4. Integrative models
4.4. THE ITERATIVE REPROCESSING (IR) MODEL
The IR model was developed by Cunningham, Zelazo and colleagues (Cunningham, Zelazo, Packer, & Van Bavel, 2007; Cunningham & Zelazo, 2007; Cunningham & Zelazo, 2010). The model is a computational one; it is informed by research in neuroscience, particularly on the development of the prefrontal cortex (PFC) in childhood (Zelazo, Carlson, & Kesek, 2008). Iterative reprocessing refers to the dynamic processing of cognitive information between
36
higher-order and lower-order processes. Lower order processes are conceptually similar to automatic attitudes; they constitute a set of stable representations in memory, activated in response to encountering or imagining an attitude object. The activated attitude is then passed on to higher-order processes which can then incorporate additional variables such as goals or context into a more complex evaluation. The evaluation is then fed back to lower-order processes for a recalculation of the affective response.
Higher-order and lower-order processes are thought to operate dynamically and in concert, with information being passed back and forth between the system networks. Top- down processing then allows for evaluative processes to alter input weights between affective associations and the attitude object. The number of total iterations taking place depends on the complexity of the evaluation and its compatibility with the automatic affective response. Awareness and controllability of attitudes in this context is viewed on a continuum, with the first few iterations being entirely non-conscious and automatic, and becoming gradually more controlled and available to introspection with each subsequent iteration.
The IR model is based on the assumption that brain systems are structured hierarchically (Bunge & Zelazo, 2006). The initial affective response is retrieved automatically; amygdala response can be detected within a few hundred milliseconds (Oya, Kawasaki, Howard, & Adolphs, 2002). Amygdala response is then relayed to the orbitofrontal cortex, which allows for a more a more nuanced analysis of expected rewards and punishments with the actual bodily state (Beer, Heerey, Keltner, Scabini, & Knight, 2003). In many cases, where no conflict is detected, these iterations will be sufficient. However, if the input stimulus triggers an
ambivalent response (strong positive and negative response activation), further iterations will take place in the lateral PFC (Bunge & Zelazo, 2006). Processing associated with the lateral PFC is higher-order rule-selection, which is then used for top-down modulation of the lower-order processes. From this perspective, lower-order processes are not simply single association paths; rather, they constitute complex networks of associations where relative connection weights are constantly updated. Stronger connection weights increase the likelihood of activation in the
37
future. At the same time, some attitudes will remain relatively stable over the course of a person’s lifetime, while others will be constantly updated and re-processed. This approach is also consistent with parallel distributed processing models (Smith, 1996), which suggest that mental representations constitute a series of connections between concepts; these can operate in parallel, and activation strength can vary depending on context.
The iterative-reprocessing approach is valuable from several perspectives. First of all, it makes the attempt to tie in cognitive theory of attitude formation with measurable
neurophysical phenomena. Data from research in neuroscience supports a biological distinction between automatic, affective processes and conscious, deliberative ones. However, it appears that the two processes are best characterised as a dialogue, with information being passed back and forth on a continuum, rather than as two independent cognitive structures. Second, as with the APE model, it suggests that evaluative judgments (‘explicit attitudes’) are motivated by resolving ambivalence (Cunningham et al., 2007); in all other cases, parsimony and resource conservation takes precedence. However, the IR model does not adequately account for studies which suggest that we are, at least to some extent, aware of our implicit attitudes, even in the absence of conflict with explicit attitudes (e.g., Monteith et al., 2001).
The principles of iterative-reprocessing are consistent with neurobiological data, although the majority of the studies have not been designed to test the model specifically. It is also necessary to test its applicability to experimental research contexts; its compatibility with implicit attitude measures currently in use and consistency with observable behaviour. The model also suggests that implicit cognition may be assessed through neuroimaging techniques; this may not be practical in most research contexts but could be intriguing as the purest form of indirect automatic attitude assessment.
Overall, the IR model allows for considerable fluidity in the process of attitude formation and is consistent with the majority of the research-revised definitions of ‘implicitness.’ It supports the re-conceptualisation of dual-process models from categorical to continuous; implicit attitudes are therefore represented as being less available to introspection or more
38
goal-independent than explicit processes. Attitudes can therefore travel along the continuum based on context demands, resource availability and motivation, which can explain the results of studies which have successfully manipulated implicit attitudes using these factors (Hermans, Crombez, & Eelen, 2000; Lowery et al., 2001; Hausmann & Ryan, 2004; Fiedler & Bluemke, 2005; Friese, Hofmann, & Wänke, 2008).
5. Summary
Research from the past decade suggests that implicit and explicit cognition may not be as categorically segregated as has been implied by dual-process models. Implicit attitudes have been demonstrated to be, to an extent, available to introspection (Olson & Fazio, 2001; Mitchell, Nosek, & Banaji, 2003) and susceptible to the effects of context (Lowery et al., 2001; Mitchell et al., 2003) and motivation (Jordan et al., 2007; Hing et al., 2008). Integrative models of cognition take these findings into account to develop models which are concerned not only with the functional and conceptual differences between implicit and explicit attitudes, but also with their interaction and behavioural consequences. Contemporary 'dual-process' models tend to view implicit and explicit cognition on a continuum, with attitudes being more or less
automatic or more or less deliberative depending on what is required for the specific context. The integrative models discussed in this section come from vastly different psychological perspectives and have very different goals. As such, they can be viewed as complimentary, rather than competitive. The MODE model is a continuum model, with attitudes classified on a spectrum from more automatic to more controlled. Attitudes become more controlled when a person is motivated to re-assess their automatic response for potential cost-and-benefit outcomes; however, most daily behaviour is viewed as more automatic than deliberative. The MODE model is most useful in the discussion of social behaviour. The APE model is conceptually closest to dual-process models, as it categorises implicit and explicit processes as qualitatively different. In this model, a discrepancy between implicit and explicit attitudes is thought to
39
indicate a state of dissonance, where an automatic association is consciously rejected as invalid. The approach consolidates much of the recent empirical findings and is therefore useful as a foundation to further research.
The REC model probably takes the most unconventional approach of the models discussed. It suggests that attitudes do not constitute stored associations in memory; rather, both implicit and explicit attitudes are behavioural expressions of the same relational network. However, at this stage the model is mostly conceptual and is in need of supporting research evidence. Finally, the IR model is aimed at qualifying the theoretical approach to dual-process cognition in neurobiological terms. Findings from fMRI studies appear to support the suggestion that attitudes are processed on a continuum, from the relatively automatic to the relatively controlled, and that different regions of the brain become increasingly involved as processing demand changes.