Age-Group Differences in the Temporal Structure of Speech and Forehead total-Hb Responses During a Phonemic Verbal Fluency Task: A Comparison of Adults in Their 40s and 70s

In populations with well-preserved cognitive function, cognitive screening total scores tend to cluster near the ceiling, making it difficult to characterize age-group differences from total scores alone. We compared the temporal structure of word production, acoustic features, and the magnitude and timing of forehead total-hemoglobin (total-Hb) responses during a phonemic verbal fluency task in adults in their 40s and 70s. A total of 254 healthy participants (115 in their 40s, 139 in their 70s) completed a 60-s phonemic verbal fluency task requiring words beginning with the Japanese syllable /ka/, administered as part of the Japanese version of the Montreal Cognitive Assessment (MoCA-J). We derived the total word count, word counts in 10-s bins, mean inter-word pause duration, speech offset time, smoothed cepstral peak prominence (CPPS), jitter, and shimmer. Area under the curve (AUC) and time-to-peak (TTP) were computed from forehead total-Hb signals recorded with a wearable single-wavelength near-infrared spectroscopy device. MoCA-J scores clustered near the ceiling in both groups, although the age-group difference was significant. The 70s group produced fewer words (14.00 vs 17.09) and showed longer inter-word pauses (1.60 vs 0.72 s). CPPS was lower, AUC was higher, and TTP was longer (32.97 vs 15.63 s) in the 70s group, whereas jitter did not differ. Word counts across 10-s bins showed an age group time-bin interaction. Within each age group, participants who produced more words showed longer TTP. Age-group differences in TTP and AUC persisted after adjustment for speech offset time (proportions mediated, 6.7% and 0.5%) and in a subsample matched on speech offset time. Even when screening scores clustered at the ceiling, the temporal structure of word production and forehead total-Hb responses differed between age groups, and these two classes of measures dissociated. Because the sample was selectively recruited and single-wavelength total-Hb signals do not index localized neural activity, the findings are descriptive and motivate longitudinal, multi-axis characterization of speech in aging.