 Culture

 Categories
Language  Linguistic relativity
and  Cognitive categories
Culture  Social categories
 Gender
  We use the term culture to refer to all the
ideas and assumptions about the nature of
things and people that we learn when we
become members of social groups.
 It can be defined as “socially acquired
Culture knowledge.” This is the kind of knowledge
that, like our first language, we initially
acquire without conscious awareness.
 We develop awareness of our knowledge,
and hence of our culture, only after having
developed language.
  With the words we acquire, we learn to
recognize the types of category
distinctions that are relevant in our social
world.
 Very young children may not initially think
of “dog” and “horse” as different types of
Culture entities and refer to both as bow-wow.
 In order to use words such as dog or horse,
rain or snow, father or uncle, week or
weekend, we must have a conceptual
system that includes these people, things
and ideas as distinct and identifiable
categories.
  A category is a group with certain
features in common and we can think of
the vocabulary we learn as an inherited set
of category labels.
Categories  However, evidence from the world’s
languages would suggest that the
organization of external reality actually
varies to some extent according to the
language being used to talk about it.
  Some languages may have lots of different
words for types of “rain” or kinds of
“coconut” and other languages may have
only one or two.
 Although the Dani of New Guinea can see
all colors of the spectrum, they only use
names for two of them, equivalents of
Categories “black” and “white.”
 Observing this difference between the
number of basic color terms in languages,
we can say that there are conceptual
distinctions that are lexicalized
(“expressed as a single word”) in one
language and not in another.
  Some of the clearest examples of lexicalized
categories are words used to refer to people
who are members of the same family, or
kinship terms.
 In some languages, the equivalent of the word
Kinship father is used not only for “male parent,” but
also for “male parent’s brother.”
Terms  Yet, we also use the same word (uncle) for
“female parent’s brother.” That distinction is
not lexicalized in English, but it is in other
languages. For example, in Turkish the word
uncle is used for both female and male
parent’s brothers as dayı and amca.
  Having words for units of time such as “two
days” or “seven days” shows that we can
think of time (i.e. something abstract) in
amounts, using noun phrases, in the same
way as “two people” or “seven books” (i.e.
Time something physical).
 In another world view, time may not be
Concepts treated in this way. In the Hopi language,
spoken in Arizona, there were traditionally no
terms equivalent to most of our time words
and phrases (two hours, thirty minutes)
because our terms express concepts from a
culture operating on “clock time.”
  Linguistic relativity: it seems that the
structure of our language, with its
predetermined categories, must have an
influence on how we perceive the world.
 This idea simply captures the fact that we not
Linguistic only talk, but to a certain extent probably also
think about the world of experience, using the
Relativity categories provided by our language.
 Our first language seems to have a definite
role in shaping “habitual thought,” that is,
the way we think about things as we go about
our daily lives, without analyzing how we’re
thinking.
  Linguistic determinism: holds that
“language determines thought.” If language
does indeed determine thought, then we will
only be able to think in the categories
provided by our language.
 For example: English speakers use one word
Linguistic for “snow”. In contrast, Eskimos look out at all
Determinism the white stuff and see it as many different
things because they have lots of different
words for “snow.”
 The category system inherent in the language
determines how the speaker interprets and
articulates experience.
  Edward Sapir and Benjamin Whorf
produced arguments that the languages of
Native Americans, such as the Hopi, led
them to view the world differently from
those who spoke European languages.
The Sapir–  According to Whorf, the Hopi perceive the
world differently from other tribes (including
Whorf the English-speaking tribe) because their
Hypothesis language leads them to do so.
 In the grammar of Hopi, there is a distinction
between “animate” and “inanimate” and
among the set of entities categorized as
“animate” are clouds and stones.
  Whorf claimed that the Hopi believe that
clouds and stones are living entities and that
it is their language that leads them to believe
this.
The Sapir–  English does not mark in its grammar that
Whorf clouds and stones are “animate” so English
Hypothesis speakers do not see the world in the same
way as the Hopi. In Whorf’s words, “We
dissect nature along lines laid down by our
native languages”.
  English does lexicalize some conceptual distinctions in
the area of “snow,” with sleet, slush and snowflake as
examples.
 However, English speakers can also create expressions,
by manipulating their language, to refer to fresh snow,
powdery snow, spring snow or the dirty stuff that is piled
up on the side of the street after the snow plough has
Example: gone through.
 These may be categories of snow for English speakers,
Snow but they are non-lexicalized (“not expressed as a single
word”).
 English speakers can express category variation by
making a distinction using lexicalized categories (It’s
more like slush than snow outside) and also by indicating
special reference using non-lexicalized distinctions
(We decorated the windows with some fake plastic snow
stuff).
  As a way of analyzing cognition, or how
people think, we can look at language
structure for clues, not for causes.
Cognitive  The fact that Hopi speakers inherit a
language system in which clouds have
Categories “animate” as a feature may tell us
something about a traditional belief
system, or way of thinking, that is part of
their culture and not ours.
  Classifiers indicate the type or “class” of noun
involved.
 The closest English comes to using classifiers is
when we talk about a “unit of” certain types of
things.
 There is a distinction in English between things
treated as countable (shirt, word, chair) and
Classifiers those treated as non-countable (clothing,
information, furniture).
 It is ungrammatical in English to use a/an or the
plural with non-countable nouns (i.e. *a
clothing, *an information, *two furnitures). To
avoid these ungrammatical forms, we use
classifier-type expressions such as “item of” or
“piece of,” as in an item of clothing, a bit of
information and two pieces of furniture.
  Words such as uncle or grandmother provide
examples of social categories.
 These are categories of social organization
that we can use to say how we are connected
or related to others.
Social  The word «brother» is similarly used among
Categories many groups for someone who is not a family
member.
 We can use these words as a means of social
categorization, that is, marking individuals as
members of a group defined by social
connections.
  ‘Brother, can you spare a dollar?’: the word
brother is being used as an address term (a
word or phrase for the person being talked
Address or written to).
Terms  By claiming the kind of closeness in
relationship associated with a family
member, the speaker’s choice of address
term is an attempt to create solidarity (i.e.
being the same in social status).
  More typically, an interaction based on an
unequal relationship will feature address
terms using a title (Doctor) or title plus last
name (Professor Buckingham) for the one
with higher status, and first name only for the
Address one with lower status, as in: Professor
Buckingham, can I ask a question?~Yes,
Terms Jennifer, what is it?
 More equal relationships have address terms
that indicate similar status of the participants,
such as first names or nicknames: Bucky, ready
for some more coffee? ~ Thanks, Jenny.
  In many languages, there is a choice between
pronouns used for addressees who are socially
close versus distant. This is known as the T/V
distinction, as in the French pronouns tu
(close) and vous (distant). A similar type of
social categorization is found in German
(du/Sie) and Spanish (tu´ /Usted).
Address  In each of these distinctions, as in older English
Terms usage (thou/you), the second form is used to
indicate that the speakers do not really have a
close relationship.
 Traditionally, these forms could be used to
mark a power relationship. The higher status or
more powerful speaker could use tu or thou to
a lower-status addressee, but not vice versa.
  Lower-status individuals had to use the
vous or you forms when addressing those
of higher status.
 This usage is described as non-reciprocal,
Address but the reciprocal use (both speakers
Terms using the same form) of the tu forms has
generally increased in Europe among
younger speakers, such as students, who
may not know each other really well, but
who find themselves in the same situation.
  In English, people without special titles are addressed as
Mr., Mrs., Miss, or Ms.
 Only the women’s address terms include information
about their social status. In fact, one of the most
frequently used address terms for a woman indicates that
she is the wife of a particular man (called “Frank Smith,”
for example), as in Mrs. Smith, and sometimes even Mrs.

Address Frank Smith.

 When the original system was put in place, women were
Terms obviously identified socially through their relationship to a
man, either as wife or daughter. These address terms
continue to function as social category labels, identifying
women, but not men, as married or not.
 A woman using Ms. as part of her address term is
indicating that her social categorization is not based on
her marital status. This type of observation leads us to a
consideration of the most fundamental difference in social
categorization, the one based on “gender.”
  Biological (or “natural”) gender is the
distinction in sex between the “male” and
“female” of each species.
 Grammatical gender is the distinction
Gender between “masculine” and “feminine,” which is
used to classify nouns in languages such as
Spanish (el sol, la luna).
 A third use is for social gender, which is the
distinction we make when we use words like
“man” and “woman” to classify individuals in
terms of their social roles.
  In Sidamo, spoken in Ethiopia, there are
some words used only by men and some
used only by women, so that the
translation of “milk” would be ado by a
Gendered man, but gurda by a woman.
Words  Many Native American languages, such as
Gros Ventre (in Montana) and Koasati (in
Louisiana), are reported to have had
different versions used by men and
women.
  These examples simply illustrate that
there can be differences between the
words used by men and women in a
variety of languages.
 There are other examples, used to talk
Gendered about men and women, which seem to
Words imply that the words for men are “normal”
and the words for women are “special
additions.”
 Pairs such as hero–heroine or actor– actress
illustrate the derivation of terms for the
woman’s role from the man’s.
  Marking this type of difference through gendered
words has decreased in contemporary American
English as firemen and policemen have become
firefighters and police officers, but there is still a
strong tendency to treat forms for the man (his) as
the normal means of reference when speaking
generally: Each student is required to buy his own
Gendered dictionary.
Words  However, alternatives that include both genders
(his or her), or avoid gendered usage (their) are
becoming more common. Other terms, such as
career woman and working mother (rarely “career
man” or “working father”) continue the pattern of
special terms for women, not men.
  The term pitch is used to describe the effect of
vibration in the vocal folds, with slower
vibration making voices sound lower and rapid
vibration making voices sound higher.
 Among women speaking contemporary
Gendered American English, there is also generally more
Speech use of pitch movement, that is, more rising and
falling intonation. The use of rising intonation ( )
at the end of statements, the more frequent use
of hedges (sort of, kind of) and tag questions
(It’s kind of cold in here, isn’t it?) have all been
identified as characteristic of women’s speech.
  Men tend to use more assertive forms and
“strong” language (It’s too damn cold in
here!).
 Other researchers have pointed to a
Gendered preference among women, in same-
Speech gender groups, for indirect speech acts
(Could I see that photo?) rather than the
direct speech acts (Gimme that photo)
heard more often from men in same-
gender groups.
  Many of the features already identified in
women’s speech (e.g. frequent question-type
forms) facilitate the exchange of turns,
allowing others to speak, with the effect that
interaction becomes a shared activity.
Gendered  Interaction among men appears to be
Interaction organized in a more hierarchical way, with the
right to speak or “having the floor” being
treated as the goal. Men generally take longer
turns at speaking and, in many social contexts
(e.g. religious events), may be the only ones
allowed to talk.
  One effect of the different styles
developed by men and women is that
certain features become very salient in
cross-gender interactions. For example, in
same-gender discussions, there is little
Gendered difference in the number of times speakers
Interaction interrupt each other.
 However, in cross-gender interactions,
men are much more likely to interrupt
women, with 96 percent of the identified
interruptions being attributed to men in
one study involving American college
students.
  In same-gender conversations, women produce
more back-channels as indicators of listening and
paying attention. The term back-channels
describes the use of words (yeah, really?) or sounds
(hmm, oh) by listeners while someone else is
speaking.
Gendered  Men not only produce fewer back-channels, but
appear to treat them, when produced by others, as
Interaction indications of agreement.
 In cross-gender interaction, the absence of
backchannels from men tends to make women
think the men are not paying attention to them.
The more frequent production of back-channels by
women leads men to think that the women are
agreeing with what they are saying.
 PSYCHOLINGUISTICS:
THE STUDY OF LANGUAGE
PROCESSING
NEXT WEEK
(O’ GRADY ET. AL., 2005)
Week 11
PSYCHOLINGUISTICS:
THE STUDY OF LANGUAGE PROCESSING
 Psycholinguistics: The Study of Language Processing

Psycholinguistics is the study of language-processing mechanisms. Psycholinguists

study how word meaning, sentence meaning, and discourse meaning are computed and
represented in the mind. They study how complex words and sentences are composed in speech
and how they are broken down into their constituents in the acts of listening and reading. In
short, psycholinguists seek to understand how language is ‘done’.
A substantial additional challenge for the psycholinguistic researcher comes from the
fact that most of language processing involves not observable physical events such as eye
movement but rather mental events that cannot be observed directly. Research in this field
therefore requires that mental language-processing events be inferred from observable behavior.
Consequently, a large part of psycholinguistic research is concerned with the development of
new (and often very clever) techniques to uncover how language processing is accomplished.
 Psycholinguistics: The Study of Language Processing

Field Methods: Slips of the Tongue

Some of the earliest and most influential studies of language processing examined the spontaneous
slips of the tongue produced during speech. Slips of the tongue are also known as spoonerisms—after
Reverend William A. Spooner, who was head of New College, Oxford between 1903 and 1925. Reverend
Spooner was famous for producing a great many, often humorous, speech errors.
What he intended: You have missed all my history lectures. What he intended: noble sons of toil
What he said: You have hissed all my mystery lectures. What he said: noble tons of soil

What he intended: You have wasted the whole term. What he intended: the dear old Queen
What he said: You have tasted the whole worm. What he said: the queer old dean
 Psycholinguistics: The Study of Language Processing

Beginning in the 1960s, Victoria Fromkin began to study these and other
naturally occurring slips of the tongue and noted that they can be very revealing of the
manner in which sentences are created in speech. Here’s the important psycholinguistic
point: in order for these exchanges to occur, the sentence would have to be planned out
before the person begins to say it.
Another important observation that Fromkin made was that speech errors
also often involve ‘mixing and matching’ morphemes within words. Consider the
following slips of the tongue.
Intended: rules of word formation Intended: I’d forgotten about that.
Produced: words of rule formation Produced: I’d forgot aboutten that.
 Psycholinguistics: The Study of Language Processing

All these errors involve morphemes being exchanged within a sentence. As is

the case for sound exchange errors, these slips of the tongue provide evidence that a
sentence must be planned out to some degree before speech begins. They also provide
evidence that the morpheme, rather than the word, is the fundamental building block of
English sentence production.
As can be seen from these examples, slips of the tongue can offer a fascinating
window to the mechanisms involved in language production and to the role that
linguistic units such as phonemes and morphemes play in that production.
 Psycholinguistics: The Study of Language Processing

Words in the Mind

One of the most intense areas of psycholinguistic research has been the
investigation of how words are organized in the mind. We are all in possession of a
vocabulary that forms the backbone of our ability to communicate in a language. In
many ways, this vocabulary must be used the way a normal dictionary is used. It is
consulted to determine what words mean, how they are spelled, and what they sound
like. But the dictionary in our minds, our mental lexicon, must also be substantially
different from a desktop dictionary. It must be much more flexible, accommodating the
new words that we learn with ease. It must be organized so that words can be looked up
extremely quickly—word recognition takes less than one-third of a second and the average
adult reads at a rate of about 250 words per minute.
 Psycholinguistics: The Study of Language Processing
Many psycholinguists conceive of the mental lexicon as a collection of individual units as in the figure. In this figure, the lexicon is shown as a space in which entries of
different types are stored and linked together.The main questions that are asked about the mental lexicon are these: (1) How are entries linked? (2) How are entries
accessed? (3) What information is contained in an entry?
 Psycholinguistics: The Study of Language Processing

Although these questions are simple and straightforward, there is no way to

answer them directly because the human mental lexicon cannot be observed. So the
psycholinguist must use special experimental methods to understand how words are
organized, accessed, and represented in the mind. We will briefly discuss the two most
common of these methods— lexical decision and priming.
 Psycholinguistics: The Study of Language Processing

Lexical Decision
In the lexical decision paradigm, the experimental participant (in this example, a
native speaker of English) is seated in front of a computer screen. A word appears in the
middle of the screen and the participant must judge as quickly as possible whether the word
is a real English word and press a button labelled ‘yes’ or a button labelled ‘no’.
Lexical decision experiments usually involve comparing participants’ performance
on one set of stimuli (e.g., nouns) to their performance on another set of stimuli (e.g., verbs).
The key to the importance of the experimental paradigm is that in order for a participant to
respond ‘no’ to a stimulus such as blove or ‘yes’ to a real word such as glove , the participant’s
mental lexicon must be accessed. The lexical decision task can therefore be used to measure
the speed and accuracy with which words in the mental lexicon are accessed.
 Psycholinguistics: The Study of Language Processing

Another way in which the lexical decision task can be used to explore language
representation and processing is to investigate the speed and accuracy with which
participants press the ‘no’ button for different types of stimuli. It has been found, for
example, that pronounceable non-words such as plib show slower ‘no’ response times
than unpronounceable non-words such as nlib . Thus, participants’ lexical decisions seem
to take into account the phonotactic constraints of the language.
 Psycholinguistics: The Study of Language Processing

The Priming Paradigm

Priming experiments typically involve the same procedure as the lexical
decision task except that the word to be judged (now called the target) is preceded by
another stimulus (called the prime ). What is measured is the extent to which the prime
influences the participant’s lexical decision performance on the target stimulus.
The priming paradigm is an excellent technique for probing how words are
related in the mind. One of the first experiments using this paradigm showed that
response time is faster when a target is preceded by a semantically related prime (e.g.,
cat-dog) as compared to when it is preceded by an unrelated prime (e.g., bat-dog).
Results of this sort lead us to the view that words are related in the mind in terms of
networks.
 Psycholinguistics: The Study of Language Processing

On the basis of evidence from these priming experiments, psycholinguists

reason that when a word such as cat is seen, its representation is activated in the mind,
and that activation spreads to other words in the lexical network that are semantically
related (e.g., dog). Because the mental representation for dog has already been activated
through the prime, it is in a sense ‘warmed up’ so that when the participant later sees it
on the screen as the target, response time is faster than it otherwise would have been.
This is called the priming effect.
 Psycholinguistics: The Study of Language Processing

Sentence Processing
It is presumed that in sentence processing (i.e., in reading or listening), a
sentence is understood through the analysis of the meanings of its words and through
the analysis of its syntactic structure. Psycholinguists refer to this type of unconscious
automatic analysis as parsing . Much of the research on sentence processing is
concerned with the principles and steps in parsing, its speed, and the manner and
conditions under which it can break down.
Two groups of experimental paradigms that have been used extensively to
study sentence processing: timed-reading experiments and eye-movement
experiments.
 Psycholinguistics: The Study of Language Processing

Timed-reading Experiments
One of the more common and revealing timed-reading experimental
paradigms is the bar-pressing paradigm, in which participants are seated in front of a
computer screen and read a sentence one word at a time. The participant begins by
seeing the first word of the sentence in the middle of the screen. When the participant
presses a bar on the keyboard, the first word disappears and the second word of the
sentence appears in its place. This process continues until all the words in the sentence
have been read. The dependent variable in these experiments is the amount of time it
takes participants to press the bar after seeing a particular word (i.e., the amount of time
they need to process that word).
 Psycholinguistics: The Study of Language Processing

Bar-pressing experiments can be very revealing about the manner in which

sentence processing occurs. Rather than producing equal bar-pressing times across a
sentence, participants show a pattern that reflects the syntactic structure of the
sentence.
The Chinese, who used to produce kites, used them in order to carry ropes across the rivers.
Participants show longer bar-pressing times for processing content words such
as nouns and verbs and relatively less time for words such as determiners, conjunctions,
and prepositions. Of particular interest is the length of the pause at the clause boundaries
—especially at the end of the relative clause (kites) and the end of the full sentence
(rivers). This increased processing time is interpreted as reflecting the extra amount of
time required to integrate preceding information into a complete clause structure.
 Psycholinguistics: The Study of Language Processing

Eye Movements
When the eyes are at rest they take a ‘snapshot’ of two or three words. These
snapshots usually last from 200 to 250 milliseconds. While the snapshot is being taken, the
language-processing system calculates where to jump to next. During a jump to the next
fixation location (usually about eight letters to the right), the reader is essentially blind.
The details of eye movements in sentence reading are studied with sophisticated
laboratory procedures in which a participant is often seated in front of a computer screen
on which text is displayed. Eye movements are tracked by a device that illuminates the
participant’s eyes with low-intensity infrared light and records the reflection. The eye-
position data are linked to the position of text on the screen so that it is possible to
determine how the eyes move from one text position to another.
 Psycholinguistics: The Study of Language Processing

This technique has revealed that fixation times are typically longer for less
frequent words and that the points of fixation are typically centered on words such as
nouns and verbs, rather than on function words such as determiners and conjunctions.
Difficult sentence structures create longer fixation times as well as many more
regressive saccades. Regressive saccades are backward jumps in a sentence and are
usually associated with mis-parsing or miscomprehension. On average, backward
saccades make up 10 to 15 percent of the saccades in sentence reading.
 Psycholinguistics: The Study of Language Processing

Brain Activity: Event-related Potentials

ERP experiments measure electrical activity in the brain. Electrodes are placed on a
participant’s scalp and recordings are made of voltage fluctuations resulting from the brain’s electrical
activity. There is a significant difference between ERP recordings and the more familiar EEG (electro-
encephalogram) recordings. In the EEG, all the electrical activity of the brain is recorded. This
electrical activity results from a very large number of background brain activities that are always going
on. The advantage of the ERP approach is that it uses a computer to calculate what part of the brain’s
electrical activity is related to a stimulus event (in our case, words or sentences on a screen). This is
done by a process of averaging. The computer records the instant at which a stimulus is presented and
compares the voltage fluctuation immediately following the stimulus presentation to the random
background ‘noise’ of the ongoing EEG. By repeating this process many times with stimuli of a
particular type, random voltage fluctuations are averaged out and the electrical potentials related to
that stimulus type can be extracted. The resulting wave forms are the event-related potentials.
 Psycholinguistics: The Study of Language Processing

It turns out that in the processing of implausible sentences, the brain displays a characteristic ERP
sign of surprise. Consider the following sentences:
a. The pizza was too hot to eat.
b. The pizza was too hot to drink.
c. The pizza was too hot to cry.
The sentences are arranged in order of semantic plausibility. In the first case, the last word fits in
perfectly well with the sentence and would typically be expected by the reader. The ERP for this sentence
shows a positive voltage associated with the last word. In the case of (b), however, in which the last word does
not make sense (people do not drink pizza), the ERP is much more negative. This negative wave occurs 400
milliseconds after the onset of the word. For this reason, this signal of semantic anomaly is called the N400
(negative wave at 400 milliseconds after stimulus presentation). The N400 is even stronger in the case of
sentence (c), which is even less congruent with the sentence context (drink is at least associated with food).
 Psycholinguistics: The Study of Language Processing

Language Processing and Linguistics

One of the most important results of such psycholinguistic investigations is
that many of the concepts and principles used by linguists to describe and understand
the structure of language in terms of phonetics, phonology, morphology, and syntax
have been found to also play an important role in the understanding of how language is
produced and comprehended during activities such as speaking, listening, reading, and
writing.
In this section, we will focus on these points of contact between theoretical
linguistics and psycholinguistics. In doing this, we will highlight the correspondence
between the study of language processing and concepts that are central to the study of
phonetics, phonology, morphology, and syntax.
 Psycholinguistics: The Study of Language Processing

Phonetics and Phonology

Language processing shows evidence that features, phonemes, and syllable
structure all capture some aspects of the way in which we process language but that
speech production and perception is a complex activity that involves much more than
these phonetic and phonological representations. To see why this is the case, consider
what might occur when you hear the sentence The dog bit the cat . Because the
utterance unfolds in time, you will first hear the segment /ð/ and then the segment / ə/.
(In fact, you do not hear these segments separately; rather, you create them out of a
continuous sound stream.)
 Psycholinguistics: The Study of Language Processing

As soon as these segments are identified, you have already accessed the
representation for the word the in your mental lexicon. When the next segment comes up in the
sound stream, you already know that it is the beginning of a new word and you also know that
this word will probably be a noun. The phonetic analysis that follows identifies the segments [d],
[ɑ], and [g] and the corresponding lexical entry dog . Now come the first segments of the word bit
. In principle, the first two phonemes /bɪ/ could be the first two segments of the word believe , but
you are not likely to consider this possibility because your developing interpretation of the
sentence is biasing you toward the word bit , which is associated in your mind with dog .
 Psycholinguistics: The Study of Language Processing

As can be appreciated from this example, language processing involves the interplay
of information that develops simultaneously at many different levels of analysis. The person
hearing the sentence The dog bit the cat is performing a phonetic analysis to isolate
phonemes and word boundaries and to relate these to items in the mental lexicon. This
inductive analysis is referred to as bottom-up processing. But we do not wait until we have
analyzed all the phonemes in a sentence before we begin to try to understand it. Rather, we
begin interpretation of a sentence spontaneously and automatically on the basis of whatever
information is available to us. For this reason, by the time we get to the word bit , we are not
only recognizing it using bottom-up processing but are also employing a set of expectations
to guide phonetic processing and word recognition. This is called top-down processing. In
normal language use, we are always engaged in both bottom-up and top-down activities. We
never just process features, or phonemes, or syllables. We process language for the purposes
of understanding each other.
 Psycholinguistics: The Study of Language Processing

Morphological Processing
The psycholinguistic study of morphology seeks to understand how word structure
plays a role in language processing.
Morpheme Activation
Words such as blackboard , happiness , and watching are made up of two morphemes.
In the case of the compound blackboard , both morphemes are roots. In the case of happiness ,
one morpheme is a root and the other is a derivational suffix. Finally, in the case of watching, one
morpheme is a root and the other is an inflectional affix.
 Psycholinguistics: The Study of Language Processing

For most multi-morphemic words, individual morphemes are

automatically activated during word recognition. One source of evidence for this
conclusion comes from priming experiments in which it is found that multi-
morphemic words will prime semantic associates of their constituents in a lexical
decision experiment. Thus, when a participant is exposed to a multi-morphemic
word such as suntan, the activation of that word in the mind will facilitate the
recognition of moon, which is highly associated with the initial morpheme of sun but
not with the whole word suntan. The facilitation, therefore, must arise from
independent activation of the morphemes within a word.
 Psycholinguistics: The Study of Language Processing

There are two proposals about how this activation of morphemes might
occur. The first is that it derives from structured morphological representations in
the mind. According to this view, multi-morphemic words such as happiness and
blackboard are represented in the mind as [happy + ness] and [black + board], and
when we access such words, first their whole word forms and then their constituent
morphemes are automatically activated. This view is termed post-lexical
decomposition, because the constituents of a multi-morphemic word are activated
only after the representation of the whole lexical item is activated. In other words,
whole word access occurs first, morphological decomposition second.
 Psycholinguistics: The Study of Language Processing

An alternative view is that the activation of constituent morphemes results

from a computational mechanism that scans a word and isolates individual
morphemes in much the same way as individual words are isolated when we see or
hear sentences. This view is termed pre-lexical decomposition, because the
constituent morphemes are activated through a computational process (called
morphological parsing ) rather than through the morphological representation of
the word in the mind. In this case, morphological decomposition occurs first, and
whole-word access occurs second.
Psycholinguistics: The Study of Language Processing
 Psycholinguistics: The Study of Language Processing

Syntax
Syntax is the system of rules and categories that underlies sentence formation in human
language. One of the fundamental insights in the study of syntax is that sentences are unique events.
They are typically made up of familiar words, but the particular combination of words and the manner
in which they are arranged are unique to each sentence.
Researchers concluded that there is at least some difference between the rules that native
speakers use to generate and comprehend sentences and the rules that linguists use to characterize
the linguistic knowledge of native speakers. It was therefore necessary to postulate a special module
for sentence processing, called the syntactic parser, and another for grammatical knowledge.
 Psycholinguistics: The Study of Language Processing

The parser is understood to be a system that makes use of grammatical

knowledge but that also contains special procedures and principles that guide the order in
which elements of a sentence are processed and the manner in which syntactic structure is
built up. Because our parsing ability is based on our grammatical knowledge of our
language, it is usually the case that there is a close correspondence between sentence
parsing and grammatical structure.
A variety of psycholinguistic studies have investigated whether this same sort of
modularity is present in syntactic processing. In other words, they look at whether
syntactic parsing operates in an automatic and obligatory manner that is relatively
independent of the activity of other processing systems. Two sources of evidence have
been very important in the exploration of modularity in sentence processing: garden path
sentences and sentence ambiguity.
 Psycholinguistics: The Study of Language Processing

Garden Path Sentences

Some sentences are extraordinarily difficult to understand even though they
are not very complex syntactically. These sentences are called garden path sentences
because they lead the syntactic parser down the garden path to the wrong analysis.
The horse raced past the barn fell.
This sentence is perfectly grammatical but almost impossible to understand.
The reason for this is that as we read the sentence, we build up a syntactic structure in
which the horse is the subject of the sentence and raced past the barn is the VP. When we
get to the word fell , we are surprised because the sentence we have built up has no
room for an extra VP.
 Psycholinguistics: The Study of Language Processing

The correct interpretation for the sentence requires that fell be the head of the main VP and that raced past the barn be
a clause (a reduced version of ‘which was raced past the barn’) that attaches to the NP the horse.
 Psycholinguistics: The Study of Language Processing

The ways in which native speakers misunderstand garden path sentences

reveals how the parser might work. It seems that we construct syntactic representations
from the beginning of the sentences to the end and that our sentence parsers are
organized so that we make a number of assumptions about how a sentence will proceed.
It has been claimed that the garden path effect results from two principles of
parsing: minimal attachment and late closure. The principle of minimal attachment
states that, all other things being equal, we do not build more syntactic structure (like an
extra embedded phrase) than is absolutely necessary. The principle of late closure states
that we prefer to attach new words to the clause currently being processed as we
proceed through a sentence from beginning to end.
 Psycholinguistics: The Study of Language Processing

Sentence Ambiguity
Another important clue to how syntactic processing is accomplished comes from the study of ambiguity.
The tuna can hit the boat = a. tuna meat that is packed in a small round can
b. a large fish swimming toward a boat
This suggests that in fact sentence processing proceeds in two stages. In the first stage, all possible
representations and structures are computed. In the second stage, one of these structures is selected and all others are
abandoned. Of course, all this happens very quickly and subconsciously, so that we as native speakers of a language are
never aware that we compute two possible interpretations. The point of the psycholinguistic experiments just described is
this: no matter which interpretation you arrived at (a or b), you probably considered both of them, chose one, discarded the
other, and forgot about the whole thing in less than a second.
 Psycholinguistics: The Study of Language Processing

Psycholinguistic Modelling
Psycholinguistic researchers present their ideas about how language is ‘done’
in terms of models. A psycholinguistic model incorporates the results of experiments
into a proposal about how processing takes place. In other words, it is a statement of
what happens when.
 Psycholinguistics: The Study of Language Processing

A serial psycholinguistic model of sentence processing

 Psycholinguistics: The Study of Language Processing

As we inspect the model, we should find that it is really very inadequate. It is

missing much important detail, it seems to characterize only one aspect of sentence
processing, and it avoids any mention of how meaning is accessed or how sentence
interpretation actually takes place.
We want models to be as detailed and comprehensive as possible, to take a
great deal of experimentation into account and, perhaps most importantly, to show how
linguistic and non-linguistic operations work together in the processing of language.
 Psycholinguistics: The Study of Language Processing
SerialVersus Parallel Processing Models

A parallel psycholinguistic model of sentence processing claims that phonological, lexical, and syntactic
processes are carried out simultaneously. Here information does not flow in a sequential manner. Rather, all modules
operate simultaneously and share information. Therefore, when we hear a sentence, we begin phonological, lexical,
and syntactic processes at the same time. As each type of processing proceeds, it informs the other.
 Psycholinguistics: The Study of Language Processing

In recent research, serial and parallel processing models have been very
important for understanding the extent to which bottom-up and top-down processing
interact. Serial models correctly characterize those aspects of language processing that
are modular and are driven by strict bottom-up procedures, such as phonetic perception.
Parallel processing models, in contrast, are more effective than serial models at
characterizing complex processes such as sentence comprehension.
 Psycholinguistics: The Study of Language Processing

Single-route versus Multiple-route Models

Put most directly, single-route models claim that a particular type of language
processing is accomplished in one manner only. Multiple-route models claim that a
language-processing task is accomplished through (usually two) competing
mechanisms. Consider, for example, the task of reading English words. Here there are
three possibilities: (1) we read a word by looking it up in our mental lexicon based on its
visual characteristics; (2) we convert a visual input into a phonological representation
first, and this phonological representation becomes the basis for comprehension; or (3)
we do both at the same time. Options (1) and (2) represent single-route models in the
sense that they claim that reading is accomplished in one way.
 Psycholinguistics: The Study of Language Processing

Option (3) represents a multiple-route model in that it claims that both mechanisms are employed. Usually such
multiple-route models employ the additional metaphor of a horse race by claiming that for some words (e.g., very
frequent short words) the direct route is faster but for others (e.g., rare words) the phonological conversion route is
faster and ‘wins the horse race’. The whole-word recognition route wins the race for frequent bimorphemic words such
as blackboard.
 Psycholinguistics: The Study of Language Processing

Symbolic versus Connectionist Models

These types of models represent fundamentally different views about the
nature of mental representations. Symbolic models claim that linguistic knowledge must
make reference to rules and representations consisting of symbols corresponding to
phonemes, words, syntactic category labels, and so forth. Connectionist models claim
that the mind can be best understood by reference to large associations of very simple
units (often called nodes) that more closely approximate the kinds of processing units
(i.e., neurons) that we know the brain to be composed of. Connectionist models typically
do not contain direct representations for language units such as words but rather
represent these units as an association of nodes.
 Psycholinguistics: The Study of Language Processing

Implementing and Testing Models Computationally

Currently, the best way to evaluate the value of a psycholinguistic model is to
implement it as a computer program and to evaluate whether the patterns of
performance of the computer program accord with the performance of humans carrying
out language tasks. Thus the merit of a proposed psycholinguistic model increases if it
makes the same kinds of mistakes that humans do, if it performs more poorly on
language structures that humans also find difficult, and if its performance changes over
time in accordance with the way in which human performance changes across the
lifespan. In this way, psycholinguistic models that incorporate human learning theory are
particularly valuable to our ability to capture the dynamic nature of human language
processing.
 NEXT WEEK

Clinical Linguistics
(Cummings, 2015)
 WEEK 12

CLINICAL LINGUISTICS
 CLINICAL LINGUISTICS

Clinical Linguistics: a broad area of academic and

clinical investigation that overlaps with medical and
other fields of study (e.g. neurology, psychology) and
that demands a sound understanding of language at
all of its levels.
 CLINICAL LINGUISTICS
Human communication is a remarkably complex
process that draws upon a diverse set of linguistic, cognitive
and motor skills. Before we even utter a single word, we
must decide what message we want to communicate to a
listener or hearer.
Deciding what that message should be is itself a
complex process that requires knowledge on the part of the
speaker of the context in which a verbal exchange is
occurring, the relevance of the message to that context and
the goals of a particular exchange. Having successfully made
these assessments, the speaker will have a clear
communicative intention in mind which he or she will wish
to convey to the hearer.
 CLINICAL LINGUISTICS
In most communication between people, that intention
is conveyed through language, although it may also be
conveyed through non-verbal means (e.g. gesture, facial
expression).
To the extent that a linguistic utterance is to be
produced, the speaker needs to select the phonological,
syntactic and semantic structures that will give expression to
this intention. This is achieved in a stage of the communication
cycle called language encoding.
CLINICAL LINGUISTICS
 CLINICAL LINGUISTICS
The speaker must select from the range of motor
activities that the human speech mechanism is capable of
performing those that are necessary to achieve the
transmission of the utterance to a listener. It is during the
stage of motor programming that these selections are
made (subconsciously, of course) and certain motor routines
are planned in relation to the utterance.
Motor programs can only be realized if anatomical
structures (e.g. lips, tongue, vocal folds) receive nervous
signals instructing them to perform particular movements.
These movements are carried out during a stage in
communication called motor execution. Assuming all
preceding stages have been performed competently, the
result of these various processes is audible, intelligible
speech.
 CLINICAL LINGUISTICS

Human communication is not just about producing

utterances but is also about receiving or understanding them.
The first step in this receptive part of the communication
cycle is called sensory processing. This is the stage during
which sound waves are converted in the ear from mechanical
vibrations into nervous impulses which are then carried to the
auditory centers in the brain.
The auditory centers are responsible for recognizing or
perceiving these impulses as speech sounds on the one hand
or non-speech (environmental) sounds on the other hand.
When recognition of speech sounds is involved, this stage of
the communication cycle is called speech perception.
 CLINICAL LINGUISTICS
Having recognized or perceived speech sounds, the
task of attributing significance or meaning to the utterance
begins. Rules begin to analyze the phonological, syntactic
and semantic structures in the linguistic utterance in a stage
called language decoding. Amongst other things, these rules
tell the hearer the grammatical constructions used in the
utterance (e.g. passive voice rather than active voice) as well
as the semantic roles at play in the utterance.
The outcome of linguistic decoding of the utterance is
not necessarily the particular communicative intention that
the speaker intended to convey. Quite often, further
processing that is pragmatic in nature is needed to recover
the particular intention that the speaker intended to
communicate.
 CLINICAL LINGUISTICS

It is only when the speaker’s intended meaning is

recovered by the hearer that the communication cycle can be
said to be complete.
The complex nature of the human communication
cycle means that there is a multitude of ways in which this
cycle may be disrupted in people with communication
disorders. The adult with schizophrenia who has thought
disorder may have difficulty formulating an appropriate
communicative intention, that is, one which is relevant, fulfils
the goals of a particular communicative exchange, and so on.
The adult with non-fluent aphasia (or dysphasia) has
impaired language encoding skills, one consequence of which
is the use of syntactically reduced utterances.
 CLINICAL LINGUISTICS

The child or adult with verbal dyspraxia may

struggle to select specific motor routines during the motor
programming of the utterance. The child with cerebral
palsy or the adult who sustains a traumatic brain injury may
both experience dysarthria, a disorder that disrupts the
execution of speech on account of failure of nervous
impulse transmission to the articulatory musculature.
 CLINICAL LINGUISTICS

Sensory processing of the speech signal may be

disrupted in the individual with hearing loss. This loss may
be either congenital (present from the time of birth) or
acquired (as a result of an infection, for example) and may
be conductive or sensorineural in nature (conductive hearing
loss describes a failure of speech sound conduction in the
middle ear, while sensorineural hearing loss is the result of
damage to auditory centers in the brain or to any of the
auditory nerves carrying nervous impulses to the brain).
 CLINICAL LINGUISTICS

Finally, the child with an autistic spectrum

disorder (ASD) may have impaired pragmatic skills and
may fail to recover a speaker’s communicative intention
in producing an utterance. These conditions and many
(stammering, voice disorder, etc.) form the complex array
of communication disorders that are examined by clinical
linguists.
 CLINICAL LINGUISTICS
The discussion of the communication cycle introduced
a number of terms that are central to work in clinical
linguistics. The cycle drew a distinction between the
expression or production of utterances (expressive
language), and the reception or understanding of utterances
(receptive language).
This distinction pervades the assessment and
treatment of communication disorders. For example, the
clinician who is asking a client with aphasia to point to the
picture in which ‘The man, who is crossing the road, is tall’ is
assessing that client’s understanding of relative clauses.
 CLINICAL LINGUISTICS

The receptive-expressive distinction allows clinicians

to characterize a number of different scenarios. One such
scenario is where there is a mismatch in receptive and
expressive skills in a client, that is, where one set of skills is
significantly better than the other. For example, in the adult
with non-fluent aphasia receptive language skills are
typically superior to expressive language skills.
Another scenario is where one set of language skills
deteriorates more rapidly than the other in a client. For
example, in the child with Landau-Kleffner syndrome
receptive language skills are first to be affected. Expressive
deficits usually occur later in the disorder and are thus
considered to be secondary to the receptive impairment.
 CLINICAL LINGUISTICS

A second important clinical distinction is that

between a developmental and an acquired
communication disorder. For a significant number of
children, speech and language skills are not acquired
normally during the developmental period. This may be the
result of an anatomical defect or neurological trauma
sustained before, during or after birth. The impact of these
events on the development of speech and language skills
varies considerably across the babies and children who are
affected by them.
 CLINICAL LINGUISTICS

The group of developmental communication

disorders is thus a large and diverse one including
children with cleft lip and palate (anatomical defect in
the pre-natal period), children with brain damage due to
oxygen deprivation during labor (neurological insult in
the peri-natal period) or children with cerebral palsy as a
result of meningitis contracted at 6 months of age
(neurological damage in the post-natal period).
 CLINICAL LINGUISTICS

The group of acquired communication

disorders is equally large and diverse. Previously intact
speech and language skills can become disrupted for a
range of reasons including the onset of disease, trauma
or injury affecting the anatomical and neurological
structures that are integral to communication. An adult
may develop a neurodegenerative condition like motor
neuron disease, multiple sclerosis, Parkinson’s disease
or Alzheimer’s disease. He or she may sustain a head
injury in a road traffic accident, violent assault, sports
accident or as a result of a trip or fall.
 CLINICAL LINGUISTICS

A previously healthy adult may sustain a stroke

(known as a cerebrovascular accident or CVA). He or she
may succumb to infection (e.g. meningitis) or develop
benign and malignant lesions on any of the anatomical
structures involved in speech production (e.g. larynx,
tongue). Any one of these events will disrupt
communication skills leading to disorders such as acquired
aphasia and dysarthria.
 CLINICAL LINGUISTICS

A third distinction that is integral to work in clinical

linguistics is that between a speech disorder and a language
disorder. These are not the same thing notwithstanding
everyday usage (people tend to use ‘speech disorder’ to refer
to both speech and language disorders). The distinction
between a speech and a language disorder can be best
demonstrated by referring to the diagram of the
communication cycle in figure 14.1. Breakdown in the boxes
in this diagram labelled ‘language encoding’ and ‘language
decoding’ typically leads to a language disorder.
 CLINICAL LINGUISTICS

So the adult with aphasia and the child with specific language
impairment have a language disorder because they are unable to
encode and decode aspects of language (e.g. syntax, semantics).
However, breakdown in the boxes labelled ‘motor programming’ and
‘motor execution’ typically lead to speech disorders, verbal dyspraxia
(or apraxia) and dysarthria, respectively.
 CLINICAL LINGUISTICS

Acquired communication disorders

Individuals with previously intact speech and language
skills can develop a communication disorder. An adult may
sustain a stroke and develop a language disorder known as
aphasia. The adult with dementia or mental illness may also
experience disruption to their language skills. A
neurodegenerative condition like motor neuron disease and
Parkinson’s disease can cause deterioration in speech skills
leading to dysarthria. Benign and malignant lesions may
compromise the neurological and anatomical structures that are
involved in communication.
 CLINICAL LINGUISTICS

Aphasia
Aphasia is an acquired language disorder that results from damage
to the language centers in the left hemisphere of the brain. Although a range
of neurological diseases and traumas can cause aphasia, the main cause of
the disorder is cerebrovascular accidents or strokes. Aphasia can compromise
language at all levels including phonology, morphology, syntax, semantics
and pragmatics.
It affects language in all its input and output modalities. So, as well
as having problems producing and understanding spoken language, the adult
with aphasia may struggle to produce and understand written language
(acquired dysgraphia and dyslexia, respectively) or produce and
comprehend signs (if the client with aphasia is a user of British Sign
Language, for example).
 CLINICAL LINGUISTICS

Historically, a number of different terms have been

used to characterize aphasia. Some of these labels reflect
neurological criteria such as the site of the lesions in the brain
that cause aphasia (e.g. Broca’s aphasia, Wernicke’s aphasia).
Other labels reflect the fact that the production or
comprehension of language can be chiefly affected in aphasia
(e.g. expressive or receptive aphasia).
In this section, the dominant classification system will
be adopted – certainly the one used by the National Aphasia
Association in the US – in which aphasia is broadly classified
into fluent and non-fluent types.
 CLINICAL LINGUISTICS
In fluent aphasia, there is a severe impairment in the
comprehension of language in the presence of effortless, fluent speech.
Utterances are well articulated and the intonational and other
suprasegmental features of normal speech are also present. This can
give the person with fluent aphasia the appearance of being a
reasonably competent communicator. However, language output is
often very incoherent due to the extensive use of jargon by the client
(hence, the term ‘jargon aphasia’ to describe this type of aphasia).
The person with fluent aphasia displays poor monitoring and
correction of incoherent language output. Other linguistic features of
the disorder include echolalia (the client echoes back another speaker’s
utterance), circumlocution (the person with aphasia is unable to
produce a target word on account of lexical retrieval problems and
proceeds to talk around it), and perseveration, where the speaker with
aphasia continues to produce a linguistic form beyond what is
appropriate. Finally, the person with fluent aphasia may also produce
errors known as phonemic paraphasias (e.g. the use of canerdillar for
‘caterpillar’).
 CLINICAL LINGUISTICS

The person with non-fluent aphasia displays a quite different

pattern of linguistic impairments. In this type of aphasia,
comprehension is relatively intact. However, the client displays
considerable struggle in producing utterances and is both aware of, and
frustrated by, his or her expressive difficulties. On account of his or her
restricted output, the person with non-fluent aphasia tends to speak in
short intonation units with the result that the suprasegmental features
of speech are disrupted. Lexical-semantic disturbances are common in
non-fluent aphasia and may manifest themselves in errors known as
semantic paraphasias. In these errors, the word uttered by the speaker
with aphasia is often semantically related to the target lexeme (e.g.
watch for ‘clock’).
Non-fluent output, which includes the retention of content
words (e.g. nouns, verbs and adjectives) and the loss of function words
(e.g. definite and indefinite articles, auxiliary verbs, prepositions), can
give spoken language the appearance of a telegram (hence, the use of
the term ‘agrammatic aphasia’ to describe this type of aphasia).
 NEXT WEEK

COMPUTATIONAL LINGUISTICS
COMPUTATIONAL
LINGUISTICS
 Computational linguistics seeks to
develop the computational machinery
needed for an agent to exhibit various
COMPU forms of linguistic behavior.
TATION
By “agent,” we mean both human
AL beings and artificial agents such as
LINGUI computer programs.

STICS By “machinery,” we mean computer

programs as well as the linguistic
knowledge that they contain.
 What does it mean for a computer to
communicate in or interpret a human
language?
COMPU
After all, computers have no inherent
TATION intelligence. Their linguistic capabilities
AL derive from programs that are written
for them. Computational linguistics
LINGUI therefore involves designing and
developing programs to carry out
STICS linguistic tasks. These programs are
based mainly on methods developed by
computer scientists, but they use
linguistic knowledge developed by
linguists.
 Language involves complex
COMPU symbol systems, and computers
are very fast mechanical symbol-
TATION processors. There are natural
connections between linguistic
AL processing and computation,
LINGUI between the complexity of
linguistic patterns and the
STICS complexity of mathematical
models of computation.
 Most people who use computers are familiar
with tools such as spelling and grammar
checkers, as well as Web search engines such
as Google, without realizing that they involve
COMPU aspects of computational linguistics.

TATION Here are some other activities that natural

AL language processing (NLP) systems can do:

LINGUI •speech processing: getting flight information

STICS or booking a hotel over the phone

•information extraction: discovering names

of people, organizations, and places from a
collection of documents
 Here are some other activities that natural
language processing (NLP) systems can do:

COMPU •machine translation: translating a

document from one human language into
TATION another

AL •question answering: finding answers to

LINGUI natural language questions in a collection of
texts or a database
STICS
•summarization: generating a short
biography of someone from one or more news
articles
 Morphology is the study of the
structure of words.

COMPU The task of an automatic morphological

TATION analyzer is to take a word in a language
and break it down into its stem form
AL along with any affixes that it may have
attached to that stem.
LINGUI In processing a sentence such as Jack
STICS reads well, the analyzer should be able
to identify Jack as a proper name, reads
as the third person singular present form
of the verb read (read _ s), and well as
either an adverb or a singular noun.
 Sometimes, instead of a full
morphological analysis, a simple
stemming algorithm is used which
COMPU strips off suffixes to arrive at a stem
form.
TATION
AL A typical strategy for applying
LINGUI morphological rules is the pattern-
action approach. A computer
STICS program identifies words that match
the rule’s pattern; it then records that
word’s morphological components,
as specified by the rule’s action.
COMPU
TATION Given a set of linguistic rules that describe
AL
how elements of a sentence can be put together (a
grammar), a computer program called a syntactic
LINGUI
parser will try to find the best grammatical analysis
of a sentence. If the sentence is ambiguous (that is,
STICS
if it has more than one possible grammatical
structure), the syntactic parser will produce all
analyses.
 Consider this short sentence:
I can fish.

COMPU This sentence could mean that I know how

TATION to fish, or that I habitually put fish in cans. In
the first reading, can is a modal auxiliary
AL verb; in the second sentence, can is the
main verb. These two different meanings
LINGUI correspond to distinct syntactic structures.

STICS We can describe the structure of sentences

in terms of a grammar, expressed as a
system of rules: phrase structure rules
(break up a sentence into its constituent
parts)
 A parser takes an input sentence
and produces one or more
syntactic representations of it. It
COMPU produces a single representation if
the sentence is syntactically
TATION unambiguous, but more than one
AL representation if there is syntactic
ambiguity (as in I can fish).
LINGUI
STICS One way to represent the
hierarchical syntactic structure of a
sentence is called a parse tree.
COMPU It is also possible to train a
TATION machine to discover the rules from
examples of linguistic analyses. For
AL example, in the case of syntax, the
LINGUI linguist may provide only the parse
trees, from which the computer
STICS can “discover” the grammar by
“studying” examples of parse trees.
 Assume that we have carried
COMPU out syntactic parsing of the
TATION utterance and have obtained a
parse tree for it. If we start with the
AL parse tree, we can associate
meanings with the words at the
LINGUI bottom of the tree and use that
STICS information and the structure of
the tree to provide a meaning for
the sentence.
COMPU
TATION
AL
LINGUI
STICS
 How is a computer to decide which
meaning is intended?

COMPU
A program called a word-sense
TATION disambiguator is used for this. A
AL word-sense disambiguator
(Kilgarriff and Palmer 2000) can
LINGUI use the context of neighboring
STICS words in the sentence as well as
other words in the document to
figure out which meaning of a
given word is most likely.
 A common assumption
COMPU underlying computational
approaches to semantics is that the
TATION meaning of the whole is
AL systematically composed of the
meaning of the parts. A computer
LINGUI program based on compositional
semantics puts together the
STICS sentence meaning from the
meanings of the words and phrases
that compose it.
 Information extraction: tries to derive
the semantic content of a document as a
whole as it relates to particular types of
events.

COMPU Information extraction programs

automatically fill out templates (or
TATION tables) from natural language input.

AL What is extracted here is not the

LINGUI meaning of individual sentences, but
information about specific entities
STICS referenced in the text, such as events,
people, organizations, locations, and
times (these are called named entities)
and particular relations between them.
 Speech recognition: uses a
dictionary that lists pronunciations
COMPU for words in its vocabulary,
expressed as sequences of
TATION phonemes.
AL
LINGUI A speech recognizer uses several
knowledge sources: an acoustic
STICS model, a pronunciation model, and
a language model (the latter
functions as a syntactic recognizer).
 Speech recognition
The acoustic model predicts the
likelihood of a particular input sound
wave matching a particular sequence
COMPU of phonetic segments.
TATION The pronunciation model uses the
acoustic model’s results to predict the
AL likelihood of a particular sequence of
LINGUI phones matching a particular word.
The language model then predicts
STICS the probability of each sequence of
words identified by the pronunciation
model.
COMPU
TATION
Speech
AL synthesis
Speech
LINGUI synthesis systems usually take
written text as input and produce speech.
STICS
 Machine translation (MT) is the
oldest application area of
computational linguistics, dating
back to the early years of the Cold
War.

COMPU Unlike the translation of literary

TATION texts, where a considerable
amount of creativity is required on
AL the part of the translator, MT is
LINGUI focused on translations which
preserve the information content
STICS of the source language as much as
possible, while rendering it in a
natural form in the target
language.
COMPU
MT poses challenges to many areas of
TATION
computational linguistics, since it involves
AL
understanding of utterances in one
LINGUI
language and generation of utterances in
the other. MT systems all tend to translate
STICS
sentence by sentence, and, with a few
exceptions, do so while ignoring discourse
context.
 There are three general
approaches to MT:

COMPU •A direct approach looks up words

TATION from the source utterance in a
AL bilingual dictionary, chooses the
most appropriate word translations
LINGUI by some method, and then
reorders the chosen word
STICS translations based on the target
language word order.
 There are three general
approaches to MT:

COMPU •A transfer approach (which is the

TATION most widely used) builds an
intermediate syntactic or semantic
AL representation of the sentence in
the source language, and then
LINGUI maps it to a syntactic or semantic
STICS representation in the target
language, from which a target
language sentence is generated.
 There are three general
approaches to MT:

COMPU •An interlingual approach

TATION attempts to decompose the
AL meaning of the source
language utterance into a
LINGUI language-neutral conceptual
STICS representation, from which
the target language sentence
is directly generated.
 Most practical systems use a
combination of these different
COMPU approaches along with corpus
TATION statistics to achieve more
robustness. While computer
AL algorithms are very well
LINGUI understood, getting computers to
acquire some of the linguistic and
STICS world knowledge needed by those
algorithms is a major challenge,
even with the very large corpora
now available.
 Assignment:
Provide three examples
COMPU (from English to Turkish or
from Turkish to English)
TATION where machine-translation
AL accidentally produce funny
LINGUI results.
STICS
 NEXTWEEK:

DISCOURSEANALYSIS
Discourse
Analysis
 "Discourse is the way in which language is used
socially to convey broad historical meanings. It is
Discourse
language identified by the social conditions of its use,
Analysis
by who is using it and under what conditions.
Language can never be 'neutral' because it bridges
our personal and social worlds" (Henry & Tator,
2002).
 Discourse: is the language beyond the sentence
Discourse
identified by
Analysis
1) the social conditions of its use,
2) by who is using it and,
3) under what conditions
Discourse The study of discourse is entirely context-
Analysisdependent because conversation involves situational
knowledge beyond just the words spoken.
(thoughtco.com)
 Cohesion: refers to the use of linguistic devices to
Discourse
join sentences together (including conjunctions,
Analysisreference words, substitution, and lexical devices
such as repetition of words, collocations and lexical
groups) and gives a written or spoken text unity and
purpose.
Discourse
Coherence: refers to the general sense that a text
Analysismakes sense through the organization of its content
and gives a written or spoken text unity and
purpose.
Discourse My father bought a Lincoln convertible. The
Analysiscar driven by the police was red. That color doesn’t
suit her. She consists of three letters. However, a
letter isn’t as fast as a phone call.
 My father bought a Lincoln convertible. He did
Discourse
it by saving every penny he could. That car would be
Analysisworth a fortune nowadays. However, he sold it to help
pay for my college education. Sometimes I think I’d
rather have the convertible.
 In a speech community, communication is
Discourse
organized in terms of events. Hymes (1972) defines
Analysisspeech events as «activities or aspects of activities
that are directly governed by rules or norms for the
use of speech» (p. 56)
 Conversation analysis is a systematic analysis of talk that is
produced as a result of normal everyday interactions …
Discourse
According to this approach, social actions hold some sort of
Analysis
significance to those who initiate them.
Conversation analysis … tries to understand the hidden
rules, meanings or structures that create such an order in a
conversation. (communicationtheory.org)
 A turn is the time when a speaker is talking
and turn-taking is the skill of knowing when to start
Discourse
and finish a turn in a conversation. There are many
Analysisways that speakers manage turn-taking and they vary
in different cultures: e.g. intonation, grammatical
structures, utterances such as 'ah', 'mm' and 'you
know', body language and gestures.
 The cooperative principle is a principle of
Discourse
conversation that was proposed by Grice (1975),
stating that participants expect that each will make a
Analysis
“conversational contribution such as is required, at the
stage at which it occurs, by the accepted purpose or
direction of the talk exchange” (p. 45).
 The four conversational maxims of Grice’s
cooperative principle:
 Quantity: Say no less or more than the conversation
Discourse
requires.
Analysis
 Quality: Do not say things for which you lack evidence.
 Manner: Do not be ambiguous. Be brief.
 Relevance: Be relevant.
Discourse
AnalysisHedges can be defined as words of phrases used to
indicate that we are not really sure what we are
saying is sufficiently correct or complete.
 A parent questions a teenage child on a Sunday
morning. He says, “What time did you come home last
night?” The teenager might respond in a number of ways:
Discourse
“I got home at midnight.”
Analysis
“I got home at around midnight.”
“I got home at midnight, I think.”
“I got home at, like, midnight.”
 As far as I know …
I am not absolutely sure …

DiscourseCorrect me if I am wrong …
Analysis almost
maybe
somewhat
think/feel
possible/likely
 Conversational implicatures:
(i) are implied by the speaker in making an utterance;
Discourse
Analysis
(ii) are part of the content of the utterance, but (iii) do
not contribute to direct (or explicit) utterance content
and (iv) are not encoded by the linguistic meaning of
what has been uttered.
 Carol: Are you coming to the party tonight?
Discourse
Lara: I’ve got an exam tomorrow.
Analysis
Lara’s response adhering to the maxims of
Relation and Quantity.
 Background knowledge is information that is not in the text,
but is used from memory by the reader to understand the
Discourse
text.
Analysis
A schema is a general term for a conventional knowledge
structure that exists in memory (e.g., classroom schema).
A script is a series of conventional actions that take place
(e.g., eating at a restaurant).
 The language with which we choose to
Discourse
express ourselves and the contexts in which we do
Analysisso displays our social identities and group
affiliations. Thus, discourse analysis can lead to a
better understanding of the values and social
practices of a community (Wennerstrom, 2003).
 To summarize, discourse analysis involves the
study of naturally occurring language in the context in
Discourse
which it is used. Discourse analysts, whether they are
Analysis
concerned with the coherence of extended structure or
with the interpretation of more minute texts, are
interested in the language choices people make to
accomplish their social goals (Wennerstrom, 2003) .
Discourse Analysis
 Critical discourse analysis (CDA) is a type of
Discourse
discourse analytical research that primarily studies the
Analysis
way social power abuse, dominance, and inequality are
enacted, reproduced, and resisted by text and talk in
the social and political context (Van Dijk, 2001).
 Fairclough and Wodak (1997: 271–80) summarize the main
tenets of CDA as follows:
1. CDA addresses social problems
2. Power relations are discursive
Discourse
3. Discourse constitutes society and culture
Analysis
4. Discourse does ideological work
5. Discourse is historical
6. The link between text and society is mediated
7. Discourse analysis is interpretative and explanatory
8. Discourse is a form of social action.
 Some research in CDA:
• Gender inequality
• Media discourse
Discourse
• Political discourse
Analysis•
From group domination to professional and
institutional power
• Discourse in educational settings
 Topics of discourse analysis in school settings:
• Classroom interaction as cultural practice
Discourse
• Classroom discourse and literacy development
Analysis
• Discourse study of second language development
• Classroom discourse as learning
• School as a venue for talk
 “Knowledge and Classroom Discourse” by Jocuns, A.
(2012)
Discourse One approach to classroom discourse has been to
Analysis
examine not only the language that is used within the
classroom or educational setting but the myriad of ways in
which knowledge is constructed, displayed, or both during
social interaction within the classroom, not to mention
content knowledge. …
 “Knowledge and Classroom Discourse” by Jocuns, A.
(2012)
Discourse
… Such studies have examined the metacommunicative
Analysis
knowledge at a student’s disposal during learning
interactions including verbal and nonverbal discourse,
and the participant structures available for a student to
use in a specific educational setting (Jocuns, 2007,
2009). …
 “Knowledge and Classroom Discourse” by Jocuns, A. (2012)
… The crux of such studies is that they emphasize that
Discourse
students learn more than just content knowledge during
Analysis
classroom interaction: they also learn the metacommunicative
means by which to construct knowledge, as well as the ability
to develop and maintain social roles and identities. What is
more, through classroom discourse students play a role in
constructing content knowledge.
Discourse
Analysis
Discourse
Analysis and
Language
Teaching It is widely accepted in the field that we
teach both “language for communication” and
“language as communication.”
Discourse Discourse analysis and pragmatics are relevant to
Analysislanguage
and teaching and language learning since they
represent two related discourse worlds that characterize
Languagehuman communication. The first represents intended
Teachingmeaning transmitted within context, and is, therefore,
concerned with sequential relationships in production; and
the other explains the interpreted meaning resulting from
linguistic processing and social interaction (Olshtain & Celce-
Murcia, 2001).
Discourse
Analysis and The competent language teacher can no
Languagelonger limit herself or himself to being an educator
Teachingand a grammarian. To a certain extent, she or he
also has to be a sociolinguist, aware of and
interested in various aspects of discourse analysis
(Olshtain & Celce-Murcia, 2001).
Discourse
Analysis and A discourse-based model for language
Languagepedagogy perceives shared knowledge as consisting
Teachingof layers of mutually understood subcategories:
content knowledge, context knowledge, linguistic
knowledge, discourse knowledge, etc. (Johns,
1997).
 The discourse analysis of oral interaction is highly relevant to
the teaching of pronunciation in a communicative classroom:
Discourse
• the difference between new and old information is signaled via
Analysisprosody,
and
• contrast and contradiction are also marked by a shift of focus in the
Language
ongoing discourse,
Teaching
• students need to be alerted to similarities and differences in rhythm
and intonation between their NL and the TL,
• the social functions of intonation, which may reveal things such as the
speaker’s degree of interest or involvement, the speaker’s expression
of sarcasm, etc. should be taught.
 Knowing grammar can no longer mean knowing only
Discourse
how a form functions within a given sentence, but must also
Analysis and discourse features of grammatical forms.
include
Language It is the context-dependent, pragmatic rules of
Teaching
grammar that play an important role in a discourse approach
to grammar. In English, such grammatical choices as passive
versus active voice, sentential position of adverbs, tense–
aspect–modality sequences, and article use, among others, are
context-dependent (Olshtain & Celce-Murcia, 2001).
 In the teaching and learning of vocabulary the discourse
perspective stands out very clearly.

Discourse Vocabulary cannot be taught or learned out of context. It is

only within larger pieces of discourse that the intended meaning of
Analysis andbecomes clear.
words
Language
Teaching
He got the ax and chopped down the tree.
He got the ax so now he is looking for another job.

(Olshtain & Celce-Murcia, 2001).

NEXT WEEK

REVISION
WEEK 14

REVISION
 First Language Acquisition
ACQUISITION
Language acquisition is the process by which humans
acquire the capacity to perceive and comprehend language (in
other words, gain the ability to be aware of language and to
understand it), as well as to produce and use words and
sentences to communicate.

Language acquisition involves structures, rules and

representation. The capacity to successfully use language
requires one to acquire a range of tools including phonology,
morphology, syntax, semantics, and an extensive vocabulary.
 First Language Acquisition
INPUT
Under normal circumstances, human infants are
certainly helped in their language acquisition by the
typical behavior of older children and adults in the
environment who provide language samples: the input.

Do not forget: there is no overt or formal instruction

involved in first language acquisition.
 First Language Acquisition
How do children acquire the language?

• Do children learn through imitation?

• Do children learn through correction and
reinforcement?
• Do children learn through structured input?
• The Innateness Hypothesis
 First Language Acquisition
Stages in Language Acquisition

The Acquisition Schedule:

• Cooing and babbling
• The one-word stage
• The two-word stage
• Telegraphic speech
Second Language Acquisition vs Learning

A “foreign language” setting means learning a

language that is not generally spoken in the surrounding
community.

A “second language” setting means learning a

language that is spoken in the surrounding community.
 Second Language Acquisition vs Learning
What is the distinction between acquisition and learning?

Acquisition is used to refer to the gradual

development of ability in a language by using it naturally
in communicative situations with others who know the
language.
Learning applies to a more conscious process of
accumulating knowledge of the features, such as
vocabulary and grammar, of a language, typically in an
institutional setting.
Second Language Acquisition vs Learning:
Focus on Method

• The grammar–translation method

• The audiolingual method

• Communicative approaches
Second Language Acquisition vs Learning:
Focus on the Learner

• Transfer
• Interlanguage
• Motivation
• Input and output
• Fossilization
Second Language Acquisition vs Learning:
Communicative Competence
 Communicative competence can be defined as the
general ability to use language accurately, appropriately,
and flexibly.

 It has three components:

 Grammatical competence
 Sociolinguistic competence
 Strategic competence
 Focus on form assumes that acquisition occurs best
when learners' attention is drawn to language items when they
are needed for communication.
Focus on forms emphasizes the role of explicit
knowledge in the acquisition process.
Types of focus on form include input flood, input
enhancement, and corrective feedback.
Types of focus on forms include present, practice,
produce (PPP) and explicit language instruction. Instructional
methods with aspects of both types of instruction include
consciousness‐raising activities and input‐based instruction.
 TYPES OF CLASSIFICATION

Within the field of linguistics, three different

approaches to language classification are used.

A first approach, linguistic

typology, classifies languages according to their
structural characteristics. A second approach, genetic
classification, categorizes languages according to their
descent. Finally, areal classification identifies
characteristics shared by languages that are in
geographical contact.
Gestures and Sign Languages
 Gestures
 Types of sign languages
 Oralism
 Signed English
 Origins of ASL
 The structure of signs
 The meaning of signs
 Representing signs
 ASL as a natural language
• Writing
• Pictograms and ideograms
• Logograms
• Rebus writing
• Syllabic writing
• Alphabetic writing
• Written English
• English orthography
 Language History and Change

• Family trees
• Comparative reconstruction
• The history of English
• Diachronic and synchronic variation
 Indo-European is the
language family with the
largest population and
distribution in the world.
Language and Regional Variation

 The standard language

 Accent and dialect
 Dialectology
 Bilingualism and diglossia
 Language planning
 Pidgins and creoles
 Language and Social Variation

 Sociolinguistics
 Speech style and style-shifting
 African American English
  Culture
 Categories
Language  Linguistic relativity
and Culture  Cognitive categories
 Social categories
 Gender
 Psycholinguistics: The Study of Language Processing
Many psycholinguists conceive of the mental lexicon as a collection of individual units as in the figure. In this figure, the lexicon is shown as a space in which entries of
different types are stored and linked together.The main questions that are asked about the mental lexicon are these: (1) How are entries linked? (2) How are entries
accessed? (3) What information is contained in an entry?
Clinical Linguistics: a broad area of academic and
clinical investigation that overlaps with medical and
other fields of study (e.g. neurology, psychology) and
that demands a sound understanding of language at
all of its levels.
CLINICAL LINGUISTICS
Computational linguistics seeks to develop the
computational machinery needed for an agent to exhibit
various forms of linguistic behavior.

By “agent,” we mean both human beings and artificial

agents such as computer programs.

By “machinery,” we mean computer programs as well as

the linguistic knowledge that they contain.
 What does it mean for a computer to communicate in or
interpret a human language?

After all, computers have no inherent intelligence.

Their linguistic capabilities derive from programs that are
written for them. Computational linguistics therefore
involves designing and developing programs to carry out
linguistic tasks. These programs are based mainly on
methods developed by computer scientists, but they use
linguistic knowledge developed by linguists.
Discourse
"Discourse is the way in which language is used socially to
Analysis
convey broad historical meanings. It is language identified by the
social conditions of its use, by who is using it and under what
conditions. Language can never be 'neutral' because it bridges
our personal and social worlds" (Henry & Tator, 2002).
  Cohesion
 Coherence
Discourse
 Conversation Analysis
Analysis
 Turn-taking
 Cooperative Principle
 Hedges
 Conversational Implicatures
 Critical Discourse Analysis (CDA)
SEE YOU NEXT YEAR!