Elsevier

NeuroImage

Volume 49, Issue 2, 15 January 2010, Pages 1837-1848
NeuroImage

Why do children make mirror errors in reading? Neural correlates of mirror invariance in the visual word form area

https://doi.org/10.1016/j.neuroimage.2009.09.024Get rights and content

Abstract

Young children often make mirror errors when learning to read and write, for instance writing their first name from right to left in English. This competence vanishes in most adult readers, who typically cannot read mirror words but retain a strong competence for mirror recognition of images. We used fast behavioral and fMRI repetition priming to probe the brain mechanisms underlying mirror generalization and its absence for words in adult readers. In two groups of French and Japanese readers, we show that the left fusiform visual word form area, a major site of learning during reading acquisition, simultaneously shows a maximal effect of mirror priming for pictures and an absence of mirror priming for words. Thus, learning to read recruits an area which possesses a property of mirror invariance, seemingly present in all primates, which is deleterious for letter recognition and may explain children's transient mirror errors.

Introduction

Many parents have made the surprising observation that their child, when learning to read and write, occasionally reverses left and right—whole words are occasionally written from right to left, and letters b, d, p, q are often confounded (Fig. 1). Although longitudinal data are lacking, the scarce evidence to date suggests that mirror reading and writing may be normal behaviors, distinct from dyslexia (Cornell, 1985, Schott, 2007). For instance, using a simple task which required subjects to write their name next to the right-hand margin of a page, Cornell (1985) found that essentially all 5- to 6-year-olds spontaneously wrote from right to left, while this behavior disappeared by age 8. Indeed, with the notable exception of Leonardo da Vinci, very few adults remain fluent in mirror reading and writing, although these abilities occasionally reappear following brain lesions (Pflugshaupt et al., 2007, Schott, 2007). It is therefore intriguing to ask why such a sophisticated behavior might be spontaneously present in young children, without having been trained, and why this competence seems to be lost in adults, although their ventral visual system remains able to recognize mirror images of objects (Eger et al., 2004, Vuilleumier et al., 2005). Understanding what neural mechanisms underlie this putative difference in mirror generalization between words and pictures in adults is an especially important goal since it has been claimed that the recognition of these two categories relies on shared visual mechanisms (Price and Devlin, 2003, Price et al., 2006). In the present paper, we address the latter question by probing mirror invariance for pictures and words in adults using behavioral and fMRI priming.

Our experiment is predicated upon a recent evolutionary and neurological perspective on reading, based on the concept of “neuronal recycling” (Dehaene, 2009, Dehaene and Cohen, 2007). Cultural inventions such as reading and mathematics are too recent to have influenced the human genome. Therefore, they must be acquired through the recycling of neuronal networks evolved for other purposes, but whose initial properties are sufficiently similar to the target function and which possess enough plasticity, particularly during childhood, for their functionality to be partially converted to this novel task (Dehaene, 2009, Dehaene and Cohen, 2007). In the case of reading, there is ample evidence that the acquisition of visual word recognition proceeds by progressively specializing a subpart of the left ventral visual system which has been termed the visual word form area (Baker et al., 2007, Ben-Shachar et al., 2007, Cohen et al., 2000, Gaillard et al., 2006). This region is reproducibly localized within a few millimeters in readers of all cultures (Bolger et al., 2005, Nakamura et al., 2005), suggesting that it may possess intrinsic biases in its retinotopic inputs (Hasson et al., 2002), preferred features (Changizi et al., 2006, Szwed et al., 2009), or connectivity (Epelbaum et al., 2008) that make it most appropriate for learning the visual shapes of letters and words. These biases, together with the general architectural properties of the ventral visual pathway for invariant object recognition (Serre et al., 2007, Ullman, 2007), seem to be co-opted for efficient visual word recognition in expert readers (Dehaene et al., 2005).

The recycling hypothesis holds that cultural learning is generally facilitated by the pre-existing properties of cortical tissue (e.g. size and location-invariant recognition). Occasionally, however, cultural learning may require the overcoming of biases that were useful in a prior environment, but are now counterproductive. Mirror errors fit in this framework. Monkey electrophysiological evidence indicates that the responses of some infero-temporal neurons show a property of mirror-image generalization, particularly across inversions of left and right (Baylis and Driver, 2001, Logothetis and Pauls, 1995, Rollenhagen and Olson, 2000). This feature presumably arose in the course of evolution because most natural visual categories are invariant across left–right changes (Corballis and Beale, 1976). It is deleterious for reading, however, where minimal mirror pairs such as p and q exist, and it may therefore impede reading acquisition, giving rise to transient mirror errors. Under this admittedly speculative hypothesis, mirror generalization would be an intrinsic property of some subpart of the visual cortex that would have to be “unlearned” as we become efficient readers.

At the behavioral level, there is considerable evidence for a progressive unlearning of symmetry generalization for letters and words. Even infants generalize across mirror views of simple objects (Bornstein et al., 1978), and this ability seems to persist into preschool years, even for letters, but to disappear when reading is acquired (Cornell, 1985, Schott, 2007). The causal role of reading acquisition in this loss is suggested by the observation that dyslexic children transiently perform better than normal in same-different judgments with mirror letters (Lachmann and van Leeuwen, 2007; see also Schneps et al., 2007), and that illiterate subjects and readers of a curvilinear language (Tamil) without mirror-image letters such as p and q continue to generalize across mirror figures, contrary to other readers (Danziger and Pederson, 1998, Kolinsky et al., 1987, Pederson, 2003).

The underlying brain mechanisms, however, remain to be studied. Obtaining direct evidence for our theoretical framework would ultimately require the longitudinal testing of mirror-image generalization in children, or alternatively the scanning of adult illiterates. Prior to conducting such complex studies, however, the present work aimed at testing a simpler prediction in normal adult readers. Our framework predicts that the visual word form area should be the site of the major difference in mirror invariance for pictures and words—it should show mirror priming for pictures, but not for words. This prediction is of general interest inasmuch as it contrasts sharply with other researchers' claim that this cortical site implements a domain-general function which is shared by words and pictures (Price and Devlin, 2003, Price et al., 2006, Wright et al., 2008). Here, using fMRI repetition priming, we show in adult readers of two different cultures (French and Japanese), that at the very same fusiform location, mirror-image generalization exists for pictures but not for written words.

Section snippets

Behavioral priming during semantic categorization

During fMRI, participants performed a primed semantic task on target words and line drawings (size comparison, Fig. 2). Each target was preceded by either the same or a different prime, which appeared either in the same orientation or in mirror image. We first examined whether behavioral priming was present—our prediction being that mirror priming should be observed for pictures but not words.

Median correct response times were entered into an ANOVA with group (French or Japanese) as a

Discussion

Taken together, the two behavioral tasks confirm that mirror-image invariance is present for pictures of faces, tools, or animals, but is absent for written stimuli in familiar or unfamiliar scripts. This difference between words and pictures is compatible with our hypothesis that symmetry generalization is partially inhibited or “unlearned” when learning to read. fMRI priming associated these effects with the left occipito-temporal VWFA: when pictures were presented, this region showed

Conclusion

Ever since Orton (1937), mirror errors have been erroneously associated with dyslexia. Our results, together with others (Baylis and Driver, 2001, Eger et al., 2004, Logothetis and Pauls, 1995, Rollenhagen and Olson, 2000, Vuilleumier et al., 2005), suggest that mirror generalization is a normal property of the primate ventral visual system. Only the excessive prolongation of letter mirroring, beyond the age of 8 or 10, may indicate a reading deficit (Lachmann and Geyer, 2003, Terepocki et al.,

Participants

We tested a total of 26 volunteers, 13 French (seven females; mean age = 23 years) and 13 Japanese (three females; mean age = 23 years). All were right-handed native speakers of their respective languages. All gave written informed consent, and the study was approved by the appropriate national and regional ethical committees.

Behavioral same-different task

The stimuli for the behavioral same-different task, performed after fMRI, were 14 French words, 14 Japanese characters, 14 pictures of tools, 14 pictures of faces, 14 unknown

Acknowledgments

This fMRI experiment was part of a general research program on functional neuroimaging of the human brain which was sponsored by the Atomic Energy Commission (Denis Le Bihan). It was supported by Institut National de la Santé et de la Recherche Médicale (INSERM), Commissariat à l'Energie Atomique (CEA), Collège de France and Agence Nationale de la Recherche (ANR). KN was supported by the Japan Society for the Promotion of Science (19500264). We are grateful to Alexis Amadon for fMRI support,

References (63)

  • HassonU. et al.

    Eccentricity bias as an organizing principle for human high-order object areas

    Neuron

    (2002)
  • LinC.Y. et al.

    Repetition priming without identification of the primes: evidence for a component process view of priming

    NeuroImage

    (2007)
  • LogothetisN.K. et al.

    Shape representation in the inferior temporal cortex of monkeys

    Curr. Biol.

    (1995)
  • MaurerU. et al.

    Coarse neural tuning for print peaks when children learn to read

    NeuroImage

    (2006)
  • McCloskeyM. et al.

    A visually based developmental reading deficit

    J. Mem. Lang.

    (2000)
  • Mochizuki-KawaiH. et al.

    Learning-related changes of brain activation in the visual ventral stream: an fMRI study of mirror reading skill

    Brain Res.

    (2006)
  • PflugshauptT. et al.

    When left becomes right and vice versa: mirrored vision after cerebral hypoxia

    Neuropsychologia

    (2007)
  • PriceC.J. et al.

    The myth of the visual word form area

    NeuroImage

    (2003)
  • PriceC.J. et al.

    How reading differs from object naming at the neuronal level

    NeuroImage

    (2006)
  • SzwedM. et al.

    The role of invariant line junctions in object and visual word recognition

    Vis. Res.

    (2009)
  • TarrM.J. et al.

    Mental rotation and orientation-dependence in shape recognition

    Cogn. Psychol.

    (1989)
  • UllmanS.

    Object recognition and segmentation by a fragment-based hierarchy

    Trends Cogn. Sci.

    (2007)
  • AllisonT. et al.

    Human extrastriate visual cortex and the perception of faces, words, numbers and colors

    Cereb. Cortex

    (1994)
  • BakerC.I. et al.

    Visual word processing and experiential origins of functional selectivity in human extrastriate cortex

    Proc. Natl. Acad. Sci. U. S. A.

    (2007)
  • BaylisG.C. et al.

    Shape-coding in IT cells generalizes over contrast and mirror reversal, but not figure-ground reversal

    Nat. Neurosci.

    (2001)
  • BealeI.L. et al.

    Confusion of mirror images by pigeons and interhemispheric commissures

    Nature

    (1972)
  • Ben-ShacharM. et al.

    Differential sensitivity to words and shapes in ventral occipito-temporal cortex

    Cereb. Cortex

    (2007)
  • BolgerD.J. et al.

    Cross-cultural effect on the brain revisited: universal structures plus writing system variation

    Hum. Brain Mapp.

    (2005)
  • ChangiziM.A. et al.

    The structures of letters and symbols throughout human history are selected to match those found in objects in natural scenes

    Am. Nat.

    (2006)
  • CohenL. et al.

    The visual word form area: spatial and temporal characterization of an initial stage of reading in normal subjects and posterior split-brain patients

    Brain

    (2000)
  • CorballisM.C. et al.

    The psychology of left and right

    (1976)
  • Cited by (123)

    • Selective Inhibition of Mirror Invariance for Letters Consolidated by Sleep Doubles Reading Fluency

      2021, Current Biology
      Citation Excerpt :

      We evaluated both short- and long-term effects of the 3-week intervention, including the impact of post-training naps. In each RCT replica, we applied our same-different visual task, as used in previous behavioral41 and neuroimaging studies15,27,28: the participants were requested to distinguish the mirror presentations of stimuli (Figure 2, top). We also applied a writing task, performed on paper with pencil, where participants were asked to reproduce letters (Figure 5, top).

    View all citing articles on Scopus
    View full text