Six Questions to Measure Sweet Addiction: Scientists Validate a New Multilingual Scale

Sugar has long been the villain of modern diets, blamed for obesity, type 2 diabetes, and a cascade of metabolic problems. But for a subset of people, the trouble with sweets goes deeper than a weak willpower moment. Clinicians in weight-management, diabetes, and eating-disorder services routinely hear patients say that it is not food in general they cannot control, but sweets specifically: the chocolate after dinner, the pastries they cannot walk past, the ice cream eaten in secret. Until now, no brief, multilingual instrument existed to capture that complaint systematically. A new study published in the International Journal of Mental Health and Addiction introduces the Brief Sweet Addiction Scale, or BSAS, a six-item questionnaire developed and validated simultaneously in English, Spanish, and Hungarian, offering researchers a compact tool for a construct that has long hovered at the edge of addiction science.

The scale is built not on the diagnostic criteria of substance-use disorders but on a different theoretical foundation: Griffiths’ components model of addiction, itself derived from earlier work on gambling. This model holds that all addictions, whether to substances or behaviors, share six components. Salience describes the activity dominating thought and behavior; mood modification refers to the arousal or relief it produces; tolerance is the need for ever-larger amounts to achieve the same effect; withdrawal covers the unpleasant states that follow abstention; conflict captures the interpersonal and intrapsychic costs incurred; and relapse describes the return to old patterns after restraint. Crucially, the model is agnostic about whether the object of addiction is a substance, which is precisely what made it usable for sweets, a target whose pharmacological addictiveness remains hotly disputed.

Each of the six components was translated into a single item referring specifically to sweet foods. Do you think about sweets constantly? Do you eat them to change how you feel? Do you need more to get the same effect? Do you feel bad when they are unavailable? Does your sweet consumption create internal conflict over health, weight, or self-esteem? Do you return to old habits after cutting down? Respondents answer on a six-point agreement scale, and the items are summed into a score ranging from 6 to 36. The team deliberately avoided reverse-worded items, which are known to load on a separate wording factor that a six-item scale cannot absorb, particularly across translations. The English items were rendered into Spanish and Hungarian through independent forward translation and blind back-translation by bilingual translators, with discrepancies resolved in discussion and pilot-tested for comprehensibility.

To test the scale, the researchers recruited 1,130 adults online between August and November 2025: 437 English speakers, 300 Spanish speakers, and 393 Hungarian speakers. The recruitment used social media platforms and snowball sampling, with no paid advertising or participant pools, and the authors acknowledge that the self-selected, predominantly female samples limit generalizability. Participants completed the BSAS alongside the Yale Food Addiction Scale 2.0, the dominant 35-item instrument in the field, which adapts DSM-5 substance-use criteria to eating as a whole. The comparison was deliberate: the YFAS 2.0 leaves the object of addiction unspecified, a weakness critics have seized upon, since an addiction ordinarily has a target and food in general is not a candidate substance. Evidence suggests addictive-like eating concentrates in processed foods high in sugar and fat, making a sweet-specific measure a plausible refinement.

The psychometric results were largely encouraging. Confirmatory factor analyses supported a single underlying dimension in the total sample, with all six items loading significantly onto one latent factor. Fit was excellent in the English sample, acceptable in the Spanish sample, and weaker in the Hungarian sample, where the RMSEA index reached .120, exceeding conventional thresholds even though other fit indices remained acceptable. The authors report this plainly as a limitation rather than evidence of a different structure. Internal consistency was good to excellent everywhere, with Cronbach’s alpha ranging from .87 to .92, remarkable for an instrument one-fifth the length of the YFAS 2.0. Average variance extracted exceeded 50 percent in every group, and inter-item correlations were uniformly strong.

Perhaps the most technically demanding test was measurement invariance, the question of whether the scale measures the same thing in each language. Using nested multigroup models, the team examined configural, metric, scalar, and strict levels of invariance, each imposing progressively stronger equality constraints across groups. All levels were supported, and the invariance was full rather than partial: no item had to be released from any constraint. This matters practically because scalar invariance is what permits meaningful comparison of scores across language groups, and strict invariance implies that even measurement error is equivalent. Differences between the three language versions are therefore unlikely to be artifacts of translation or response style, though the authors caution that the self-selected samples differed substantially in age, sex distribution, and body mass index.

Convergent and discriminant validity told a nuanced story. The BSAS correlated strongly with YFAS 2.0 severity in the total sample, r = .61, and this association remained positive within each language group even after adjusting for age, sex, and BMI. Yet the scale was not merely a miniature YFAS. The heterotrait-monotrait ratio of correlations, a modern discriminant-validity criterion, came in at .745 in the pooled sample, below the .85 threshold that would suggest the two instruments measure the same construct. The authors are careful, however, noting that the language-specific correlation matrices showed numerical irregularities and that differences in theoretical framework and response format could contribute to the observed discrimination. Neither do these findings prove the BSAS adds predictive information beyond general food-addiction measures.

The associations with weight-related variables added further texture. In the pooled sample, BSAS scores correlated moderately with BMI and highest lifetime weight, weakly with current weight, and essentially not at all with lowest lifetime weight or self-nominated ideal weight. But the language-stratified analyses qualified this picture: highest lifetime weight was significantly associated with BSAS scores only in the English group, and current weight only in the English and Spanish groups. Regression models confirmed a positive BMI association after demographic adjustment, but the authors stress that cross-sectional data cannot establish causality. Correlations with self-reported eating-disorder history were numerically smaller for the BSAS than for the YFAS 2.0, hinting that the two instruments may have different external-validity profiles, though formal comparisons were not performed.

What the study deliberately does not claim is as important as what it does. Sweet addiction appears in neither the DSM-5 nor the ICD-11, and even food addiction lacks an accepted criterion standard. The authors consider a diagnostic cut-off premature: a threshold is a classification claim requiring sensitivity and specificity estimates against a clinical criterion, and deriving one from a self-selected, non-clinical sample would be an artifact. Low base rates in behavioral-addiction research also inflate false-positive rates, a pitfall commentators have warned against specifically. The BSAS is therefore offered as a dimensional research instrument, with high scores treated as an indication for further assessment rather than a diagnosis. The scientific dispute over whether sugar is addictive in the pharmacological sense remains unresolved, with reviews finding human evidence weak while others argue that highly processed foods meet established criteria for addictive substances. Because the components model describes a behavioral pattern rather than a mechanism, the scale can be useful regardless of how that debate ends. The authors call for independent replication, test-retest reliability studies, criterion validation against clinical interviews, and comparisons with related constructs such as sweet craving, emotional eating, and binge eating before the tool finds broader application.

Subject of Research: Development and psychometric validation of a brief multilingual scale for measuring addiction-like sweet consumption

Article Title: Development and Psychometric Validation of a Brief Sweet Addiction Scale (BSAS) in English, Spanish, and Hungarian

Article References: Avena, N., Kovács, R. E., de la Vega, R., Soraci, P., Somogyi, A., & Szabo, A. (2026). Development and Psychometric Validation of a Brief Sweet Addiction Scale (BSAS) in English, Spanish, and Hungarian. International Journal of Mental Health and Addiction. https://doi.org/10.1007/s11469-026-01731-2

Image Credits: AI Generated

DOI: 10.1007/s11469-026-01731-2

Keywords: sweet addiction, Brief Sweet Addiction Scale, psychometrics, food addiction, behavioral addiction, measurement invariance, components model, Yale Food Addiction Scale, sugar consumption, scale validation, cross-cultural research, eating behavior