Mutations in the gap junction β2 (GJB2) gene, encoding the connexin26 (CX26) protein, are the most common cause of non-syndromic hearing loss (HL) in many populations. In the East Asian population, two variants, p.V27I (c.79G>A) and p.E114G (c.341G>A), are considered benign polymorphisms since these variants have been identified in both HL patients and normal hearing controls. However, some studies have postulated that homozygotes carrying both p.V27I and p.E114G variants could cause HL. To elucidate possible roles of these variants, we used in vitro approaches to directly assess the pathogenicity of four haplotypes generated by the two polymorphisms: VE (wild type), I*E (p.V27I variant only), VG* (p.E114G variant only), I*G* (both variants). In biochemical coupling assays, the gap junctions (GJs) composed of VG* and I*G* types displayed defective channel activities compared with those of VE wild types or I*E types, which showed normal channel activities. Interestingly, the defect in hemichannel activity was a bit less severe in I*G* type than VG* type, suggesting that I* variant (p.V27I) may compensate for the deleterious effect of G* variant (p.E114G) in hemichannel activities. Our population studies using 412 Korean individuals showed that I*G* type was detected at around 20% in both HL patients and normal controls, suggesting that I*G* type may not be a pathogenic polymorphism. In contrast, VG* type was very rare (3/824) and detected only in HL patients, suggesting that VG* homozygotes (VG*/VG*) or compound heterozygotes carrying VG* type with other mutations may cause HL.