Elsevier

NeuroImage

Volume 141, 1 November 2016, Pages 350-356
NeuroImage

There is more into ‘doing’ than ‘knowing’: The function of the right inferior frontal sulcus is specific for implementing versus memorising verbal instructions

https://doi.org/10.1016/j.neuroimage.2016.07.059Get rights and content

Abstract

In the present study we examine the mechanism underlying the human ability to implement newly instructed stimulus-response mappings for their future application. We introduce a novel procedure in which we can investigate the processes underlying such implementation while controlling for more general working-memory demands. The results indicate that a region within the dorso-lateral prefrontal cortex (DLPFC) in the vicinity of the inferior frontal sulcus (IFS) is specifically recruited when new instructions are implemented compared to when new instructions are memorised. In addition, we observed that this area is more strongly activated when task performance is effective. Together, these findings suggest that the DLPFC, and more specific the IFS, plays an important role during the formation of procedural representations in working memory.

Introduction

The human ability to learn on the basis of instructions is a key example of cognitive flexibility. It allows us to act appropriately in a given situation without the need of past experience with similar situations and enables us to bypass the time-consuming process of learning via trial and error. Most likely, this fast form of learning relies on a mechanism that translates instructions that specify how to act in a particular situation into ready-to-use procedural representations. Already several decades ago Woodworth (1938) reported evidence that procedural if-then-rules akin to stimulus-response (S-R) mappings can be established merely on the basis of instructions. More specifically, he showed that a stimulus could reflexively trigger an appropriate response through instructed S-R mappings. It is only during the last decade, however, that researchers started to investigate the processes underlying instruction learning in a systematic way (Cohen-Kdoshay and Meiran, 2007, Cohen-Kdoshay and Meiran, 2009, Cole et al., 2013, De Houwer et al., 2005, Liefooghe et al., 2012, Liefooghe et al., 2013, Waszak et al., 2008, Wenke et al., 2007, Wenke et al., 2009).

Conceptually, the mechanisms underlying instruction learning are strongly related to the architecture of working memory (WM). WM is typically described as a system that temporarily stores and manipulates information in order to perform complex cognitive tasks such as language comprehension, learning and reasoning (e.g. Baddeley, 1992). Recently, it was proposed that declarative and procedural information are represented in different parts of WM (e.g. Logan and Gordon, 2001, Oberauer, 2009, Oberauer, 2010, Rubinstein et al., 2001, Vandierendonck, 2012). According to this view, declarative WM is at play when, for example, one is asked to memorise the different steps that are needed to make a dish. By contrast, procedural WM is at play when you actually want to use this information to prepare the dish. This functional distinction proposed in WM relates to the proposed distinction between declarative and procedural knowledge by Anderson (1983). Anderson argued that all knowledge starts as declarative knowledge and that procedural knowledge is acquired through interferences of already existing declarative knowledge. From this perspective, instruction implementation can be seen as prototypical behaviour in which declarative information is transformed into procedural response rules (Oberauer, 2009). In the literature, this implementation process has been called ‘task-set formation’ (Cole et al., 2010). Measuring instruction implementation during the instruction phase with fMRI can thus inform us about the neural mechanisms that are involved when declarative representations are transformed into procedural representations.

Already in the 1960s, it was shown that lesions to the lateral prefrontal cortex (LPFC) have a strong detrimental effect on the ability to learn through verbal instructions while the ability to remember the instructions remained intact (Milner, 1964, Milner, 1965, Luria, 1973, Luria et al., 1964). This phenomenon has been referred to as the dissociation of knowing and doing. More recently, neuroimaging studies with healthy volunteers have shown that a broad fronto-parietal network is involved during instruction learning. This network consists of parts of the lateral prefrontal cortex (LPFC), including the inferior frontal sulcus/junction (IFS); the posterior parietal cortex (PPC); the premotor/motor cortices and pre-supplementary motor area (pre-SMA; Cole et al., 2010, Dumontheil et al., 2011, Hartstra et al., 2011, Hartstra et al., 2012, Ruge and Wolfensteller, 2010). Especially the finding that the LPFC plays a key role in transforming verbal instructions into procedural response rules parallels with aforementioned lesion studies. Ruge and Wolfensteller (2010) found that the stronger LPFC is activated during the encoding of newly instructed S-R mappings, the better performance is in the subsequent application of these mappings (Ruge and Wolfensteller, 2010). In addition, Hartstra et al. (2012) reported that the LPFC is the only region that is sensitive to the integration of task-relevant information.

However, other lines of research have shown that similar parts of the LPFC are also involved in tasks that simply require the maintenance of declarative information in WM, without the demand to implement this information into a procedural representation (e.g. Cohen et al., 1997, Courtney et al., 1998, Jonides et al., 1997, Owen et al., 2005, Rypma et al., 1999, Zanto et al., 2010, Zanto et al., 2011). The possibility thus remains that LPFC activation during instruction learning is not a proxy for the active implementation of instructions into procedural response rules as such, but that this activation simply relates to the maintenance of the instructions.

Until now, most studies investigating instruction learning have not directly distinguished between the maintenance and implementation of response rules. One notable exception is a behavioural study of Liefooghe et al. (2012). In this study, it was observed that, in contrast to memorised S-R mappings, implemented S-R mappings cause response interference in a secondary task, which shares the same stimuli and responses, but uses a different relevant stimulus feature. In a follow-up study, Liefooghe et al. (2013); also see Wenke et al., 2009) furthermore demonstrated that this interference effect only shows up when participants actively prepare for an upcoming task on the basis of the instructed S-R mappings for that task. In other words, instruction implementation does not seem to be a process that automatically occurs when response rules are memorised, but rather instructed response rules have to be actively “translated” to a procedural representation (see also Meiran et al., 2015 for a similar conclusion). These studies thus suggest that implemented S-R rules functionally differ from S-R rules that are held in declarative working memory. However, these studies only offer behavioural evidence and it remains unclear whether neuro-physiological dissociations can be obtained between implemented and memorised S-R mappings.

Here, we directly investigated this issue by contrasting a condition in which instructed S-R rules are implemented for future application to a condition in which instructed S-R rules are simply maintained in declarative memory. By doing so, we can control for processes that are involved when simply maintaining information in declarative WM. We learned from the studies of Liefooghe et al., 2012, Liefooghe et al., 2013 that crucial processes for instruction implementation occur immediately after instruction presentation. Therefore, in our fMRI analyses, we focused on brain activation during the phase in which the instructions were presented, not during the phase in which the instructions were executed.

We argue that the regions that are more involved during instruction implementation than during instruction memorisation are related to a mechanism that translates instructions into procedural response rules. In addition, we expect that higher levels of activity in regions that are involved during instruction implementation will relate to improved performance during task execution (see also, Ruge and Wolfensteller, 2010). In sum, this experiment allows to identify the brain regions that are involved in the mechanism proposed by Anderson (1983) that is responsible for translating declarative knowledge into procedural knowledge.

Section snippets

Participants

Thirty-seven, neurologically normal, right-handed volunteers (15 men, age range = 19 to 36, mean age = 22.8) with (corrected-to-) normal vision and normal colour vision were recruited at Ghent University. None of the participants reported a history of neurological or psychiatric disorder or any other past major medical issue. Participants were randomly divided into two experimental groups; the ‘implementation group’ (N = 19, 7 men) and the ‘memorisation group’ (N = 18, 8 men). All participants gave

Behavioural results

One participant of the implementation group was excluded from all analyses because the proportion of errors (0.25) in the Go trials exceeded the mean proportion of error in the total group by more than two standard deviations. Of the remaining 36 participants, we analysed the reaction times (RT) in Go trials and error rates in Go, NoGo and Catch trials in the probe phase.

For the reaction time analyses, erroneous Go trials, NoGo trials and catch trials with a response were removed from the data

Discussion

In the present study, we investigated the neural correlates of instruction learning. More specifically, we were interested in brain regions involved in the formation of procedural WM representations on the basis of instructions. By comparing the implementation of newly instructed S-R mappings with the mere memorisation of newly instructed S-R mappings we found evidence that an area within the right IFS plays a specific role in this process. First, we observed that this is the only brain area

Acknowledgements

The research reported in this article was supported by a grant P7/33 of the Interuniversity Attraction Poles Program (IAP) grant P7/33 of The Belgian Science Policy Office and further supported by the Special Research Fund (BOF) of Ghent University (Grants BOF08/GOA/011 and BOF09/01M00209), and DFG grant WE 2852/3-1. The authors declare no competing financial interests.

References (49)

  • WaszakF. et al.

    Cross-talk of instructed and applied arbitrary visuomotor mappings

    Acta Psychol.

    (2008)
  • ZantoT.P. et al.

    Top-down modulation of visual feature processing: the role of the inferior frontal junction

    Neuroimage

    (2010)
  • BaddeleyA.

    Working memory

    Science

    (1992)
  • BrassM. et al.

    Who comes first? The role of the prefrontal and parietal cortex in cognitive control

    J. Cogn. Neurosci.

    (2005)
  • BrettM. et al.

    Region of interest analysis using the MarsBar toolbox for SPM 99

    NeuroImage

    (2002)
  • BuckleyM.J. et al.

    Dissociable components of rule-guided behavior depend on distinct medial and prefrontal regions

    Science

    (2009)
  • CohenJ.D. et al.

    Temporal Dynamics of Brain Activation During a Working Memory Task

    (1997)
  • Cohen-KdoshayO. et al.

    The representation of instructions in working memory leads to autonomous response activation: evidence from the first trials in the flanker paradigm

    Q. J. Exp. Psychol.

    (2007)
  • Cohen-KdoshayO. et al.

    The representation of instructions operates like a prepared reflex: flanker compatibility effects found in first trial following S–R instructions

    Exp. Psychol.

    (2009)
  • ColeM.W. et al.

    Prefrontal dynamics underlying rapid instructed task learning reverse with practice

    J. Neurosci.

    (2010)
  • ColeM.W. et al.

    Rapid instructed task learning: a new window into the human brain's unique capacity for flexible cognitive control

    Cogn. Affect. Behav. Neurosci.

    (2013)
  • CourtneyS.M. et al.

    An area specialized for spatial working memory in human frontal cortex

    Science

    (1998)
  • De HouwerJ. et al.

    Further evidence for the role of mode-independent short-term associations in spatial Simon effects

    Percept. Psychophys.

    (2005)
  • DerrfussJ. et al.

    Involvement of the inferior frontal junction in cognitive control: meta-analyses of switching and Stroop studies

    Hum. Brain Mapp.

    (2005)
  • Cited by (26)

    • Instruction-based learning: A review

      2022, Neuropsychologia
      Citation Excerpt :

      Indeed, subjects have to control their action to make a correct response during implementation. Regarding the IFS, Demanet et al. (2016) found that the right IFS is specially activated when new instructions are implemented compared to when new instructions are memorized. Moreover, the activities in this region are correlated with the task performance (Demanet et al., 2016), which is consistent with the role of the IFS in application of arbitrary S-R associations (e.g., Passingham et al., 2000; Toni et al., 2001a, b).

    View all citing articles on Scopus
    View full text