Language production - summary of chapter 12 of Cognitive Psychology by Gilhooly, K & Lyddy, F, M

Cognitive Psychology
Chapter 12
Language production

Syntax: to rules governing the ways words can be combined to create meaningful sentences.
Content words: words that provide meaning to the sentence; these contrast with function words which do the grammatical work of the sentence.
Language production: a number of processes by which we convert a thought into language output, in the form of speech, sign language or writing.
Social cognition: the ways in which people make sense of themselves and of others in order to function effectively in a social world.

Conceptually driven or top-down processes reflect the influence of higher order cognitive processes such as thought, beliefs and expectations.

Language and communication

Communication: any means by which information is shared.

Language has been particularly important for human evolution because it promotes social bond and social interaction and because it provides an effective means of persuading others.

Two ways in which we can use language to communicate:

• Writing
  Involves converting thoughts or speech to print.
  Plays a vital role in language survival, by allowing a record of the language to be retained across generations.
• Speech
  A feature of human cognition for tens of thousands of years, and without a parallel in the animal kingdom.

People also communicate non-verbally. Non-language vocalizations an convey information, and gestures can supplement or substitute for spoken language.
Gesture is so closely tied to human language that we continue to gesture even when we cannot be seen.
Subtler non-verbal signals such as body language and tone of voice also communicate to others.

Language universals

Mental lexicon: our store of knowledge about words and their uses.

Languages vary in number and type of sounds used, in basic word order, in the size of their vocabularies and in their rules for sentence construction.
However, all languages are capable of expressing complex and new ideas. Non are primitive. The expression of complex ideas is evident in all languages and in all human groups.

Linguistic universals: linguistic features said to be found in all languages.
All languages:

• Have consonants and vowels
• Combine basic sounds into larger units
• Have nouns
  (like book)
• Have verbs
  (like read)
• Can combine words in meaningful ways
• Can express who did what to whom
• Can express sentences as negatives
• Can express sentences as questions
• Involve a syntactic structure of grammar
• Allow reconstruction

There are immediate problems with this listing

• Sing languages
• Some languages do not reliability distinguish between classes of nouns and verbs
• Tonal languages use changes in tone to alter the meaning of the word
• Some languages have unusual classes of consonant words (like click languages)

Hockett’s design features for human language

• Vocal-audiory communication channel
• Broadcast transmission and directional reception.
  The speech signal is transmitted out from the source and is localized in space by the receiver.
• Rapid fading
• Interchangeability
  The sender can also be receiver and vice verse. The role is not fixed.
• Feedback
  The speaker has access to the message and can monitor its content.
• Specialization
  The energy expended in producing the message does not alter the meaning of the message (we can change the emphasis).
• Semanticity
  Sounds within speech refer to objects and entities in the world. Words have meaning.
  Functional reference: the use by animals of a specific call to stand for a specific object or threat.
• Arbitrariness
  The relationship between the spoken words and its referent in the world is arbitrary. The form of the signal does not relate to its meaning directly.
• Discreteness
  The speech signal is composed of discrete units. Vocabulary is built up from smaller meaning units and so on.
• Displacements
  We can use language to refer to things that are displaced from the present situation, either in time or space.
• Productivity
  Language allows us to create novel utterances
• Cultural transmission
  A language is learned through interaction with more experienced users of the language within a verbal community
• Duality
  Meaningful elements are created by combining a small set of meaningless units
• Prevarication
  Language can be used to deceive and lie
• Reflexiveness
  We can use language to communicate about language
• Learnability
  A language can be learned by a speaker of another language

These features are not independent.

The design features apply to spoken language and do not apply fully to sign languages or written languages.

Components of language

Language is a structured system which uses a finite set of sounds to construct words, sentences and ultimately conversations.

Phonemes

Phonemes are the basic sounds that make up speech within a language.
There are about 100 basic sounds units or phones (the basic speech sounds) that can be used to make up words.
Phonetics: the study of speech sounds.
Languages use a subset of phones, and languages vary in the number of sounds used.
Phoneme: the smallest meaningful sound unit within a language.
The number of phonemes within a language varies.

Some basic vowel sounds occur in all languages, but consonants can be used differently and are perceived differently.
Phonemes also change as a function of the surrounding sounds in words and in sentences, an effect referred to as co-articulation.
Allophones: phonetic variants of the same phoneme. Different phones that are treated as the same phoneme within a language.
Phonemes therefore do not correspond directly to physical sounds, rather they are abstract representations of the phonological units of a language, the units used to represent the forms of words in our mental lexicons.
A phoneme is a subjective category that is recognized as meaningful by the speakers of a language, but is not necessarily constant as objectively measured.

The tendency to perceive the difference between two allophones decreases with age, suggesting a critical period beyond which the adult is tuned to the sound of his or her native language.
While a child can discriminate between the full set of phones, an adult generally cannot.

Phonotactic rules: stipulate which combinations of sounds are permitted in language.
These rules differ across languages so that sounds that are ‘natural’ and easy to produce and discriminate in one language may not be so easy for adult speakers of another language.
Speech segmentation relies on knowledge of word boundaries using information about phontactic probabilities in a language.

Changing a phoneme within a word will change the meaning of that word.

Morphemes

Morphemes: the meaning units of language.
They are the building blocks of words.
Morphology: the level of linguistic analysis concerned with morphemes and their role within words. The study of rules in a language according to which words can be constructed.

Free morpheme: a morpheme that can stand alone as a word (car)
Bound morpheme: a morpheme that cannot form a word on its own, but forms a word when attached to a free morpheme. (cars)

Inflectional morpheme serves a grammatical function but does not change the syntactic category of the word to which it is attached (-ed and -ing and -s and so on)

Dervational morphemes create new words with new meaning when added to a stem (ify, ish, able)
They can change the grammatical category of the word.

Prefix: a morpheme to the start of a word
Suffix: a morpheme to the end of a word.

Language-specific rules govern the ways in which words can be altered.
Such alterations apply only to content words.

Function words: provide grammatical structure that shows how content words relate on each other within a sentence.

Semantics and the lexicon

Words: the smallest unit of grammar that can be meaningfully produced on its own. It can consist of one or more morphemes.

People also use other meaningful sounds, though not all are words. (like a groan).
A few words are not referential, they have no clear referent (like saying hello).

Semantics: the study of meaning of words and morphemes and the relationship between the words we use and the objects they refer to in the world.

Syntax

We construct novel sentences when we speak.

Productivity of language: the ability to generate novel utterances.

Two aspects of the language system allow us to use language productivity:

• Syntax
  The rules that determine the construction of phrases (a group of words referring to a particular idea) and sentences in a language.
  It relates to grammar. It reflects how language is used.
• Morphology

Slang: describes an informal pattern of speech that is considered to be non-standard.
Syntactically correct.

Sentences follow a hierarchical structure and are made up of two parts:

• A noun phrase
  Which contains a noun, the subject of the sentence
  The subject: the word or words that gives what the sentence is about or performs the action.
• A verb phrase
  Contains the verb and conveys the action of the sentence.

Object of a sentence: the word or words that receives the action, or is acted on, by the subject of the sentence.

One key property of syntax underlies the productivity of sentence construction.
Recursion: the ability to extend sentences infinitely by embedding phrases within sentences.
Recursion would seem to be a resilient property of human language as even young children who have been deprived of language input retain the ability to use recursion .

Discourse

Discourse: multi-sentence speech and includes dialogue, conversation and narrative.

Pragmatics: the understanding of the communicative functions of language and the conventions that govern language use.
At the level of discourse, the function of language in communicating directly and indirectly comes to the fore.
Distinction:

• Linguistic competence
  Our ability to construct sentences
• Communicative competence
  Our ability to communicate a message effectively.

Effective discourse is based on a shared understanding between those engaging in a conversation.

Conversations require turn-taking and cooperation and participants follow a set of implicit social conventions.
A variety of verbal and non-verbal signals serve to regulate the conversation by indicating who speaks when and for how long.

Four conversation rules or maxims that describe effective conversations and reflect expectations of . listeners:

• Maxim of quantity: the speaker should provide enough information in order to be understood but not too much information.
• Maxim of quality: the speaker should provide accurate information
• Maxim of relevance: the speaker should provide information that is relevant to the current topic of conversation.
• Maxim of manner: ambiguity and vagueness should be avoided.

If the maxims are violated, more cognitive processing is required to determine the response, or the participants may have to backtrack or repair the conversation.
Violations of these rules can also form the basis of humor.

Speech errors

Speech is produced at a rate about 15 speech sounds and 2-4 words per second. For the most part this is fluent and well-formed.

Aphasia: the term given to a group of speech disorders that occur following brain injury.

Hesitations and pauses

Disfluency: a hesitation or disruption to the normal fluency of speech.
Dysluency: refers to an abnormal disruption to fluency, such of following brain damage.

Disfluencies are more common than actual errors and vary with the situation and individual. They are a natural characteristic of fluent speech.
These pauses can be silent or filled (like um).
Filled pauses occur with less frequency than silent pauses and may serve to announce a delay in speech.
During pauses the speaker plans the articulation of their next words.

Clause: a part of a sentence containing a subject and verb.

Some disfluencies may facilitate comprehension.
Some disfluencies may act as cues that direct listeners’ attention towards a particular word.

Slips of the tongue

Parapraxes: slips of the tongue or other actions originally thought to reflect unconscious motives.

When errors occur, they are not random.
They are systematic and highly informative as to the nature of the underlying processing.

Analysis of speech errors points to the importance of the phrase as a unit of production, as errors rarely jump across phrase boundaries.
Errors preserve the consonant-vowel distinction and phonological errors are in keeping with the phonological constraints of a given language.
Exchange errors show a lexical in that they are more likely to result in a word than in a non-word.
The lexical bias effect reflects both immediate feedback between speech sounds and word forms and monitoring of inner speech producing a real word bias.
Non-word errors are more readily detected and repaired, while real word errors can slip through the net and remain undetected before being uttered.

Content words tend to exchange with content words and function words with other function words.

Stressed and unstressed syllables do not exchange with each other.
Transpositions generally stay within the same syntactic or morphological class.
There is a systematic process whereby a sentence is constructed such that the word’s skeleton or frame and its segmental content are independently generated.
Speech productivity is highly rule governed.

Speech errors are generally collected from spontaneous speech.

The tip of the tongue state

Tip-of-the-tongue state (TOT): a temporary inability to access a known word.

When we experience a TOT, we generally say whether we know the word and we may have access to some information about the word.
In languages in which the noun had gender, that information can be available in the TOT state, showing that access to syntactic category information is preserved, although access to the specific phonological word form is unavailable.

Feeling-of-knowing: a subjective sense of knowing that we know a word, and is an example of meta-memory (our knowledge about the contents of our memories).
This is:

• Universal
• Occurs about once a week
• Increases in frequency with age
• Frequently affects recall of proper names
• Often involves an available initial letter
• Often accompanied by other words
• Resolved on almost half of occasions

Bilinguals would seem to be more prone to the TOT experience.

Theories of speech production

There are a number of stages to speech production.

1. Conceptualization
   A poorly understood process by which a thought forms and is prepared to be conveyed through language.
   An abstract thought or idea becomes a verbal thought remain elusive.
2. Formulation of a linguistic plan.
   The concept must be translated so that the thought becomes language and the sentence that we want to output is planned.
   Comprising two sub-stages
   1. Lexicalization; the words are selected from the mental lexicon
      Lemma: an abstract word form that contains syntactic and semantic information about the word.
      The concept must connect with the lemma. The lemma dis not yet specify its phonological form.
   2. Syntactic planning
      The order in which the selected words will be output is decided
3. Articulation of the plan.
   The sounds for the words are accessed, the lexeme is specified and the motor program for speech output is planned and articulated.
   Lexeme; the basic lexical unit that gives the word’s morpho-phonological properties.
4. The output is monitored so that corrections can be made if necessary.

Modular theories of speech production

Garrett’s model

Garrett’s hierarchical model.
Speech is produced via a series of stages, proceeding in a top-down manner so that processing at lower levels does not influence that at higher levels.

• Inferential level.
  Conceptualize the massage that we want to express.
• Functional level
  The syntactic and semantic framework of the sentence is constructed.
• Positional level
  The words are allocated to positions within the syntactic frame.
• Phonological level
  The sounds for the content words are acquired
• Articulation level
  Sounds are prepare for speech

Garrett’s model suggests that content and function words are treated differently and this is supported by the data on errors.

Is does not predict errors that occur ‘across levels’.

Non-plan internal error: occurs wen concepts from the message level intrude when articulating the sentence, specifying the words at the phonological level.

Levelt’s model

Weaver++
focuses on the production of single words rather than the construction and output of whole sentences.
Focuses on the lexical access aspects of speech production.
The first two stages of processing involve lexical selection.
Three stages from encoding follow, before articulation occurs.
These two systems seem to involve quite different processes and involve different areas of the brain.

Stages:

1. Conceptual preparation
   The process leading up to the activation of lexical concept.
2. Lexical selection
   A lemma or abstract word is retrieved from the mental lexicon and its syntactic category is activated. A number of words might be primed based on meaning, with selection dependent on relative activation so that the more appropriate selection occurs.
3. Morphological encoding
   Once the lemma is selected, processing proceeds from the conceptual/ syntactic domain to the phonological/ articulatory domain. At this point, a TOT state can be produced.
4. The syllables that make up the word are computed.
5. Phonetic encoding
   The actual speech sounds activate at this stage.
6. Articulation
   The phonological information is transferred to a motor plan and executed by the articulatory system and speech musculature.

Sometimes one can access phonological information without accessing syntactic information, which dos not support the notion of two separate stages for syntactic and phonological information.
Processing at this stage involved three types of information:; the word’s morphology, its stress patterns, and the segments that make up the words activated.

The model considers the role of self-monitoring at multiple levels throughout the processing stages.
This provided a mechanism for the detection of errors that allows us to repair speech and involves the cognitive mechanisms involved in speech comprehension.
The precise means by which the mechanism operates and the attentional systems governing it are not elaborated.

Levelt’s account shows how a series of specialized modules contribute to the process of speech production.

As a modular account, feedback between levels is limited.
The retrieval of the word form occurs only after the lemma has been selected, there is no feedback from the word form to lemma levels.
Some types of speech errors suggest that feedback does occur.

Interactive theories of speech production

Dell’s model

Dell’s cascaded or spreading activation account.
Based on connectionist principles.
Uses the concept of spreading activation in a lexical network to show how competing activation across different levels might predict speech errors.
Activation from one level can affect processing at other levels, processing is interactive. Processing is also parallel such that information can be processed at different levels at the same time.

There are four levels

• A semantic level
  Where we conceptualize what it is we want to say
• A syntactic level
  The structure of the sentence is devised
• A morphological (word) level
  The morphemes that make up the target words are selected
• A phonological (sound) level.
  The sounds that make up those words are activated

The connections between the layers allow bidirectional spreading of activation
Aword unit can activate the phonological units at the layer below and the semantic units at the layer above.
Lexical access involves six steps:

1. The semantic units are activated by an external source
2. Activation spreads through the network
3. The word unit with the highest level of activation is selected and linked to the syntactic frame for the sentence, in the appropriate slot. Once the word has been placed in the frame, its activation reduces to zero.
4. When the time is right, based on the slot in the syntactic frame the word is assigned to, the phonological information activates. If a single word is to be produced, selection of the word triggers the phonological information.
5. Activation continues to spread, but phonological units linked to the selected word become more highly activated.
6. The most active phonological units are selected, and these are linked to slots in a phonological frame for the word. This allows the concept phoneme to drop into the correct slot in the word so that the sounds are output in the correct order.

During the planning stage, the various words selected for the sentence become active. Activation drops off once the word is placed in the sentence.
This is an interactive account and feedback can occur from later to earlier levels such that phonological level activation can inform processing at earlier stages.
A monitoring process to account for self-corrections and repairs.
Errors occur when activation for a non-target overrides that of the target morpheme, phoneme or word.
Word substitutions occur because a semantically related, but incorrect, choice achieves a higher activation than the target word.

Neuroscience of language production

Language involves a number of interacting brain areas, with many of the key language areas located within the left cerebral hemisphere in the majority of people.
Language involves a number of cognitive processes, interacting with systems for attention, memory, perception and motor function.
Sociocultural knowledge informs the ways in which we use language with others.
Neurolinguistics: the study of the relationship of brain function to language processing.

Lateralization of function

Sensory information coming into one side of the body is processed on the contralateral (opposite) side of the brain.

Different functions are associated with the left and right cortical hemispheres.
Language is largely a left hemisphere function while the right hemisphere is specialized for functions related to spatial/ holistic processing.
When a cognitive function is lateralized, one cortical hemisphere is dominant for that function.

Lateralization of function: the asymmetric representation of cognitive function in the cerebral hemispheres of humans and higher primates.
Particularly apparent when we consider the effects on cognitive processing of a set of conditions that gives rise to the split-brain phenomenon.

The left hemisphere and language

In the majority of people, speech is lateralized in the left hemisphere of the brain, and the left hemisphere is dominant for the majority of language functions.

Evidence from the typical population

Dichotic listening task: one where different stimuli are presented to each ear.

While auditory processing involves both contralateral and ipsilateral (same side) connections from ear to brain, contralateral connections are dominant.
There is a right-ear advantage for verbal stimuli.
This left hemisphere specialization seems to be in place at quite young age, as children as young as two years of age show a right-ear advantage for speech sounds.
Right-ear advantage may be restricted to consonant sounds. Consonants sounds and vowels may be treated differently.
The rapid changes in consonant sounds evident in human speech are complex auditory patterns and require high level sequential processing.
Different areas within the left hemisphere process information relating to meaning and to syntax.

Transcranial magnetic stimulation. A non-invasive method of temporarily exciting or inhibiting cortical areas.

Broca’s area plays a crucial role in grammar processing.
Potential applications to remedial intervention ins cases of language disorder after brain injury.

The right hemisphere is involved in emotional aspects of speech, prosody and aspects of non-literal speech.

Evidence from aphasia

Speech production results from the processing that occurs in a number of language areas located around the Sylvian fissure of the left hemisphere, an area referred to as the peri-Sylvian language region.
Damage to any of these areas can impair the ability to produce speech or writing.

Electrocortical stimulation of the surface of the cortex allows a surgeon to locate, and avoid damage to, brain regions associated with a particular cognitive function.

Wernicke-Geschwind model: a simplified model of language function used as the basis for classifying aphasia disorders.
We repeat a heard word boy processing of the following sequence of brain areas.
The word in the auditory cortex, information about the word meaning in Wernicke’s area and the output is sent to Broca’s area via a band of connecting fibers called the arcuate fasciculus.
Broca’s area prepares the speech output and a motor program for output is than articulated via the motor cortex.

Aphasia is the term used to describe a deficit in language following brain injury.
It generally refers to spoken language.
In aphasia the internal processing of language has broken down.

Crossed aphasia: language dysfunction following right hemisphere damage in a right-handed individual.

Aphasic disorders can be classified according whether they are fluent, non-fluent or pure type.

• In the pure disorder a particular faces of language is affected, while other language function remain intact.
• Fluent but empty speech. The person produces fluent sentences, but the content of the utterances is not as they intended.
• Non-fluent. Reduced speech output, slow or effortful speech.

Broca’s aphasia

Broca’s aphasia: an acquired language disorder characterized by non-fluent speech, reduced speech output and problems with grammar processing.
Global aphasia: an acquired language disorder involving extreme impairment of language function.

Patient with Broca’s aphasia show deficits ranging from serve mutisms to dysfluency or laboured speech.
Broca’s aphasia is categorized as non-fluent, expressive or productive aphasia.
Function words rather than content words tend to be compromised.
People with non-fluent aphasia are aware of their speech problems.

Wernicke’s aphasia.

Associated with damage further back in the brain than the region associated with Broca’s aphasia.
Classified as fluent aphasia.
These patients are generally unaware of the problem with their speech output.
People with Wernicke’s aphasia are likely to produce phonemic paraphasias, substitution errors in which a similar sounding word (or non-word) is produced instead of the target word.
The function words and the grammar structures of the sentence produced are relatively intact.

The arcuate fasciculus was identified as the band of fibers that connects Broca’s and Wernicke’s areas and disconnection of this band of fibers is associated with a specific deficit in repetition, a disorder known as conduction aphasia. When the patient as a specific difficulty affecting the repetition of speech.

• Severely impaired repetition
• Frequent phonemic paraphasias
• Repetitive self-corrections
• Word finding difficulties

Repetition deficits are a key feature.

Anomic aphasia

Anomic aphasia is when the patient has specific difficulty with word retrieval.
Relatively small lesions within the language areas can produce anomia.
For the individual with anomia, access to the word that he or she is searching for is denied, but the patient has not lost knowledge of the word or of its meaning.
This reflects a problem with translation between word forms and their conceptual representations.

Writing

The processes involved in writhing are similar to those involved in speech production, but writing requires access to the orthographic (written) form of a word.
When we write, we have more time to think about what it is we want to express and to ‘translate’ it into written form.
We can also monitor the output more easily.
Writing is often a solitary activity, while the writer will have a reader in mind when writing, he or she lacks the immediate feedback.
Writing makes fewer demands on memory and therefore more complex ideas can develop through writing. The act of writing ideas down can facilitate thinking and bring about deeper understanding of the subject matter.

Research examining writing focuses on the later stages of the process, including composition and revision processes, as the earlier processes such as lexical retrieval and structuring a sentence have much in common with speech processes.

Composition is a process by which ideas are turned into symbols.

The Hayes and flower model of writing

A cognitive model of writing that focuses on three main domains affecting the writing process.

• Task environment
  The topic of focus, the intended readership and the purpose of the writing task
• Long-term memory
  The availability and accessibility of relevant information in long-term memory supports the writing process.
  The long-term memory system stores schematic information that will shape the writer’s view of readers’ expectations.
• The immediate cognitive aspects of the writing process.
  The writing process itself and the immediate demands it brings. Working memory demands are relevant to this component.

Three general stages of writing

1. Planning
   Includes the sub-operations of generating, organizing and global setting
2. Translating
   Convert ideas from memory into sentences on paper
3. Reviewing
   Reading and editing. A key stage in writing.

This model considers writing as a metacognitive act, with an executive process monitoring the key processes of planning, translating and reviewing.
The goal of writing is to create ‘reader-based propose’ as opposed to ‘writer-based propose’.