PY2505.7 ~ {Literacy}
the problem
any pattern can only convey meaning after we have learnt to associate visual patterns with meanings to gain access to mental lexicon
mental lexicon
part of brain that store's a person's knowledge of words + meanings [grows as new words are learnt]
process of lexical access
the cognitive process of retrieving a word's meaning + associated info from the mental lexicon, essentially "looking up" a word in your brain to access its semantic + phonological properties, allowing you to understand or produce a word during language comprehension or production
how many words do adults have in their mental lexicon
estimate of 20k
lexical access
accessing our knowledge about a word [reteieving words from the mental lexicom + finding word forms to express a meaning
stroop task
naming a colour in which a word is written is more difficult if the colour name + word conflict - difficult because it creates a conflict between our automatic tendenc to read words + intentional task of indentifying colour of ink
main methods for examining lexical access
RT - naming/pronunciation time + lexical decision + tachistoscopic recognition
lexical decision
how long does it take you to say a string of letters is[n't] a word
tachistoscopic recognition
how accurately can you identify a very briefly presented word
what makes lexical decision easier
semantic priming [prior exposure to a related word in meaning]
serial search model
(Foster, 1976,1979) - a model of leical access that assumes people search through lexical entries one by one [to find target]
how does ssm explain data
we know that if you repeat a word, you've already looked for it so it is moved to the top of bind which speeds up processing
fundamental problem with ssm
assumes attention is focused on one thing ast a time [unrealistic in real world scenarios] - so it oversimplifies representation of visual processing
logogen model
(Morton, 1979) - a theory that explains how people recognize words when they read or hear them
how does logogen model work
model assumes that each word has a specialized recognition unit called a logogen + when a word is encountered, the relevant logogen accumulates activation + when the activation level reaches a threshold, the logogen "fires" and the word is recognized + model takes into account contextual information, which helps people recognize words that may differ from the exact acoustic input.
how does logoden model explain data
freq effects - more freq words have a higher 'resting activation' so reach 'recognition threshold' more quickly
[semantic priming also explained where activated logogen spreads to related words]
logogen model + semantic priming
activated logogens don't return to their resting level immediately so require less perceptual input to be activated [so it's less far to reach recognition threshold for other words]
connectionist models
use multi-layer architechture [including hidden layers] + activation & inhibition [you can activate some words but also inhibit others if they're deffo gonna be relevant]
early connectionist model
McClelland & Rumelhart (1981) + Rumelhart & McClelland (1982) to account for the superioty effect - so you have multiple players: input layer + middle layer + output level [once unit is activated sends activation in parallel to all other ocnnected units
top down activation
cog process that uses higher-level brain functions to interpret new info
IAC model pros
explains superiority effect + heavily interactive + good at explaining context effects
IAC model cons
limited in theoretical value without postulating something like sequential stages of processing in which some categorization occurs before response selection
reading challenge
writing + reading is unlikely to have evolved brain systems dedicated to it cause its a human invention
preliminary model
most english words have a regialr spelling to sound correspondence wealso have irregular words [pint/vase/steak] so children are taught to read in a way that supports this complexity
according to preliminary model how are children meant to read
phonics [sounding out] for regular words + by sight for freq irregular words [for adults we read by sight]
dual route model
theres a fast route [direct lexical route + irr words can only be read this way] so from print to lexicon [meanings] to pronunciation
slow route [indirect] for unfamilair + new words so from print to grapheme-phoneme conersion rules then pronunciation
multiple route model
Grainger et al (2012) - all children start with a serial letter identification phase [phonological recoding] where you sound out the word then get to semantics and blend those tg and as they become more familiar with orthography you then start recognising chunks of words
connectionist models
inspired by neural networks - aim to model the way in which the brain learns to process info - so cm is large numbers of units [artificial neurons] with connections [weighted as excitatory/inhibitory]
what are connectionist models aiming to do
training network to correct errors using process called back propagation - you present an input to the networks [letter a] then allow it to produce an output [you can give model feedback + adjust weights to reduce likelihood of repetition of error]
Seindenberg & McClelland (1989) model
trying to show that you could read both regular + irr words in English using a single mechanism - its a probabilistic model - task of model is to learn to put an othological input with phonological output
does seindenberg model work
trained network to read 2897 words, after training, model tested by presenting letter strings + computing the error - measured accuracy using discrepancy between output vs what its supposed to be - results showed model performed better for regular words
Turkeltaub et al. (2003)
recorded the brain activity of adults and children whilst they viewed words, they weren't asked to read aloud instead they had to judge whether a word had a 'tall letter' [sauce=no alarm=yes] and they gave them real letters vs false font
Turkeltaub et al. (2003) - results
children primarily engaged areas used for phonological processing whereas adults showed additional extensive engageent of areas involved with semantic processing
children vs adults temporal
c [9y]- Lposterior temporal cortex - activity associated with maturation of phonological abilities
a - L parietal temporal + frontal cortices
correlations between activity + reading ability
increased involvement of left temporal + frontal regions [phonology + semantics]
decreased right posterior activation [occipital lobe]- possibly as text becomes more familiar so less activation in systems for recognizing non-lexical forms