Two groups of Tamil-speaking older adults (≥60 years) participated, group 1 included cognitively healthy elderly and elderly with Mild NCD, with 30 participants in each group. Older adults were defined as individuals aged 60 years and above, reflecting the age range commonly addressed in the UN Population Ageing 2019 and the WHO Global Health Estimates 2019, and considering the context of population aging in India. Healthy controls scored >22 on the montreal cognitive assessment-Tamil version (MoCA-TAM); [12] and showed intact performance on all Tamil cognitive linguistic assessment protocol (T-CLAP) [13] domains. The MoCA-TAM assesses seven cognitive domains including visuospatial/executive functions, attention, naming, language, abstraction, delayed recall, and orientation designed for differentiating healthy individuals from those with Mild NCD. The T-CLAP assesses attention, perception and discrimination (auditory and visual), memory (episodic, working, semantic), reasoning and problem solving, and organization. These domains complement MoCA-TAM by capturing additional cognitive-linguistic abilities relevant for early detection of Mild NCD. Mild NCD participants met DSM-5 criteria, with MoCA-TAM scores between 19.02 and 21.96 (1~2 standard deviations below norms) and mild deficits on T-CLAP without significant functional impairment. Groups were matched for age, gender, and education. Inclusion criteria included native Tamil speakers, age ≥60, minimum 8thgrade education, normal or corrected vision/hearing, and adequate task comprehension. The education levels of the participants are listed (Table 1) according to the Indian Standard Classification of Education [14]. Exclusion criteria ruled out neurological, psychiatric, or medical conditions affecting cognition; dementia, stroke, neurodegenerative disease; language disorders, head injury, developmental disabilities; cognition-impairing medications; and recent substance abuse. Demographic details of the participants are represented in Table 1.

Informed consent was obtained from all participants in accordance with ethical research guidelines. The naming test was administered using the test of naming in Tamil, a semi-automated E-naming tool developed by the first author as part of a doctoral thesis. The tool was developed using Next.js (Vercel, Sanfrancisco, CA, USA) and TypeScript (Microsoft, Redmond, WA, USA) as a lightweight, portable application optimized for personal computers. It employs Webkit Speech Recognition (Google, Mountain View, CA, USA) for real-time voice detection and transcription in Tamil and English, enabling measurement of naming accuracy and latency while accommodating linguistically diverse and low-literacy populations. The stimuli represent highly frequent & familiar noun targets across common lexical categories in Tamil language. The tool displays 60 black-andwhite line drawings on-screen, and participants are required to name each image. In Figure 1, target word is “cow” and facilitators are “milk man” and “cart”. Latency in millisecond is shown below the test item image in the bottom-left corner. Options of “accept” and “skip” is shown in the bottom-right corner of the test item image.

Each test item is shown for 10 seconds, in case of correct answer, the next test item gets displayed. In case of “incorrect response” or “don’t know” response, facilitators 1 & 2 pops up which are connected to target image in terms of scene co-occurrence or any other association. When the participant names the test item correctly, after 1st facilitator, the next test item run starts automatically, if failed, 2nd facilitator pops up. Following this, either a correct response or incorrect response may be expected. Any which way, after the 2nd facilitator, the next target test item gets displayed. Responses are automatically classified as correct or incorrect and naming latency was calculated as the time interval between stimulus presentation and correct target word production. Correct naming on the first attempt or after facilitator prompts, was accepted as accurate, with latency recorded up to the point of correct retrieval. This occurs for all 60 test items. Accuracy (percentage accuracy) included correct responses, synonyms, borrowed translations, and self-corrections. Both accuracy and latency data were logged locally for each participant. At the end of the session, the tool generated total scores and a graphical display of item-wise performance. Testing was conducted in a quiet setting with stable internet (≥250 kb/s), with participants seated 2~3 feet from a 35.56 cm (14-inch) laptop monitor.

All statistical analyses were performed using JASP version 0.19.3 (University of Amsterdam, Amsterdam, Netherlands). Kolmogorov-Smirnov test indicated non-normal distribution (p < 0.05), and therefore, Mann-Whitney U-tests were used to compare group differences in naming accuracy and latency.

Healthy older adults achieved near-ceiling naming accuracy (mean = 97.99%), whereas Mild NCD participants demonstrated significantly reduced performance (mean = 79.72%). This decline may reflect breakdowns in lexical-semantic access, causing disruptions in semantic memory networks and retrieval strategies [16]. Recent studies also emphasize that accuracy deficits may appear subtly in Mild NCD and progress rapidly in conversion to Alzheimer’s disease. Studies highlight that item-level lexical-semantic retrieval, such as delayed recall of proper names, is sensitive to early Alzheimer’s pathology. These subtle deficits may appear before broader cognitive decline, supporting the use of targeted tasks like confrontation naming as early behavioral indicators of neurodegeneration [17]. Our results corroborate this pattern, underscoring the clinical relevance of naming accuracy as a sensitive index of cognitive-linguistic vulnerability. Literature evidence states that semantic fluency tasks have resulted to be crucial to differentiate Mild NCD group from, healthy ageing and preclinical and clinical groups [18].

Latency analysis revealed a striking contrast, with Mild NCD participants taking approximately three times longer to produce correct responses than controls. Prolonged latencies likely reflect slowed cognitive processing, impaired lexical search, and weaker inhibitory control which are considered hallmarks of early neurodegeneration [11,16]. The greater variability observed within the Mild NCD group suggests heterogeneity in disease trajectory, supporting earlier reports that latency-based measures can differentiate Mild NCD from healthy aging and identify individuals at higher risk for progression [19-21]. Recent studies further corroborate these findings, showing that naming impairments are sensitive markers not only even in atypical Alzheimer’s presentations [22], and that examining both accuracy and latency enhances the discrimination of Mild NCD from healthy aging.

The study provides several implications for clinical practice. First, the marked latency differences indicate that naming latency may serve as a more sensitive clinical marker than accuracy alone, especially in high-functioning individuals who can compensate for accuracy but not speed. Second, the use of a Tamil-specific, culturally adapted tool (E-naming) ensures ecological validity for underrepresented populations, supporting calls for linguistically inclusive screening tools. Finally, the tool’s semi-automated capture of accuracy and latency offers a practical avenue for large-scale digital screening in community and clinical settings, potentially aiding early identification of Mild NCD where access to neuropsychological testing is limited.

Future research should adopt longitudinal designs to determine whether latency and accuracy changes predict conversion from Mild NCD to major neurocognitive disorders. Integration with neuroimaging or electrophysiological measures could clarify the neural bases of slowed lexical retrieval. Expanding sample size and including different subtypes of Mild NCD would allow finer-grained profiling of language decline. Additionally, cross-linguistic studies comparing Tamil with other Indian languages could enrich understanding of how cultural and linguistic factors mediate naming performance in neurocognitive disorders.