Transcript The neural basis of unconscious orthographic priming

C35
The neural basis of unconscious orthographic priming
Joseph. T. Devlin, Sharon Geva, Hannah Devlin, Mark W. Woolrich
Centre for Functional Magnetic Resonance Imaging of the Brain
University of Oxford, Oxford, U. K.
Background
Predictions
Previously we have demonstrated that neural priming effects for
morphologically related word pairs (e.g. teacher-teach) overlap with the
adaptation observed for visually related words (e.g. ponder-pond) and
semantically related words (e.g. idea-notion) (1). We argued that these results
were consistent with the claim that morphology is an emergent property of
interactions between form and meaning rather than a separate level of linguistic
representation (2). Davis (3) noted, however, that the visual priming effect in
the left posterior occipito-temporal cortex could be due to apparent
morphological structure as the visually related word pairs looked as if they
ended in a valid morpheme (e.g. ponder ≠ pond + -er). In other words, if
morphemes are stored at a purely orthographic level (4), then the visual priming
effect may have been due to repeated access to visual morphemes (3) rather than
shared visual form more generally (1).
The two hypotheses generate different predictions regarding activation of the left
posterior fusiform region in visual masked priming:
Imaging results: Reading Words
Visual form hypothesis: If the left posterior fusiform computes abstract visual form
information more generally, then all three conditions (orthographically, pseudo-affixed,
and morphologically related) pairs will show a neural priming effect relative to unrelated
pairs as all shared equivalent orthographic overlap.
 To verify the efficacy of the visual masking, a behavioural pre-test was conducted
where participants were asked to read masked words aloud as best possible.
A. Reading aloud paradigm
The current results are inconsistent with the hypothesis that the left
occipito-temporal cortex is performing morphological decomposition at an
orthographic level (3). The fact that non-affixed (spinach), apparently
affixed (ponder) and affixed words (teacher), all show neural adaptation
when paired with their embedded stems, suggests that this region
processes abstract, visual form rather than apparent morphological
structure. Together with previous findings (7, 9), we argue against the
notion of either pre-lexical (10), lexical (8), or morphological (3) visual
word forms stored in a particular patch of cortex. Instead, this region
appears to integrate abstract, visual form with higher order properties of
the stimulus such as meaning or sound (but not limited to these), based on
functional (11, 12) and anatomical (13) connections to higher order
association cortices (14). This account leads away from cognitive-based
parcelations of cortex and towards an understanding of brain function in
terms of information processing grounded in known anatomical and
neurophysiological properties of the region.
pSTG
PTr
pOT
aSTG
Figure 5: Group results for unrelated words relative to consonant letter strings. Based on
a mixed-effect group analysis shown on a single subject’s T1 image.
 Activation was present in
 the triangular (PTr) and opercular (POp) regions of Broca’s area,
 ventral premotor cortex (PMv),
 areas of the anterior and posterior superior temporal gyrus (a/pSTG), and
 the posterior occipito-temporal region (pOT).
 Words were presented for either 33 or 200msec and forward and backward masked with
visual noise
 The aim of the current study was to evaluate these two hypotheses of left occipitotemporal involvement in orthographic processing
Discussion
PMv
Morphological structure hypothesis: If orthographic representations are organised
morphologically and stored in the left posterior fusiform gyrus, then one should observe
neural priming (i.e. reduced activation) for pseudo-affixed (ponder-pond) and
morphologically related word-pairs (teacher-teach) because both have apparent
morphological structure. Orthographically related pairs such as spinach-spin should not
show neural priming because they do not have apparent morphological structure.
Behavioural pre-testing
Current Study
Email: [email protected]
References
Word reading in individuals
B. Reading accuracy
1.
Devlin, J. T., Jamison, H. L., Matthews, P. M., & Gonnerman, L. M. (2004).
Morphology and the internal structure of words. Proc Natl Acad Sci U S A, 101(41),
14984-14988.
2.
Seidenberg, M. S., & Gonnerman, L. M. (2000). Explaining derivational morphology
as the convergence of codes. Trends Cogn Sci, 4(9), 353-361.
3.
Davis, M. H. (2004). Units of representation in visual word recognition. Proc Natl
Acad Sci U S A, 101(41), 14687-14688.
4.
Rastle, K., Davis, M. H., & New, B. (2004). The broth in my brother's brothel:
morpho-orthographic segmentation in visual word recognition. Psychon Bull Rev,
11(6), 1090-1098.
5.
Devlin, H., Devlin, J. T., Woolrich, M., Miller. K. and Jezzard, P. (2006) An efficient
method for obtaining subject-specific HRF estimates in event-related fMRI. Poster at
ISMRM 2006.
6.
Cohen, L., Lehericy, S., Chochon, F., Lemer, C., Rivard, S., & Dehaene, S. (2002).
Language-specific tuning of visual cortex? Functional properties of the visual word
form area. Brain, 125, 1054-1069.
7.
Devlin, J. T., Jamison, H. L., Gonnerman, L. M., & Matthews, P. M. (2006). The role
of the posterior fusiform in reading. Journal of Cognitive Neuroscience, 18(6).
8.
Kronbichler, M., Hutzler, F., Wimmer, H., Mair, A., Staffen, W., & Ladurner, G.
(2004). The visual word form area and the frequency with which words are
encountered: evidence from a parametric fMRI study. Neuroimage, 21(3), 946-953.
9.
Price, C. J., & Devlin, J. T. (2003). The myth of the visual word form area.
Neuroimage, 19(3), 473-481.
98%
#$%%#$
500ms
1500ms
33 or 200ms
500ms
1000ms
PASS
33ms
500ms
1000ms
passive
$%##&@#@
*
 The design included four test conditions, two non-word conditions and a fixation baseline
200ms
33ms
Duration
Figure 2: A. Schematic of the reading paradigm with visual noise patterns before and
after the word. For short (33msec) but not long (200msec) durations, the
masking blocked conscious awareness of the word. B. Results of the
behavioural pre-test.
 Performance was at ceiling in for words presented for 200msec. In contrast, words
presented for 33msec were only occasionally read aloud.
Table: Sample stimuli in the lexical decision task
Example
journal – HAZE
spinach – SPIN
planet – PLAN
poetry – POET
florze – HALDA
pchmmv – WLPBX
9%
Time
Figure 1: A schematic diagram of the visual masked priming paradigm. The prime was
forward masked by a visual noise pattern and backward masked by the target so that
participants were not consciously aware of the prime.
Condition
1. Unrelated
2. Orthographic
3. Pseudo-affixed
4. Morphological
5. Pseudowords
6. Consonant letter strings
Time
Semantic
Similarity Rating
1.3
1.4
1.5
7.6
–
–
 These results suggest that visually masked primes were not consciously perceived,
consistent with post-hoc self-report in all experiments.
10. Dehaene, S., Cohen, L., Sigman, M., & Vinckier, F. (2005). The neural code for
written words: a proposal. Trends Cogn Sci, 9(7), 335-341.
Words
Nonwords
99%
99%
 For each individual, the mean percent BOLD signal change relative to fixation was
computed per condition:
97%
Accuracy
(% correct)
97%
Neural priming effects
93%
Words
 Trials were presented in a pseudo-randomized, event-related design with “null events”
and a mean ISI of approximately 5s
Unrel
Data Analyses
 Median RTs for correct responses were calculated per condition per subject and used in
the behavioural analyses. Data from two subjects were excluded due to atypical
performance: one had RTs > 200msec longer than any other participant and one fell
asleep during the task
 The functional imaging data were realigned to correct for small head movements,
spatially smoothed with a 6 × 6 × 6mm FWHM Gaussian filter, and registered into
standard space with an affine transformation
 First level analyses used individually-tailored HRF estimates convolved with stimulus
onsets to maximize sensitivity to BOLD signal change (5)
 Briefly, run A was used to determine HRF shape and then used to analyse run B
 Run B was used to determine the HRF shape for the analysis of run A
 Produced an unbiased, efficient estimate of individual subjects’ HRF
 In each subject, a functional ROI was defined as a sphere (6mm radius) surrounding the
peak activation for reading [Unrelated > fixation] in the left occipito-temporal area (a
standard space sphere around [-42 -56 -18] ). The ROI was then used to evaluate priming
effects
Orth
PsAff
Morph
PsWord
Cons
Figure 3: Accuracy to words (left) and nonwords (right) in the lexical decision task.
There were no significant differences between lexical conditions in accuracy
(F3,39=2.0, p=0.13)
 There was a significant main effect of Relation (Unrelated, Orthographical, Pseudoaffixed, Morphological) between prime and target on reaction times (F3,39=9.1,
p<0.001).
Words
Reaction Times (msec)
 Sixteen native British English speakers (8F, 8M) participated in two runs with the order
counter-balanced across subjects (B0=3T, gradient echo EPI, TR = 3sec, TE = 30msec,
FOV = 192 x 256mm, matrix = 64 x 64)
11. Bokde, A. L., Tagamets, M. A., Friedman, R. B., & Horwitz, B. (2001). Functional
interactions of the inferior frontal cortex during the processing of words and word-like
stimuli. Neuron, 30(2), 609-617.
 Accuracy across conditions was 97.3%, indicating that subjects had no difficulty
performing the task.
98%
 Lexical stimuli were also matched for familiarity, frequency, concreteness, imageability
and syllable length, while all stimuli were matched for letter length
 Demonstrates that the lexical decision task engages the specific occipito-temporal
region observed in previous studies of the “visual word form area” (6-9)
Behavioural results: Lexical decision
 Unrelated, orthographically related and pseudo-affixed pairs were matched for rated
semantic similarity across conditions (1.0 = unrelated to 9.0 = identical meanings)
 All related prime-target pairs were phonologically transparent in British English and had
statistically equivalent orthographic overlap (mean = 4.4 letters, F2, 87<1)
Figure 6: In all cases, unrelated words relative to consonant strings activated left posterior
occipito-temporal region near the occipito-temporal sulcus (white arrows). A
dotted line marks the collateral sulcus between the fusiform and
parahippocampal gyri.
+25ms
Nonwords
Orth
Nonwords
*
*
Morph
0.56
0.41
0.37
PsAff
Morph
0.36
Unrel
Orth
0.36
PsWord
Cons
 Within the left occipito-temporal region activated by words, all three conditions which
shared orthographic overlap showed a significant neural priming effect:
 Orthographic
(spinach-SPIN)
t13=3.2, p=0.011
 Pseudo-affixed
(planet-PLAN)
t13=3.0, p=0.017
 Morphological
(poetry-POET)
t13=2.7, p=0.027
PsWord
Cons
Figure 4: RTs for words (left) and nonwords (right). The magnitude of the priming effect
relative to unrelated pairs is indicated above each condition in msec. The *
indicates a significant priming effect at p=0.05.
 Post-hoc comparisons revealed that the only significant priming effect was for
morphologically related words (16 msec priming, t13=2.7, p=0.05 after Bonferroni
correction for multiple comparisons, two-tailed).
13. Distler, C., Boussaoud, D., Desimone, R., & Ungerleider, L. (1993). Cortical
connections of inferior temporal area TEO in macaque monkeys. Journal of
Comparative Neurology, 334(1), 125-150.
14. Price, C. J., & Friston, K. J. (2005). Functional ontologies for cognition: The
systematic definition of structure and function. Cognitive Neuropsychology, 22(3-4),
262-275.
Email: [email protected]
0.63
+21ms
PsAff
12. Mechelli, A., Crinion, J. T., Long, S., Friston, K. J., Lambon Ralph, M. A., Patterson,
K., et al. (2005). Dissociating reading processes on the basis of neuronal interactions. J
Cogn Neurosci, 17(11), 1753-1765.
*
Figure 7: Neural repetition priming effects. There was a significant difference between
lexical conditions in BOLD signal change (F3,39=3.9, p=0.015)
*
–16ms
Unrel
BOLD signal change (%)
200ms
reptile
$%##$%
*
Accuracy
(% correct)
 A visual masked priming paradigm with a lexical decision task was used to engage the
region and fMRI was used to measure neural adaptation
 The effects were case-independent, consistent with previous studies (6, 7).
 In other words, words which shared visual form, independent of apparent morphological
status, led to a significant reduction in BOLD signal relative to unrelated words.