Figure 10. Two visual targets create two implicit error sources.

(A) Data reported in Mazzoni and Krakauer, 2006. Blue shows error between primary target and cursor during adaptation and washout. Red shows the same, but in a strategy group that was instructed to aim to a neighboring target (instruction) to eliminate target errors, once participants experienced two large errors (two cycles no instruction). (B) The error between the cursor and the aimed target during the adaptation period. These curves are the same as in A except we use the aimed target rather than primary target, so as to better compare learning curves across groups. (C) The washout period reported in A. Here, error is relative to primary target, though in this case aimed and primary targets are the same. (D) We modeled behavior when implicit learning adapts to primary target errors e₁. Note that the no-strategy learning group resembles data. However, strategy learning exhibits no drift because the implicit system has zero error. Note here that the primary target error of 0° is a 45° aimed target error in the strategy group. (E) Similar to D, except here the implicit system adapts to errors between the cursor and aimed target, termed e₂. (F) In this model, the strategy group adapts to both the primary target error and the aimed target error (e₁ and e₂ at top). The no-strategy group adapts only to the primary target error. Learning parameters are identical across groups. (G) We show how aiming target and primary target errors evolve in the strategy group in F. (H) A potential neural substrate for implicit learning. The primary target error and aiming target error engage two different sub-populations of Purkinje cells in the cerebellar cortex. These two implicit learning modules combine at the deep nucleus. (I) Data reported in Taylor and Ivry, 2011. Before adaptation, subjects were taught to re-aim their reach angles. In the ‘nstruction with target’ group, participants re-aimed during adaptation with the aid of neighboring aiming targets (top-left). In the ‘instruction without target’ group, participants re-aimed during adaptation without any aiming targets, solely based on the remembered instruction from the baseline period. The middle shows learning curves. In both groups, the first two movements were uninstructed, resulting in large errors (two movements no instruction). Note in the ‘instruction with target’ group, there is an implicit drift as in A, but participants eventually reverse this by changing explicit strategy. There is no drift in the ‘instruction without target’ group. At right, we show the implicit aftereffect measured by telling participants not to aim (first no feedback, no aiming cycle post-adaptation). Greater implicit adaptation resulted from physical target. Error bars show mean ± SEM. Statistics: *p < 0.05, ***p < 0.001.