Questions: Working Memory in Sentence Comprehension

Question 1

A native English speaker reads 'The rat the cat chased died' slowly and carefully but still finds it difficult to parse. She reads 'The cat chased a rat, and it died' with no difficulty. Both sentences are grammatically correct and describe the same event. What best explains the processing difference?

• A. The first sentence uses passive voice, which is harder to process than active voice
• B. Center-embedding requires holding an incomplete noun phrase ('the rat') in working memory while processing a full embedded clause before the matrix verb can be resolved — exceeding normal working memory capacity
• C. The first sentence contains a garden-path effect where 'chased' is initially misread as a past-tense main verb
• D. The second sentence uses a simpler vocabulary, reducing lexical processing demands

Answer: (B)

Both sentences use active voice and identical vocabulary. The difference is structural: 'The rat the cat chased died' is a center-embedded relative clause — the relative clause ('the cat chased') is embedded inside the matrix clause ('The rat… died'), forcing the reader to hold an unresolved noun phrase in working memory while processing the embedded material. Right-branching constructions ('The cat chased a rat, and it died') resolve the matrix clause first, minimizing memory load. The difficulty is not about vocabulary, voice, or garden-path reanalysis — it is about working memory capacity under structural dependency.

Question 2

In processing 'Which book did you think the author intended to write about?', what memory demand does 'which book' impose on the parser?

• A. It must be stored as a semantic concept because its meaning is ambiguous until the end of the sentence
• B. It functions as a filler that must be kept active in working memory while the parser searches through intervening embedded clauses for the corresponding gap position
• C. It triggers a garden-path effect because 'which' initially signals a question about the subject rather than the object
• D. No special memory demand — the parser resolves 'which book' immediately and moves on

Answer: (B)

Wh-movement constructions create long-distance filler-gap dependencies: 'which book' is the filler — a displaced element that must be linked to a gap elsewhere in the structure (the object position of 'write about'). The parser must keep an active representation of 'which book' throughout the entire intervening string ('did you think the author intended to write about') while scanning for where it semantically fits. The longer and more deeply embedded the intervening material, the greater the memory load and the slower/more error-prone comprehension becomes. This is a core finding in psycholinguistics: dependency length predicts processing difficulty.

Question 3

Sentences that are difficult for native speakers to comprehend, such as doubly center-embedded sentences, reveal gaps in their grammatical competence — they don't know the rules for those constructions.

• T. True
• F. False

Answer: False

False. This conflates competence with performance. Grammatical competence — the abstract internalized knowledge of a language's rules — is distinct from performance, which is what you can actually process under real-time cognitive constraints. When presented with 'The rat the cat the dog bit chased died,' most native speakers recognize upon analysis that it is grammatically well-formed and can identify the correct interpretation if given enough time. The difficulty is not ignorance of the rule; it is a working memory bottleneck that prevents real-time processing. The grammar permits unlimited center-embedding; working memory creates the ceiling.

Question 4

Because working memory limits are universal cognitive constraints, the difficulty of center-embedded sentences is roughly equal across most languages and grammatical structures.

• T. True
• F. False

Answer: False

False. While working memory capacity is a universal constraint, languages differ in how they organize dependencies, and the same memory load can be distributed differently across a sentence. Some languages allow pro-drop or use morphological marking that reduces the need to hold fillers actively; others have head-final structures that create different dependency profiles. The center-embedding vs. right-branching contrast is most striking in English and similarly structured languages. Psycholinguists study cross-linguistic variation in processing difficulty to understand exactly how grammar and memory interact — and languages with different word orders often show different processing profiles for structurally equivalent sentences.

Question 5

What is the key theoretical distinction between linguistic competence and linguistic performance, and why does the study of working memory in sentence comprehension make this distinction unavoidable?

Model answer: Linguistic competence is abstract knowledge — the internalized system of grammatical rules that defines which sentences are well-formed. Performance is the actual real-time processing of language under cognitive constraints including working memory limits, attention, and processing speed. The distinction is unavoidable in working memory research because center-embedded sentences are both grammatically well-formed (competence says 'legal') and effectively incomprehensible in real time (performance says 'fails'). A theory that conflates the two cannot explain this divergence. Any adequate theory of language must specify both the grammatical rules (what is legal) and the resource architecture (what is actually processable) — and account for cases where they come apart.

Chomsky's competence/performance distinction was originally motivated partly by observations like this: the grammatical system is idealized and unlimited, but actual speakers process under real constraints. Psycholinguistics lives in the performance space — it studies the mechanisms by which competence knowledge is deployed in real time. Working memory research has been particularly productive because it identifies a specific, measurable resource that creates systematic gaps between what the grammar allows and what humans can actually process.