1. Forming a ratio with respect to a standard (the normal). A normalized value usually is dimensionless. Normalizing often consists of scaling such that ‘‘something equals one,’’ the ‘‘something’’ being the rms value, the maximum value, etc. For example, an array response may be normalized by dividing each value by the rms value; hence for the array X=(x1,x2,...,xn), the rms value is Y=[(x12+x22+...+xn2)/n]1/2 and the normalized array is (x1/Y,x2/Y,...,xn/Y). Autocorrelations are normalized by dividing by the value at zero time lag so that the maximum value of ‘‘one’’ indicates perfect correlation. Type curves (calculated effects for a model body) often are normalized so that the maximum effect is one. 2. To adjust a floating point number so that the most significant bit (or digit) is held in the highest position of the mantissa, thereby permitting maximum precision to be represented. 3. Scaling so that the rms amplitudes of all traces are equal. 4. To adjust measurements so that they fall within a prescribed range.