Exam Language Production

The 4 Stages of Language Production

1. Conceptualization (The “What”): You decide what you want to say. This is "pre-verbal"—you have the idea (a "message"), but no words yet.

• Macroplanning: Your overall goal (e.g., "I want to ask for a coffee").

• Microplanning: Choosing the specific perspective (e.g., "Could I have..." vs. "Give me...").

2. Formulation (The “How”): This is where the idea gets "translated" into linguistic code. It has two sub-steps:

• Grammatical Encoding: You pick the Lemmas (abstract words with meaning and grammar but no sound yet) and build a sentence frame.

• Phonological Encoding: You retrieve the Lexemes (the actual sounds) and build a phonetic plan.

3. Articulation (The Physical): Your brain sends motor commands to your lungs, larynx, tongue, and lips to physically produce the sound waves.

4. Self-Monitoring (The “Quality Control”): You "listen" to your internal speech (before you speak) and your external speech (after you speak) to catch errors.

Morpheme

 = the smallest unit of meaning or grammatical function in a language

• free morphemes → these can stand alone as complete words (e.g. cat, house, run, the)

• bound morphemes → these cannot stand alone; they must be attached to a free morpheme (a “root” or “stem”) (e.g. -s, un-, -ed)

inflection

= different forms of same word, makes it grammatically fit into the sentence (like tense, number, or person); never changes category

derivation

= creating a new lexeme (= a new "dictionary entry”); usually changes category of word (like verb to noun); and changes meaning

content word

= words that carry meaning

function word

= words that make grammar

Constituent tests

= are "diagnostic tools" linguists use to figure out how words are grouped together in a speaker's mind. (Tells whether something is a NP, VP, etc.)

• question test

• moving part of sentence test

• omission test (leaving it out)

constituency

= sentence is not just a linear sting of words, but a hierarchical structure of nested groups, these groups are called constituents

Three Types of Subordinate Clauses

1. Complement Clauses: These fill a “slot” required by the verb.

   1. Example: “I know [that you are tired].” (The clause is the object.)

2. Adverbial Clauses: These act as adjuncts, telling us when, why, or how. 

   1. Example: “[Because it rained], we stayed inside.” (Reason)

3. Relative Clauses: These modify a noun.

   1. Example: “The man [who lives there] is kind.” (Modifies ‘man’)

Syntactic knowledge

• constituency, syntactic knowledge

• syntactic ambiguity

• subcategorization information: the kinds of arguments a verb takes

• syntactic (grammatical) functions: subject, object, etc.

intransitive verb

= a verb that does not require a direct object

transitive verb

= only requires 1 NP argument in addition to subject

ditransitive verb

= requires 2 NP objects (plus the subject)

Thematic Role: Agent

= The intentional initiator of the action. – Can they choose to do it?

Thematic Role: Patient

= The entity that undergoes a change in state. – Is the entity different after the verb happens?

Thematic Role: Theme

= The entity being moved or whose location is described. – Is it just “there” or moving without changing form?

Thematic Role: Experiencer

= The entity that perceives or feels something. – Used with verbs of emotion senses (love, hear, see)

Thematic Role: Goal

= The endpoint of a movement or transfer. – Look for the preposition “to”T

Thematic Role: Instrument

= The tool used to carry out the action. – Often the subject in “The key opened the door.”

step 1: conceptualization = generating a pre-verbal message

• concepts/ relations the speaker intends to express

• what the listener is likely to know

• communicative goals

Conceptualisation a) Macroplanning: the overall goal

• speech acts

• linearization

• main vs. side information

• amount and type of information

• bookkeeping

• instrumentality → intention recognition

Conceptualisation b) microplanning: choosing the specific perspective

• perspective

• information structure

Evidence for conceptualization

Pausing

in reading aloud:

• planning prosody

• planning articulation

• breathing

in producing memorized utterances:

• all of the above, plus

• planning syntactic structure

in spontaneous speech:

• all of the above, plus

• selecting words, sentence structure

• monitoring, revising

How does macroplanning show in pausing?

choosing the order of big ideas/ linearization → Long pauses at the start of new topics or story sections

How does microplanning show in pausing?

choosing the specific perspective/ focus of a phrase → shorter, frequent pauses within a sentence or before complex words

What is formulation?

= idea translated into linguistic code

Formulation: functional processing

• lexical selection

• functional assignment

• result: unordered set of lemmas (= semantic aspect of words)

Formulation: prositional processing

• creating constituent frames

• inserting content words

• specifying function words and grammatical endings

→ encoding grammatical aspects

evidence: speech errors, syntactic aspects

Successful communication

• concepts / relations the speaker intends to express → communicative goal / intentions

• illocutionary intentions = what the speaker wants to express using language

• crucially, the addressee must recognise the speaker’s communicative intentions

Intentions and speech acts

• illocutionary intentions: specific “goals” in the speaker’s mind – informing, persuading, inquiring, requesting

• speech acts: the labels/ buckets linguists use to classify these goals – assertiveness, directness, expressives, declarations

Notation

→ making relations explicit

• notation refers to the “formal language” or “mental code” used to represent the pre-verbal message

• notation makes relations explicit by stripping away the ambiguity of human language and using a logical structure

• Example: 

  • KNOW (HANNA, INTEND (SIMON, BELIEVE (HANNA, ASTRONAUT (WOBBO)))) → Hanna knows (that) Simon intends Hanna to believe (that) Wubbo is an astronaut.

Bookkeeping: Discourse record

> structured dynamic representation of what has happened in the conversation so far

• type of discourse: register, style, turn-taking

• discourse topic: what the discourse is about

  • hierarchical structure: subtopics

  • topic shift

• discourse model: what the speaker believes to be shared knowledge in the discourse 

1. common ground: shared knowledge before the conversation

2. own contribution

   1. a + b: presuppositions

3. interlocutor’s contribution

4. information still to be conveyed

• attentional focus: speaker’s vs. hearer’s focal center

• exact wording: involvement, priming

Cooperation

> cooperation in conversation: adhering to rules of turn-taking & making appropriate contributions to the conversation

Grice (1975): Cooperative principles

“Make your conversation contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.” (Levelt (1989, p.39)

maxims of…

• quality: truthfulness

• quantity: right level of informativity

• relation: being relevant

• manner: clarity, brevity, orderliness

Instrumentality

instrumental message: changes the hearer’s discourse model in the intended way

→ make sure information is instrumental

how to make the message instrumental: 

• (make sure Grice maxims are met)

• selecting appropriate speech acts

• selecting appropriate referring expressions (a expression that refers to something, usually a noun phrase)

• separating information into a main-structure and side-structure

Selecting referring expressions

 → choice depends on alternative referents in the context 

• (like black ball of 2 balls differing in color)

• redundant information is included (contra Grice)

  • to help identify the object (“gestalt”)

  • to contrast with previously mentioned referent, i.e. the focal center (“endophoric redundancy”)

→ some referring-information (against Grice norms) can be helpful to provide extra information that allows to focus on specific goal

• collaborative reference: establishing names for things

• functional reference: the ham sandwich

Constructing request: factors:

• speaker’s firmness of wanting the action

• legitimacy of request

• hearer’s willingness and ability to perform the action

→ politeness: loopholes for the hearer to say No

Main-structure vs. side-structure information

main-structure / foreground information: hearer should detect intention

side-structure / background information: for the speaker

• giving associations, reasons, embellishments

• non-communicative reasons: appearing knowledgeable or pleasant

Pausing as evidence for planning phases

experimental evidence, findings:

→ amount of planning depends on complexity

→ verb and object closer interlinked than verb and subject

→ pause between 2 major constituents (subject – verb & object)

Pausing examples

• breathing

• reading aloud: planning prosody, articulation

• memorized or spontaneous utterances: also planning syntactic structure

• spontaneous speech: also selecting words, sentence structure

• monitoring, repair

→ alternating between fluent and hesitant phases

experimental evidence: route descriptions

• with familiar routes = little / no pausing

• with less familiar routes = more pausing

• pausing decreases when description is repeated

Macroplanning: Linearization

ordering the information units/ speech acts based on content: principle of natural order

• temporal, spatial (sequence of things told seen as temporal)

• entities / concepts that belong together (e.g. table + chair)

based on process: influenced by short-term memory limitations

• principle of connectivity (connected points, “tour”)

• stack principle: back to most recent choice point

• minimal load principle: simple before complex

Formulation: Assigning

• accessibility status ← speaker’s discourse model (we know this one, but the listener’s discourse model is unknown)

• topic status ← bookkeeping: topic structure (how relations are developing, keeping track of topic, of digression)

• propositional structure (preposition: subject - object structure)

Microplanning: Accessibility status

• inaccessible = not expected to be in listener’s discourse model, wasn’t mentioned ever before, not common knowledge

• accessible = expected to know, common knowledge, talked about before, in fruit from something else that was mentioned before

• in the discourse model / outside the discourse model

• in focus (part of the discourse model, that we are currently attending to → particular way of referring to it)

Referring expression of accessibility status as jusged by the speaker

 determined from the referring expression

• inaccessible → indefinite NP

• accessible → definite NP, name

• in the discourse model → definite NP, name

• in focus → name, pronoun

choice of referring expression influenced by Gricean maxim of quantity

• the more accessible the referent, the more reduced the referring expression (shorter description/ name, pronoun; deaccenting)

• reduced expression indicated high accessibility

Prominence

“newsworthiness”

• new referent, 

• old referent in new role

• non-prominent → deaccented

Microplanning: Topicalization

topic: what the message is about

→ where to store the information in the discourse model

(Example: The dog bit the cat. / The cat was bitten by the dog.)

Microplanning: Perspective — in directions

• speaker-centred, deictic: gaze tour (up/ down/ left/ right)

• intrinsic perspective: body tour (“turn around”, “straight on”)

Microplanning: Perspective — when describing scenes, figures - ground

• smaller entity described in relation to larger one

• proximal object described in relation to more peripheral one

• moving entity described in relation to stationary one

Formulating

• grammatical encoding

• lexically driven: lemmas + their syntactic “needs”

→ surface structure

• phonological encoding: lexical form, prosodic structure

fragments of message → Formulator → phonetic / articulatory plan

Lemmas

= the dictionary form of a lexical item, usually a Grundform that might even not be used in “normal” sentences

Grammatical encoding

• functional processing: lemma selection, grammatical relations

• positional processing: sentence frame, lexemes

• evidence: speech errors

Surface structure

• result of grammatical encoding;

• lemmas, grammatical relations, diacritic features (morphological and prosodic marking)

→ input into phonological encoding

Encoding grammatical functions

• word order (English) → syntactic configuration

• case-making morphology

Input to phonological encoding:

• lexical pointers: word-form information for lemma

• diacritic features: morphological (case, definiteness, tense, etc.) → one doesn’t get the fully formed word from the dictionary, phonological 

• phrase information: prosodic structure, weakening, etc.

3 Factors affecting prosody

mood: "Mood" refers to the grammatical category that reflects the speaker's purpose. Even with the exact same words, prosody changes to signal the mood. 

• Examples: declarative, interrogative, imperative

modality (commitment, desirability): Modality shows the speaker's attitude toward the proposition. Are you sure it's true? Do you want it to be true?

• Example: John came? vs. Did John come?

prosodic focus: This is arguably the most important part of Microplanning. You use prosody to highlight the "New" information and downplay the "Old" information.

• Example: prominence ← newsworthiness

The lexical entry includes the word’s…

• meaning

• form: pronunciation (and spelling)

• form: morphological makeup

• morphosyntactic information: syntactic category, inflectional properties etc.

Evidence for Lexical Access

• pausing

• speech errors

• brain imaging (detecting brain areas that are particularly active during some aspect of language production)

Pausing when producing words:

• frequency → activation level

• predictability: transitional predictability

Lexicalization

 = cognitive process of turing a pre-verbal message (conceptualization) into surface structure

• two-stage Process: grabbing meaning (lemma-retrieval) and then grabbing phonological encoding (lexeme-retrieval)

• this is proven by brain imaging

• alternative: network model, activation flow

The organization of the lexicon — reltions within lexical entries are

• relations within lexical entries are paradigmatic (different form of a word)

The organization of the lexicion — relations between items:

• intrinsic: based on features (meaning, form, morphosyntactic properties)

• associative

word association test: chair

• taxonomic relations:

  • table (coordinate)

  • furniture (hypernym)

  • stool (hyponym)

• meronyms: seat, cushion, legs

• attributive relations: comfortable, upholstered

• functional relations: sitting, rest, rocking

Lexical network model (Book and Levelt 1994)

• conceptual level

  • concepts linked by intrinsic and associative relations

• lemma retrieval

  • morphosyntactic properties: syntactic category, arguments, gender

• lexeme level

  • phonological properties

Interactive Activation

• facilitation between levels 

  • SHEEP activated WOOL

  • activating the word form facilitates the individual sounds

• inhibition within a level

  • activating SHEEP reduces the activation of GOAT

  • interactive; activation spreads in both direction

Speech errors involving lemma selection

• blends – synonyms

  • Example: paddle (spank + paddle), valify (validate + verify)

• substitutions – near-synonyms, antonyms

  • Example: Close it so it doesn't go fresh (stale).

• malapropisms – partly shared from, different meaning

  • Example: If these two vectors are equivocal (equivalent), then…

What does speech errors involving the lemma selection show:

→ these errors are systematic and can be sorted into 3 big classes, this is why we are interested in them: looking at (nearly) native speakers, people who usually don’t do this mistake, who know the rules of the language

conceptual intrusion: two concepts activate lemmas → blend

1. failure to choose between two appropriate concepts, the lemmas have roughly equal activation levels

2. distraction: potentially unrelated thought interferes during lemma retrieval

Assicoative intrusion

= substitutions

cause: activation spreading among connected nodes in the mental lexicon

Ease of lemma retrieval depends on:

• strength of connections

• frequency

• predictability: priming (facilitation, inhibition)

full listings vs. words-and-rules

• mostly morphemic storage, but

• some words stored whole:

  • frequent affixed words (impossible vs. imperceptible)

  • opaque compound (butterfly)

  • irregular forms (go/ went)

Exchange errors show what?

(1) Are my doors in the key? (Are my keys in the door?)

(2) I’d hear one if I knew it. (I’d know one if I heard it.)

→ evidence for parallel processing

→ words of same category are exchanged

Steps of Grammatical Encoding

1. lemma retrieved, grammatical information becomes available

2. phrase category constructed with lemma as head

3. complements, specifiers and diacritic information identified in the message

4. features and positions from (3) added to the phrase marker

5. elements corresponding to (4) are incorporated in the phrase marker

6. the phrase receives a grammatical function

7. the next (higher-order) categorial procedure is called → construction of the next phrase begins

Call hierarchy vs. destination hierarchy

> the logical thought (the “call”) and the final sentence structure (the “destination”)

• Complement of believe: Always a “proposition” (a thought with a subject and a verb)

• Destination: The physical “slot” in the sentence

The Big Takeaway: Grammatical encoding is “clausal”. The brain works in chinks of one subject/ predicate at a time. We know this because the “cognitive load” (measured by tones and pauses) stay the same regardless of whether we use a “that” clause or a “to be” structure.

Topic

• determined by speaker’s communicative purpose

• encoded in syntactically prominent position

• early in the sentence ← early retrieval

• evidence: elicitation of passive voice

• factors affecting saliency: 

  • animacy/ humanness (biased toward humans and animals) 

  • change of state (change implies newsworthiness) 

  • figure/ ground (noticing moving objects more than static background)

  • vividness (anything emotionally charged or visually striking)

• early constituents not always subject, e.g. object fronting in German

Cohesion

• relating to context: referent’s discourse status

• repetition of structure: syntactic priming

The “Strict” View: Information Encapsulated

• Levelt: system is encapsulated (this means each stage is a black box that does its job without knowing what is happening in other stages)

  • No Influence of Lexical Accessibility: This theory claims that just because a word is "easy to find" (accessible), it shouldn't force its way to the front of the sentence.

  • The Logic: Word order is decided by the Conceptualizer (based on Topic and Saliency). By the time your brain starts looking for the actual word in the Mental Lexicon, the grammatical "frame" has already been built.

The Result: If you can't find a word, you don't change the sentence; you simply pause until the word is retrieved.

The “flexible” view: Feedback and Interaction

• The second part suggests a “leakier” system where the stages are incremental and interactive.

  • Phonological Influence on Grammar: This is the idea that if the sound of a word is hard to retrieve (low phonological accessibility), your brain might realize this mid-sentence and "re-calculate" the grammar.

The “Availability” Strategy: Your brain prefers to put words that are ready into the sentence first. If a word is "difficult-to-access," the brain might choose a syntactic structure that pushes that difficult word to the end of the sentence to buy more time for retrieval.

Perseveration (Lingering Activation)

• A perseveration error is opposite. It occurs when a sound from an earlier word “hangs around” too long and replaces a sound in a later word.

• What it proves: This shows that once a sound is activated and spoken, it must be inhibited (turned off). A perseveration happens when the “turn-off” signal fails, and the old sound remains more active than the new one.

Anticipation (Planning ahead)

• An Anticipation error occurs when a sound that is supposed to appear later in the sentence “jumps the gun” and replaces a sound that comes earlier.

• What it proves: This is evidence that language production is incremental and involves look-ahead planning. Your brain has already activated the “future” word while you are still trying to say the “current” word.

Gestures

• accompanies speech but are not a full-fletched language (like sign language is)

• sign languages are just as intricate as spoken language

• spoken language also relies on gestures and iconic components

Co-Speech gestures

• representational gesture

  • carries content, information, adds to content of utterance

  • symbols/ emblems → conventional, culture-specific

  • indixes/ deictic gestures (expl. pointing → connected to situation)

  • iconic (lexical) gestures – more universal

  • example: thumbs up, pointing 

• non-representational gestures

  • does not carry content, more mood, but mostly puts emphasis on what is said

  • beats, batonic gestures

  • example: scolding and showing fist, doing something just to emphasise what is said

Functions of gestures for the listener

• communicating context

• discourse management (exmpl. her engaging with class)

• emphasis

Functions of gestures for the speaker

• lexical retrieval

  • more gestures with spontaneous speech

  • less fluent speech when gestures are not possible

  • gestures start before pronunciation

• planning and organization

  • more gestures with higher linguistic or processing complexity

• gestures also aid learning and memory, reduce cognitive load

speakers monitor the correctness/ appropriateness of…

• conceptual information 

  • intended message

  • best linearization

  • contect: cohesion, topic

  • register

• encoding

  • lexical errors

  • correct syntax and morphology

  • prosody and pronunciation

Evidence of monitoring

evidence: spontaneous self-correction

loudness, speech rate, etc. = adjustments but no/ few self-corrections

selective attention – context matters:

• non-word errors filtered among real words but less so / not among nonwords

• semantic priming: more bad – mug → mad bug-type errors following the sequence irate – wasp, angry – insect

• syntactic priming, e.g. errors resulting in Adj.+N pairs

• situation/ conversational context: “electricity” vs. “sexy” errors

more resources for monitoring towards constituent boundary

Connectionist theory of monitoring

• looks at language production as one big network

• when lemma is accessed → possible morphological components are activated, they might also be components of other lemmas, they are activated as well

Connectionist theory of monitoring: McKay’s node structure theory

• activation of muscle because of activation of lema

•  this muscle activation is felt, but maybe the specific activated sound is not needed → sensory feedback

• or you start pronouncing the word and realise something is wrong about (pronunciation/ word) only then

• catching errors short before or when it happens all part of one system

• one network capable of production and conception, and by that correction

• single system for production and monitoring (evidence: delayed auditory feedback)

• error detection: activation via bottom-up priming → perception

• prediction: error detection should always be immediate and automatic

• listening to ourself talk is similar to ordinary perception (right bottom of tree)

• detecting errors always immediately and automatically (because the perception mechanism is automatic as well)

Editor theories of monitoring

(here it is more than one system)

• external editor/ monitor

• distributed: following different levels of processing

• problem: reduplication of knowledge → everything twice not very helpful

Only prearticulatory monitoring (Motley, Camden & Baars)

evidence: “taboo” errors (tool kits; fits tall)

• socially appropriate partial speech error more frequent than inappropriate one

• socially appropriate partial speech error more frequent when taboo comes second

• emotional response even when error is intercepted

Levelt: external vs. internal self-monitoring loop

• same mechanism applying to internal speech and overt speech

• evidence: experiments with syllable sequences with and without auditory masking & brain imaging

  • overt-speech (external) loop slower

  • some types of errors (voicing, vowel quality) better detected via external loop

• internal feedback is faster

• monitoring system looking at end-product just before going out and shortly after

• internal is partially better because it is faster, but it can’t catch all errors

Repair process

• interruption

• editing expression (doesn’t always need to be included, but often is)

• making the repair

Main interruption rule: immediate interruption

• incorrect words are interrupted, not even phonological constraints are respected

• inappropriate words might be completed before being corrected (words that you are gonna use, but it would be better to use something before that as an example)

  • socially inappropriate are interrupted immediately

Editing expression

• placeholder in general, not just expression

Continuation problem for listener

• need to know what is being fixed, not only that something is being fixed

• might not be obvious how far one has to go back

• speaker is not consciously aware of that, but unconsciously provide this information

Help from speaker: structured repair

• syntactically regular: utterance + repair from a coordination (Well-Formedness Rule), less frequently: lemma substitution

• different ways of restarting for error vs. inappropriateness error: mostly instant repair or anticipatory retracing inappropriateness: often a fresh start → reformulation

• marking the position in utterance: Word-Idenity Convention, Category.Identity Convention

• prosodic marking of correction

• → considerate speaker

Speakers anticipating comprehension difficulty

• reduce connected speech processes that would affect low-frequency words

• pronounce more clearly words that are new to the discourse

• use more explicit syntactic structure

• complementiser (that) is used if it helps with ambiguity, or if the sentence is more complex in general → doesn’t mean speakers always avoid ambiguity (possibly because they are not aware of ambiguity)