1.1 What is language? What is knowledge of language?

When we think about what makes us human, one of the first things that might come
to mind is language. Language is our special ability to speak and understand words
so we can communicate with each other. It's something that only humans can do,
and it's very important for how we evolved as a species. A well-known psychologist
named Elizabeth Spelke once talked about this in a TV show called The Human
Spark. She explained that even though young human babies are similar to young
animals in many ways, it's language that really shows what's unique about humans.
This starts to show when babies begin to pick up words at around 9 or 10 months
old and then start putting these words together a little later.
Another unique thing about humans is our ability to learn more than one language—
not just the language we grow up speaking at home, but also other languages.
Some people grow up knowing two or more languages from childhood, but many
learn new languages as kids or even as adults. The ability to know and use more
than one language, called multilingualism, is a big part of being human.
Research shows that there are more people in the world who speak two or more
languages than people who only speak one. For example, in the European Union,
over half the people can speak at least one other language besides their native
language. In North America, many Canadians and a good number of Americans can
do this too. These numbers are expected to rise as the world becomes more
connected.
This book focuses on how different languages are stored in our minds and brains
when we speak, write, listen, and understand them. Linguistics, which is the study
of language, asks some big questions like:
 What do we really know when we know a language?
 How do we learn this knowledge?
 How do we use this knowledge in everyday life?
 How is this knowledge organized in our brain?
 How did the ability to use language develop in humans over time?

In this text, we don't have complete answers to all the big questions about
language. The very last question we asked won't really be discussed much
here. But, the first four questions are very important to understand how
people learn a second language, so we will start with the simple parts.

A key idea about language is that we can make endless sentences in the
languages we know. This amazing ability comes from a limited number of
sentences we hear while learning and a set of rules about what makes a
sentence correct. Here, I use the word "sentence" as an example of a
language unit, but this idea also applies to smaller or bigger language units
like phrases or longer discussions.
Since the 1950s, thanks to Noam Chomsky, these five questions have been
central in the study of generative linguistics. This approach suggests that our
ability to use language is just another normal feature of humans, similar to
seeing or hearing. For Chomsky and other scientists who follow his ideas,
language is a "natural object." This perspective led to the creation of a new
field called "biolinguistics."

In 2005, Chomsky wrote an important article named "Three factors in

language design." He described three main things that help language grow
in a person:
a. Genetic endowment
b. Experience
c. Principles not specific to language.

Language knowledge is stored in our mind and brain. It mainly depends on

something we are born with, called Universal Grammar, which means we
have some language knowledge that we don't need to learn because it's
already in us and is the same for all languages. This is what Chomsky means
by "genetic endowment." The second thing is our environment: we need to
hear language used around us to learn it properly. The third thing includes
general rules of good design that help us analyze data and compute
efficiently, and these aren't just about language.

In this book, we will talk a lot about these three things because they are
essential to understand how people learn a second language, often
shortened to L2A or SLA.

The second factor, which is the language we hear around us, is very important for
learning language. We call this "linguistic input." An example of this could be when
we talk directly to babies in a way they can understand—this is known as "infant-
directed speech." When we can understand what is being said because it connects
with what we are seeing or experiencing, this is called "comprehensible input." For
instance, if someone says "The dog wants to go out" while a dog is waiting by the
door, we connect these words with what is happening. This connection between
words and the world around us is essential for learning language because neither
words alone nor the world alone can teach us language without the other.
Children learn the language of their environment. For example, a child born to
American parents but adopted by Brazilian parents will grow up speaking Brazilian
Portuguese. A child in Quebec with one French-speaking parent and one English-
speaking parent might grow up speaking both languages. Similarly, a child of Polish
parents in the UK might start by speaking Polish, but will likely learn English when
starting school, becoming bilingual. How much they use Polish and English later
might change depending on the language they hear most often.
While it's clear that the language we learn comes from our experiences, whether we
also have a built-in language ability called Universal Grammar is debated. Some
experts think we are born with a specific skill just for learning language. Others
believe our ability to learn comes from general skills that help us notice patterns in
what we hear, which helps in learning many things, not just language. This idea is
called the "emergentist" or "usage-based" view, meaning our ability to use
language develops from our experiences with it.
In this book, we focus on the "generative" approach, which has helped discover
many things about how we learn a second language. This approach talks about the
specific rules that make up a language, aiming to explain all the correct ways of
using the language. The term "generative" comes from the idea that these rules
generate all the correct sentences in a language.
Back in the late 1950s, linguists thought they almost knew everything about
languages and that describing them completely was an achievable goal. They also
thought that languages were very different from each other. But Chomsky
challenged these views, showing that languages actually share many similarities.
This led to a realization that we didn't know as much as we thought about these
common features, which are part of Universal Grammar. This started a new phase in
linguistics, almost like starting over to understand these basics.
During the 1970s and 80s, linguists were trying to understand how children learn
languages so quickly and easily without much help. They realized that to explain
this, Universal Grammar (the language knowledge we're born with) must be very
complex and specific. The Principles and Parameters approach, which started in the
early 1980s, suggested that learning a language involves using universal principles
common to all languages and figuring out specific settings, or "parameters," for the
language being learned. These parameters are thought to be preset options
provided by Universal Grammar.
For example, linguist Hagit Borer suggested that these parameters are like
grammatical choices every language has already made. They are stored in our
mental dictionary of functional morphemes—these are the parts of language like
past tense and plural endings. While basic properties of words, such as nouns and
verbs, are learned from experience, functional morphemes are also shaped by this
experience.
How does Universal Grammar help us learn languages? It gives us a set of universal
rules and properties automatically. As linguist Mark Baker says, the more we see
languages as different, the more we find they are actually similar. For example,
most languages have rules about subjects and objects in sentences. A child doesn't
need to learn these rules from scratch because they come naturally, although they
will see these rules confirmed through everyday language use.
What about the differences between languages? Universal Grammar limits and
defines the choices we have when learning a new language. Generative linguists,
who study these patterns, say that language differences are very controlled. A good
way to think about this is like a light switch that can be turned on or off. For
instance, consider the Null Subject Parameter, which was explored by Nina Hyams
in 1986. In English, every sentence needs a clear subject, like in "She ate the pizza."
You can't just say "Ate the pizza." However, in languages like Italian and Spanish,
you can leave out the subject (it's "null") because the verb ending already shows
who is doing the action, as in the Italian "Ha mangiato la pizza," which means
"She/he ate the pizza." The information about who ate the pizza is not in the verb
ending itself but is understood from the context—similar to how we use "she" in
English. This difference between languages like Italian and English is what linguists
call a parameter, which can be either on (null subjects are possible) or off (null
subjects are not possible).
To sum up, according to generative linguistics, Universal Grammar provides a
complete set of rules needed for creating all the correct sentences in a language.
This includes universal rules that apply to all languages, as well as options for
different rules, called parametric options.
It's important to note that the number of sentences we can create in any language
is unlimited, but the rules we use to create these sentences are limited. This creates
a special situation in language: we can say far more things than the rules and words
we've specifically learned. This difference between what we know about language
(our competence) and how we use it (our performance) applies to our first language
and any other languages we learn.
What do we actually know when we know a language? We know enough rules and
words to make any acceptable sentence in that language. This knowledge is what
we call linguistic competence. We don't need to have heard or said every possible
sentence ourselves—that would be impossible because we can always think of new
things to say. This ability to make new sentences is part of our linguistic
performance. Language allows us to be creative, putting together new sentences
whenever we speak, write, or even think.
This concept of limited input but unlimited output also applies to learning
languages. Learners can understand and create sentences they've never heard
once they know the basics of the language. This idea can be compared to knowing a
recipe for baking bread. Once you know the recipe, you can bake many loaves of
bread, each slightly different.
1.2 The language architecture
Earlier, we mentioned that cognitive scientists and linguists find language
fascinating because it's a structured product of the human mind, which helps them
study the mind itself. Understanding how language is structured, or the "language
architecture," helps scientists figure out how language is learned and used in daily
communication. From the start of generative grammar, linguists have been trying to
understand the parts of language and how they come together to form messages.
From now on, I'll use the term "grammar" to refer to the system of rules that forms
the basis of our language knowledge.
Linguistics covers the processes of creating and understanding messages. For
instance, when we speak a sentence, it's made of sounds that turn into sound
waves, like the word "bag" which consists of the sounds [b], [æ], and [g]. In sign
language, similar messages are made through gestures. The study of these sound
waves and speech sounds is called phonetics. When we hear someone talk, we use
rules specific to our language to organize these sounds into meaningful units, which
phonology studies. Phonology looks at the sounds available in a language and the
rules for how they can be combined and affect each other. While phonetics and
phonology are closely related, phonetics focuses on the physical properties of all
human speech sounds, and phonology focuses on how these sounds function within
a particular language.
Let’s break down how we understand and create language, step by step, using
some special terms from the study of language, called linguistics.
First, when we hear someone speaking, our brains pick up the sound waves and
start to break them down into smaller meaningful parts called morphemes. These
are the smallest units in language that have meaning. For example, in the word
“works” from the sentence "Josh works in the library," there are two morphemes:
“work” (which is the main verb) and “-s” (which shows it’s happening now and it’s
Josh who is doing it). This part of language study is called morphology. Inside our
brains, we have a mental dictionary called a lexicon, where we store all the words
and grammatical bits we know.
Now, let’s see how these words and bits come together. Morphemes and words
must be arranged in a certain order to make sense in any language. For example, in
English, the correct order is “Josh works in the library,” but if you jumble it like "Josh
in library the works," it doesn’t make sense—that’s why we mark it with a star (*) to
show it’s wrong. The rules that tell us how to order these words and phrases
correctly in sentences are studied in a field called syntax. Sometimes we talk about
morphosyntax because the rules for ordering words often involve those
grammatical bits, or morphemes, we talked about.
After we put words into the right order in our minds, we then need to understand
what they mean together. This is handled by our Conceptual-Intentional system,
which works out the meanings based on the rules of semantics. Semantics deals
with the meanings of words and sentences. But meanings can change depending on
where and how something is said—that part is studied in another field called
pragmatics.
Let’s look at an example to understand semantics and pragmatics better: Imagine
someone asks, “Have you seen my gym bag?” The basic question seems to be
about whether you’ve seen the bag or not. That’s the semantic part—the literal
meaning. However, pragmatically, the speaker might be asking if you know where
the bag is right now because they need it, or they might be showing off a new gym
bag and looking for your opinion. The context or situation changes the meaning,
and that’s what pragmatics studies.
From the earliest days, linguists have tried to explain how all these parts—sounds,
meanings, word order—fit together. They use the idea that language connects
sounds (which we make and hear) and meanings (which we think and understand).
There’s a classic model in linguistics that shows this connection as a Y shape, where
words are pulled from the mental dictionary and put into order (syntax) until they
split into the sound part (phonetics) and the meaning part (semantics) at a point
called "spell-out." After this point, they don’t mix but go their separate ways to
complete our understanding and speaking process.
This system of breaking down and studying language helps us see how amazing
human language is, allowing us to express limitless ideas with a limited set of
sounds and rules.
In the 1990s, a new approach in linguistics called the Minimalist Program introduced
the idea that sentences are too complex to process all at once. So, linguists started
thinking about breaking down the process of creating sentences into smaller, more
manageable parts. This idea led to what's known as Phase Theory, developed by
Noam Chomsky. According to this theory, constructing a sentence happens in stages
or "phases." Each phase builds a part of the sentence, checks it, and then moves on
to add more parts until the sentence is complete. The main goal of this theory is to
make the process as simple as possible, which is called Minimizing Computation.
Now, not everyone agrees with this approach. Another linguist, Ray Jackendoff, has
a different view. He thinks the model that focuses too much on syntax—the
arrangement of words in sentences—is too limited. He suggests that structure is
created not just in syntax but also in phonology (the study of sounds) and
conceptual structures (ideas and meanings). This means that all parts of language,
from sounds to meanings, are structured at their own levels.
Tanya Reinhart proposed yet another model where semantics (the meaning of words
and sentences) and discourse-pragmatics (how context affects meaning) are treated
as separate layers. This model suggests that meanings are first checked at the
semantics level and then might be adjusted based on the conversation's context, as
we saw in the gym bag example.
Reinhart's model is not syntacto-centric because it doesn't focus only on syntax.
Instead, it gives equal importance to semantics and discourse-pragmatics, showing
that understanding language involves more than just arranging words.
Next, we'll look at how these different parts of language—like syntax and semantics
—interact. These interactions happen at interfaces, places where different language
modules meet and work together. For example, the syntax-semantics interface
ensures that the structure of sentences matches their meaning. This matching is
not always straightforward and requires specific processes that only target relevant
structures for each module. These processes are like a careful matching game,
where each part of language has to align perfectly with another to make sense
overall.

It has been suggested that grammatical operations that occur at the interfaces
between linguistic modules are more complex and demanding because they have to
consider information from two modules, not just one. Moreover, when we look at
languages from around the world, we often find that the same ideas might be
expressed differently. For example, while the English past progressive tense (like
"was going") refers to an action that was ongoing in the past, the Spanish Imperfect
tense can describe both ongoing actions and habitual actions (things done
regularly). On the other hand, the English simple past tense (like "went") can
describe a one-time action or habitual actions, whereas the Spanish preterit tense
only describes one-time actions. So, the basic concepts of ongoing, habitual, and
one-time actions are expressed differently in different languages.
Whether these operations at the interfaces are actually more challenging than other
linguistic operations is a question that can only be answered through experimental
research.
1.3 What exactly has to be acquired?
Why do we focus on these models of language architecture in a textbook about
learning a second language? Understanding the structure of language helps us
grasp what exactly needs to be learned when acquiring a new language. Knowing
about language architecture is especially important for language teachers as it aids
them in guiding students through the learning process.
The structure of language is crucial because it shows us where languages differ. This
understanding allows us to make detailed plans on how someone can learn the
differences between two languages. For instance, under the Minimalist assumption,
differences between languages are mainly in the functional lexicon, while the basic
syntactic operations are the same across languages. Jackendoff argues that the
meaning structures are universal. So, when a child learns her first language, she
doesn’t need to learn syntactic operations anew but does need to learn the
grammatical features specific to her language found in the functional morphology.
This makes learning the functional lexicon one of the key tasks in language
acquisition.
When learning a second language, the situation builds on the familiar structure from
the first language. The learner starts with whatever universal properties of language
they already know, either through Universal Grammar or their native language. The
first step in learning a second language is understanding the new lexicon—the
words of the second language. Then, they must learn the differences in grammatical
structures between their first language (L1) and the second language (L2), like
different word orders or grammatical functions. For instance, in some languages,
question words can be at the beginning of a sentence or remain in their original
position, and these details are related to how sentences are structured in those
languages.
Finally, learning a second language can involve different ways of linking meanings
with grammatical structures, as seen in the differences between English and
Spanish in expressing aspects of actions. Understanding these specific differences
helps learners and teachers anticipate the challenges of learning a new language.
In the following sections, I will provide information on how these theories connect to
teaching practices. For instance, we know that the phonetics or pronunciation of a
second language, if learned after childhood, might not sound exactly like a native
speaker’s. However, the semantics or the meaning and structure of sentences can
often be learned more completely. Therefore, knowledge of these different areas of
language is crucial for teachers to help them understand and address the
challenges learners might face.
1.4 The Scientific Method in Second Language Acquisition (SLA) Research
The scientific method is a system used to study natural phenomena, learn new
things, and improve or integrate existing knowledge. To be considered scientific, a
method must rely on evidence that can be observed and measured and follow
specific reasoning principles. In the field of generative linguistics, this method helps
researchers study language and how it is learned. The process usually involves
looking at data, noticing patterns, creating hypotheses to explain these patterns,
and then testing these hypotheses with more data. The hypotheses are then
updated to reflect new findings and tested again. Different scientific fields might
adjust this method to suit their specific needs, and we'll see how this works in
linguistics.
When linguists want to describe a language that hasn't been documented before,
they start by collecting sentences from native speakers. From this initial data, they
develop hypotheses. For example, they might hypothesize that in this language, the
subject always comes before the object and the verb, because this is common in
many languages. Then, they check this hypothesis with more data. Linguistic theory
helps them predict other language features, like suggesting that in languages with
lots of verb agreement, subjects might often be dropped (not said out loud).
In the study of how people learn a second language using generative methods
(GenSLA), researchers start with basic assumptions and the structure of language
we've discussed earlier, as well as theories on how second languages are learned,
which we will cover in upcoming chapters. Researchers pick a language feature to
study that fits with a specific theory. They look for descriptions of this feature in
both the native language and the second language, or they provide these
descriptions themselves if they are native speakers of these languages. They
typically focus on features that are different between the first (native) and second
language because if there are no differences, there might not be much new to learn.
GenSLA research is driven by theories. Researchers are testing theories or models
that explain how second language learning happens. They don't just look at errors
learners make or how they learn a specific language structure. Their research
questions are always based on theoretical predictions about language behavior and
development. At the end of their studies, researchers aim to contribute knowledge
about the competence in the second language and how it develops.
A typical result of these studies is either support for existing theories or evidence
that challenges them. If a theory is challenged, researchers might propose an
alternative theory that must explain everything the old theory did, plus the new
data.
For example, consider the theory of L1 transfer. This theory suggests that learners
are influenced by the grammar of their native language when learning a new
language. If a learner's first language often drops subjects in sentences, they might
initially do the same in a second language that usually requires subjects. This
prediction fits with the principle of Occam’s razor because it offers a simple and
efficient explanation.
If we were to predict what happens when a learner whose native language always
uses subjects learns a language that often drops them, what would we expect? To
test a prediction like this, researchers gather measurable data from learners over
time or from many learners at once. They also compare these learners to native
speakers to check the accuracy of their methods. If they want to examine different
levels of language skill, they test learners who are beginners, intermediates, and
advanced, and they often use a separate test to measure each learner's proficiency
level.
One of the earliest studies in generative second language acquisition (GenSLA) was
conducted by Lydia White in 1985. She focused on how native Spanish speakers
learn English, particularly looking at a language feature known as the Null Subject
Parameter. In Spanish, it's common to leave out the subject of a sentence (the
person or thing doing the action), but English usually requires the subject to be
stated clearly.
White theorized that Spanish speakers would initially leave out subjects in English
too because of their native language habits. However, she believed that over time,
as they learned more English (the L2 input), and with the help of Universal
Grammar, these learners would adjust and start including subjects in all their
English sentences. She tested learners across five different levels of English
proficiency, from beginners to advanced. She also included a group of French native
speakers learning English for comparison, since French, like English, typically
requires subjects in sentences.
The experiment asked participants to judge sentences with and without subjects.
For example, the sentence "John is greedy. *Eats like a pig." The star indicates that
the second sentence is incorrect in English because it lacks a subject.
The results showed that learners with low English proficiency, especially the Spanish
speakers, had trouble recognizing that the sentences without subjects were
incorrect. Compared to French learners, Spanish speakers found this more
challenging initially. However, as their English improved, so did their ability to
identify sentences that were missing subjects. This supported the hypothesis that
L1 (first language) influences, or "transfers," to L2 (second language) learning, but
can be overcome with enough exposure and learning.
In summary, this study illustrates how the scientific method is fundamental to
GenSLA. This field examines the mental and psychological processes involved in
learning and using a new language, providing insight into how language functions in
the human mind. It demonstrates that GenSLA not only helps us understand how
people learn languages but also offers a glimpse into the complexities of the human
brain.