Week 5: Psycholinguistics and Language Processing
Lecture 5.1: Cognitive Theories of Language Processing
Lecture 5.1: Cognitive Theories of Language Processing
Introduction
This lecture will delve into cognitive theories of language processing, which investigate how the human mind processes, comprehends, and produces language. Central to this exploration are the mechanisms by which we store, retrieve, and manipulate language, as well as how we process sentences during both comprehension and production. We will examine key cognitive theories related to word recognition, syntax processing, and language models, comparing the more traditional generative models with newer connectionist models. This will help us understand how language operates in the brain and the different ways in which linguistic information is processed.
By the end of this lecture, you will have a deep understanding of the mental lexicon, theories of word recognition, sentence processing, and the relationship between comprehension and production models of language.
I. The Mental Lexicon and Sentence Processing
1.1 The Mental Lexicon: Definition and Function
The mental lexicon refers to the mental dictionary that stores words and their meanings, forms, and syntactic properties.
It is believed that the mental lexicon is organized not only by alphabetical order but also by semantic relationships (e.g., synonyms, antonyms) and syntactic structures (e.g., verb-noun collocations).
Lexical Access: The process of retrieving words from the mental lexicon. This occurs when we recognize a word's meaning during reading, listening, or speech.
1.2 Word Recognition
Word Recognition Models:
Logogen Model: Proposed by Jerry Morton (1969), this model suggests that every word has a "logogen" in the brain, a unit that accumulates evidence for word recognition. When enough evidence is accumulated (from hearing or seeing the word), the logogen "fires," and the word is recognized.
Interactive Activation Model (IAM): A connectionist model proposed by McClelland and Rumelhart (1981), which explains word recognition as a process where different levels (e.g., visual features, letters, words) interact to activate the word. This model emphasizes parallel processing.
1.3 Sentence Processing
Sentence Processing is the way our brain structures and interprets a series of words to derive meaning from sentences. This involves recognizing words, determining syntactic structures, and computing semantic interpretations.
Two Key Aspects of Sentence Processing:
Parsing: The syntactic structure of a sentence is identified (e.g., noun phrases, verb phrases) to understand how words relate to one another.
Integration: The different elements of the sentence are integrated to construct meaning.
II. Theories of Word Recognition and Syntax Processing
2.1 Theories of Word Recognition
Dual Route Model: Proposes two different routes for recognizing words:
Direct Route: For familiar words, recognition happens directly from the visual form of the word to its meaning.
Indirect Route: For unfamiliar or non-words, we use phonological processing to convert the word into sound before linking it to its meaning.
Frequency Effects: Words that are more frequently encountered are processed faster than rare words. This is known as the frequency effect in word recognition.
Semantic Priming: A phenomenon in which the presentation of a word (e.g., doctor) facilitates the recognition of a semantically related word (e.g., nurse), demonstrating how semantic associations in the mental lexicon speed up word processing.
2.2 Syntax Processing
Syntactic Parsing: The brain uses syntactic rules to assign roles and relationships to words in a sentence (e.g., subject, verb, object). This involves constructing a mental parse tree.
Garden Path Theory: A sentence comprehension model proposed by Frazier (1979) which posits that sentences are parsed incrementally, often leading to temporary misinterpretations or "garden path" sentences, which force the reader to re-parse the sentence as new information is received.
Example: The old man the boats can initially confuse readers, as it seems syntactically incorrect. However, the correct interpretation is made when the reader realizes that "man" is a verb, not a noun.
Minimal Attachment: A strategy that says the brain prefers simpler syntactic structures when parsing a sentence. It avoids unnecessary complexity unless the sentence structure demands it.
III. Language Comprehension and Production Models
3.1 Models of Language Comprehension
Interactive Models of Comprehension: These models suggest that language comprehension is not a strictly step-by-step process. Instead, top-down and bottom-up processes occur simultaneously, interacting in a dynamic way.
Top-down processing refers to the influence of higher-level knowledge (e.g., context, world knowledge) on comprehension.
Bottom-up processing refers to the analysis of individual linguistic units (e.g., phonemes, words) starting from the sensory input.
Connectionist Model of Language Comprehension: Based on parallel processing and neural networks, this model suggests that language comprehension results from the simultaneous activation of various linguistic and contextual representations in the brain. Each unit (e.g., word, concept) is connected to others, and understanding emerges through the interaction of these connections.
Evidence for Interactive Models:
Contextual Influence: Studies in psycholinguistics have shown that our comprehension of language is heavily influenced by prior knowledge, context, and expectations. For example, the sentence The cat chased the... will likely lead to the expectation of a word like mouse based on context.
3.2 Models of Language Production
Modular Models of Production: These models argue that the processes of language production are divided into distinct, sequential stages: concept formation, lexical selection, syntactic planning, and phonetic realization.
Interactive Models of Production: These models propose that language production is more interactive and fluid. Bidirectional feedback occurs between different stages of production, meaning that later stages (e.g., phonetic realization) can influence earlier stages (e.g., lexical selection).
Example: The activation of a word’s phonetic structure might influence the choice of the word itself in the process of speech production.
Levelt’s Model of Speech Production: This influential model suggests that speech production follows a series of steps:
Conceptualization: Formulating the message.
Lexicalization: Selecting words.
Syntactic Structuring: Organizing words into syntactic structures.
Phonological Encoding: Assigning phonetic forms to the words.
Articulation: Producing the speech sounds.
Evidence for Interactive Models: The phenomenon of tip-of-the-tongue (TOT) states, where a person cannot immediately retrieve a word, suggests that feedback loops between the phonological and lexical levels are at work during production.
IV. Connectionist vs. Generative Models
4.1 Connectionist Models
Overview: Connectionist models are based on the idea that language processing is the result of parallel activation across networks of interconnected processing units (neurons), rather than the result of hierarchical rules.
Strengths:
Emphasizes distributed processing and learning from experience.
Models can account for complex patterns like language acquisition and error patterns in language learning (e.g., overgeneralization).
Example: Plaut et al. (1996) demonstrated how connectionist models can simulate children’s development of word reading by gradually learning mappings between orthography, phonology, and semantics.
4.2 Generative Models
Overview: Generative models (e.g., Chomsky’s theories of Universal Grammar) emphasize the role of innate structures and rules in language processing.
Strengths:
Focus on the syntactic structure of language.
Provide clear, formalized rules for how sentences are generated in the mind.
Criticism: Generative models often struggle to account for the gradual learning process observed in children’s acquisition of language, as well as the flexibility of language use in real-life contexts.
Conclusion
In this lecture, we explored cognitive theories of language processing, including the role of the mental lexicon, the theories of word recognition and syntax processing, and the models of language comprehension and production. We also contrasted connectionist and generative models, discussing their strengths and limitations in explaining how the human mind processes language.
In the next lecture, we will look into language acquisition, specifically how children acquire language and the various theories that attempt to explain this fascinating process.
Thank you for your attention, and I look forward to our next session!
Riaz Laghari (Riaz Hussain) is an independent scholar, linguist, and writer based in Islamabad, Pakistan. His primary research focus is comparative and theoretical syntax, particularly the structure and variation of English, Urdu, and Saraiki, followed by psycholinguistics, with an interest in language processing and cognition. His wider writing explores education, culture, and society.
“Learn to live, live to learn.” — Riaz Laghari
