TICS-1417; No. of Pages 7
Opinion
Degree of automaticity and the prefrontal cortex Hyeon-Ae Jeon and Angela D. Friederici Max Planck Institute for Human Cognitive and Brain Sciences, Department of Neuropsychology, Leipzig, Germany
The dorsolateral prefrontal cortex (PFC), with more anterior areas [Brodmann area (BA) 45, 47, and 10], has been known to be activated as cognitive hierarchy increases. However, this does not hold for highly automatic processes such as first language (L1), where the posterior region (BA 44) is known as the key area for the processing of complex linguistic hierarchy. Discussing this disparity, we propose that the degree of automaticity (DoA) is a crucial factor for the framework of functional mapping in the PFC: the posterior-to-anterior gradient system for more controlled processes and the posterior-confined system for automatic processes. We support this view with previous findings and provide a new perspective on the functional organization of the PFC. Hierarchical processing in the PFC Hierarchical processing refers to the view that processes in the superordinate level control, modify, and modulate processes in the subordinate level over a longer timescale [1–3]. Since the early 1950s, when the concept of hierarchical processing was introduced in relation to the domain of action planning [4], it has been further extended to various cognitive domains such as language and music. Within this scope, action has been characterized as being goal directed and hierarchically structured such that simple action gestalts integrate into progressively more elaborate actions in a hierarchy that involves different processes at the upper and lower levels corresponding to sequences and action chunks, respectively [5]. Language has been characterized as a hierarchical structure; syllables comprise phonemes, words are made up of syllables, phrases comprise words, and then phrases are assembled to build a sentence [3,6,7]. In the music domain, rule-based arrangement of musical sets results in a hierarchical structure; discrete acoustic sounds are assembled into harmonic sequences following a certain arrangement of chord functions [8,9]. Taking these observations together, it is suggested that human behavior is organized in several levels of hierarchy and that hierarchical processing is essential in human cognition [10–15]. Along with the accumulation of findings from various functional neuroanatomy and lesion studies, the neural Corresponding author: Jeon, H.-A. ([email protected]). Keywords: degree of automaticity; hierarchical processing; cognitive control; syntactic hierarchy; BA 44; prefrontal cortex. 1364-6613/ ß 2015 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tics.2015.03.003
basis of hierarchical processing has been found to involve the dorsolateral PFC [16]. Furthermore, the possible system for generating different levels of hierarchy in human cognition, more specifically in cognitive control (see Glossary), has been proposed to reside in the PFC [2,17–26]. One influential perspective on the cognitive framework suggests a topographic organization along a posterior-toanterior axis in the PFC, with progressively anterior regions being involved in higher levels of hierarchical processing than posterior regions (Figure 1A) [23,27–30]. However, it is not always the case that this posterior-to-anterior gradient system stands to reason. The neural underpinning for hierarchical processing in L1 has been shown to involve a posterior region of the PFC; that is, BA 44 (Figure 1B) [31–33]. An increase in syntactic hierarchy leads to an increase of activation in BA 44 rather than to recruitment of more anterior parts of the PFC as suggested earlier in hierarchical processing for cognitive control (Figure 1C) [32]. What, then, is the reason for these disparate representations between hierarchy of cognitive control (the posterior-to-anterior gradient system) and hierarchy in L1 (the posterior-presiding system)? Here, based on two overarching notions – hierarchy of cognitive control and automaticity – we propose a novel conceptual framework that DoA may play a critical role in the functional organization of the PFC: the posterior-toanterior gradient system for more controlled processes with a low DoA and the posterior-confined system for automatic processes with a high DoA. The essential role
Glossary Center-embedded sentence: a linguistic description of a sentence where phrases/clauses are inserted in other phrases/clauses [66,67]. It is described as having the most demanding hierarchical and recursive structure, with longdistance dependencies [35]. Cognitive control: a cognitive mechanism responsible for coordinating or guiding thoughts and behaviors in relation to current goals and intentions [2,27]. Dorsolateral PFC loop: corticostriatal–thalamocortical loops refer to five loops that are topographically organized and functionally segregated with selected cortical areas in the frontal lobe. These are reported to be involved in motor functions with motor and oculomotor loops and in nonmotor functions with dorsolateral, ventral/orbital, and anterior cingulate loops [68]. In particular, the dorsolateral PFC loop is recruited in cognitive aspects, including working memory, planning, rule-based learning, and sequence learning [58,61,69–77]. Early left-anterior negativity (ELAN): a language-relevant event-related brain potential component found between 120 and 200 ms in response to a syntactic phrase-structure error, thereby reflecting initial syntactic structure-building processes [78]. Early right-anterior negativity (ERAN): known as an electrophysiological marker of musical structure building that is mediated by the inferior frontal gyrus and superior temporal gyrus [79,80].
Trends in Cognitive Sciences xx (2015) 1–7
1
TICS-1417; No. of Pages 7
Opinion (A)
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
condition, whereas the second qualification (complexity) is not. This prominent viewpoint, in harmony with our proposal, also emphasizes the value of automaticity rather than complexity. In the following section, we briefly review the influential models of hierarchical processing in cognitive control, demonstrating the organization of the posterior-to-anterior gradient in the PFC. Next, we examine a study comparing hierarchical processing and the role of automaticity in its functional mapping in the PFC for L1, second language (L2), and the non-language (NL) domain. We substantiate our proposal with previous neuroimaging studies in language development and L2. Lastly, we discuss the DoA and its relation to subcortical structures; that is, the basal ganglia.
Posterior
Anterior
Cognive control Badre et al. (2007) Koechlin et al. (1999, 2003)
(B)
First language Friederici et al. (2006) Makuuchi et al. (2009)
Key: Low
Medium
High
(C)
BA 44
% signal change
0.4
0.2
0
–0.2 0
5
10
15
20
Time in seconds 3.09
Friederici et al. (2006) TRENDS in Cognitive Sciences
Figure 1. Posterior-to-anterior gradient system for cognitive control and posteriorpresiding system for first language. In (A) and (B), the approximate locations of the activation foci are provided in a schematic display with spheres centered on the coordinates of peak activation from previous studies. These spheres indicate activation from different levels of hierarchy (red, high; green, medium; blue, low). (A) Subregions in the prefrontal cortex (PFC) selectively contributed to the different levels of hierarchical processing; as the level of hierarchy became higher, activation was observed in a more anterior region of the PFC [23,29,30]. (B) Activation for hierarchical processing in first language (L1) was positioned in the posterior region of the PFC [Brodmann area (BA) 44] even when the level of hierarchy was high, indicating the posterior-presiding system within the PFC for L1 [32,33]. (C) There was a systematic increase of activation in BA 44 as the level of hierarchy in sentences of L1 increased. Different colored lines in the plot of percentage blood-oxygen level-dependent (BOLD) signal change indicate activation from different levels of hierarchy (red, high; green, medium; blue, low). Adapted, with permission, from [32].
of the DoA and its relation to the functional organization of the PFC can be traced back to the identification of two qualifications, ‘newness’ and ‘complexity’ [16]. Goal-directed behavior, which requires the involvement of the PFC, is characterized as being novel and demanding whereby the former qualification (newness) is a sufficient and necessary 2
Hierarchical processing in cognitive control Numerous studies have demonstrated possible frameworks for generating different levels of hierarchies in cognitive control and their topographic maps in the PFC [2,17,18]. Here we discuss two influential theories: temporal abstraction and policy abstraction. The first framework, temporal abstraction, suggests that a significant fraction of cognitive control is based on temporal framing and context, with three different levels in increasing ranking order: contextual, episodic, and branching control [17,28]. Contextual control denotes a synchronic dimension, which means that selecting an action depends on the stimulus response associations according to current context. Episodic control indicates a diachronic dimension, meaning that cognitive control depends on a discrete past event. Branching control refers to a polychronic dimension, or that selecting actions depends on the information conveyed by past events, which are maintained in a pending state and then reactivated later. These three levels form a cascade of top-down selective processes operating along the posterior-to-anterior axis of the lateral PFC, with the contextual, episodic, and branching controls subserved by the posterior lateral PFC, the anterior lateral PFC, and the frontopolar PFC, respectively [17,23,25,30,34]. The other framework, policy abstraction, describes a cognitive hierarchy ranked by different levels of abstractness [27]. Policy abstraction is related to the rules that govern other subrules such that a higher level of policy abstraction has more subcategories or more subordinate representations than a lower one. Multiple levels of abstractness were proposed by summing several constraints for appropriate action selections: the first level for selection between multiple competing actions (least abstract), the second level for selection between different response-level policies (more abstract), and so on up to multiple levels of increasing abstraction [29]. This multistage system of cognitive control is subserved by different areas supporting each level from the left dorsal premotor cortex (the lowest level) to the left anterior dorsal premotor cortex, to the left inferior frontal sulcus, and to the left frontopolar cortex (the highest level) [2,27]. Although the characteristics of the levels are defined to some extent in different ways in these two frameworks, both postulate a similar topographic mapping of the
TICS-1417; No. of Pages 7
Opinion
DoA and language development Automaticity of language processing and its relation to the functional segregation of the PFC can be investigated in
Co
lex mp
Posterior
ity
Jeon & Friederici (2013)
Anterior
Key:
Low, L2 Medium, L2 Medium, NL High, L2 High, NL
(B)
Au
a tom
cit
Posterior
y
r = –0.62 P < 0.01
1.0
% signal change
Anterior
BA 44 BA 10
BA 47
0.5
0.0
–0.5 0.09
0.10
0.11
0.12
0.13
Index for DoA 1.0
1.0
% signal change
Influence of DoA on hierarchical processing Current knowledge of the posterior-to-anterior gradient view of hierarchy in cognitive control (Figure 1A) does not accord with the posterior-presiding system for hierarchical processing of L1 even when the level of hierarchy is high (Figure 1B,C). The processing of syntactically complex and center-embedded sentences reflecting linguistic hierarchy puts Broca’s area, in particular BA 44, at the center of interest [32,33,35–38]. A recent study reconsidered the disparity in the topographic mapping between hierarchy in cognitive control and in L1, investigating the impact of the DoA on hierarchical processing [39]. The crucial idea behind this study originated from the fact that native language processes in adult language users are highly automatic [40] and rely on BA 44 when processing syntactic information of different levels of hierarchical complexity, but on more anterior portions of the inferior frontal gyrus (BA 45) during development [41]. The impact of the DoA on hierarchical processing was investigated in three different domains: L1 for high DoA and L2/ NL for low DoA. L1 sentences varied in their levels of hierarchy (high with center-embedded sentences and low with non-embedded sentences). Both L2 sentences and NL stimuli exhibited three levels of hierarchy (high, medium, and low) following the three cognitive control levels (branching, episodic, and contextual) as the framework of temporal abstraction suggested [28]. Hierarchical processing for the low-DoA systems (L2/NL) resulted in a posterior-to-anterior pattern of activation as the level of hierarchy increased (Figure 2A). For L1, however, syntactic activation was confined in the posterior region (BA 44) even for the highest level of hierarchy (area with black dots in Figure 2B). Further analyses were limited to the same (highest) levels of hierarchical processing for L1, L2, and NL. The results of these analyses indicate that the anterior region (BA 10/47) showed more activation when involved in the process inherent to the low DoA, whereas the posterior region (BA 44) was more activated when automaticity was high (Figure 2B). Thus, the DoA appears to be a crucial factor for the organization of the PFC: the posterior-to-anterior gradient system characterized by more anterior involvement for the more controlled process (L2/NL) and the posterior-dominant system characterized by posterior involvement for the automatic process (L1). Interestingly, BA 44 was involved not only in the lower levels of hierarchy in studies on cognitive control during action [17,23,25,30] but also in the highest level of hierarchy in L1 [39], suggesting supramodal involvement of BA 44 in human cognition, although at different levels of hierarchy. The question here is: to what extent can this proposal of the DoA and its relation to the functional organization of the PFC explain data from previous studies? The answer can be found from studies of L1 development and L2 learning.
(A)
% signal change
activation slanting toward the anterior region of the PFC as the process becomes more effortful due to an increase in the level of hierarchy (Figure 1A).
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
0.5
0.0
r = 0.56 P < 0.01
–0.5 0.09
0.10
0.11
0.12
0.5
0.0
0.13
0.09
Index for DoA
Key:
r = 0.54 P < 0.01
–0.5 0.10
0.11
0.12
0.13
Index for DoA
L1
L2
NL TRENDS in Cognitive Sciences
Figure 2. Degree of automaticity (DoA) and its influence on first language (L1), second language (L2), and non-language (NL) domains. (A) In L2 and NL, a gradient pattern of activation was observed depending on the level of hierarchy, as predicted in the posterior-to-anterior gradient system. Colors indicate activation from different levels of hierarchy. Activation in the low level in NL was found in the posterior region [Brodmann area (BA) 6; not shown in the figure), but only with a more lenient threshold. (B) Correlation between DoA and blood-oxygen leveldependent (BOLD) signals of the high level of hierarchy depicted for the L1, L2, and NL domains. BA 44 for L1 showed more activation when processing centerembedded sentences (high level of hierarchy) than when processing nonembedded sentences (low level of hierarchy). BA 47 for L2 and BA 10 for NL produced more activation in the branching condition (high level of hierarchy) compared with the episodic and contextual conditions (medium and low levels of hierarchy). A scatter plot of each participant’s index for DoA (as the value of the index decreases, DoA increases) as a function of the percentage BOLD signal was obtained from the three peak activations of L1, L2, and NL. Adapted, with permission, from [39].
studies of language development in children. What are the neural correlates of syntactic processing as children develop their language processes? Do children show a developmental shift toward an adult-like posterior-residing system as their DoA in syntactic processing reaches adult level? 3
TICS-1417; No. of Pages 7
Opinion Some imaging studies indicate that language-related brain activation is more extended in children than in adults [42–44]. In a recent fMRI study [45], three groups of children underwent a sentence–picture matching task with different levels of syntactic hierarchical complexity. Only the group of 9–10-year-old children showed an adultlike effect of complexity in the posterior region of the PFC (i.e., pars triangularis and pars opercularis). Another study showed that younger children (5–6 years of age), whose ability for the process was not fully developed, recruited not only the inferior frontal gyrus but also additional areas of the frontal operculum, which is located more ventrally and medially to BA 44 and BA 45 [42]. This additional involvement of the anterior region was attributed to the higher processing demands described by longer reaction times from children than from adults. The low DoA in children’s syntactic processing at around age 6 years has been evidenced in the temporal dynamics of the bloodoxygen level-dependent (BOLD) signal, with a longer time-to-peak BOLD latency than in adults [46]. Moreover, it is reflected in the lack of early left-anterior negativity (ELAN) in event-related brain potential data from the group of 6-year-old children in response to syntactic errors in complex sentences, which is usually elicited by highly automatic structure-building processes in adults [41]. However, children showing adult-like syntactic processing in a more automatic way demonstrate the involvement of the posterior-residing system [44]. In summary, these studies indicate the additional involvement of the anterior region in younger children whose ability for syntactic processing is not fully developed and therefore occurs at a more effortful and controlled level. DoA and L2 The nature of L2 processing, compared with L1, is less automatic and more effortful such that it is considered to be a controlled process [47,48]. L2 processing manifests in increased fMRI responses in the left PFC for switching to a language in which one is less proficient [49] or inhibiting the processing of an irrelevant language [50]. It has been suggested that improvement in proficiency is connected to an increase in automaticity [48] and thus the influence of the DoA on the neural underpinnings of language processing has been studied using L2 with proficient or unskilled learners. Highly proficient L2 learners show an overlap in activation between L1 and L2 whereas those with a low level of proficiency show less overlap between the two languages and, rather, recruit additional brain areas [51]. Low-proficiency groups usually show activation clusters that are widely distributed in the anterior PFC extending to BA 45/46 [48,52–54]. For instance, when Korean sentences and English sentences were presented to native Korean speakers for auditory comprehension, broader regions in the PFC were activated, including the pars triangularis, pars opercularis, and superior frontal gyrus, for English [55]. This pattern was also found in a study where Russians judged the acceptability of German sentences (L2), yielding activation in BA 45/47 [56]. Taking these observations together, the additional involvement of the anterior PFC in L2, corresponding with that observed in NL tasks and less-proficient children’s language 4
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
Box 1. DoA in mathematics and music Two studies were conducted to investigate the neural underpinnings of building a hierarchical structure using arithmetic. In one study, university students were recruited for hierarchical processing with abstract mathematical formulae [81]. This study found significant activation in anterior areas in the PFC including BA 45, 47, and 10. The other study, which used well-trained participants with a high DoA in mathematics, found that such hierarchical processing elicited activation in the posterior PFC as BA 44 (x, y, z coordinates: 45, 9, 30 based on the Ju¨lich histological atlas [82,83]) as well as the bilateral ventral occipitotemporal cortices [84]. Notably, in this experiment simple arithmetic calculations could be automatized in highly proficient participants, consequently being involved in the posterior PFC and compiled in the visual cortex. Therefore, processing of mathematical formulae also seems to implicate the anterior region of the PFC for low DoA and posterior activation for high DoA. In the music domain, a posterior-dominant contribution to the processing of high DoA can be supported by a comparison between two groups of participants based on years of musical training. For example, well-trained and less-trained groups who had taken formal piano lessons identified melodies by watching silent movies of hierarchically organized hand movements playing familiar and unfamiliar melodies, resulting in increased activation in Broca’s area in the unfamiliar melody condition in the well-trained participants only [85]. A compatible finding was reported in eventrelated potential studies; when professional pianists with a minimum of 14 years of musical training imitated silent videos of a right hand playing congruent or incongruent sequences of chords, ERAN was evoked for incongruent chords, reflecting music-syntactic processing [86]. In summary, the influence of the DoA on the functional organization of the PFC may be generalized to cognitive domains outside language.
mentioned earlier, coincides with our proposal that a more anterior region is recruited for the more controlled process with a low DoA. We have discussed the pivotal role of the DoA in the functional dissociation of the PFC mostly in language processing (L2 and children’s language). However, can the idea of DoA be applied to domains other than language? The DoA seems to be able to explain the topographic mapping of the PFC in other cognitive domains such as mathematics and music (Box 1). DoA and basal ganglia Along with the PFC, the basal ganglia implemented within the dorsolateral PFC loop also deserve discussion in terms of the DoA. The DoA has been found to modulate the posterior-to-anterior system even in the subcortical structures, mirroring the gradient pattern of the PFC in the basal ganglia [57]. More specifically, the functional specificity of the caudate nucleus has been discussed in terms of its segmentation into anterior (head) and posterior (body) components, with the head being mediated by more controlled processes compared with the body [58–61]. In a recent high-resolution fMRI experiment, the posterior-toanterior gradient system observed in the PFC was also manifested in the caudate nucleus, more specifically, with the anterior region being activated for the more controlled process [34]. A similar result was observed in semantic/ syntactic ambiguity resolution, with the most complex condition requiring higher-order control processes showing activation in the anterior dorsomedial striatum [62,63]. Similarly, processing a hierarchical rule compared with
TICS-1417; No. of Pages 7
Opinion Box 2. DoA and its influence on basal ganglia The DoA and its relation to the basal ganglia have been substantiated in studies of L2 learning [48,87,88] where the globus pallidus as well as the anterior PFC (BA 47) were activated in more effortful and controlled processes (e.g., translating L1 into L2). Likewise, when a simple condition (naming a picture in L1) was compared with a more controlled condition (naming a picture in either L1 or L2 depending on the cue), for the more controlled condition the caudate nucleus was significantly activated as well as there being anteriorly extended activation along the inferior frontal gyrus and the middle frontal gyrus [89]. Additionally, direct comparisons between L1 and L2 naming also yielded extensive neural activation in the anterior PFC (BA 44, 45, and 47) and the caudate nucleus in L2 naming [89]. Involvement of the basal ganglia in more controlled processes has also been found in motorsequence learning in the action domain. Learning complex sensory–motor sequences, compared with simple sequences, yielded broad activation not only in the frontoparietal area but also in the caudate nucleus and cerebellum [90]. Taken together, the involvement of these subcortical structures linked by corticosubcortcial or corticocerebellar loops may orchestrate the processing of cognitive control depending on the DoA.
Box 3. Outstanding questions Do people with a high level of proficiency resort to the posteriorconfined system in hierarchical processing in the NL domain (i.e., music, action, mathematics)? Is the posterior-to-anterior system transformed into the posteriordominant system as one’s DoA increases in an initially controlled task? Besides the DoA, what can be a possible explanation for functional organization in the PFC? Why do the basal ganglia show the same posterior-to-anterior system pattern as the PFC? What is the advantage of having this mirrored system between the cortical and subcortical regions? How are the temporal dynamics manifested between the highand low-DoA processes in the PFC as well as the basal ganglia? Do behavioral deficits resulting from lesions located at different sites in the posterior-to-anterior system in the PFC vary?
processing a flat rule manifested with significant activation in the anterior portion of the caudate nucleus [64,65]. To summarize, the posterior-to-anterior system also seems to contribute to the topographic organization of the substructures of the basal ganglia, which suggest that the PFC and the basal ganglia work in harmony, forming an interwoven network between the cortical and subcortical areas for the cognitive processes in more complex and demanding tasks (see Box 2 for more examples) [34]. Concluding remarks Several theoretical accounts have been suggested as a framework for generating different levels of hierarchy in cognitive control along with the posterior-to-anterior axis of the PFC [27]. Incorporating this current knowledge, we propose that the DoA may be the critical factor in the functional organization of the PFC: the posterior-to-anterior gradient system with progressively anterior regions supporting a low DoA in more controlled processes and the posterior-confined system with posterior regions being involved in the process with a high DoA in highly automatic processes. To make possible broad generalizations of our proposal, further research should test hierarchical
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
processing in various cognitive domains in groups of participants with varying levels of DoA and also scrutinize the change of topographic mapping in the PFC as DoA increases (Box 3). Acknowledgments The authors thank Kerstin Flake and Andrea Gast-Sandmann for graphics and Elizabeth Kelly and Shameem Wagner for proofreading. This work was supported by a grant from the European Research Council (ERC-2010-360 AdG 20100407) awarded to A.D.F.
References 1 Welford, A.T. (1951) Skill and age; an experimental approach. In On the Nature of Skill. pp. 11–27, Oxford University Press 2 Badre, D. (2008) Cognitive control, hierarchy, and the rostro-caudal organization of the frontal lobes. Trends Cogn. Sci. 12, 193–200 3 Jeon, H.A. (2014) Hierarchical processing in the prefrontal cortex in a variety of cognitive domains. Front. Syst. Neurosci. 8, 223 4 Lashley, K.S. (1951) The problem of serial order in behavior. In Cerebral Mechanisms in Behavior (Jefress, L.A., ed.), pp. 112–146, Wiley 5 Klapp, S.T. and Jagacinski, R.J. (2011) Gestalt principles in the control of motor action. Psychol. Bull. 137, 443–462 6 Fitch, W.T. (2010) The Evolution of Language, Cambridge University Press 7 Berwick, R.C. et al. (2013) Evolution, brain, and the nature of language. Trends Cogn. Sci. 17, 89–98 8 Thompson-Schill, S. et al. (2013) Multiple levels of structure in language and music. In Language, Music and the Brain: A Mysterious Relationship. pp. 289–303, MIT Press 9 Patel, A.D. (2010) Syntax. In Music, Language, and the Brain. Oxford University Press 10 Fitch, W.T. and Martins, M.D. (2014) Hierarchical processing in music, language, and action: Lashley revisited. Ann. N. Y. Acad. Sci. 1316, 87–104 11 Azuar, C. et al. (2014) Testing the model of caudo-rostral organization of cognitive control in the human with frontal lesions. Neuroimage 84, 1053–1060 12 Koelsch, S. et al. (2013) Processing of hierarchical syntactic structure in music. Proc. Natl. Acad. Sci. U.S.A. 110, 15443–15448 13 Udden, J. and Bahlmann, J. (2012) A rostro-caudal gradient of structured sequence processing in the left inferior frontal gyrus. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 367, 2023–2032 14 Bahlmann, J. et al. (2008) Hierarchical artificial grammar processing engages Broca’s area. Neuroimage 42, 525–534 15 Bahlmann, J. et al. (2009) Neural circuits of hierarchical visuo-spatial sequence processing. Brain Res. 1298, 161–170 16 Fuster, J.M. (1989) A theory of prefrontal functions: the prefrontal cortex and the temporal organization of behavior. In The Prefrontal Cortex: Anatomy, Physiology and Neuropsychology of the Frontal Lobe. pp. 157–192, Lippincott Williams & Wilkins 17 Koechlin, E. and Summerfield, C. (2007) An information theoretical approach to prefrontal executive function. Trends Cogn. Sci. 11, 229–235 18 Botvinick, M.M. (2008) Hierarchical models of behavior and prefrontal function. Trends Cogn. Sci. 12, 201–208 19 Zilles, K. and Amunts, K. (2012) Neuroscience. Segregation and wiring in the brain. Science 335, 1582–1584 20 O’Reilly, R.C. (2010) The what and how of prefrontal cortical organization. Trends Neurosci. 33, 355–361 21 Badre, D. and D’Esposito, M. (2009) Is the rostro-caudal axis of the frontal lobe hierarchical? Nat. Rev. Neurosci. 10, 659–669 22 Botvinick, M.M. (2007) Multilevel structure in behaviour and in the brain: a model of Fuster’s hierarchy. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 362, 1615–1626 23 Koechlin, E. et al. (2003) The architecture of cognitive control in the human prefrontal cortex. Science 302, 1181–1185 24 Stout, D. (2010) The evolution of cognitive control. Top. Cogn. Sci. 2, 614–630 25 Koechlin, E. and Jubault, T. (2006) Broca’s area and the hierarchical organization of human behavior. Neuron 50, 963–974 26 Szczepanski, S.M. and Knight, R.T. (2014) Insights into human behavior from lesions to the prefrontal cortex. Neuron 83, 1002–1018 5
TICS-1417; No. of Pages 7
Opinion 27 Badre, D. (2013) Hierarchical cognitive control and the functional organization of the frontal cortex. In The Oxford Handbook of Cognitive Neuroscience, Volume 2: The Cutting Edges. pp. 300–317, Oxford University Press 28 Koechlin, E. and Summerfield, C. (2008) The cognitive architecture of the human lateral prefrontal cortex. In Sensorimotor Foundations of Higher Cognition (Haggard, P. et al., eds), pp. 487–509, Oxford University Press 29 Badre, D. and D’Esposito, M. (2007) Functional magnetic resonance imaging evidence for a hierarchical organization of the prefrontal cortex. J. Cogn. Neurosci. 19, 2082–2099 30 Koechlin, E. et al. (1999) The role of the anterior prefrontal cortex in human cognition. Nature 399, 148–151 31 Friederici, A.D. (2011) The brain basis of language processing: from structure to function. Physiol. Rev. 91, 1357–1392 32 Friederici, A.D. et al. (2006) Processing linguistic complexity and grammaticality in the left frontal cortex. Cereb. Cortex 16, 1709–1717 33 Makuuchi, M. et al. (2009) Segregating the core computational faculty of human language from working memory. Proc. Natl. Acad. Sci. U.S.A. 106, 8362–8367 34 Jeon, H.A. et al. (2014) Functional network mirrored in the prefrontal cortex, caudate nucleus, and thalamus: high-resolution functional imaging and structural connectivity. J. Neurosci. 34, 9202–9212 35 Friederici, A.D. et al. (2011) The neural basis of recursion and complex syntactic hierarchy. Biolinguistics 5, 87–104 36 Poeppel, D. (2014) The neuroanatomic and neurophysiological infrastructure for speech and language. Curr. Opin. Neurobiol. 28, 142–149 37 Poeppel, D. et al. (2012) Towards a new neurobiology of language. J. Neurosci. 32, 14125–14131 38 Fedorenko, E. (2014) The role of domain-general cognitive control in language comprehension. Front. Psychol. 5, 335 39 Jeon, H.A. and Friederici, A.D. (2013) Two principles of organization in the prefrontal cortex are cognitive hierarchy and degree of automaticity. Nat. Commun. 4, 2041 40 Abutalebi, J. et al. (2013) Language proficiency modulates the engagement of cognitive control areas in multilinguals. Cortex 49, 905–911 41 Hahne, A. et al. (2004) Brain signatures of syntactic and semantic processes during children’s language development. J. Cogn. Neurosci. 16, 1302–1318 42 Brauer, J. and Friederici, A.D. (2007) Functional neural networks of semantic and syntactic processes in the developing brain. J. Cogn. Neurosci. 19, 1609–1623 43 Wilke, M. et al. (2005) Comprehensive language mapping in children, using functional magnetic resonance imaging: what’s missing counts. Neuroreport 16, 915–919 44 Nun˜ez, S.C. et al. (2011) fMRI of syntactic processing in typically developing children: structural correlates in the inferior frontal gyrus. Dev. Cogn. Neurosci. 1, 313–323 45 Skeide, M.A. et al. (2014) Syntax gradually segregates from semantics in the developing brain. Neuroimage 100, 106–111 46 Brauer, J. et al. (2008) Temporal dynamics of perisylvian activation during language processing in children and adults. Neuroimage 41, 1484–1492 47 Indefrey, P. (2006) A meta-analysis of hemodynamic studies on first and second language processing: which suggested differences can we trust and what do they mean? Lang. Learn. 56, 279–304 48 Abutalebi, J. (2008) Neural aspects of second language representation and language control. Acta Psychol. 128, 466–478 49 Wang, Y. et al. (2007) Neural bases of asymmetric language switching in second-language learners: an ER–fMRI study. Neuroimage 35, 862–870 50 Rodriguez-Fornells, A. et al. (2002) Brain potential and functional MRI evidence for how to handle two languages with one brain. Nature 415, 1026–1029 51 Higby, E. et al. (2013) Multilingualism and the brain. Annu. Rev. Appl. Linguist. 33, 68–101 52 Sebastian, R. et al. (2011) Meta-analysis of the neural representation of first language and second language. Appl. Psycholinguist. 32, 799–819 53 Golestani, N. et al. (2006) Syntax production in bilinguals. Neuropsychologia 44, 1029–1040 54 Hernandez, A.E. et al. (2007) Age of acquisition modulates neural activity for both regular and irregular syntactic functions. Neuroimage 36, 912–923 6
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
55 Jeong, H. et al. (2007) Effect of syntactic similarity on cortical activation during second language processing: a comparison of English and Japanese among native Korean trilinguals. Hum. Brain Mapp. 28, 194–204 56 Ru¨schemeyer, S-A. et al. (2006) Native and non-native reading of sentences: an fMRI experiment. Neuroimage 31, 354–365 57 Frank, M.J. and Badre, D. (2015) How cognitive theory guides neuroscience. Cognition 135, 14–20 58 Grahn, J.A. et al. (2008) The cognitive functions of the caudate nucleus. Prog. Neurobiol. 86, 141–155 59 Desrochers, T.M. and Badre, D. (2012) Finding parallels in frontostriatal organization. Trends Cogn. Sci. 16, 407–408 60 Robinson, J.L. et al. (2012) The functional connectivity of the human caudate: an application of meta-analytic connectivity modeling with behavioral filtering. Neuroimage 60, 117–129 61 Grahn, J.A. et al. (2009) The role of the basal ganglia in learning and memory: neuropsychological studies. Behav. Brain Res. 199, 53–60 62 Mestres-Misse, A. et al. (2014) Dorsomedial striatum involvement in regulating conflict between current and presumed outcomes. Neuroimage 98, 159–167 63 Mestres-Misse, A. et al. (2012) An anterior–posterior gradient of cognitive control within the dorsomedial striatum. Neuroimage 62, 41–47 64 Badre, D. and Frank, M.J. (2012) Mechanisms of hierarchical reinforcement learning in cortico-striatal circuits 2: evidence from fMRI. Cereb. Cortex 22, 527–536 65 Frank, M.J. and Badre, D. (2012) Mechanisms of hierarchical reinforcement learning in corticostriatal circuits 1: computational analysis. Cereb. Cortex 22, 509–526 66 Corballis, M.C. (2014) Language and recursion. In The Recursive Mind: The Origins of Human Language, Thought and Civilization. pp. 19–35, Princeton University Press 67 Carnie, A. (2012) Extending X-bar theory to functional categories. In Syntax: A Generative Introduction. pp. 197–217, John Wiley & Sons 68 Alexander, G.E. et al. (1986) Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annu. Rev. Neurosci. 9, 357–381 69 Koziol, L. and Budding, D. (2009) The basal ganglia: beyond the motor system – from movement to thought. In Subcortical Structures and Cognition. pp. 27–68, Springer 70 Goldman-Rakic, P.S. (2011) Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In Comprehensive Physiology. John Wiley & Sons 71 Liljeholm, M. and O’Doherty, J.P. (2012) Contributions of the striatum to learning, motivation, and performance: an associative account. Trends Cogn. Sci. 16, 467–475 72 Barbas, H. et al. (2013) Frontal–thalamic circuits associated with language. Brain Lang. 126, 49–61 73 Llano, D.A. (2013) Functional imaging of the thalamus in language. Brain Lang. 126, 62–72 74 Metzger, C.D. et al. (2013) Functional mapping of thalamic nuclei and their integration into cortico-striatal–thalamo-cortical loops via ultrahigh resolution imaging – from animal anatomy to in vivo imaging in humans. Front. Neurosci. 7, 24 75 Chan, S-H. et al. (2013) Role of the striatum in language: syntactic and conceptual sequencing. Brain Lang. 125, 283–294 76 Dominey, P.F. (2013) Recurrent temporal networks and language acquisition – from corticostriatal neurophysiology to reservoir computing. Front. Psychol. 4, 500 77 Brovelli, A. et al. (2011) Differential roles of caudate nucleus and putamen during instrumental learning. Neuroimage 57, 1580–1590 78 Friederici, A.D. (2002) Towards a neural basis of auditory sentence processing. Trends Cogn. Sci. 6, 78–84 79 Garza Villarreal, E.A. et al. (2011) Distinct neural responses to chord violations: a multiple source analysis study. Brain Res. 1389, 103–114 80 Koelsch, S. (2009) Music-syntactic processing and auditory memory: similarities and differences between ERAN and MMN. Psychophysiology 46, 179–190 81 Friedrich, R. and Friederici, A.D. (2009) Mathematical logic in the human brain: syntax. PLoS ONE 4, e5599 82 Eickhoff, S.B. et al. (2005) A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. Neuroimage 25, 1325–1335 83 Amunts, K. et al. (1999) Broca’s region revisited: cytoarchitecture and intersubject variability. J. Comp. Neurol. 412, 319–341
TICS-1417; No. of Pages 7
Opinion 84 Maruyama, M. et al. (2012) The cortical representation of simple mathematical expressions. Neuroimage 61, 1444–1460 85 Wakita, M. (2013) Broca’s area processes the hierarchical organization of observed action. Front. Hum. Neurosci. 7, 937 86 Sammler, D. et al. (2013) Syntax in a pianist’s hand: ERP signatures of ‘‘embodied’’ syntax processing in music. Cortex 49, 1325–1339 87 Kotz, S.A. (2009) A critical review of ERP and fMRI evidence on L2 syntactic processing. Brain Lang. 109, 68–74
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
88 Friederici, A.D. (2006) What’s in control of language? Nat. Neurosci. 9, 991–992 89 Abutalebi, J. et al. (2008) Language control and lexical competition in bilinguals: an event-related FMRI study. Cereb. Cortex 18, 1496–1505 90 Jouen, A.L. et al. (2013) Discrete sequence production with and without a pause: the role of cortex, basal ganglia, and cerebellum. Front. Hum. Neurosci. 7, 492
7
Opinion
Degree of automaticity and the prefrontal cortex Hyeon-Ae Jeon and Angela D. Friederici Max Planck Institute for Human Cognitive and Brain Sciences, Department of Neuropsychology, Leipzig, Germany
The dorsolateral prefrontal cortex (PFC), with more anterior areas [Brodmann area (BA) 45, 47, and 10], has been known to be activated as cognitive hierarchy increases. However, this does not hold for highly automatic processes such as first language (L1), where the posterior region (BA 44) is known as the key area for the processing of complex linguistic hierarchy. Discussing this disparity, we propose that the degree of automaticity (DoA) is a crucial factor for the framework of functional mapping in the PFC: the posterior-to-anterior gradient system for more controlled processes and the posterior-confined system for automatic processes. We support this view with previous findings and provide a new perspective on the functional organization of the PFC. Hierarchical processing in the PFC Hierarchical processing refers to the view that processes in the superordinate level control, modify, and modulate processes in the subordinate level over a longer timescale [1–3]. Since the early 1950s, when the concept of hierarchical processing was introduced in relation to the domain of action planning [4], it has been further extended to various cognitive domains such as language and music. Within this scope, action has been characterized as being goal directed and hierarchically structured such that simple action gestalts integrate into progressively more elaborate actions in a hierarchy that involves different processes at the upper and lower levels corresponding to sequences and action chunks, respectively [5]. Language has been characterized as a hierarchical structure; syllables comprise phonemes, words are made up of syllables, phrases comprise words, and then phrases are assembled to build a sentence [3,6,7]. In the music domain, rule-based arrangement of musical sets results in a hierarchical structure; discrete acoustic sounds are assembled into harmonic sequences following a certain arrangement of chord functions [8,9]. Taking these observations together, it is suggested that human behavior is organized in several levels of hierarchy and that hierarchical processing is essential in human cognition [10–15]. Along with the accumulation of findings from various functional neuroanatomy and lesion studies, the neural Corresponding author: Jeon, H.-A. ([email protected]). Keywords: degree of automaticity; hierarchical processing; cognitive control; syntactic hierarchy; BA 44; prefrontal cortex. 1364-6613/ ß 2015 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tics.2015.03.003
basis of hierarchical processing has been found to involve the dorsolateral PFC [16]. Furthermore, the possible system for generating different levels of hierarchy in human cognition, more specifically in cognitive control (see Glossary), has been proposed to reside in the PFC [2,17–26]. One influential perspective on the cognitive framework suggests a topographic organization along a posterior-toanterior axis in the PFC, with progressively anterior regions being involved in higher levels of hierarchical processing than posterior regions (Figure 1A) [23,27–30]. However, it is not always the case that this posterior-to-anterior gradient system stands to reason. The neural underpinning for hierarchical processing in L1 has been shown to involve a posterior region of the PFC; that is, BA 44 (Figure 1B) [31–33]. An increase in syntactic hierarchy leads to an increase of activation in BA 44 rather than to recruitment of more anterior parts of the PFC as suggested earlier in hierarchical processing for cognitive control (Figure 1C) [32]. What, then, is the reason for these disparate representations between hierarchy of cognitive control (the posterior-to-anterior gradient system) and hierarchy in L1 (the posterior-presiding system)? Here, based on two overarching notions – hierarchy of cognitive control and automaticity – we propose a novel conceptual framework that DoA may play a critical role in the functional organization of the PFC: the posterior-toanterior gradient system for more controlled processes with a low DoA and the posterior-confined system for automatic processes with a high DoA. The essential role
Glossary Center-embedded sentence: a linguistic description of a sentence where phrases/clauses are inserted in other phrases/clauses [66,67]. It is described as having the most demanding hierarchical and recursive structure, with longdistance dependencies [35]. Cognitive control: a cognitive mechanism responsible for coordinating or guiding thoughts and behaviors in relation to current goals and intentions [2,27]. Dorsolateral PFC loop: corticostriatal–thalamocortical loops refer to five loops that are topographically organized and functionally segregated with selected cortical areas in the frontal lobe. These are reported to be involved in motor functions with motor and oculomotor loops and in nonmotor functions with dorsolateral, ventral/orbital, and anterior cingulate loops [68]. In particular, the dorsolateral PFC loop is recruited in cognitive aspects, including working memory, planning, rule-based learning, and sequence learning [58,61,69–77]. Early left-anterior negativity (ELAN): a language-relevant event-related brain potential component found between 120 and 200 ms in response to a syntactic phrase-structure error, thereby reflecting initial syntactic structure-building processes [78]. Early right-anterior negativity (ERAN): known as an electrophysiological marker of musical structure building that is mediated by the inferior frontal gyrus and superior temporal gyrus [79,80].
Trends in Cognitive Sciences xx (2015) 1–7
1
TICS-1417; No. of Pages 7
Opinion (A)
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
condition, whereas the second qualification (complexity) is not. This prominent viewpoint, in harmony with our proposal, also emphasizes the value of automaticity rather than complexity. In the following section, we briefly review the influential models of hierarchical processing in cognitive control, demonstrating the organization of the posterior-to-anterior gradient in the PFC. Next, we examine a study comparing hierarchical processing and the role of automaticity in its functional mapping in the PFC for L1, second language (L2), and the non-language (NL) domain. We substantiate our proposal with previous neuroimaging studies in language development and L2. Lastly, we discuss the DoA and its relation to subcortical structures; that is, the basal ganglia.
Posterior
Anterior
Cognive control Badre et al. (2007) Koechlin et al. (1999, 2003)
(B)
First language Friederici et al. (2006) Makuuchi et al. (2009)
Key: Low
Medium
High
(C)
BA 44
% signal change
0.4
0.2
0
–0.2 0
5
10
15
20
Time in seconds 3.09
Friederici et al. (2006) TRENDS in Cognitive Sciences
Figure 1. Posterior-to-anterior gradient system for cognitive control and posteriorpresiding system for first language. In (A) and (B), the approximate locations of the activation foci are provided in a schematic display with spheres centered on the coordinates of peak activation from previous studies. These spheres indicate activation from different levels of hierarchy (red, high; green, medium; blue, low). (A) Subregions in the prefrontal cortex (PFC) selectively contributed to the different levels of hierarchical processing; as the level of hierarchy became higher, activation was observed in a more anterior region of the PFC [23,29,30]. (B) Activation for hierarchical processing in first language (L1) was positioned in the posterior region of the PFC [Brodmann area (BA) 44] even when the level of hierarchy was high, indicating the posterior-presiding system within the PFC for L1 [32,33]. (C) There was a systematic increase of activation in BA 44 as the level of hierarchy in sentences of L1 increased. Different colored lines in the plot of percentage blood-oxygen level-dependent (BOLD) signal change indicate activation from different levels of hierarchy (red, high; green, medium; blue, low). Adapted, with permission, from [32].
of the DoA and its relation to the functional organization of the PFC can be traced back to the identification of two qualifications, ‘newness’ and ‘complexity’ [16]. Goal-directed behavior, which requires the involvement of the PFC, is characterized as being novel and demanding whereby the former qualification (newness) is a sufficient and necessary 2
Hierarchical processing in cognitive control Numerous studies have demonstrated possible frameworks for generating different levels of hierarchies in cognitive control and their topographic maps in the PFC [2,17,18]. Here we discuss two influential theories: temporal abstraction and policy abstraction. The first framework, temporal abstraction, suggests that a significant fraction of cognitive control is based on temporal framing and context, with three different levels in increasing ranking order: contextual, episodic, and branching control [17,28]. Contextual control denotes a synchronic dimension, which means that selecting an action depends on the stimulus response associations according to current context. Episodic control indicates a diachronic dimension, meaning that cognitive control depends on a discrete past event. Branching control refers to a polychronic dimension, or that selecting actions depends on the information conveyed by past events, which are maintained in a pending state and then reactivated later. These three levels form a cascade of top-down selective processes operating along the posterior-to-anterior axis of the lateral PFC, with the contextual, episodic, and branching controls subserved by the posterior lateral PFC, the anterior lateral PFC, and the frontopolar PFC, respectively [17,23,25,30,34]. The other framework, policy abstraction, describes a cognitive hierarchy ranked by different levels of abstractness [27]. Policy abstraction is related to the rules that govern other subrules such that a higher level of policy abstraction has more subcategories or more subordinate representations than a lower one. Multiple levels of abstractness were proposed by summing several constraints for appropriate action selections: the first level for selection between multiple competing actions (least abstract), the second level for selection between different response-level policies (more abstract), and so on up to multiple levels of increasing abstraction [29]. This multistage system of cognitive control is subserved by different areas supporting each level from the left dorsal premotor cortex (the lowest level) to the left anterior dorsal premotor cortex, to the left inferior frontal sulcus, and to the left frontopolar cortex (the highest level) [2,27]. Although the characteristics of the levels are defined to some extent in different ways in these two frameworks, both postulate a similar topographic mapping of the
TICS-1417; No. of Pages 7
Opinion
DoA and language development Automaticity of language processing and its relation to the functional segregation of the PFC can be investigated in
Co
lex mp
Posterior
ity
Jeon & Friederici (2013)
Anterior
Key:
Low, L2 Medium, L2 Medium, NL High, L2 High, NL
(B)
Au
a tom
cit
Posterior
y
r = –0.62 P < 0.01
1.0
% signal change
Anterior
BA 44 BA 10
BA 47
0.5
0.0
–0.5 0.09
0.10
0.11
0.12
0.13
Index for DoA 1.0
1.0
% signal change
Influence of DoA on hierarchical processing Current knowledge of the posterior-to-anterior gradient view of hierarchy in cognitive control (Figure 1A) does not accord with the posterior-presiding system for hierarchical processing of L1 even when the level of hierarchy is high (Figure 1B,C). The processing of syntactically complex and center-embedded sentences reflecting linguistic hierarchy puts Broca’s area, in particular BA 44, at the center of interest [32,33,35–38]. A recent study reconsidered the disparity in the topographic mapping between hierarchy in cognitive control and in L1, investigating the impact of the DoA on hierarchical processing [39]. The crucial idea behind this study originated from the fact that native language processes in adult language users are highly automatic [40] and rely on BA 44 when processing syntactic information of different levels of hierarchical complexity, but on more anterior portions of the inferior frontal gyrus (BA 45) during development [41]. The impact of the DoA on hierarchical processing was investigated in three different domains: L1 for high DoA and L2/ NL for low DoA. L1 sentences varied in their levels of hierarchy (high with center-embedded sentences and low with non-embedded sentences). Both L2 sentences and NL stimuli exhibited three levels of hierarchy (high, medium, and low) following the three cognitive control levels (branching, episodic, and contextual) as the framework of temporal abstraction suggested [28]. Hierarchical processing for the low-DoA systems (L2/NL) resulted in a posterior-to-anterior pattern of activation as the level of hierarchy increased (Figure 2A). For L1, however, syntactic activation was confined in the posterior region (BA 44) even for the highest level of hierarchy (area with black dots in Figure 2B). Further analyses were limited to the same (highest) levels of hierarchical processing for L1, L2, and NL. The results of these analyses indicate that the anterior region (BA 10/47) showed more activation when involved in the process inherent to the low DoA, whereas the posterior region (BA 44) was more activated when automaticity was high (Figure 2B). Thus, the DoA appears to be a crucial factor for the organization of the PFC: the posterior-to-anterior gradient system characterized by more anterior involvement for the more controlled process (L2/NL) and the posterior-dominant system characterized by posterior involvement for the automatic process (L1). Interestingly, BA 44 was involved not only in the lower levels of hierarchy in studies on cognitive control during action [17,23,25,30] but also in the highest level of hierarchy in L1 [39], suggesting supramodal involvement of BA 44 in human cognition, although at different levels of hierarchy. The question here is: to what extent can this proposal of the DoA and its relation to the functional organization of the PFC explain data from previous studies? The answer can be found from studies of L1 development and L2 learning.
(A)
% signal change
activation slanting toward the anterior region of the PFC as the process becomes more effortful due to an increase in the level of hierarchy (Figure 1A).
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
0.5
0.0
r = 0.56 P < 0.01
–0.5 0.09
0.10
0.11
0.12
0.5
0.0
0.13
0.09
Index for DoA
Key:
r = 0.54 P < 0.01
–0.5 0.10
0.11
0.12
0.13
Index for DoA
L1
L2
NL TRENDS in Cognitive Sciences
Figure 2. Degree of automaticity (DoA) and its influence on first language (L1), second language (L2), and non-language (NL) domains. (A) In L2 and NL, a gradient pattern of activation was observed depending on the level of hierarchy, as predicted in the posterior-to-anterior gradient system. Colors indicate activation from different levels of hierarchy. Activation in the low level in NL was found in the posterior region [Brodmann area (BA) 6; not shown in the figure), but only with a more lenient threshold. (B) Correlation between DoA and blood-oxygen leveldependent (BOLD) signals of the high level of hierarchy depicted for the L1, L2, and NL domains. BA 44 for L1 showed more activation when processing centerembedded sentences (high level of hierarchy) than when processing nonembedded sentences (low level of hierarchy). BA 47 for L2 and BA 10 for NL produced more activation in the branching condition (high level of hierarchy) compared with the episodic and contextual conditions (medium and low levels of hierarchy). A scatter plot of each participant’s index for DoA (as the value of the index decreases, DoA increases) as a function of the percentage BOLD signal was obtained from the three peak activations of L1, L2, and NL. Adapted, with permission, from [39].
studies of language development in children. What are the neural correlates of syntactic processing as children develop their language processes? Do children show a developmental shift toward an adult-like posterior-residing system as their DoA in syntactic processing reaches adult level? 3
TICS-1417; No. of Pages 7
Opinion Some imaging studies indicate that language-related brain activation is more extended in children than in adults [42–44]. In a recent fMRI study [45], three groups of children underwent a sentence–picture matching task with different levels of syntactic hierarchical complexity. Only the group of 9–10-year-old children showed an adultlike effect of complexity in the posterior region of the PFC (i.e., pars triangularis and pars opercularis). Another study showed that younger children (5–6 years of age), whose ability for the process was not fully developed, recruited not only the inferior frontal gyrus but also additional areas of the frontal operculum, which is located more ventrally and medially to BA 44 and BA 45 [42]. This additional involvement of the anterior region was attributed to the higher processing demands described by longer reaction times from children than from adults. The low DoA in children’s syntactic processing at around age 6 years has been evidenced in the temporal dynamics of the bloodoxygen level-dependent (BOLD) signal, with a longer time-to-peak BOLD latency than in adults [46]. Moreover, it is reflected in the lack of early left-anterior negativity (ELAN) in event-related brain potential data from the group of 6-year-old children in response to syntactic errors in complex sentences, which is usually elicited by highly automatic structure-building processes in adults [41]. However, children showing adult-like syntactic processing in a more automatic way demonstrate the involvement of the posterior-residing system [44]. In summary, these studies indicate the additional involvement of the anterior region in younger children whose ability for syntactic processing is not fully developed and therefore occurs at a more effortful and controlled level. DoA and L2 The nature of L2 processing, compared with L1, is less automatic and more effortful such that it is considered to be a controlled process [47,48]. L2 processing manifests in increased fMRI responses in the left PFC for switching to a language in which one is less proficient [49] or inhibiting the processing of an irrelevant language [50]. It has been suggested that improvement in proficiency is connected to an increase in automaticity [48] and thus the influence of the DoA on the neural underpinnings of language processing has been studied using L2 with proficient or unskilled learners. Highly proficient L2 learners show an overlap in activation between L1 and L2 whereas those with a low level of proficiency show less overlap between the two languages and, rather, recruit additional brain areas [51]. Low-proficiency groups usually show activation clusters that are widely distributed in the anterior PFC extending to BA 45/46 [48,52–54]. For instance, when Korean sentences and English sentences were presented to native Korean speakers for auditory comprehension, broader regions in the PFC were activated, including the pars triangularis, pars opercularis, and superior frontal gyrus, for English [55]. This pattern was also found in a study where Russians judged the acceptability of German sentences (L2), yielding activation in BA 45/47 [56]. Taking these observations together, the additional involvement of the anterior PFC in L2, corresponding with that observed in NL tasks and less-proficient children’s language 4
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
Box 1. DoA in mathematics and music Two studies were conducted to investigate the neural underpinnings of building a hierarchical structure using arithmetic. In one study, university students were recruited for hierarchical processing with abstract mathematical formulae [81]. This study found significant activation in anterior areas in the PFC including BA 45, 47, and 10. The other study, which used well-trained participants with a high DoA in mathematics, found that such hierarchical processing elicited activation in the posterior PFC as BA 44 (x, y, z coordinates: 45, 9, 30 based on the Ju¨lich histological atlas [82,83]) as well as the bilateral ventral occipitotemporal cortices [84]. Notably, in this experiment simple arithmetic calculations could be automatized in highly proficient participants, consequently being involved in the posterior PFC and compiled in the visual cortex. Therefore, processing of mathematical formulae also seems to implicate the anterior region of the PFC for low DoA and posterior activation for high DoA. In the music domain, a posterior-dominant contribution to the processing of high DoA can be supported by a comparison between two groups of participants based on years of musical training. For example, well-trained and less-trained groups who had taken formal piano lessons identified melodies by watching silent movies of hierarchically organized hand movements playing familiar and unfamiliar melodies, resulting in increased activation in Broca’s area in the unfamiliar melody condition in the well-trained participants only [85]. A compatible finding was reported in eventrelated potential studies; when professional pianists with a minimum of 14 years of musical training imitated silent videos of a right hand playing congruent or incongruent sequences of chords, ERAN was evoked for incongruent chords, reflecting music-syntactic processing [86]. In summary, the influence of the DoA on the functional organization of the PFC may be generalized to cognitive domains outside language.
mentioned earlier, coincides with our proposal that a more anterior region is recruited for the more controlled process with a low DoA. We have discussed the pivotal role of the DoA in the functional dissociation of the PFC mostly in language processing (L2 and children’s language). However, can the idea of DoA be applied to domains other than language? The DoA seems to be able to explain the topographic mapping of the PFC in other cognitive domains such as mathematics and music (Box 1). DoA and basal ganglia Along with the PFC, the basal ganglia implemented within the dorsolateral PFC loop also deserve discussion in terms of the DoA. The DoA has been found to modulate the posterior-to-anterior system even in the subcortical structures, mirroring the gradient pattern of the PFC in the basal ganglia [57]. More specifically, the functional specificity of the caudate nucleus has been discussed in terms of its segmentation into anterior (head) and posterior (body) components, with the head being mediated by more controlled processes compared with the body [58–61]. In a recent high-resolution fMRI experiment, the posterior-toanterior gradient system observed in the PFC was also manifested in the caudate nucleus, more specifically, with the anterior region being activated for the more controlled process [34]. A similar result was observed in semantic/ syntactic ambiguity resolution, with the most complex condition requiring higher-order control processes showing activation in the anterior dorsomedial striatum [62,63]. Similarly, processing a hierarchical rule compared with
TICS-1417; No. of Pages 7
Opinion Box 2. DoA and its influence on basal ganglia The DoA and its relation to the basal ganglia have been substantiated in studies of L2 learning [48,87,88] where the globus pallidus as well as the anterior PFC (BA 47) were activated in more effortful and controlled processes (e.g., translating L1 into L2). Likewise, when a simple condition (naming a picture in L1) was compared with a more controlled condition (naming a picture in either L1 or L2 depending on the cue), for the more controlled condition the caudate nucleus was significantly activated as well as there being anteriorly extended activation along the inferior frontal gyrus and the middle frontal gyrus [89]. Additionally, direct comparisons between L1 and L2 naming also yielded extensive neural activation in the anterior PFC (BA 44, 45, and 47) and the caudate nucleus in L2 naming [89]. Involvement of the basal ganglia in more controlled processes has also been found in motorsequence learning in the action domain. Learning complex sensory–motor sequences, compared with simple sequences, yielded broad activation not only in the frontoparietal area but also in the caudate nucleus and cerebellum [90]. Taken together, the involvement of these subcortical structures linked by corticosubcortcial or corticocerebellar loops may orchestrate the processing of cognitive control depending on the DoA.
Box 3. Outstanding questions Do people with a high level of proficiency resort to the posteriorconfined system in hierarchical processing in the NL domain (i.e., music, action, mathematics)? Is the posterior-to-anterior system transformed into the posteriordominant system as one’s DoA increases in an initially controlled task? Besides the DoA, what can be a possible explanation for functional organization in the PFC? Why do the basal ganglia show the same posterior-to-anterior system pattern as the PFC? What is the advantage of having this mirrored system between the cortical and subcortical regions? How are the temporal dynamics manifested between the highand low-DoA processes in the PFC as well as the basal ganglia? Do behavioral deficits resulting from lesions located at different sites in the posterior-to-anterior system in the PFC vary?
processing a flat rule manifested with significant activation in the anterior portion of the caudate nucleus [64,65]. To summarize, the posterior-to-anterior system also seems to contribute to the topographic organization of the substructures of the basal ganglia, which suggest that the PFC and the basal ganglia work in harmony, forming an interwoven network between the cortical and subcortical areas for the cognitive processes in more complex and demanding tasks (see Box 2 for more examples) [34]. Concluding remarks Several theoretical accounts have been suggested as a framework for generating different levels of hierarchy in cognitive control along with the posterior-to-anterior axis of the PFC [27]. Incorporating this current knowledge, we propose that the DoA may be the critical factor in the functional organization of the PFC: the posterior-to-anterior gradient system with progressively anterior regions supporting a low DoA in more controlled processes and the posterior-confined system with posterior regions being involved in the process with a high DoA in highly automatic processes. To make possible broad generalizations of our proposal, further research should test hierarchical
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
processing in various cognitive domains in groups of participants with varying levels of DoA and also scrutinize the change of topographic mapping in the PFC as DoA increases (Box 3). Acknowledgments The authors thank Kerstin Flake and Andrea Gast-Sandmann for graphics and Elizabeth Kelly and Shameem Wagner for proofreading. This work was supported by a grant from the European Research Council (ERC-2010-360 AdG 20100407) awarded to A.D.F.
References 1 Welford, A.T. (1951) Skill and age; an experimental approach. In On the Nature of Skill. pp. 11–27, Oxford University Press 2 Badre, D. (2008) Cognitive control, hierarchy, and the rostro-caudal organization of the frontal lobes. Trends Cogn. Sci. 12, 193–200 3 Jeon, H.A. (2014) Hierarchical processing in the prefrontal cortex in a variety of cognitive domains. Front. Syst. Neurosci. 8, 223 4 Lashley, K.S. (1951) The problem of serial order in behavior. In Cerebral Mechanisms in Behavior (Jefress, L.A., ed.), pp. 112–146, Wiley 5 Klapp, S.T. and Jagacinski, R.J. (2011) Gestalt principles in the control of motor action. Psychol. Bull. 137, 443–462 6 Fitch, W.T. (2010) The Evolution of Language, Cambridge University Press 7 Berwick, R.C. et al. (2013) Evolution, brain, and the nature of language. Trends Cogn. Sci. 17, 89–98 8 Thompson-Schill, S. et al. (2013) Multiple levels of structure in language and music. In Language, Music and the Brain: A Mysterious Relationship. pp. 289–303, MIT Press 9 Patel, A.D. (2010) Syntax. In Music, Language, and the Brain. Oxford University Press 10 Fitch, W.T. and Martins, M.D. (2014) Hierarchical processing in music, language, and action: Lashley revisited. Ann. N. Y. Acad. Sci. 1316, 87–104 11 Azuar, C. et al. (2014) Testing the model of caudo-rostral organization of cognitive control in the human with frontal lesions. Neuroimage 84, 1053–1060 12 Koelsch, S. et al. (2013) Processing of hierarchical syntactic structure in music. Proc. Natl. Acad. Sci. U.S.A. 110, 15443–15448 13 Udden, J. and Bahlmann, J. (2012) A rostro-caudal gradient of structured sequence processing in the left inferior frontal gyrus. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 367, 2023–2032 14 Bahlmann, J. et al. (2008) Hierarchical artificial grammar processing engages Broca’s area. Neuroimage 42, 525–534 15 Bahlmann, J. et al. (2009) Neural circuits of hierarchical visuo-spatial sequence processing. Brain Res. 1298, 161–170 16 Fuster, J.M. (1989) A theory of prefrontal functions: the prefrontal cortex and the temporal organization of behavior. In The Prefrontal Cortex: Anatomy, Physiology and Neuropsychology of the Frontal Lobe. pp. 157–192, Lippincott Williams & Wilkins 17 Koechlin, E. and Summerfield, C. (2007) An information theoretical approach to prefrontal executive function. Trends Cogn. Sci. 11, 229–235 18 Botvinick, M.M. (2008) Hierarchical models of behavior and prefrontal function. Trends Cogn. Sci. 12, 201–208 19 Zilles, K. and Amunts, K. (2012) Neuroscience. Segregation and wiring in the brain. Science 335, 1582–1584 20 O’Reilly, R.C. (2010) The what and how of prefrontal cortical organization. Trends Neurosci. 33, 355–361 21 Badre, D. and D’Esposito, M. (2009) Is the rostro-caudal axis of the frontal lobe hierarchical? Nat. Rev. Neurosci. 10, 659–669 22 Botvinick, M.M. (2007) Multilevel structure in behaviour and in the brain: a model of Fuster’s hierarchy. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 362, 1615–1626 23 Koechlin, E. et al. (2003) The architecture of cognitive control in the human prefrontal cortex. Science 302, 1181–1185 24 Stout, D. (2010) The evolution of cognitive control. Top. Cogn. Sci. 2, 614–630 25 Koechlin, E. and Jubault, T. (2006) Broca’s area and the hierarchical organization of human behavior. Neuron 50, 963–974 26 Szczepanski, S.M. and Knight, R.T. (2014) Insights into human behavior from lesions to the prefrontal cortex. Neuron 83, 1002–1018 5
TICS-1417; No. of Pages 7
Opinion 27 Badre, D. (2013) Hierarchical cognitive control and the functional organization of the frontal cortex. In The Oxford Handbook of Cognitive Neuroscience, Volume 2: The Cutting Edges. pp. 300–317, Oxford University Press 28 Koechlin, E. and Summerfield, C. (2008) The cognitive architecture of the human lateral prefrontal cortex. In Sensorimotor Foundations of Higher Cognition (Haggard, P. et al., eds), pp. 487–509, Oxford University Press 29 Badre, D. and D’Esposito, M. (2007) Functional magnetic resonance imaging evidence for a hierarchical organization of the prefrontal cortex. J. Cogn. Neurosci. 19, 2082–2099 30 Koechlin, E. et al. (1999) The role of the anterior prefrontal cortex in human cognition. Nature 399, 148–151 31 Friederici, A.D. (2011) The brain basis of language processing: from structure to function. Physiol. Rev. 91, 1357–1392 32 Friederici, A.D. et al. (2006) Processing linguistic complexity and grammaticality in the left frontal cortex. Cereb. Cortex 16, 1709–1717 33 Makuuchi, M. et al. (2009) Segregating the core computational faculty of human language from working memory. Proc. Natl. Acad. Sci. U.S.A. 106, 8362–8367 34 Jeon, H.A. et al. (2014) Functional network mirrored in the prefrontal cortex, caudate nucleus, and thalamus: high-resolution functional imaging and structural connectivity. J. Neurosci. 34, 9202–9212 35 Friederici, A.D. et al. (2011) The neural basis of recursion and complex syntactic hierarchy. Biolinguistics 5, 87–104 36 Poeppel, D. (2014) The neuroanatomic and neurophysiological infrastructure for speech and language. Curr. Opin. Neurobiol. 28, 142–149 37 Poeppel, D. et al. (2012) Towards a new neurobiology of language. J. Neurosci. 32, 14125–14131 38 Fedorenko, E. (2014) The role of domain-general cognitive control in language comprehension. Front. Psychol. 5, 335 39 Jeon, H.A. and Friederici, A.D. (2013) Two principles of organization in the prefrontal cortex are cognitive hierarchy and degree of automaticity. Nat. Commun. 4, 2041 40 Abutalebi, J. et al. (2013) Language proficiency modulates the engagement of cognitive control areas in multilinguals. Cortex 49, 905–911 41 Hahne, A. et al. (2004) Brain signatures of syntactic and semantic processes during children’s language development. J. Cogn. Neurosci. 16, 1302–1318 42 Brauer, J. and Friederici, A.D. (2007) Functional neural networks of semantic and syntactic processes in the developing brain. J. Cogn. Neurosci. 19, 1609–1623 43 Wilke, M. et al. (2005) Comprehensive language mapping in children, using functional magnetic resonance imaging: what’s missing counts. Neuroreport 16, 915–919 44 Nun˜ez, S.C. et al. (2011) fMRI of syntactic processing in typically developing children: structural correlates in the inferior frontal gyrus. Dev. Cogn. Neurosci. 1, 313–323 45 Skeide, M.A. et al. (2014) Syntax gradually segregates from semantics in the developing brain. Neuroimage 100, 106–111 46 Brauer, J. et al. (2008) Temporal dynamics of perisylvian activation during language processing in children and adults. Neuroimage 41, 1484–1492 47 Indefrey, P. (2006) A meta-analysis of hemodynamic studies on first and second language processing: which suggested differences can we trust and what do they mean? Lang. Learn. 56, 279–304 48 Abutalebi, J. (2008) Neural aspects of second language representation and language control. Acta Psychol. 128, 466–478 49 Wang, Y. et al. (2007) Neural bases of asymmetric language switching in second-language learners: an ER–fMRI study. Neuroimage 35, 862–870 50 Rodriguez-Fornells, A. et al. (2002) Brain potential and functional MRI evidence for how to handle two languages with one brain. Nature 415, 1026–1029 51 Higby, E. et al. (2013) Multilingualism and the brain. Annu. Rev. Appl. Linguist. 33, 68–101 52 Sebastian, R. et al. (2011) Meta-analysis of the neural representation of first language and second language. Appl. Psycholinguist. 32, 799–819 53 Golestani, N. et al. (2006) Syntax production in bilinguals. Neuropsychologia 44, 1029–1040 54 Hernandez, A.E. et al. (2007) Age of acquisition modulates neural activity for both regular and irregular syntactic functions. Neuroimage 36, 912–923 6
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
55 Jeong, H. et al. (2007) Effect of syntactic similarity on cortical activation during second language processing: a comparison of English and Japanese among native Korean trilinguals. Hum. Brain Mapp. 28, 194–204 56 Ru¨schemeyer, S-A. et al. (2006) Native and non-native reading of sentences: an fMRI experiment. Neuroimage 31, 354–365 57 Frank, M.J. and Badre, D. (2015) How cognitive theory guides neuroscience. Cognition 135, 14–20 58 Grahn, J.A. et al. (2008) The cognitive functions of the caudate nucleus. Prog. Neurobiol. 86, 141–155 59 Desrochers, T.M. and Badre, D. (2012) Finding parallels in frontostriatal organization. Trends Cogn. Sci. 16, 407–408 60 Robinson, J.L. et al. (2012) The functional connectivity of the human caudate: an application of meta-analytic connectivity modeling with behavioral filtering. Neuroimage 60, 117–129 61 Grahn, J.A. et al. (2009) The role of the basal ganglia in learning and memory: neuropsychological studies. Behav. Brain Res. 199, 53–60 62 Mestres-Misse, A. et al. (2014) Dorsomedial striatum involvement in regulating conflict between current and presumed outcomes. Neuroimage 98, 159–167 63 Mestres-Misse, A. et al. (2012) An anterior–posterior gradient of cognitive control within the dorsomedial striatum. Neuroimage 62, 41–47 64 Badre, D. and Frank, M.J. (2012) Mechanisms of hierarchical reinforcement learning in cortico-striatal circuits 2: evidence from fMRI. Cereb. Cortex 22, 527–536 65 Frank, M.J. and Badre, D. (2012) Mechanisms of hierarchical reinforcement learning in corticostriatal circuits 1: computational analysis. Cereb. Cortex 22, 509–526 66 Corballis, M.C. (2014) Language and recursion. In The Recursive Mind: The Origins of Human Language, Thought and Civilization. pp. 19–35, Princeton University Press 67 Carnie, A. (2012) Extending X-bar theory to functional categories. In Syntax: A Generative Introduction. pp. 197–217, John Wiley & Sons 68 Alexander, G.E. et al. (1986) Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annu. Rev. Neurosci. 9, 357–381 69 Koziol, L. and Budding, D. (2009) The basal ganglia: beyond the motor system – from movement to thought. In Subcortical Structures and Cognition. pp. 27–68, Springer 70 Goldman-Rakic, P.S. (2011) Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In Comprehensive Physiology. John Wiley & Sons 71 Liljeholm, M. and O’Doherty, J.P. (2012) Contributions of the striatum to learning, motivation, and performance: an associative account. Trends Cogn. Sci. 16, 467–475 72 Barbas, H. et al. (2013) Frontal–thalamic circuits associated with language. Brain Lang. 126, 49–61 73 Llano, D.A. (2013) Functional imaging of the thalamus in language. Brain Lang. 126, 62–72 74 Metzger, C.D. et al. (2013) Functional mapping of thalamic nuclei and their integration into cortico-striatal–thalamo-cortical loops via ultrahigh resolution imaging – from animal anatomy to in vivo imaging in humans. Front. Neurosci. 7, 24 75 Chan, S-H. et al. (2013) Role of the striatum in language: syntactic and conceptual sequencing. Brain Lang. 125, 283–294 76 Dominey, P.F. (2013) Recurrent temporal networks and language acquisition – from corticostriatal neurophysiology to reservoir computing. Front. Psychol. 4, 500 77 Brovelli, A. et al. (2011) Differential roles of caudate nucleus and putamen during instrumental learning. Neuroimage 57, 1580–1590 78 Friederici, A.D. (2002) Towards a neural basis of auditory sentence processing. Trends Cogn. Sci. 6, 78–84 79 Garza Villarreal, E.A. et al. (2011) Distinct neural responses to chord violations: a multiple source analysis study. Brain Res. 1389, 103–114 80 Koelsch, S. (2009) Music-syntactic processing and auditory memory: similarities and differences between ERAN and MMN. Psychophysiology 46, 179–190 81 Friedrich, R. and Friederici, A.D. (2009) Mathematical logic in the human brain: syntax. PLoS ONE 4, e5599 82 Eickhoff, S.B. et al. (2005) A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. Neuroimage 25, 1325–1335 83 Amunts, K. et al. (1999) Broca’s region revisited: cytoarchitecture and intersubject variability. J. Comp. Neurol. 412, 319–341
TICS-1417; No. of Pages 7
Opinion 84 Maruyama, M. et al. (2012) The cortical representation of simple mathematical expressions. Neuroimage 61, 1444–1460 85 Wakita, M. (2013) Broca’s area processes the hierarchical organization of observed action. Front. Hum. Neurosci. 7, 937 86 Sammler, D. et al. (2013) Syntax in a pianist’s hand: ERP signatures of ‘‘embodied’’ syntax processing in music. Cortex 49, 1325–1339 87 Kotz, S.A. (2009) A critical review of ERP and fMRI evidence on L2 syntactic processing. Brain Lang. 109, 68–74
Trends in Cognitive Sciences xxx xxxx, Vol. xxx, No. x
88 Friederici, A.D. (2006) What’s in control of language? Nat. Neurosci. 9, 991–992 89 Abutalebi, J. et al. (2008) Language control and lexical competition in bilinguals: an event-related FMRI study. Cereb. Cortex 18, 1496–1505 90 Jouen, A.L. et al. (2013) Discrete sequence production with and without a pause: the role of cortex, basal ganglia, and cerebellum. Front. Hum. Neurosci. 7, 492
7
