Purpose: Children with congenital hearing loss (HL) have auditory impairments that may place them at increased risk for delays or variability in language development. However, obtaining reliable brain markers for early classification of young children with HL versus those with normal hearing (NH), as well as for precise assessment of HL children's language ability, remains a challenge due to limitations in traditional neuroimaging techniques and theoretical frameworks. To address this gap, we propose the maternal mirror hypothesis, which suggests that brain activities of mothers might mirror or indirectly reflect children's auditory language ability, offering an additional and useful approach for obtaining brain markers of HL children in clinical assessment. Method: Children aged 2-5 years with HL (n = 105) and NH (n = 89), along with their mothers, participated in the study. Brain activity in each mother--child dyad was simultaneously measured using functional near-infrared spectroscopy (fNIRS) while they watched a silent video together. From these data, we derived maternal and child intrapersonal brain functional connectivity (FC), as well as mother-child intersubject correlation (ISC). Children's language comprehension and production ability were assessed at baseline with a follow-up of their changes over 6 months. Results and Conclusions: We found that maternal brain FC or mother-child ISC outperformed child-based FC in predicting HL children's language comprehension and production, as well as their plastic changes across 6 months. Moreover, brain markers predicting HL children's language ability did not differ between groups of HL and NH, whereas those brain markers that classified HL versus NH group status were not correlated with HL children's language ability. This dissociation suggests distinct neural mechanisms underlying HL pathology with brain deficits versus the compensatory mechanisms with the functional recovery of HL children. These findings support the maternal mirror hypothesis, having the potential to address traditional challenges in early functional assessment and prediction of HL children by providing a novel neuroimaging approach and an original theoretical framework.