Mitchell et al., in their interesting and provocative treatment of human associative learning, link propositional learning with awareness. Specifically, they conclude that “The available evidence largely supports the propositional approach to learning. Thus, learning does not take place outside of awareness; it requires cognitive resources, and it is affected by verbal instructions, rules, and deductive reasoning processes” (sect. 8, para. 2). Our research strongly supports this conclusion regarding human autonomic classical conditioning, as the authors noted. Therefore, our comments focus on the role of awareness of the CS-US (conditioned stimulus-unconditioned stimulus) contingency in human autonomic classical conditioning.
Confirming findings
Research from our laboratories is even more supportive than Mitchell et al. indicate of the position that human autonomic classical conditioning is propositional and requires significant cognitive resources. As they point out, much of our research has embedded conditioning within a distracting cognitive masking task with the result that subjects who become aware of the CS-US contingency successfully condition, and subjects who remain unaware do not. They then raise the “devil's advocate” possibility that an automatic associative-link mechanism might exist, but the cognitive load imposed by a masking task may act to prevent conditioning by reducing the degree to which the CS and US are processed, and hence reduce the input to the link mechanism. However, in the cognitive masking task that we have used, the CS is specifically the focus of the subject's attention (it must be judged on some dimension and remembered), and expectancy of the US is constantly reported by the subject.
For instance, subjects were presented with a series of tones on each trial, one of which was the CS, and were required to determine which tone matched a preceding tone in pitch. Subjects indicated expectancy or non-expectancy of the US by pressing a series of buttons continuously during the tones (Dawson & Biferno Reference Dawson and Biferno1973). Hence, the failure to condition without awareness cannot be attributed to the failure to attend and process the CS and US. We have found this necessity of awareness not only with typically used CSs, such as tones or colored lights, but also with odor CSs, stimuli often thought to be capable of eliciting conditioned emotional responses without a supporting conscious memory (Marinkovic et al. Reference Marinkovic, Schell and Dawson1989). Another line of evidence in support of the importance of cognitive resources not mentioned by Mitchell et al. is that performance on a secondary reaction time task performed during the conditioning session shows deterioration during the CS exactly when the conditioned responses are elicited (Dawson et al. Reference Dawson, Schell, Beers and Kelly1982).
Perplexing exceptions
Although we are strong advocates of the position that human classical conditioning cannot occur without awareness of the CS-US relationship, we find ourselves in the unusual position of noting that there may be limitations and exceptions to this general proposition. First, Dawson and Furedy (Reference Dawson and Furedy1976) reviewed evidence in support of a “necessary-gate” hypothesis that included the following propositions: (1) contingency awareness is necessary, but not sufficient, for human autonomic classical conditioning (e.g., researchers in this field often observe participants who are aware of the CS-US relationship, give strong URs [unconditioned responses], but do not show conditioning); (2) the degree of contingency awareness has a gate, but not analog, relation to the strength of the conditioned response (i.e., once a critical minimum level of awareness has developed, there is little or no relationship between the strength of the conditioned autonomic response and the degree of accuracy or certainty in the learned proposition); and (3) contingency awareness is not necessary for performance of a response that has been previously conditioned. This second proposition is contrary to the position of Mitchell et al. in section 3.1 that the strength of the CR is related to the “strength of belief” in the CS-US contingency.
Second, conditioning using biologically prepared CS-US relations (e.g., angry faces associated with aversive events) may be possible without awareness of the presence of the CS, as demonstrated by Öhman and his colleagues (see Esteves et al. Reference Esteves, Parra, Dimberg and Öhman1994). Esteves et al. (Reference Esteves, Parra, Dimberg and Öhman1994) used as CSs in a discrimination conditioning paradigm pictures of angry or happy faces that were backwardly masked to prevent conscious awareness of their presence. These CSs were paired with an electric shock US. Conditioned skin conductance responses were observed following the angry face but not the happy face CS (see also Öhman & Soares Reference Öhman and Soares1998; Morris et al. Reference Morris, Buchel and Dolan2001). These results indicate that humans can be conditioned to a stimulus they do not consciously perceive, if that stimulus is evolutionarily prepared to be associated with the US. However, when biologically prepared CS-US relations were embedded in a distracting cognitive masking task, which ensured conscious perception of the individual CS, autonomic conditioning only occurred among aware subjects and only after they became aware (Dawson et al. Reference Dawson, Schell and Banis1986).
Thus, there is conflicting evidence of whether autonomic conditioning can occur without awareness when biologically prepared CS-US relations are involved. When biologically prepared CSs are backwardly masked and subjects are presumably unaware of the CSs' presence, there is evidence of unaware conditioning. When the CSs are part of a distracting masking task, one which ensures that the CSs are perceived and discriminated, there is no evidence of unaware conditioning.
A possible explanation of these conflicting results is that when higher cortical processes become involved, as when the CSs are perceived during a distracting cognitive task, propositional learning is the dominant force in controlling autonomic conditioning. Propositional learning will be dominant even if an incorrect proposition has been tacitly learned – that the CS has no particularly predictive value. However, when these higher cortical processes concerning the CSs are prevented from occurring, as when the CSs are effectively backwardly masked, and when they are biologically prepared, then the automatic associative learning processes become the dominant force. Therefore, under conditions where the CSs are perceived, propositional learning is necessary for human autonomic classical conditioning.
