is strongly embedded in its users’ overall social lives (Manago, Taylor, & Greenfield, 2012; Wilson, Gosling, & Graham, 2012). Online social networking sites such as Facebook now serve as an extension of real-life social context, allowing individuals to express their actual selves (Back et al., 2010). Facebook profiles have been found to provide fairly accurate portrayals of their users’ personalities and behaviors (Kosinski, Stillwell, & Graepel, 2013; Schwartz et al., 2013; Wang, Kosinski, Stillwell, & Rust, 2012), including socially undesirable aspects (Garcia & Sikström, 2014), such as self-promotion (Waggoner, Smith, & Collins, 2009; Weisbuch, Ivcevic, & Ambady, 2009), narcissism (Buffardi & Campbell, 2008), and low self-esteem (Zywica & Danowski, 2008).

In this work, we detected dishonesty by analyzing Facebook users’ status updates that were used to broadcast messages to their online social network. Using language to tap into people’s psyches dates back to Freud (1901), who analyzed patients’ slips of the tongue, and Lacan (1968), who argued that the unconscious manifests itself in language use. A growing body of literature has since demonstrated that the language that people use in their daily lives can reveal hidden aspects of their personalities, cognitions, and behaviors (Pennebaker, Mehl, & Niederhoffer, 2003). The linguistic approach is especially useful in the case of dishonesty, which—though prevalent—is frowned upon when detected, and therefore leads those who are acting dishonestly to try to hide it from others (Hancock, 2009; Toma, Hancock, & Ellison, 2008). In the case of Facebook, the dishonesty we refer to is not necessarily blunt deception aimed at exploiting or harming others but rather a mild distortion of the truth intended to construe a more socially desirable appearance (Whitty, 2002; Whitty & Gavin, 2001).

Method

Participants and Procedure

A total of 153,716 participants were recruited using the myPersonality Facebook application (Kosinski, Matz, Gosling, Popov, & Stillwell, 2015). Participants voluntarily chose to use this application and provided opt-in consent to record their Facebook profiles, including demographic data and their status updates (more information about the myPersonality Facebook application is available at http://mypersonality.org). This analysis is limited to users who used the English version of Facebook, had more than 50 Facebook status updates, and had more than 30 friends (an indication of being an active Facebook user). The final sample included 73,789 participants (62.0% female, M _age = 25.34, M _{network size} = 272.37; M _{status updates} = 201.28, SD _{status updates} = 167.33; M _words = 3,181.82, SD _words = 3,014.44).

Measures

We used Linguistic Inquiry and Word Count (LIWC Version 2007; Tausczik & Pennebaker, 2010) in order to analyze participants’ status updates. The analysis was conducted by aggregating all the status updates of every participant into a single file and executing a LIWC analysis on each user’s combined status updates. The LIWC software reported the percentages of the words in each LIWC category out of all of the words used in the combined status updates, as follows:

$$\text{LIWC category rate user }X=\frac{\text{User }X\text{ word count for LIWC category in all status updates}}{\text{User}X\text{ word count in all status updates}}.$$

Honesty

The honesty of the status updates written by the participants was assessed following the approach introduced by Newman, Pennebaker, Berry, and Richards (2003) using LIWC. Their analyses showed that liars use fewer first-person pronouns (e.g., I, me), fewer third-person pronouns (e.g., she, their), fewer exclusive words (e.g., but, exclude), more motion verbs (e.g., arrive, go), and more negative words (e.g., worried, fearful; Newman, Pennebaker, Berry, & Richards, 2003). The explanation was that dishonest people subconsciously try to (1) dissociate themselves from the lie and therefore refrain from referring to themselves, (2) prefer concrete over abstract language when referring to others (using someone’s name instead of “he” or “she”), (3) are likely to feel discomfort by lying and therefore express more negative feelings, and (4) require more mental resources to obscure the lie and therefore end up using less cognitively demanding language, which is characterized by a lower frequency of exclusive words and a higher frequency of motion verbs. Equation and usage rates in this study are summarized in Table 2.

Table 2.

Study 2: Word Analysis of LIWC Categories and Key Words.

LIWC Dimensions	Sample LIWC Key Words	Honesty Coefficients βs	Percentage (M, %)	Percentage (SD, %)
First-person pronouns	I, me, mine	.260	4.21	1.71
Third-person pronouns	She, her, him, they, their	.250	0.84	0.33
Exclusive words	But, without, exclude	.419	1.78	0.63
Motion verbs	Arrive, car, go	−.259	1.57	0.53
Anxiety words	Worried, fearful, nervous	−.217	0.21	0.14

Note. LIWC = Linguistic Inquiry and Word Count.

Newman et al. (2003) achieved up to 67% accuracy when detecting lies, which was significantly higher than the 52% near-chance accuracy achieved by human judges. Their approach has been successfully applied to behavioral data (Slatcher et al., 2007) and to Facebook status updates (Feldman, Chao, Farh, & Bardi, 2015). Other studies have since found support for these LIWC dimensions as being indicative of lying and dishonesty (Bond & Lee, 2005; Hancock, Curry, Goorha, & Woodworth, 2007; see meta-analyses by DePaulo et al., 2003 and Hauch, Masip, Blandón-Gitlin, & Sporer, 2012).

To calculate the honesty score, we first computed LIWC scores to obtain participants’ use rate of first-person pronouns, third-person pronouns, exclusive words, motion verbs, and anxiety words and then applied average regression coefficients from Newman et al. (2003). Here, we note that we focused on anxiety words rather than general negative words (which include anxiety, anger, and sadness) due to two considerations. First, it has been suggested that anxiety words may be more predictive of honesty than overall negative emotions (Newman et al., 2003). Second, measuring honesty using negative emotions with anger words may bias the profanity–honesty correlations because anger has been shown to have a strong positive relation with profanity. Holtzman et al. (2010) reported a correlation of .96 between anger and profanity, and Yarkoni (2010) found swearing to be strongly associated with anger but not with anxiety, which is not surprising given the conclusion by Jay and Janschewitz (2008) that profanity is mostly used to express anger.¹

Profanity

We used the LIWC dictionary of swear words (e.g., damn, piss, fuck) to obtain the participants’ use rate of profanity. This approach was previously used to analyze swearing patterns in social contexts (e.g., Holtgraves, 2011; Mehl & Pennebaker, 2003). Profanity use rates were calculated per each participant using LIWC, with rates indicating the percentage of swear words used in all status updates by the participant overall. Profanity use rates were then log-transformed to normalize distribution (ln[profanity + 1]).

Results

The descriptive statistics and zero-order correlations of all variables are provided in Table 3. The mean of profanity use was 0.37% (SD = 0.43%; 7,969 [10.8%] used no profanity at all), which is in line with previous findings (Jay, 2009). Profanity and honesty were found to be significantly and positively correlated (N = 73,789; r = .20, p < .001; 95% CI [.19, .21]; see Figure 1 for an aggregated plot), indicating that those who used more profanity were more honest in their Facebook status updates. Controlling for age, gender, and network size resulted in a slightly stronger effect (partial r = .22, p < .001; 95% CI [.21, .22]).

Table 3.

Study 2: Descriptive Statistics of Honesty, Profanity, and Demographics.

Variables	Mean	SD	Skewness	Kurtosis	Honesty	Profanity	Age	Gender
Honesty (raw)	0 (1.60)	1 (0.60)	0.03	0.02	(—)	.22
Profanity (raw)	0.28 (0.37)	0.26 (0.43)	1.37 (2.51)	2.00 (9.49)	.20	(—)
Age	25.34	8.78	1.90	3.96	−.05	−.18	(—)
Gender	0.62	0.49	−0.49	−1.76	.12	−.23	.08	(—)
Network size (raw)	5.30 (272.37)	0.79 (249.71)	−0.03 (4.18)	−0.25 (39.82)	.18	−.09	−.13	.00 (ns)

Note. Gender coding: 0 = male, 1 = female. ns indicates a nonsignificant correlation coefficient; remaining coefficients were significant at p < .001 level; honesty was standardized; profanity and network size were log transformed. Males used more profanity than females, d = .12 [0.12, 0.13], t(4,6884.67) = 59.26, p < .001, d = .47, and were less honest, d = −0.14 [−0.15, −0.13], t = −31.69, p < .001, d = −0.23. Raw lines indicate statistics for variables before transformations or standardizing. Values above the diagonal are partial correlations controlling for age, gender, and network size.

Figure 1.

Study 2: the relationship between profanity and honesty (Model 2). The first two scatterplots are of two randomly chosen 1% subsets of the total population (Plot 1: n = 750; Plot 2: n = 721). The third graph is a plot of aggregated honesty groups, and average profanity was computed for five equal groups of participants based on their honesty. The honesty score was standardized to the mean of 0 and standard deviation of 1. Error bars indicate a 95% confidence interval. The profanity rate is in percentages (e.g., 0.25 is 0.25% use).

Study 3—State-Level Integrity

Studies 1 and 2 demonstrated that the use of profanity is a predictor of honesty at the individual level. Study 3 sought to extend these findings by taking a broader view and examining the possible implications that individual differences in use of profanity have for society (as suggested by Back & Vazire, 2015). If the use of profanity is indeed positively related to honesty, then it can be argued that societies with higher profanity rates may be characterized by a higher appreciation for honesty and genuineness. Study 3 examined whether the state-level use of profanity is predictive of state-level integrity as reported by the State Integrity Index 2012.

Measures

State-level profanity

State-level profanity scores were computed by averaging the profanity scores of the American participants in Study 2 (29,701 participants) across the states. The state profanity scores are detailed in Table 4.

Table 4.

Study 3: State-Level Profanity and Integrity Rates.

State	Profanity Rate	Integrity	State	Profanity Rate	Integrity	State	Profanity Rate	Integrity
Alabama	34	72	Maine	33	56	Oregon	36	73
Alaska	42	68	Maryland	46	61	Pennsylvania	42	71
Arizona	41	68	Massachusetts	46	74	Rhode Island	44	74
Arkansas	29	68	Michigan	41	58	South Carolina	29	57
California	44	81	Minnesota	39	69	South Dakota	38	50
Colorado	39	67	Mississippi	33	79	Tennessee	32	76
Connecticut	52	86	Missouri	37	72	Texas	38	68
Delaware	51	70	Montana	35	68	Utah	26	65
Florida	41	71	Nebraska	42	80	Vermont	35	69
Georgia	36	49	Nevada	47	60	Virginia	40	55
Hawaii	45	74	New Hampshire	36	66	Washington	36	83
Idaho	31	61	New Jersey	50	87	West Virginia	34	68
Illinois	45	74	New Mexico	34	62	Wisconsin	39	70
Indiana	35	70	New York	46	65	Wyoming	34	52
Iowa	40	87	North Carolina	37	71
Kansas	39	75	North Dakota	37	58
Kentucky	37	71	Ohio	39	66
Louisiana	35	72	Oklahoma	33	64

Note. Integrity is the State Integrity Investigation 2012 index. Profanity rates were aggregated to the state level from the Study 2 Facebook profanity rates for American participants.

State-level integrity

State-level integrity was obtained from the State Integrity Investigation 2012 (SSI2012), the year that the myPersonality data collection was concluded. Estimating state levels of integrity and corruption is a complicated and controversial issue. For example, corruption was sometimes measured with the number of corruption convictions per state, yet a higher conviction rate can be indicative of better policing and thus lower corruption. We therefore used an index of integrity that is less affected by possible conflicting interpretations of crime and conviction statistics: the SSI2012. The SSI2012 ranks the states on 14 broad integrity criteria, including stance on honesty and transparency; the presence of independent ethics commissions; and executive, legislative, and judicial accountability. State integrity scores are detailed in Table 4. More information about how the State Integrity scores were obtained can be found in the Online Supplemental Materials.

Results

A scatterplot of profanity and integrity rates for all states is provided in Figure 2. We found a positive relationship between profanity and integrity on a state level (N = 50; r = .35, p = .014; CI [.08, .57]). States with a higher profanity rate had a higher integrity score.² For example, two of the three states with the highest profanity rate, Connecticut and New Jersey, were also two of the three states with the highest integrity scores on the index.

Figure 2.

Study 3: scatterplot presenting integrity and profanity rates across 50 U.S. states.

We also conducted a spatial regression analysis to address possible spatial-dependence regional confounds (Ward & Gleditsch, 2008). We calculated spatial distance matrices (Merryman, 2008) for the distance between states’ centroids using the following formula for Euclidean distance between State A and State B (y and x denote the y coordinate and x coordinate, respectively):

$$d(x_A,y_A;x_B,y_B)=\sqrt{{(y_A-y_B)}^2+{(x_A-x_B)}^2}.$$

We then inverted the distances (1/X) to form a proximity measure, multiplied the proximity matrix by the state profanity column, and divided by the sum to create a measure of spatial lag—a spatial weighted profanity per each state (Webster & Duffy, 2016). Excluding Hawaii and Alaska for their geographical isolation, the spatial profanity measure had a correlation of r = .55 with the state profanity measure (n = 48; p < .001; CI [.32, .72]; Moran I statistic = .15, p < .001), indicative of spatial dependence. After controlling for the spatial profanity, the partial correlation between profanity and integrity was r = .33 (p = .025, CI [.05, .56]).

General Discussion

We examined the relationship between the use of profanity and dishonesty and showed that profanity is positively correlated with honesty at an individual level and with integrity at a society level. Table 5 provides a summary of the results. Study 1 showed that participants with higher profanity use were more honest on a lie scale, and in Study 2, profanity was associated with more honest language patterns in Facebook status updates. In Study 3, state-level profane language usage was positively related to state-level integrity.

Table 5.

Summary of the Results.

#	Sample Size	Sample Type	Level of Analysis	Profanity Measure(s)	Honesty Measure	Effect
1	276	American English native MTurk workers	Individual	1–2: Counts of written profanity 3: Self-report	Eysenck Personality Questionnaire Revised short scale	.20/.13/.34
2	73,789	English version Facebook users	Individual	Rate of profanity in language used in status updates	Derivative of standard LIWC dimensions (Newman, Pennebaker, Berry, & Richard, 2003)	.20 (.22)
3	50 (48)	States in the United States	State	Average profanity in language used in status updates	State Integrity Investigation 2012 index	.35 (.33)

Note. Effects in parentheses are effects while controlling for other factors (Study 2: age, gender, and network size; Study 3: spatial distance). LIWC = Linguistic Inquiry and Word Count.

Challenges in Studying Profanity and Dishonesty in Naturalistic Settings

The empirical investigation of the relationship between dishonesty and profanity poses a unique challenge. The behavioral ethics literature has been successful in devising ways to examine unethical behavior in the lab, yet observing dishonesty and unethical behavior in the field remains an ongoing challenge, and so far only a few studies were able to devise innovative methods to overcome that challenge (e.g., Hofmann et al., 2014; Piff, Stancato, Côté, Mendoza-Denton, & Keltner, 2012). The indirect linguistic approach for the detection of dishonesty with an analysis of spoken and written language patterns paves the way for more behavioral ethics research on actual dishonest behavior in the field.

Unlike behavioral ethics, the study of profanity is still very much in its infancy (Jay, 2009). Profanity is a much harder construct to measure and even more difficult to effectively elicit or manipulate, whether it is in the lab or in the field. The relatively low use rates of profanity decrease even further when people know that they are observed or that their behavior is studied. Therefore, to be able to gain an understanding of profanity use, it is important that the behavior observed is genuine and in naturalistic settings. The current investigation has been able to address this challenge by applying a linguistic analysis approach to a unique large-scale naturalistic behavior data set.

The linguistic approach to detecting dishonesty used in Study 2 has been used and verified in a number of previous studies (e.g., Feldman et al., 2015; Slatcher et al., 2007). In Study 2, the linguistic analysis showed that men tended to be more dishonest than women, which is in line with a large body of literature presenting similar findings (Childs, 2012; Dreber & Johannesson, 2008; Friesen & Gangadharan, 2012). Also, those with larger networks had a higher likelihood for dishonesty and a lower likelihood for profanity, which supports the notion of dishonesty online as a means of creating a more socially desirable profile. Both findings contribute to the construct validity of the linguist honesty measure by demonstrating previously established nomological networks. The consistency in the direction and effect size of the profanity–honesty relationship across the three studies further raises confidence in this approach to measuring dishonesty.

Extending to Society Level

Our research offers a first look at the use of profanity at a society level. Using the large-scale sample of American participants from Study 2, we were able to calculate state-level rates of profanity for use in Study 3. Addressing calls for psychological research to attempt to examine the social implications of psychological findings (Back, 2015; Back & Vazire, 2015), we used this measure in order to examine whether the positive relationship between profanity and honesty found at the individual level could be extended to the society level. Such an attempt involves many challenges, as there are many variables that may intervene or offer competing explanations for a detected relationship. Yet we believe that this is an important first attempt to provide a baseline for further investigation. The consistent findings across the studies suggest that the positive relation between profanity and honesty is robust and that the relationship found at the individual level indeed translates to the society level.

Implications and Future Directions

We briefly note several limitations in the current research and these are further discussed in the Online Supplemental Materials with implications and future directions. First, the three studies were correlational, thus preventing us from drawing any causal conclusions. Second, the dishonesty we examined in Studies 1 and 2 was mainly about self-promoting deception to appear more desirable to others rather than blunt unethical behavior. We therefore caution that the findings should not be interpreted to mean that the more a person uses profanity, the less likely he or she will engage in more serious unethical or immoral behaviors. Third, the measures in Study 2 were proxies using an aggregation of linguistic analysis of online behavior using Facebook over a long period of time. Finally, Simpson’s Paradox (Simpson, 1951) points to conceptual and empirical differences in testing a relationship on different levels of analysis, and therefore the state-level findings of Study 3 are conceptually broader than the findings in Studies 1 and 2.

These limitations notwithstanding, our research is a first step in exploring the profanity–honesty relationship, and we believe that the consistent effect across samples, methodology, and levels of analysis contributes to our understanding of the two constructs and paves the way for future research. Future studies could build on our findings to further study the profanity–honesty relationship using experimental methods to establish causality and incorporating real-life behavioral measures with a wider range of dishonest conduct including unethical behavior.

Conclusion

We set out to provide an empirical answer to competing views regarding the relationship between profanity and honesty. In three studies, at both the individual and society level, we found that a higher rate of profanity use was associated with more honesty. This research makes several important contributions by taking a first step to examine profanity and honesty enacted in naturalistic settings, using large samples, and extending findings from the individual level to a look at the implications for society.

Author Biographies

Gilad Feldman is a postdoctoral research fellow with the work and social psychology department at Maastricht University. His research focuses on morality, personal values, lay-beliefs, and judgment decision-making.

Huiwen Lian is an associate professor of management at the Hong Kong University of Science and Technology. Her research focuses on leadership, abusive supervision, and deviant behavior.

Michal Kosinski is an assistant professor in organizational behavior at Stanford Graduate School of Business and his research focuses on individual differences in behavior, preferences, and performance.

David Stillwell is a lecturer in big data analytics and quantitative social sciences at Judge Business School University of Cambridge.

Handling Editor: Gregory Webster