Exploiting redundancy for flexible behavior: unsupervised learning in a modular sensorimotor control architecture

Autonomously developing organisms face several challenges when learning reaching movements. First, motor control is learned unsupervised or self-supervised. Second, knowledge of sensorimotor contingencies is acquired in contexts in which action consequences unfold in time. Third, motor redundancies must be resolved. To solve all 3 of these problems, the authors propose a sensorimotor, unsupervised, redundancy-resolving control architecture (SURE_REACH), based on the ideomotor principle. Given a 3-degrees-of-freedom arm in a 2-dimensional environment, SURE_REACH encodes 2 spatial arm representations with neural population codes: a hand end-point coordinate space and an angular arm posture space. A posture memory solves the inverse kinematics problem by associating hand end-point neurons with neurons in posture space. An inverse sensorimotor model associates posture neurons with each other action-dependently. Together, population encoding, redundant posture memory, and the inverse sensorimotor model enable SURE_REACH to learn and represent sensorimotor grounded distance measures and to use dynamic programming to reach goals efficiently. The architecture not only solves the redundancy problem but also increases goal reaching flexibility, accounting for additional task constraints or realizing obstacle avoidance. While the spatial population codes resemble neurophysiological structures, the simulations confirm the flexibility and plausibility of the model by mimicking previously published data in arm-reaching tasks.     

What might philosophy of science look like if chemists built it?

Had more philosophers of science come from chemistry, their thinking would have been different. I begin by looking at a typical chemical paper, in which making something is the leitmotif, and conjecture/refutation is pretty much irrelevant. What in fact might have been, might be, different? The realism of chemists is reinforced by their remarkable ability to transform matter; they buy into reductionism where it serves them, but make no real use of it. Incommensurability is taken without a blink, and actually serves. The preeminence of synthesis in chemistry could have led philosophers of science to take more seriously questions of aesthetics within science, and to find a place in aesthetics for utility. The necessary motion twixt macroscopic and microscopic views of matter in modern chemistry leads to the coexistence of symbolic and iconic representations. And in another way to the deliberate, creative violation of categories.     

Global-local orientation congruency effects in visual search

In a visual search task, subjects had to decide which of 2 possible target letters was presented among 12 distractor letters. The 13 characters were arranged to form a global Navon-type letter. The global letter and the local letters (target and distractors) were independently presented in four different viewer-related orientations. When the global letter and the target were frequently congruently oriented, the response times increased with growing orientation disparity between them. This global-target congruency effect was independent from target identity (Experiment 1), and it diminished when global and target orientations were not correlated (Experiment 2). The results indicate that the orientation of the global letter can be deliberately used in order to facilitate the processing of congruently oriented local targets. The alignment of a spatial frame of reference is discussed as the most probable process underlying this facilitation.     

Response specific temporal expectancy: Evidence from a variable foreperiod paradigm

When a stimulus-response event is frequently paired with a specific foreperiod, response performance for this event is improved after this foreperiod. This phenomenon is referred to as specific temporal expectancy. In four experiments, we investigated whether stimulus- or response-related processing benefits from specific temporal expectancy. In a speeded choice reaction task, different features of the imperative stimuli were frequently paired with foreperiods in such a way that only in some experiments were the responses also frequently paired with foreperiods. Participants revealed evidence for specific temporal expectancy when responses were frequently paired with foreperiods, but not when only the stimuli were frequently paired with foreperiods. We concluded that specific temporal expectancy affects response-related processing.     

Patterns, chunks, and hierarchies in serial reaction-time tasks

The impact of relational structures (i.e., the systematicity of relations between successive items) on incidental sequence learning was investigated in a serial reaction-time (SRT) task while keeping constant the statistical structure. In order to assess the influence of relational structures in stimulus and response sequences separately, the strength of relational patterns in sequences of digits as stimuli and of keystrokes as responses was orthogonally varied. In Exps. 1 and 2, the variation of relational patterns was mainly effective in the keystroke sequence. In Exp. 2, in addition to the variation of relational patterns, the presentation of stimuli was delayed at serial positions that were incongruent with the relational structure. The results show that these incongruent pauses reduced the learning of strongly structured sequences of keystrokes but improved the learning of weakly structured sequences. Experiment 3 suggests that even higher-order relations between elementary patterns are utilized to accelerate responses. The data are interpreted as evidence for the impact of relational patterns, in addition to statistical redundancies, on the formation of chunks. Reasons are discussed for the finding that relational chunking was more pronounced in the keystroke than in the digit sequences. Received: 25 June 1998 / Accepted: 3 December 1998