The association between short-term emotion dynamics and cigarette dependence: a comprehensive examination of dynamic measures

Abstract

Background:

The association between short-term emotion dynamics and long-term psychopathology has been well established in the psychology literature. Yet, dynamic measures for inertia and instability of negative and positive affect have not been studied in terms of their association with cigarette dependence. This study builds an important bridge between the psychology and substance use literatures by introducing these novel measures and conducting a comprehensive examination of such association with intervention implications.

Methods:

This study conducted secondary analysis on the data from a community sample of 136 dual users (e-cigarette+cigarette) and 101 exclusive smokers who completed both the two-week ecological momentary assessment (EMA) and cigarette dependence assessments in a recent study.

Results:

Among dual users, a higher average level of negative affect, lower inertia of negative affect (i.e., less sustained negative affect), and higher instability of positive affect (i.e., greater magnitude of changes in positive affect) were associated with higher cigarette dependence. The patterns of associations among exclusive smokers were, however, different. Higher inertia of negative affect, lower instability of positive affect, and higher variability of negative affect were associated with higher dependence.

Conclusions:

The results illustrate the importance of examining not only negative affect but also positive affect in order to fully understand the association between emotion dynamics and cigarette dependence. The different patterns of association between emotion dynamics and cigarette dependence across the two groups of cigarette users also call for future research that is designed to compare cigarettes and e-cigarettes in terms of their effects on emotion regulation.

1. Introduction

Short-term fluctuations in psychological states and long-term developmental processes both lie at the heart of behavioral research (Sliwinski, 2008). Yet, they have mostly been investigated in separate studies. In the recent decade, a growing number of psychological studies have begun to examine them simultaneously. Houben et al. (2015) conducted a meta-analysis on 79 psychological papers that studied the association between short-term emotion dynamics (including variability, inertia, and instability) and psychological well-being (e.g., depression, externalizing behavior). In general, the literature indicated that high variability (a large variance of emotion over time), high inertia (high correlation of emotions over time), and high instability (large magnitude of emotional changes over time) are all associated with worse psychological well-being. Furthermore, the effect of short-term negative emotion dynamics on psychological well-being is stronger than the effect of short-term positive emotion dynamics. These results imply that short-term emotion dynamics may serve as a harbinger of longer-term psychopathology. Although the meta-analysis did not include studies with substance use related outcomes such as nicotine dependence, the implication is highly relevant to the field as some theories of addiction focus on the impact of affect (e.g., Baker et al., 2004). As more and more substance use studies conduct ecological momentary assessment (EMA) that collects behavior samples multiple times per day in the natural environment using widely accessible mobile technology (Buu et al., 2017), researchers are more able to capture and study short-term emotion dynamics. In fact, previous EMA studies have found within-person variability in emotion to be associated with the quantity of cigarette use (Bares et al., 2018), nicotine dependence (Piasecki et al., 2016), and progression from experimentation to higher smoking frequency (Weinstein et al., 2013).

In the field of nicotine and tobacco research, within-person mean and variance are commonly adopted to summarize EMA data (e.g., Sheets et al., 2015). However, the field has not yet examined two other measures of emotion dynamics developed in the psychology literature: inertia and instability. While inertia (or temporal dependency) can be measured straightforwardly by the first-order autocorrelation (i.e., the correlation between consecutive observations) with a higher value indicating that emotion is sustained over time and shows less homeostatic recovery, instability refers to the magnitude of moment-to-moment emotional changes and is a more complex characteristic with multiple indices proposed to measure it. The most popular index of instability is the mean square successive difference (MSSD) which is calculated by averaging the squared difference between consecutive observations. MSSD was shown to be a function of the variance and the first-order autocorrelation with a high value of instability corresponding to high variability and low inertia (Jahng et al., 2008). An alternative approach to assessing instability is to capture dramatic changes (i.e., exceeding a threshold) instead of average change (like the one captured by MSSD). Johns et al. (2019) adopted this approach to developing the fragmentation measures that involve two steps: the first step conducts participant-level normalization (by the mean and standard deviation) to facilitate the comparison across participants; the second step dichotomizes the normalized EMA scores according to whether the score is inside or outside the participant-specific standard range (defined as one standard deviation of the mean). After these two steps, the fragmentation measures can be calculated as:

$${\lambda }_o=\frac{n_o}{t_o}$$

$${\lambda }_s=\frac{n_s}{t_s}$$

where n_o is the number of transitions from outside to inside the standard range; t_o is the number of assessments with a score outside the standard range; n_s is the number of transitions from inside to outside the standard range; and t_s is the number of assessments with a score inside the standard range. Based on the above calculations, λ_o indicates the conditional probability of transitioning to inside the standard range, given being outside, whereas λ_s measures the conditional probability of transitioning from inside to outside. Although the capacity of these novel indices to differentiate mood disorder subtypes (Bipolar I, Bipolar II, and major depressive disorder) has been demonstrated (Johns et al., 2019), the potential applications to nicotine and tobacco research have not yet been explored.

In the U.S. adult population, about 21% (51 millions) reported smoking in past month; among those smokers, 18% (9 millions) were dual users of combustible and electronic cigarettes, according to our secondary analysis of the publicly accessible Wave 4 data (2016–2018) from the Population Assessment of Tobacco and Health (PATH) Study (USDHHS, 2019). Limited literature of nicotine and tobacco research suggests that in comparison to exclusive smokers, dual users are more likely to be white, younger, have more than a high school education, report a psychiatric history (especially depression and/ or attention-deficit hyperactivity disorder), smoke fewer cigarettes per day (Piper et al., 2019), have greater smoking expectancy on negative affect reduction (Peltier et al., 2019), and have higher levels of substance abuse/problems (Leventhal et al., 2016). Interestingly, an EMA study (Jorenby et al., 2017) demonstrated a larger jump in negative affect due to attempted smoking cessation among exclusive smokers compared to dual users, when the female participants had higher cigarette dependence. Such a group difference was, however, reversed for the women with low dependence. This implies that the association between emotion dynamics and cigarette dependence may differ across the two groups.

This study addresses the current knowledge gaps by conducting secondary analysis on EMA data collected from a recent study (Piper et al., 2020) to examine the association between short-term emotion dynamics and cigarette dependence among a community sample of dual users and exclusive smokers. Specifically, this study explores whether specific emotion dynamics are related to cigarette dependence and specific facets of cigarette dependence (i.e., do smokers with higher levels of dependence also have higher levels of affective variability, higher levels of sustained negative or positive affect, or greater magnitude of changes in negative affect). Such findings would suggest that short-term emotion dysregulation may serve as a harbinger of longer-term cigarette dependence. Furthermore, the patterns of association between emotion dynamics and cigarette dependence are compared across dual users and exclusive smokers.

2. Material and Methods

2.1. Study Sample and Procedures

The original study recruited 256 dual users and 166 exclusive smokers in Wisconsin via television and social media from 2015 to 2017 and conducted multi-wave assessments with 2-week EMA nested within wave over 2 years (see details in Piper et al., 2019). The study protocol was approved by the Institutional Review Board (IRB) of the University of Wisconsin (UW HS-IRB # 2014–0093). Eligibility criteria included: 1) at least 18 years old; 2) able to read and write English; 3) no plans to quit smoking and/or e-cigarette use in the next 30 days; 4) not currently taking smoking cessation medication; and 5) not currently in treatment for psychosis or bipolar disorder. Additionally, the dual users had to report using e-cigarettes at least once a week for the past month and smoking daily for the last 3 months, whereas the exclusive smokers had to report smoking at least 5 cigarettes per day for the past 6 months and not using e-cigarettes within the last 3 months. Participants were required to meet the minimum use criteria for combustible cigarettes described above but were not excluded for using other combustible tobacco products. The current study conducted secondary analysis on the baseline data from 136 dual users and 101 exclusive smokers who completed both the cigarette dependence assessments and the first 2-week EMA.

The indices for emotion dynamics developed in the psychology literature (Jahng et al., 2008; Johns et al., 2019) were adopted to characterize short-term changes in negative and positive affect captured by EMA. The relationship among various indices was examined comprehensively to inform their future applications in nicotine and tobacco research. Because a large number of potential covariates (6 indices × 2 affects) were considered simultaneously, a computationally efficient procedure for variable selection, the Least Absolute Shrinkage and Selection Operator (LASSO; Tibshirani, 1996), was adopted to select the covariates that were associated with cigarette dependence. Within each group (i.e., dual users vs. exclusive smokers), the associations between selected covariates and dependence were also estimated and tested for statistical significance while adjusting for smokers’ sociodemographic characteristics.

2.2. Dependence Assessments & EMA Items

The dual users and exclusive smokers both completed three cigarette dependence assessments: the Penn State Cigarette Dependence Index (PS-CDI; Foulds et al., 2015), the Fagerstrӧm Test of Cigarette Dependence (FTCD; Fagerstrӧm, 2012), and the Brief Wisconsin Inventory of Smoking Dependence Motives (WISDM; Smith et al., 2010). The WISDM has two major subscales: the Primary Dependence Motives (PDM) measure automaticity, loss of control, craving, and tolerance; whereas the Secondary Dependence Motives (SDM) capture affiliative attachment, cognitive enhancement, cure exposure/associative processes, social/environmental goads, taste, weight control, and affective enhancement.

During the 2-week EMA, the participants reported their mood, withdrawal symptoms, and substance use behaviors during at most two cigarette or e-cigarette use events and prior to going to bed per day through a study smartphone. They were trained to press a button marked with a “C’ every time they started a new cigarette and to press one marked “E” every time they vaped. At most two use events for each product per day were quasi-randomly selected with follow-up questions about the context of the event (with 20 items for about 3 minutes). Among the 237 participants in the analytic sample, the median number of days responding to EMA was 5 and the median number of events per day with EMA reports was 2. The analysis used all the available EMA data from each participant so no participants were excluded because of missing assessment. In this study, we used the reports on emotion during cigarette use events to capture emotion dynamics in both groups of cigarette users. The average of the following three items on a scale of 1 (strongly disagree) to 7 (strongly agree) was used to measure negative affect in the last 15 minutes: 1) feeling anxious, worried or stressed; 2) feeling angry or irritated; and 3) feeling sad or unhappy. Positive affect was measured by the mean of two items: excited; and enthusiastic (both on a scale of 1 to 7). These were all the items available in the parent study and have been consistently used in prior work (e.g., Kim et al., in press).

2.3. Indices of Emotion Dynamics

Four characteristics for short-term emotion dynamics were examined. Average level was measured by the within-person mean; variability was indexed by the variance; inertia was measured by the first-order autocorrelation; and instability was characterized by three indices: the MSSD (Jahng et al., 2008) and the two fragmentation measures (λ_o and λ_s; Johns et al., 2019).

2.4. Statistical Analysis

T test or Chi-square test were used to compare dual users and exclusive smokers on sociodemographic characteristics, EMA dynamic measures for negative and positive affect, and cigarette dependence measures. Pearson’s correlation coefficients were calculated to estimate the correlations among different indices of emotion dynamics as well as the correlations between the indices for negative affect and their positive affect counterparts. Further, we applied the LASSO to conduct variable selection to identify which characteristics of emotion dynamics during smoking events are associated with cigarette dependence, controlling for the effects of sociodemographic variables including race (1=White; 0=other), gender (1=male; 0=female), age, and education on an ordinal scale of 1–6 (1=never attended school; 2=elementary school; 3=some high school; 4=high school graduate; 5=some college or technical school; 6=4-year college graduate). The LASSO was adopted because of its popularity and superior performance (Tibshirani, 1996).

3. Results

Table 1 shows descriptive statistics of the sociodemographic characteristics, emotion dynamics, and dependence measures by the two groups. Among the 136 dual users, 51% were male; 70% were White; the average age was 37; and the education was at the level of some college or technical school (Level 5 on an ordinal scale of 1–6). In comparison, the group of 101 exclusive smokers had lower percentages of male (47%) and White (57%) participants, higher average age (40), and about the same education level. Nevertheless, the two groups were not significantly different on any of the sociodemographic variables. Furthermore, there were no significant group differences on any of the indices for emotion dynamics or dependence measures. The average negative affect was around 2 and the average positive affect was about 3 on a scale of 1–7. Clinically, the average scores of PS-CDI (≈ 12) and FTCD (≈ 5) were both classified as medium dependence. Yet, the standard deviations for both measures were large.

Table 1.

Descriptive statistics of sociodemographic characteristics, emotion dynamics, and dependence measures by cigarette user groups.

	Dual users (N=136)		Exclusive smokers (N=101)		Group comparison
	Mean (%)	SD	Mean (%)	SD	T (χ2) statistic	p-value
Sociodemographic characteristics
Gender (male)	51.47%		46.53%		0.38	0.54
Race (White)	69.85%		57.43%		3.39	0.07
Age	37.31	12.84	39.57	13.58	−1.30	0.20
Education (1–6)	4.71	0.87	4.51	0.78	1.84	0.07
Emotion dynamics
Negative affect:	2.24	1.41	2.05	1.21	0.71	0.48
mean
Negative affect: variance	1.00	1.33	1.22	1.51	−0.71	0.48
Negative affect: autocorrelation	−0.16	0.30	−0.11	0.25	−0.89	0.37
Negative affect: MSSD	2.02	3.16	2.26	2.89	−0.39	0.69
Negative affect: λo	0.65	0.43	0.66	0.42	−0.08	0.93
Negative affect: λs	0.39	0.23	0.36	0.20	0.68	0.49
Positive affect: mean	2.76	1.59	2.87	1.49	−0.36	0.72
Positive affect: variance	1.10	1.27	1.38	1.23	−1.11	0.27
Positive affect: autocorrelation	−0.18	0.28	−0.11	0.28	−1.24	0.22
Positive affect: MSSD	2.17	2.46	2.63	2.47	−0.90	0.37
Positive affect: λo	0.60	0.43	0.61	0.39	−0.12	0.90
Positive affect: λs	0.38	0.21	0.40	0.25	−0.42	0.67
Dependence measures
PS-CDI (0–20)	11.53	4.01	11.55	4.06	−0.03	0.98
FTCD (0–10)	4.47	2.39	4.78	2.29	−0.65	0.51
WISDM PDM (1–7)	4.55	1.39	4.66	1.44	−0.40	0.69
WISDM SDM (1–7)	4.21	1.16	4.18	1.26	0.13	0.89

PS-CDI: Penn State Cigarette Dependence Index; FTCD: Fagerstrom Test of Cigarette Dependence; WISDM: Brief Wisconsin Inventory of Smoking Dependence Motives; PDM: Primary Dependence Motives; SDM: Secondary Dependence Motives.

Table 2 depicts the correlation matrices of the 6 indices of emotion dynamics for negative affect (top) and positive affect (bottom) among dual users. The corresponding matrices for exclusive smokers have similar patterns and thus are not presented here. The correlations among the 6 indices related to negative affect were small to moderate except for the correlation between MSSD and variance (r = 0.93). The other two indices of instability (λ_o, λ_s), however, were only weakly correlated with variance (r = 0.16 and 0.21, respectively). The same pattern applied to positive affect. We, thus, decided not to include MSSD in the variable selection procedure because of its substantial overlap with variability. Table 2 also shows the correlation between each of the indices for negative affect and its positive affect counterpart (the middle row between the two correlation matrices). The correlations for variance, autocorrelation, and MSSD tended to be smaller (r = 0.25 − 0.32) than those for mean, λ_o, and λ_s (r = 0.51 − 0.61).

Table 2.

Correlation matrices of the indices for emotion dynamics in negative affect and positive affect among dual users (N=136).

	Neg aff: mean	Neg aff: variance	Neg aff: autocorrelation	Neg aff: MSSD	Neg aff: λo	Neg aff: λs
Negative affect: mean	1	0.48	−0.17	0.43	0.31	0.45
Negative affect: variance		1	−0.11	0.93	0.16	0.21
Negative affect: autocorrelation			1	−0.31	0.06	−0.31
Negative affect: MSSD				1	0.15	0.19
Negative affect: λo					1	0.37
Negative affect: λs						1
Positive affect counterpart	0.51	0.27	0.32	0.25	0.58	0.61
	Pos aff: mean	Pos aff: variance	Pos aff: autocorrelation	Pos aff: MSSD	Pos aff: λo	Pos aff: λs
Positive affect: mean	1	0.32	−0.13	0.30	0.28	0.46
Positive affect: variance		1	0.00	0.93	0.12	0.15
Positive affect: autocorrelation			1	−0.24	0.02	−0.36
Positive affect: MSSD				1	0.11	0.22
Positive affect: λo					1	0.31
Positive affect: λs						1

Table 3 displays the fitted linear regression models of PS-CDI, FTCD, and the two WISDM subscales on emotion dynamic indices (selected by LASSO) controlling for sociodemographic characteristics among dual users and exclusive smokers, separately. White participants tended to have significantly lower WISDM scores than participants of other races, but such a racial difference was not significant on PS-CDI and FTCD. Male smokers had significantly lower scores than females on WISDM SDM. Older age was associated with higher scores on PS-CDI and FTCD among both groups, whereas younger age was associated with a higher score on WISDM SDM among dual users.

Table 3.

Linear regression models of cigarette dependence measures on emotion dynamics among dual users (N=136) and exclusive smokers (N=101) resulting from variable selection via LASSO.

	PS-CDI		FTCD		WISDM PDM		WISDM SDM
	Dual users	Excl. smokers	Dual users	Excl. smokers	Dual users	Excl. smokers	Dual users	Excl. smokers
Intercept	9 38*** (2.45)	10.31*** (2.96)	4.25** (1.29)	2.46 (1.54)	4 35*** (0.77)	4 00*** (0.96)	4.83*** (0.64)	4 37*** (0.80)
Sociodemographic characteristics
Race (White)	−0.24 (0.73)	−1.25 (0.82)	−0.14 (0.45)	−0.78 (0.47)	−0.36 (0.27)	−0.79** (0.29)	−0.47* (0.22)	−0.80** (0.24)
Gender (Male)	0.70 (0.68)	−0.45 (0.79)	0.74 (0.40)	0.36 (0.44)	−0.29 (0.24)	−0.22 (0.28)	−0.46* (0.19)	−0.53* (0.23)
Education (1–6)	−0.43 (0.38)	−0.56 (0.53)	−0.41 (0.23)	−0.02 (0.29)	0.01 (0.14)	0.05 (0.18)	0.08 (0.11)	0.11 (0.15)
Age	0.10*** (0.03)	0.11*** (0.03)	0.05** (0.02)	0.05** (0.02)	0.01 (0.01)	0.02 (0.01)	−0.02* (0.01)	−0.01 (0.01)
Emotion dynamics
Neg Aff mean	0.85** (0.28)	0.43 (0.46)	--	--	--	--	--	--
Neg Aff variance	--	0.01 (0.35)	--	--	--	0.18 (0.10)	--	0.21* (0.08)
Neg Aff autocorrelation	−2.41* (1.13)	5.11** (1.72)	--	2.13* (0.92)	−0.67 (0.40)	--	--	--
Neg Aff λo	--	--	--	--	--	--	--	--
Neg Aff λs	−3.59 (1.87)	4.39 (2.48)
Pos Aff mean	--	−0.17 (0.29)	--	--	--	--	--	--
Pos Aff variance	0.20 (0.27)	0.25 (0.38)	--	0.11 (0.19)	--	--	--	--
Pos Aff autocorrelation	--	−1.63 (1.50)	--	--	--	--	--	--
Pos Aff λo	--	−2.43* (1.15)	--	--	--	--	0.52* (0.22)	--
Pos Aff λs	−2.29 (1.94)	--	--	1.83 (0.95)	--	--	--	--
Model R2	0.22	0.28	0.11	0.19	0.07	0.17	0.13	0.25

Note 1: PS-CDI: Penn State Cigarette Dependence Index; FTCD: Fagerstrom Test of Cigarette Dependence; WISDM: Brief Wisconsin Inventory of Smoking Dependence Motives; PDM: Primary Dependence Motives; SDM: Secondary Dependence Motives.

Note 2: In each cell, the numbers are regression coefficient (standard error).

^*p < 0.05

^**p < 0.01

^***p < 0.001

In general, the models for exclusive smokers explained more variance in dependence measures (R² = 0.17 − 0.28) than those for dual users (R² = 0.07 − 0.22). Based on the fitted models for PS-CDI, a higher average level (mean) and lower inertia (autocorrelation) of negative affect in dual users were associated with higher dependence, whereas among exclusive smokers, higher inertia (autocorrelation) of negative affect and lower instability (λ_o) of positive affect were associated with higher dependence. In terms of FTCD, while none of the emotion dynamic indices were associated with dependence in dual users, higher inertia (autocorrelation) of negative affect was associated with higher dependence among exclusive smokers. For WISDM PDM, none of the emotion dynamic indices were significantly associated with the dependence score among both groups. The fitted models for WISDM SDM indicate that among dual users, higher instability (λ_o) of positive affect was associated with higher dependence, whereas higher variability (variance) of negative affect was associated with higher dependence among exclusive smokers.

4. Discussion

This is the first study that conducted a comprehensive examination of emotion dynamics in negative and positive affects among exclusive smokers and dual users. The patterns of correlations among dynamic measures were similar across these two types of users. The result indicates that the most commonly adopted measure for emotion instability, MSSD, was indeed highly correlated with variance for both negative affect and positive affect. Unlike MSSD that captures the average moment-to-moment changes, the recently developed fragmentation measures focus on dramatic changes from moment to moment and only had small correlations with variance. The regression models after variable selection also demonstrate the potential use of these new measures of emotion instability in nicotine and tobacco research. For example, instability of positive affect as indicated by λ_o may carry some information about the level of cigarette dependence. Furthermore, each of the measures for negative affect dynamics was positively correlated with the positive affect counterpart, with the mean and the fragmentation measures having higher correlations. This implies that the average level and the pattern of dramatic changes for negative affect are consistent with rather than the opposite of those of positive affect during smoking episodes. The stability of these affective regulation markers across the affective spectrum may be important in developing intervention that might capitalize on a smoker’s ability to recover from negative affect but also their inability to sustain positive affect.

This study introduced two types of measures for emotion dynamics, inertia and instability, that were shown to be associated with a variety of psychological well-being measures, but have not yet been examined with respect to their relations with cigarette dependence. We demonstrate that these dynamic measures could sometime inform cigarette dependence better than within-person mean or variance. Among dual users, a higher average level of negative affect and lower inertia of negative affect (i.e., less sustained negative affect) were associated with higher cigarette dependence as indicated by PS-CDI. This suggests that a measure of dependence that focuses more on heaviness of smoking is related to greater negative affect but less sustained negative affect, possibly due to the use of cigarette smoking to reduce such negative affect. Conversely, higher instability of positive affect (i.e., greater magnitude of changes in positive affect) was associated with higher cigarette dependence as indicated by WISDM SDM, suggesting that instrumental use of cigarettes (i.e., use for secondary dependence motives such as enhancing affect or concentration) is associated with greater peaks in positive affect.

The patterns of associations among exclusive smokers were, however, different. Higher inertia of negative affect (i.e., greater sustained negative affect) and lower instability of positive affect (i.e., lower magnitude of changes in positive affect) were associated with higher scores of PS-CDI or FTCD suggesting higher levels of dependence among assessments that focus on heaviness of use. These results imply that a smoker with negative affect that lingers and shows less homeostatic recovery is more likely to depend on cigarettes to regulate mood. Additionally, a lower probability of transitioning from outside the standard range of positive affect to inside (indicated by a lower value of λ_o) may indicate poor mood regulation and perhaps higher dependence on cigarettes to aid in such regulation. Furthermore, the finding that higher variability of negative affect was associated with higher WISDM SDM is consistent with the EMA literature (Piasecki et al., 2016). In general, the regression models of emotion dynamics tended to explain the variances in cigarette dependence better among exclusive smokers than among dual users. This may be partially explained by dual users’ flexibility to use e-cigarettes as an alternative product to regulate their mood, especially given that combustible cigarette use is prohibited in more settings in everyday life. However, it is clear that more research is needed to understand the relations between short-term emotion dynamics and cigarette dependence as well as its facets.

Our findings also revealed sociodemographic differences on the three cigarette dependence scales. Among exclusive smokers, our findings were mostly consistent with previous findings that whites, compared to other races, had lower cigarette dependence measured by the WISDM scales (Piper et al, 2008; Bacio et al., 2014). Our results were also accordant with previous findings that no significant differences in the FTCD scores between white and non-white exclusive smokers were observed (St.Helen et al., 2012; Schnoll et al., 2014). Although our study showed significant gender differences in cigarette dependence measured by the WISDM SDM, caution is needed when comparing our results with previous findings that were mostly based on the total score rather than the SDM subscale of the WISDM (e.g., Allen et al, 2015; Piper et al., 2008; Bacio et al., 2014). Furthermore, we showed consistent results that older exclusive smokers had higher FTCD scores (Park et al., 2012; John et al., 2003). It is notable that, however, there have been very few studies investigating these sociodemographic differences in dual users’ cigarette dependence using these three scales. Future studies are warranted to disentangle the complexities of cigarette/e-cigarette dependence among dual users and investigate associated sociodemographic disparities.

Some limitations of the present study are important to note. First, the inclusion criteria of dual users in the original study were stricter for smoking (smoked daily for the last 3 months) than for vaping (used e-cigarettes at least once a week for the past month). These unavoidably resulted in a sample of dual users who tended to use combustible cigarettes as the primary product. Thus, the findings based on data from the dual users participating in this study may not be generalizable to all dual users. Second, these data were collected prior to the advent of Juul e-cigarettes and other nicotine salt products. For this reason, the cigarette/e-cigarette use patterns among the dual users in this study may be different from those among dual users adopting newer products. Third, in spite of the advantage of ecological validity, the EMA data used in the secondary analysis are still self-reported and thus are not free from bias. Fourth, the original study only randomly sampled up to two smoking events a day during the two-week EMA. Thus, the emotion dynamics derived from these EMA data may not fully characterize the emotion dynamics of all use events. Future EMA studies with higher assessment frequencies are warranted to verify the current findings.

In spite of the aforementioned limitations, this study has built an important bridge between the psychology literature and the literature of nicotine and tobacco research that offers some important insight into the association between short-term emotion dynamics and cigarette dependence. Our findings demonstrate that in addition to the commonly calculated mean and variance, the novel measures for inertia and instability may be important to characterize emotion dynamics across cigarette use events. Further research is needed to understand the relations between such emotion dynamics and continued smoking over time. The results also illustrate the importance of examining not only negative affect but also positive affect in order to fully understand the association between emotion dynamics and cigarette dependence. Furthermore, the different patterns of association between short-term emotion dynamics and cigarette dependence across the two groups of cigarette users call for future research that is designed to compare cigarettes and e-cigarettes in terms of their effects on emotion regulation.

Highlights.

• Inertia and instability are important emotion dynamics indicating dependence.

• Negative and positive affect are both essential for understanding dependence.

• Dual users and exclusive smokers are different on emotion-dependence association.