Reading and the brain Flashcards
the ‘visual word form’ area (Cohen et al., 2000)
Ventral occipito-temporal cortex.
‘Brain letter box’, or gate to the reading system.
Activated specifically by letter strings acceptable in the language (e.g. NGTH but not TGNH, in English). - combination of letters useful in language
The proximity of this area to other regions coding faces and objects supports the idea that the brain has evolved to perceive letters and words as complex objects.
see notes
Stage 1: Processing restricted to the contralateral hemisphere (Visual area 4, occipital pole). - code mere visual properties
Stage 2: Transfer of information from both occipital poles to a more ventral region of the left occipital lobe (Visual word form area).
Activation moved forward – more ventrally – entering visual word form area – this area is left localised
Because the visual word form area is located in the left hemisphere, information coded in the primary visual areas in the right hemisphere (i.e., the left visual hemi-field) has to cross to the left hemisphere.
This transfer from right to left is done via the corpus callosum, which connects the two hemispheres.
In contrast, the information already coded in the left hemisphere (i.e.., the right visual hemi-field) does not have to cross. It is sent within the same hemisphere instead.
Asymmetry
In case of a lesion at an early age, the visual word form area can swap to the right hemisphere.
why is the right visual field coded by the left hem and vice versa?
see notes
The left retina of each eye captures the right visual field, whereas the right retina of each eye captures the left visual field.
But, and this is the key element, the inner retinas send information across, to the opposite hemispheres, whereas the outer retinas send information to the hemisphere on the same side (for this, follow the optical nerves coded in orange vs. green).
As a result, what is presented in the right hemi-field ends up in the left hemisphere, whereas what is presented in the left hemi-field ends up in the right hemisphere.
(This is not that right eye projects to the left, and vice versa for the left eye.)
This relates to the upper graph of my demonstration on the previous slide, before the information in the right occipital pole crosses via the corpus callosum to reach the brain letter box (or visual form area), normally developing in the left hemisphere.
Info captured by outer hemispheres stays in the same hemisphere and vice versa
spatiotemporal dynamics of word processing in the human cortex (Marinković et al., 2003)
Looked at the timecourse of visual word processing, using magneto-encephalography (MEG).
Estimated the peaks of cortical activity and its progression over time.
During reading, activation starts in both occipital poles.
At about 170ms it shifts to the left occipito-temporal region.
At about 230ms activity explodes in regions of both temporal lobes.
From 300ms onwards it extends over prefrontal and other temporal regions especially in the left hemisphere, before falling back in part to the posterior visual areas (occipital pole).
Catani et al. (2003)
Fibres depicted in red convey information from port-to-port, in an omnibus fashion; those depicted in green work like motorways.
In the reading system, obviously, these two types of fibers are especially important in the left hemisphere (unlike on the picture), to transfer information from the ventro-occipital regions (VWFA) to both posterior areas of the frontal lobe and temporal regions (see previous slide).
can cog models explain brain activation during word and pseudo word reading? a meta-analysis of 36 neuroimaging studies - Taylor et al. (2012)
Cognitive models make predictions about the functional overlap between brain regions involved in reading.
Thus, brain regions sensitive to these contrasts should correspond to the model’s components (e.g., input lexicon, grapheme-to-phoneme transcoding, etc. ).
Multiple options for reading the same word out loud
Conversions of letters –> sounds good for unknown words
see notes
They examined whether 36 neuroimaging studies of reading pointed at the same brain regions regarding two main contrasting dimensions in DRC:
- lexical status (i.e., words vs. nonwords; Is the stimulus part of your long-term memory?)
- regularity (i.e., regular/irregular; Can the word be read correctly by both routes?)
They distinguished between engagement vs. effort in how these dimensions translate into BOLD (Blood Oxygen Level Dependent) signal (i.e., amount of oxygen needed by a region). – no linear r’ship – inverted U shape r’ship
Given that cognitive models were partly derived from observing patients with reading disorders, you will see that lexical status and regularity provide also two contrasts that distinguish between various types of acquired dyslexia.
What time of task was the person doing?
Note: peaks in BOLD signal (brain activity) do not mean that a region “likes” the stimulus, but instead, that it is having a hard time processing it.
Engagement refers to whether or not the brain region in question is able to deal with the stimulus. In other words, it refers to the capacity of the stimulus to evoke knowledge in that region.
Effort refers to the amount of resources/fuel required to code and process the stimulus in that region, once the region is engaged.
Is given region good for given stimulus?
One region engaged, how effortful will processing be? – hard to process = need more oxygen – what fMRI meausres
Distinguish between those easy and painful to process
see notes
But first, they tested whether the distinction between engagement and effort made sense in DRC
Check their relevance by looking at the activity in the input lexicon (in the computerized version).
These results were obtained by giving the computer-implemented version of the dual-route model a list of words and nonwords, which the model had to recognize or reject.
Words in the input lexicon earn points in proportion to how well they match the stimulus (i.e., in terms of letters in the correct positions), and the more they earn points and also the more they are frequent in the language, the more able they are to deplete their competitors.
The word is recognised once its activation level is higher than that of its competitors by a minimum distance.
the graphs shows how much activity there is in the input lexicon.
As can be seen, firstly, existing words generate a strong engagement from that part of the model, in comparison to unknown words (i.e., pseudowords).
Secondly, in the word case, it takes longer for low frequency strings to reach the same activity level compared to high frequency words (Fig. A).
Thirdly, activity remains for longer in that region of the system for low- (as opposed to high-) frequency words, which altogether reflects that recognition is not as easy for low-frequency words (Fig. B).
see notes
the subtraction logic
Left with pool of photos taken while brain doing task– measure engagement of various regions
Where is vocab knowledge?
If want to locate given region, have to take photo of activation for words and subtract images for pseudowords
see notes
words > pseudowords
see notes
In order to isolate the regions specifically involved in recognizing existing words, one should take the brain activity elicited by word stimuli and subtract from it the activity elicited by unknown words.
This removes the activity common to processing the two kinds of stimuli from the brain map generated by existing words.
Hence, this identifies the neural activity specific to lexical processing (i.e., the lexical route).
Depends on what the target is
pseudowords > words
see notes
Conversely, in order to isolate the regions specifically involved in dealing with pseudowords, one does the opposite: subtract from the neural activity generated by pseudowords that generated by existing words.
That identifies the brain regions particularly engaged (i.e., effortfully) in reading unknown (though linguistically plausible) letters strings: the print-to-sound conversion route!
red route/lexical pathway - see notes
The lexical pathway (in red and yellow) engages words, but not or only weakly pseudoword strings.
It is also taxed more by low-frequency words compared to high-frequency words. We have therefore here two contrasts that point towards the lexical route: (a) words minus pseudowords, (b) low-frequency words minus high frequency words.
How will a given pathway react to the variables of interest
Lexical – access LTM for meaning
Relevant region should let in words and not nonwords
Subtract word images from pseudowords/nonwords
Split inverted U and consider which variables could apply to one side or the other
Semantic system common to many modalities
yellow route/irregular words - see notes
Irregular words (pint) should engage the lexicon more than non-orthographic strings (&$^£%).
The latter are not represented in lexical memory and obviously are not going to activate any word neighbour.
In addition, the mental dictionary (in yellow) should be taxed more by irregular words than by regular words, especially for words that occur less often in the language (these have faint lexical presentations).
As lexical memory is the only way to read exception words, low frequency irregular words will take time to be evoked compared to high-frequency regular words.
In sum, here we have two contrasts that point towards the lexicon: (a) irregular words minus non-orthographic strings (b) (low-frequency) irregular minus (high-frequency) regular words.
Note: In DRC, the regularity contrast points specifically to the lexicon and excludes the semantic system.
This is because the model does not assume more engagement of semantics for irregular words compared to regular words.
Irregular words – have to consult exception in mental dictionary
Exceptions must be coded and so there must be a region in the brain that will code this region
Region engaged but words might be low in freq and so region will need more oxygen
Engagement extreme – use nonsense symbols – wont activate region
Subtract low freq and irregular from nonsense symbol words
Subtract symbols from low freq words
blue route/non-lexial route - see notes and summary
In contrast, the non-lexical, print-to-sound conversion route should engage both words and pseudowords, but not non-orthographic strings.
The latter are not linguistically relevant.
The non-lexical route should also be taxed more by pseudowords than by words.
This is because even though pseudowords engages the non-lexical route, the latter is less accustomed to pseudowords compared to regular words.
And the same applies to the phoneme output (i.e., inner speech) system, at the next level.
Likewise, both the non-lexical route and the phoneme output system should be struggling with irregular words.
This is because for these, DRC produces two conflicting responses.
With this conflict arising, it is likely that activity should linger particularly at the level of the phoneme output system, with some reverberation upwards into the print-to-sound conversion region.
Blue – rely on generative knowledge
Phoneme buffer – prepare articulation – common to both pathways
non-lexical pathway
Interested in sounds, strings and non-strings
How deal with pseudowords
Focus on pseudowords and subtract region common to reading normal words
Subtract control from target
Guess words by using knowledge of English
Subtract regular words from pseudowords – effort pathway – regular are easiest words to read – conversion route reads them correctly as well – activation more spread and even – every given route should be less engaged
Engagement pathway
Pseudowords only read by non-lexical route
Irregular words need more fuel – conflict in model – correct pronunciation in lexical route but NL tries to also find pronunciation – conflict at response level so activation lasts longer to resolve conflict
lexico-semantic pathway
engagement: words minus pseudowords – crude contrast
effort: low-frequency minus high-frequency words
input lexicon
Enagagement: irregular words minus non-orthographic strings
Effort: (low-frequency) irregular words minus (high-frequency) regular words – imbedded into crude contrast
print-to-sound conversion/phoneme output buffer
pseudowords and words - non-orthographic strings - engagement
pseudowords - regular words - effort
irregular words - regular words - effort
Taylor et al. (2012)
36 fMRI or PET studies included in the meta-analysis.
Same task on both words and pseudowords, or regular and irregular words.
Tasks
- reading aloud or silently
- lexical decision (Is BRAIN a word?) - Recognition but also partially familiarity
- visual feature detection (Is there a | in BRAIN?)
- phonological lexical decision (Does BRANE sound like a word?) – pseudo-homophone task
- stimulus repetition detection (e.g., BRAIN BRAIN)
- rhyme judgement (do BRAIN and TRAIN sound the same?)
Set of regions that deal with vocab knowledge, or semantic system or output buffer
Set of converging evidence on v. distinct regions/specific regions light up for particular contrast – consistency in regions
see notes
