Language Functions: Communication vs. Thought – Detailed Study Notes

Language is uniquely human, emerging $100{,}000{-}1{,}000{,}000$ years ago.
Two main proposals about its function:
- Language-for-Communication: primary purpose is to transmit information, emotions, cultural knowledge.
- Language-for-Thought: language provides the representational medium or scaffolding for (some/all) cognition.
- Claims range from “all propositional thought needs language” to “language merely facilitates specific reasoning processes or helps only during development.”
Evolutionary fitness could favor either (or both) functions.
Modern tools—high-resolution neuro-imaging, large multilingual corpora, information-theoretic analyses—now allow empirical tests of the predictions each hypothesis makes.

Left-lateralized, fronto-temporal network that supports:
- Storage of lexical items & syntactic rules.
- Production & comprehension across modalities (speech, sign, writing).
- Modality-independence ⇒ representations are abstract.
HLN components respond to both lexical semantics and syntactic dependencies.
Distinct from lower-level speech-specific regions (Broca’s articulatory area, Wernicke’s speech-perception area).
Corresponds to Hauser, Chomsky & Fitch’s “Faculty of Language in the Narrow Sense”.

Classic model: Broca (production) + Wernicke (comprehension) linked via arcuate fasciculus.
Current view:
- Adds distributed higher-level language areas (frontal + temporal) that handle abstract form-meaning mappings.
- Speech articulation & perception areas are functionally distinct and not sensitive to meaning.
- Damage to circumscribed HLN portions seldom yields lasting global language loss because of distribution & redundancy.

Severe aphasia patients (lexical + syntactic deficits) can:
- Solve arithmetic & algebra (Varley et al.).
- Perform formal logic, causal chains, scientific reasoning.
- Execute executive-function tasks, planning, navigation.
- Understand others’ beliefs (Theory of Mind) and make pragmatic inferences.
- Categorize objects & judge event semantics non-verbally.
Neuroimaging: HLN silent during non-linguistic reasoning tasks; instead, tasks recruit other networks:
- Multiple-Demand (MD) network – fluid intelligence, working memory, math, computer-code parsing.
- Theory-of-Mind (ToM) network – social reasoning.
Pre-verbal infants and many animals show sophisticated inference without language.

Intellectual impairments (Down, Williams syndromes, schizophrenia, focal MD lesions) can spare language yet undermine reasoning/problem-solving.
Hence, intact HLN ≠ intact cognition.

HLN and MD/ToM already dissociated by age $\approx4$ (fMRI localizers).
Deaf children with late exposure to any language still acquire rich causal, numerical & relational cognition, albeit with some delays.
Training studies: labels or syntactic constructions can facilitate certain tasks (e.g.
complement clauses aiding false-belief reasoning), but facilitation ≠ necessity.

Many non-human primates, corvids, elephants, cephalopods exhibit planning, tool use, social mind-reading without language.
Human brain evolution shows parallel expansion of multiple association networks (language, MD, ToM, Default) rather than one language-driven cascade.

An efficient channel must be easy to produce/comprehend, robust to noise, & learnable (Shannon 1948).
Empirical hallmarks across linguistic levels:
- Phonetics: vowels/consonants occupy dispersed acoustic space; inventories adapt to climate, bite configuration.
- Lexicon:
- Frequent/low-information words are shorter (Zipf-ian economy).
- Words re-use efficient phonetic clusters & allow polysemy for compression.
- Semantic domains (kinship, color, seasons, grammatical markers) obey complexity–informativeness trade-off: systems cluster near Pareto frontier (Fig 2a).
- Syntax:
- Compositionality balances expressiveness vs. learnability.
- Dependency Length Minimization (DLM): average distance between syntactically linked words is shorter than random baseline across >30 languages.
  - Cognitive cost of long dependencies $\uparrow$ working-memory load ⇒ selection pressure.
  - Explains Greenbergian word-order universals.
- Word-order shifts (SOV → SVO) accounted for by noisy-channel models: listener leverages positional cues under information loss $\bigl(C = I(S;\hat{S})\bigr)$ .
- Ambiguity: facilitates reuse of short forms; mathematically optimal under context-sensitive compression.
- A non-ambiguous code would require exponentially larger lexicon $|L| \propto 2^{H(M)}$ where $H(M)$ is message entropy.
- Constructed “unambiguous” language Ithkuil is unlearnable.

Dimensions:
1. Scope: all vs. specific thought domains.
2. Strength: necessary vs. facilitative.
3. Linguistic Target: universal mechanisms (lexicon, syntax) vs. language-specific categories.
4. Timing: developmental only vs. lifelong.
Strong necessity claims easiest to falsify; none survive current neuropsychological & neuroimaging evidence.

What precise representations/computations live in HLN? Alignment with large language models may offer clues.
Nature of thought representations: symbolic ( $\text{LOT}$ ) vs.
sub-symbolic vs.
hybrid.
Developmental timeline: what do proto-HLN regions do before language onset (6–30 months)?

Challenges human exceptionalism: cognition likely expanded via multiple parallel neural systems, not one linguistic leap.
Language’s power lies in inter-generational cultural transmission, explaining rapid accumulation of technology & social complexity.
Education & therapy: restoring or augmenting language aids communication and learning but need not rebuild reasoning circuits.
AI analogy: Large Language Models show language competence without genuine reasoning ⇒ mirrors biological dissociation.

Empirical double dissociation: thought survives without language; language may flourish without full reasoning ability.
Linguistic structures bear fingerprints of pressure for efficient, learnable, noise-tolerant communication—not of internal code for thought.
Language, reasoning, social cognition co-evolved, but current evidence supports communication as the primary function.
The cumulative cultural niche enabled by external language use, rather than an internal mentalese, underlies Homo sapiens’ ecological dominance.