I have long held (Evans Reference Evans and Claxton1980) that while we can introspect on our mental experiences, we have no access to the processes which underlie our behaviour, and I have equally long held the view (Evans Reference Evans1989; Wason & Evans Reference Wason and Evans1975) that strategy reports frequently reflect confabulations. Crossing the disciplinary divide with philosophy, this accords well with Carruthers' claims that (a) we have no introspective access to our propositional attitudes and (b) that we may have an illusion of conscious control resulting from applying our “mindreading” abilities to ourselves as well as others. Although Carruthers seeks to reconcile his massively modular view of the mind with dual-process theories of reasoning (see also Carruthers Reference Carruthers2006), it is not clear how many tenets of the standard theory he would accept. I therefore examine his “mindreading is prior” argument with respect to this standard dual-process approach.

Dual-process theorists propose that humans have two distinct forms of cognitive processing: one fast, automatic, and high capacity (Type 1), and another slow, controlled, and low capacity (Type 2) (Evans Reference Evans2008). It is commonly assumed that these two modes of thought reflect two distinct forms of knowledge: implicit and explicit (Carruthers does not appear to endorse this distinction). Implicit knowledge may be encapsulated in cognitive modules or acquired from associative or procedural learning. Explicit knowledge has some propositional format and can be “called to mind.” Neuroscientific evidence strongly supports the existence of dissociable implicit and explicit memory systems (Eichenbaum & Cohen Reference Eichenbaum and Cohen2001). Intuitive and reflective judgments are assumed to reflect access to these two forms of knowledge and to comprise two distinct cognitive systems. This can explain, for example, why people seem to possess implicit attitudes and stereotypes, which may conflict and compete with their explicit social attitudes (Smith & DeCoster Reference Smith and DeCoster2000).

There is, however, a point of discomfort within dual-process theory to which the current target article is highly relevant. Since theorists claim that there are two systems with functional control of behaviour, System 2 – the “conscious” one – cannot be epiphenomenal. On the other hand, as Carruthers correctly states, the evidence that people frequently confabulate explanations for their behaviour is overwhelming. Evans and Over (Reference Evans and Over1996) were clearly struggling with this issue, when they stated (p. 160) that “we do not regard explicit thinking as simply serving to rationalise behaviour, and believe that decisions and actions can result from explicit processes.” Does this mean that Type 2 processes sometimes really control a response and at other times confabulates an explanation for a Type 1 response? Could we not instead argue that all strategy reports are self-interpretations in just the same way as Carruthers argues for propositional attitudes?

The problem as I see it (Evans Reference Evans2008; Reference Evans, Evans and Frankish2009) is that it is a mistake to use consciousness in the definition of the distinction between Systems 1 and 2. A more satisfactory definition is that the latter requires access to central working memory, whereas the former does not. For sure, this implies that something about a Type 2 process is conscious, as the contents of working memory tend to reflect in conscious experiences (Andrade Reference Andrade and Andrade2001). However, it can still be the case that (a) most of the workings of System 2 are unconscious – for example, the processes that direct our current locus of attention, and those that retrieve memories relevant to the current context, and (b) that we lack introspective access to the nature of System 2 processing. (Note that [b] may be at odds with Carruthers' claim that we do have a conscious mind.) Introspection provides no access to Type 1 processes, that either (preconsciously) pass information into working memory for further processing, or by-pass it altogether (Evans Reference Evans, Evans and Frankish2009). Because System 2 uses working memory, something about it can be introspected – the locus of our attention, for example, or relevant perceptual and emotional experiences. These will be used as inputs to the self-interpretation process that Carruthers discusses, together with contextual knowledge and any current goals we are pursuing. Because we generally have good theories of our own behaviour, we can often produce veridical reports. On other occasions, we confabulate, but I suggest that mental processes in either case are exactly the same.

I suggest that it is unhelpful to describe dual-process theories as contrasting conscious and nonconscious processing, and that the theory in no ways rests upon either introspective access or any notion of conscious control. Dual-process research methods depend instead on the assumptions that (a) Type 1 and Type 2 processing are qualitatively and measurably distinct in their characteristics and outputs, and (b) that the balance between the two forms of processing is related to a number of well-defined variables. These include cognitive ability, motivation, time available for reasoning, and the presence of competing or distracting tasks. However, we still need some account of why confabulation is common and why we do feel as though we have conscious control of our behaviour. This is where I find Carruthers' “mindreading is prior” argument helpful. If our brains have an in-built folk psychology for understanding the minds of others, why would the same system not interpret our own behaviour?

In summary, Carruthers' account helps to resolve a problem for standard dual-process accounts of higher cognition. People, he argues, are conscious of perceptual and quasi-perceptual experiences, the latter being formed by mental rehearsal that involves inner speech and/or imagery. It is precisely such processes – that involve working memory – that cause people to confabulate. People do not confabulate accounts of Type 1 processes such as those underlying recognising a face, or understanding a sentence; they do confabulate explanations for Type 2 processes such as decision making. As Carruthers says, people will also confabulate on the basis of behavioural data, but they omit reference to relevant Type 1 processes when they do so. For example, people never refer to a number of well-documented perceptual, cognitive, and social biases in their verbal reports, for folk psychology knows nothing of such matters. This also explains our chronic tendency to overly attribute conscious reasons for actions in ourselves and in others.