Chapter 5: Automaticity in social interaction
5.1 From models of the Speaker to models of dialogue: the argument so far
5.3.6 Are language processes System-1 operations?
In a later paper, Mercier & Sperber present their own version of the Addressee- Speaker feedback loop: ‘For communication to be stable, it has to benefit both senders and receivers; otherwise they would stop sending or stop receiving, putting an end to communication itself’ (Mercier & Sperber, 2011: 60). Although their focus is on the influence of the Speaker rather than the Addressee, their proposal
nonetheless describes one of the possible mechanisms whereby the two systems envisaged in dual-process theory may help to ensure sustained – and therefore successful – communication. However, there is a further reason why their discussion of communication in a dual-process context is of interest here. Their hypothesis appears to be one of the few recent dual-process accounts that focuses on language use specifically and in detail: a gap in the narrative that can perhaps be explained by the fact that linguists themselves have not been heavily involved in the development of dual-process accounts. This is not to say that the subject is entirely ignored in current dual-process models, but it figures more as an incidental than a dominant theme.
For instance, Carruthers (2009: 118-20) proposes a model of System-2 operation in which beliefs and goals emerge into consciousness as a result of cyclical operations of System 1, among which he includes the mental rehearsals of bodily and speech actions. In the latter case, a wide variety of System-1 subsystems are involved, including perceptual systems, bodily states, belief/desire generating systems and ‘practical reason systems’ as well as language production and comprehension. The output is ‘inner speech’, a set of (mainly auditory) representations of the form the rehearsed speech item would take which, in a process reminiscent of that proposed by Levelt, the language comprehension sub-system can then process. Language also plays a key role in supporting the tripartite model of the mind proposed by
Stanovich, in which the architecture of human rationality (i.e. System 2) is restructured into two levels, ‘algorithmic’ (calling on fluid intelligence) and ‘reflective’ (calling on goals/ epistemic values). A crucial link between the two – cognitive decoupling, or the preparation of representations for use in ‘what-if’, hypothetical reasoning – is performed at the algorithmic level, and draws heavily on
language for its functioning. ‘Decoupling…’ Stanovich comments (2009: 63), ‘is a cognitively demanding operation. Any mindware that can aid this computationally expensive process is thus immensely useful, and language appears to be one such mental tool. Language provides the discrete representational medium that greatly enables hypotheticality to flourish as a culturally acquired mode of thought. For example, hypothetical thought involves representing assumptions, and linguistic forms such as conditionals provide a medium for such representations.’
Although Stanovich is focussing here on the use of language as a medium for System-2 thinking, he had earlier included it, along with other domain-specific processes such as perception, in the range of automatic systems that make up System 1 (Stanovich, 2004). Here he is not alone. There seems to be agreement on the role of language as a mediator/enabler for cognitive processing at the conscious (System- 2) level: as Frankish suggests, ‘the conscious mind [is] a language-dependent virtual machine’ (Frankish & Evans, op cit: 22). But it is also clear that some language processes themselves are System-1 operations, working according to the standard System-1 pattern of processing: fast, effortless, automatic, undemanding of working memory, with only the final output accessible to conscious awareness and
manipulation. This is the clear implication of Frankish’s explanation of the distinction between personal reasoning (i.e. System-2 mental processing) and the processes underpinning it:
Although personal reasoning itself is conscious, the beliefs and desires that motivate it typically will not be. Actions can be consciously performed even if we do not consciously reflect on the reasons for performing them… Most of our behaviour is generated without the involvement of personal reasoning. Think about the actions involved in such everyday activities as driving a car,
holding a conversation [my emphasis], or playing sports. These are intelligent
actions, which are responsive to our beliefs and desires (think of how beliefs about the rules of the game shape the actions of a football player), and a great deal of complex mental processing must be involved in generating them. Yet, typically, they are performed spontaneously with no prior conscious thought or mental effort. Indeed, giving conscious thought to such activities is a good way to disrupt their fluidity.
Something like automaticity of language output also seems to be implied in Smith & DeCoster’s discussion of time constraints on cognitive processing (see Chapter 5.3.4): ‘Responses that are made quickly or when the perceiver is busy or distracted likely will be controlled by the associative system…’ Evans, too, allocates to Type/System 1 the functioning of the ‘perceptual, language [my emphasis] and memory systems that provide content for working memory form’ (op cit: 48), although he acknowledges that this can present problems where language itself is concerned3.
Effortlessness; speed; spontaneity … These are features of naturally-occurring
speech that we have met before. They characterise the speech on ‘something close to autopilot’, illustrated in Chapter 3 by the examples of phatic communion. There are, for instance, few signs of careful reflection in the garden-lovers’ dialogue (Chapter 3.2, Example 3) and none in my exchange with the weary passer-by (Chapter 3.4, example 7). And this effortlessness is not entirely the result of the automaticity ascribed by Levelt to the lower-level processes in his model; there is a message-level element in it as well. Responding to the elicitatory eye-contact made by their
(future) Addressees, the interactants in these examples decide to talk; they therefore have to decide what to say. And these decisions are made against a background of many other demands on their cognitive resources, notable among which is the simple need to keep progressing through a public space.
Taken together, these circumstances seem to indicate that the lower speech
production levels are not the only candidates for System-1 status in the dual-process framework; the message level is also a candidate, and not just because it can draw on stock phrases easily retrievable from long-term memory4. Both the intention to speak and the selection of appropriate content appear to be at least partially
supported by System-1 processes as well: high-speed, automatic processes that can handle large amounts of data in parallel and, via a ‘learn slow, retrieve fast’ memory system, can respond to even a limited stimulus by recalling a mass of associated
3
Dual-system theorists, he comments, generally address the issue of modularity by siting modules within System 1 due to their manner of functioning: rapid, unconscious, automatic. ‘But this now seems problematic to me. Language appears to be modular in the strong sense but is (a) uniquely human [therefore a System-2 candidate] and (b) is an essential prerequisite for type 2 processing, facilitating higher order and explicit representations of knowledge’ (ibid: 40).
4
Although these certainly play a part in supporting ‘effortless’ verbal interaction; see Chapter 6 for a discussion of speech routines and routinisation.
material. And if, for casual speech at least, the Conceptualizer operates according to System-1 principles, do we here have the essential link in the Addressee-Speaker feedback loop which enables a Speaker to retain Addressee interest by producing relevant utterances – all the while complying with the heavy time pressures of normal dialogue?
At the moment, direct evidence for such a link from dual-process work seems thin on the ground. However, as I hope to show in Chapter 6, one can approach a dual- process account of speech production from other perspectives and, in the meantime, research in the field of neuroscience offers valuable insights into the crucial
mechanisms supporting dialogue: those underpinning social interaction itself. These developments are outlined in the following section.