Research report
Spatial imagery in deductive reasoning: a functional MRI study

https://doi.org/10.1016/S0926-6410(01)00116-1Get rights and content

Abstract

Various cognitive theories aim to explain human deductive reasoning: (1) mental logic theories claim syntactic language-based proofs of derivation, (2) the mental model theory proposes cognitive processes of constructing and manipulating spatially organized mental models, and (3) imagery theories postulate that such abilities are based on visual mental images. To explore the neural substrates of human deductive reasoning, we examined BOLD (blood oxygen level dependent) contrasts of twelve healthy participants during relational and conditional reasoning with whole-brain functional magnetic resonance imaging (fMRI). The results indicate that, in the absence of any correlated visual input, reasoning activated an occipitoparietal–frontal network, including parts of the prefrontal cortex (Brodmann’s area, BA, 6, 9) and the cingulate gyrus (BA 32), the superior and inferior parietal cortex (BA 7, 40), the precuneus (BA 7), and the visual association cortex (BA 19). In the discussion, we first focus on the activated occipito-parietal pathway that is well known to be involved in spatial perception and spatial working memory. Second, we briefly relate the activation in the prefrontal cortical areas and in the anterior cingulate gyrus to other imaging studies on higher cognitive functions. Finally, we draw some general conclusions and argue that reasoners envisage and inspect spatially organized mental models to solve deductive inference problems.

Introduction

Reasoning is a cognitive process that yields conclusions from given premises. It occurs whenever human beings make implicit information explicit. This study is about one form of reasoning, deduction. By definition, in deductive reasoning, the truth of the premises ensures the truth of the conclusion. (In contrast to inductive reasoning, in which the truth of the premises does not warrant the truth of the conclusion.)

Many people report that they often think by visualizing objects and events. They typically experience reasoning as seeing the information from the premises and scanning this vivid mental image to find new, not explicitly given information. Various sorts of evidence are compatible with this assumption, including the well-known studies of the mental rotation and the mental scanning of images [40], [60].

However, in the behavioral sciences, the question of how people reason deductively is still open. Cognitive psychologists conducted behavioral experiments to investigate the cognitive processes underlying different kinds of deduction, such as conditional reasoning, syllogistic reasoning, relational reasoning, etc. Nevertheless, there is still controversy on how the experimental findings can be integrated into a general theory of human reasoning. Mental proof theories completely deny that reasoning is based on mental imagery, but rather on the application of language-like formal rules of inference [6], [55]. In contrast, the mental model theory postulates that reasoning does not rely on syntactic operations as in rule-based approaches, but rather on the construction and manipulation of spatially organized mental models [19], [29]. Such mental models represent situations spatially, but they can abstract away from visual details such as colors, textures, and shapes, which are not relevant to the problem.

The third approach is the visual mental imagery theory. Proponents of this account describe mental images used in reasoning as structurally similar to perceptions. Like visual precepts, visual mental images represent colors, shapes, and metrical distance, can be rotated and scanned, have a limited resolution [14], [40], and sometimes are so similar to real perceptions that the two can be confused [25]. Reasoning, from this point of view, is to ‘look’ mentally at a visual mental image to find new information not explicitly given in the premises.

The notion of spatial mental models and visual mental images is related to Kosslyn’s model in which mental imagery is composed of two different kinds of processes, one visual and one spatial [41]. The latter is concerned with what an image looks like from a certain point of view; the former depends on where an object is located relative to other objects. The role of visual mental images and spatial mental models in deductive reasoning has been studied extensively, for instance, in Refs. [33], [34].

All three cognitive approaches have been implemented in computational models. Hagert [24], for instance, proposed a computational account of relational reasoning that is based on the application of formal inference rules. Schlieder [58] implemented relational reasoning as the construction and inspection of spatial mental models [23]. A computational approach of visual mental imagery in reasoning has been developed by Glasgow and Papadias [21]. Further computational visual imagery approaches can be found in Glasgow et al. [22]. A comparison of spatial mental models and visual mental imagery in reasoning is given in Schlieder and Berendt [59].

In recent years, the debate within cognitive and computational theories of reasoning regarding sentential mental proofs, spatial mental models, and visual mental images also started in the cognitive neurosciences. On the neuroanatomical level, the sentential theory predicts that the language processing regions of the brain are involved in reasoning, whereas the spatial theory predicts that the cortical areas involved in spatial working memory, perception, and movement control are evoked by reasoning. The sentential theory, furthermore, predicts a dominance of the left hemisphere, whereas the spatial theory assumes a right hemispheric prevalence [19], [20], [27]. According to the visual theory, the primary visual cortex, or at least nearby visual regions, should be evoked by reasoning without a specific assumption concerning hemispheric differences [41].

While past studies on the neural basis of human reasoning were restricted to investigations with brain damaged patients [7], [9], [16], [17], [51], [54], more recent neuroimaging techniques, such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI), were used to study reasoning processes in the intact brain. However, the results of these studies are contradicting. On the one hand, Goel et al. [19], [20] investigated deductive (and inductive) reasoning problems in two PET studies and reported activation in temporal and prefrontal regions, mainly in the left hemisphere. These results appear to support the sentential theory. On the other hand, there are results that at least indirectly support the spatial and visual theories of reasoning. Prabhakaran et al. [53] studied problems selected from the Raven’s Progressive Matrices Test, which elicit reasoning, and found increased activity in right frontal and bilateral parietal regions. Osherson et al. [48] compared probabilistic and deductive thinking and found in the latter increased activation in right-hemisphere parietal regions. The visual theory is related to a series of studies that found activity in the primary visual cortex when participants manipulated objects and scenes in working memory [42], [43], [44], [57].

The aim of the present fMRI study is to explore the neural substrates of human deductive reasoning, and specifically, its visual and spatial components. We selected two essential sorts of human deductive reasoning: relational and conditional reasoning. In a typical relational reasoning problem, at least two relational terms X r1 Y and Y r2 Z are given as premises, and the goal is to find a conclusion X r3 Z that follows from the premises. In a conditional reasoning problem, the first premise consists of an ‘if p, then q’ statement and the second premise refers to the truth of the antecedent (‘if’ part) or the consequent (‘then’ part). The goal is to find a conclusion that follows from both premises. The two valid inferences are ‘if p, then q, and p is true, then q is true’ (modus ponens), and ‘if p, then q, and q is false, then p is false’ (modus tollens).

Section snippets

Participants

Twelve right-handed male students of Freiburg University (mean age=23.9, S.D.=3.3) participated in the experiment. None had any history of neurological or psychiatric disorders. They were paid for their participation and informed consent was obtained in writing. Before the brain imaging study started, participants attended a 20-min training experiment in which they solved 12 conditional and 12 relational sample reasoning tasks. Participants were not instructed in any way to choose a particular

Behavioral data

Overall, participants’ performance in the behavioral experiment was slightly better (86.7% correct) than inside the scanner (81.9% correct). However, since the patterns of results were identical outside and inside the scanner, in the following we report only data from the scanning. The analysis of response latencies shows that participants needed the same time for correct responses in the relational inferences (2.0 s) and the conditional problems (2.1 s) (t-test, t=1.074; P>0.285) and for modus

Discussion

The reported results can be summarized under two headlines. First, we focus on the most striking result of the present study, namely that reasoning activated the occipito-parietal pathway in the absence of any correlated visual input. Second, we briefly relate the activation in the prefrontal cortical areas and in the anterior cingulate gyrus to other imaging studies on higher cognitive functions. Finally, we draw some general conclusions an spatial mental models, visual imagery, and reasoning.

Conclusion

The aim of the present study was to investigate the neural correlates of human deductive reasoning. From the different sorts of deductive reasoning we selected two: relational and conditional reasoning. As proposed by several authors, we identified the spatial accounts of reasoning with (right) parietal activity, mental proof theories with activity in (left) temporal regions, and the visual account with activation of the primary visual cortex.

The present study yielded a surprising result: as a

Acknowledgements

The research was supported the German National Research Foundation (Deutsche Forschungsgemeinschaft; DFG) under contract numbers Str 301/5-2 (MeMoSpace project), Kn 465/2-1, and Kn 465/2-3 (WorkSpace project).

References (69)

  • S.M Kosslyn et al.

    Neural systems shared by visual imagery and visual perception: a positron emission tomography study

    NeuroImage

    (1997)
  • D Osherson et al.

    Distinct brain loci in deductive versus probabilistic reasoning

    Neuropsychologia

    (1998)
  • S Pigott et al.

    Capacity of visual short-term memory after unilateral frontal or anterior temporal-lobe resection

    Neuropsychologia

    (1994)
  • V Prabhakaran 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)
  • D.E Read

    Solving deductive-reasoning problems after unilateral temporal lobectomy

    Brain Lang.

    (1981)
  • C Schlieder

    The construction of preferred mental models in reasoning with interval relations

  • E.E Smith et al.

    Working memory: a view from neuroimaging

    Cogn. Psychol.

    (1997)
  • K Stenning et al.

    A cognitive theory of graphical and linguistic reasoning: Logic and implementation

    Cogn. Sci.

    (1995)
  • A.D Baddeley

    Working Memory

    (1986)
  • S.C Baker et al.

    Active representation of shape and spatial location in man

    Cereb. Cortex

    (1996)
  • P.A Bandettini et al.

    Processing strategies for time-course data sets in functional MRI of the human brain

    Magn. Reson. Med.

    (1993)
  • B Berendt

    Explaining preferred mental models in Allen inferences with a metrical model of imagery

  • M D’Esposito et al.

    The neural basis of the central executive system of working memory

    Nature

    (1995)
  • J.St.-B.T Evans et al.

    Human Reasoning

    (1993)
  • R.A Finke

    Principles of Mental Imagery

    (1989)
  • M.S Gazzaniga

    The Social Brain

    (1985)
  • G.C Geminiani et al.

    Deductive reasoning in right-brain damaged

  • M.S George et al.

    An H2150 PET study on the stroop and the emotional stroop

    Hum. Brain Mapp.

    (1994)
  • V Goel et al.

    The seats of reason: a localization study of deductive and inductive reasoning using PET (O15) blood flow technique

    NeuroReport

    (1997)
  • V Goel et al.

    Neuroanatomical correlates of human reasoning

    J. Cogn. Neurosci.

    (1998)
  • C. Hagen, C. Schlieder, M. Knauff, R. Rauh, G. Strube, A cognitive modelling of preferred mental models, 1998. Online...
  • G Hagert

    Modeling mental models: Experiments in cognitive modeling of spatial reasoning

  • Cited by (0)

    View full text