A fixed test did not confirm shared concavity of load responses in EEG and fNIRS from 20 reserved participants

We tested whether a workload measure calibrated at 0-back and 2-back follows a straight line at 3-back. Public simultaneous EEG and fNIRS data supplied two distinct signals in the same people. Six people supplied development. A local code lock preceded tests in 20 reserved people. EEG used posterior alpha suppression. fNIRS used frontal oxygenated-haemoglobin change. Each person model adjusted session and block order. Curvature was A3-1.5*A2+0.5*A0, in person block-SD units. Mean EEG and fNIRS curvatures were -0.343 and 0.03491, with 95% person t intervals [-0.5563 to -0.1298] and [-0.2851 to 0.355]. A shared claim also required positive 2-back-minus-0-back responses in both signals. The fixed rule failed, p=1. All controls and failures remain, including a previous-load check and the initial failed EEG quality rule. Two behavior folders had different internal person IDs and were excluded from that check. Exploratory different-session person agreement was -0.03835, with bootstrap range [-0.2202,0.1444]. A curve below a straight line does not prove a flat response or a memory limit. One study, paired people, different sensors, eye signals, scalp blood flow, task rules and carryover limit scope. No common neural cause or separate-study transfer is established. The package gives new block results, source pins and terms, all statistics, figures, code and repeat instructions. Original neural arrays remain external.

working memory · n-back · EEG · fNIRS · calibration · concavity · replication