To determine if P300 latency changes precede and correlate with memory and mental status, patients (N=1506 aged 20-100 years) who received medical and psychiatric diagnoses (from 1997 to 2002), were assessed for P300 (N=1496), WMS-III (N=694), and MMSE (N=456). Patient and control groups included, a) normal WMS-III on all 4 subscales (N=36), b) normal WMS-III and MMSE (N=189) with subjective memory/mental status complaints, and c) medical patients with normal WMS-III and no memory complaints (N=205), and d) P300 control group without medical, psychiatric or memory problems for ROC. Patients with impaired/borderline memory had a prolonged P300 latency (P<0.02) compared to age matched non-impaired controls; in patients with normal WMS-III/MMSE, with subjective mild memory/mental status impairment, P300 latency was prolonged compared to controls (P=0.0004). The P300 latency increased by 0.72ms per year (P=7.9x10(-65)) and voltage decreased by 0.03dV per year (P=6.7x10(-10)), and both parameters were linearly correlated with the age of the subjects. Male subjects had an average voltage of 6.1dV and female 6.8dV (P=0.00009). Statistically, prolonged latency began at age range 41-50 (P=0.0002); reduced P300 voltage began at age range 51-60 (P=0.003). WMS-III memory decline for all measures began in females at age range 61-70 (P value at least=0.02) and for males at age range 61-80 (P=0.02). Prolonged P300 latency (P<0.0001) and memory impairment (at least <0.02) were greater for females than males. MMSE memory decline, male and female, began at age range 81-90 (P value of at least 0.00007). In our logistic regression model P300 latency was more predictive of WMS-III impairment than MMSE > 24. In patients whose WMS-III score is impaired < or = 69, or borderline < or = 79 (P at least=0.004), a P300 latency more prolonged than the norm (> or = 300 + 30 + Age) identifies these patients, whereas a MMSE > 24 failed. With the ROC curve, we confirmed that P300 latency could accurately identify borderline/impaired memory.