Knowledge coded for language: written word identification

What are the steps of identifying a word?

1. Fixate on the word

2. Identify component letters

3. Encode the order of the letters

4. Compare the encoded string with mental dictionary (lexicon) of previously learned words and select the best matching word

5. Retrieve the meaning of the word

6. Compute pronunciation of the word

Why must we directly fixate on words?

Due to the anatomy of the eye - high acuity vision is limited to a small area on the retina (the fovea)

Written words require high acuity vision to process

How many letters are seen with 100% acuity around the fixation point?

Only 4-5

How many fixations needed per word?

Typically 1 for smaller words and 2 for longer words

What are letter representations like?

They are abstract letter codes that treat A/a as the same despite visual differences

Masked priming provides evidence for this → equal priming between kiss/KISS (visually similar) and read/READ (visually dissimilar)

What are the general processing levels for visual word identification?

1. low level visual processing - not specific to words → detecting and processing edges (simple, complex cells)

2. letter processing in orthographic system → detecting shape-specific letters, e.g. letters of a specific font (shape-specific letter cells); general letter detection (complex letter cells)

Do only external letters in words matter?

NO - still need central letters to distinguish words that only differ by internal transposition → e.g. form, from

What is the context-dependent view of letter coding?

Letters are coded by both letter-type and position in LTM

So, there are different letter detectors for A1, A2, A3, etc

e.g. - the A in SALT has a completely different letter code than the A in STAN (A2 vs A3) so would be encoded by a completely different unit

How would SALT and LAST be distinguished using a context-dependent model of letter coding?

They would be distinguished by the fact that they activate different letter detectors → S1 vs S3, L3 vs L1

Is context-dependent letter coding localist or connectionist?

Localist

What is the context-independent view of letter coding in LTM?

Letters are coded independently from position.

Letter order is instead coded by the relative level of activation

So, order of letters = rate of neuron fire

Letters coded left→right so first letter of word has greater activation than last letter

Same letter units (although there are multiple clones to account for multiple of same letter in one word) are being activated regardless of position and RATE OF FIRE dictates position in word

Under a context-independent view, how are SALT and LAST coded?

With the same set of letter detectors

BUT, letter detectors differ in level of activation/rate of fire as S and L are in different positions