Ling 423/640G: Cognitive Linguistics

Ben Bergen

 

Assignment 2

Distributed: September 26, 2008

Due: October 2, 2008

 

Instructions:

Feel free to discuss this assignment with your classmates. But perform the statistical tests and write up your discussion notes and answers independently. Double-space all work and use a font size no smaller than 12.

 

Part I. Statistics

 

You will read about three hypothetical experiments and will be given imaginary data that could result from them. For each experiment, your tasks are:

(1)  identify the independent and dependent variables in the experiment

(2)  say whether these are categorical or continuous

(3)  say how many levels each categorical variable has

(4)  say whether each independent variable is manipulated between or within participants

(5)  select the correct statistical test to apply to these data

(6)  perform that statistical test, using one of the online statistics calculators referred to in the notes from Meeting 10,

(7)  report the results of that test, in the prescribed manner, and

(8)  say whether the measured effect was statistically significant (equal to or better than p=0.05) or not.

 

Hints:

á      For reporting results of t-tests, see here: http://www.ilstu.edu/~jhkahn/apastats.html

á      For regression, use this page (not the one from Meeting 10): http://www.wessa.net/esteq.wasp . Your two columns of data go in "Data X", and your column titles go in "Names of X Columns", then press "Compute". A new page will pop up. The options are all correct, so hit "Estimate" to get the results. Present your results as: R2=a; F(b,c)=d; p=e . The values that fill in a-e appear on the results page. First, find the box labeled "Multiple Linear Regression - Regression Statistics". The R-squared value in this box is a. Then go to "Multiple Linear Regression - Analysis of Variance", and take the regression degrees of freedom (regression DF) as b, total degrees of freedom as c, F-test value as d, and p-value as e. E.g., R2=0.33; F(1,29)=2.14; p=0.12.

á      When you run Fisher's Exact Test, report only the p-value, e.g. p=0.001.


 

Writeup

 

While there will be a lot for you to do behind the scenes, your answers will be very brief. For instance:

 

0.   Independent variables (2)

            Amount of Light (categorical; two levels; within participants)

            Amount of Water (categorical; two levels; between participants)

 

      Dependent variable (1)

            Plant height (continuous)

 

      Test: repeated-measures ANOVA

 

      Results: F(1,19)=4.90; p=0.03; this effect is significant.

 

 

Exp 1. Age of acquisition and speech rate

 

This experiment aimed to determine whether the earliest age at which a person began learning a second language affects the rate of their speech. The prediction is that the earlier a person began learning the language, the faster they will talk. In order to test this, 20 participants who had started learning English after another language were brought in to the lab for interviews, and were asked ten questions about their current beliefs and activities, each of which elicited a long response. Their responses were recorded and coded for speech rate, in words per second. The age at which each participant started learning English is in one column below, and their overall speech rate is in the other column.

 

 

Exp 2. Producing metaphorical language

 

This experiment tested whether activating a source domain for a metaphor would make people more likely to produce metaphorical language using that source domain. For instance, people who are thinking about being inside containers might be more likely to talk about being in trouble, while people thinking about possessing objects might be more likely to talk about having trouble. To test this, the experimenters brought participants into the lab and either gave them several billiard balls to hold in their hands (meant to activate possession), or had them enter and sit down inside a very small recording booth (meant to activate containment). They then asked them to read two words and make a sentence using them: Mary and trouble. They recorded whether each participant used metaphorical language about containers (like Mary is in trouble, Mary is trying to get out of trouble, etc.) or metaphorical language about possession (like Mary is having trouble with her work, Mary is trying to get rid of her troubles, etc.). Below you see the results: for each participant, you have which condition they were exposed to – container or possession – and which type of sentence they produced – containment or possession.

 

Exp 3. Different types of metonymy

Words referring to a concept a can be used to identify b if there's a pragmatic relation between a and b. This experiment aimed to investigate whether different sorts of metonymy lead to different mental access to a. Specifically, it measured the extent to which intrinsic versus extrinsic metonymical expressions using language for a to describe b led language understanders to mentally imagine a. The prediction was that intrinsic metonymies, like handle for bottle of liquor with a handle would produce less detailed mental imagery of a – in this case, the handle - than extrinsic metonymies would, like hash browns for person who ordered hash browns. The experimenters created ten ordinary sentences using 5 extrinsic and 5 intrinsic metonymies. For each, they also selected a word that described some perceptual aspect of a, for instance salty for hash browns or curved for handle. Fifteen participants came into the lab, and were presented with all ten metonymical sentences. After each sentences, they saw the target word, and had to decide if it was a word of English or not. (In another ten trials, they saw filler sentences followed by non-words). Faster responses to the target words in the extrinsic condition would imply that participants had already activated a mental image of a, which primed access to words describing its perceptual properties. The average time (in milliseconds) it took each participant to respond to words following the five intrinsic and five extrinsic metonymies, respectively, are shown in the data.

 

Data

Exp 1

Age

Rate

13

1

18

1

4

1

15

1.2

14

1.2

5

1.1

5

1.3

6

1.2

18

1.2

7

1.1

9

0.9

5

1.2

12

1.2

12

1.5

5

1.3

13

0.3

13

0.8

16

1.3

18

1.4

12

0.5

 

Exp 2

Participant

Prime

Sentence

1

Container

Container

2

Possession

Container

3

Container

Possession

4

Possession

Possession

5

Container

Container

6

Possession

Possession

7

Container

Container

8

Possession

Possession

9

Container

Possession

10

Possession

Possession

11

Container

Container

12

Possession

Possession

13

Container

Container

14

Possession

Container

15

Container

Container

16

Possession

Possession

17

Container

Container

18

Possession

Possession

19

Container

Container

20

Possession

Possession

 

Exp 3

Participant

Intrinsic

Extrinsic

1

414

426

2

510

480

3

609

482

4

590

505

5

550

530

6

508

400

7

560

570

8

684

640

9

743

721

10

462

460

11

534

514

12

538

468

13

600

579

14

604

615

15

609

443