Topic 6: Language

Philosophical Foundations and Defining Natural Language

• Why Human Language Is Unique:

  • Scientists generally explain human intellectual uniqueness in two ways:

  • The large growth of our prefrontal cortex, which allows advanced reasoning, planning, and problem-solving.

  • Our unique ability to use natural language, which lets us share ideas, build culture, and express endless new thoughts.

  • An 18th-century cardinal once watched a chimpanzee and said, "Speak and I shall baptize thee." This highlights the long-standing belief that language is the ultimate divide between humans and animals.

  • René Descartes believed animals were simply biological machines. He was willing to explain animal behavior purely through mechanical rules, but stopped at humans because of language. He could not imagine how any physical machine could understand or create an endless number of new, meaningful sentences suited to any context.

  • Modern cognitive science solves Descartes' puzzle using computational theory and recursive grammar.

• Communication vs. Language:

  • Communication: Sharing thoughts or feelings with another person. Because we cannot transfer thoughts directly, we turn internal thoughts into external symbols (sounds, written letters, gestures).

  • The Core Difference: All language communicates, but not all communication is language. Many systems share information without using true language (such as paintings, dance, or body language).

  • Comparing Real-World Examples:

  • A Baby's Cry or Smile: Communicates physical needs or emotions, but is not language because the sounds cannot be broken down and rearranged into new phrases.

  • A Computer Prompting for a Login ID: A borderline case of communication, but it uses formal programming language that follows strict structural rules.

  • Paintings (e.g., Botticelli's The Birth of Venus): Communicates emotions and ideas clearly, but lacks grammar, words, or sentence structure.

  • Textbooks and Academic Writings: Full examples of both communication and natural language.

  • A Dog Wagging Its Tail: Communicates feelings (like happiness or submission) through body language, but has no grammar or words.

  • Washoe the Chimp Using Sign Language: Learned over $100$ signs (like "want more water"). This is intentional communication, but it lacks the grammatical structure of true human language.

  • Sarah the Chimp: Used hundreds of signs. When she saw ducks land on a pond, she signed "water bird". Some viewed this as an early form of language because she combined two basic words into a new idea, though critics noted it lacked complex grammar.

  • Architectural Blueprints: Both communication and a formal visual language, using standardized symbols assembled under strict design rules.

  • American Sign Language ($\text{ASL}$): A complete, natural human language with its own grammar, word structure, and rules.

• The Four Core Properties of Natural Language:

  • In cognitive psychology, any true natural language must have four key properties:

  1. Systematicity:

     • If you understand one sentence structure, you automatically understand related structures.

     • Language is not just memorizing a list of phrases. Anyone who understands "John loves Mary" can automatically understand "Mary loves John", even if they have never heard that specific sentence before.

  2. Compositionality:

     • The meaning of a whole sentence depends directly on the meanings of its individual words plus the grammatical rules connecting them.

     • Words keep their core meaning across different contexts (for example, "loves" holds a similar relational meaning whether talking about family, friends, pets, or hobbies).

  3. Productivity (Generativity):

     • A finite brain can produce and understand an endless number of novel sentences.

     • Across Sentences: You can create an infinite number of distinct sentences. If you pick a random sentence from a book and search it online, it almost never appears anywhere else, showing that most sentences we produce are unique.

     • Within a Single Sentence: Sentences can expand forever by adding clauses (e.g., "The girl hit the boy" $\rightarrow$ "The girl hit the boy and he cried" $\rightarrow$ "The big girl hit the boy who was misbehaving and he cried loudly…").

     • The Combinatorial Engine: We achieve this infinite variety by applying a small set of repeating rules to a limited set of parts (in English: $26$ letters, about $40$ to $41$ speech sounds, and thousands of words).

  4. Inferential Coherence:

     • Sentences with similar logical forms must lead to similar conclusions.

     • If hearing "John and Peter and Susan went to the store" allows you to conclude "John went to the store", your brain uses the same logic to conclude "John and Peter went to the store".

• Hockett's Criteria for Human Language:

  • Arbitrariness and Meaning: In human speech, the link between a word's sound and its meaning is arbitrary (there is no physical reason the word "dog" represents a canine). In contrast, animal signs are often directly tied to their meaning (a dog growling and showing teeth directly prepares for biting). However, some animals use arbitrary calls, such as vervet monkeys making different warning sounds for snakes, leopards, and eagles.

  • Displacement in Time and Space: Humans can talk about things that are far away or happened in the past or future. Vervet monkeys only call out when danger is right in front of them. Honeybees show some displacement through their waggle dance (telling the hive the direction and distance of flowers), though their dance is continuous rather than made of separate words.

  • Discreteness and Regularity: Language is built from separate, distinct units (sounds, words) arranged by strict grammatical rules. Jumbled strings of random words are rejected by native speakers because they lack proper structure.

Neurobiology and Structural Components of Language

• Brain Lateralization:

  • In about $92\%$ of right-handed people, language is controlled by the left side of the brain.

  • For the $8\%$ who are left-handed, roughly half still process language on the left side. Overall, about $96\%$ of all humans process language primarily in the left cerebral hemisphere.

  • Split-brain studies show that the isolated right hemisphere has only very basic language skills.

  • This specialization is not caused by the left brain being physically larger; it comes from differences in microscopic neural wiring.

  • Non-human primates also show brain asymmetry, even though they do not develop natural language.

• Key Brain Areas for Language:

  • Broca's Area: Located in the left frontal lobe, right next to the motor area controlling the mouth and vocal cords. Discovered in $1861$ by Paul Broca. It handles grammar, sentence structure, and planning speech movements.

  • Wernicke's Area: Located in the left temporal lobe, behind the hearing center. Discovered by Carl Wernicke. It handles language comprehension and word meanings.

  • Parietal Centers (Angular and Supramarginal Gyri): Help connect visual words, written text, and spoken sounds together.

  • Language Disorders (Aphasias):

  • Broca's Aphasia: Speech is halting, slow, and lacks grammatical words, but the person generally understands what others say.

  • Wernicke's Aphasia: Speech sounds smooth and effortless but makes no sense ("word salad"), and the person struggles to understand spoken language.

  • Modern View: Real-world speech depends on broad, interconnected brain networks rather than strictly isolated spots.

  • Brain Plasticity: Young children who suffer damage to the left brain can often develop normal language skills using the right side of their brain instead.

• The Levels of Language:

  • Phonetics: The physical sounds of speech and how the mouth produces them.

  • Phonology: The rules for organizing speech sounds. A phoneme is the smallest sound unit that changes a word's meaning (such as the difference between the $/d/$ in duck and the $/t/$ in tuck).

  • Morphology: The rules for building words from basic units of meaning (morphemes), including prefixes, roots, and suffixes.

  • Syntax: The grammar rules that decide how words are ordered into proper phrases and sentences.

  • Semantics: The study of meaning and truth. A sentence can follow grammatical rules perfectly while being completely meaningless (for example, Chomsky's famous line: "Colorless green ideas sleep furiously").

  • Pragmatics: How social rules and context shape meaning. For example, if someone asks "Can you pass the salt?", social convention tells us to hand them the salt shaker rather than just answering "Yes, I can."

Phrase Structures, Grammar Rules, and Speech Production

• Linguistic Intuitions and the Competence vs. Performance Split:

  • Linguistic Intuitions: Native speakers naturally know whether a sentence sounds correct or awkward without needing formal grammar training.

  • Paraphrase Intuition: Knowing that two differently worded sentences mean the same thing (e.g., "The girl hit the boy" means the same as "The boy was hit by the girl").

  • Ambiguity Intuition: Spotting the difference between a double-meaning word (Lexical Ambiguity, like a river bank vs. a money bank) and a double-meaning sentence structure (Structural Ambiguity, like "They are cooking apples").

  • Competence vs. Performance (Chomsky):

  • Competence: The ideal, perfect knowledge of grammar stored inside a speaker's mind.

  • Performance: How people actually speak in daily life, which is often messy due to memory lapses, distractions, stumbles, and pauses.

  • Scientific Debate: Researchers sometimes disagree on whether speech errors show a flaw in a person's underlying grammar rules (competence) or are simply temporary slips caused by tiredness or memory limits (performance).

• Phrase Structure Trees:

  • Sentences are organized hierarchically into smaller chunks called phrases (noun phrases, verb phrases).

  • Shifting phrase order changes the meaning entirely, even when using identical words:

  • "Bill told the men to deliver the piano on Monday." (The delivery happens Monday).

  • "Bill told the men on Monday to deliver the piano." (The conversation happened Monday).

• Evidence That Phrase Structures Are Real in the Mind:

  • Speech Pauses: People pause longer between major grammatical phrases (averaging about $1.03\,\text{s}$) than within phrases ($0.75\,\text{s}$), showing that the brain plans speech phrase by phrase.

  • Speech Corrections: When people make a mistake and correct themselves, they restart at the beginning of the whole phrase ("Turn on the stove / the heater switch"), rather than repeating only the single wrong word.

  • Speech Slips (Spoonerisms): Swapping sounds accidentally (e.g., "bake my bike" instead of "take my bike").

  • Sound Errors (Positional Level): Sound swaps almost always stay within the same phrase ($87\%$ of the time).

  • Word Errors (Functional Level): Swapping entire words usually happens across different phrases ($83\%$ of the time, such as "That kid's toy makes a great mouse").

  • Laboratory Slip Tests: When researchers prime people to make sound slips, participants are far more likely to make errors that produce real words rather than nonsense words, and their brains automatically suppress taboo or vulgar words.

Transformational Grammar and Deep Structure

• Recursion:

  • Recursion occurs when a rule feeds back into itself, allowing a small set of instructions to create an endless variety of outputs (similar to rules in formal logic and mathematics).

• Why Simple Phrase Trees Fall Short:

  • Simple phrase trees cannot explain why a single structure can have two completely different meanings (such as "The shooting of the hunters was terrible", which can mean the hunters were shot or the hunters were bad at aiming).

  • They also fail to explain why two completely different sentence layouts (active vs. passive) can mean the exact same thing.

• Deep Structure vs. Surface Structure:

  • Russell's Logical Solution: Bertrand Russell showed that surface grammar can hide underlying logic. A sentence like "The king of France is bald" really claims that a King of France currently exists. Because France has no king, the claim is false at its deep logical level.

  • Chomsky's Model:

  • Deep Structure: The basic underlying meaning of the sentence, formed by core phrase rules.

  • Surface Structure: The final spoken or written sentence after words have been moved around.

  • Transformations: Mental movement rules that turn deep structures into surface sentences.

Animal Communication and Cognitive Limits

• Teaching Apes Language:

  • Vocal Limitations: Early attempts to teach chimps to speak failed because their throats and vocal cords cannot physically produce human speech sounds.

  • Gestures and Symbols:

  • Projects using sign language (such as Washoe) showed that chimps could learn many signs, but later studies (like Nim Chimpsky) revealed that the animals were mostly imitating trainers rather than generating original grammar.

  • Kanzi the Bonobo: Learned to use symbolic keyboard pictures (lexigrams) simply by watching his mother's lessons. By age $5.5\,\text{years}$, Kanzi understood spoken English commands about as well as a $2\text{-year-old}$ human child.

  • The Language Instinct (Pinker): Humans are biologically wired to learn grammar, just as songbirds are wired to sing their species' songs. While smart animals can learn associations, they lack the innate grammatical system found in humans.

  • Ethics: If apes can use true language, philosophers like Descartes argued they deserve moral and legal protections similar to humans, which led to movements limiting invasive animal testing.

Language and Thought

• Three Theoretical Possibilities:

  1. Thought depends on language.

  2. Language depends on thought.

  3. Language and thought are separate, independent systems.

• Does Thought Depend on Language?

  • Watson's Behaviorist View: Claimed thought is nothing more than tiny, silent movements of the vocal cords (subvocal speech).

  • The Curare Experiment: A researcher was temporarily paralyzed using the drug curare. With his muscles completely unable to move, he remained fully conscious, solved math problems in his head, and remembered conversations. This proved that thinking happens in the brain without needing muscle movements.

  • The Whorfian Hypothesis (Linguistic Determinism): The idea that the language you speak shapes how you perceive reality.

  • Whorf claimed Eskimo languages have dozens of words for snow, but later studies showed this was exaggerated.

  • Color Studies: English has $11$ basic color words, while the Dani people of New Guinea have only two (mili for dark/cool and mola for light/warm). Yet Dani speakers remember universal "focal" colors just as easily as English speakers, proving that visual biology shapes color thinking more than words do.

  • Subtle Influences: Comparing English and Berinmo speakers showed that having a specific color boundary in your language helps you sort colors faster near that boundary. Language does not blind us to reality, but it can bias our attention and memory.

• Does Language Depend on Thought?

  • Evolutionary Priority: Complex non-verbal thinking (like navigation and tool use) evolved long before spoken language.

  • Word Order Across Languages: In $98\%$ of the world's languages, the Subject comes before the Object ($\text{SOV}$, $\text{SVO}$, or $\text{VSO}$). This matches human thought, where we naturally think of an actor (Subject) doing something to an item (Object).

  • Kinship Terms: Languages balance memory effort with communication needs. They create single words for the family relationships people talk about most often.

  • Modifier Placement: Languages place describing words (adjectives) next to the nouns they describe so the brain's short-term memory is not overloaded.

• The Modularity View (Language as an Independent System):

  • Proposed by Jerry Fodor and Noam Chomsky, arguing that language is a separate mental module that works on its own rules.

  • Double Dissociations:

  • Williams Syndrome: People have low general intelligence and struggle with basic math, but speak fluently and expressively.

  • Specific Language Impairment ($\text{SLI}$): People have normal intelligence and hearing, but have severe, lifelong trouble learning grammar.

  • The FOXP2 Gene: Mutations in this gene cause speech and motor control problems. However, because humans share this gene variant with Neanderthals, it is part of a broader motor and neural system rather than a single "grammar gene."

Language Acquisition

• Universal Stages of Child Language Development:

  • 0 to 6 Months: Crying, basic sounds, and vowel noises.

  • 6 Months (Babbling): Producing consonant-vowel combinations. Infants are born able to tell apart all human speech sounds (for instance, Japanese infants can distinguish English $/l/$ and $/r/$ sounds). Over the first year, they lose the ability to distinguish sounds they do not hear regularly.

  • 12 Months (One-Word Stage): First single words appear (e.g., Mommy, bye, hot). Children often overextend words (calling all furry animals "dog").

  • 18 Months (Two-Word Stage): Telegraphic speech starts, combining two words to show basic relationships (like "more milk" or "Daddy go").

  • Multiword Stage: Children skip a strict three-word stage and quickly begin using longer sentences (three to eight words), gradually learning adult grammar.

• The Past-Tense Debate (Rules vs. Associations):

  • Children learn past-tense verbs in a U-shaped pattern:

  1. Stage 1: Memorize irregular words correctly by imitation (went, came).

  2. Stage 2: Learn the general rule ($-ed$) and overapply it to everything (goed, singed).

  3. Stage 3: Master both the general rules and the exceptions (went, played).

  • Connectionist View: Computer neural networks learned this U-shaped curve purely through pattern associations without needing programmed grammar rules.

  • Dual-System View: Modern brain scans suggest we use two systems: Broca's area handles grammar rules for regular verbs, while the temporal lobe uses memory networks to recall irregular verbs.

• Input Quality and the Poverty of the Stimulus:

  • No Negative Evidence: Parents rarely correct a child's grammar; they usually only correct whether a statement is true or false. Direct grammar corrections are almost entirely ignored by young children.

  • Motherese Is Optional: Simplified baby talk is not used in all cultures, yet children in those cultures still learn language at normal rates.

  • Homesign: Deaf children raised without sign language input invent their own structured gestural systems with regular grammar.

  • Chomsky's Conclusion: Because the speech children hear is messy, incomplete, and full of mistakes, children must have an innate biological toolkit—the Language Acquisition Device ($\text{LAD}$) containing Universal Grammar ($\text{UG}$)—to guide their learning.

• The Critical Period:

  • Learning a language fluently is easiest during childhood (from age $2$ to $12$).

  • Immigration Studies: Older learners can learn vocabulary and grammar well, but their accent and phonetics rarely match native levels.

  • Brain Scans: People who learn a second language in early childhood process grammar predominantly in the left hemisphere, while late learners show more scattered, bilateral brain activity.

  • Late Sign Language Learners: Deaf individuals who first learn sign language as teenagers or adults show lasting difficulties with complex grammar compared to native childhood signers.

• Universal Constraints and Setting Parameters:

  • Island Constraints: Grammar has invisible universal boundaries. For example, you can say "Which woman did John meet who knows the senator?", but you cannot say "Which senator did John meet the woman who knows?" Children never make these mistakes, suggesting innate structural boundaries.

  • Parameter Setting: Universal grammar acts like a control board with about $100$ binary switches:

  • The Pro-Drop Parameter: In Spanish and Italian, you can drop the subject pronoun ("Voy al cine"). In English, you must keep the pronoun, using placeholder words like "it" ("It is raining").

  • Toddlers may start with the switch set to drop pronouns ("No walk"), but hearing placeholder words like "it" flips the switch to the English setting.

Additional Topics in Language Research

• Latent Semantic Analysis ($\text{LSA}$):

  • Computer models can figure out the meanings of words simply by tracking which words frequently appear near each other across massive collections of text, even without seeing or touching real-world objects.

• Speech Fillers (Uh vs. Um):

  • Hesitation sounds are not random mistakes. Speakers use uh to announce a brief pause, while um announces a longer delay needed to plan complex phrases or sentences.

• Grammatical Gender and Thought:

  • In languages with arbitrary noun genders, speakers describe objects using gendered adjectives. For instance, German speakers (where key is masculine) call keys hard and jagged, while Spanish speakers (where key is feminine) describe them as shiny and tiny.

• Pidgins and Creoles:

  • When adults from different language backgrounds work together, they build a simplified contact jargon (pidgin) that lacks complex grammar.

  • When children grow up hearing a pidgin, their brains naturally add grammar, structure, and complex rules, turning it into a complete, sophisticated natural language (creole) within a single generation.