Language and the Brain: Language Processing, Production, and Disorders

General Issues in Experimental Research

1. Choose a task or an experimental protocol that will address the research question.

2. Assemble appropriate materials. In Linguistics, they usually consist of words, sentences, and/or pictures presented to subjects either visually or auditorily. Participants are often 

   asked to respond or perform a particular task upon being presented with the materials.

• Two kinds of materials are required. In addition to the experimental critical stimuli (those words, sentences, and/or pictures that the researcher is interested in), filler stimuli are needed. These are other materials interspersed with the critical stimuli, used to prevent participants from guessing which aspect of the stimuli the researcher is interested in.

• Some experiments require one or more control conditions, which serve as a baseline against which results from the experimental condition(s) may be compared.

3. Collect a large enough number of participants to generate statistically significant data in order to be able to generalize findings — it’s impossible, for example, to draw conclusions about “what speakers of English do” if only three people have been tested

Choosing a Method

• Task that measures the process vs. the final result

  • Methods that allow us to catch language processing as it is happening are referred to as online tasks (note: this has nothing to do with the Internet).

• Methods in which subjects are asked about the results of a completed linguistic process are called offline tasks

• Naturalness of the task

  • Naturalistic task: something a person might do in the normal course of producing or perceiving language in their everyday environment

• Metalinguistic task: requires a person to use explicit knowledge about their language to make judgments about a word or sentence (Compare: Asking a person to explain exactly how to catch a baseball rather than to just catch one). For this reason, 

  metalinguistic tasks are sometimes dispreferred

Measuring Activity in the Brain

• In Magnetic Resonance Imaging (MRI), brain activity is investigated by scanning the brain every 1 to 5 seconds with a set of powerful magnets.

• Advantage: MRI studies generally provide very good spatial information.

• Disadvantages: 1) Relatively slow, i.e., inappropriate for understanding time course of language processing; 2) Participants cannot move, so tasks are somewhat restricted.

• Different types of MRI reveal different information about the structure of the brain.

  • Example, fMRI (“functional” MRI): reveals areas of increased blood flow in the brain, which can be related to increased brain activity because active nerve cells consume oxygen, in turn increasing the blood flow to that region

• Electroencephalography (EEG) and magnetoencephalography (MEG) detect changes in the electrical or magnetic field given off by the brain.

• Process: A stimulus is presented to the participant, and researchers look for differences between the responses to target and filler stimuli. Because the responses are a reaction to a particular event (the stimulus), the result is called an event-related potential, ERP

• Advantage: Detects changes in electrical or magnetic activity at the millisecond level, so can be used to study the time course of an event. MEG can provide high spatial resolution in addition to this temporal resolution, but it is extremely expensive

Tasks for Lexical Processing

• Lexical decision tasks

  • Procedure: A participant is asked to identify stimuli as words or nonwords, and the time that it takes the participant to make a decision is measured.

• Example findings: More-frequent words are recognized faster than less-frequent words.

• Naming tasks

  • Procedure: The participant responds to a stimulus by saying something aloud. 

    (Remember: Basket vs. Syringe in Lecture 11.)

• Example findings: More-frequent words are produced more quickly than less-frequent words

• Many tasks can be combined with techniques such as priming. In priming tasks, participants are presented with one stimulus, the prime, right before the stimulus of interest, the target. with the idea that having just accessed the prime will affect response times for accessing the target.

• Priming is often used to study the structure of the mental lexicon.

  • Example findings: Participants are faster to confirm that a stimulus is a word when the prime is semantically related to the target (i.e., confirming that nurse is a word when 

    the prime is the semantically related word doctor compared to butter).

    • → The mental lexicon is partially organized by semantic relatedness. The 

      prime doctor partially activated words semantically related to it, such that nurse was already partially activated when the target word appeared

Tasks for Lexical/Sentence Processing

• Eye-tracking: Method to study online lexical/sentence processing

• Motivation: Because we usually look at the objects we are thinking about, this method lets us study things like the timing of speech perception or how certain aspects of the linguistic input affect what we pay attention to.

  • Example study: Contrastive Inference

Tasks in Sentence Processing

• End-of-sentence comprehension task: Used to study globally ambiguous forms.

• Process: Participants read ambiguous strings of words and answer a comprehension question 

  after reading each one.

• Example: Participants read The cop saw the man with the binoculars and answer the comprehension question Who had the binoculars? The answer to this question can tell 

  us how participants interpreted the ambiguous form.

• Disadvantage: Cannot tell us what happened during the actual reading and processing. It’s possible that a participant initially used parsing strategies such as late closure while reading, but later decided that cops are more likely to have binoculars and therefore changed his 

interpretation

• Self-paced reading: For measuring online comprehension and can be used to study temporarily ambiguous forms.

• Process: Participants read a string of words in small chunks, usually one word at a time. Whenever they have read and understood a word, they push a button to move on to the next word.

  • Example: Someone shot the servants of the actress who was standing on the balcony. Someone shot the servants of the actress who were standing on the balcony.

  • Temporary ambiguity until after who

    • → The choice of was or were disambiguates the form: it tells us what who modifies because the rules about verb agreement say that was must go with actress and were must go with servants

• Measure the time it takes a participant to read each word (the time between button pushes).

  • If participants thought that who modified servants, they should take longer to read was than were

  • If participants thought that who modified actress, they should take longer to read were than was

• The reason is that participants would have to change their analysis if their initial interpretation turned out to be incorrect, and changing the analysis takes time.

  • → Unlike the end-of-sentence comprehension task, self-paced reading allows us to see what happens during processing

Insights from Ambiguities: Lexical Ambiguity

• The judge gave the criminal a very long sentence, but he parsed it just fine.

• The same phonological string [sɛntɛns] has two possible meanings, and we must figure out 

  which one is applicable.

• Linguistic context helps here! While context of a judge and a criminal to push your mind toward the ‘prison term’ meaning of sentence. But research reveals that, for a split second, we tend to consider all of the possible meanings

• Insights

  • Frequency matters (Lecture 11): When the frequencies are roughly the same, both meanings are considered, otherwise the more frequent meaning is preferred.

• Frequency interacts with context: We might ignore less frequent meanings if context biases us against them.

• Neurolinguistic studies suggest that the left hemisphere settles on the intended meaning pretty quickly (~200 ms), while the right hemisphere keeps considering both 

  meanings for up to a full second

Insights from Ambiguities: Is Syntax Independent From Semantics?

• In sentence comprehension, the process of assigning a syntactic structure to words is called

  parsing

• A key debate in sentence processing: How much is parsing based solely on the syntactic

  properties of the words or is additionally influenced by semantic properties of words?

• Evidence for the initial computation of a sentence structure comes from garden-path sentences. The fact that there is an ambiguity suggests that not all possible sentence constructions are considered (consistent with a structure-driven parse)

• Garden-path sentences → Evidence for a structure-driven parse. The fireman told the man that he had risked his life for to install a smoke detector

• However, in some instances, semantics can actually mediate the garden-path effect. For example by introducing linguistic context:

• A fireman braved a dangerous fire in a hotel. He rescued one of the guests at great danger to himself. A crowd of men gathered around him. The fireman told the man that he had risked his life for to install a smoke detector. (Altmann et al., 1994)

• On balance, it seems that the setting up of a sentence structure is, to some degree, dependent on both syntactic and contextual factors

• Whether you perceive something as semantically vs. syntactically incoherent can be tracked

  in your brain!

• Event-related potentials (ERP) analysis refers to certain patterns of electrical or magnetic activity, usually a positive or a negative peak.

  • N400: A negative peak around 400 ms after parsing something that's semantically odd, e.g., #He spread the warm bread with socks

  • P600: A positive peak around 600 ms after facing a garden-path or a very

    ungrammatical choice, e.g., agreement violations like She loves himself and I are smart

Maze Task

• Goal: Measuring how quickly (and without errors) people can process and navigate through a sentence.

• Setup: Two words are presented on the screen. One is the correct continuation of the sentence (the target word), the other one is incorrect (the distractor word).

• Questions:

  • 1. What do you predict we'll see? (What are your hypotheses?)

  • 2. What do you think are the advantages over a self-paced reading task?

  • 3. Why did I choose to show the completed parts of the sentence on screen?

  • 4. Were there filler stimuli? Why/why not?

  • 5. Why did I ask you in the study to place your left index finger on a left key and your right index finger on a right key?

• 6. Do you think it might matter whether the key was on the left vs the right?