Exp Brain Res (2000) 133:55–65 Digital Object Identifier (DOI) 10.1007/s002210000400 REVIEW Lisa Koski · Tomáö Paus Functional connectivity of the anterior cingulate cortex within the human frontal lobe: a brain-mapping meta-analysis Published online: 10 May 2000 © Springer-Verlag 2000 Abstract A database of positron-emission-tomography testable hypotheses about functional and effective con- studies published between January 1993 and November nectivity within the human frontal lobe. 1996 was created to address several questions regarding the function and connectivity of the human anterior cin- Key words Anterior cingulate · Positron emission gulate cortex (ACC). Using this database, we have previ- tomography · Meta-analysis · Frontal cortex ously reported on the relationship between behavioural variables and the probability of blood-flow response in distinct subdivisions of the ACC. The goal of the current Introduction analysis was to discover which areas of the frontal cortex show increased blood-flow co-occurring consistently The anterior cingulate cortex (ACC) is a heterogeneous with increased blood-flow in the ACC. Analyses of the region in terms of its cytoarchitecture, connectivity and frequency distributions of peaks in the ACC and the re- function. Since Brodmann’s initial classification of areas maining frontal cortex (FC) yielded several important 24, 25 and 32, evidence has emerged for the further sub- findings. First, FC peaks in the precentral gyrus, superior division of the ACC into limbic and paralimbic regions frontal gyrus, middle frontal gyrus, inferior frontal gy- (Sanides 1964). Anatomists working in both the human rus, medial frontal gyrus and orbitomedial frontal gyri and monkey brain have identified several cytoarchitec- were more frequent in subtractions that also yielded a turally distinct regions within area 24, denoted as 24a, peak in the ACC than in those that did not yield an ACC 24b and 24c (human: Petrides and Pandya 1994; Vogt peak. Second, regional differences in the frequency dis- et al. 1995; Zilles et al. 1995; monkey: Barbas and Pan- tribution of these FC peaks were observed when the dya 1989; Matelli et al. 1991; Petrides and Pandya 1994; ACC peaks were subdivided into the rostral versus cau- Vogt and Pandya 1987; Vogt et al. 1987). Different corti- dal ACC and supracallosal versus subcallosal ACC. cal zones can also be distinguished along the rostral- Peaks in the precentral gyrus and in the vicinity of the caudal plane of the human ACC. The most obvious of supplementary motor area were more prevalent in sub- these is the gigantopyramidal field located in the caudal- tractions with co-occurring peaks in the caudal than with most portion of area 24c, in the dorsal bank of the cingu- the rostral ACC. Peaks in the middle frontal gyrus were late sulcus (Braak 1976). Subtle variations in cytoarchi- more frequent in subtractions with co-occurring peaks in tecture define further subdivisions of area 24 as it curves the paralimbic part of the supracallosal ACC, relative to around rostral to the genu of the corpus callosum (Vogt the subcallosal or limbic supracallosal ACC. These ob- et al. 1995). Cytoarchitectural differences along the ros- servations are consistent with known differences in the tral-caudal plane are also seen in area 32, a transitional anatomic connectivity in these cortical regions, as de- cortex that shares many features of its neighbouring neo- fined in non-human primates. Further analyses of the in- cortical areas. Sarkissov’s brain map (Sarkissov et al. fluence of behavioural variables on the relationships be- 1955) emphasises the resulting gradations from caudal to tween the ACC and other regions of the frontal cortex rostral to subcallosal area 32 in the use of the labels suggested that this type of meta-analysis may provide 32/8, 32/9, 32/10 and 32/12 (see Fig. 1). The anatomic and functional specificity of these sub- divisions is illustrated by the presence of one or more L. Koski · T. Paus (✉) cingulate motor areas in the caudal portion of area 24 Montreal Neurological Institute, 3801 rue University, Montreal, Quebec H3A 2B4, Canada (Luppino et al. 1991; Shima et al. 1991) that contain di- e-mail: [email protected] rect topographic connections to the spinal cord (Dum Tel.: 514-398-8504, Fax: 514-398–1338 and Strick 1991; He et al. 1995). Functional neuroimag- 56 Fig. 1 Cytoarchitectonic divi- sions of the frontal cortex ac- cording to Sarkissov et al. (1955). Note the ventral to dor- sal differentiation of parts a, b, and c within area 24. Also note that the paralimbic anterior cin- gulate area 32 is labelled as a transitional cortex that shares some of the features of medial areas 8, 9, 10 and 12 ing studies in humans support the existence of regions ten used when referring to simple inter-regional correla- within the ACC that are differentially involved in manu- tions, whereas “effective” connectivity refers to a more al, oculomotor and vocal responses (Paus et al. 1993; causal relationship between two systems such that one Picard and Strick 1996). neural system exerts direct influence over the other Although the empirical evidence is less striking, a (Friston et al. 1996). Second, new experimental methods distinction has also been made between the functions of that combine transcranial magnetic stimulation with posi- the supracallosal, or “cognitive”, and the subcallosal, or tron emission tomography (Paus et al. 1997), functional “visceral”, parts of the ACC (Mayberg 1997; Vogt et al. magnetic resonance imaging (Bohning et al. 1998) or 1995). In the monkey, these regions show somewhat dis- electroencephalography (Ilmoniemi et al. 1997) promise tinct patterns of connectivity. Areas 24 and 32 are pre- to demonstrate effective connectivity, without the behav- dominantly supracallosal and are interconnected with ioural confound inherent in traditional imaging approach- dorsolateral frontal areas (Barbas and Pandya 1989; es. Finally, with the growing body of PET experiments, it Morecraft et al. 1993), which have been implicated in is now possible to analyse large sets of data in which cognitive functions such as attention (Mesulam 1981; multiple brain regions may show activation peaks within Posner and Petersen 1990) and working memory (Gold- the same subtraction. The present study demonstrates man-Rakic 1987; Petrides 1996). Subcallosal area 25 how these data may be used to establish which areas of is interconnected with posterior orbitofrontal area 13 the brain show changes in activity that co-occur in a con- (Barbas and Pandya 1989) and has been implicated in sistent manner with changes in a particular region of in- the control of respiration, blood pressure and other auto- terest, i.e. across studies and across various behavioural nomic functions (see Kaada 1960 for a review). manipulations. Our knowledge of the connections between the ACC We recently compiled a database of 107 published and other parts of the frontal cortex has been restricted to PET studies and were able to demonstrate a quantitative non-human primates (Barbas and Pandya 1989; Bates and relationship between increases in the rate of behavioural Goldman-Rakic 1993; Luppino et al. 1993; Morecraft and response across tasks and increases in activation specific Van Hoesen 1992, 1993; Vogt and Pandya 1987), as the to the caudal part of the ACC (Paus et al. 1998). We also invasiveness of anatomical tracing techniques precludes found that increases in task difficulty affected cingulate their use in humans. Research into the connectivity of the blood-flow in a regionally differentiated way, with the human brain relies on the recent development of new greatest increases seen in the paralimbic part of the su- techniques using functional neuroimaging. First, statisti- pracallosal section of the ACC, smaller increases seen in cal methods such as principal components analysis the limbic part of the this section, and negligible changes (Friston et al. 1993), structural equation modelling seen in the subcallosal portion of the ACC. (Horwitz et al. 1995; McIntosh et al. 1994) and multivari- For the present study, we looked for evidence of func- ate partial least squares (McIntosh et al. 1996, 1999) have tional connectivity between the ACC and other parts of been used to evaluate co-activations of brain regions dur- the frontal cortex by returning to our database and as- ing the performance of cognitive tasks. Depending on the sessing which areas of the frontal cortex contained acti- approach, different aspects of inter-regional interactions vation peaks in the same subtractions that produced are emphasised: the term “functional” connectivity is of- peaks in subregions of the ACC. This approach was vali- 57 dated by the confirmation of strong predictions of func- Difficulty Level for the paralimbic, rather than limbic, portions of tional connectivity between the caudal ACC and the pri- the supracallosal cingulate. Only those subtractions containing at least one peak in the mary and supplementary motor areas. Our approach also ACC and one peak elsewhere in the frontal cortex were included suggested connections between supracallosal cingulate in the dataset for the present study. Therefore, the first step of this and the middle frontal gyrus. Finally, we investigated analysis was to conduct a set of tests to determine whether the re- whether these relationships between subregions were de- lationships between the ACC subdivisions and the behavioural variables observed in the earlier study remained significant.