Elsevier

Brain Research

Volume 1189, 16 January 2008, Pages 97-114
Brain Research

Research Report
Activation of the posterior cingulate by semantic priming: A co-registered ERP/fMRI study

https://doi.org/10.1016/j.brainres.2007.10.095Get rights and content

Abstract

Although the N400 is the best understood semantically sensitive component of the event-related potential (ERP), others have been observed as well. In an earlier lexical decision study, an N300 ERP was found to be enhanced to unprimed targets, although the effect could also be characterized as a prolonged P2 to primed targets as described in other reports. Because its scalp topography suggested its neural source might be of interest, a source localization was conducted that suggested that this component emanated from the dorsal posterior cingulate cortex (dPCC). In order to confirm this word N300 localization, a functional magnetic resonance imaging (fMRI) study was conducted to replicate the ERP study with a separate sample of 17 participants in an event-related design, using a 3-T scanner. A significant activation in the right dPCC was found corresponding to the N300 localization. The activation was greater on the related prime trials, supporting the characterization of the ERP component as being a P2 rather than an N300. A review is provided which suggests that a number of separate lines of ERP research regarding the word N300, the picture N300, the word P2, the phonological mismatch negativity, and the word midline frontal negativity may be most parsimoniously regarded as dealing with the same ERP component and that they all therefore emanate from the dPCC. It is suggested that this region plays a role in stimulus–response mapping in polymodal fashion. It is also suggested that the ERP component be termed a P2-dPCC.

Introduction

One of the chief tools for investigating the neural basis of semantics has been the N400 response, a component of the event-related potential (ERP) that is sensitive to deviations from the current semantic context (Kutas and Federmeier, 2000, Kutas and Hillyard, 1980), as in semantic priming tasks. Semantic priming is said to have occurred when a stimulus is recognized more quickly or accurately because of prior exposure to another related stimulus. A common semantic priming paradigm is the lexical decision task in which the participant decides whether a letter string is a word (Meyer and Schvaneveldt, 1971). If the target is a word, it will be recognized more quickly if it is preceded by a related word than if it is preceded by an unrelated word. Even the simple lexical decision task can involve multiple semantic processes (McNamara, 2005, Neely, 1991), let alone the more complex case of sentence comprehension. Given the complexity of language processing, there has been increasing interest in whether other ERP components might provide additional insights into semantic processing.

It was therefore of interest when a recent study of semantic priming reported an additional semantic effect, dubbed an N300, in addition to the N400 (Franklin et al., 2007). An N300 component was first noted in a picture-matching task where it was larger for mismatches (Barrett and Rugg, 1990). It has been suggested it might reflect a system dedicated to processing semantics of pictures (Barrett and Rugg, 1990) or integrating pictorial-semantics into higher-level representations (McPherson and Holcomb, 1999). However, the finding that an apparently identical N300 can be observed in a word stimulus experiment (Franklin et al., 2007) with a scalp topography very similar to the picture N300 as seen in a report using the same recoding equipment (Hamm et al., 2002) suggests that the picture N300 may not be specific to pictures. If the word N300 and the picture N300 component are the same, then the hypothesis that the picture N300 reflects a picture-specific semantic process (Barrett and Rugg, 1990, McPherson and Holcomb, 1999) would have to be revised.

Furthermore, there have also been reports in other visual word experiments of a similar-looking component which has variously been termed part of an N330 (Nobre and McCarthy, 1994), an N310 (Hill et al., 2005, Hill et al., 2002), and a mid-frontal negativity or MFN (Frishkoff et al., 2004, Frishkoff, 2007). These appear to be the only current reports of such an N300 to words, which is not surprising because in all three cases it required careful topographic analysis of high-density data to clearly distinguish from the N400; without such an examination it can appear to be simply an anterior portion of the N400 effect. Note that this midline N300 is different from an N300 that has a left anterior temporal scalp distribution and is larger for related, rather than unrelated, words (Dien et al., 2003, Frishkoff, 2007, Nobre and McCarthy, 1994).

It is not known at this point exactly what the word N300 (and the picture N300) reflects, other than being sensitive to semantics under some circumstances. In the preceding ERP study (Franklin et al., 2007), there was a trend toward it being more significant for a 500-ms stimulus onset asynchrony or SOA (not significant for the short 150 ms SOA when tested separately), suggestive of it reflecting a controlled process rather than automatic spreading activation (ASA). Also, one of the MFN studies, which may or may not be the same as the word N300, reported finding it to be stronger for long versus short SOAs (Frishkoff, 2007). On the other hand, a different lexical decision study (Hill et al., 2002) reported that the word N300 was only significant for a short SOA condition (without explicitly testing the priming effect for an interaction with SOA) and suggested it therefore reflected ASA (but see Hill et al., 2005).

Another putative component may provide more information. A number of studies (Barnea and Breznitz, 1998, Carreiras et al., 2005, Landi and Perfetti, 2007, Liu et al., 2003) have reported a P2 effect which, upon close inspection (as helpfully pointed out by a reviewer of this manuscript), has the same appearance as the N300 effect (see also Coulson et al., 2005). It is often difficult to determine whether an ERP effect is an enhanced negativity in one condition or an enhanced positivity in the other and this one seems to be particularly ambiguous. Furthermore, the long duration of this effect (starting close to the peak of the P2 but extending past the P2, on to the end of the N400) makes it hard to determine its peak latency. Although the original reports (Barnea and Breznitz, 1998, Carreiras et al., 2005, Liu et al., 2003) implicated the P2 effect in phonological processing, the most recent report (Landi and Perfetti, 2007) found P2 effects in a semantic task as well, leaving its nature unclear. Given the close correspondence between the P2 and the word N300 effects in terms of scalp topography, time course, and semantic effects, this manuscript will make the parsimonious assumption that it is the same component, although one report (Landi and Perfetti, 2007) made the unsupported observation that the P2 effect was not as frontal as the MFN (Frishkoff et al., 2004, Frishkoff, 2007).

A source analysis of the P2 effect using LORETA placed it in the right hemisphere prefrontal region in the vicinity of BA 6 and BA 8 (Liu et al., 2003), whereas the MFN appeared to localize to the anterior cingulate (Frishkoff et al., 2004, Frishkoff, 2007) using BESA. These contrasting effects could indicate that they are different components (Landi and Perfetti, 2007) or it could reflect differences in the source analysis procedure. The source of the effect or effects would be important both for helping determine whether these are the same component and for interpreting their nature. For example, the anterior cingulate is a region of great current interest in studies of executive function (Bush et al., 2000, Holroyd and Coles, 2002, Isomura and Takada, 2004, Posner and Rothbart, 1991, Rushworth et al., 2004).

The present study was therefore conducted to perform source analysis on the N300 and to try to confirm results with an fMRI replication. In order to maximize comparability with the ERP dataset, the same experimental design in all respects was utilized. This was done with the recognition that the original N300 finding was a serendipitous observation made in a study (Franklin et al., 2007) designed to contrast backward and forward associative priming. The effect was observed primarily for symmetrically related filler pairs that were included to establish desired listwise parameters for the task. These symmetric pairs were conventional semantically related items (both associative and semantic features) and were not matched to the asymmetric items because they were not originally intended to be analyzed. Although it might be of interest to also be able to analyze the backward and forward pairs, existing datasets (e.g., Dien et al., in press) show that the N300 effect is not always present (an argument against it reflecting ASA) and thus appears to be controlled by as yet undetermined parameters. Thus, changing the stimulus list could potentially eliminate the subject of interest. There was also the concern that increasing the number of backward and forward priming trials would lengthen the experiment past the session time available. The decision was therefore made to proceed with the project using the current experimental design rather than let it be possibly derailed by this side interest.

The only comparison of a priori interest is that of the symmetrical versus unsymmetrical word pairs, based on the ERP results. The forward and backward pairs are not compared with the symmetric pairs because they are not of primary interest and are not properly matched either in terms of number of stimuli or psycholinguistic parameters. Furthermore, there are too few of them to provide sufficient statistical power for the present fMRI analyses (although they did yield some effects of interest for the ERP analyses, for which more subjects could be collected). The forward and backward priming pairs were included in the stimulus set solely to maximize correspondence with the ERP dataset due to concern that changing the stimulus set could possibly adversely affect the N300 effect and are treated as filler pairs for the purposes of the present analysis. To confirm that nothing of interest was being overlooked, however, the forward-related versus forward-unrelated and backward-related versus backward-unrelated contrasts were examined.

There is a need to first consider the co-registration strategy implemented in this study. The first issue is whether it is preferable to collect data concurrently or in separate samples. There are both advantages and drawbacks to each approach. The chief advantage of collecting the data concurrently is that it eliminates any concern that the data might not be comparable due to differences in the sample or in the context of the experimental sessions. On the other hand, effects that do not generalize out of the scanner setting or the sample would be of reduced interest to ERP researchers. Drawbacks to concurrent collection also include strong electroencephalographic (EEG) artifacts induced by the rapidly alternating magnetic fields, loss of expensive fMRI data when the corresponding ERP data prove to be unusable (especially likely due to the EEG scanner artifacts and the weight of the prone head on the electrodes), and longer fMRI scanning sessions (that again increases the costs). Thus, concurrent collection risks both higher costs and reduced data quality compared to that obtained when separate recording sessions are used. This report therefore takes the approach of optimizing data quality via separate recording sessions, with the caveat that divergent results must be taken skeptically; conversely, results that converge point toward robust generality even under differing experimental contexts and samples.

Another issue is whether it is even reasonable to expect that results might correspond across two such very different recording modalities. An initial examination of this issue in an animal model (Logothetis et al., 2001) reported that both electrophysiological and fMRI data reflect the same aspect of neuronal activity, the modulation of dendritic inputs. Although there was a subsequent report that they did not correspond linearly (Devor et al., 2003), it has been reported that linearity is a reasonable approximation for their relationship (Sheth et al., 2004); furthermore, it has been reported that these nonlinearities are primarily observed when rats are anesthetized, not awake (Martin et al., 2006), consistent with reports in awake humans (Arthurs and Boniface, 2003).

Although overall, findings support the feasibility of identifying corresponding effects in ERP and fMRI data, reports suggest that the relationship may have a number of subtleties. For example, it has been reported that attentional distraction can reduce the fMRI response to somatosensory stimuli without reducing the ERP response that is thought to originate in the same region (Arthurs et al., 2004), although it is possible that this observation is instead an indication that they do not, in fact, correspond. It has also been reported in humans that ERP and fMRI relationships may not be the same in different cortical regions (Huettel et al., 2004). Overall, these reports suggest that it is reasonable to seek correspondences between the two data modalities but reinforce the normal hesitancy to interpret null effects.

A final issue is the strategy for localizing the ERP activity prior to the co-registration process. Two difficulties that are encountered when trying to perform source analysis (with point equivalent algorithms) is that specifying the wrong number of dipoles (Achim et al., 1991) or the presence of overlapping ERP components (Zhang and Jewett, 1993) can lead to substantial localization errors. The strategy pursued here is to try to isolate the ERP component using principal components analysis (PCA), a procedure that addresses both these issues. Results with both simulation datasets (Dien et al., 2007) and real data (Dien et al., 2003) suggest that it can be an effective approach. Although it is also possible to use a combination of expert judgment and appropriate experimental manipulations to generate compelling source solutions (e.g., Scherg et al., 1989), it is suggested that a procedure that can be more readily replicated by others is preferable.

It is the experience of the present experimenters that using PCA does not by any means guarantee success. A simple guideline for identifying a promising source solution is that the residual variance (the amount of the waveform not accounted for by the source solution) should be no more than 10%. Furthermore, another guideline is that even a good fit should be considered skeptically if the equivalent dipole is located at a deep level, such as the brain stem, since, in the experience of the present authors, the broadly distributed scalp topographies corresponding to such solutions allow for a very wide range of locations that still meet the 10% criterion. Finally, it should be recognized that point equivalent dipole solutions correspond to the location of the generator if it is a dimensionless point; the point solution therefore corresponds to a set of more superficial solutions with increasingly wider circumferences along the line determined by the orientation of the dipole (Scherg and Von Cramon, 1986). To be a plausible solution, a suitable cortical surface should be positioned along the line ranging from the point to the surface of the skull.

Because there is no current consensus on how to evaluate confidence intervals for such point equivalent dipole solutions, the procedure was adopted of only evaluating the best fit location for the source solution and only judging the fit a success if it was located in relatively close proximity to an appropriate fMRI activation. The decision was therefore made to avoid the common practice of seeding the dipole locations according to fMRI activations out of concern that it might allow too much latitude for correspondence. Seeding also has the drawback that doing so ignores the point source nature of point equivalent dipoles; for anything other than an activation with zero radius (impossible of course), seeding the dipole within the fMRI activation area is likely to be inappropriate. Instead, we followed the procedure of deriving the best fit location and then tracing the line from the point location to the cortical surface along the line of the dipole orientation to determine if an fMRI activation lay along this line; such a procedure is more consistent with the nature of point equivalent dipole solutions.

Whereas the lack of a significant fMRI activation at the location of the source solution would not be interpretable for a number of reasons already considered, as well as due to the always present possibility of a Type II error due to insufficient statistical power, it is suggested that the presence of correspondence would provide strong support for the N300 source solution as well as the status of the word N300 as being a genuine ERP component separate from the N400. In doing so, it should be noted for the ERP readership that fMRI analyses (using SPM) incorporate strong multiple comparison controls.

To summarize, the same lexical decision task paradigm as the previous ERP study was used (Franklin et al., 2007), using only the long SOA (500 ms) condition. Thus, backward associative, forward associative, and symmetrical associative + categorical prime pairs were utilized. The focus of analysis is the symmetrical priming contrast that was significant for the ERP data. The intention is to leave the issues regarding ASA, SOA, and priming type for future ERP studies. The first chief question is whether a source analysis of the word N300 effect (and thus potentially of the MFN and/or the P2 effects) could be verified with an fMRI replication, given the uncertainties of ERP source analysis methods. Potential sites are the anterior cingulate (Frishkoff et al., 2004, Frishkoff, 2007) and dorsal prefrontal cortex (Liu et al., 2003). The second chief question is whether the fMRI data could provide guidance as to whether the word N300 effect is indeed an enhanced N300 for unrelated targets or an enhanced P2 for related targets by determining the direction of the corresponding BOLD response. If a plausible source is identified, then the corresponding functional neuroanatomical literature will be utilized to generate a preliminary hypothesis for this ERP effect(s) that thus far has no specific cognitive function associated with it (them).

Section snippets

Behavioral data

A one-way repeated measure ANOVA was conducted for only correct “word” responses for participants' reaction times to forward-related, forward-unrelated, backward-related, backward-unrelated, and symmetrical primes. Arithmetic means of the individual participants' median RTs are presented in Table 1. It was found that the median reaction times for the individual prime types significantly differed, TWJt/c(4, 12.8) =8.90, p = .013. Additionally, further analysis showed that there were significant

Discussion

Using a lexical decision task, a significant change in activity in response to symmetrically related versus unrelated word pairs was found in the dPCC. A right-lateralized portion of this cluster corresponds closely to the source analysis location for the word N300 component reported earlier by this lab (Franklin et al., 2007). Thus, the fMRI study fully bore out the prediction by the ERP data that there would be a PCC semantic priming effect. This study therefore suggests a possible role for

Conclusion

The significant activity in the dPCC found in this study mirrors the word N300 ERP activity seen in a previous study (Franklin et al., 2007), with both having a stronger response to symmetric priming word pairs in a lexical-decision task and co-registering to the same region. This finding is important as it provides additional clues regarding the nature of this poorly understood region. This review also shows how a number of apparently independent lines of electrophysiological studies

Participants

Seventeen participants (F = 13, M = 4) aged 20–41 years were recruited for monetary compensation. All participants were right-handed, native English-speaking, had no history of neurological damage or disorders, and were not taking psychotropic medications.

Stimuli

The experiment consisted of 360 prime–target pairs, divided into four blocks of 90 trials each, in an event-related design. The stimuli included asymmetrically associated word pairs consisting of 40 compound items (i.e., compound words such as

Acknowledgments

This study was supported by a grant from the University of Kansas General Research Fund and pilot funding from the Hoglund Brain Imaging Center. We also thank the anonymous reviewers who materially improved the manuscript with their insightful comments.

References (135)

  • J.F. Connolly et al.

    Event-related potential sensitivity to acoustic and semantic properties of terminal words in sentences

    Brain Lang.

    (1992)
  • J.F. Connolly et al.

    The effects of phonological and semantic features of sentence-ending words on visual event-related brain potentials

    Electroencephalogr. Clin. Neurophysiol.

    (1995)
  • D.A. Copland et al.

    Brain activity during automatic semantic priming revealed by event-related functional magnetic resonance imaging

    NeuroImage

    (2003)
  • D.A. Copland et al.

    Neural correlates of semantic priming for ambiguous words: an event-related fMRI study

    Brain Res.

    (2007)
  • G.I. de Zubicaray et al.

    Cerebral regions associated with verbal response initiation, suppression and strategy use

    Neuropsychologia

    (2000)
  • A.E. Desjardins et al.

    Removal of confounding effects of global signal in functional MRI analyses

    NeuroImage

    (2001)
  • J.T. Devlin et al.

    Susceptibility-induced loss of signal: comparing PET and fMRI on a semantic task

    NeuroImage

    (2000)
  • A. Devor et al.

    Coupling of total hemoglobin concentration, oxygenation, and neural activity in rat somatosensory cortex

    Neuron

    (2003)
  • M.T. Diaz et al.

    Electrophysiological differentiation of phonological and semantic integration in word and sentence contexts

    Brain Res.

    (2007)
  • J. Dien et al.

    Localization of the event-related potential novelty response as defined by principal components analysis

    Cogn. Brain Res.

    (2003)
  • J. Dien et al.

    Is “blank” a suitable neutral prime for event-related potential experiments?

    Brain Lang.

    (2006)
  • K.D. Federmeier et al.

    Picture the difference: electrophysiological investigations of picture processing in the two cerebral hemispheres

    Neuropsychologia

    (2002)
  • M.S. Franklin et al.

    Semantic priming modulates the N400, N300, and N400RP

    Clin. Neurophysiol.

    (2007)
  • G.A. Frishkoff

    Hemispheric differences in strong versus weak semantic priming: evidence from event-related brain potentials

    Brain Lang.

    (2007)
  • G.A. Frishkoff et al.

    Frontal and posterior sources of event-related potentials in semantic comprehension

    Cogn. Brain Res.

    (2004)
  • M.A. Goodale et al.

    Separate visual pathways for perception and action

    Trends Neurosci.

    (1992)
  • M.A. Goodale et al.

    An evolving view of duplex vision: separate but interacting cortical pathways for perception and action

    Curr. Opin. Neurobiol.

    (2004)
  • P. Hagoort et al.

    ERP effects of listening to speech: semantic ERP effects

    Neuropsychologia

    (2000)
  • J.P. Hamm et al.

    Comparison of the N300 and N400 ERPs to picture stimuli in congruent and incongruent contexts

    Clin. Neurophysiol.

    (2002)
  • H. Hill et al.

    Automatic vs. controlled processes in semantic priming—differentiation by event-related potentials

    Int. J. Psychophysiol.

    (2002)
  • H. Hill et al.

    SOA-dependent N400 and P300 semantic priming effects using pseudoword primes and a delayed lexical decision

    Int. J. Psychophysiol.

    (2005)
  • P. Koenig et al.

    The neural basis for novel semantic categorization

    NeuroImage

    (2005)
  • S.A. Kotz et al.

    Modulation of the lexical–semantic network by auditory semantic priming: an event-related functional MRI study

    NeuroImage

    (2002)
  • A. Kujala et al.

    Activation in the anterior left auditory cortex associated with phonological analysis of speech input: localization of the phonological mismatch negativity response with MEG

    Brain Res. Cogn. Brain Res.

    (2004)
  • M. Kutas et al.

    Electrophysiology reveals semantic memory use in language comprehension

    Trends Cogn. Sci.

    (2000)
  • N. Landi et al.

    An electrophysiological investigation of semantic and phonological processing in skilled and less-skilled comprehenders

    Brain Lang.

    (2007)
  • R.J. Maddock

    The retrosplenial cortex and emotion: new insights from functional neuroimaging of the human brain

    Trends Neurosci.

    (1999)
  • R.J. Maddock et al.

    Activation of left posterior cingulate gyrus by the auditory presentation of threat-related words: an fMRI study

    Psychiatry Res.

    (1997)
  • P. Marien et al.

    The lateralized linguistic cerebellum: a review and a new hypothesis

    Brain Lang.

    (2001)
  • C. Martin et al.

    Investigating neural–hemodynamic coupling and the hemodynamic response function in the awake rat

    NeuroImage

    (2006)
  • A. Matsumoto et al.

    Linking semantic priming effect in functional MRI and event-related potentials

    NeuroImage

    (2005)
  • M.B. Miller et al.

    Brain activations associated with probability matching

    Neuropsychologia

    (2005)
  • C.J. Mummery et al.

    Dual-process model in semantic priming: a functional imaging perspective

    Neuroimage

    (1999)
  • M. Nakic et al.

    The impact of affect and frequency on lexical decision: the role of the amygdala and inferior frontal cortex

    NeuroImage

    (2006)
  • F.A. Nielsen et al.

    Mining the posterior cingulate: segregation between memory and pain components

    NeuroImage

    (2005)
  • A. Pfefferbaum et al.

    ERPs to stimuli requiring response production and inhibition: effects of age, probability, and visual noise

    Electroencephalogr. Clin. Neurophysiol.

    (1988)
  • A. Pfefferbaum et al.

    ERPs to response production and inhibition

    Electroencephalogr. Clin. Neurophysiol.

    (1985)
  • N.A. Phillips et al.

    ERP measures of auditory word repetition and translation priming in bilinguals

    Brain Res

    (2006)
  • A. Raposo et al.

    Repetition suppression and semantic enhancement: an investigation of the neural correlates of priming

    Neuropsychologia

    (2006)
  • O.J. Arthurs et al.

    Attention differentially modulates the coupling of fMRI BOLD and evoked potential signal amplitudes in the human somatosensory cortex

    Exp. Brain Res.

    (2004)
  • Cited by (28)

    • Bilingual language processing: A meta-analysis of functional neuroimaging studies

      2020, Neuroscience and Biobehavioral Reviews
      Citation Excerpt :

      Although less obvious for language processing, the engagement of the posterior cingulate cortex during semantic tasks is documented in the literature. In a semantic priming experiment with a lexical decision task, O’Hare et al. (2008) reported that the activation of the posterior cingulate cortex was modulated by semantic priming, concluding that this region works as a mediator between the stimulus category and the response program. More recently, Krieger-Redwood et al. (2016) have investigated the functional coupling between posterior cingulate cortex and the dorsolateral prefrontal cortex during semantic decisions of varying difficulty, suggesting that the co-activation of these regions may underlie the ability to perform semantic judgements.

    • At least eighty percent of brain grey matter is modifiable by physical activity: A review study

      2017, Behavioural Brain Research
      Citation Excerpt :

      Anterior cingulate is active for identifying response conflicts [106], problem-solving [107], and processing of cognitive information [108], and it mediates executive functions [26], coordinates complex movements [92] and is associated with spatial memory acquisition [109]. Besides, posterior cingulate is reported to help for reading comprehension [110], retrieval of information [111], semantic priming [112], and moral judgments [113]. Among the deep structures of the brain, insula is a multisensory region that has enhanced activity with increasing intensity of exercise [114], sense of effort [115], and during exercise imaginary [116].

    • Architecture of the Cerebral Cortex

      2012, The Human Nervous System, Third Edition
    • Conscious intention to speak proactively facilitates lexical access during overt object naming

      2011, Journal of Memory and Language
      Citation Excerpt :

      We must point out that the underlying functionality of this P2 deflection is unclear and we certainly do not wish to claim that it reflects the process of lexical access by itself. In fact, P2 modulations have been reported before in language comprehension studies illustrating sensitivity to anticipatory effects, phonological priming and word class (e.g., Federmeier, 2007; O’Hare, Dien, Waterson & Savage, 2008; ter Keurs, Brown, Hagoort, & Stegeman, 1999). Most of these studies have related the P2 effects to the ones encountered in visual attention, namely a modulation engendered by an attention-driven analysis related to the physical properties of a visual stimulus (e.g., Luck & Hillyard, 1994).

    • Neural enhancement and attenuation induced by repetitive recall

      2011, Neurobiology of Learning and Memory
      Citation Excerpt :

      Correct retrieval and execution of intentions are also associated with the lateral parietal region (Hashimoto, Umeda, & Kojima, 2011). Enhanced activity associated with related words/semantic priming (O’Hare, Dien, Waterson, & Savage, 2008; Rossell, Bullmore, Williams, & David, 2001) in the medial parietal region suggests explicit retrieval (Henson et al., 1999). Successful retrieval responses in both cued recall and recognition tests have been associated with the left inferior parietal area (Salami et al., 2010).

    View all citing articles on Scopus
    View full text