Components of human reasoning. Numerous kinds of reasoning and component processes have been outlined on theoretical and philosophical grounds. Only few, however, have been based additionally on neuroanatomical considerations. One theoretical framework that joins both the psychological and neurological approaches regards the process of reasoning in terms of manipulating information at different levels of relational complexity. Using this framework it is possible to distinguish among different component processes of reasoning, mapped in a systematic fashion onto prefrontal cortex subregions (Christoff et al., 2001, NeuroImage). In this study, fMRI was used to examine brain activation during 0-relational, 1-relational, and 2-relational problems adapted from the Ravens Progressive Matrices. Effects of increasing difficulty were accounted for separately and excluded from the activation results. The process of relational integration, corresponding to highest levels of relational complexity, recruited selectively the most anterior regions of the prefrontal cortex, the rostrolateral prefrontal cortex (RLPFC; area 10), a result suggesting that the process of integrating multiple relations is associated with the manipulation of abstract, internally generated information.
Subregional analysis of prefrontal function. Knowledge about prefrontal cortex organization at the subregional level could serve as a useful basis for subdividing components of human thought and reasoning. Patient studies throughout the last century have repeatedly linked complex reasoning and problem solving to the prefrontal cortex, but systematic analysis of prefrontal function at the subregional level was only made possible with the advent of neuroimaging methods such as PET and fMRI. A review of the neuroimaging literature on reasoning and episodic memory (Christoff and Gabrieli, 2000, Psychobiology) reported a meta-analysis aimed to differentiate between two subregions, dorsolateral (DLPFC) and rostrolateral (RLPFC), in terms of their contributions to complex cognitive processes. The review suggested the two regions are functionally distinct, with DLPFC being involved when externally generated information is being evaluated, and RLPFC becoming recruited when internally generated information is evaluated. This distinction appeared to be hierarchical in nature: DLPFC is sufficient for the evaluation and manipulation of externally generated information, whereas RLPFC is additionally required during the evaluation and manipulation of internally generated information.
Anterior prefrontal contributions to human cognition. Theoretical conclusions from this review were empirically supported in a follow-up event-related fMRI study (Christoff et al., 2003, Behavioral Neuroscience), showing that RLPFC is specifically involved in the evaluation of internally generated information or information that cannot be readily perceived from the external environment but has to be inferred or self-generated. The findings were also consistent with the hierarchical model of lateral prefrontal organization outlined above, with RLPFC contributing only at the highest orders of cognitive transformations. The anterior prefrontal cortex (or RLPFC) is frequently recruited during complex cognitive tasks across a wide range of domains, including reasoning, long-term memory retrieval, and working memory. One advantage of this characterization of RLPFC function is that it helps to link and explain seemingly disparate findings across multiple cognitive domains and provides a unified account of this regions contribution to human cognition.
Cognitive and neural basis of spontaneous thought processes. The study of human thought has focused upon deliberate, goal-directed mental processes occurring during complex cognitive tasks. A different type of thought processes, however, dominate the internal mental world when no external task is present or when task demands are low: the spontaneously arising undirected thought, yielding the feeling of a ceaseless thought flow. A review of accumulating behavioral and neuroimaging findings (Christoff et al., 2004, Cortex), suggests that such spontaneous thought processes share common cognitive and neural mechanisms with purposeful, task-related thought processes. Furthermore, a novel fMRI study reported in the same paper compared rest to a simple but continuously engaging baseline task of minimal cognitive demands. Rest was associated with greater activation in prefrontal cortex (RLPFC), temporopolar cortex, and parahippocampus, as well as parietal and visual cortical areas. Activation of temporal lobe structures was particularly extensive and robust, suggesting that long-term memory processes may form the core of spontaneous thought. By considering such long-term memory processes as an essential part of thought mechanisms, it may be possible to gain better understanding into spontaneous thought phenomena that have remained unaccounted for until now.
fMRI and complex cognition. Using fMRI as a tool to study complex cognitive processes requires some particular considerations that differ from those involved in the study of other types of processes such as visual perception for instance. Activations from pairwise comparisons in studies of complex cognition are frequently assumed to reflect novel processes uniquely associated with one of the conditions. If there is a systematic difference in response latency between conditions, however, some activations may reflect not novel processes, but longer engagement of processes shared between conditions. The traditional approach models each event-related response using a fixed canonical hemodynamic response function (HRF) and cannot be used to distinguish between these two sources of activation. An alternative approach (Christoff et al., 2001, Neuroimage; also abstracts in Christoff et al., 2001; Christoff, 2002) to modeling event-related responses is to construct the model by convolving the canonical HRF with the response time associated with each trial. Using these two alternative analyses, we observed a dissociation between parietal (BA40, BA7) and prefrontal (lateral BA10) regions, during reasoning and mental rotation tasks. This novel RT-convolved modeling approach can also be used to separate novel processes from shared processes, both across tasks (Christoff et al., 2001, Neuroimage) and across individuals (Shelton, Christoff et al., 2001, manuscript in preparation).
Executive processes at different levels of abstraction. The organization of executive and prefrontal function remains a matter of considerable debate, especially where sub-region and sub-process functions are concerned. The work described here can be put together and further extended within a framework of executive processes organized by levels of abstraction (Christoff, 2003, Neuroimage AbsTrak). Within this framework, prefrontal cortex is viewed as organized in terms of executive control functions at different levels of abstraction so that with increasing level of abstraction, attentional control is implemented by more anterior subregions. Experimental work to test this hypothesis and further extend this framework is in progress (Christoff et al. 2003, SFN Abstracts).
This framework could also allow to distinguish between two different types of thought that humans appear to be capable of, the controlled, deliberate, and effortful thought that seems to be implemented through local interactions within the prefrontal cortex, on the one hand and the spontaneous, more automatic thoughts that occur to us, which could be based on neural influences from subcortical and posterior cortical areas exerted upon the prefrontal cortex.