21 - Word Meaning 1

Question 1

How do infants become more perceptive and making distinctions in their own language?

Answer 1

Distributional information (statistical learning)

Question 2

Why can’t adults find distinctions in other languages?

Answer 2

Because they have developed a categorical perception of phonemes through statistical learning

Question 3

Does word comprehension or word production come first in infants aged 1-2 years?

Answer 3

Comprehension

Question 4

Do infants aged 1-2 years have a sharper increase with their comprehension or word production?

Answer 4

Word production occurs rapidly towards the end of first year, comprehension is more steady

Question 5

How many words do adults know on average?

Answer 5

More than 60,000

Question 6

How many words do children learn on average each day at 12-16 months?

How many at 8-10 years?

Answer 6

12-16 months = .3 words per day

8-10 years =- 12.1 words per day

Question 7

Adults tend to stop learning words at a certain age, what are the two arguments put forth for why?

Answer 7

The critical period argument, we are no longer able

The environment: we know most words in our immediate environment

Question 8

What was the Werker et al. (1998) infant word learning experiment (what was the method)?

Answer 8

Infant are able to associate words with objects in the lab

8, 12, and 14-months olds trained with novel word-object pairings

Repeated presentations

Tested with habituation

• Correct word/object pairings given until habituation
• Incorrect word/ object pairings then presented

Question 9

What were the results of the Werker et al. (1998) infant word learning experiment?

Answer 9

All the infants looked at the incorrect word/object pairings

When there were 2 word/object pairings only 14-month olds could do this for both

Younger groups could only do this for one or the other

Question 10

Difficulty to map onto a label with what is referring to in the world when there are multiple objects present in the field of view is called…

Answer 10

Referential Uncertainty

Question 11

It is hard to determine whether infants understand what we think they understand because of

Answer 11

Referential Uncertainty

Question 12

Infants will often hear unique words in a sentence rather than in isolation, why is this a challenge for learning?

Answer 12

Word segmentation and referential uncertainty makes it hard for infants to understand what the word is referring to

Question 13

How do we learn word meanings?

Answer 13

Through a combination of statistical learning and a bias towards communication

Question 14

What are the 5 suggested mechanisms for how we learn word meanings in the real world?

Answer 14

Fast mapping

Cross-situational word learning

Zipfian distributions

Mutual Exclusivity

Communicative intent

Question 15

What is fast mapping?

Answer 15

The fact that biologically prone to quickly refer words to objects and remember them well

Question 16

Describe the Markson and Bloom (1997) experimental method which they used to show ‘fast mapping’?

Answer 16

Experiment with 2 phases

Training phase:

Playing measuring games with 10 objects

3 conditions for 3 objects. Casually mention ONCE ONLY:

“This is a Koba. Let’s put the Koba here.”

“This is from my uncle. Let’s put it here.”

“Let’s put this sticker here [on object].”

Test phase:

ONE MONTH LATER:

10 items, pick out the referenced item

Question 17

What were the results of the Markson and Bloom (1997) fast mapping experiment?

Answer 17

Children could learn the item based on its name, but they were even stronger at identifying the social aspect item (that it was given by an uncle). This is not language-specific but social connections is another valuable learning tool.

Sticker was not helpful in remembering, only language and social connection

Question 18

What is ‘cross-situational word learning’?

Answer 18

Learning a word by seeing it across different situations

Aggregating information after hearing it be used over multiple scenarios

- e.g. hear “ball and bat” infant is unsure what is ball and what is bat. However, if they later hear “ball and dog” they will be able to aggregate the info the determine which one is the ball

Question 19

What is this an example of?

A bottle of tesguïno is on the table.

Everybody likes tesguïno.

Tesguïno makes you drunk.

We make tesguïno out of corn

Answer 19

Cross-situational word learning.

We have learned that Tesguino is an alcoholic drink without it ever being described to us.

Question 20

12-14 month infants have been shown to be able to learn new words through cross-situational word learning.

How was this shown?

Answer 20

By pairing artificial words with different shapes

Question 21

What are ‘Zipfian distributions’?

Answer 21

Words are not distributed evenly but are skewed.

Some words are very common whereas other words are more rare.

Low-frequency words are hard to learn because they are not heard often

High-frequency words might help

Question 22

How has it been suggested that we use zipfian distributions to learn words?

How was it shown?

Answer 22

We use ‘high-frequency words’ to learn ‘low-frequency words’

Adults learned new words at a high or low frequency

They had to match 4 objects to 4 certain words

They were better at matching objects when high-frequency words were included rather than just low-frequency words

I.e. we know two words so we only need to figure out the last two word-object pairings

Question 23

What is ‘mutual exclusivity’?

Answer 23

We have a bias that states that one word is equal to one object

(one to one mapping bias)

Question 24

How does ‘mutual exclusivity’ lead to statistical learning?

Answer 24

If there are two objects and we are asked (which one is Dax) and we know one is a carrot, we know that the other one must be Dax

Question 25

What is the **‘one-to-one mapping bias’**?

Answer 25

That we believe that each name refers to only one object

Question 26

Are object-word relationships mutually exclusive?

Answer 26

No, there are multiple different words we can use to portray an object

Question 27

What have experiments shown in regards to our performance when 1-to-1 mapping is available?

Answer 27

**We are better at doing 1-to-1 mapping** We can do it without 1-to-1 mapping, but it is harder This means that it is important to our learning but it is not a hard and fast rule

Question 28

What is **‘communicative intent’**?

Answer 28

Infants and adults are sensitive to communication and pick up on social cues. They primarily learn things through watching others or in reference to others.

Question 29

What are three experimental results that showed ‘communicative intent’?

Answer 29

**Only learn labels if the speaker is confident** *(Sabbagh & Baldwin, 2001)* **Only learn labels in person, not through a voice recording** *(Fulkerson and Haaf, 2003)* **Only learn labels when the speaker is looking at the object** *(Baldwin et al., 1996)*

Question 30

What explanation do theorists give for ‘communicative intent’?

Answer 30

Evolutionary bias. Children learn through communicative context and form bonds with adults. They learn what to learn and what to avoid.

Question 31

What was the Frank, Goodman and Tenenbaum (2009) experiment regarding communicative intent? What were the results?

Answer 31

Tracked how often the words a parent said co-occurred with that object being in view This was done with and without the parents communicative intent (intentionally labelling object) The best model for each was: Co-occurence – 11% accuracy Co-occurrence with intent – 55% accuracy

Question 32

What is **generalising word meanings**?

Answer 32

When words refer to more than one thing (e.g. mammal, fish - larger categories)

Question 33

How did **Landau, Smith and Jones (1992)** test (methodology) how we generalise word meanings?

Answer 33

Tested on both adults and children The stimuli used were physical objects They presented the ‘dax’ then other objects when the texture, size, or the shape was different (all other properties were the same). Tested on whether they believed it ‘is a dax’

Question 34

What were the results of the **Landau, Smith and Jones (1992)** generalising word meanings experiment?

Answer 34

Most adults and children believe that it is still the same word if the size or texture changes However, if the shape changes less children and very few adults believe it is the same word *This is called **shape bias** and it becomes more strong over time*

Question 35

What is **shape bias**?

Answer 35

The tendency to view things of a different shape as outside of a category

Question 36

What are the three degrees of which we generalise word meanings?

Answer 36

Superordinate Basic Subordinate

Question 37

What does this experiment on generalising word meanings show us?

Answer 37

**How we generalise words is sensitive to the context we have learnt them.** * That if we use one-word, we are likely to generalise to the ‘basic level’ (i.e. might select a different dog if shown a dalmation)* * 3 sub examples we wont generalise to basic level* * 3 basic example we are not likely to generalise to the superordinate level* * If we are given 3 items from the super level they will generalise to the other category hierarchies.*

Question 38

People generalise words more tightly when there are _____ \_\_\_\_\_\_ available.

Answer 38

More examples

Question 39

Children are more likely to generalise to the superordinate level than adults - true or false

Answer 39

False, both aren’t overly likely to generalise to the superordinate level, kids less so

Question 40

Why are people less likely to want to generalise to the superordinate level?

Answer 40

Because of **‘communicative intent’** * The superordinate level is too wide to be useful typically - subordinate level is most specific and is, therefore, most used* * The fact that someone is communicating a label to us gives us context of where to apply it* * People are very sensitive to communicative intent and not just tracking co-occurrences*

Question 41

If experiments pick training items that are similar in some way and label them something, participants are more likely to apply to label only to the subordinate level. If participants pick the training items they are likely to generalise to the basic level. Why is this?

Answer 41

**Communicative intent and distribution matters** If experimenter chooses the item, it suggests that there is meaning in those items and that they should be strict with their definition as these cover the full extent of the variation in items that get the label. If participants choose items randomly from a selection they believe that these examples are less likely to cover the full extent of the label and they can be more broad.

Question 42