Elsevier

NeuroImage

Volume 48, Issue 1, 15 October 2009, Pages 291-302
NeuroImage

Cerebral correlates of analogical processing and their modulation by training

https://doi.org/10.1016/j.neuroimage.2009.06.025Get rights and content

Abstract

There is increasing interest in understanding the neural systems that mediate analogical thinking, which is essential for learning and fluid intelligence. The aim of the present study was to shed light on the cerebral correlates of geometric analogical processing and on training-induced changes at the behavioral and brain level. In healthy participants a bilateral fronto-parietal network was engaged in processing geometric analogies and showed greater blood oxygenation dependent (BOLD) signals as resource demands increased. This network, as well as fusiform and subcortical brain regions, additionally showed training-induced decreases in the BOLD signal over time. The general finding that brain regions were modulated by the amount of resources demanded by the task, and/or by the reduction of allocated resources due to short term training, reflects increased efficiency – in terms of more focal and more specialized brain activation – to more economically process the geometric analogies. Our data indicate a rapid adaptation of the cognitive system which is efficiently modulated by short term training based on a positive correlation of resource demands and brain activation.

Introduction

Analogical processing is a key component of fluid intelligence. An analogy is defined as the “agreement or equivalence between the ratios or relationships present in different cases or objects” (Baldwin, 1901). Analogical processing thus refers to the mapping of two domains or mental representations of concepts and involves different core processes such as (i) building a representation, (ii) retrieving a source for the analogy (selecting relevant features and inhibiting irrelevant ones) and building/identifying relations, (iii) mapping relations between source and target, and (iv) evaluating the analogy (Gentner, 1983, van der Meer and Klix, 1986, Holyoak and Thagard, 1995, van der Meer, 1996, Holyoak and Thagard, 1997, French, 2002, Kokinov and French, 2002, Holyoak, 2005). A formalization of an analogy is A׃A′ = B׃B′, that is, an identical or similar relation has to be detected in two distinct representations, structures, or domains (i.e., the source domain and the target domain, respectively). In other words, analogical thinking is the process of transferring knowledge from one representation to another based on analogical similarities between the two representations. This process comprises the identification of (at least partially) identical relations in both analogy domains or the mapping of critical relations from source domain to target domain or vice versa (van der Meer, 1996).

Despite the relevance of analogical processing in our daily cognitive processing, little is known about its cerebral correlates and how it is modulated by training or learning. Wharton et al. (2000) used Positron-Emission-Tomography (PET) and a delayed matching test to characterize the neural basis of processing geometric analogies. Participants either had to judge analogies or identities of successively presented spatial arrangements of four simple, nameable geometric objects of a certain shape, size, color, and texture (e.g., two triangles above a rectangle and a triangle). In distractor trials, target and source domain differed in respect to spatial relationship (position) or object relationship (shape, texture, color). In the identity condition, the same picture was present successively. There was a greater increase in cerebral blood flow in left hemispheric frontal, parietal, and occipital brain regions in the analogy condition as compared to the identity condition. In a repetitive transcranial magnetic stimulation (rTMS) study using the same task, reaction times in the analogy condition were faster after the application of rTMS pulses over the left prefrontal cortex (Boroojerdi et al., 2001). These studies suggest that the left frontal and the parietal cortex are engaged in solving geometric analogies. Using non-geometric analogical tasks such as letter-string analogies or verbal/semantic analogies, recent functional magnetic resonance imaging (fMRI)-studies have reported activation of the frontal and anterior cingulate cortices, as well as inferior parietal and temporal brain regions (Luo et al., 2003, Bunge et al., 2005, Geake and Hansen, 2005, Green et al., 2006, Wright et al., 2007, Wendelken et al., 2008). These studies used different ways to manipulate the difficulty of analogical processing. While Wharton et al. (2000) compared analogical and identity judgments, Geake and Hansen (2005) manipulated the difficulty of letter-string analogy problems by varying the number of transformations required to solve the analogy task. Processing more difficult as compared to easier analogies resulted in greater increases in blood oxygenation dependent (BOLD) signals in the bilateral frontal, parietal, and cingulate cortex (Geake and Hansen, 2005). Bunge et al. (2005) manipulated the difficulty of semantic retrieval via semantic association strengths of word pairs that had to be judged either as semantically analogous or related. The frontopolar cortex showed greater increases in BOLD signal during analogical than during semantic processing. More importantly, the left hemispheric anterior inferior prefrontal cortex showed greater activation in more difficult (i.e., less associated) than easier (i.e., stronger associated) stimulus pairs. These results indicate a positive correlation between task demands and increases in BOLD signal. Conversely, this also implies that the processing of tasks which are less demanding results in weaker signal changes. Such a reduction of task demands could occur as a result of learning or practicing the task. Correspondingly, if a task is practiced over the course of the experiment, one would expect a general reduction in BOLD signal changes, because the task difficulty should decline over time due to learning and training. If the behavioral advantage (i.e., reduced processing time and/or increased performance or accuracy) is accompanied by a reduced allocation of resources – the limited, system-specific set of activations available to enable the maintenance of representations and the execution of cognitive operations – it is often described as “increased cognitive efficiency” (Kahneman, 1973, Haier et al., 1992, Just and Carpenter, 1992, Just et al., 2003, Rypma et al., 2006, Toffanin et al., 2007). Notably, the term “cognitive efficiency” is mostly used as a concept related to individual differences in cognitive abilities introduced to describe the relation between brain activation and inter-individual differences in cognitive performance. Hence the concept of “cognitive efficiency” describes that individuals scoring higher in – for instance – fluid intelligence (Haier et al., 1988) display a more efficient brain functioning because of better myelination (e.g., Miller, 1994) or higher axonal and/or dendrite plasticity (e.g., Garlick, 2002). We argue, however, that higher efficiency of the cognitive system can also be characterized by short term training effects and underlying immediate changes in brain activation.

To solve a cognitive task, a certain amount of cognitive resources need to be allocated. In general, the amount of brain activation used in a given task is related to both resource consumption (i.e., resource demand of the task) and resource supply (i.e., resources provided by the individual). In other words, cognitive processes are modulated by the amount and type of resources demanded by the task relative to the resources available and supplied by the individual (Just et al., 2003).

Different approaches can be used to expand our understanding of the relations among task demands, behavioral performance, and brain activation in analogical processing. One can either correlate individual differences in analogy processing and brain activation across subjects (i.e., manipulation of resource supply, Just et al., 2003), or analyze the relationship of analogy tasks of varying difficulty and brain activation within subjects (i.e., manipulation of resource demands, Just et al., 2003), or relate training-induced changes within an analogy task to respective changes in brain activation within subjects (i.e., manipulation of resource supply of the individual due to training or learning). An increased efficiency would be indicated by increased behavioral performance together with stable or reduced BOLD signals, that is, a negative correlation between performance and the BOLD signal. Previous functional imaging studies have used the individual differences approach correlating brain activation (i.e., the BOLD signal) and measures of fluid intelligence (Gray et al., 2003, Lee et al., 2006; for review Jung and Haier, 2007), mathematical competence (O'Boyle et al., 2005, Grabner et al., 2007), or processing speed (Rypma et al., 2006); or they have studied the effect of varying task difficulties within subjects (for reasoning and analogy tasks see for instance Prabhakaran et al., 1997, Christoff et al., 2001, Kroger et al., 2002, Bunge et al., 2005, Geake and Hansen, 2005, Kalbfleisch et al., 2007, Kroger et al., 2008) to analyze the relationship between brain activation and behavior. Thus far, however, only a few functional imaging studies have focused on the mechanisms underlying short term learning or training related changes in brain activation within a given task (for a review on short and long term learning effects see Kelly and Garavan, 2005). Kelly and Garavan (2005) distinguish two types of training- or practice-related activation change: There may be a redistribution of brain activation in task related brain areas or there may be a functional reorganization of brain activity, that is, decreases in some brain areas and increases in other brain areas. Decreases of brain activity are attributed to increased neural efficiency, strengthening of particular neural networks, or more efficient use of specific neuronal circuits. Increase of brain activation is suggested to reflect recruitment of additional cortical units with practice. Functional reorganization is observed when the contribution of specific brain areas to task performance changes as a result of practice. None of the studies, however, focused on short term training effects in analogical processing and whether it goes in line with a redistribution or a reorganization of brain activity.

Because analogical processing and fluid reasoning underlies learning to solve new problems it is central to cognitive development (Wright et al., 2007). To enhance our understanding of the core mechanisms underlying analogical processing we need to advance our understanding of learning/training-induced brain plasticity in the shorter time domain. A recent study observed that over the course of learning arithmetic problems, the BOLD signal decreases in frontal, parietal and occipital regions and in the basal ganglia, whereas it increases in the angular gyrus and the temporal lobe (Ischebeck et al., 2007). Neubauer et al. (2004) combined the individual differences approach and a pretest-learning-posttest design and found a negative correlation between intelligence and brain activation reflected by event-related desynchronizations of the upper alpha band in the posttest only. Given the importance of analogical processing and learning in human cognition, it is striking that there is such a lack of profound knowledge regarding the cerebral correlates underlying analogical processing and training/learning-induced neural modulations in higher cognitive functions in the shorter time domain.

Wharton et al. (2000) studied the cerebral correlates of solving geometric analogies by comparing blocks of geometric analogy decisions to blocks of identity decisions. Two important questions, however, remain open: How does the difficulty of the analogy task, that is, changes in resources demanded by the task, affect BOLD signal changes in left frontal and parietal brain regions? And how does training over the course of the experiment, that is, changes in resources supplied by the individual, modulate the BOLD signal? The goal of the present study was to advance our knowledge about cerebral mechanisms underlying the processing of geometric analogies with varying difficulty, and to find out whether and how training modulates cerebral activation patterns over the course of the experiment. Based on the above mentioned literature we expect frontal and parietal brain regions, mainly in the left hemisphere, to be engaged in processing geometric analogies. Over the time-course of the current experiment we expect a training-induced behavioral improvement (decreased reaction times, increased accuracy) in line with a significant decrease or shift of the BOLD signal during analogical reasoning (within frontal, parietal, occipital regions and basal ganglia; see Ischebeck et al., 2007) indicating an increased processing efficiency.

Section snippets

Participants

Fifteen healthy male high school students (age 18.7 ± .5 years [mean, standard deviation]) participated. All attended the 12th grade of a Berlin school which has a special emphasis on mathematics and natural sciences. All participants were right-handed, had normal or corrected-to-normal vision, and no past neurological or psychiatric pathology. All participants gave written informed consent before the investigation and were paid for their participation. The study was approved by the local Ethics

Behavioral data

Within analogy trials repeated-measures analysis of variance (ANOVA) showed a significant main effect of task difficulty (easy, medium, and difficult analogies) on both reaction times (F(df) = 135.9(2), p < .001) and accuracy (percent correct: F(df) = 7.9(2), p = .002; dprime: F(df) = 12.0, p < .001). Paired t-tests showed significant faster reaction times for easy compared to medium trials (T =  8.8, p < .001), as well as for medium compared to difficult trials (T =  8.4, p < .001). Paired t-tests on accuracy

Discussion

The goal of the present study was to characterize the cerebral correlates underlying geometric analogy processing and to find out whether and how they are modulated by short term training effects. Our results indicate that a fronto-parietal network is engaged and modulated by task difficulty when solving geometric analogies. Our data replicate previous results on visual-spatial and analogical reasoning. Specifically, and in line with our data, Wharton et al. (2000) found a very similar left

Acknowledgments

This research was supported by grants of the BMBF (NIL – Neuroscience Instruction Learning, BNIC – Berlin NeuroImaging Center), EU (NEST 012778), Berlin School of Mind and Brain, and the Stifterverband für die Deutsche Wissenschaft (Claussen-Simon-Stiftung). We thank P. Kazzer for his help in data acquisition.

References (79)

  • GreenA.E. et al.

    Frontopolar cortex mediates abstract integration in analogy

    Brain Res.

    (2006)
  • HaierR.J. et al.

    Intelligence and changes in regional cerebral glucose metabolic rate following learning

    Intelligence

    (1992)
  • HaierR.J. et al.

    Cortical glucose metabolic rate correlates of abstract reasoning and attention studied with positron emission tomography

    Intelligence

    (1988)
  • HugdahlK. et al.

    Sex differences in visuo-spatial processing: an fMRI study of mental rotation

    Neuropsychologia

    (2006)
  • IschebeckA. et al.

    Imaging early practice effects in arithmetic

    NeuroImage

    (2007)
  • JenkinsonM. et al.

    A global optimisation method for robust affine registration of brain images

    Med. Image Anal.

    (2001)
  • JenkinsonM. et al.

    Improved optimisation for the robust and accurate linear registration and motion correction of brain images

    NeuroImage

    (2002)
  • KlingbergT.

    Development of a superior frontal-intraparietal network for visuo-spatial working memory

    Neuropsychologia

    (2006)
  • KrogerJ.K. et al.

    Distinct neural substrates for deductive and mathematical processing

    Brain Res.

    (2008)
  • LammC. et al.

    The functional role of dorso-lateral premotor cortex during mental rotation. An event-related fMRI study separating cognitive processing steps using a novel task paradigm

    NeuroImage

    (2007)
  • LeeK.H. et al.

    Neural correlates of superior intelligence: stronger recruitment of posterior parietal cortex

    NeuroImage

    (2006)
  • LittleD.M. et al.

    Prototype-distortion category learning: a two-phase learning process across a distributed network

    Brain Cogn.

    (2006)
  • LuoQ. et al.

    The neural substrate of analogical reasoning: an fMRI study

    Brain Res. Cogn. Brain Res.

    (2003)
  • MelroseR.J. et al.

    An fMRI investigation of the role of the basal ganglia in reasoning

    Brain Res.

    (2007)
  • MillerE.M.

    Intelligence and brain myelination: a hypothesis

    Pers. Individ. Differ.

    (1994)
  • NeubauerA.C. et al.

    Intelligence and individual differences in becoming neurally efficient

    Acta Psychol. (Amst)

    (2004)
  • O'BoyleM.W. et al.

    Mathematically gifted male adolescents activate a unique brain network during mental rotation

    Brain Res. Cogn. Brain Res.

    (2005)
  • PausT.

    Location and function of the human frontal eye-field: a selective review

    Neuropsychologia

    (1996)
  • PrabhakaranV. et al.

    Neural substrates of fluid reasoning: an fMRI study of neocortical activation during performance of the Raven's Progressive Matrices Test

    Cogn. Psychol.

    (1997)
  • PriceC.J. et al.

    Reading and reading disturbance

    Curr. Opin. Neurobiol.

    (2005)
  • QiuJ. et al.

    The neural basis of analogical reasoning: an event-related potential study

    Neuropsychologia

    (2008)
  • RypmaB. et al.

    Neural correlates of cognitive efficiency

    NeuroImage

    (2006)
  • SmithS.M. et al.

    Advances in functional and structural MR image analysis and implementation as FSL

    NeuroImage

    (2004)
  • SmithR. et al.

    Localizing the rostrolateral prefrontal cortex at the individual level

    NeuroImage

    (2007)
  • TurnerM.S. et al.

    Distinct roles for lateral and medial rostral prefrontal cortex in source monitoring of perceived and imagined events

    Neuropsychologia

    (2008)
  • UnterrainerJ.M. et al.

    Planning and problem solving: from neuropsychology to functional neuroimaging

    J. Physiol. Paris

    (2006)
  • WhartonC.M. et al.

    Toward neuroanatomical models of analogy: a positron emission tomography study of analogical mapping

    Cogn. Psychol.

    (2000)
  • WoolrichM.W. et al.

    Temporal autocorrelation in univariate linear modeling of FMRI data

    NeuroImage

    (2001)
  • WoolrichM.W. et al.

    Multilevel linear modelling for FMRI group analysis using Bayesian inference

    NeuroImage

    (2004)
  • Cited by (0)

    View full text