The reliability of a two-item scale is a topic of debate among researchers. While Cronbach's alpha is commonly used to assess reliability, some argue it is inappropriate for two-item scales. Others suggest using the Pearson correlation coefficient or the Spearman-Brown formula. This article clarifies the appropriate reliability coefficient for two-item scales. The reliability of a two-item scale depends on the measurement model. Under parallel measures, Cronbach's alpha equals the squared correlation between the true score and the scale score. The Pearson correlation between the two items is lower than the reliability of the scale. The Spearman-Brown formula can convert split-half reliability into a reliability estimate with the coefficient alpha interpretation. However, it is not equivalent to coefficient alpha unless the items are parallel. For tau-equivalent measures, Cronbach's alpha equals the squared correlation between the true score and the scale score. The Spearman-Brown coefficient is always larger than coefficient alpha for two-item scales, except when the items are parallel. For congeneric measures, Cronbach's alpha may significantly underestimate the true reliability. The Spearman-Brown coefficient is more accurate in this case. The bias of the coefficient is the difference between the true reliability and the estimate obtained by using either Cronbach's alpha or the Spearman-Brown formula. The article concludes that the Spearman-Brown formula is a more appropriate reliability coefficient for two-item scales. It is less biased than Cronbach's alpha and always provides a reliability estimate greater than or equal to alpha. The Pearson correlation is not an adequate measure of reliability for two-item scales. The article emphasizes that two-item scales are not recommended, as they may not adequately represent the construct. However, if only two items are available, the Spearman-Brown reliability estimate is the best option.The reliability of a two-item scale is a topic of debate among researchers. While Cronbach's alpha is commonly used to assess reliability, some argue it is inappropriate for two-item scales. Others suggest using the Pearson correlation coefficient or the Spearman-Brown formula. This article clarifies the appropriate reliability coefficient for two-item scales. The reliability of a two-item scale depends on the measurement model. Under parallel measures, Cronbach's alpha equals the squared correlation between the true score and the scale score. The Pearson correlation between the two items is lower than the reliability of the scale. The Spearman-Brown formula can convert split-half reliability into a reliability estimate with the coefficient alpha interpretation. However, it is not equivalent to coefficient alpha unless the items are parallel. For tau-equivalent measures, Cronbach's alpha equals the squared correlation between the true score and the scale score. The Spearman-Brown coefficient is always larger than coefficient alpha for two-item scales, except when the items are parallel. For congeneric measures, Cronbach's alpha may significantly underestimate the true reliability. The Spearman-Brown coefficient is more accurate in this case. The bias of the coefficient is the difference between the true reliability and the estimate obtained by using either Cronbach's alpha or the Spearman-Brown formula. The article concludes that the Spearman-Brown formula is a more appropriate reliability coefficient for two-item scales. It is less biased than Cronbach's alpha and always provides a reliability estimate greater than or equal to alpha. The Pearson correlation is not an adequate measure of reliability for two-item scales. The article emphasizes that two-item scales are not recommended, as they may not adequately represent the construct. However, if only two items are available, the Spearman-Brown reliability estimate is the best option.