*Background:* The 2x2 factorial model has been recently proposed as a promising framework to measure individual differences in interoception. The first factor addresses which domain is being measured (interoceptive accuracy vs. attention), while the second distinguishes how it is being measured (self-report beliefs vs. objective performance). *Aim: *The current study examined the association between self-reported interoceptive accuracy and attention. We hypothesized no linear association between these constructs, although a quadratic U-shaped association was expected. Furthermore, alexithymia should be differentially related to interoceptive accuracy and attention. *Methods:* An online community sample (*n *= 515) completed the Interoceptive Accuracy Scale (IAS), the Body Perception Questionnaire (BPQ) - indexing self-reported interoceptive attention, and the Toronto Alexithymia Scale (TAS). Pearson correlations, Steiger’s Z-test, polynomial regression analysis, and two-lines testing were used for statistical analysis. *Results:* IAS was positively correlated with BPQ, r = .204, p < .001. In the polynomial regression analysis, the linear model indicated a positive association between IAS and BPQ (4.2%), but the quadratic term explained an additional 12.7% of the variance. Two-lines testing indicated a U-shaped association between self-report interoceptive accuracy and attention. IAS was negatively correlated with TAS, r = -.291, p < .001, while there was no significant association between BPQ and TAS, r = -.030, p = .500, as these correlations were statistically different. *Implications:* These results suggest that interoceptive accuracy and attention can be dissociated using self-report measures and may display a quadratic U-shaped association, providing further evidence for the 2x2 factorial model. Future studies should explore the non-linear relationship between interoceptive accuracy and attention using alternative questionnaires and performance-based measures.