00:00:00
So let’s talk about signs. When you see
a new sign, you have to figure out what it
00:00:04
means if you want to follow it. The same thing
happens when you run into a word that you’ve
00:00:07
never seen before - you just have to work out what
it does. Every one of us takes up new words
00:00:12
all the time, from infancy onwards, but how
do we actually learn what those words mean?
00:00:16
I’m Moti Lieberman, and this is the
Ling Space.
00:00:33
So all of us have tons and tons of words in
our heads, in one language or across many.
00:00:38
We rarely think about our individual words
when we’re trying to say them - they’re just
00:00:42
there in our heads, ready for us to express
our thoughts. But at some point, each and
00:00:45
every one of those words had to be learned.
00:00:48
This situation is made more complicated by
the fact that the words that we use for things
00:00:51
don’t usually have anything to do with the
things themselves, except for onomatopoeia
00:00:55
like plop or achoo. Like, there’s nothing
particularly wall-like about the word “wall”,
00:00:59
or pie-ish about pies.
This concept is often called the Arbitrariness
00:01:03
of the Linguistic Sign, and the groundbreaking
linguist Ferdinand de Saussure came up with
00:01:08
it a century ago.
00:01:09
The sounds of words just aren’t predictable
based on the information that we have about the
00:01:13
world. That information needs to be stored
somewhere.
00:01:15
But where? Well, we can imagine that we have a kind of
mental dictionary in our brains, where we keep all
00:01:21
information about words.
We call these mental files a Lexical Entry,
00:01:25
and it contains everything you need to know about a word: how it's pronounced, where it goes, and
00:01:29
what it means. Let’s look at an example.
00:01:31
The Lexical Entry for the word “hound”
tells you that it’s a noun that's pronounced like [hawnd],
00:01:35
and it means “large domesticated canine”.
So for every language you speak, you fill up
00:01:40
your brain with tens of thousands of these
lexical entries. But how do you manage to
00:01:44
build that dictionary up? Learning words
is complicated! And in the real world, words
00:01:48
are flying at you like a flock of ravens.
00:01:51
The first thing you have to do is separate
that stream of language into
00:01:54
its component parts.
And that’s hard enough, but even when you do have
00:01:57
those little chunks, your job’s not over.
You still need to map those phonological forms onto
00:02:02
their meanings.
But if meanings aren’t predictable from
00:02:04
the sounds, then how do you even do that?
00:02:06
Here’s a concrete example, based on the
work of philosopher W.V.O. Quine. So let’s
00:02:11
say you just happen to be standing around
with a friend of yours, and a rabbit runs
00:02:14
by. Your friend points at the rabbit and says,
“Hodor!” What does hodor mean? Well, maybe
00:02:20
it means rabbit. That’s a decent assumption.
But if you think about it a minute, it could
00:02:23
also mean “run”, or “fast”, or “dinner”,
or “look over there”.
00:02:27
There’s a huge list of things that hodor
could mean. But you probably thought that
00:02:31
it meant “rabbit”, right?
Well, fortunately, that’s one of the basic
00:02:34
assumptions that pretty much everyone makes,
so it’s a natural guess.
00:02:38
Linguists have figured out four easy tricks that people use to associate arbitrary sounds and symbols
00:02:43
to meanings in the real world.
00:02:44
The first of these guidelines is known as the
Whole Object Assumption. Now, this says that every
00:02:49
new word that you encounter will probably refer
to the whole object, and not one of its parts
00:02:53
or attributes. So hodor should mean rabbit,
not just the rabbit’s foot, or fluffy.
00:02:58
We’ve also got the Type Assumption. This
means that a new word refers to a type of
00:03:03
thing, and not to any particular thing.
So your friend pointing and saying hodor probably
00:03:07
doesn’t mean, “Look, there’s my pet
rabbit, Hodor!”
00:03:10
So linguists ran an experiment on 4 and 5
year olds, to test the Type Assumption.
00:03:14
They showed them a picture of a cow, and said,
“This is a dax. Can you find another dax?”
00:03:19
And then they showed them two other pictures: one of a pig, and one of milk. Now cows and pigs go together
00:03:24
by type, like if “dax” meant “barnyard
animal” or something. Cows and milk aren’t
00:03:28
related by type - there’s a theme there,
but nothing else. So what did kids pick?
00:03:33
Well, those kids that were told that the cow was a dax picked the pig 65% of
00:03:37
the time. So they definitely
were preferring to interpret by type! But interestingly,
00:03:42
when they weren’t given a new label for things, and they just got shown the cow and then told something like,
00:03:47
“look at this, now can you find another one?”
Kids only picked the pigs a quarter of the time.
00:03:52
So they were really using that label
to help them group things together.
00:03:55
Another arrow we have in our word-learning
quiver is the Basic Level Assumption. The
00:03:59
idea here is that some categories are more
fundamental than others. So, a word like
00:04:03
“rabbit” is a basic-level category,
whereas words like “mammal” or “animal” are higher
00:04:07
level categories. We usually assume that a
new word will apply to a basic-level idea:
00:04:12
so hodor probably doesn’t mean “mammal”.
00:04:14
We’ve got one more weapon when it comes
to picking up new words.
00:04:18
Let’s say you managed to learn that hodor
really did mean rabbit. But the next day
00:04:22
you run into your friend again, and he’s holding his new
pet rabbit, and he says something different,
00:04:26
like… bodor. Now, will you think that bodor means
rabbit? Not if you follow the Mutual Exclusivity
00:04:32
Assumption.
This says that objects have just one label.
00:04:36
So maybe bodor means something like soft, or fluffy, or it could be the rabbit's name, but it won’t just mean rabbit.
00:04:42
So let's look at an example. Imagine
an experiment where kids get shown
00:04:46
two things - like, a tree,
and then a shiny, funky new object they’ve never
00:04:49
seen before. Then you ask them, “Show me the zib”.
They’re not going to point at the tree,
00:04:54
because they already know that trees are trees.
00:04:56
They nearly always assign the new label to
the new object, because they assume that terms
00:05:00
are mutually exclusive. This is so powerful
that it can actually override the other assumptions,
00:05:05
which we’ll talk about back on our website.
Okay. So that’s the big stuff. But there are
00:05:09
a few other things that we can use if those four
assumptions don’t give us all the information we need.
00:05:13
For example, we can use the information we already have about syntax and morphology for our language.
00:05:18
So, for example, we already know that -s usually gets stuck on nouns to make them plural,
00:05:24
or that common nouns can get preceded by determiners, like ‘a’ or ‘the’.
00:05:27
We’ve got some cool experiments about this,
too. So researchers showed some kids who were
00:05:32
about a year and a half old some dolls. And
the two dolls looked similar, but they only
00:05:36
talked about one of them, making the other doll really jealous. For some of the kids, they said “This is a zav”,
00:05:42
with the determiner stuck in there.
Other kids were just told “This is Zav”,
00:05:46
without the ‘a’. After a while of this,
researchers asked the kid to put some dress-up
00:05:50
clothes on the doll.
00:05:51
If they’d been hearing “This is Zav”,
then they were asked to dress Zav, and they only
00:05:55
went to dress that one particular doll. But
if they’d been hearing “a zav” the
00:05:59
whole time, then they were asked to dress “a
zav”, and then they would dress either that one
00:06:03
doll or the other one. Both of them were okay, because
they'd been using that little ‘a’ the whole time.
00:06:08
Little kids who weren’t even two and had
never studied any grammar already knew that proper
00:06:13
nouns refer to one individual and don’t
take ‘a’, whereas common nouns refer to a
00:06:17
type of thing, and do.
00:06:19
Before we wrap up, let’s go back to our
idea about lexical entries. We already said that they
00:06:23
have information about pronunciation and meaning
in them, as well as whether they’re a noun or a
00:06:28
verb or whatever. But now we can add some
more information. Lexical entries include
00:06:32
facts about what situations a word can show
up in. Like, is a morpheme free to stand
00:06:36
on its own, or does it have to get bound to
something in order to show up? Does it need to have a
00:06:41
determiner like ‘a’, or not? Stuff like that.
00:06:43
Oh, and then, you have to wonder about plurals.
Like, do you have two separate lexical entries
00:06:47
for “spear” and “spears”? Or do you
have just one?
00:06:50
Well, the kind of plurals that you make just
by adding an -s use a rule that people store in
00:06:54
their heads, even little kids. Remember the
Wug Test, that we talked about back in episode 21?
00:06:58
You show a kid a drawing of some unfamiliar
00:07:01
critter and tell them it’s a wug. Then, you show them two of them, and ask them what they're called.
00:07:06
Kids will usually say two wugs, with the -s,
even though they never heard the word wug
00:07:11
before. So this means that plurals, along with a whole bunch of other things, are stored separately in our
00:07:16
brains from the mental dictionary. We don’t
need multiple lexical entries for spear and
00:07:20
spears, or for dragon and dragons. We only have
the one, and our rules take care of the rest.
00:07:25
So even if Hodor is causing you some problems,
you can use these assumptions to work
00:07:29
out what your friend is trying to get across.
You build your mental dictionary using these
00:07:33
strategies, and you solve the problem of what
meanings go where without ever even really thinking
00:07:37
about it much. That’s just another sign
of how well our minds are built for language.
00:07:42
So we’ve reached the end of the Ling Space
for this week.
00:07:45
If you assigned the right meanings to my sounds,
you learned that what words refer to
00:07:49
isn’t as obvious as you might think;
that we construct our mental dictionaries
00:07:53
out of lexical entries with information about
pronunciation, usage, and meaning; that we
00:07:57
use four assumptions to help us guide our task: Whole
Object, Type, Basic Level, and Mutual Exclusivity;
00:08:04
and that things that are driven by rules sit
outside the lexical entries for the words
00:08:07
themselves.
00:08:08
The Ling Space is produced by me, Moti Lieberman.
It’s directed by Adèle-Elise Prévost,
00:08:13
and it’s written by both of us. Our editor
is Georges Coulombe, our production assistant
00:08:17
is Stephan Hurtubise, our music and sound
design is by Shane Turner, and our graphics
00:08:21
team is atelierMUSE.
00:08:22
We’re down in the comments below, or you
can bring the discussion back over to our
00:08:26
website, where we have some extra material
on this topic. Check us out on Tumblr, Twitter
00:08:30
and Facebook, and if you want to keep expanding
your own personal Ling Space, please subscribe.
00:08:35
And we’ll see you next Wednesday. Hodor!
