Langston
Psycholinguistics
Lecture 6
What is a Word?
A word is… (write your definition).
From Pinker: Two approaches…
What is a Word?
A syntactic atom: A unit that can’t be
divided further by syntactic rules. A word
may be a product of rules, but it is an atom
from the perspective of syntax.
Electric (root)
Shoes (shoe + plural)
Crunchable (capable of crunching)
Toothbrush (compund)
Yugoslavia report
What is a Word?
A rote-memorized chunk of “linguistic stuff”
paired with an arbitrary meaning. The
elements are listemes (entries in your
mental dictionary; any element whose
meaning and form have to be associated).
How Many Words?
High school graduate: 45,000.
Pinker: We’re not playing Scrabble.
Counting proper names, foreign words, etc.
60,000.
How do you learn all of that? That is 10 words a
day, every day, from your first birthday. It’s a lot
for a totally arbitrary pairing.
How do you store and access all of that
information?
Word Formation
Make words out of smaller elements the
way sentences are made out of words.
The wug test: “This is a wug. Now there
are two of them, now there are two
_____.”
Word Formation
The wug test:
Word Formation
The wug test:
If kids can answer this question, it must be a
rule (add -s).
Word Grammar
Elements + rules?
N -> Nstem + Ninflection (a noun is a
noun stem plus a noun inflection).
Dogs -> Dog + -s
Nstem -> Nstem + Nstem
Toothbrush-holder fastener box
Word Grammar
Elements + rules?
Nstem -> Nroot + Nrootaffix (some
morphemes go with roots, some with
stems).
Darwinian, Darwinianism, Darwinianisms.
Darwinism, Darwinismian.
Word Grammar
Inflectional morphology: Inflect the
meanings of words. Change the
meaning.
Two noun forms: duck, ducks.
Four verb forms: quack, quacks, quacked,
quacking.
English not rich in this.
Word Grammar
Derivational morphology: The meaning
can be derived from the bits.
English offers a lot more to choose from
here.
Word Grammar
-able -ate -ify -ize
-age -ed -ion -ly
-al -en -ish -ment
-an -er -ism -ness
-ant -ful -ist -ory
-ance -hood -ity -ous
-ary -ic -ive -y
Word Grammar
What is the longest word? Pinker (2000)
says that this is a meaningless question.
Floccinaucinihilipilification
Floccinaucinihilipilificational
Floccinaucinihilipilificationalize
Floccinaucinihilipilificationalization
Meaning
Once we know what a word is we still
have a big problem: What does a word
mean? There is an arbitrary association
between form and meaning, so how do
you make those connections?
Meaning
We discussed Aslin, Saffran, & Newport
(1998) as an example of how word
boundaries can be detected in speech.
Now, the gavagai problem. A rabbit runs
by and someone says “gavagai.” Do
they mean rabbit, furry, tail, running?
Meaning
Pinker (2000) suggests two “legs up”
that help with this problem:
A predisposition to chop the world into
individuals, classes, and actions.
Categorization:
○ Superordinate.
○ Basic.
○ Subordinate.
Meaning
Categorization: You can tell basic level
because they have the most feature
overlap.
What features (unique to furniture) do all
members of the category furniture have in
common?
What features do all chairs have in common?
What features do all desk chairs have in
common?
The longest list should be at the basic
level.
Meaning
Pinker (2000) claims that adults and
children both operate at the basic level.
Rabbit runs by, adult more likely to say “rabbit”
than “animal” or “Beveren.”
Kids also seem to expect this.
○ Dax task.
○ Show tongs, call them dax, ask for more dax, they pick
a different set of tongs (assume the word is the basic
level category).
○ Show cup, call it dax (they know the word cup), they
assume dax refers to what the cup is made of.
Meaning
http://www.straightdope.com/columns/read/49
1/whats-the-origin-of-kangaroo-court (3/30/11)
I never imagined that the day would come when I would spot an
error in your witty and admirably researched column, but your
recent discussion of the etymology of kangaroo, alas, shows
you aren't up to date on the research in this area. In the Guugu
Yimidhirr language, spoken by the aboriginals of the area where
Captain Cook's party recorded the term kangooroo (the original
spelling), this word (more accurately pronounced something like
kang-ooroo) refers to a particular species of kangaroo, namely
the large black kangaroo. The only error Cook's party can be
accused of is mistaking the name of one variety of kangaroo for
the generic term. I hope you will be able to bring your readers
up to date on this question and disillusion them regarding the
widespread mythology surrounding it.
Meaning
Four levels of meaning:
Referential: What the word refers to (the
actual thing). If I say “the book is too long”
the referential meaning is the particular book
we are discussing.
Denotative: The generic concept that
underlies the word. There is a lot of stuff you
know about book besides some particular
book.
Meaning
Four levels of meaning:
Denotative: How organized? Features…
Feature Man Woman
Living + +
Animal + +
Mammal + +
Human + +
Female - +
Meaning
Four levels of meaning:
Denotative: How organized? Features…
One problem is that you need a lot of features.
(Fewer than the number of things classified,
hopefully.)
Also, some features seem like they need
features themselves (e.g., mammal).
Getting the right set that classifies everything
with the fewest possible is tough and somewhat
arbitrary (calling them transducible helps a little).
Meaning
Four levels of meaning:
Denotative: Add boy…
Feature Man Woman Boy
Living + + +
Animal + + +
Mammal + + +
Human + + +
Female - + -
Meaning
Four levels of meaning:
Denotative: Add boy…
We’d prefer not to add a feature for every new
concept.
How would you add computer to that table? It
seems like features vary in “importance” to a
particular concept, how is that captured?
There are other ways of organizing denotative
information that get around some of these
issues.
Meaning
Four levels of meaning:
Associative: What you think of when you hear
the word (other words associated with it).
Origins:
○ Common expressions (“coffee, tea, or milk”).
○ Experience (we usually see tables and chairs together).
○ Antonyms (good-bad).
○ Units (ding-dong).
Meaning
Four levels of meaning:
Associative:
Assessing: Count concepts in common, the
more things they share the higher the
associative similarity.
Producing associations:
○ Network models: Assume distance in semantic space is
meaningful. Two things that are associatively related
are closer in semantic space.
○ Feature flipping: Find associates by flipping features.
Meaning
Four levels of meaning:
Affective: How a word makes you feel. We
will come to this a bit later…
A fifth level: What to do with
embodiment? If understanding a word or
sentence involves a motor component,
is that meaning? What about images,
metaphors, and spatial relationships
(iconicity)?
Meaning
How do you access meaning? The lexicon
must be organized in such a way as to
have direct access.
Tip of the tongue state (TOT; Frick-Horbury
& Guttentag, 1998):
An arch or hoop in in croquet that the balls have
to be hit through.
A frame or latticework for climbing plants.
A black cutout of paper to represent the outline
of a person’s head.
Meaning
The lexicon needs at least this information:
Vision: Appearance of word.
○ Words.
○ Partial words (a--i--in).
Vision: Access meaning from appearance of
object.
Audition:
○ Words.
○ Partial words (phonemic restoration).
Audition: The sounds things make.
Meaning
The lexicon needs at least this information:
Touch.
Smell.
Taste.
Things that affect access:
Frequency.
Morphology.
Syntactic category.
Priming.
Ambiguity.
Whatever model we come up with needs to know
this.
Lexical Organization
Network models (e.g., Collins & Loftus,
1975): Arrange information into a
network.
Lexical Organization
Collins and Loftus (1975):
Spreading activation; decreasing gradient.
The longer you process a concept the longer it
sends activation.
Activation decreases over time.
“Intersection” has a threshold for firing.
Organize network around semantic similarity.
Link network to lexicon with phonemic and
orthographic information in it.
Lexical Organization
Collins and Loftus (1975):
These kinds of models can account for
categorization phenomena and a lot of other
data.
Plausible?
Lexical Organization
Landauer and Dumais (1997): Latent
semantic analysis.
“A typical American seventh grader knows the
meaning of 10-15 words today that she did not
know yesterday. She must have acquired them
as a result of reading because (a) the majority of
English words are used only in print, (b) she
already knew well almost all the words she
would have encountered in speech, and (c) she
learned less than one word by direct instruction”
(p. 211).
Lexical Organization
Landauer and Dumais (1997):
That’s the problem. But, it gets harder:
“Studies of children reading grade-school text
find that about one word in every 20 paragraphs
goes from wrong to right on a vocabulary test.
The typical seventh grader would have read less
than 50 paragraphs since yesterday, from which
she should have learned less than three new
words. Apparently, she mastered the meanings
of many words that she did not encounter” (p.
211).
Lexical Organization
Landauer and Dumais (1997):
Problem: “how people acquire as much
knowledge as they do on the basis of as
little information as they get” (p. 212).
Solve this problem with “a high-dimensional
linear associative model that embodies no
human knowledge beyond its general
learning mechanism” (p. 211).
Lexical Organization
Landauer and Dumais (1997):
Anti-instinctivist in the sense that they don’t
think Pinker’s approach actually explains
how it gets done (“it’s biology” is not an
answer).
Instead, it’s a constraint satisfaction
problem. Evaluate their hypothesis with a
model and compare it to people.
Lexical Organization
Landauer and Dumais (1997):
Create a matrix with rows representing event
types (e.g., words) and columns
representing contexts (e.g., paragraphs).
The numbers in the cells are the numbers of
times a word appears in a particular context.
Lexical Organization
Landauer and Dumais (1997):
Compress this matrix to an optimal
dimensionality.
This allows latent knowledge to emerge.
Lexical Organization
Landauer and Dumais (1997):
Evaluation:
○ Learn language from encyclopedia entries.
Answer Test of English as a Foreign
Language (TOEFL) items. Model 64.4%
correct, applicants 64.5%. “Closely mimicked
the behavior of a group of moderately
proficient English readers” (p. 220).
○ Etc.
Lexical Organization
Landauer and Dumais (1997):
Conclusions (from vocabulary simulations):
○ “LSA learns a great deal about word meaning
similarities from text” (p. 226).
○ “About three quarters of LSA’s word knowledge is the
result of indirect induction, the effect of exposure to text
not containing words used in the tests” (p. 226).
○ “There is enough information present in the language to
which human learners are exposed to allow them to
acquire the knowledge they exhibit on multiple-choice
vocabulary tests” (p. 226).
Lexical Organization
Landauer and Dumais (1997):
Note: Did not use spoken language,
morphology, syntax, logic, or perceptual world
knowledge.
Not claiming this is what people do, but it does
show how much information is there.
I guess our question could be: Is there a need
for more than this, or is the information all there?
Symbol grounding problem (contrast to
embodiment; Glenberg & Gallese, in press).
Taboo
FCUK
Taboo
Jay (2009; doi:10.1111/j.1745-
6924.2009.01115.x):
“What are taboo words and why do they exist?
What motivates people to use taboo words?
How often do people say taboo words, and who
says them?
What are the most frequently used taboo
words?” (p. 153)
Taboo
Jay (2009):
What are taboo words and why do they
exist?
“sanctioned or restricted on both institutional
and individual levels under the assumption
that some harm will occur if a taboo word is
spoken” (p. 153).
Taboo
Jay (2009):
What are taboo words and why do they
exist?
Proposes that aversive classical
conditioning gives taboo words their taboo.
Taboo boundaries are fuzzy, even when
legally defined.
Taboo
Jay (2009):
What are taboo words and why do they exist?
Taboo semantic range limited in English (sexual,
profane or blasphemous, scatalogical, some
animal names, ethnic-racial-gender slurs,
perceived psychological, physical, or social
deviations, ancestral allusions, slang).
Offensiveness determined by context.
Taboo
Jay (2009):
What motivates people to use taboo words?
“Swearing is like using the horn on your car” (p.
155).
Two thirds of swearing data linked to anger and
frustration.
“Can intenstify emotional communication to a
degree that nontaboo words cannot” (p. 155).
Taboo
Jay (2009):
How often do people say taboo words and who
says them?
Jay (1980) 0.7% of words. Other estimates
similar.
Also individual differences from 0% per day to
3.4% per day (me).
Average speaker 80-90 taboo words per day.
Affected by personality and social factors.
Taboo
Jay (2009):
What are the most frequently used taboo
words?
20 years of data.
70 types recorded, top 10 get 80% (f-ck, s-it,
hell, g-ddamn, Jesus C-rist, ass, oh my God,
b-tch, sucks). Fu-k and sh-t get 1/3 to 1/2.
Taboo
Jay (2009):
Pinker asserted that swearing is not genuine
language.
Can we really afford to ignore emotion in
language? Or social factors?
Taboo
Taboo Stroop task (McKay, Shafto,
Taylor, Marian, Abrams, & Dyer, 2004):
Name the colors of words, including taboo
words. Taboo takes longer (first 100 trials,
neutral 704 ms, taboo 768 ms; surprise
recall neutral 27%, taboo 67%).
They are special.
Taboo
Maybe taboo also arises from magical
thinking (sympathetic magic).
Rozin, Millman, & Nemeroff (1986;
doi:10.1037/0022-3514.50.4.703):
Contact contagion: Roached juice.
Similarity: Sodium cyanide, baby darts.
Taboo
Rozin, Markwith, and McCauley (1994;
doi:10.1037/0021-843X.103.3.495):
Contact aspect of sympathetic magic:
People who have worn clothing somehow
contaminate it.
Measured willingness to wear a sweater
worn by various people.
Taboo
Previous user (no Effect
photo)
Man -24
Accident -33
Homosexual man (no AIDS) -47
Convicted murderer -62
Man with AIDS due to -63
transfusion
Homosexual man (with AIDS) -66
Man with tuberculosis -64
Taboo
Could sympathetic magic underlie some
aspects of taboo?
Look at some of the words.
In a way, contact with contexts could be
seen as “rubbing off on” the word.
Taboo
The first chapter of Kennedy (2002) (a
book whose main title is the N-word)
explains some of the reasons why the N-
word is so bad:
Slavery
Lynchings
Jim Crow
When a word derives its power from
contexts like that, can it be rehabilitated?