Language and Communication

16 min readJul 30, 2021

Introduction

The root of the word “communication” in Latin is communicare, which means to share, or to make common. It is defined as the process of understanding and sharing meaning.
Language is the principal method of human communication, consisting of words used in a structured and conventional way and conveyed by speech, writing, or gesture.

It is quite clear that language is everywhere, from casual conversations to literature and social media, language is essential for human communication to take place. The question arises when one thinks about what exactly a language is, is it just a dictionary of words? a dictionary of words along with a set of grammatical rules? or something more? meanings? Context? In this paper, we will be analyzing language as a whole, how it is divided into interconnected levels and structures, and how the structures are never sufficient for successful communication to take place. if language is just a set of rules and words, can a person simply learn them and engage in successful communication? Some people might agree, but we will see why this is not the case.

Let’s code a robot that can communicate with humans, by giving it the power of language. Our focus is mainly on the linguistic features that we would like the bot to have, so let's not get into the technical details. Let’s start by looking at few basic design features of language (Also known as Hockett’s features) that we would like our bot to have before we get deeper into language.

Displacement: The ability to talk about things beyond here and now. Bee’s perform waggle dance to communicate with fellow bees whenever they find food, but they cannot communicate about the time and place of the event, we wouldn’t want our bot to communicate so less would we?
Reflexivity: Using language to talk about language. Emojis lack the reflexivity feature, meaning you cannot talk about emojis using emojis. I don't think we would like a bot that cant talk about itself or language.
Duality of Patterning: words are made up of 2 main levels of structures: forms and combinations. Parrots can mimic what you say but they don't really understand it when they say it, they lack the duality of patterning. This is one of the main reasons why robots fail to communicate with humans.
Arbitrary sign: Every Language has its own alphabet and dictionary, which people use when they speak or write. A dog in English is a kutta in Hindi and a kukka in Telugu, but they all mean the same thing.

Recursion: Repeating the same process again and again. Language has this feature which was once strongly considered a universal grammar rule. The process of Embedding one noun phrase/sentence into another noun phrase/sentence is called recursion.

eg: [[[[The cat's] tail's ] color's] brightness], Another beautiful example for recursion is This Is the House That Jack Built - Wikipedia.

Now that we know the basic design features of language, let's get into more detail and study language layer by layer.

Language and its layers

Morphology

This is the wug test.

If our bot has answered Wugs, Then we can rest assure that it already has the knowledge of morphology, the study of forms. Let’s look at a tougher example in a different language, In Telugu, Chadavadam = To Study so what is She is studying = ?. The answer to it is Chaduvutundi = She is studying. It's not as simple as adding a letter to the end of a word to transform it, morphology is a bigger domain and deals with all such forms of words. An example of more forms of the same word is given below

Every language has its own morphology, one word in Telugu can even be equivalent to more than 5 words in English.

Syntax

The Study of how languages express relationships between words is called syntax. If I was coding a bot to learn English, I would program it to output sentences(Not all) to be in the form of subject-verb-object, but I would program it to output subject-object-verb if it was a Hindi bot. This tells us that knowing the words of a language is simply not enough, you need to know the grammatical rules of the language as well.

The Syntax of a sentence conveys its meaning. For example:

Prayanikulu(Travelers) oka ammayini camera tho chusaru in Telugu could mean

The travelers saw a girl who has a camera with her, or
The travelers saw a girl through a camera

We can use Phrase Structure Grammars to disambiguate these sentences,

Prayanikulu [oka ammayini camera tho] [chusaru]
Prayanikulu [oka ammayini] [camera tho chusaru]

by bracketing [oka ammayini camera tho] = [girl who has a camera with her] and

[camera tho chusaru] = [Saw through a camera] we are significantly changing the meaning of the sentence.

We would like our bot to understand and bracket sentences properly so that it doesnt interpret the wrong meaning.

Now that we have looked at how each language has its own morphological, phonetic, and syntactic features, let’s assume we have coded the bot to successfully be able to use all these words and grammar of a language. It’s time to test it now, let’s try to generate a random sentence using the words and grammar of Telugu(A parallel example in English is given).

Nalla Paalu Lechi Arustunnayi. (Black Milk woke up and is shouting)
colorless green ideas sleep furiously.

This was the bot's output. Did we fail?…

Semantics

The sentence Nalla Paalu Lechi Arustunnayi(Black Milk woke up and is shouting) is perfect when u think about the grammar and rules of the Telugu language, but does it really make any sense? Nope. It’s missing the very essence of communication, meaning. If the bot knew that:

Milk is not Black.
Milk is not something that can wake up, since it cannot sleep.
Milk cannot shout.

It would not have made this sentence.

The branch dealing with meanings is Semantics. For the bot to know what each word means, it would have to refer to a dictionary, not a big deal! but dictionaries are not always the only things needed to understand what a word means. To understand that, let's think about the different types of meaning:

Conceptual meaning(Denotative meaning): The logical/denotative/cognitive meaning of a word; indicates how we describe the outer world in a linguistic system.
Contrastiveness: meanings can be analyzed as being constituted of collections of features. For example: woman: +human –male +adult man: +human +male +adult
Structure: The principle of building up to larger linguistic units out of smaller units. for example: In the system of animals, rabbits and elephants are subdivisions known as hyponyms and animals are the hypernyms. These terms are relative, meaning rabbits is the hypernym of brown rabbits
Connotative meaning: This meaning associates real-world experience with an expression. for example: the word “Desham” in Telugu has a typical feature of “woman”, Putative meaning like “Mother” and Psychological property like “Home”. Unlike a conceptual meaning, connotations are variable across time and space; they are peripheral/unstable
Social meaning: This is the aspect of communication that has to do with the social situation or context. Languages vary along with different parameters: (i) Dialects (regional/social): Regional examples are Andhra/Telangana Telugu
Affective meaning: Reflects the personal feelings of the speaker.

Sense Relations

If I were to tell the bot Icecream naaku in Telugu, It would not know whether to lick the ice cream or to give me the ice cream, since naaku could mean "To me" or "Lick".

Sense Relations: Two or more words can

Have the same phonetic form but unrelated meaning like not - knot, sun-son, lap = thigh/racing lap, Cheppu = Slipper/Tell, Raayi = Stone/Write [Homonymy]
Have same semantic form but related meaning eg: Shiksha, Phalam, Dikku [Polysemy]
include same or similar meanings like Euphemisms: Paiki Povadam(Die, pass away, expire), Social dialects: kuko/kurcho [Meaning Inclusion]
be similar in some contexts but not in the other eg: maarna, kottadam [Near Synonymy]
have opposite meanings eg: vedi, chali [Antonymy]
have a part-whole relationship eg: car and engine. [Meronymy] etc.

It is often very difficult to know what each word means due to this. Eleanor Rosch came up with the Prototype theory which says each word can have a prototype associated with it. for example A chair can be visualized like this

but A chair can have 3 legs or be in the form of a banana.

Words are not the only ones one needs to care about, Sentence meaning plays a major role in communication as well. Sentence Relations are examined in terms of truth values(Logical).

Synonymy: X and Y are synonymous if they have the same truth values. For example, I own the phone and The phone belongs to me.
Entailment: X entails Y if X implies Y holds. For example, She bought a new car entails She bought a new automobile. Hyponymy is a common source of entailment.
Contradiction: X is invariably false. For example, This bachelor has a pretty wife.
Presupposition: It appears to be like entailment, but it borders on the realm of real-world content — the presupposed statement is always true. For example, I found the meeting boring presupposes I attended the meeting; but so does I did not find the meeting boring.

Another important difference between presupposition and entailment is that entailment only applies to assertions, while presupposition can hold between other types of utterances as well.

Tautology: X is semantically vacuous and carries no real information as it is invariably true. For example, This bachelor has no wife. Such sentences can be used in speech because of connotations; for instance, Boys will be boys.
Anomaly: X presupposes a contradiction. For example, He chased his own widow out of the house.

Now that we have completed associating meaning with language, assuming we incorporate all these features in our bot, it should be ready to communicate, or is it…?

Let’s test it! (the examples at right are in Telugu and have the same meaning as the English ones)

me: can I have a cookie? [Naaku oka cookie dorukutunda?]
bot: I dont know [Naaku teliyadu]
me: Look there [Akkada choodu]
bot: where? [Ekkada?]
me: what a wonderful bot!
bot: Thank you

Pragmatics

There is still one essential thing missing for meaning, Context. Pragmatics is the study of how context contributes to meaning. Semantics was all about “What does X mean?”, Pragmatics is “What do u mean by X?”.

Deixis

Pointing via Language is called Deixis, In the below question, it is obvious that context is needed to understand what is going on. “there” could mean anywhere!

me: Look there
bot: where?

Deixis is all about the usage of words and phrases to refer to a specific time, place, or person in context.

There are three main types of deixis:

Spatial deixis is related to the usage of words to refer to spatial locations (physically or psychologically close or away from either speaker or hearer)
[idhi]this, [ikkada]here(can be seen) and [adhi]that, [akkada]there(can't be seen anymore), [raa]come and [po]go
Temporal deixis is related to the usage of words to refer to various times involved with respect to the speaker.
[ippudu]now (near speaker), [appudu]then, away from the speaker (both future and past)
Personal deixis is related to the usage of words to refer to grammatical persons involved in an utterance(1st person, 2nd person, 3rd person)
[nenu]I, you[nuvvu/miru], he[athanu], she[thanu], it[adhi], they[vaaru/vaallu].
special forms:
social deixis, Sie (T-V): the words/phrases used in one language need not refer to the exact same grammatical person in another language.
eg: In Telugu, you can be either nuvvu (informal) or miru (formal). Hindi has 3 such distinction where you = tum/tu/aap
The paper “The Pronouns of Power and Solidarity” by R. Brown and A. Gilman talks about these relations in detail.
If a person/bot is learning and willing to communicate in a particular language, it is important that they also have knowledge over the social aspects of language, like social deixis in this example.

Anaphora

You might have not noticed why many words like “these”, “him”, “her” show up in the text, but it's actually quite often that this happens. anaphora is the repetition of a word or words at the beginning of successive phrases, clauses, or lines to create a sonic effect. For example: If I were to say “It is a relative of Deixis”, you might have thought It = Anaphora quite immediately, this is anaphora and it is often used in political speeches and prose and poetry. It is a rhetorical device that relies on context to express its meaning.

Speech Acts

Language is used to do things, This term(Speech Acts) was introduced in How to Do Things with Words (Austin, 1975)

me: can I have a cookie?
bot: I dont know

Here, Language is being used to make a request(please give me the cookie), which the bot fails to recognize and interprets it as if the speaker is questioning their ability to have the cookie.

Language can be used for acts such as orders, requests, warnings, advice, in other words, do things that go beyond the literal meaning of what we say. Examples:
Eg: No Smoking = order
andar toh bahot garmi hai(It’s so hot in here) = turn on the air conditioner
kya mai aapka order le sakta hu?(Can I take your order?) = mujhe batayi aapko kya chahiye(Please tell me what you want)
can I have a whopper with egg = Please give me a whopper with egg
Do u fancy a tea, darling? = Would you be willing to pay a visit to our house? (inviting)

Force

Austin gives 3 basic senses in which “saying something” is “doing something”:

Locutionary act/force: the speaker utters a sentence with determinate sense and reference.
Illocutionary act/force: It is the action intended by the speaker. This is what the term “speech acts” often refers to.
Perlocutionary act/force: It is the effect or “take-up” of an illocutionary act on the participants.

Perlocutionary force is not particularly linguistic in nature. For example, for the utterance Shoot her! the illocutionary force is of ordering the addressee. But it has the perlocutionary effect of forcing or frightening the addressee into shooting her.

Searle gives 5 types of speech acts named representatives, directives, commisssives, expressives, declarations. Each of these speech acts represents various kinds of “Acts” the speaker would like to perform with language.

Searle also developed Austin’s Felicity conditions into a classification of conditions necessary for a successful speech act: preparatory, propositional, sincerity, and essential. For example, for the act of promising, the conditions are:

preparatory
H would prefer S’s doing A to S’s not doing A, and S believes this.
it is not obvious to both S and H that S will do A in the normal course of events.
propositional — in expressing P, S predicates a future act A of S.
sincerity — S intends to do A.
essential — the utterance e counts as an undertaking to do A.

The ability to “do something” with speech acts, like a warning is also an essential feature for our bot to have, since it needs to communicate, either for performing speech acts or understanding speech acts

Cooperative Principle

me: what a wonderful bot!
bot: Thank you

If the bot had understood that the sentence was sarcastic, it would have responded in a different way. This is because it does not follow the same cooperative principle as humans, Let’s see how.

The cooperative principle is divided into Grice’s four maxims of conversation, called the Gricean maxims. It is assumed that these maxims are required for two people in a conversation to cooperate.

The Maxim of Quality(assumption of truth): Make your contribution one that is true.
The Maxim of Quantity — Do not make your contribution more informative than is required.
The Maxim of Relevance: Make your contribution relevant to the conversation.
The Maxim of Manner: Be perspicuous; specifically, avoid obscurity and ambiguity, and be brief and orderly.
but wait!, when I say “what a wonderful bot!”, I’m directly violating the maxim of quality because I’m being sarcastic(lying). Then I expect the bot to understand what I’m saying, in this case, I’m asking for Conversational Implicature.

Conversation is a 2-way thing and not a 1-way thing, Cooperation is definitely needed for communication. Although Grice had suggested these maxims, no one follows them all the time, the maxims are violated and flouted very often. This happens with the assumption that the speaker and hearer both understand that certain maxims are violated/flouted and respond according to them, this is called conversational implicature. If this does not hold, cooperation breaks, and communication will be hard.

When we look at pragmatic features of language like speech acts, deixis, and cooperative principle, It’s quite clear that these are Extremely hard to encode into a bot and expect it to understand and communicate with humans. This clearly shows us how deeper of a concept Language, compared to words and grammatical rules which were covered earlier. Clearly, understanding context is a difficult task even for humans quite often, Imagine a bot doing it!

Is This Enough?

Let’s think about more complex situations where not only bots, but humans themselves find it very hard to communicate even after having complete knowledge in the words and grammatical rules of a language

Linguistic relativity

Linguistic Relativity, also known as Sapier-Whorf Hypothesis initially said that “Our language determines how we see the world”, we call it the strong hypothesis. But if this was the case, communication between people from different languages would have been too difficult to even imagine. But the theory was not completely false, after careful research, more and more languages had different ways of thinking, for Example:

The Hopi people were believed to have no concept of time in their language.
A Bridge “die Brucke” is considered female in German which is why they use words like [ Beautiful, Slender ] to describe it, in Spanish “El puente “ is considered male so they use words like [ Strong, dangerous] to describe it.
Russians have 2 words for blues (light and dark), and they can easily distinguish between light and dark blues compared to English speakers.
This is also observed when we compare kinship terms of Languages, each of them has different terms to address different people in the family. Ex: Telugu has pinni, peddanaanna, attha, mama, maradalu, bava etc. whereas in English it's just aunt, uncle, and cousins. Each of these relations is treated differently when it came to the religious customs of these languages.
The kuuk thaayorre people only have the concept of direction in their language, and a strong sense of it. They see the world in a different perspective since the relative concept of left, right, straight, etc doesn’t exist. if you were to ask a kuuk thaayorre person to close their eyes and point to the north, all of them would be able to do it very easily, but we don't have such an ability.
The Nootka people referred to “The stone falls” as “It stones down”

All These examples proved that “Our language influences how we see the world”, this is called the weak hypothesis, which is more accepted till today.

When we think about the concept of linguistic relativity, we see how deep language can get and how much it can influence a person, now imagine a conversation between a kuuk thaayorre person and an English person, it would be very difficult to communicate, even if they both know each other's languages, wouldn’t it? This shows us how important semantics are for successful communication.

Associative and Conceptual Engineering

Have you ever noticed how advertisements are engineered in a certain way to build the image of their product or company? This happens through a process called Associative Engineering. One can play with words to change their associations in order to change the connotative meaning of their utterance. Euphemisms are generally used for such tasks, Euphemisms are words that are used to avoid directly addressing things that might be deemed to be negative or embarrassing e.g. death, sex, excretory bodily functions.

thanu chanipoyindi(she died) — thanu devudi deggariki vellipoindi (She Went to visit god)
Toilet — Washroom/Water closet

It’s quite obvious that the right sentences sound better than the left sentences in these examples, why? because we have bad associations with few words which we tend to avoid using, we find better words to denote these words and use them to make our sentences sound better. However, Euphemisms themselves get bad associations from their previous holders eventually, and newer Euphemisms are discovered to keep countering this.

In a similar way, one can manipulate their sentences (and people) using connotative engineering, which is basically using dictionary meaning to either cover/avoid the truth or play safely in a conversation. These are also very common in every language and vary accordingly, which is why it is important to also understand the engineering behind any sentence we see in order to completely understand it.

Politeness Theory

Erving Goffman proposed the term face, as the public self-image one has established for others to recognize, the concept of face is considered fundamental for social interactions as they are about saving or losing face. People usually cooperate in maintaining face in an interaction, since everyone’s face is dependent on everyone else’s.

A face-threatening act(FTA) is an act that damages the face of the addressee or speaker. Our current focus is on speech acts. FTAs on hearers can be things like creating pressure through acts like orders or putting them in a tough spot by compliments, forcing them to accept or reject a proposal, ignoring them, ridicule, or even wrongly addressing them. Cases, where FTAs are on speakers, are acts like apologies or excuses or even accepting compliments.

One simple example would be: Addressing an elderly person as (him)”vaadu”, instead of “vaaru”(him). As we can see both the words of Telugu translate to “him” in English, that's because in this case, the power semantic is not present in English. This was just a simple example, but there are many ways one can perform an FTA without even knowing, especially non-native speakers who find it challenging to adapt to such changes. This is another important part where language is to be understood in more than words and grammatical rules, in the above example, if one does not understand the social reasons behind addressing elders differently, it becomes harder for them to communicate or be polite. There could have been several and worse situations as well.

There are several ways to avoid FTAs, but FTAs occur very often, and to reduce the damage, we use politeness strategies. politeness strategies such as going off the record (Trying to communicate indirectly through linguistic tools like idioms and metaphors). redressing the speaker's utterances by trying to protect a person's Negative face(their desire to be free) or positive face(Their desire to be liked by others) can be used to avoid the damage caused by FTAs.

Clearly, these strategies are very important in communication and are things that are beyond arbitrary symbols (words) or grammatical rules.

Conclusion

For successful communication to take place, it is not sufficient to just know the words and grammatical structures of a Language. We have seen a number of examples where words and grammars mean little to nothing, whereas things like relativity, meaning, context, politeness help in communication. Structures and words are merely tools for communication, which is why the classification “humans” comes above “language”. From linguistic relativity, we have seen how language influences humans, but language itself has its own core and enables communication between humans. Speech acts show that Language can not only say things but do things. From politeness strategies, we see how being polite does not depend on things like morphology or syntax but semantics or pragmatics. Language can be engineered to express ourselves in different ways, all of this with the same dictionary.

Coding a robot that has a list of all words and follows all the syntactic rules of a language is much simpler than making the robot understand the context or meaning of words and sentences, this clearly shows how deep language is, and how it extends beyond arbitrary signs and symbols. So the next time when you want to say “I know this language”, you might want to reconsider what you’re saying…