The study was preregistered on the Open Science Framework ( Cox et al., 2025, see: https://osf.io/s8jze/).

This was an observational study involving cross-sectional and 12-month follow-up survey data.

The study is part of the ongoing Smoking Toolkit Study (STS), a household survey of adults aged 16+ in England ( Fidler et al., 2011b). The sample is recruited via a hybrid of a random probability and quota sampling, which results in a sample that is broadly representative of the adult population of smokers in England. This hybrid method is considered superior to conventional quota sampling ( Stephenson, 1979). The choice of households to approach is limited by the random allocation of small geographical areas and one advantage is that interviewers can choose which households within these areas are most likely to fulfil their quotas, rather than being allocated households in advance. Interviews are held with one household member. Respondents provide informed consent prior to each interview. Baseline data were collected between May 2015 and April 2022 and follow-up data were collected from respondents contactable at a 12-month follow-up (May 2016 and April 2023). The STS is a representative survey as baseline, but natural selection bias at follow-up (i.e., those who respond versus those who do not) reduces the representativeness of the survey. Data are usually collected monthly through face-to-face computer assisted interviews. However, physical distancing restrictions under the COVID-19 pandemic meant that no data were collected in March 2020, data from April 2020 onward were collected via telephone, and the minimum age for participation was increased from 16 to 18 years (because of changes in consenting procedures), however both face-to-face and telephone collection provide similar key performance indicator estimates and were broadly comparable over these time points (e.g., Jackson et al., 2020). The telephone-based data collection relied on the same combination of random location and quota sampling and weighting approach as the face-to-face interviews.

For the current study, at baseline participants (aged ≥16 years) who have smoked in the past year were asked whether they had made a serious attempt to quit in the last year, this is defined as per the literature as “by serious, we mean that you made a conscious attempt to stay off cigarettes/give up smoking for good” ( Chaiton et al., 2016). Those who self-reported they were not smoking at baseline and who self-reported as to whether the attempt was successful at 12-month follow up constituted the analytic sample.

Ethical approval was granted by UCL’s Research Ethics Committee (0498/001). Our study was conducted in accordance with the relevant guidelines and regulations or in accordance to the Declaration of Helsinki. As data is collected over the telephone (as described above) participants provide informed verbal consent to take part in the study provided to the surveyor over the telephone at the time of survey, and all methods are carried out in accordance with relevant regulations. The data are not collected by UCL and are anonymised when received by the research team.

Outcome variable

Smoking status indicated at 12-month follow up, by agreement to “I’m still not smoking”, binary coded ‘smoking’ (0)/’still not smoking’ (1).

Explanatory variable

Desire and obligation are measured on a single adapted semantic differential 7-point scale. At baseline, participants are asked: “To what extent do you agree or disagree with the following statement: “I am stopping smoking because I want to rather than because I feel I have to.” Response options range from 1 (strongly disagree) to 7 (strongly agree) and was treated continuously. A higher score with a positive association with the outcome indicates that desire (over obligation) is more instrumental in explaining maintenance of quit success, and the inverse would be true for a negative association.

Baseline covariates

Smoking-related characteristics

Strength of urges to smoke, is a validated measure of dependence ( Fidler et al., 2011a), because urges to smoke vary in intensity between former smokers and can exist even after sustained periods of cessation ( Ussher et al., 2013), we included this as assessed by self-reported ratings of the strength of urges to smoke over the last 24 hours [not at all (coded 0), slight (1), moderate (2), strong (3), very strong (4) and extremely strong (5)]. This question was also coded ‘0’ for smokers who responded ‘not at all’ to the (separate) question: ‘How much of the time have you spent with the urge to smoke? This six-point scale was treated as continuous. We measured the number of prior quit attempts in the past year (categorised as 1, 2, 3 or ≥ 4), and time since quit attempt started (coded as last week (0), between a week and a month (1), 1-2 months (2), 2-3 months (3), 3-6 months (4) and 6-12 months (5), with this six-point scale treated as continuous in logistic regression models). We measured the number of prior quit attempts in the past year (categorised as 1, 2, 3 or ≥ 4), and time since quit attempt started (categorised as last week (1)/between a week and a month (2)/1-2 months (3)/2-3 months (4)/3-6 months (5) and 6-12 months (6)). We also measured if people have others who regularly smoke in their presence yes (0), no (1), this was also treated continuously. Use of evidence-based support in the most recent quit attempt (i.e., prescription NRT, e-cigarettes, behavioural support, or the medications varenicline, bupropion and telephone/written/digital), coded no (0), yes (1). Quit approach at their most recent attempt (gradual/abrupt) was assessed with the question, ‘Did you cut down the amount you smoked before trying to stop completely at your most recent serious quit attempt?’ and two options: ‘Stopped without cutting down’ (abrupt (0)) or ‘Cut down first’ (gradual (1)), this is included because observational studies often report abrupt quitting as superior for cessation (e.g., Garnett et al., 2022).

Sociodemographic characteristics

Sex was categorized as women (0) versus other (1) and age was treated continuously (owing to small cell sizes). Socio-economic positioning was operationalised by housing tenure (as per a recent analysis ( Beard et al., 2021), as mortgage; owned outright; rented from private landlord or other coded as private tenure (0); belongs to housing association or rented from local authority, or social housing coded as social housing (1). Occupational social grade was operationalised using the British National Readership Survey’s Social Grade Classification Tool, which comprises the categories AB (higher and intermediate managerial, administrative and professional), C1 (supervisory, clerical and junior managerial, administrative and professional), C2 (skilled manual workers), D (semi-skilled and unskilled manual workers) and E (state pensioners, casual and lowest-grade workers, unemployed with state benefits), with responses coded into routine and manual (1) (C2DE) and all other (ABC1) social grades (0). Year of survey was also included as a categorical variable.

Analyses were conducted in R version 4.2.

Descriptive analyses

Descriptive statistics (% (n) or means ± standard deviations (SDs), as appropriate) are presented to report the variables included in the analyses (socio-demographic characteristics and smoking and quitting behaviour).

Data were weighted with raking such that the analytic sample matched the English population profile on the dimensions of age, social grade, region, housing tenure, ethnicity and working status within sex.

Inferential analyses

To assess the predictive value added to the model through the inclusion of the motivation (desire vs. obligation) factor, the analysis was conducted before and after including it as a predictor. To address the research question, unweighted binary logistic regression analyses were conducted with adjustment for all covariates. Analysis was conducted on complete cases at follow-up only. Adjusted odds ratios (AOR) are presented with 95% confidence intervals and p-values. An Akaike information criterion (AIC) was conducted on both models to assess model fit.

This study aimed to assess, in people who report having stopped smoking within the past 12 months, whether those who report stopping because of desire rather than obligation are more likely to still be abstinent 12 months later, after adjusting for a range of other well-known predictors of the outcome. We found evidence that people motivated to quit by desire had similar or lower odds of continued 12-month abstinence than people motivated by obligation. Including a measure of desire and obligation strengthened the model fit, allowing better prediction of which individuals would remain abstinent at 12 months.

Further understanding of the relationship between desire and obligation (and other motivational facets) for sustaining quit attempts is required. To date, the literature suggests those reporting desire over obligation would be more likely to report abstinence (e.g., Smit et al., 2011), our results indicated that people motivated by desire had similar or lower odds of continued abstinence a year later than people motivated by obligation. There are several possible explanations for this result. We surveyed people who had already quit smoking at baseline, and this may create a selection bias in that desire may have already helped them to achieve a period of success — meaning that those people who relapsed to smoking shortly after attempting to quit, who may have been motivated by obligation, are not included in the sample (as is shown by the low proportion of participants in this study motivated by obligation).

PRIME theory proposes that different facets of the motivational system are likely key at different stages during the quitting process (i.e., time-varying effects) and may be useful for interpreting the findings here ( West & Brown, 2014; West & Michie, 2020). Over time, as one quits smoking and experiences the benefits to oneself and others, the strongest wants and needs and impulses in response to smoking related cues or stress, which are often automatic, can be adequately responded to even given challenging psychological and environmental circumstances. Though speculative, one interpretation could be that desire is necessary for the early phase of quitting but as that is realised, even in the short to medium term, obligation to either oneself or indeed others is more important for sustaining abstinence. This is an empirical question and further analysis – ideally using longitudinal survey data with multiple measurements (e.g., Ecological Momentary Assessment in people’s daily lives; ( Shiffman et al., 2008) and through greater qualitative exploration over time – is needed to test this theory.

We also tested if our measure of desire and obligation was predictive over and above other known markers of smoking relapse. Removing the measure did reduce the fit of the model but the size of the association with the measure was small and not statistically significant. Time since quit was an independent predictor in both models, and strength of urges became significant only when the measure of desire and obligation was removed (model 2).

The main strengths of this study are that data derive from a nationally representative sample of people who had quit smoking in England, and this was the first analysis to examine desire over obligation over a 12-month period. Such longer-term examinations are needed because less than 5% of those making a quit attempt maintain this for one year without relapsing, and the chances of remaining quit vastly improve thereafter. There are several limitations. The sample was small and although the STS does generally capture a representative sample of adults, those were followed up were more likely to be older and female. Asking people who quit smoking up to 12 months ago to retrospectively report their motives for quitting at that point in time also constitutes a limitation. Indeed, past similar studies have measured motivation prior or at the time of quitting, not post like we have here. Furthermore, people misremember quit attempts and misreport quit attempts especially those that are unaided, so time since quit should be treated with caution in its association with motivational factors ( Perski et al., 2022). A longitudinal survey with multiple measurements over time might be better able to capture trajectories of relapse and abstinence. People’s recall of their motivations may also change. We did not adjust for the number of quit attempts, as we included people who had recently quit only measured cross-sectionally at baseline, therefore those who had quit for a longer period would not have had the opportunity to report many previous quit attempts so measure would not have been valid. We cannot ascertain here whether people interpret “should quit” within the survey question as obligation to oneself or to others, or both. Disentangling this is useful from both an empirical point but also to shape public health messages, i.e., is it more powerful to create messages around quitting for one’s own benefit or benefit towards others? Motivation is also dynamic and complex, and PRIME theory suggests is changing over time, we forced people into a choice (because we wanted to establish importance of these constructs), but it cannot be discounted that people felt both obligated by others and had desire for themselves to quit. Lastly, our measure of desire and obligation has not been validated, and therefore how true a measure it is and how reliable it is, is unknown.

Future research should try to ascertain the meaning of obligation, whether this is to oneself or to others, or both. Our measure is a simple one, and receiving public involvement on how this is interpreted may benefit future surveys looking to use the same or a similar measure. As noted above, in-depth qualitative research can also help with this. It is likely of course, that a range of motivational factors vary in intensity and salience across the time someone has quit. As suggested above, future research using real-time Ecologically Momentary Assessment can help capture time while also monitoring within-person changes in both internal and external factors and their potential interactions (e.g., see ( Perski et al., 2023). Motivation is largely influenced by external factors outside of the persons control; for example, while only speculative we saw a non-significant rise in the number of people reporting desire during the pandemic years. It would be useful to know what it was about those circumstances that increased one’s desire to quit which was not present in other years (if this finding is not an artefact). It is also plausible that certain types of quit aids increase or sustain motivation, either through the successful reductions of urges to smoke or other factors. Therefore, understanding the role of aids could also be a useful consideration within future studies.

Ethical approval was granted by UCL’s Research Ethics Committee (0498/001). Our study was conducted in accordance to the relevant guidelines and regulations or in accordance to the Declaration of Helsinki. Respondents provide informed consent prior to each interview.