Elsevier

Brain and Language

Volume 143, April 2015, Pages 81-96
Brain and Language

Decomposition, lookup, and recombination: MEG evidence for the Full Decomposition model of complex visual word recognition

https://doi.org/10.1016/j.bandl.2015.03.001Get rights and content

Highlights

  • We report an MEG study of morphological decomposition in visual word recognition.

  • We test the predictions of Taft’s Full Decomposition model, using corpus measures.

  • We demonstrate a late effect of a novel statistical measure, semantic coherence.

  • The results support Taft’s model of decomposition, lookup, and recombination.

Abstract

There is much evidence that visual recognition of morphologically complex words (e.g., teacher) proceeds via a decompositional route, first involving recognition of their component morphemes (teach + -er). According to the Full Decomposition model, after the visual decomposition stage, followed by morpheme lookup, there is a final “recombination” stage, in which the decomposed morphemes are combined and the well-formedness of the complex form is evaluated. Here, we use MEG to provide evidence for the temporally-differentiated stages of this model. First, we demonstrate an early effect of derivational family entropy, corresponding to the stem lookup stage; this is followed by a surface frequency effect, corresponding to the later recombination stage. We also demonstrate a late effect of a novel statistical measure, semantic coherence, which quantifies the gradient semantic well-formedness of complex words. Our findings illustrate the usefulness of corpus measures in investigating the component processes within visual word recognition.

Introduction

According to the Full Decomposition model of morphologically complex visual word recognition (Taft, 1979, Taft, 2004, Taft and Forster, 1975), complex visual words, composed of a stem and affix, are recognized via a multi-stage process of decomposition, lookup, and recombination. In the first stage, a complex visual word is decomposed into its component morphemes, based on the visual form of these morphemes. In the second stage, the lexical entries for the component morphemes are accessed from their form. Finally, after the lexical entries for the morphemes have been looked up, the separate morphemes are then recombined into the complex form. For example, according to the Full Decomposition model, a complex visual word like cats would first be decomposed into cat + -s, the lexical entries for the stem cat and the suffix -s would then be consulted, and finally, the meanings of the stem (i.e., a particular type of furry animal) and suffix (i.e., plural) would be combined to obtain the meaning of the whole form cats (i.e., several of a particular type of furry animal). A more recent version of a decompositional model was proposed by Crepaldi, Rastle, Coltheart, and Nickels (2010), in which they used a behavioral masked morphological priming effect for irregular verbs to argue for an initial morpho-orthographic stage of decomposition (for regular and pseudo-regular words), followed by a lemma-level stage of lexical access for the component morphemes of a word (irrespective of orthographic regularity); notably, this theory predicts differential effects for regular and irregular verbs, since the former should benefit from priming at both the orthographic level and the lemma level, while the latter should only benefit from priming at the lemma level. Following up on this study, Fruchter, Stockall, and Marantz (2013) found an early (i.e., ∼170 ms) MEG masked morphological priming effect for irregular verbs, which, contrary to Crepaldi et al. (2010), they used to argue for an initial stage of form-based decomposition for both regular and irregular verbs.1

In contrast to the decompositional view of complex visual word recognition, some models allow for recognition of complex words via their entire surface forms without decomposition (i.e., without access to the component morphemes of the word). In particular, Giraudo and Grainger’s (2000) supralexical model argues for an initial stage of whole-word processing, followed by a later stage of decomposition. Other models that fall under this general umbrella include Pinker and Prince’s (1988) dual-route model, which argues for non-decompositional processing for irregular verbs, and Baayen, Milin, Đurđević, Hendrix, and Marelli’s (2011) amorphous model, which explains seemingly morphological effects during visual word recognition without recourse to a specifically morphological level of representation (but see Marantz, 2013, for an interpretation of Baayen et al., 2011, that is in fact consistent with contemporary linguistic theories of morphology). Thus, there arises a crucial dividing line in the various theoretical accounts of morphologically complex visual word processing: is morphological decomposition taken as a prerequisite of lexical access for complex words? In the present study, we consider the theories with the clearest predictions on this issue, namely the Full Decomposition model and the supralexical model, in order to ask the question: does morphological decomposition takes place prior to, or subsequent to, lexical access? We use MEG data from a visual lexical decision experiment to provide new evidence that, as predicted by the Full Decomposition model, there is an initial stage of morphological decomposition, a later stage of lexical access for the component morphemes of a complex word, as well as a final recombination stage.

In previous psycholinguistic and neurolingusitic work, the processing of morphologically complex words has been shown to involve a number of properties crucially tied to the identity, derivational and inflectional morphological family, and combinatoric potential of their component morphemes. At 80–100 ms after visual presentation of a complex word, there is an early MEG evoked response known as the M100 (associated with the Type I response of Tarkiainen, Helenius, Hansen, Cornelissen, & Salmelin, 1999), localized to the occipital lobe, which is sensitive to visual features of the stimulus. At 150–200 ms post-stimulus onset, the transition probability from stem to derivational affix (e.g., p(teacher|teach)) modulates an MEG evoked response known as the M170, localized to the Visual Word Form Area in the left fusiform gyrus (Solomyak & Marantz, 2010). Behavioral work has shown that even pseudo-morphological structure plays a role in visual word recognition: in a masked priming paradigm, both genuinely affixed (e.g., teacher-TEACH) and pseudo-affixed (e.g., corner-CORN) pairs exhibited significant priming effects in lexical decision, while comparable orthographic controls (e.g., brothel-BROTH) failed to show such behavioral facilitation (Rastle, Davis, & New, 2004). Following up on this work, Lewis, Solomyak, and Marantz (2011) found that the transition probability from pseudo-stem to pseudo-affix (e.g., p(corner|corn)) also modulates the M170, suggesting that this evoked response represents a brain index of an early stage of visual form-based morphological decomposition. Several EEG masked priming studies have demonstrated morphological priming effects at the N250 response, both in the case of derived forms (Lavric et al., 2007, Morris et al., 2007, Morris et al., 2008), as well as inflected forms (Morris and Stockall, 2012, Royle et al., 2012). Recently, Beyersmann, Iakimova, Ziegler, and Colé (2014) demonstrated EEG effects of morphological priming at 100–250 ms, while effects of semantic priming were not yet visible at that latency. Using MEG, Lehtonen, Monahan, and Poeppel (2011) found similar effects for regular derived words at a latency of ∼225 ms, and Fruchter et al. (2013) demonstrated effects of morphological priming for irregular past tense verbs at an even earlier latency (∼170 ms). In summary, there is extensive evidence for an early stage of orthographic decomposition into the component morphemes of a complex visual word, regardless of whether the word is derived or inflected, regular or irregular, or even if the word is only apparently complex (as in the case of pseudo-affixed items, such as corner).

After the initial decomposition stage, lexical access for the decomposed morphemes arguably takes place in the left superior and middle temporal regions at 300–400 ms post-stimulus onset, as indexed by the M350 evoked response (Pylkkänen & Marantz, 2003). The localization of lexical access to the left temporal lobe, and specifically the left middle temporal gyrus, is supported by the findings of many previous studies (e.g., Binder et al., 1997, Friederici, 2012, Hickok and Poeppel, 2007, Indefrey and Levelt, 2004). In the fMRI domain, Devlin, Jamison, Matthews, and Gonnerman (2004) observed effects of masked morphological and semantic priming in the left middle temporal gyrus; similarly, Gold and Rastle (2007) found effects of masked semantic priming in the same region. In the MEG literature, the relationship between the M350 and lexical access is supported by the fact that measures related to the lexical identity and morphological family of the stem modulate the M350; in particular, Solomyak and Marantz (2010) showed an effect of lemma frequency on the M350, and Pylkkänen, Feintuch, Hopkins, and Marantz (2004) showed that family frequency and family size modulate the M350, albeit in ways that were not entirely predicted. In the EEG domain, many studies have demonstrated lexical effects at the ERP analogue of the M350, namely the N400 response (e.g., Van Petten & Kutas, 1990; see Lau, Phillips, & Poeppel, 2008, for a review of findings relating the N400 to lexical access). Recently, Laszlo and Federmeier (2014), using a single-trial correlational analysis of ERPs, demonstrated a separate time course of effects for variables that track relatively perceptual features (e.g., bigram frequency) as compared to variables that track relatively semantic features (e.g., number of lexical associates), with the former variables becoming relevant as early as ∼130–150 ms, while the latter variables became relevant only at ∼300–340 ms.

Finally, after visual morphological decomposition, as well as lexical access for the decomposed morphemes, the last stage of processing for complex words arguably involves the recombination of stem and affix (Taft, 1979, Taft, 2004). Several previous studies have provided evidence suggestive of the presence of such a recombination stage for complex words. In particular, Domínguez, de Vega, and Barber (2004) compared morphological and pseudo-morphological overt (i.e., unmasked) priming effects in Spanish, and found that the former led to a sustained attenuation of the N400, while the latter produced a similar attenuation initially (i.e., at 250–350 ms), but which differentiated itself over time to form a delayed N400. Lavric, Rastle, and Clapp (2011) found comparable effects in English for morphological (e.g., teacher-TEACH) and pseudo-morphological (e.g., corner-CORN) pairs. Extending these findings to a non-priming paradigm, Lavric, Elchlepp, and Rastle (2012) demonstrated similar effects for single word recognition of affixed (e.g., teacher) and pseudo-affixed (e.g., corner) words, providing further support for the notion of a two-stage process of complex visual word recognition, in which affixed and pseudo-affixed words are initially processed in a similar fashion but diverge from each other over time. Although these studies did not specifically refer to Taft, 1979, Taft, 2004 recombination stage as an explanation of their results, they nevertheless provide initial empirical support for the presence of an early morpho-orthographic decomposition stage, followed by a later stage in which the semantics of the complex word become relevant.

In the present study, we report findings from an MEG visual lexical decision experiment with suffixed words. Our results build on the prior experimental findings from the literature in a number of important ways. First, we investigate whether a statistical measure derived from the lexical frequencies of the morphological family members of a stem, termed the derivational family entropy, plays a role in modulating the brain activity associated with lexical access (i.e., the M350). Our use of this variable is inspired by a model of morphological processing (Moscoso del Prado Martín, Kostić, & Baayen, 2004) that argues that complex word recognition is most affected by the statistical relations among members of a morphological family, rather than mere stem frequencies. Here, we test the applicability of this model of morphological processing to neural activity, as measured with MEG: in particular, we examine the role of derivational family entropy in modulating activity within the cortical regions and time windows associated with lexical access for the decomposed morphemes.

In addition to investigating the lexical access stage of complex word recognition, we also aim to provide neural evidence for the subsequent recombination stage. In particular, derivational family entropy, as an index of stem lookup, should modulate brain activity prior to the effects of surface frequency, as an index of recombination of stem and affix. Furthermore, we argue that the recombination stage must involve a semantic evaluation of the combination of the component morphemes of a given complex word. In other words, regardless of the fact that a particular complex word (e.g., predictable) has been previously encountered, the Full Decomposition model predicts that during recognition, one must nevertheless evaluate the well-formedness of the combination of its constituent parts (e.g., predict + -able). In the present study, we develop a novel methodology for quantifying the degree of semantic well-formedness, which we call semantic coherence. Given the timing of the various stages within the Full Decomposition model, we expect that the effects of semantic coherence should occur at the later stages of the visual recognition process: in particular, the semantic effects, like the effects of surface frequency, should occur after the effects of derivational family entropy, which relate to stem lookup.

In our analysis of the morpheme lookup stage of the Full Decomposition model, we examine the effects of the information content of a word’s morphological families upon lexical access for its previously decomposed stem. We use the derivational family entropy measure to capture this information content. Derivational family entropy, as will be defined formally within Section 2.1.1.1, is a statistical measure of the distribution of lexical frequencies within the derivational family of the stem, such that more balanced distributions of frequencies correlate with higher entropy, and conversely, distributions of frequencies dominated by a few family members have lower associated entropy.2 Our use of this measure follows the work of Moscoso del Prado Martín et al. (2004), who used a measure of morphological family entropy which also included the inflectional families nested within the derivational families; here, we simplify the measure to only model the distribution of frequencies within the derivational families.

Moscoso del Prado Martín et al. (2004) studied the entropy of morphological families primarily as a component of the total information residual of a word, which is simply the sum of the surface frequency and family entropy measures. Thus, the information residual measure encapsulates both the information content of the word itself, as well as the support it receives from its morphological families. Taking advantage of the temporal resolution of MEG, here we separately examine both the hypothesized earlier derivational family entropy effect on lexical access for a stem, as well as the hypothesized later surface frequency effect on recombination of stem and affix.

In our analysis of the recombination stage of the Full Decomposition model, we also examine the effects of semantic composition, that is to say the building of complex meanings from the meanings of individual parts, within morphologically complex words. Previous work on semantic composition within words has shown that the semantic constraints with respect to combining a particular prefix with a stem play a role in modulating an MEG response at around the same latency as the M350 evoked response, but from a medial frontal area shown to be sensitive to variables associated with semantic composition during sentence comprehension (Pylkkänen, Oliveri, & Smart, 2009). The characteristics of this “Anterior Midline Field” (AMF) in the study, namely its latency and its modulation by the semantic relationship between morphemes, are consistent with the properties of the recombination stage of complex word recognition. However, the notion of semantic selection involved in the study was categorical: Pylkkänen et al. (2009) studied the reversative prefix un-, which can be attached to verbs whose action can be undone in a particular way, but which produces unacceptable forms when attached to verbs of the wrong semantic type. For reasons elaborated upon within their article, phrases like the toilet was being unclogged were taken as examples of acceptable semantic forms, while phrases like the toilet was being unflushed were taken as examples of semantic violations. Thus the Pylkkänen et al. (2009) study used a violation paradigm and a binary comparison between acceptable and unacceptable forms: the unacceptable forms yielded increased activation in the brain region and in the time window associated with semantic composition from previous studies (though these prior studies focused on semantic composition between words, as will be discussed further below). Their work thus leaves open the question of whether the semantic AMF effects for derivationally complex words are only obtained within a pure violation paradigm, or whether comparable effects would also be obtained during a more ecologically valid paradigm with fully acceptable complex words, distributed over a range of semantic well-formedness.

Here, we are interested in exploring precisely this question, namely whether the relevant semantic variable modulating the AMF is more general and would be sensitive to a gradient of semantic coherence between stem and affix for fully acceptable (and existing) words. Varying semantic well-formedness is expected even for productive affixation since the affixes themselves are associated with semantic properties that make them more or less compatible with different stems. Taking one illustrative example from the theoretical literature on derivational morphology, consider Riddle’s (1985) observation that words containing the suffix -ness (as opposed to -ity) tend to denote embodied traits; this may lead to the prediction that words such as acuteness, which refers to a mental ability rather than an embodied trait, have a lower semantic coherence of stem and affix than words such as awkwardness, a canonical embodied trait. Riddle points to pairs of words built on the same stem like brutalness and brutality as evidence for the claim that -ness carries semantic implications that make it “cohere” better to some stems than other. The word brutality is used to describe a behavior (“his brutality in that situation was shocking”) while the word brutalness describes an embodied state (“He’s got a brutalness the likes of which I’ve never seen,” Riddle p. 439). Semantic coherence in this sense could be expected to modulate the frequency of word use; if acuteness is less semantically coherent than awkwardness, we may also expect it to be used less often by English speakers, and thus have a lower surface frequency, after controlling for other relevant differences between the two words. This prediction leads us to develop a new model for the semantic coherence of a derived word, based on the deviation of the surface frequency from its expected value (given the stem frequency and a phonological measure of the transition from stem to suffix). Our statistical estimation of semantic coherence (described formally in Section 2.1.1.4) offers a novel methodology for evaluating the semantic well-formedness of complex words, as compared to other possible techniques, including the corpus-based measure of Latent Semantic Analysis (Landauer, Foltz, & Laham, 1998), behavioral judgments of the decomposability of complex words (Hay, 2000), as well as the corpus-derived semantic coherence of a word’s morphological family used by Hauk, Davis, Ford, Pulvermüller, and Marslen-Wilson (2006). In contrast with the latter measure, which relates more directly to the consistency of meanings within the morphological family of a stem, our instantiation of the semantic coherence measure instead utilizes a lexical frequency-based statistical model to estimate the semantic well-formedness of the combination of stem and affix.

Our study of semantic composition within words also builds upon previous studies of semantic composition between words. For example, a previous MEG study (Pylkkänen & McElree, 2007) looked at sentence-level “complement coercion” effects; these effects relate to the repair of an initial conflict between a verb and its complement. In sentences such as The author began the book, in which the verb began selects for an event-denoting object (e.g., writing the book), the noun phrase the book is regarded as being coerced from an entity to an event, in order to satisfy the semantic constraints of the verb. The sentences within the coercion condition were compared to control sentences, such as The author wrote the book, in which the verb does not conflict with its complement. A second study (Brennan & Pylkkänen, 2008) examined sentence-level aspectual coercion effects, in which a punctual verb conflicts with a durative modifier, such as The clown jumped for ten minutes (which yields a repetitive reading of the verb). In both of these studies, an AMF was observed as a neural correlate of the coercion condition. These experiments focused on coercion effects, since they arguably represent a purely semantic manipulation, without a corresponding change in the syntax (Pylkkänen & McElree, 2007); consequently, these paradigms enabled initial investigation into the neural basis of sentence-level semantic composition. Subsequent studies have examined the neural basis of semantic composition more directly; for example, Bemis and Pylkkänen (2011) analyzed the neural response to adjective-noun pairs, such as red boat, thus confirming an AMF effect at ∼400 ms for cases of semantic composition in minimal linguistic phrases. The present study explores the notion of a gradient intra-word semantic coherence, in the framework of our current understanding of the recognition of morphologically complex words. Within this framework, the brain activity around 300–500 ms post-stimulus presentation associated with the AMF indexes the stage of processing when the constituent morphemes of a complex word, having been identified as word forms and having their lexical entries activated, are recombined to determine the meaning and well-formedness of the combination. In other words, the semantic coherence of stem and affix should not be relevant for complex word recognition until this later recombination stage.

The neural generator of the AMF in the aforementioned studies is typically taken to be ventromedial prefrontal cortex (or orbitofrontal cortex). This localization for prefrontal semantic effects differs from what might be expected given the fMRI literature. For example, in their reviews of the neuroimaging findings on language processing, Friederici (2012) and Price (2012) both identified the left inferior frontal gyrus, and not orbitofrontal cortex, with semantic processing. Additionally, Thompson-Schill, D’Esposito, Aguirre, and Farah (1997) demonstrated effects of semantic selection in the left inferior frontal gyrus. Lehtonen, Vorobyev, Hugdahl, Tuokkola, and Laine (2006) argued that the ventral portion of the left inferior frontal gyrus (i.e., pars orbitalis) is responsible for semantic integration during processing of morphologically complex words (but see Tyler, Stamatakis, Post, Randall, & Marslen-Wilson, 2005, in which the authors use fMRI data on the English past tense to argue that left inferior frontal cortex subserves morpho-phonological segmentation). In a particularly relevant study, Vannest, Newport, Newman, and Bavelier (2011) observed effects of morphological complexity in the left superior temporal gyrus and the left inferior frontal gyrus, which they interpreted in light of Taft, 1979, Taft, 2004 stages of decomposition and recombination. Nevertheless, while there is indeed conflicting evidence regarding the prefrontal areas most associated with semantic (and morphological) processing, given the extensive MEG literature on semantic composition, here we decided to investigate the orbitofrontal cortex as our prefrontal semantic region of interest.

In summary, our study extends the prior research on the neural basis of semantic composition in several important ways: (i) we develop a novel statistical measure of semantic well-formedness for morphologically complex words (semantic coherence); (ii) we use this measure to examine the gradient effect of semantic well-formedness for fully acceptable items, as opposed to the binary comparison of acceptable and unacceptable items; and (iii) we focus on semantic composition within words (e.g., teach + -er), as opposed to the more traditional examples of semantic composition between words (e.g., red + boat; Bemis & Pylkkänen, 2011).

Our experiment consists of an MEG visual lexical decision task, with the stimuli of interest being 200 bimorphemic suffixed words, distributed over a range of frequencies and semantic coherence, though all are existing words. For the analysis of the MEG data, the anatomical regions of interest are the left superior and middle temporal regions, associated with the M350 (Pylkkänen and Marantz, 2003, Pylkkänen et al., 2002, Pylkkänen et al., 2004, Solomyak and Marantz, 2010), and the orbitofrontal cortex, associated with the AMF (Brennan and Pylkkänen, 2008, Pylkkänen and McElree, 2007, Pylkkänen et al., 2009). We will investigate the role of derivational family entropy and surface frequency in modulating activity in the left temporal lobe during the post-M170 time window. In accordance with the results of a previous behavioral experiment (Moscoso del Prado Martín et al., 2004), we expect to find facilitatory effects of entropy and surface frequency; in the context of an MEG experiment, this means that we expect to find a decrease in neural activity correlated with an increase in these variables. Additionally, we expect the entropy effect to occur during lexical access for the stem, which should precede the surface frequency effect, since that is hypothesized to occur during the later recombination stage. These findings would be consistent with the predictions of the Full Decomposition model (Taft, 1979, Taft, 2004, Taft and Forster, 1975).

We will also investigate the role of semantic coherence in modulating activity in the orbitofrontal cortex, the brain region associated with the AMF, during a later time window (i.e., during and after the M350 response). We expect to find a facilitatory effect of semantic coherence, due to the previous MEG study of un-prefixation (Pylkkänen et al., 2009), in which the anterior midline region showed increased amplitude for the semantic violation condition. Finally, we expect to find a positive correlation between semantic coherence and the alternative corpus-based measure of Latent Semantic Analysis, as well as facilitatory effects of the latter measure in orbitofrontal cortex.

Section snippets

Design and stimuli

The experiment consisted of a visual lexical decision task, with simultaneous MEG recording of the magnetic fields induced by electrical activity in the brain. The target stimuli were words chosen from four suffix families: -er, -ness, -ly, and -able. Membership in these suffix families was determined by using the morphological parsing of words in the CELEX database (Baayen, Piepenbrock, & Gulikers, 1995). We excluded all compound words from our analysis. We also excluded words if any of the

Behavioral results

The mean accuracy rate across all subjects was 89.7% (±4.2%). The mean RT across all subjects was 699.2 ms (±241.7 ms). RT was found to be inversely correlated with surface frequency (t = -7.06, p = 0.0001), stem frequency (t = −7.25, p = 0.0001), derivational family entropy (t = −2.48, p = 0.0076), and semantic coherence (t = −3.47, p = 0.0002), indicating that higher values for these frequency-based variables correlated with shorter RTs. As expected, the information residual variable was positively correlated

Discussion

The results of our experiment show that statistical measures based on lexical frequency can provide a window into the spatiotemporally-differentiated stages of neural processing during complex visual word recognition.

The behavioral analysis confirmed the role of our linguistic variables of interest in the visual word recognition process. In particular, surface frequency, stem frequency, derivational family entropy, information residual, and semantic coherence were all shown to significantly

Conclusion

In summary, the results of this experiment provide evidence of the effects of morphological decomposition on lexical access, as indexed by derivational family entropy, and they also illuminate the recombination stage, as indexed by surface frequency and semantic coherence. With respect to the theoretical debate about visual processing of morphologically complex words, our results favor the Full Decomposition model (Taft, 1979, Taft, 2004, Taft and Forster, 1975) over non-decompositional models,

Acknowledgments

This work was supported by the National Science Foundation under Grant No. BCS-0843969, and by the NYU Abu Dhabi Research Council under Grant No. G1001 from the NYUAD Institute, New York University Abu Dhabi. We thank Gwyneth Lewis and Jeff Walker for their assistance in collecting the experimental data, Todd Gureckis for his comments on the original draft of the manuscript, David Poeppel, Liina Pylkkänen, Adam Buchwald, and Christian Brodbeck for discussion of the study, and several anonymous

References (67)

  • P. Indefrey et al.

    The spatial and temporal signatures of word production components

    Cognition

    (2004)
  • M. Lehtonen et al.

    Neural correlates of morphological decomposition in a morphologically rich language: An fMRI study

    Brain and Language

    (2006)
  • G. Lewis et al.

    The role of visual representations during the lexical access of spoken words

    Brain and Language

    (2014)
  • E. Maris et al.

    Nonparametric statistical testing of EEG- and MEG-data

    Journal of Neuroscience Methods

    (2007)
  • J. Morris et al.

    Early, equivalent ERP masked priming effects for regular and irregular morphology

    Brain and Language

    (2012)
  • F. Moscoso del Prado Martin et al.

    Putting the bits together: An information theoretical perspective on morphological processing

    Cognition

    (2004)
  • S. Pinker et al.

    On language and connectionism: Analysis of a parallel distributed processing model of language acquisition

    Cognition

    (1988)
  • C.J. Price

    A review and synthesis of the first 20 years of PET and fMRI studies of heard speech, spoken language and reading

    Neuroimage

    (2012)
  • L. Pylkkänen et al.

    Neural correlates of the effects of morphological family frequency and size: A MEG study

    Cognition

    (2004)
  • L. Pylkkänen et al.

    Tracking the time course of word recognition with MEG

    Trends in Cognitive Sciences

    (2003)
  • L. Pylkkänen et al.

    Neuromagnetic evidence for the timing of lexical activation: An MEG component sensitive to phonotactic probability but not to neighborhood density

    Brain and Language

    (2002)
  • P. Royle et al.

    The temporal dynamics of inflected word recognition: A masked ERP priming study of French verbs

    Neuropsychologia

    (2012)
  • O. Solomyak et al.

    Lexical access in early stages of visual word processing: A single-trial correlational MEG study of heteronym recognition

    Brain and Language

    (2009)
  • M. Taft et al.

    Lexical storage and retrieval of prefixed words

    Journal of Verbal Learning and Verbal Behavior

    (1975)
  • L.K. Tyler et al.

    Temporal and frontal systems in speech comprehension: An fMRI study of past tense processing

    Neuropsychologia

    (2005)
  • J. Vannest et al.

    Interplay between morphology and frequency in lexical access: The case of the base frequency effect

    Brain Research

    (2011)
  • Y. Adachi et al.

    Reduction of non-periodic environmental magnetic noise in MEG measurement by continuously adjusted least squares method

    IEEE Transactions on Applied Superconductivity

    (2001)
  • R.H. Baayen et al.

    An amorphous model for morphological processing in visual comprehension based on naive discriminative learning

    Psychological Review

    (2011)
  • R.H. Baayen et al.

    The CELEX Lexical Database (Release 2) [CD ROM]

    (1995)
  • D.A. Balota et al.

    The English Lexicon Project

    Behavior Research Methods

    (2007)
  • Bates, D., & Maechler, M. (2009). lme4: Linear mixed-effects models using S4 classes (R package version 0.999375-32)...
  • D.K. Bemis et al.

    Simple composition: A magnetoencephalography investigation into the comprehension of minimal linguistic phrases

    The Journal of Neuroscience

    (2011)
  • D.K. Bemis et al.

    Basic linguistic composition recruits the left anterior temporal lobe and left angular gyrus during both listening and reading

    Cerebral Cortex

    (2013)
  • Cited by (64)

    • Neural Processing of Morphology During Reading in Children

      2022, Neuroscience
      Citation Excerpt :

      Unlike inflected forms, derived forms constitute new lexical representations, raising the question of whether they are stored and retrieved as whole lexical items or decomposed into their morphological units like inflected forms (Matthews, 1991; Kirkici and Clahsen, 2013). The Full decomposition model (Taft and Forster, 1975; Taft, 1979, 2004; Fruchter and Marantz, 2015) argues that all morphologically complex words are decomposed before lexical access. In contrast, Hybrid models (Bozic et al., 2013a,b; Klimovich-Gray et al., 2016) argue that, unlike inflected words, derived words are stored and may undergo decomposition only after lexical access.

    • Electrophysiological correlates of character transposition in two-character Chinese word identification

      2021, Journal of Neurolinguistics
      Citation Excerpt :

      The two-character lexical decision task revealed word level semantic processing at 450 ms (Lo, McBride, Ho, & Maurer, 2019). In addition, some magnetoencephalography (MEG) studies investigating complex word processing further suggested that the effects of the visual form emerged in the first 100–250 ms after the stimulus presentation, while the semantic effects emerged later at 250–500 ms poststimulus (Fruchter & Marantz, 2015; Solomyak & Marantz, 2010). There has been substantial research effort dedicated to the electrophysiological correlates of visual word identification.

    • Morphological decomposition compensates for imperfections in phonological decoding. Neural evidence from typical and dyslexic readers of an opaque orthography

      2020, Cortex
      Citation Excerpt :

      Finally, studies with written derived words also show morphological effects in left middle temporal cortex (Devlin et al., 2004; Meinzer et al., 2009). These effects are evident around 350–400 msec after word presentation, as shown by MEG finding (Fruchter & Marantz, 2015; Hakala, Hulten, Lehtonen, Lagus, & Salmelin, 2018; Neophytou et al., 2018; Solomyak & Marantz, 2010) and hence are attributed to semantic re-composition of morphologically derived words. The tendency to decompose written words depends not only on linguistic properties of the words but also on the readers' language and reading skills.

    View all citing articles on Scopus
    View full text