Learning Outcomes from last year:
- Understand and describe the challenges of comprehension using appropriate terminology
- Explain how words are recognised in the mental lexicon
- Explain how we comprehend sentences
- Understand and describe models of word comprehension
- Explain how the evidence supports current assumptions
Learning Outcomes from this year:
- Understand and describe the process of reading
- Explain how written words are recognised in the mental lexicon
- Describe the Dual Route Cascaded (DRC) Model of reading (Coltheart et al., 2001)
- Describe the self-teaching hypothesis of learning to read
- Evaluate the evidence for the self-teaching hypothesis
The mental lexicon
Our mental lexicon contains different representation of words.
Today we are focusing on the orthographic representation – the letters and writing systems that make up a language, to a certain extent the phonological representation may also become involved when reading (sounds).
What is reading?
When we are reading we are trying to get insight into what they have written, which is a representation of what is in their minds – when reading we are forming our own representations of what they are trying to say.
We do this using writing systems, orthography is the writing system of a language, graphemes are it’s constitutional units (visual representations of a phoneme) – there are about 44 phonemes/ graphemes in English.
Writing systems
Different types of writing systems exist…
• Logographic systems (based on whole characters/ concepts/ structures)
- Japanese 愛 ai (love)
- Chinese 香港 heung gong (Hong Kong; fragrant harbour)
• Alphabetic Systems (A single letter represents a distinctive sound in spoken language)
- Korean 학교 hakkyo (‘school’) ㅎ ㅏ ㄱ ㄱ ㅛ
Grapheme correspondence with phonemes
So, remember a grapheme is a unit in a writing system, and a phoneme is a unit of sound. A grapheme is a representation of a phoneme, a phoneme can be represented by one or more than one letter.
A grapheme can be made up of a number of letters e.g.,
- that /ðæt/
- Night /naɪt/
- Through /θru/
The words are transcribed on the write as how we write the sounds of the words (phonetic transcript, IPA). In through, the 4 letter phoneme ough is represented by a single grapheme – lots of structures are possible.
1 grapheme can represent more than 1 phoneme
e. g., the ‘i’ in Mint /mɪnt/ Pint /paɪnt/
e. g., ‘th’ /ð/ (in that) or /θ/ (in through)
1 phoneme can be represented by more than 1 grapheme
e.g., /k/ can be represented by ‘c’ ‘k’ ‘ck’
Types of Writing Systems –
Transparent
The spelling of each word maps directly on to its pronunciation (e.g., Finnish or Italian)
Opaque
The spelling of each word does not map directly on to its pronunciation (e.g., English)
English has some transparency
DOG, PRINT, COBWEB = regular words
But a lot of opaque spellings
YACHT, KNIGHT, COLONEL = irregular words
Regular and irregular words ‘mint’ and ‘pint’
Mint is the regular word here
Models of Visual Word Recognition
Two competing models we are only going to look at one;
DRC: Dual Route Cascaded model of visual word recognition and reading aloud (Coltheart, Rastle, Perry, Langdon & Ziegler, 2001)
DRC: Dual-Route Cascaded model of visual word recognition and reading aloud
Overview of the model structure and how it works
SEE IMAGE IN NOTES
The two routes are the lexical route and non-lexical route for reading (remember lexical means words, whole words in this context – so non-lexical means not whole words)
So, when we are trying to process the printed word we can activate a whole word orthographic representation of the word in the lexical route by activating a print representation that we have stored in our lexicon. Alternatively, we can activate a non-lexical route which will do a process called grapheme-phoneme correspondence (where it will convert the graphemes into phonemes (letters into sounds), so we can access a whole word phonological representation (whole word sound) that we have stored in our phonological lexicon). By doing that we can access the semantic meaning of the word via activation of phonological or orthographic representations of the word.
Lexical route –> Orthographic representation – representation of the whole word –> phonological lexicon –> semantics
Non-lexical route –> Grapheme-phoneme correspondence –> phonological lexicon –> semantics
Lexical Route
Lexical route –> Orthographic representation – representation of the whole word –> phonological lexicon –> semantics
Non-Lexical Route
Non-lexical route –> Grapheme-phoneme correspondence –> phonological lexicon –> semantics
IRREGULAR WORDS CAN’T USE THE NON-LEXICAL ROUTE
There are certain words that we can access the semantics of via either route e.g. text/ mint because they are regular words, the word “pint” however would only be accessible via the lexical route as it is irregular – using the non-lexical route and breaking it down into sounds would convert the graphemes into phonemes that wouldn’t access the correct phonemes (would end up pronouncing pint like “mint”. So to access the correct phonological representation of pint the lexical route must be used.
DRC: Dual-Route Cascaded model of visual word recognition and reading aloud
Theory behind it
- DRC is a cascaded model (like Dell’s model of speech production) and not a discrete model (like Levelt et al.) – generally info passes down through the model from one level to the next, but you don’t have to finish processing at one level to begin processing at the next
- Imaging evidence of different routes for reading supports the model (e.g., Seghier, Lee, Schofield, Ellis & Price, 2008)
- Provides a clear account of disordered reading e.g., different types of dyslexia (see Eysenck & Keane for a review)
Lexical Access for written words (aka visual word recognition)
The system accesses lexical items based on
The system accesses lexical items based on
– Visual input (e.g. graphemes and their correspondence with phonemes)
– Lexical characteristics (e.g., frequency, familiarity)
– Context
– Prediction
– Less impacted by temporal information than speech (e.g., able to ‘read’ a whole word, do not have to wait for it to ‘unfold’)
Lexical Access for written words (aka visual word recognition)
There are some specific lexical characteristics that can impact the speed or ease you can recognise written words…
(List)
- Familiarity
- Frequency
- Length
- Neighbourhood Density
- (Context)
Lexical Access for written words (aka visual word recognition)
Lexical characteristics
Familiarity
Familiarity:
Eichelman presented participants with word pairs which we either words or non-words. Ppts made same/different decisions faster when the pairs were words compared to when they were nonwords. Words (e,g., SEAT) are read faster than pseudohomophones (e.g., SEET)
Lexical Access for written words (aka visual word recognition)
Lexical characteristics
Frequency
Frequency:
Words more frequently accessed are read + processed more quickly
Naming speed was slower with increasing length for pseudowords and low frequency but not for high frequency words
Lexical Access for written words (aka visual word recognition)
Lexical characteristics
Length
Length:
Larger impact of length on low compared to high frequency words
The less familiar the word, more decoding of graphemes, slows recognition of word
Lexical Access for written words (aka visual word recognition)
Lexical characteristics
Neighbourhood density
Lexical access will be slower for words with lots of neighbours
e.g.
“REED”
Orthographic neighbours;
reek, rend, peed, heed, deed, reel, reef, weed, seed, feed, read, need
Phonological neighbours;
rowed, cede, wreath, ream, wreak, knead, keyed, heed, mead, bead, rode, raid, she’d, rude, we’d, he’d, ride, reach, lead, road
Neighbourhood density effect in speech recognition
Lexical access will be slower for words with lots of neighbours e.g. yacht will be faster than peach (neighbours like pea, piece, peel, peace). In spoken word recognition, lots of neighbours is inhibitory.
Neighbourhood density effect in visual word recognition
Where when we are listening to someone speak words with lots of neighbours might be slower to be recognised we don’t tend to get that with written words. Lexical access faster for low frequency words with lots of neighbours but no effect for high frequency words with lots of neighbours. In written word recognition, lots of neighbours is facilitative (if the word is low frequency, no difference for high frequency due to Hebbian learning - neuronal activation is high, threshold is low, and neurones are practiced).
Context and Visual Word Recognition
Context:
Ambiguous words can have dominant and subordinate meanings attached to them. For example an ambiguous word such as bark could have a dominant meaning of dog attached to it and a subordinate meaning of tree. So in an exp we would present the word bark, then either dog or tree and test the reaction – respond more quickly to dominant meaning.
Learning to Read
What does the DRC predict about how children learn to read?
- DRC Predicts that readers start by decoding words using the non-lexical route (breaking words down)
- Gradually build up an orthographic lexicon
- Enabling direct access of meaning via the lexical route without accessing phonology
SO the DRC model predicts that the link between the orthographic lexicon and the phonological lexicon becomes weakened over time as readers become more skilled
Learning to Read
Self-teaching hypothesis (Share, 1995)
What he said
Share said children use Grapheme-Phoneme correspondences to teach themselves to read, and orthographic (whole word) representations are added after 1-6 exposures of the same word
