Questions & explanations
1. How do speakers produce derived words like 'teacher' from 'teach'?
To produce 'teacher' from 'teach', speakers add the suffix '-er' which changes the verb 'teach' into a noun meaning 'one who teaches'. This is derivation, which creates new words with different meanings or word classes. Unlike inflection (like past tense), derivation often changes the core meaning or part of speech. Speakers must know which suffixes combine with which base words; for example, '-er' usually attaches to verbs to make agent nouns. Some derivations are less predictable, like 'arrival' from 'arrive' (adding '-al'). In production, speakers retrieve the base word and then select the appropriate derivational suffix, sometimes needing to adjust the spelling or pronunciation (e.g., 'teach' + '-er' = 'teacher' with no change, but 'create' + '-ion' = 'creation' with a stress shift).
2. Why might some researchers argue that negative evidence is not necessary for first language acquisition, and what evidence supports this view?
Some researchers, like those following Chomsky's theory, argue that negative evidence is not necessary for first language acquisition because children receive little direct correction in natural conversations. They point out that parents often correct the truth of a statement rather than its grammar, for example, saying 'Yes, you saw a dog' even if the child says 'I seed a dog.' Despite this, children still learn correct grammar. This suggests that children have an innate ability to learn language from positive evidence alone—just hearing correct sentences. However, other studies show that parents do sometimes correct grammar, and that correction can speed up learning. The debate continues, but it's clear that children can learn without negative evidence, though it may help.
3. How does language impairment in frontotemporal dementia differ from that in Alzheimer's disease?
Frontotemporal dementia (FTD) often affects language more severely and earlier than Alzheimer's. In one type of FTD called semantic dementia, people lose the meaning of words, so they might not know what a 'dog' is, even though they can speak fluently. In another type, progressive nonfluent aphasia, they have trouble forming sentences, with slow, effortful speech and grammar errors. In Alzheimer's, word-finding problems are common but comprehension of word meanings is usually preserved until later. Also, Alzheimer's typically affects memory first, while FTD often starts with language or behavior changes. The specific brain regions affected differ: FTD damages the frontal and temporal lobes, while Alzheimer's affects the hippocampus and temporal-parietal areas.
4. How did B.F. Skinner's behaviorist theory explain language acquisition, and what was Noam Chomsky's main criticism of it?
B.F. Skinner, a behaviorist, proposed that children learn language through imitation, reinforcement, and conditioning. For example, when a child says 'milk' and gets milk, the word is reinforced. Skinner argued that language is just a set of verbal behaviors learned like any other behavior. Noam Chomsky strongly criticized this, pointing out that children produce sentences they have never heard, like 'I goed,' which cannot be explained by imitation. Chomsky argued that children must have an innate 'language acquisition device' (LAD) that allows them to deduce grammar rules from limited input. He also noted that parents do not consistently correct grammar, yet children still learn complex rules. This debate shifted linguistics toward nativist theories.
5. What is Noam Chomsky's concept of 'Universal Grammar' and how does it explain the ease and speed of first language acquisition?
Noam Chomsky proposed that all humans are born with a 'Universal Grammar' (UG), an innate set of principles that underlie all languages. UG explains why children learn language so quickly and with little explicit teaching: they already have a framework for how language works. For example, UG includes the idea that sentences have a subject and a verb, and that words can be moved to form questions. Children need only to learn the specific settings for their language, like word order. This nativist theory accounts for the fact that children everywhere follow similar stages of language development, despite different environments. Chomsky's ideas revolutionized linguistics and sparked decades of research on language acquisition.
6. What was John Locke's empiricist view on how children learn language, and how does it differ from the nativist view?
John Locke, a 17th-century philosopher, argued that the mind is a 'blank slate' (tabula rasa) at birth, and all knowledge comes from experience through the senses. For language, he believed children learn words by hearing them and associating them with objects and actions. This empiricist view contrasts with nativism, which says language ability is innate. Locke's idea suggests that language is learned entirely from the environment, through imitation and reinforcement. Modern behaviorists like B.F. Skinner later developed this idea, proposing that language is a set of habits learned through conditioning. However, nativists like Chomsky argue that children learn too quickly and make too few errors for this to be true.
7. What is a key challenge in integrating neural networks with symbolic reasoning for language comprehension?
A key challenge is that neural networks work with continuous numbers (vectors), while symbolic reasoning works with discrete symbols (like 'cat' or 'subject'). Bridging this gap requires converting neural representations into symbolic form, which can lose information. For example, the neural network might represent 'cat' as a vector of numbers, but the symbolic reasoner needs a clear category. Also, training the hybrid model end-to-end is difficult because the symbolic part may not be differentiable (cannot use backpropagation). Researchers often use separate modules that communicate via interfaces, which can be inefficient. Another challenge is scaling symbolic reasoning to handle the complexity of natural language.
8. In a classroom where a teacher says 'Good try, but we say 'she doesn't like it' not 'she don't like it',' what is the likely effect on the student's future use of that grammar rule?
The teacher's explicit corrective feedback is likely to help the student learn the correct form 'doesn't' instead of 'don't' for third-person singular. Research shows that such feedback can improve accuracy, especially when the student is attentive and understands the rule. The student may remember the correction and apply it in future sentences. However, the effect may not be immediate; the student might still make the error sometimes, especially in spontaneous speech. Repeated correction over time can lead to lasting change. The feedback also signals that the teacher values correct grammar, which can motivate the student to self-correct. Overall, explicit correction is a useful tool in language teaching.
9. Why is syntactic priming important for understanding how we produce sentences?
Syntactic priming shows that sentence production is not just about choosing words; it also involves activating abstract grammatical structures. This suggests that our mental grammar includes separate representations for sentence patterns, like passive or dative structures. Priming effects occur even when speakers are not aware of the repetition, indicating that structure activation is automatic. This helps explain why we often reuse structures in conversation, making speech more fluent. It also supports models of language production where syntactic planning happens before word selection. Overall, syntactic priming reveals that our brains prepare sentence frames independently of the specific words.
10. How do neural network models of ambiguity resolution typically handle the 'garden path' effect, where an initial interpretation turns out to be wrong?
Neural networks often do not explicitly reanalyze; instead, they compute probabilities for different interpretations incrementally. For a garden path sentence like 'The horse raced past the barn fell', the network might initially assign high probability to 'raced' as the main verb. When 'fell' appears, the probability of that interpretation drops, and the network may shift to the reduced relative clause interpretation. However, unlike humans, networks do not have a conscious reanalysis process; they simply update their hidden state or attention weights. Some models incorporate a reanalysis mechanism that backtracks, but most current models do not explicitly model the garden path experience.
11. How do researchers test for syntactic priming in an experiment?
In a typical syntactic priming experiment, participants take turns with a confederate (a helper) describing pictures. The confederate first describes a picture using a specific structure, like a double-object dative (e.g., 'The man gave the woman a book'). Then the participant describes a different picture. The researcher counts how often the participant uses the same structure (e.g., 'The boy handed the girl a flower') compared to an alternative (e.g., 'The boy handed a flower to the girl'). If participants use the primed structure more often, that is evidence of syntactic priming. The pictures are carefully chosen to have different words, so priming is due to structure, not vocabulary.
12. Why might bilingualism improve theory of mind? Give a possible reason.
One reason is that bilinguals constantly monitor which language to use with whom, which trains them to pay attention to others' perspectives. For instance, a bilingual child must know that their grandmother speaks only Spanish, so they must speak Spanish with her. This requires understanding the listener's language ability. Additionally, bilinguals often have to inhibit one language while speaking another, which strengthens cognitive control. Good cognitive control helps in theory of mind tasks because you need to suppress your own knowledge to consider someone else's false belief. These everyday experiences may sharpen the ability to see things from another person's point of view.