Page 12 - dan_sun

Introduction and Overview

are not in terms of causal knowledge and deliberate decisions, but rather a mechanistic
process that could explain goal-directed behavior based on merely the formation of associ-
ations. In short, it assumed that cooccurrence of actions and resulting events could lead to
1 the formation of bidirectional associations, by which activation of the representation of the
event or outcome could readily activate the representation of the action that produced it,
without any deliberation, reflection, or decision.

Although there are similarities with the notion of automatized goal-directed be-
havior (Bargh, 2006; Custers & Aarts, 2005; Kruglanski et al., 2002), this approach not only
removed deliberate conscious processes from the execution of goal-directed behavior, it
also removed the need for complex cognitive processes such as inferences of causal rela-
tions from the learning process. This account, therefore, became highly popular as an expla-
nation for goal-directed behavior in agents that seemed to lack these capacities.
A great example is a notion that the development of goal-directed behavior in
infants starts with exploratory/random movements (“motor babbling”) (e.g., Meltzoff &
Moore, 1997; Demiris & Dearden 2005). It is assumed that this behavior starts as random
activation of motor commands (activation patterns in the brain that lead to overt behaviors
of limbs) that produce certain behavioral outcomes (e.g., an arm shooting up, a leg kick-
ing). After enough repetition, the specific motor programs would become associated with
their outcomes, which could lead to activation of motor programs in the opposite direc-
tion: Thinking of the behavioral outcomes could then lead to the activation of the relevant
motor programs, rendering goal-directed behavior possible just based on associations that
developed over times. Indeed, spontaneous bidirectional associations between actions and
outcomes have been demonstrated in in 9-month-old infants (Verschoor et al., 2010).

Such a link between the representations of perceived outcomes and motor pro-
grams has become a common explanation for motor mimicry (Iacoboni, 2009). It has even
been suggested to provide an explanation of the origin of mirror neurons: single neurons
fire both when a specific action is executed and when the result of this action is perceived.
It was suggested that such neurons are not innate, but rather the result of this rudimen-
tary learning process that would create representations of goal-directed actions. Motor
programs and outcomes share a common code during such learning stage, which could be
measured even at the single neuron level (Heyes, 2010).

The most likely candidate for such a rudimentary process would be Hebbian learn-
ing. Based on Donald Hebb’s (2002) notion: “neurons that fire together, wire together.”,
mere coactivation of representations would be enough to support this form of learning.
From this perspective, the only features necessary for goal-directed action would be the
ability form bidirectional associations between actions (or motor programs) and the out-
comes (or sensory effects) they produce, simply based on the fact that the two always occur
in temporal contiguity.

In sum, then, the ideomotor theory provides a mechanistic account for how people
and other organisms can acquire the (rudimentary) capability to act in a goal-directed man-

10