Week 7 - Brain and Cognition - Language I

Orthography

= the written structure of words, including their spelling and arrangement of letters

eg, you see CAT and recognise the familiar spelling

Phonology

= speech sounds. in reading, it allows us to connect written letters to their corresponding sounds

eg, MIP - a made up word, but you can still pronounce it

Semantics

= meaning of a word. Allows you to understand what a written word represents

eg, DOG - you understand what a written word represents

Grapheme

= a letter or combination of letters representing a speech sound

eg, SHip

Phoneme

= the smallest distinctive unit of sound in a language

eg, shIP

Orthographic Transparency

= describes how consistently written letters correspond to sounds

eg, break and freak - same spelling, different sounds

• matters because the consistency of spelling-to-sound relationships can affect reading strategies

Lexical Decision Task + Naming Task

• lexical - decide whether a letter string is a real word

• naming - reading a word aloud

Saccades

= rapid movements of the eyes from one location to another

• jumps between fixations

• approx 20-30 milliseconds

• not a lot of useful reading info is extracted during them

Fixations

= the periods when the eyes remain relatively still

• this is when readers extract useful visual info

• usually at the beginning of the word

Perceptual Span

= refers to the area around fixation from which a reader can obtain useful info

• 3-4 letters to the left, up to 15 to the right

• the span is asymmetrical because information ahead in the direction of reading is useful

Eye-tracking research - De Leon Rodriguez et al. (2016)

Aim = to investigate where readers first fixate when reading words, including whether language experience affects this

Procedure = researchers tracked the eye movements of German-French bilinguals and French-german bilinguals while reading real words and pseudowords in their first and second language

Findings = readers generally fixated closer to the centre of a word rather than directly on its first letter. Reading strategies also differed depending on the language

Conclusion = reading does not operate through a simple letter-by-letter sequence. Readers position their eyes to gather info efficiently

Rayner et al. (2006) - Jumbled Letters

Aim = investigate how changing letter positions affects reading

Procedure = ppts read normal sentences and sentences containing different types of letters transportations while researchers recorded their eye movements

Findings:

• normal condition = 10.4 fixations, 236 ms mean fixation duration

• internal letter swaps = 11.4, 244 ms

• end-letter swaps = 12.6, 246 ms

• beginning-letter swaps = 13.0, 259 ms

Conclusion = jumbled words require additional processing. Swapping letters at the beginning was the most disruptive in this experiment

• letter position still matters, even though we do not read every letter separately

Automatic Processing

= some mental processes occur rapidly and without deliberate conscious control

• connected with the Stroop effect - when a colour word is printed in an incongruent ink colour, its meaning can interfere with naming the ink colour

Semantic Priming

= occurs when seeing one word makes it easier to recognise another related word

• eg, nurse > doctor - faster recognition as related words. Library > doctor - slower recognition as unrelated

The Word Superiority Effect (WSE)

= the finding that a letter is easier to recognise when presented as part of a meaningful word than in isolation or among unrelated letters

• eg, recognising the letter K may be easier in WORK than in an unrelated letter string

= this happens because the surrounding letters provide info about what the target letter is likely to be

• supports the idea that higher-level word info can influence lower-level letter recognition

Bottom-up Processing

= info begins with the sensory input, such as the lines and shapes that form letters and processes towards word recognition

visual features > letters > words > meaning

Top-down Processing

= existing knowledge and expectations help interpret the incoming info

• eg, reading “she spread butter on her brea_” - you are likely to predict bread using the sentences meaning

Rumelhart (1985)

= describes a strictly linear, bottom-up model

visual input > character > phonetic representation > lexicon > semantics

• main limitation - cannot adequately explain how expectations and word context affect what letters we perceive

McClelland & Rumelhart’s (1981) Interactive Activation Model

= explains how bottom-up and top-down processes interact

Basic Idea:

1. Feature level = recognises lines and shapes

2. Letter level = recognises individual letters

3. Word level = recognises whole words

• these levels interact rather than working in a strictly one-way sequence

How does the Interactive Activation Model work?

• eg, seeing the word CAT

• step 1 = feature detection = visual system detects curves. straight lines + other features that could form letters

• step 2 = letter activation = these features activate possible letter representations, such as C, A, T

• step 3 = word activation = recognised letters activate possible words, including CAT

• step 4 = feedback = word-level representation sends activation back to the letter level, helping confirm which letters are present

Two connections the Interactive Activation Model uses

• excitatory connections = increase activation of compatible interpretations

• inhibitory connections = reduce activation of incompatible interpretations

Strengths of Interactive Activation Model

• explains the WSE

• combines bottom-up and top-down processing

• makes testable predictions

Limitations of Interactive Activation Model

• does not include phonological processing

• does not explicitly model meaning and wider context

• original model focuses on four-letter words

Orthographic Neighbours

= words that differ by one letter in the same position

• eg, stem, step, stew

• these words share most of their letters, so their representations can become activated together

• more neighbours = slower recognition

Weak Phonological Model

= Coltheart et al. (2001)

• phonology is not necessary to identify a word

• meaning can be assessed without first retrieving pronunciation

Strong Phonological Model

= Frost (1998)

• phonology plays a central role in recognition

• phonological info is strongly involved in assessing words

Evidence for Phonological Processing - Van Orden (1987) - Homophones

Aim - investigate whether phonological info affects word recognition

Procedure - ppts answered questions such as whether a displayed word belonged to a particular category

• eg, Is it a flower? ROWS

Finding - ppts sometimes made errors because ROWS sounds like ROSE

Conclusion - the pronunciation of written words can influence meaning-based decisions

Evidence for Phonological Processing - Rastle & Brysbaert (2006) - Masked Phonological Priming

Aim - investigate whether phonological info is activated automatically

Procedure - brief, masked nonword primes preceded target words

Finding - recognition was faster following phonologically identical primes than following orthographically similar primes with different sounds

Conclusion - phonological processing can occur rapidly without conscious awareness of the prime

Evidence for Phonological Processing - Daneman et al. (1995) - eye movements

• ppts read sentences containing correct words or incorrect homophones

• eg, he was in his stocking feat/feet

Finding - incorrect homophones produced longer fixations and more backward eye movements

Conclusion - phonological info is activated during reading, but the pattern suggested that meaning may have been accessed first in this task

Evidence for Phonological Processing - Patient PS - Hanley & McDonnell (1997)

• patient PS could understand and define printed words, but struggled to pronounce them accurately

Conclusion - word meaning can sometimes be assessed without successful phonological access

The Dual-Route Cascaded Model (DRC)

= attempts to explain how we read words aloud

• key idea is that different pathways can be used to generate pronunciation

Three Routes:

• Route 1 = non-lexical route

• Route 2 = lexical-semantic route

• Route 3 = lexicon-only route

Route 1 - Non-lexical route

= converts graphemes into phonemes using spelling-to-sound rules

• particularly useful for unfamiliar words and pseudowords

• eg, never seen MIP, but can work out how to pronounce it

Route 2 - Lexical-semantic route

= uses stored knowledge of familiar words and their meanings

• eg, recognise PINT and retrieve familiar pronunciation rather than relying on regular spelling rules

Route 3 - Lexican-only route

= uses stored knowledge of a word’s pronunciation without passing through its meaning

Strengths of Dual-route Cascaded Model (DRC)

= model explains different reading difficulties, including surface and phonological dyslexia, and can account for performance in reading-aloud tasks

Limitations of DRC

= lack of explicit learning mechanism and difficulties accounting for some pronunciation patterns and other languages

Three Reading Impairments

• surface dyslexia

• phonological dyslexia

• deep dyslexia

Surface Dyslexia

= difficulty reading irregular words

• eg, mispronouncing PINT using regular rules

McCarthy & Warrington (1984) - Patient KT

• 100% accuracy for nonwords

• 81% regular words

• 41% irregular words