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

1
Neural correlates of morphological decomposition
in a morphologicallyrich language An fMRI study

• Lehtonen, M., Vorobyev, V.A., Hugdahl, K.,
  Tuokkola T., Laine M.

2
Morphologically Complefl Words

• SINGERS
• Most Indo-European languages, such as English or
  French, use inflectional affixes to quite a
  limited extent.
• Finnish has abundant inflectional possibilities
  a single noun in Finnish can have as many as 2000
  different forms
• HUONEISSANNEKIN
• room plural in your even
• even in your rooms

3
Morphological Decomposition

• When compared to otherwise matched monomorphemic
  words, Finnish morphologically complex words
• are usually processed more slowly
• with a greater probability of error.
• This suggests that morphological decomposition
  takes place during recognition of these words
• The recognition process of inflected words is
  assumed to require two major processing steps
• early morphological decomposition at the visual
  word form level
• integration of the meaning of the morphological
  constituents at the semantic-syntactic level.

4
Where is the Processing Cost?

• It could be at the early visual level or the
  semantic integration level.
• Hyönä et al. (2002)
• suggests that the morphological effect derives
  from the more central, semantic-syntactic level.
• They observed that the morphological effect seen
  with isolated inflected words vanished when the
  participants read the very same items embedded in
  neutral sentence contexts.
• This study uses fMRI data to answer this question.

5
Previous Neuroimaging Studies

• Two specific brain areas were identified as being
  involved in morphological processing.
• The left OTC (occipitotemporal cortex),
  especially the fusiform gyrus, involved in
  decomposition at the visual input level
• The LIFG (left inferior frontal gyrus) and/or the
  left pSTG/pMTG (posterior superior/middle
  temporal gyri), involved in integration of
  meaning of the morphological constituents that
  calls for semantic-syntactic processing

6
OTC and Fusiform Gyrus
7
left inferior frontal gyrusLIFG
8
posterior superior temporal gyruspSTG
9
posterior middle temporal gyrus pMTG
10
So

• Increased activation in the OTC should indicate
  work on morphological decomposition at the visual
  input level.
• Increased activation in the LIFG and/or pSTG/pMTG
  should indicate work on integration of meaning.

11
Methods

• Participants
• 12 right-handed students
• Materials
• 85 monomorphemic words
• 85 case-inflected nouns
• 170 pseudowords (half w/ inflection-like endings)

12
Methods

• Procedure
• Lexical decision task
• ½ responded with left hand, ½ with right (why?)
• The stimuli were projected onto a screen that the
  participants saw through an angled mirror fixated
  on the head coil.
• Asterisk for 500 ms, 500ms blank, then stimulus
  work for up to 2000ms

13
Methods

• Procedure
• 34 30s task blocks
• 10 items per block (5 pseudowords, 5 words)
• 17 monomorphemic (MM) blocks
• 17 inflected (INFL) blocks

MM (30s)
INFL (30s)
INFL (30s)
MM (30s)
REST (20s)
REST (20s)
REST (20s)
REST (20s)
14
MRI Scanning

• fMRI is the use of MRI to measure the
  hemodynamic response related to neural activity
  in the brain
• http//www.howstuffworks.com/mri.htm
• http//www.fmrib.ofl.ac.uk/fmri_intro/brief.html

15
Results Behavioral Data

• As expected, the inflected words elicited
  significantly longer reaction times than the
  monomorphemic words
• (mean for monomorphemic, 763ms, SD 80 for
  inflected, 827, SD 88 t (11)D12.0, plt.001).
• The inflected items received also significantly
  higher error rates than the monomorphemic ones
• (mean for monomorphemic, 2.84, SD 3.93 for
  inflected, 5.36, SD 5.80 t (11)2.75, p.02).

16
Results fMRI Data
17
Discussion

• the contrast between inflected and monomorphemic
  items elicited clear activation increases in two
  of the three VOIs in the LIFG and in the left
  temporal region.
• Because the two stimulus groups were carefully
  matched and the only relevant difference between
  them was morphological complexity, it is
  plausible that the behavioral and activation
  differences between the two item types (inflected
  vs. monomorphemic words) were caused by
  morpheme-based processing of the inflected items.

18
Discussion

• Semantic unification is assumed to be related to
  the function of the LIFG
• LIFG activation in the present study may reflect
  the operations required for constructing a
  semantic-syntactic interpretation of the stem and
  affix combination on-line.
• Temporal areas have been assumed to play a
  crucial role in the memory component of language
• Activation in the temporal VOI was expected to
  reflect activation of the representations of the
  stem and the affix
• may also reflect the activation of the rules of
  the stemaffix combinability.
• The fact that recognizing an inflected word
  involves accessing two types of information may
  also bring about increased activation in temporal
  areas when compared to monomorphemic words.

19
Discussion

• results do not rule out the possibility that the
  left OTC is indeed involved in segmenting the
  orthographic patterns into morphemic units and
  relaying the information for temporal regions for
  further analysis
• This process may simply be very fast and
  automatic, and the main processing load with
  inflected words stems from the later stages.

20
Discussion

• The neural correlates of morphological
  decomposition reflect either form- or
  meaning-related lexical access processes that are
  part and parcel of the recognition of any given
  word (albeit more complex for inflected words).

21
Questions

• In an fMRI study activation patterns seem to be
  compared in between a baseline condition (do
  nothing?) and a task condition. Then how do we
  compare two (or more than) two different task
  type conditions directly?
• My only question was whether (and how much) it
  matters that the words were all low and medium
  surface frequency - that places considerable
  limit on the extension of the findings, I think.
   They didn't really address the significance of
  that, rather, they just noted that high surface
  frequency helps predict full-form/decompositional
  processing.

22
Questions

• I don't know Finnish at all. So, it is amazing
  that Finnish can have as many as 2000 different
  forms of a single noun. How does it work?
• Why does Finnish benefit finding out the neural
  correlaton of morphological decomposition?
• Also, I am wondering if the effect could be
  resulted from working memory load but not
  morphology itself.

23
Questions

• Compared with Finnish, which has lots of
  inflectional possibilities, Mandarin, on the
  other hand, is inflection impoverished.
  Characters such as "LE", "ZHE" and "ZAI" is used
  in Mandarin to mark inflection. However, unlike
  other Indo-European languages, these characters
  in Mandarin are separate from the previous stem
  or words. Thus coming back to the topic of this
  paper, which states that left frontal regions and
  left superior or middle temporal areas the ones
  most commonly observed in studies of
  morphological processing, I was wondering if the
  same areas will be activated too?

24
Questions

• Since this paper was aiming to look at the
  morphological decomposition, I'm thinking maybe
  using words with similar orthography but
  different morphological structures would speak
  more to this issue?
• Other questions?