Higher Order Conditioning of the Eyeblink Reflex
Classical conditioning of the eyeblink response is a well-established model of cerebellar-dependent associative motor learning. In this paradigm an initially neutral conditioned stimulus (CS - e.g. an auditory tone or light) that elicits no response is paired with an aversive unconditioned stimulus (US -e.g. an air puff to the eye ) that elicits an innate reflex blink response. A CS paired with the US at a fixed latency comes to elicit a predictive blink, if the CS begins sufficiently in advance of the US. A predictive blink is such that eyelid closure is maximal at the time of expected air puff delivery. Physical and pharmacological manipulation experiments have determined that this behaviour is dependent on the cerebellum, and theoretical accounts based on the structure of the cerebellum have proposed that it represents a mechanism for this type of learning (Marr, 1969; Albus, 1971).
The cerebellum has two distinct afferent pathways. The first are the mossy fibres which have diverse origins from various nuclei, including sensory structures. The second pathway is that of the climbing fibres that originate from the inferior olivary nuclei. In Marr’s proposal, the climbing fibres represent instructive signals for plasticity, and the mossy fibres carry "context" information on which plasticity operates so that the contextual signals can activate learned movements. Studies in the eyeblink paradigm have confirmed that the climbing fibre activity is correlated with the US and that they encode error signals. Unexp ected aversive stimuli cause an increase in activity of climbing fibres, whereas the absence of a predicted aversive stimulus causes a decrease in the activity.
Recent evidence has revealed that after training the climbing fibres become responsive to the conditioned stimulus, firing shortly after its onset. This suggests that the conditioned stimulus now has a prediction error value and could itself become a conditioned reinforcer. To test this experiments were devised in both rabbit and human subjects using second-order conditioning schedules, where conventional pairings of a CS1 and the US were interleaved with pairings of a different CS2 and the CS1 (but no US). Consistent results were obtained from both sets of experiments: some subjects learned to respond to the CS2, but more weakly than to the CS1. Strikingly the responses to the CS2 were more closely tied to the onset of CS1 (the time at which climbing fibres respond after training) rather than to the time of expected US. These findings confirm that the CS1 can become a conditioned reinforcer, but have implications for the prevailing view that eyeblink conditioning as a paradigm is primarily about the timing of the response. The relevance of the experimental eyeblink paradigms to natural behaviour is discussed.