OUP user menu

`Theory of mind' and the prefrontal cortex

Tim Shallice
DOI: http://dx.doi.org/10.1093/brain/124.2.247 247-248 First published online: 1 February 2001

Ten years ago there was little knowledge of the functions of the subregions of the human prefrontal cortex (see Fuster, 1989). Over the last 10 years there has been a great change in the amount of information available on the topic. The major source has been functional imaging. The journals now contain a flood of studies on a variety of different types of task where there are significant differences in activation in complex patterns across prefrontal regions. Thus, a process such as episodic memory (autobiographical memory), previously thought not to be strongly related to frontal functions, now seems to involve five, if not six, anterior regions, which are differentially involved depending on subtle variations in the tasks (see Henson et al., 1999; Lepage et al., 2001). This suggests that a large number of different types of subprocesses are frontally localized.

This flood of information has not, however, led to much closure on the nature of the individual processes involved. The difficulties involved in interpretation may be at least twofold. The first is that the subprocesses involved may be much too abstract to map simply onto the perceptual input or motor output. Secondly, the tasks which activate prefrontal cortex effectively may often involve a number of such subprocesses and there may be no simple observable manifestations of the successful completion of any one stage in normal performance.

If these difficulties do exist, then progress will be considerably quicker if functional imaging of standard tasks can be supplemented by other methods. One would be to employ apparently more complex tasks which, however, are directly linked to more abstract potential subfunctions. The second is to resurrect the older lesion approach, as the effects of a lesion to an underlying process can be more transparent to interpretation than are the patterns of activation obtained when scanning normal subjects. However, to be usefully relatable with imaging studies, an anatomically based group study approach must use a finer localization grain than that typical of 10 years ago. How lesion sites should be clustered to form subgroups is a complex methodological problem. The Toronto–Boston collaboration of Stuss, Alexander and co-workers has been developing a number of alternative patient grouping strategies over the last few years. Their approaches have shown the importance of medial frontal structures in standard frontal tasks such as Wisconsin Card Sorting and Verbal Fluency, previously generally thought to involve principally dorsolateral regions (see Stuss et al., 2000). In the study of Stuss and colleagues in the present issue, they contrasted a bifrontal group with left and right frontal groups, with the bifrontal group overwhelmingly having relatively pure medial lesions and they supplemented the overall analysis by a `hot spot' one (Stuss et al., 2001). Again, the most striking finding involves the medial prefrontal region.

The tasks they were studying involve `theory of mind' or `mentalizing', the cluster of abilities held to be necessary to understand the mental processes of others. Originally discussed by Premack and Woodruff, it was the demonstration by Baron-Cohen and colleagues of their selective impairment in autism that made it plausible that they have a specific brain basis (Baron-Cohen et al., 1985), which more recent studies have directly investigated (see Frith and Frith, 1999). The first study to specifically investigate patients with frontal lesions, that of Stone and co-workers, suggested that inferior medial damage was critical (Stone et al., 1998). These authors also found deficits following dorsolateral lesions but held that working memory demands could be responsible. However, they included no unilateral right frontal patients. A very recent study again found a left frontal effect in understanding `theory of mind' stories (Channon and Crown, 2000).

The current paper uses two main tasks. Both require the patients to make an inference about the location of a ball or a coin they cannot see, from the direction in which experimenters point. In the first task there are two experimenters who point to different places. The patients need to realize that, as they are sitting next to one experimenter, this person cannot see either possible position of the ball. Thus the patient should rely on the pointing of the other experimenter. The frontal patients produced a much higher error rate on the task, so it is now clear that the frontal lobes are important in mentalizing. It appeared that the right frontal lobe was the most critical region; however, the small numbers in that subgroup make this only a suggestion.

The second task involved deception. Here there is only one experimenter who points. In this case, though, she/he always points to the wrong position. It is this task that gives rise to the striking right medial prefrontal difference between the impaired and unimpaired patients. There are certain elements of this task which resemble the older go-no go and alternation tasks that involve prefrontal cortex (e.g. Drewe, 1975) and medial lesions affect voluntary suppression of a very strongly potentiated response as in the anti-saccade task (Paus et al., 1991). However, the current task involves an inference about a specifically human ability: interpretation of pointing by another person. So understanding of deception, which depends on the ability to mentalize, may well be the critical process. The frontal region which appears most critical to the ability to perform this task is the right medial prefrontal region. Strikingly it has been the medial prefrontal cortex, in particular the region around the paracingulate sulcus, that has been involved in reports of mental state in a number of imaging studies. Figure 1 presents a meta-analysis by C. Frith (personal communication) which is an update of their evidence reviewed in Frith and Frith (Frith and Frith, 1999). In some studies (e.g. Fletcher et al., 1995), the activated region is left areas 8, 9 and 10 rather than right, which may reflect the verbal nature of the tasks employed. There appears to be a promising correspondence in localization (within the hemisphere) using two very different methodologies for establishing the material basis of certain complex but critical higher-level cognitive processes.

Fig. 1

Location of peak activations in medial prefrontal regions during tasks where subjects think about their own or others' mental states (carried out by C. Frith, personal communication). This is an updated version of the analysis presented in Frith and Frith (Frith and Frith, 1999).