Speech Perception & Sentence Comprehension Flashcards

1
Q

Which of these is incorrect? There can be more than one answer.

Prosody is a general term that refers to the aspects of an utterance not specific to words themselves. It does several things:
A) Distinguishes words
C) Determines the semantic meaning of a word
B) Signals emotional content
D) Allows for sentence parting

A

Incorrect:
C) Determines the semantic meaning of a word
D) Allows for sentence parting

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

——— is the study of speech sounds. It studies place and manner of articulation as well as voicing.

A

Phonetics

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What are some challenges our brain has to overcome when it comes to spoken word recognition?

A

There are quite a few challenges it must overcome, the most notable is that there is no 1:1 relationship between a certain waveform and a phoneme. Additionally, there is great speaker variability (gender, dialect, age, …) and there is also the cocktail party effect - how do we distinguish background noise, speech, … from one another?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Researchers often used sound spectrograms in audio signal processing. Series of dark bands on a sound spectogram are called ———-. The first one signals the lowest frequency. The darker it is, the more ——– it is. It also varies by speaker.

A

Researchers often used sound spectrograms in audio signal processing. Series of dark bands on a sound spectogram are called formants. The first one signals the lowest frequency. The darker it is, the more intense it is. It also varies by speaker.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Choose one or more correct answers.

Which aspects of formants are important for word recognition?
A) Frequency
B) Steady state
C) Transition
D) Reverberation

A

B) Steady state
C) Transition

Transition: movement of formants. They occur either at the beginning or the end of a syllable.
Steady state: the period between the transition
→ Transitions correspond to the consonantal portion of the syllable and the steady state to the vowel
Seek → see the difference between the duration of ‘ee’ outside of a context and how within a context like ‘seek’, the duration of ‘ee’ is much shorter. ‘Ee’ doesn’t show the same steady state (flat) formant patterns

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Which of these is not correct?

There are three levels of word recognition processing:
A) Phonetic
B) Semantic
C) Phonological
D) Auditory

A

B) Semantic

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Fill in the blanks.

There are two mechanisms for perceiving sound. The ————- perception is found most strongly with consonants, while the —————— perception is more commonly used for vowels. This is because with consonants, ——–.

A

There are two mechanisms for perceiving sound. The categorical perception is found most strongly with consonants, while the gradual perception is more commonly used for vowels. This is because with consonants listeners are unable to perceive within category differences while with vowels, the change is more gradual

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Fill in.

The ————- is the point at which voal cords begin vibrating.

A

voice onset time (VOT)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is the advantage of having a categorical perception of speech sounds?

A

It allows us to ignore irrelevant information and be better at speaker variation (distinguishing between speakers).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

We can manipulate the voice onset time in categorization tasks. Researchers often do this with /ta/ and /da/. For example, they play /da/ and gradually move to /ta/ and the participants have to decide what sound they are hearing. For about 30ms, there is an uncertainty point. What is this uncertainty point? What does it show?

A

Experimenters manipulate the voice onset time, which is when the vocal cords vibrate. Vocal cords vibrate with voiced sounds, like /da/ but not with voiceless like /ta/. This uncertainty point is when participants have about a 50/50 response as to wheter they are hearing da or ta, but otherwise they are very accurate at deciding if they’re hearing ta or da. This is because the way we recognise consonants is categorical and this uncertainity point signifies the shift from one category to another.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What is the lack of invariants problem?

A

The lack of invariants problem is that there is no 1:1 mapping between phonemes of language and their acoustic instantations. A certain phoneme may have a number of different wave forms but our brains are really good at distinghuishing one phoneme from another regardless of this.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Due to the lack of invariants problem, two theories cropped up to try and solve it. Name the theories, who proposed them and compare them - how do they compare, what are some strong points, what are some problems, …?

A

The theories are the auditory theory by Stevens and the motor theory by Liberman. Stevens states that the invariance is normalized by our systems in the auditory step. Liberman’s theory, on the other hand, states that there is no variability in the way we move our mouth and how we use our vocal tract. As such, listening is actually a process of subvocalisation (the listener silently articulates the sound s/he is hearing). The link between articulation and perception is much more direct than the link between the acoustic structure and perception. Steven’s theory is a more general model, which can be applied to all sensory modalities and focuses on perceptual normalization while Liberman’s is more language specific. Additionally, only Liberman’s motor theory of speech adresses the McGurk effect (listeners see ga, hear ba, report da) as it is the only theory in which our motor movement is relevant.

While Liberman’s theory circumvents the problem of finding acoustic invariants for sounds, minimizes the mechanisms used in production and perception and explains the McGurk effect, it has some problems. The first is allophonic variations (bad at discriminating allophones but good at producing distinguished ones) and also, infants cannot produce much but can perceive a lot of contrasts. Steven’s theory also has some problems because it does not adequately adress acoustic variability because it is more than just a power law transformation of intensity.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Fill in.

Syntactic parsing obeys the —————– principle, which states that me make decisions immediately as we encounter each word.

A

Syntactic parsing obeys the immediacy principle, which states that me make decisions immediately as we encounter each word.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Answer and explain.

Linguistic elements have two syntactic levels. What are the levels?

A

Linguistic elements have:
* a deep level (syntax)
* surface level (what we see/hear)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Describe incremental structure building

A

We process sentences bit by bit and during that, we build a structure. The first two things we hear form a constituent - the background against which ou evaluate new information. We have a constituent, ‘the actor’, then you get a new one, the first gets destroyed and a new, bigger one is formed. Because whenever we encounter things in language, we have to understand the thematic roles. When we speak, language comprehension is exactly this, ‘who does what to who’. This is repeated until the sentence is spoken.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What are garden path sentences? What do they show us? Give an example and how we would process it (use incremental structure building)?

A

Garden path sentences are sentences that ‘’lead you up the garden path’’. These are sentences in which you build a structure, and interpretation and then there is a new element that destroys the interpretation so you must go back and re-analyze.
Garden path sentences show us that our processor is very impatient and processes things immediately (it doesn’t wait until the end of a sentence)
Example:
The old man the boat
First you see ‘the old man’ and presume it’s the sentence subject. Then you read on and oh-uh, there is no verb?!! Which is when you destroy your initial structure and go back and see aaaa, man is actually the verb and now everything makes sense again.

17
Q

How do we process this sentence? What kind of a sentence is it?
The horse raced past the barn fell.

A

In order for us to understand that the main verb is not ‘raced’, we have to come to the last word, ‘fell’. You realise that you mis-analyized (horse as subject, raced as vp, past the barn a), soyou go back and realize that ‘raced’ is not the main verb but the verb of a reduced relative clause (the horse that raced past the barn fell)

18
Q

Compare the two sentences:
1) Enraged cow injures farmer with ax
2) The girl told the story cried.

A

The first is structurally ambigous while the other is a true garden path sentence. The difference between the two is that with garden path sentences, you only have a single way of understanding it (syntactically and semantically) while in structurally ambigous sentences, you can interpret it either way syntactically and sometimes semantically (though here, there is only one semantically sound interpretation).

19
Q

The late closure parsing strategy states that…

A

whenever possible, we prefer to attach new items to the current constituents.
Since Jay always jogs a mile it seems like a short distance to him. → when you arrive to ‘seems’, you analyse this phrase ‘a mile’ not as the object of jogs but as the object of ‘seems’. But why do we do garden path analysis? Because of late closure (but this is not universal to all languages)

20
Q

There are different parsing strategies, like late closure, ————– attachment, the —————- approach and the constraint based model.

A

There are different parsing strategies, like late closure, the minimal attachment and the modular approach and the constraint based model.

21
Q

What does the minimal attachment approach to language parsing state?

A

That we prefer to use as few nodes as possible (a).
The boy saw the man with the telescope. The way we analyse depends if we have the prepositional phrase to ‘the man’ or if we attach it to the boy. (minimal attachment would say that the boy saw a man that had a telescope because it has one node fewer than the other interpretation)

22
Q

Explain the constraint based model of parsing.

A

Constraint-based model: we immediately use all information available, both syntactic and contextual.

23
Q

How would you test the contraint based model of parsing?

A

Having two garden path sentences in which analysis could be facilitated by context. For example:
The florist sent the flowers was very pleased.
The performer sent the flowers was very pleased.

The first one → much higher reaction times than the second one, we are more aware that performers receive flowers; this shows the big garden path effect (influence from higher levels, from context, word knowledge; the fact that we have different noun phrases influences garden path)

24
Q

How does punctuation help comprehension?

A

Commas can serve the same purpose as prosodic boundaries. Since Jay jogs, a mile and a half seems like a short distance to him. Steinhauer (2003) also looked at effect of commas on ERP signal, he used the same sentences with or without commas.
Peter verspricht Anna zu arbeiten und das Buro zu putzen.
Peter verspricht , Anna zu arbeiten und das Buro zu putzen. → seen as ungrammatical
Peter verspricht, Anna zu enlasten und das Buro zu putzen. → is grammatical only with comma
Peter verspricht Anna zu entlasten und das Buro zu putzen. → not grammatical without comma