How to Study Language in Psychology

Language psychology organises around four core questions.

1. Language processing: How does the mind/brain convert sounds and symbols into meaning in real time? Key mechanisms: categorical perception of phonemes; the cohort model of word recognition; syntactic parsing (and its failures, as in garden-path sentences); semantic access and priming; pragmatic inference (Grice's cooperative principle). The phonological loop (Baddeley, 1986) keeps phonological strings active during comprehension of long or complex sentences.

2. Language development: How do children acquire language, and what does the trajectory reveal about its nature? The canonical sequence: canonical babbling (6–8 months) → first words (12 months) → vocabulary spurt with fast mapping (~18 months) → two-word combinations → grammatical morphology (Wug test, 1958). Key phenomena: overextension, overregularisation (U-shaped curve), and the critical period (Johnson & Newport, 1989).

3. Language and thought: Does language influence cognition? The Sapir-Whorf hypothesis (weak form = linguistic relativity) has empirical support: Russian speakers discriminate blues faster at the linguistic boundary (Winawer et al., 2007); Mandarin speakers use vertical spatial frames for time more readily than English speakers (Boroditsky, 2001). But the strong form (linguistic determinism) is falsified — the Dani (Rosch, 1972) and other cross-cultural data show that perceptual categories exist without corresponding lexical items.

4. Language and the brain: Broca's area (BA 44/45, left inferior frontal gyrus) — discovered via Tan (Leborgne, 1861) — is involved in production but also syntax, phonological working memory, and hierarchical structure building. Wernicke's area (left posterior superior temporal gyrus) is critical for comprehension. The arcuate fasciculus connects them; damage causes conduction aphasia (fluent, comprehension intact, repetition impossible). The dual-stream model (Hickok & Poeppel, 2007): ventral stream for meaning ("what"); dorsal stream for sensorimotor articulation ("how").

This is the central controversy in language acquisition and the one most likely to appear in examinations.

Nativist (Chomsky): Language is too complex, too fast, and too universal to be learned from the impoverished input children receive. Children must come equipped with Universal Grammar (UG) — innate constraints on possible grammars — and a Language Acquisition Device (LAD) that triggers on exposure to any human language. Key arguments: poverty of the stimulus (structure-dependence is never taught but always respected); linguistic universals (Greenberg, 1966); the critical period (Lenneberg, 1967).

Behaviourist (Skinner): Language is shaped by operant conditioning — reinforcement of correct utterances, imitation, and caregiver correction. Demolished by Chomsky's 1959 review: cannot explain productivity, the poverty of the stimulus, or the universality of grammar.

Statistical learning / connectionism (Saffran; Rumelhart & McClelland): Language is learned by a domain-general mechanism that detects statistical regularities. Saffran et al. (1996) showed 8-month-olds segment words from continuous speech using transitional probabilities after 2 minutes of exposure. Rumelhart & McClelland (1986) showed U-shaped past-tense learning without explicit rules. Limitations: models require carefully crafted inputs and struggle with rare structures.

Usage-based / social-pragmatic (Tomasello): Children learn language in social interaction by reading communicative intentions within joint attention frames, then abstracting constructions from item-specific exemplars. No innate grammar required — joint attention and intention-reading are the species-specific adaptations. Strength: explains why language development is so tightly coupled to social development.

Where the debate stands: Strong nativism has weakened — UG is now more minimalist (Chomsky's 2002 Minimalist Program); the debate has shifted to what exactly is innate (hierarchical recursive merge? prosodic preferences?) rather than whether there is any innate component. Most researchers accept some interaction between domain-specific priors and domain-general learning.

Categorical perception: Liberman et al. (1957) — VOT continuum; phoneme boundaries are sharp even when acoustic signal is continuous.

McGurk effect: McGurk & MacDonald (1976) — /ba/ + /ga/ lip movements = /da/. Speech is multimodal.

Cohort model: Marslen-Wilson (1987) — words recognised at the uniqueness point, before the final phoneme.

Semantic priming: Meyer & Schvaneveldt (1971) — "doctor" primes faster recognition of "nurse". Semantic memory is associatively organised.

Garden-path sentences: Bever (1970); Ferreira & Clifton (1986) — syntax is parsed incrementally with costly reanalysis.

Grice's maxims: Grice (1975) — cooperative principle: quantity, quality, relation, manner. Conversational implicature explains what is communicated beyond what is literally said.

Wug test: Berko (1958) — "wugs" shows productive rule use, not rote memorisation.

Statistical learning: Saffran et al. (1996) — 8-month-olds segment words from continuous speech via transitional probabilities.

Johnson & Newport (1989) — critical period: pre-7 arrivals reach native-speaker level; post-puberty arrivals do not.

Winawer et al. (2007) — Russian colour advantage at linguistic boundary; eliminated by verbal interference (proves linguistic mediation).

Broca (1861) — Tan case; left inferior frontal gyrus = speech production.

Wernicke (1874) — posterior superior temporal gyrus = speech comprehension.

Hickok & Poeppel (2007) — dual-stream model: ventral (meaning) + dorsal (articulation).

Compare the aphasia types: Broca's (non-fluent production, telegraphic speech, good comprehension) vs Wernicke's (fluent but meaningless output, poor comprehension, paraphasias) vs conduction (fluent, good comprehension, impaired repetition — arcuate fasciculus lesion). Understanding the triple dissociation is more powerful than memorising each in isolation.

Apply linguistic relativity: When evaluating a claim about language and thought, ask: (a) is there a direct perception measure or only a linguistic one? (b) has verbal interference been tested? (c) is the effect cross-culturally replicated? Winawer et al. (2007) is the gold-standard design because it includes the verbal interference control — that's what lets you conclude the effect is genuinely linguistic.

Critique the poverty of the stimulus: Pullum & Scholz (2002) argue that the relevant input (sentences requiring structure-dependence) may not be as absent as Chomsky claims — corpora contain it. Statistical learning models can sometimes acquire structure-dependent rules from realistic input sizes. The debate is unresolved.

Compare acquisition theories on the same evidence: Overregularisation ("goed") is interpreted differently by each theory: Skinner (error that should be extinguished by correction); Chomsky (evidence of innate rule application); connectionism (emergent from statistical patterns); usage-based (transition from memorised item to abstracted construction). Any single piece of evidence supports multiple positions — this is what makes the debate productive.

Dual-stream vs classic Broca/Wernicke: The classic model predicts sharp dissociations; neuroimaging shows overlap and bilateral activation. Broca's area activates during comprehension of complex syntax; Wernicke's during phonological tasks. The dual-stream model better accounts for the full neuroimaging literature, but classical localisation remains useful for clinical lesion work.

Confusing Broca's and Wernicke's aphasia: Broca's = non-fluent, effortful, good comprehension ("telegraphic man who understands everything"). Wernicke's = fluent, meaningless jargon, poor comprehension ("talking perfectly but making no sense"). The mnemonic: Broca broke his speaking; Wernicke wrecked his understanding.

Describing Sapir-Whorf as one position: There are two versions. Linguistic determinism (strong) — language determines thought — is largely rejected. Linguistic relativity (weak) — language influences thought — has empirical support (Winawer, Boroditsky). Always specify which you mean.

Treating the critical period as absolute: Johnson & Newport (1989) show a gradient, not a cliff. Children before 7 perform like natives; those 7–puberty show progressive decline; those after puberty perform poorly but not uniformly. The window closes gradually.

Saying "Chomsky proved language is innate": Chomsky proposed an argument (poverty of the stimulus); empirical debate continues. Statistical learning and usage-based accounts provide alternative explanations for much of the relevant data. Describe Chomsky as the leading nativist position and the alternatives — don't treat it as settled.

Describing statistical learning as "just memorising": Saffran et al. (1996) showed infants generalize — they discriminate new word/non-word pairs built from the same syllable set. The mechanism is not memorisation of specific sequences; it is sensitivity to distributional regularities — a genuinely different kind of learning.

Conflating the LAD and Broca's area: The LAD is a theoretical construct. Broca's area is an anatomical region. They are not the same. Chomsky never claimed a specific neural correlate for the LAD.