Intro to comp theory of mind (1)

midterm psych-ling

modularity

Language - highly modular system

Modularity

1. General

   • language is a different system form vision

2. Specific

   • within language there are different systems (phonology, syntax, morphology etc)

cognition

process that takes place in the human brain

computational theory of the mind

•Mind: Analogous to software

•Brain: Analogous to Hardware

3 levels of understanding / Marr’s  Neurological basis of information processing

1. computations

   • What problem is being solved? → Understanding and producing language.

   • Why is this important? → Language enables communication, expression, and thought.

   • This level looks at the abstract goal of language processing

   • Theoretical linguistics

2. Algorithmic Level (How)

   • How does the brain process language?

   • This level describes the mental representations and rules used to understand and generate speech.

   • real time processing

3. Implementation Level (Physical Realization)

   • How does the brain (or a computer) physically carry out language processing?

   • In humans:

     • Brain regions like Broca’s area (speech production) and Wernicke’s area (language comprehension) are involved.

     • Neural circuits process phonemes, words, and sentences.

   • patterns of neural activation

   • relevant brain structures

the unification problem

how to relate the three levels of understanding

temporal and spatial summation

Temporal: The sum of several impulses form the same presynaptic neuron can create action potential in the postsynaptic neuron
Spatial: impulses form several neurons incoming to one neuron, the sum of the impulses reaches the threshold of action potential

refractory period

the amount of time it takes for an excitable membrane to be ready for a second stimulus once it returns to its resting state following an excitation.

fasciculus

a bundle of neurons with a common destination, neural pathway

glial cells

1. provide myelin

2. organize growth

3. absorb dead neurons

Cerebellum

•The cerebellum is involved in the coordination of voluntary motor movement, balance and equilibrium and muscle tone. It is located just above the brain stem and toward the back of the brain. It is relatively well protected from trauma compared to the frontal and temporal lobes and brain stem.

•Functions:

•Coordination of voluntary movement

•Balance and equilibrium

•Some memory for reflex motor acts.

Thalamus

•a routing station for all incoming sensory impulses except those of smell, transmitting them to higher (cerebral) nerve centers.

•connects various brain centers with others. Thus the thalamus is a major integrative complex, enabling sensory stimuli to evoke appropriate physical reactions as well as to affect emotions. 

Neocortex

•Neocortex: The newer portion of the cerebral cortex that serves as the center of higher mental functions for humans.

•Contains some 100 billion cells, each with 1,000 to 10,000 synapses (connections), and has roughly 100 million meters of wiring, all packed into a structure the size and thickness of a formal dinner napkin.

Lobes and functions

1. Frontal lobe

   • decision making

   • problem-solving

   • planning

2. Parietal lobe

   • reception and processing of the sensory information

   • spatial processing

3. Occipital lobe

   • concerned with vision

4. Temporal lobe

   • memory

   • emotion

   • hearing

   • language

   • object recognition

areas in the brain

Bromann’s areas

Gestalt Laws

• proximity

• similarity

• good continuation

• common fate

David Marr

familiar objects are configurations of simple components

Biederman

Geons

Log term potentiation (LPT)

Language processing requires the brain to form and strengthen neural connections to recognize words, interpret meaning, and produce speech. LTP plays a role in this by reinforcing frequently used neural pathways, making communication more efficient.

LTP occurs at the synapses, which are the junctions where two neurons communicate. When a neuron is repeatedly stimulated by another neuron, the synaptic strength between them increases, meaning the postsynaptic neuron (the receiving neuron) becomes more responsive to the same signal in the future.

• Increase in Synaptic Strength: The post-synaptic neuron becomes more sensitive to the pre-synaptic neuron's signal.

• Reduced Activation Threshold: With more AMPA receptors and other modifications, the post-synaptic neuron requires less input to reach the activation threshold, increasing the likelihood of firing.

• Persistent Changes in Synaptic Structure: Over time, dendritic spines (small protrusions on the post-synaptic neuron) may grow or change shape, physically altering the synaptic structure, making communication between neurons more efficient.

SAME STIMULUS CAUSES MORE POLARIZATION, BUT THAT PLATEAUS OVER TIME

multi-model perception

different systems interact - Mcgurc effect

human language capacity

translating acoustic words into meaning and back

Language is domain specific

• specific language impairment

• idiot savant

• Williams syndrome

• brain imaging research

• aphasia

memory trace

Definition:

• A memory trace (engram) is the neural representation of a memory, stored through physical and chemical changes in the brain.

Key Processes:

1. Encoding – Sensory input is processed and stored in neural circuits (hippocampus, cortex).

2. Storage – Strengthened through Long-Term Potentiation (LTP); memories move from short-term (hippocampus) to long-term (cortex).

3. Retrieval – The stored memory trace is reactivated when recalling information.

How Memory Traces Are Strengthened:

✅ Repetition & Practice – Strengthens neural connections.
✅ Association – Linking new info to known concepts enhances retention.
✅ Emotion – Strong emotional events create deeper memory traces.
✅ Sleep & Consolidation – Reinforces learning and memory stability.

Forgetting & Weakening of Memory Traces:

❌ Lack of Use – "Use it or lose it" (synapses weaken over time).
❌ Interference – New learning can overwrite old memories.
❌ Brain Damage – Conditions like Alzheimer’s disrupt memory traces.

Nativist view

Humans are innately predisposed to acquire language, due to genetic programm

Empiricist view

Tabula rasa - nurture, learning comes from experience

3 problems for learning from observation

1. too many encodings

2. False encodings

3. Abstract meanings

Universal Grammar Hypothesis

• All languages have the same basic, core properties

• These properties are innately available to children

• Language-specific input (polish vs dutch)

Dissociation between language and general intelligence

•Intact general intelligence - language impaired:

–Aphasia patients (later in this course)

–SLI, specific language impairment (later in this course)

•Intact language – cognitive retardation:

–Williams syndrome (later in this course)

–savant syndrome (later in this course)

Critical period

If a child does not acquire first language by approximately the age of puberty, it will never be able to acquire it as a mother tongue.

Reason: Brain loses its plasticity with age

Evidence: Multiple studies with immigrants and my personal experience…

Language is NOT imitation, why?

The phrase "Language learning is NOT imitation" suggests that language acquisition is more than just copying words and sentences. It highlights that:

1. Children Create New Sentences

   • If language learning were purely imitation, children would only repeat what they hear.

   • However, they often form new, grammatically correct sentences they’ve never heard before.

2. Errors Show Rule-Based Learning

   • Children make predictable errors like "I goed to the park" instead of "I went".

   • This shows they are applying learned grammar rules, not just imitating adults.

3. Chomsky’s Universal Grammar

   • Noam Chomsky argued that humans have an innate ability for language learning.

   • This explains why children can acquire complex structures without direct teaching.

4. Context, Meaning & Interaction Matter

   • Language learning involves understanding meaning and context, not just repeating words.

   • Social interaction, problem-solving, and cognitive processing play a key role.

5. Poverty of stimulus

Positive and negative evidence

1. Positive Evidence (Exposure to Correct Forms)

• Definition: Positive evidence refers to the correct language input that learners hear or read.

• How it works: It provides examples of what is possible in the language, helping learners identify correct structures.

• Examples:

  • A child hears: “She is running fast” → Learns correct verb usage.

  • A language learner sees: "I have eaten breakfast." → Learns correct past participle structure.

• Key point: Positive evidence helps learners form hypotheses about how the language works, but it does not directly tell them what is incorrect.

2. Negative Evidence (Indication of Errors)

• Definition: Negative evidence provides learners with information about what is incorrect in a language.

• Types of Negative Evidence:

  1. Explicit Correction (Direct Feedback)

     • Someone corrects an error directly.

     • Example:

       • Child: "He goed to the park."

       • Parent: "No, we say ‘He went to the park.’"

  2. Implicit Correction (Indirect Cues) – Also called Reformulation or Recasting

     • The error is not explicitly corrected, but the correct form is modeled.

     • Example:

       • Child: "He goed to the park."

       • Parent: "Oh yes, he went to the park!" (Without directly saying it was wrong)

positive: example of overt approval (you hypothesize, someone provides evidence)
negative: example that does not fit your theory (you assume SVO structure but then you hear John Mary likes), overt disapproval

Statistical learning

Child is innately sensitive to the statistical regularities in the input (not only linguistic)

The child is sensitive to how often (on average) it is a sunny day, how often (on average) a visual image of a dog is accompanied by a sound string /DOG/, etc

Sensitivity to infrmation

Information comes from things / events that are not typical (a sudden appearance of a kangaroo in this class, an unusual co-occurrence of consonants, etc.)

A learner is particularly sensitive to such unusual events and interprets them as meaningful.

psycholinguisitic experimentation categorization

1. off-line

   • Usually interested in speakers’ knowledge of certain rules, or the availability of a correct interpretation, or ability to process structures in general.

     It is not concerned with how difficult it is, or how fast/slow the underlying process is.

     Often used with children and brain damaged patients.

   • repetition task

     • Can experimental subjects repeat the sentences equally well?

       Do they replace the initial structure

   • picture selection task

     • Often used to see if the subjects get the right interpretation.

   • Truth Value Judgment Task (TVJT)

     • you show one picture at a time and then ask if its true of false

     • Kermit the frog says sentences, a child gives him ice cream when he’s right, shoe when wrong

   • Sentence completion Task

     • used to see if subject has access to a particular morpheme

     • This bear walks. This bear …

2. on-line

   • Interested in the time course of a certain process

     Sometimes interested in the localization – in brain imaging

     Used to see when the reaction takes place (e.g. how long it takes to detect anomaly)

   • CMLD (cross-modal lexical decision)

     • Cross-modal: Visual and auditory presentation

       Lexical decision: Is something a word?

     • reaction time in comprehension (secondary and primary task)]

     • common pool of resources for both primary and secondary - if one take too much energy there isn’t left for another

Comparison 1st vs 2nd language acquisition

broca’s are and syntactic complexity (slajd 86 class 5)

Brain and language research methods

Static and dynamic

Static Methods

1. Brain tumors

2. Penetrating wounds

3. Aphasia

   • all identify the relevance of a particular (damaged) are for a particular (linguistic) function

   • advantages: well-understood

   • disadvantages: we can’t compare with the patient prior to a lesion, lesion is much larger than the area we’re interested in

4. CT (CAT) scan - computed tomography

CT scan

Computed tomography

• 3D image

• The X-ray beam passes through the head

• cross-sectional images

• only structure

Advantages:

• very common, cheap, easily available

Disadvantages:

• doesn’t show the activity, only a still image

Dynamic methods

Haemodynamic:

1. PET Scan

2. fMRI

3. OT

Electrophysiological

1. EEG

2. MEG

EEG

Electroencephalography

• Electrical sensors placed on the scalp measuring the electrical activity (firing) of neurons

• ERP - event related potential

Advantages:

• noninvasive

• Excellent TEMPORAL RESOLUTION

Disadvantages:

• false information from echoes

• bad spatial resolution

ERPs

ELAN - early left anterior negativity - first signal of early language access 160 ms

P200 - Linked to phonetic processing and early word recognition.

N400 - Negative, Linked to meaning (semantics) and expectation violations.

• If a word doesn’t fit the expected meaning, your brain says: “Wait, that doesn’t make sense!” → N400 appears.

• “She spread the butter on the bread.” → Low N400
  “She spread the butter on the socks.” → High N400

P600 - Related to syntax (grammar) processing and reanalysis.

• If a sentence has weird grammar or structural ambiguity, the brain struggles → P600 appears.

• "The girl enjoys the movie." → Low P600 (correct grammar).
  "The girl enjoy the movie." → High P600 (grammatical error detected).

• Garden Path Sentences - Reanalysis Needed

Attributes:

• polarity - (N vs P)

• latency (100 vs 400ms)

• distribution over the scalp

iEEG

intracerebral EEG

• very high both temporal and spatial resolution

• only implemented in strictly clinical purposes

• only 5 to 9 electrodes in the region

MEG

Magnetoencephalography

• Neurons communicate using electrical signals → these signals create tiny magnetic fields

  The key principle: Whenever an electric current flows (neuronal activity), a magnetic field is generated.

  MEG uses highly sensitive sensors to detect these tiny magnetic signals.

• HIGH TEMPORAL RESOLUTION (no BOLD 0 blood oxygen level-dependent)

• GOOD SPATIAL RESOLUTION (number 3 on the list)

Advantages

• noninvasive and silent

Disadvantages

• Expensive

• difficult to maintain

• requires isolation fro noise, vibration and magnetic fields

PET

Petrision Emission Tomography

• A radioactive glucose tracer is injected, as the brain uses glucose in its activity. As the tracer breaks down it produces particles that collide with electrons, producing gamma rays that get picked up by the PET scanner

• Regions that are more active consume more glucose

• More activity = brighter colors (red, yellow)

• Less activity = darker colors (blue, purple)

• Reasonable spatial resolution (2nd place)

• Baaad temporal resolution (the worst out of the dynamic ones)

• tracks NEUROTRANMITTERS

Disadvantages

• it’s literally a radioactive injection

• New experimental conditions can be introduced only once every 10-15 minutes

• Subjects cannot be given mixed stimulus

MRI

Magnetic resonance imaging

• A huge magnetic field is introduced, forcing the hydrogen atoms in the body to align in the same direction, then a radio pulse is sent, knocking them out of that alignment. When it’s done, the hydrogen atom release a radio signal that is detected by the machine

• Greater detail than a CT scan

• not dynamic!

BOLD imaging

Blood oxygen level dependent

fMRI

• The brain sends oxygen-rich blood to the regions working hardest.

• Example: If you’re reading, Wernicke’s area (language comprehension) will show increased oxygen use.

• Oxygenated blood = Stronger MRI signal

• Deoxygenated blood = Weaker MRI signal

Advantages:

• safe, noninvasive

• no special prep

• number 1! on SPATIAL RESOLUTION list

Disadvanages

• expensive

• cannot be used with patients with metallic devices

• cannot be used with uncooperative patients

• claustrophobic

• loud

• TEMPORAL RESOLUTION - slower (seconds) but more detailed, number 4/5

Block design vs odd-ball design

• Both are experimental paradigms used in fMRI and ERP studies to measure brain activity under different conditions.

• Stimuli are grouped into blocks (e.g., 30 seconds of one task, then 30 seconds of another).

• Rare, unexpected stimuli (the “oddballs”) are mixed into a stream of frequent stimuli.

Block example:

• Seq. 1: wrote letters, ate apples, sang songs, etc.

  Seq. 2: wrote sang, ate rang, bring swam, etc.

Odd-ball example:

• wrote letters, ate apples, bring swam, sang songs

OT

Optical Topography

• BOLD, the inferred light diffuses differently on differently oxygenated and deoxygenated neurons

• Moderate/slow TEMPORAL RESOLUTION (like every BOLD)

• Moderate 4/5 SPATIAL RESOLUTION (only works near the surface)

Advantages:

• safe

• silent

• portable

• can be used with babies

Disadvantages:

• only cortex/near the surface

main properties of language

1. Creativity

2. Structure

3. Meaning

4. Reference

5. Interpersonality

creativity

Language allows us to generate infinite new sentences from a limited set of words & rules.

• You’ve probably never seen the sentence "Unicorns bake cookies on Mars" before, but you immediately understand it!

structure

Language follows systematic rules for combining sounds, words, and sentences.

• English has word order rules: "The cat chased the dog" ≠ "The dog chased the cat" (same words, different meaning).

Meaning

Words and sentences carry meaning, allowing us to convey thoughts, emotions, and information.

• "Love", "run", and "freedom" have meanings that exist beyond just their sound or letters.

Reference

Words stand for things in the real or imaginary world, even if they aren’t present.

• The word "apple" 🍏 refers to an actual apple, even if none are in sight. We can also talk about unicorns, which don’t exist!

Interpersonality

Language is used in social interactions to express ideas, influence others, and build relationships.

• Saying "Can you pass the salt?" isn’t just about salt—it’s a polite request in a conversation.

Phonetics

Speech is a translation of physical stimuli (acoustic waves) into something cognitively interpretable.

Phonetics deals with the human capacity to process sound, to separate it from the background noise, and to parse it into units interpretable by the next system – phonology.

Phonetics – intermediate position between physics and linguistics.

• Variability of the acoustic wave (male vs female)

• Separation from the background

• Language-specific (whats relevant for one doesn’t have to be relevant for another)

• Parameters: place and manner of articulation

Phonology

unit: phoneme

First purely linguistically meaningful level (distinctive phonemes worm vs warm)

Morphology

Morpheme - the smallest meaningful unit of speech

Free (stand-alone) and bound (attaches itself) morphemes, functional morphemes

Rules of morphological computations (ed and the end not beginning)

Syntax

Unit: phrase

phrases organized hierarchically

Syntactic computations are followed unconsciously by native speaker

INTUITION! of native speakers

Tranformations

D - structure, S - structure

Deep structure is SAD - simple, active, declarative (S,V,O, not a question)

• First, you access the words form the lexicon (words and morphemes), which requires efficient use of LTM and WM

• Syntax: you put words together in phrases

• that you from the D-structure

• then you transform it into an S-structure

• There could be wrong retrieval or wrong computations

Target S-structure: who was chased by the dog?

D-structure: the dog chased who

Semantics

Lexical and compositional

Lexical: word meaning, problem: almost impossible to give a finite definition (except biological terms)

Compositional: how we derive the meaning of sentences from individual words. The meaning of the whole is determined by the meaning of the parts + the structure.

• Truth coditions

Discourse, speaker-internal vs conversation internal knowledge

In discourse, meaning is not just about individual words or sentences—it’s about context, shared knowledge, and how speakers interact.

• Speaker-internal knowledge = What a speaker knows before entering a conversation (background knowledge, beliefs, assumptions).

• Conversation-internal knowledge = What speakers learn and negotiate during a conversation (new information, shared understanding).

• Speaker internal example:

  • Before starting a conversation, you already know that "Paris is the capital of France" or that "Water is wet."

  • If someone says "I just came back from Paris!", you already internally know what Paris is and what that might mean (travel, sightseeing, French culture, etc.).

• Conversation interanal example

  • If someone says, "I just came back from Paris, and it was raining the whole time!", you didn’t know about the rain until they said it—this is conversation-internal knowledge.

  • If they say, "My flight was delayed because of a strike," you just learned something new, which changes the course of the conversation.

If you say: the bride was young, it grammatically makes sense but not in a conversation internal discourse. ( I went to a wedding, the bride was young is okay)

BRIDGING

Bridging

Bridging is the process of linking new information to previously mentioned information in a discourse.

Involves lexical semantics, Lexical semantics = The study of word meanings and their relationships.

Since bridging requires linking new information to previous information, it depends heavily on lexical meaning and relationships between words.

Example: "I bought a new car. The vehicle is red."

• Bridging inference: "The vehicle" refers to "the car".

• This works because "car" and "vehicle" are semantically related (synonyms).

Pragmatics

The use of language

• What’s on TV tonight? --- Nothing

• Do you have time? --- Yes.

• She walks on the thin ice.

• John kicked the bucket.

Normal, expected interpretation appears to be impaired in right brain damaged patients.

Relevant area for memory

prefrontal cortex and hippocampus

Model for memory

sensory imput—> senory memory—→(attention) Short term memory (maintanance rehearsal)—→ (enocding/retrival) Long term memory

Memory stores (function, capacity ad storage time)

1. Sensory memory:

   • Sensory memory makes sure that the information is held in a buffer, long enough to be processed if necessary.

   • large capacity

   • very short (0,1-0,5 seconds)

   • all sensory modules have their own kind of sensory memory

     1. Iconic memory

        • sperling

2. Short term memory

   • Working Mememory

   • small capacity (limited memory span)

   • short duration (several seconds)

   • active

   • 7+2 items

   • word length effect

3. Long term memory

   • ACTIVATION

     • determines speed and accuracy of access

   • memories are activated when associated with present concepts —→ necessary to bring the element to the threshold

   • POWER OF LAW

   • large capacity

   • long lived

   • passive (does not require continuous effort to stay there)

Baddeley’s theory of working memory

1. phonological loop

   • 1-2 s

   • verbal thoughts

2. visuospatial sketchpad

   • visual thoughts

   • how does “p” look like when turned upside down?

3. episodic buffer

   • navigates

4. central executive

   • connects to LTM

loop and sketchpad independent = possible to do double tasks

power law

ower Law of Practice states that the more we practice something, the faster and more efficient we become—but improvements slow down over time.

long term potentiation

LTP

Long Term Potentiation

"Neurons that fire together, wire together."

• Step 1: A neuron sends a signal (Action Potential).

• Step 2: If this happens repeatedly, the receiving neuron becomes more responsive.

• Step 3: The synapse strengthens—making future signals easier & faster!

• increase of synapse surface

• more ion channels

• more transmitter vesicles

• also: new synapses

Prefrontal and hippocampal regions show decreased activation as participant become more practiced.

Base level of activation

The base level of activation is how "ready" a memory is to be retrieved.

• High activation = Memory is easy to recall.

• Low activation = Memory is harder to access.

• Every memory has a baseline activation level.

• Frequently used memories stay highly active

• Infrequent memories require extra effort to retrieve.

• When we comprehend language (listening/reading), the external stimulus (sound waves or written text) provides the activation energy to trigger word recognition.

• when we produce language (speaking/writing), we have to generate the activation internally, starting from our thoughts.

memory trace

the neural representation of a memory in the brain. It’s basically the physical "footprint" left in your brain when you experience something.

• Strengthened through Long-Term Potentiation (LTP) → The more a memory is used, the stronger the trace.

• Stored in neural networks → A combination of synaptic connections & patterns of activation.

• The distance of the memory trace to the threshold is measurable (reaction time)

Primed lexical decision

Prime and target

The Primed Lexical Decision Task (LDT) is an experiment where participants:

1. See a prime word first (e.g., doctor).

2. See a target word and must decide ASAP if it’s a real word (nurse or blork).

3. Reaction times are measured to see if the prime influenced word recognition.

• If the prime is related to the target → Faster response

• If the prime is unrelated → Slower response

• If the target is a non-word → Slowest response

Cross-modal lexical priming

Cross-Modal Lexical Priming is when a word in one modality (e.g., spoken) influences how quickly we recognize a word in another modality (e.g., visual).

• ou hear the word "bank" (auditory).

• A second later, you see the written word "money" or "river" (visual).

• If you recognize one faster than the other, it tells us which meaning was activated in your brain first!

David Swinney’s experiment

David Swinney’s experiment showed that when we hear an ambiguous word, our brain briefly activates all possible meanings, even if context makes one more likely.

Step 1: Participants listened to a sentence containing an ambiguous word.
Step 2: Right after hearing the ambiguous word, they were shown a visual word on a screen.
Step 3: They had to do a lexical decision task (decide if the visual word was a real word).
Step 4: Reaction times were measured.

Example Sentence Used in the Experiment:

• "The government building had a large bug in the office."

• The word "bug" can mean:

  • Insect (literal meaning)

  • Listening device (spy bug)

  • A general problem

What happened?

• Right after "bug", participants responded equally fast to words related to both meanings ("ant" and "spy").

• This showed that both meanings of "bug" were activated at first, even though context favored only one.

• What happened a second later?

  • After a slight delay (200-300ms), only the contextually appropriate meaning ("spy") remained active, while the unrelated meaning ("ant") faded.

What Does This Mean for Language Processing?

1. Lexical access is automatic → When we hear a word, our brain activates all possible meanings instantly.

2. Context takes time to narrow down meaning → After a short delay, only the relevant meaning stays active.

3. Cross-modal priming proves that lexical access happens in real-time across different sensory modes (hearing + vision).

Spreading activation

Spreading Activation is the process where activating one concept in memory "spreads" and activates related concepts.

1. Our mental lexicon (word storage in the brain) is organized like a network, where words are connected by meaning, sound, or experience.

2. When one word is activated, nearby related words also get partially activated, making it easier to retrieve them.

Factors influencing memories

1. Elaborative Processing

   • when we actively connect new information to things we already know. This makes memories stronger, richer, and easier to retrieve later.

2. The use of mental imagery

3. Personal and emotional relevance

Sentence superiority

Sentence Superiority Effect = Words are processed more easily & accurately when they appear in a meaningful sentence, rather than in isolation or in random word sequences.

• Context boosts word recognition.

• Sentences create expectations, helping predict upcoming words.

• Our brain is wired to process language holistically, not just as individual words.

Flashbulb memory

lashbulb Memory = A highly detailed, vivid memory of an emotionally significant event.

• Feels like a mental "snapshot" of the moment.

• Often involves major public or personal events.

• People remember not just the event, but also where they were, who they were with, what they were doing, and how they felt.

declarative memory

• General knowledge about the world

• knowledge of language, notably words (mental lexicon)

• irregular verbs

Procedural memory

• procedures

• skills

• regular verbs

Fun fact about occipital lobe

Same activation when looking at sth when just imagining sth

4 lobes and relevant fissures and 1 thing that connects 2 things

Malfunction in the monitoring

• stuttering (oversensitive and overdeveloped in the monitoring)

  • is the acoustic signal good enough representation of thought?

  • vicious cycle hypothesis

  • threshold of what’s acceptable is too high

  • distraction beneficial

Production side of the model

1. Conceptualizer

   • prepares a pre-linguistic, conceptual, structured message

   • information about individuals, their properties and events

   • adding thematic roles (agent, theme), not syntax yet!

   • Monitoring – Before speaking, you check if your message makes sense.

   • the end result is a pre-verbal message that goes into the formulator

2. Formulator

   • Grammatical Encoding – Selecting words (lemmas) and structuring them grammatically. = Deep structure is built from the lexicon

   • Surface Structure – Arranging the sentence correctly (word order, syntax).

     • Argument structure→ argument structure of a verb means how many specific arguments it can take (1 - John is jumping, 2 - John ate an apple, 3 - John gave an apple to Mary), also what kind of arguments (thematic roles, agent, theme etc)

     • Projection principle → reducing uncertainty

   • Phonological Encoding – Adding sounds to words (how they should be pronounced).

   • Sound structure

3. Articulator

   • Phonetic Plan → Muscle Movements – Your brain sends signals to your tongue, lips, and vocal cords to produce sound. (cortical homunculus)

   • Overt Speech – Finally, you say "I’m hungry!" out loud.

Lemma vs Lexeme

1. Lemma is the "base" of a word that carries its meaning and grammatical properties.

• It does NOT include morphological inflections (e.g., tense, plural, conjugation).

• It’s like the mental blueprint of a word before you actually say or write it.

Example:

• The lemma for "run" is the base form RUN (verb) 🏃‍♀

• But it does NOT specify if it’s "ran" (past), "runs" (third person), or "running" (progressive).

2. What is a Lexeme?

   • A lexeme is a word family—the base word plus all its inflected forms.

   • It includes all morphological variations of a word.

   • This is what we think of when we look up a word in a dictionary.

   Example:

   • The lexeme RUN includes:

     • Run (base form)

     • Runs (third-person singular)

     • Running (progressive form)

     • Ran (past tense)

First, the brain selects the lemma (choosing the base word & grammar).
Then, the lexeme is activated (choosing the right form for pronunciation).

ToT

Tip of the tongue: accompanied by a strong feeling of knowing the intended meaning and grammatical characteristics of the message

This phenomenon supports the idea of multiple stage model of lexicalization: A word is a complex system that has several faces: meaning, grammar, sound (and orthography when written.)

garret’s model

First vs second language acquisition

L1 acquisition:

• Happens Naturally – No formal instruction is needed.

• Critical Period Hypothesis (CPH) – If a child isn’t exposed to language by early childhood, they may never fully acquire it (e.g., Genie case study).

• Universal Stages – Babies go through cooing, babbling, one-word, two-word, and full-sentence stages.

• Mostly left hemisphere:

  • Broca’s Area (speech production)

  • Wernicke’s Area (comprehension)

  • Superior Temporal Gyrus (auditory processing)

• Highly automatic (no cognitive effort required)

• Low WM load (fast & efficient retrieval)

• Stronger, more efficient connections'

• Unsupervised

L2 aquisition

• More Effort Required – Learning is influenced by age, motivation, environment, and L1 influence.

• Interference from L1 – Grammar and pronunciation rules from the first language can carry over.

• More Variability – Some people achieve native-like fluency, others never fully do.

• More distributed activation:

  • Left hemisphere (if fluent)

  • Right hemisphere (if learned late)

  • More reliance on Prefrontal Cortex (working memory)

• Requires more cognitive control, esp. in early learners

• Higher WM load (more effortful retrieval)

• Weaker, more effortful connection

Critical period hapothesis

• L1 must be acquired before puberty for full fluency.

• L2 is harder to acquire after puberty because brain plasticity decreases.

Comprehension side model

Relevant areas:

Wernicke’s area

Auditory cortex

Key Components:

• Phonetic String → Parsed Speech – The sounds of language are analyzed and understood.

• Connection to Mental Lexicon – You match words to their meaning.

• Connection to Conceptualizer – You interpret speech in context.

Example:

• Your friend hears you say "I’m hungry."

• Their auditory system processes the phonetic string.

• Their speech comprehension system recognizes words and meaning.

• Their conceptualizer figures out how to respond: "Let’s get pizza!"

Self-Monitoring: You can also hear yourself talk and correct errors if needed.

Parsed speech is the structured version of spoken language after the brain has identified words, syntax, and meaning.

• First, speech is just a raw sound wave (phonetic string).

• Then, the brain processes it to recognize words & sentence structure.

• Finally, it turns into parsed speech—ready for comprehension.

The model includes feedback loops to monitor and correct speech errors.

How It Works:

• Before you speak, Monitoring checks if the message makes sense.

• After you speak, you hear yourself and can correct mistakes.

• If you say, "I’m thirsty—uh, I mean hungry!", that’s your speech comprehension system catching a mistake and fixing it in real-time

Categorical perceprion

Voiced vs unvoiced

voiced - first vocal cords activated, then lips released

unvoiced - first lips released then vocal chords activated

Understanding speech

1. Differentiation of speech from other sounds

2. Recognizing words

3. Activating their syntactic and semantic properties

4. Building their grammatical structure

5. Interpreting this structure

Proposed parsing algorithms

1. Wait and see

2. parallelism

3. conservative guessing

Wait and See

Slow but accurate – Avoids premature errors.
Works well for ambiguous sentences.
Prefers to avoid making incorrect assumptions.

Example Sentence:

• "The horse raced past the barn fell."

  • Wait-and-See Strategy: Does NOT immediately assume "raced" is the main verb.

  • Waits for more words to confirm whether "raced" is a verb or a reduced relative clause.

  • Delays commitment until full information is available.

Parallel parsing

The brain considers ALL possible interpretations simultaneously and waits to see which one is correct.

• Keeps multiple syntactic structures active at the same time

• Efficient for complex or ambiguous sentences.

• Used in computational models of language processing.

Example Sentence:

• "I saw the man with the telescope."

  • Parallel Parsing: Brain keeps BOTH possible meanings active:
    1. I used a telescope to see the man.

    2. The man had a telescope.

  • Waits for disambiguation later in the sentence.

Advantage: More flexibility, handles ambiguity well.
Disadvantage: Requires more cognitive effort & memory to keep multiple structures in mind.

Conservative guessing

The brain makes an immediate decision based on early input and sticks with it—sometimes leading to errors.

• Fastest parsing strategy – Prioritizes efficiency over accuracy.

• Leads to "garden path" effects when a sentence is misleading.

• Uses probabilistic cues & past experience to make quick guesses.

Example Sentence:

• "The old man the boats."

  • Conservative Guessing: Immediately assumes "old" is an adjective (not a noun).

  • WRONG! The sentence actually means "Old people are the ones who man the boats."

  • Brain needs to backtrack & reanalyze the sentence.

Advantage: Super fast processing.
Disadvantage: More errors in ambiguous or tricky sentences.

Incremental parsing

Incremental Parsing = The brain processes sentences word-by-word as they arrive, without waiting for the full sentence.

Super fast – We don’t wait until the sentence ends to start making sense of it.
Efficient for everyday speech – Allows real-time conversation & prediction.
BUT… it can lead to parsing errors!

Garden path

A Garden Path Sentence is a sentence that initially leads the reader to an incorrect interpretation, requiring reanalysis to understand the correct meaning.

• Happens because of Incremental Parsing → The brain processes words as they come in, making predictions.

• When the prediction is wrong, the brain has to backtrack & repair the structure (Later Repair/Reanalysis).

• Causes a moment of confusion!