but increasing spread, so that in the lower right panel near superposition is obtained, similarly to the lower right panel of Figure 26.9, for scalar pacemaker rate variance.
In both threshold cases, however, the function forms are different They are sharply peaked and, in fact, discontinuous at T = S+. The discontinuity stems from the possibility in theory that occasional negative threshold values may preclude responding for any discrepancy. If the account is altered to constrain all thresholds positive, then an alternative difficulty arises. Accuracy at T = S+ is always perfect. Unless further modified, these features make it unlikely that the smooth shoulders and less than perfect accuracy we see in Figure 26.2 may be reconciled with threshold variance alone.
Our conclusion from this analysis is that an information processing model that satisfies the proportionality or superposition finding may be obtained from an appropriate selection of a response rule and an adroit introduction of variance in the structures we describe. An absolute response rule