Radio Inspire

How To Learn Sign Language

CGSL Sign Language Identity Colloquium: Marie Coppola


We’re all here together on this great Friday afternoon. It is my pleasure to be hosting this joint colloquium for the Psych department and for the Linguistics department by my dear colleague and
friend, Marie Coppola, Assistant Professor at the
University of Connecticut. She’s just given one presentation and now she will be giving another. Just a brief introduction
about some of Marie’s kind of academic history. So, she and I met back in the ’90’s in ’94, so we’ve known each other quite
a number of years already. She has been one of the most prolific researchers in Nicaragua doing work on the emerging sign language in Nicaragua and the entire, sort of,
linguistic situation there starting with homesigners
and following the subsequent cohorts or generations of Nicaraguan sign. She has recently also
been awarded a prestigious NSF career award grant
for the next five years to be working with deaf children on math and numeration
skills and the relationship between language and numeration. And that work grew out
directly from her contact with the homesigners and
the emerging language of Nicaraguan Sign Language,
so it’s basic science becoming, sort of, useful
in a practical way. And so it’s kind of the
best of both worlds. So this afternoon, she’s
going to talk about unexpected routes to
language, evidence from child and adult homesign systems and I now give the floor over to Marie. – Thank you so much Diane,
for that really warm and lovely introduction, I
really appreciate the invitations to both events, thank
you all for coming late on a Friday afternoon. I would like to share some of
the research that I’ve done in Nicaragua with all of you
and I really would like this discussion to be a dialogue
and a way for people with different perspectives
on how language works and the language faculty to bring together our complementary areas of expertise and really try and get at
these fundamental issues of what language is about,
how it’s represented, what is the nature of
the language faculty. As Diane mentioned, I am going
to be talking quite a bit about my research with homesigners. Those of you who were at the panel earlier already know what I’m talking
about when I talk about a homesigner, some deaf
people may have recollections of their own experience using homesigns with their hearing family members and when people are thinking about that they’re thinking about sort of different vocabulary
signs that are different from the conventional ASL signs that they later learned
when they became part of the Deaf community and acquired ASL. When I’m talking about homesign, I’m talking about a whole gesture system that basically functions
like sign language for the deaf people I’ll
be talking about today. So homesigners are deaf people, they haven’t acquired a
conventional language, either a spoken one, a written
one, or a sign language. And they create gesture systems and they continue to use
these gesture systems throughout their lives as
their primary language, that is what they have. They’re socially integrated
into their communities, so I want to make sure
that everyone understands that these are not cases
of people, like Genie, the girl who was found in
her home in California after many, many years of neglect. She also did not acquire language, but she was not exposed to language. She was also not exposed to
typical social interactions and affection and love
and family relationships. Okay, so the people
I’m talking about today have all of those things, the
thing that they’re missing is specifically linguistic input. Okay, they get input, right,
they see things in the world, they’re interacting with
people all the time, every day, but those interactions don’t
contain linguistic information from the outside.
I’m gonna talk a lot about the linguistic information
that homesigners innovate and produce. And in fact, here are a few
examples of some of that structure, so some, we
have evidence for the grammatical relation of subject, plural morphology, and of course, the work done here at
University of Chicago by Susan Goldin-Meadow, Carolyn Mylander, and they’re my, all of us are colleagues, which all homesigns systems
have really formed the basis for this continuation of
what other kinds of devices and complexity might develop
when a homesign system continues to be used through adulthood. I wanna now show you an example of what a mature homesign system looks like. I’m sure almost everybody in this room has seen examples of child homesign. And adult homesign is different and I, myself was surprised when I began looking for adult
homesigners in Nicaragua. And I will let you just see for yourself, I’m gonna show this video once, just as it is and then
I’ll show it to you again with some help, okay, with some captions and slower so you can pick
up on some of the things you might have missed the first time. Is it possible to, I don’t
know how visible that is, it looks like there’s a lot of glare. Yes, great. So, this is me, this is the
homesigner I’m talking about, this is the same homesigner I mentioned in my panel presentation
who found out about the deaf community, was kinda like, meh, decided to just keep on with his homesign, he was happy with it. This is his mother. Okay, so one clue I didn’t tell you is that this video was
made in the spring of 2002, which is relevant for the topic, those of you who know what this is about, please restrain yourselves, okay, I’m gonna play it again, now. Does anyone feel like they, I don’t want you to answer, I
just want to get a sense of, does anybody think they
know anything about what he just said? Okay, anybody who is not a signer who knows anything
about what he just said? Alright. My point is made. Okay, this is not mime, okay it’s not getting up
and acting things out. Okay. So, he’s talking about some
pretty complicated things, right, this is not something
that he was present for, right? He learned about this event through life and he is now talking about it. I want to mention here, his
mother who he lives with, right, he’s in his late 20’s here, her comment after this conversation, which went on for quite some time, I’m just showing you a
very small snippet of it, said, “The Nicaraguan civil
war was really awful.” Indeed it was, but it is
not close to the topic that he is talking about here. So that, I’m gonna return to that point, that sort of lack of
successful communication using the homesign system
as part of the theoretical implications of the work that
I want to tell you about. Okay, here’s a brief schematic
that shows you the role of homesign systems in the emergence of Nicaraguan Sign Language. When a critical mass of
deaf children came together in the 1970’s at a center
for special education, that was the first time
that we see the beginnings of a deaf community
and a sign language in Nicaragua, prior to that time, there was no sign language
in the country, okay, there were very small groups of children, deaf children, who came together, they were not allowed to
sign, the conditions were not suitable for a sign language to emerge until the late ’70’s in this center. As new children entered the school and saw this rudimentary
sign communication from that first group, they passed, they took that in and they produced a language that was
more complex, and yeah, more complex. And that
process has repeated itself, it’s continuing to repeat
itself up until now. So, when these small houses
represent homesign systems of the first children and adolescents who entered that community
and the homesign systems were the basis, okay, the
roots of the emergence of that new language. The homesign systems I’m
gonna be talking about today are represented by these houses, okay, these houses don’t have little arrows that go into the community,
these are homesigners, deaf people, who have not
had the opportunity to join, or have had it and rejected it, as the gentleman you just saw, they have not acquired
Nicaraguan Sign Language. They’re using their own homesign systems that they have developed. Okay, so these roots are very distinct. So I wanna raise a puzzle. What does it mean to have a language? How should we think about
a system that has some, but maybe not all of the
properties that we associate with natural language? And where do these structures come from? And I think that the study of homesign really uniquely offers
perspectives on these questions that can’t be answered in any other way. I’d like to frame this
discussion by considering the relationship between
meaning and structure, or, I know these don’t exactly overlap, by considering the
relationship between semantics and syntax, obviously a
focus of the department here at the University of Chicago, the Linguistics department. There have been several proposals about how these are related
in acquisition of language. So, Tomasello proposes that
children acquire words first and then they acquire
structures using those words. Pinker suggested that semantics,
or the meaning of words, bootstraps the syntax, it
helps children learn verbs. And then syntactic bootstrapping, basically holds the converse, right, that the structure of a
sentence helps you learn the meanings of the words. Now, for my purpose, and so these people have
lots of other kinds of disagreements in their approach
to how kids learn language that we will also address
later in today’s talk. But for right now, what I want
to draw your attention to, is that all of these proposals assume that there’s linguistic
input in the environment. And the phenomena that we’ll see, that I’ll show you
evidence for in homesign, created a bootstrapping
problem that doesn’t have an obvious solution. Okay,
so none of these things can be working in homesign
because there’s nothing to bootstrap from. So not only these proposals,
but also other proposals that talk about distributional learning or statistical learning,
that is distribution over an input, right, that is not present for homesigners. And Susan Goldin-Meadow has made that point recently. So the specific questions
that I will talk about today are whether morphological development precedes syntactic development, and whether it’s possible
for an emerging language to fail to develop more abstract
or arbitrary structures such as phonology, this
really hasn’t been addressed in the previous literature. And when I say emerging language here, I mean to include both homesign and Nicaraguan Sign Language,
although I’ll only be talking about the homesign data today. Third, I’ll talk about some
studies looking at whether a conventional lexicon has
to be the starting point for future morphological and
grammatical developments. And then finally, I want
to return to that point that I eluded to when I talked about how the homesigner’s mother
doesn’t understand him. What is the role of communication
and social interaction? And it’s a complicated role,
okay, it’s not straightforward, it’s not that it matters not at all and it’s not that it
is totally crucial for, that it contributes structure. Okay, so here’s my little
bit of ASL discourse, I have to tell you what I’m going to say before I tell you what I’m going to say. So I’m gonna answer those
questions very briefly and then I’ll describe the
evidence in a more detailed way. So what we’ll see is
that handshape complexity marks morphological distinctions
and I’ll be referring to that as morphophonology. And that morphophonological structure emerges before morphosyntax, which may seem a little counterintuitive. With respect to how fast a
lexicon conventionalizes, we see that the patterns
of social interactions in a network affect how
quickly that happens. And we have both behavioral evidence as well as computational
modeling evidence for that. But those social interaction patterns don’t provide the structure. Okay, they’re necessary
for it, they influence it, but they’re not actually
providing the structure. And I want to argue, and also discuss, the implications of these findings for theories of the role
of meaning, reference, and interaction in creating
structure in homesign. So first, I’ll address
this question of how handshape comes to be used linguistically. And this work is
primarily a collaboration with Diane Brentari, Susan
Goldin-Meadow, and Laura Horton and Annie Senghas also
collaborate on this question. So Brentari and colleagues
have shown that different handshape types express
different classes of events. And those patterns appear in signers but not in gesturers, so
the people who are fluent in sign language will
already know about this, but I’ll give you some
examples for those of you who are not signers. So I want to contrast transitive events, those events that have agents, that a person’s doing something,
manipulating something, with events that do not have an agent. What we see in morphosyntax
is that handshape type expresses this contrast.
So here you see examples of handshapes that are used
to manipulate a small object, okay, so that handshape is
the handshape that you use when you’re talking about
picking up a Tic-Tac, something very small. In contrast, when you’re
talking about an event when there’s no agent
manipulating anything, the handshape represents
the object itself. And so you see here,
three different handshapes that take on the properties of the object that they refer to. So I want to turn now to another level of linguistic structure
that’s also in evidence here, and this is the morphophonological level. And here we’re talking
about selected finger group. These three handshapes up to here are examples of handshapes
where the selected fingers are very easy to describe, so
here, it’s these two fingers that are selected, that are
involved in forming a sign, here it’s all the fingers,
here it’s all but the thumb, we ignore the thumb, you can
talk to Diane about that later. Here, we have the medium and
high complexity handshapes where you need more nodes in the structure to represent which fingers
are being selected. Okay? So there’s a complexity range here, it’s not that that all handling handshapes are low complexity in
terms of selected fingers, but there’s a strong tendency for that. But only the object handshapes have medium and high
complexity handshapes. Okay, so these distinctions
are really important ’cause they serve as the foundation for the next several studies
that I’ll talk about. Okay, so I just wanted to highlight the thing I already said. So some of you may be asking
why are we calling this morphophonology or maybe
you’re not asking that because you all know Diane, but basically what our definition here, of what counts as morphophonology, is a case where you have
a set of classifiers that form a morphological class and they also form a phonological class. So here, all of the handshapes
in this set on the left have this fully opened
joint specification, I haven’t talked about
joints, I’m not gonna say more about joints, but they’re
all open on the joints, you don’t see any bending, okay, so on that joint feature,
they’re all the same. So they’re a morphological class and then they also function
as a phonological class. On the other hand, you can
imagine a morphological class that is not a phonological class, where the joint feature,
the joint specification differs across the handshapes, okay? So this is the kind of
combination of features that we’re talking about functioning as a morphophonological class. I wanna give you an idea now, of what the data look like and the stimuli as well as the data I’ll be talking about in the results coming up. So the first study I’m going to talk about is a longitudinal study
of one child homesigner. So, I was able to study
him between the ages of seven and twelve years of
age, this is towards the end of that longitudinal
study and the video clip shows the stimulus item
that he’s talking about and then his response, and
then you’ll see it again slowed down. So the analysis I’m about to show you, we’re only considering
the event description in these analyses, so in
that response you just saw, that’s the first part, the action part, and then you notice at the
end, he labels the plane, he tells you what kind of object was participating in that event. But the results I’m showing you now are only about this event part and what we did is we transcribed
all of these handshapes and we selected, in a very detailed way, and then we took all of that detail, we mushed it into one,
two, or three, basically, set of categories that
reflect the complexity level. So these handshapes reflect the fact that these are, it takes a
lot of description, okay, to arrive at how those
fingers get selected to produce this handshape. So what we found in our
initial study of a comparison of how homesigners and hearing
gesturers perform this task, we had them do exactly the same task, we asked the hearing people to not speak and just describe these
events using their hands and we coded them according to the the exact same coding scheme. And what we found is that the homesigners but not the hearing gesturers showed that cross-linguistically observed
sign language pattern where that handling handshapes were used for manipulation events
and the object handshapes were used when there was no agent. So what we learned from that result is that that pattern that we
see cross-linguistically in sign languages, is not
just about using your hands to talk about events because
we see different patterns in homesigners and hearing gesturers. It’s also not an elaboration
of the patterns that we see, it’s not fancy gesture, okay, it’s qualitatively different in the way it uses these features
to relate to the meanings that are being described. And, because we see that
pattern in homesigners, it means that it doesn’t
require linguistic community to develop because
homesigners don’t participate in a linguistic community,
they have a social, communicative community, but
not a linguistic community. Okay, so I want to turn now
to the patterns that we saw in adult homesigners with
respect to the morphophonology. So I was talking before
about how this handshape gets that three rating, okay,
that’s a high complexity handshape, and that’s
what’s being plotted here on the y-axis, the
complexity of the handshape, and then we have each of
four adult homesigners, you can see they were all in
their 20’s when we tested them on this task and you’ll
see the complexity ratings for the object handshapes in light gray and the handling handshapes in the black. And what you see is in three
of the four adult homesigners the selected finger complexity is higher for those object handshapes
than it is for the handling hanshapes, there’s
variability in how high it is, but what’s important is that distinction, that contrast between
the complexity levels across those two types of handshapes that mark morphosyntax. So where does that child homesigner fit in with respect to this pattern? Well, at the age of 12 he was already showing that adult homesign pattern. So not only does it not, this morphophonological development, not require linguistic community, it also doesn’t require a whole lifetime of using a homesign system,
he was able to do it at 12. Okay, so does that
morphophonological pattern appear before or after the
morphosyntactic pattern that we talked about? The association of a handling handshape with an agentive event
and an object handshape without an agent. So as I mentioned, we
tested him from the ages of 7 and a half to 12 and a half, and again, you see the
selected finger complexity on the y-axis here,
and what we see is that that morphophonolocial pattern
appeared at the age of 11. Okay it wasn’t present before
that, it appeared sometime before he was 11 and after he was 10. And this is a little bit
later than this development, the corresponding
development in deaf children who are acquiring ASL
as a native language, okay, so they acquire this pattern between the ages of four and six. But he’s doing it on his own, taking him a little bit longer, fair. Okay, I want to just
briefly comment on the morphosyntactic pattern that we see in established sign languages. I already described this in words, this is just what individual
ASL adults would look like. They would be using handling handshapes for agentive events and object handshapes with non-agentive events,
this just shows you what the typical established
sign language pattern is. Now I want to show you what the child homesigner looked like. So, for agentive events he
is looking pretty much like adult ASL signers but interestingly, for the events without an agent, he doesn’t really have a preference for one kind of handshape over the other. And this is a little bit, I found it counterintuitive
when we saw these data. This makes sense, you see an agent, you do a handling handshape, he’s all over the place
with the object handshapes, he’s producing the two handshape
types equally often. But importantly, he’s making a distinction between the two kinds of events even though he hasn’t fully gotten to the established sign language pattern. Okay… I’m gonna skip this point,
we can return to it, this is the seeds of a
noun-verb distinction, we can return to this in the discussion but I want to move on to the
next kind of question. And I’m gonna skip the summary. Okay, so now I want to
turn to another domain where you might think
that being able to observe meaningful events in the world might, and this, having interactions with other people in your family, talking about the same
things all the time, might actually help a homesigner pick, or might help that group
decide, conventionalize, on a particular form to
represent a particular meaning. So I want to turn now to
some studies we’ve done on the conventionalization
of the the lexicon. What we asked people to
do is produce gestures in response to very simple
images of everyday objects and what we measured is the distance between the responses
of any two participants. I can explain the details
of how we did that in the question period, if you like. But we looked at this both
in a homesign type network where the deaf person is the center node and then they’re using
that homesign system with each of their communication partners, who are hearing, but those
communication partners don’t use the homesign with each other. If they’re using, they’re
gesturing with the homesigner, these hearing people are
really great gesturers, but they don’t use it with each other. They’re speaking Spanish to each other ’cause it’s easier for them,
it’s their native language. So that’s the star type network that we see in homesign. I want to contrast that with
a sign language type network. In this case, we tested Nicaraguan signers and what you see here
is all of the signers have the possibility of interacting with all of the other signers, so this is a fully interactive network. And what we did is we wanted to compare the Nicaraguan signers and the homesigners to see how conventionalized
their lexical items were after the same amount
of time of being used. So we picked 25 years, because
that was basically how old the homesigners were at that point, if you assume that the homesigners, we picked two, okay, we just picked two, because that was when they
started using their homesign, we tested them 25 years later
and so we found a comparable group of Nicaraguan signers
who had been interacting for 25 years and we tested them after the same period of time. So here is what we found,
you’ll see the results for the Nicaraguan signers here
and the homesigners here, same amount, same period
of time and experience, almost the same in age, and
what I’m showing you here on the y-axis is a measure
of how conventionalized their lexical items were, their words
to refer to the same thing. So a score of zero here means they all produced the same sign when they saw a picture of a cow, they all produced the same sign when they saw a picture
of an orange, okay. So zeros here mean they’re
all on the same page, they’re all using the same sign. That’s what we see across all of the pairs of Nicaraguan signers. In contrast, what we see
in the homesign systems is that none of those
pairs of homesigners, a homesigner and their
communication partner, were producing exactly the same sign for those very common objects
after 25 years of interacting on a daily basis, this
was very surprising to me. I’m not showing you here, but they’re getting closer, okay, so the slopes of these go down after time, but they’re not fully conventionalized. They’re certainly nowhere
near as conventionalized as the Nicaraguan signers are. So, we decided to test
whether that network structure is the thing that might
be driving this difference in the rate of conventionalization. So, we and I mean the royal we here, I did not do the
computational modeling work, but the simulation, my collaborators, Russel Richie, my graduate student, and Charles Young, devised
a computational model that had all of these properties. It was an agent based
model with a very simple learning algorithm, I can talk
about some of the details, again, in the discussion period, but what we found was some
very striking differences between these two networks. So the average number of
interactions to convergence, which you can think of
as conventionalization, when they agree on what the sign should be for a particular meaning, was significantly lower
for the Nicaraguan signer, these are not real people, right, these are agents, but
for the NSL type network, it’s much lower than for
the homesign type network that’s not fully connected. And in fact all of the
sign language type networks converge, eventually,
whereas only 80% converged in the star type network. So to sum up, what we
learned from that study is that conventionalization
of lexical items happens more rapidly given
typical, richly connected linguistic community structure. So I want to briefly
describe some findings from a study we did on
narrative development. And again, you might think
that having access to observing meaningful events in the world and then retelling them
might help homesigners develop more structure, but
it turns out that it doesn’t compared to their peers
who are participating, not their peers, sorry,
compared to signers who are part of a deaf community. So, back to our good old
friends Tweety and Slyvester from the Canary Road cartoon. We asked people to watch that cartoon and retell the story, and then
we analyzed their productions and we got a measure of story goodness. So this measure says how
well organized your story is and how complete it is based
on its episodic structure. So a complete episode
here has to have all three of these components, an initiating event, an attempt, and a direct consequence. So, for those of you who saw the examples that I showed earlier
in the panel discussion, what we want to know is
were the differences, I think most of you were able to see the sort of global differences
in the amount of information and the quality of
information that was produced in the homesign narrative versus the Nicaraguan Sign Language narrative, and now I want to
quantify that in terms of how many complete episodes people in each of these groups produce. So here you’ll see the homesigners, so we’re sort of increasing in
terms of language evolution. On the x-axis, we have
each individual homesigner, four members of cohort one, they were the input to
cohort two, et cetera. And then we have ASL just as
a sort of baseline measure of what an established deaf
community sign language looks like on this measure. So what I’d like you to
notice is that the number of complete episodes increases with between homesigners and cohort one, in terms of participating
in a linguistic community, and then as language evolution proceeds, we see corresponding increases
in the quality of the story, how much information is being conveyed, as well as how it’s being organized. So the takeaways from that study are that social interaction and
being able to see stuff happen in the world, doesn’t compensate
for a lack of linguistic input or not participating
in a linguistic community. And those experiences mediate, well, I didn’t show you direct
evidence for this, but we can discuss this, both the encoding and the expression of observable events in the world, right, things that everybody has
the ability to perceive. All of the people in this study, but what they’re doing with
that information they perceive is mediated by these factors. So far, what we’ve seen
is two examples of domains where participating in
a linguistic community appears to be really
important for certain kinds of linguistic structure to develop, a conventionalized lexicon
and narrative abilities. But maybe we’re missing something by looking only at production, okay, maybe communicating about these events and these meanings might be more revealing about how structure develops in homesign. So I want to turn now to
systematic examination of that point I raised earlier, what’s the role of
communicating and understanding each other in the context
of a homesign family group and what role does that play in developing the structure we see in homesign. So this is a study that
I’ve done in collaboration with my doctoral student, Emily Kerrigan, and it’s focused on how
family members comprehend the utterances that homesigners produce. So I had noticed going way back many years that the homesigners’ family
members really struggled to understand many of the
things that they said. So what we did is we
decided to really pair this down. We had the homesigners
describe very simple events and then, to their communication partners, and then their partners
had to describe, I’m sorry, had to pick the picture that matched. We compared the scores of
the homesigners’ mothers and other family members,
although I’m going to focus on the mothers, here, to
the comprehension scores of ASL signers here in the United States, deaf native signers, who
had never met a Nicaraguan homesigner, they had no
experience with homesign systems in Nicaragua, just in the context of that, of doing this task. So we had several different event types, these are just, I’m not going
to go through all of these, but I just want to give you a sense of the kinds of events, these
are fairly simple events, it’s not that hard to
pick out the right answer if you’re following what the gestures are. So I’ll just pick this one to describe, so here you have a reversible
event with two entities. So the correct event, the
video that the homesigner is describing is a man kisses a woman and then the pictures that
the communication partner has to choose from, or the ASL signer, are the reverse, so where
the woman is kissing the man, a totally unrelated one, where it’s, the man’s doing something
completely different, and another unrelated one where he’s acting on an inanimate object. Okay, so this isn’t like
a man kisses a woman wearing a brown sweater
versus a man kisses a woman wearing a navy blue sweater,
these are not the kinds of contrasts that we are
asking people to make. Here are the results for
comparing the homesigners’ mothers in how well they understood,
how often they picked the right picture, compared
to how often the ASL signers, who have never met these people, okay, how well they did on this test. So you see in three out of the four cases, the ASL signers performed
significantly better. Not just statistically,
but like pretty freaking significantly better, okay. This, well, whatever, I
mean it is what it is. This mom is really good,
okay, she’s a really good comprehender and everyone in
that family is pretty good, so that’s a lucky homesigner. Okay, but I found this
result incredibly striking and quite consistent with my observations of their behavior in real life. I don’t want to go
through the charts here, I just want to point out
that when you consider all of the family members,
the factor that related to their comprehension score was the age that they were tested and how old they were
when they first started interacting with the homesigner,
those were the things that related to comprehension performance and not how many years experience they had communicating with the
homesigners, so that means that on average, younger siblings
were much better at this task than mothers who are
much older at the time that they started interacting
using the homesign system. And it doesn’t matter what kind of event, you see the same pattern, I don’t want to spend
any more time on this. So, the takeaway messages
from that set of studies is that successful communication
does not seem to be driving the development
of linguistic structure in homesign systems because, I
hope that I’ve convinced you, there’s not a lot of
successful communication that’s based on the gestures
that the homesigner produces. There’s successful communication happening in lots of other ways, based on routines, based on shared knowledge,
based on all kinds of contextual information,
but it’s not coming from the actual things that the
homesigner says with their hands. However, it’s clear, okay,
it’s clear that having someone to interact with and
engage with is necessary for this kind of structure to emerge. It stimulates linguistic structure, but it’s not directly creating
that linguistic structure. The mothers are not
inventing a gesture system and kind of giving it to
their deaf son or daughter. Because if that were the case,
seems reasonable to expect that they would understand
it when the homesigner says it back to them. Okay, so I want to just
sum up the findings that I’ve described to you from
these four sets of studies. First, going back to the analyses of child and adult homesign, we see that handshake complexity marks
morphophonological distinctions and that that morphophonological
structure emerges earlier in development than
morphosyntactic structure. I skipped this part, but in
work that Susan and Diane and Laura and Anne
Senghas and I did together, based on those same stimuli, those same kinds of events and coding, we also found evidence for
a very robust distinction between nominals and predicates. We saw that when you
look at how the lexicon, the list of words that
are used by homesigners and Nicaraguan signers, and
whether they’re all agreed on, we see that the social
interaction patterns influenced that development,
both in real life, in a naturalistic kind of situation, as well as from a computational
modeling perspective. We saw from the studies of narrative
that homesign structure does not faithfully
reproduce event features or the structure of
events out in the world. Even though you can see
something happening, that doesn’t automatically
mean that you’re encoding it in terms of cause and effect relationships or other conceptual
groupings based on temporal and causal relationships. And then from the set of
studies looking at comprehension of homesign
systems, I’m arguing that successful communication
doesn’t drive the development of the linguistic structure
that we see in homesign. Okay, let’s take a minute to look at the bigger picture here. I’m a big fan of homesign research, I’ve been doing this for 25 years. It’s really great for linguistics
and for cognitive science and for furthering our knowledge
about ourselves as humans. Being a homesigner and
being deprived of language is not great, okay, it’s very limiting. We talked earlier in the
day about how homesigners, nevertheless, despite
their lack of participation in a linguistic community,
they do develop an identity, so I don’t intend for
this to be robbing them of their dignity as people,
but I think it’s pretty clear that they’re not able to
reach their full capacity in the absence of being
exposed to linguistic input. So to address that, I
have two initiatives. One is founding an
organization here in Chicago at University of Chicago, in
the context of Susan’s lab with some students called
Manos Unidas, Hands Together, that works to promote educational access, equal access to language
and education and vocational opportunities for deaf
people in Nicaragua. And at UCONN, I strongly
encouraged a group of students to form a student
organization that also works both to raise money, but
also to advocate for access to high quality language
input, not just for deaf kids in Nicaragua, but for all kinds of groups that are vulnerable to not
getting that good input. Kids from bilingual families, kids from low socioeconomic
status families, we all know about the
gaps in language input that kids get, so they’re
advocating and educating people about that to improve language
outcomes and of course outcomes in other cognitive
domains for all kids. And of course, I have to acknowledge all of the collaborators
who make this work possible and the funding entities who
supported it financially. Thank you. (applause) (indistinct talking) – [Translator] I saw that
you gave another talk this afternoon, not this morning, with the Tweety Bird and Sylvester, and you did, you showed some results about the homesigners and Nicaraguan
Sign Language users, and that they used less,
the homesigners used less classifiers to describe the scenes, I was able to capture some of
the expressions on their faces and it seems that the
homesigners use minimal facial expression where cohort three
used more facial expressions, like they became the cat,
they became the bird, or the old lady, so I
noticed the distinction between those two groups. – Thanks, Drucie, for your observation, that is in fact true, we did,
that analysis that I showed didn’t specifically look
at the use of classifiers in each of the groups
although we did code that, but I think it’s a
really, really interesting and it brings me to a
point that I usually make when I talk about those
results which is that even the things that seem really gestural and mimetic, that are part
of a fluent sign language description of that kind of story, you see more of that the
more evolved the language is. Right, so the homesigners
are actually relying, they’re more on the gesture
continuum end of that, so you might expect
more of that from them. But that has already become
taken up and made part of the linguistic
structure of the language and it’s not a basic sort of basic thing that you get for free, okay,
is to represent characters in those ways even though
it’s totally observable, it’s right there, it’s what
people sometimes naturally do when they’re forced to use only
their bodies to communicate. But that coordination that you
see in the Nicaraguan signers reflects the fact that that
is a linguistic development, even though it looks very
acty-outy and kind of more gesturey and basic. So the other
point that I wanted to make is that when we counted up
both the number of events that participants
expressed in the narrative, as well as how they were packaged, we did not care how they did it. They could have literally
gotten up and acted it out like charades and we
would’ve given them credit, but the homesigners didn’t do that and that’s not what they do
in their regular life, either. So it’s really striking
to me that even though they have the basic
linguistic nuts and bolts to put together sentences that
would express those events, there’s something clearly
limiting about their underlying conceptual structure that prevents that from being expressed, it’s
not the linguistic means of expression that I think
is the limiting factor, I think it’s the underlying, conceptual, like I said, encoding… that affects how well they
can express both the quantity and quality of the events. Would you come up? – [Man] When you study homesigners, is it a possible research question, how many distinct items do
they have in their lexicons and if you can make sense
out of that question, how does it relate to the
other things that you study? – That’s a good question. So from Susan’s work
with young homesigners, we have a sense that they
have a stable lexicon even fairly early in development. Now those data, those
analyses are based on spontaneous productions, right, these are play sessions, the
kids are just engaged in, and kids tend to talk about
the stuff that’s right there, right, they’re kids, that’s
what little kids talk about whether they’re acquiring
a language or not. When you’re looking at, well… So if you try to do the same
kind of analysis with adult homesigners, if you haven’t
controlled their production, it’s very difficult to map one to one, whether a difference in
form really does indicate a difference in meaning
if you don’t know exactly, if you haven’t fully
constrained the meaning. So, I would say they have
fewer stable conventionalized items in their inventory
than Nicaraguan signers, and I think they probably
vary individually, but to get to the second
part of your question, does that relate to other aspects of their linguistic structure. That’s something I haven’t
looked at systematically, but I have the data to do that. It makes me think about a study
that some of my colleagues did while we were in graduate school. Amy Marie Aixty. A University of Chicago alum, and Carla Hudson Cam. They had hearing people, this
is all about hearing people, hearing people come in and
they taught them some words in Persian and then they
tested them on their knowledge of those vocabulary items
and then they had them solve communication problems,
give each other directions in this little toy world, kinds of things. And what they found is
that the people who had better vocabulary scores
actually had better syntax. It’s like as if, they call
it the lexical competence hypothesis, so that already
having premade words makes you not have to work
so hard to develop structure. So I think that’s a very intuitive idea, whether that’s true in the
case of the homesigners, I haven’t specifically
asked that question. But you will be the first to know. Anastasia, please
come up, please come up, I think it’s hard for people to hear you. – [Woman] I’m so loud. – Yeah. (laughter) – [Woman] Thank you so much,
this was very interesting. I have two questions, so
one is about the input, so I don’t know, having
known you for so many years, we say that in the
homesigns there is no input, but in fact there is no linguistic input, in the classical sense, but there is input in the gestural sense. – Absolutely. – [Woman] And we know now
from Susan’s work again, that co-speech gesture is
an integral part of speech, so it’s not necessarily
the case that we know exactly where language stops
and the co-speech gesture starts, I mean, it may
actually be that there is even a continuum there, so it’s
not fully accurate to say that they don’t have any input,
so they have this gestural input, and I was wondering
if you have studied or if you had any new thoughts about whether that kind of input
drives their structures. I mean, we saw a little
bit that it doesn’t fully, because we have the same
synonyms and the mothers, and the output of what
the signers produce, but then there’s this other
general question that I wanted to ask, so what does drive
the homesign, any idea? Thank you. – Thanks for your question. So, you’re right that co-speech gesture is very tightly integrated
with the speech that it accompanies but you can only interpret it in the context of that speech, right, it’s an integrated system as
long as you have both pieces. If you only have one piece,
as you would if you were a homesigner, that’s, there’s
very little structure there, especially if you consider
that, and McNeill has shown that, people, when they’re
speaking, tend to only produce one gesture per clause, so
you’re not really getting gestures combined with
each other that would offer the kind of structure that a
homesigner would be able to use. So I’m less worried about
that than I am about the way the homesigners
communication partners actually gesture with
them, which is often without using their voice,
they do use the homesign system back to the homesigner. – [Woman] Right, so this then
becomes part of the language. – So, in principle, is out
there in the world for the homesigner to see, but the
fact that the communication partners don’t understand
the homesign, argues to me that they don’t possess the structure that the homesigner has,
so it seems unlikely that they’re producing
it but they’re not able to comprehend it. Or that any, and of course now these guys are all 30 years old, they’re
all at least 30, right, so there’s no way to go back in time except by studying younger
homesigners to compare whose development is
preceding whose, right? So presumably the communication
partners have learned something, right, it’s equally plausible if we see structure in the
gestures of the communication partners, that they have
learned that structure from the homesigner, right,
it’s not obvious that the homesigner learned that from them. We have to actually see
it happening in real time, so Diane and I are actually
engaged in that longitudinal study of the child homesigner,
we have collected data both from him and from his
primary communication partner, his younger brother who is
fortunately young enough to be really kind of
co-creating the system in a way that might approach
what you’re talking about, but we don’t have those,
I can’t report any results from that, but that’s
certainly a question that we’re interested in, whether the
homesigns change over time, whether the communication
partners change over time, whether it looks like one
is trying to match the other in terms of structure, I
mean, I take those questions very seriously, but the available results don’t support that kind
of getting input argument. – [Woman] I see, okay, well
it would require, of course, as you said, that we
actually get to see the, you know, partners that
are relatively close to be able to assess accurately,
on the other hand, I was thinking that, you
know, if we actually think of language acquisition,
also when the children are very young, they are
exposed to a lot of input, but what they actually
extract from that input is not necessarily all of
it, so they extract very little things, there’s of
course also the general comprehensibility and
whatnot, where they are yet, they don’t really extract full sentences, you know, there’s some mimic
comprehension is perhaps not what we think it is,
so if we mimic this and we transport it into a situation in which what the learner reads from input is merely the gestural part,
and also the gestural part that if we think of responding
to their own gestural part, maybe what we see with the
homesigner could be an experiment to see what happens if the
whole language acquisition was just that part instead of the spoken. – I was with you almost until the end, what was, if it were only that part? Can you just say what you mean? – [Woman] So, if you have a
child, so like if you take a small child, you know,
just within the community, the average comprehensibility
of a small child, they hear input, they hear words, but they don’t understand
what you’re saying, they hear sounds,
likewise, the deaf child, you know, hears exactly the same thing, they don’t hear the sounds,
but they see the gestures, so they extract from the
input this particular aspect, which is the miming aspect. So as the hearing child
grows, they replace that, the co-speech role, the
co-speech gestural part, with actual spoken language,
but that never happens in the case of homesigners,
but they are, though, trained in extracting information and there are some
irregularities in frequencies and everything else, they
are trained in extracting information from that manual. – Okay, here’s my response that’s based on the evidence I presented today, if they’re doing that kind
of analysis of the input, then shouldn’t they use
the same lexical items that they’re parents are using to them? But we don’t see that. – [Woman] But that system,
the gestural system, is not as stable as the linguistic system, so the kinds of gestures
that the parents use, you know, may differ also through time because before they establish
a homesign with a child, they may gesture you know,
from meaning x, y, z, but as they communicate,
that ways, it changes into way harder, it
requires a different form, unless the system breeds itself, you know, the way the adapt waving gesture is going to more unstable
than the linguistic input, which is kind of stable,
you see what I mean? – No, I didn’t see what you mean. – [Woman] What’s cat
remains cat, you know, even, cat, as you know I
say, cat or home or house. – So the parents keep
using the spoken words. – [Woman] They keep using all that. – But the gesture for cat might change, that’s what we see, that is what we see, the gesture for cat continues to change. – [Woman] Right, so that
is the reason why we have this asimile, so they do acquire
something from the input, but the input itself is not a
set of stable lexical items. – I don’t disagree with that. – [Woman] Okay, okay. – Yeah, but we don’t see them converging. – [Woman] But that’s what I’m suggesting, that maybe we don’t see them converging because there is no stable system that they have to converge with. – That’s, yeah, that’s the
point I’m trying to make. That’s the point I’m
trying to make, right, there’s nothing, when I say
there’s no linguistic input, I mean there’s no
conventional stable structure that they are taking in
that they use the way that kids who are exposed
to linguistic input do. – [Woman] But as, I’m
sorry to, like, fully, but as they develop,
you know, the homesign, doesn’t that become partly stable? – So that’s what we see,
we see that they converge over time, the homesigner and their mom and their sister and their brother, the responses to those
items get closer together, but they’re never at the
point where they’re all using the same, 25 years later,
they live with each other, they’re not using the same form, and I think that’s because
their communication is very contextually based,
they are relying primarily on habits and routines and
this is what we always do, I mean, they don’t have a lot, I mean, so there’s this paradox where
they’re engaging gesturally so much more than the
parents of the deaf children that Susan has studied, but
they’re still really limited on what they can really communicate. But the homesigners are much less limited on what they can communicate, because they communicate
all kinds of stuff to me and to other people
who’ve had the benefit of growing up with a sign language. That seems to give us an
advantage in understanding what they’re saying, as it
does for the ASL signers who have never met them, right, there’s something there to
be understood, it’s just not being taken up, so that’s what, again, what leads me to argue
that there’s not so much going on out there, because otherwise, you would see more in the homesign. You would see more stability
and more consistency. – [Woman] I see. – [Translator] I’m really
fascinated with this, but I really want to now
move over to narrative. Okay. So from the two videos. – I can show you more. – [Translator] I would love to see more! But the two videos, the young
man that was talking about the plane and the crash and
the other boy was talking about the plane and how the plane moved around. So I was noticing in the
story what would happen is their eye gaze, that was
very interesting to me, so it actually talked a
little more about the action, it acted like a narrative
shift, the eye gaze. So that was something that I noted and I noticed that for a long time now, whether it’s first person or third person, so what was fascinating was the
boy, when he talked about it he looked at the plane, he
looked at how the plane moved. And I saw that there and
then the concept of the plane and the pilot and the gun at the head, and he went from first
person to third person, well that’s a complex concept, where does that come from
and how does it become the eye gaze, the literate relationship, where does that all stem from, so that’s one of my fascinations. – Peter thank you so
much for your comments. So we did in fact code things
like whether the participants used changes in eye gaze to mark
changes in… Sorry, I’m running on steam here. Whether it was a change from
first person to third person or a changing character for
example, so we can look at those data more closely, I didn’t report them
in so much detail here. In terms of the other, I agree with you, homesigners are doing
a lot of complex things that I haven’t even begun to describe. There’s these very, very, very rich data and I haven’t been able
to look at every aspect of all of their productions. Where does that, in terms of,
where does that come from? That’s the $64,000
question which, I think, with inflation is worth,
like, $3 million now. And I would say that clearly, I mean here’s the interesting thing about the homesign results, they clearly point to a biological basis for linguistic structure, but
they also clearly point to the fact that not everything
is gonna emerge, right, in the absence of a linguistic community. Even if you have someone
who’s really trying to communicate with you and
trying to talk to you everyday. So, some things depend on
certain kinds of interactions, some things depend on having
a linguistic community, some things depend on the
expression of that structure passing through multiple
iterations of child human brains, so all of the work that Annie Senghas and all of our collaborators are doing is trying to identify what
conditions are necessary for different kinds of structure to emerge and that’s kind of one of the goals I had hoped to achieve by
raising these results with all of you, is that the
biggest proponents of a biological basis for
a linguistic structure aren’t taking on those
really difficult questions, they’re very sticky, right,
how do we measure interaction? I mean, we just now paying attention to those kinds of things in
typical language acquisition, but clearly they matter a
lot in these kinds of cases as well, even more, and we can
see them so much more clearly because there’s that pesky
input getting in the way. I hate to put it in that glib way ’cause I’ve already talked about the limitations of language deprivation, but from a scientific perspective, this is a really amazing
opportunity to ask those questions, but it’s not easy and clearly
these worlds are colliding. The importance of biology, the importance of this interaction, and those people don’t
talk to each other enough, in my view, and they’re kinda divided and I’m hoping that we can break
down some of those barriers and really get to the hard questions. Please come up. – [Woman] I’ve looked at cases where a homesigner did decide
to join the NSL community at the age of a teenager or adult, and how did that, having
grown up on homesign rather than a full fledged language affect their NSL acquisition? – Thanks for raising that point. One thing I didn’t mention is that all of the Nicaraguan
signers in the studies that I talked about
today are early learners. It turns out that in
Nicaragua, there are almost no deaf people born to deaf families, the way that we have here
in the United States, so pretty much the earliest
age that most people are exposed is when they get to school. That age is getting lower
and lower over time, so the younger people in the
studies that I showed you entered the school and were exposed to Nicaraguan Sign Language by
four or five years of age, but all of the people I
showed you were exposed by five or six, so what we’re not seeing, the differences in the cohorts that we see are not about late acquisition. So having said that, ’cause that’s what we want to know, we want to know about whether the version of the language you learn affects the complexity of your
linguistic structure, not whether the age of your
acquisition affects that. We have an enormous body
of evidence that shows that the later you acquire a language, the less proficient you will be at it. So we have not focused
in our work on that, although I know that Annie
and her collaborators, especially Amber Martin,
have looked at the effects on spatial cognitive abilities, based on the age that people were exposed to Nicaraguan Sign Language. But since the earliest
work that Annie and I did, looking at production
of spatial modulation in a grammatical way,
we did replicate that, you know, many, many
prior findings showing that the later people are exposed, the less well they control
the grammatical aspects of their language and we did find that. But that’s not news, the news is, this is a new language
and the language itself is changing over time, so we have focused on those differences. – [Diane] I think that’s
a good time for stop. Let’s thank our speaker
for her stamina. (laughter and applause) – Thank you.

Leave a Reply

Your email address will not be published. Required fields are marked *