Speech perception models

Question 1

Logogen model

Answer 1

Early model, proposed the logogen as a cognitive unit that mediated the access of lexical items based on input and context

Question 2

Cohort model

Answer 2

Marlsen-Wilson and Welsch

Three stages in which a cohort of possible word candidates is formed, which is gradually narrowed as more info is available:

1. Access:
   Initial activation of multiple word-candidates based on match w first part.
2. Selection
   One of the word-candidates is selected based on speech input and context against activated words.
3. Integration
   Meaning of selected word is activated and integrated to the discourse/message

highlights integration of top-down and bottom-up processess

Question 3

TRACE model

Answer 3

McClelland & Elman

Connectionist model, based on parallel distributed processing (PDP), operating trough interconnected nodes in different layers;

From feature, to phoneme, to word nodes activation spreads though network. Bidirectional, so not only bottom-up but also top down.

Competitive inhibition for nodes that are activated weaker, so stronger nodes are facillitated.

The model also adapts when new info is learned.

Question 4

Word superiority effect

Answer 4

People are better at recognizing letters when they are presented within words compared to when they’re in isolation.

Word context provides facilitation for letter recognition as it works bidirectonally.

Question 5

Context effect

Answer 5

If the context is sufficiently restrictive we hear phonemes even if they’re missing (eg covered by a cough).