Background: In nursing homes, most of the patients with dementia are affected by severe cognitive disorder. Care interventions follow an accurate and recurring multidimensional assessment, including cognitive status. There is still a need to develop new performance-based scales for moderate-to-advanced dementia. Objectives: The development of the Residual Cognition Assessment (RCA) responds to the need to create new scales for global cognitive screening in advanced dementia, with some peculiar features: performance based, brief (<5 m), available without specific training, and suitable for nonverbal patients with minimal distress. Methods: Two raters have administered the RCA and the Severe Impairment Battery-short version (SIB-S) to 84 participants with MMSE = 0. After 2–3 weeks, the same sample has been retested. The RCA has been also administered to 40 participants with MMSE 1–10 for a comparison. Results: The RCA has exhibited excellent values for test-retest reliability (intraclass correlation [ICC] = 0.956) as well as for inter-rater reliability (ICC = 0.997). The concurrent validity analyzes have shown strong correlations between the RCA and the SIB-S with ρ = 0.807 (p < 0.01), and the RCA and the Clinical Dementia Rating (CDR) with ρ = −0.663 (p < 0.01). Moderate correlation has been found between the RCA and the Functional Assessment Staging Scale with ρ = −0.435 (p < 0.01). The instrument has showed high internal reliability, too (total: α = 0.899). The RCA has low floor effect (2%) with respect to the SIB-S (58%) but shows ceiling effect in the MMSE 1–10 sample (50%). The ROC curve analyses demonstrate that the RCA is acceptably able to discriminate between subjects with CDR 4/5 with an AUC of 0.92. Exploratory factor analysis shows 3 factors, defined as three major degrees of cognitive performance in advanced dementia, indeed hierarchically structured in three possible levels of decline. Conclusions: The RCA has showed excellent validity and reliability as well as good sensitivity to identify advanced cognitive impairment in dementia, without floor effect. The RCA seems complementary to the MMSE, so advisable when the latter reaches 0. Administration and scoring are simple, and only few minutes are required to assess the patient. The RCA can discriminate at least 3 different major stages in advanced dementias: severe, profound, and late.