How does language work?
1
Language in the empirical paradigm
Natural language is a sign system serving cognition and communication that developed
naturally during social interactions between community members. When researching language
in the paradigm of empirical sciences, i.e., assuming it has material causes, we could postulate
either that it is predominantly innate, (genetically transmitted), or that it is a sort of cultural
adaptation, (a result of human development and activity). We questioned the rationale of the
former assumption, when mentioning Chomsky’s theory of language.
Assuming the latter alternative, that of language being a sort of cultural adaptation, as
explicated by Gabriel Altmann in 1978, language must be viewed as “a self-organizing and a
self-regulating system all of whose properties are brought into line with one another in some
way or other”. In other words, language belongs to a wide range of phenomena, constituting a
class of natural self-organizing and self-regulating complex systems, which self-organize and
self-regulate not as a result of some central planning, but spontaneously as a result of
elementary parts’ interacting locally with each other (directly or indirectly) and, typically, also
with their surroundings. Altmann’s proposal was made more specific by Reinhard Koehler,
when he stated that “language system develops in reaction to the properties and requirements
of its environment by adaptation mechanisms in analogy to biological evolution.” (*This
elaborates on Zipf’s approach to modeling language by searching for its material causes,
which was formulated still in the first half of the 20th Century.)
1.1 Language — a self-regulating system evolving in analogy to biological evolution
In Darwin’s theory of evolution, the adaptation of species takes place through preferential
reproduction of organisms with genes promoting a certain type of fitness beneficial in a given
environment. In the case of language adaptation, as reflected in an increasingly more
economic distribution of its linguistic units and creation of new ones, language optimizes
through their speakers’ preferential production of utterances consisting of linguistic entities and
patterns that are the fittest, given the current form of the given language, predominant
communicative needs of its users, and their cognitive capabilities.
Preferential production of the fittest linguistic entities is the result of the omnipresent
Zipf’s principle of The Least Effort, which states that in any human action, including verbal
communication, people attempt to minimize the effort needed to accomplish a given task. In
the case of using language, this implies the drive to minimize language processing effort
understood broadly as the effort required to acquire, keep in memory, comprehend, and
produce language needed for successful verbal communication, taking into account our
biological and social constrains and communicative needs.
1
Note, that what counts towards the fitness of a linguistic entity is its statistical efficiency.
The fittest linguistic solutions for a given language are the ones minimizing overall language
processing effort of all its users over time. This means that not every linguistic utterance
needs to, or even can, be optimal.
1. 2 Reaching a compromise between the speaker and listener
The degree of fitness of a linguistic item may be different from the perspective of the
speaker and that of the listener. For instance, speakers will deem fit constructions requiring as
little effort to construct them as possible. To this end, speakers tend to avoid redundancy,
shorten frequent, predictable words and word patterns, and take linguistic shortcuts, often
merely indicating what is to be said. They prefer to pronounce successive sounds in a more
alike, blurred fashion, would rather have to remember a smaller vocabulary, even at the cost of
these words having multiple meanings. Listeners, on the other hand, want to save effort when
comprehending a given utterance – they prefer to hear utterances, which are precise, given the
purpose of the conversation, and easy to decode even at the cost of their increased length.
Therefore, listeners favor sentences that are more explicit, unambiguous and ones constructed
with some degree of redundancy to balance the noise present during communication process.
They deem fit words pronounced without blurring successive phonemes into each other and
favor language made up of the phonemes distinct enough to be easily identifiable, even at
speakers’ increased cost of producing them. And so on.
In other words the optimization of language requires reaching a compromise between
the needs of speakers and those of the listeners. Importantly, the speaker may attend to his own
needs only as far as this does not curtail listener’s preferences too much and does not result in
the break-down of communication. When communication fails, the speaker usually rephrases
his message using linguistic entities, which are, hopefully, more accommodating to the
listener’s needs.
1.3 Choosing the fittest of linguistic items
Now, how do humans actually go about selecting the fittest entities optimizing language
processing effort? It turns out that only some optimizing choices are conscious. People, for
instance, consciously introduce borrowings from other languages to fill up gaps in labels for
novel objects or actions in their native tongue, or sometimes to show off. Speakers, also
consciously, coin new names for new products, for instance, by making acronyms such as
RADAR, or producing blends such as brunch (breakfast+lunch), in Polish przywitul
(przywitaj+przytul) [welcome +hug], or using brand names as labels for a given category of
items (e.g., Electrolux in Polish for a vacuum cleaner). Occasionally, governing bodies
optimize language by introducing spelling reforms, such as that introduced by Webster in the
USA at the beginning of 19th Century, or a similar one that took place more recently in
Germany. Speakers also voluntarily indulge in linguistic fades, using extensively buzzwords,
or euphemisms, which expands language by giving words additional shades of meanings
— the use of buzzwords and euphemisms informs the listener that the speaker is an insider in
a certain field or … likes to show off. Most of such novelties introduced voluntarily optimize
language by providing the means to convey the content more adequate for a given situation
— more precise and concise, easy to learn and remember.
2
In most part, however, self-regulation of language, ensuring both desired precision of the
information conveyed and reducing language processing effort of the interlocutors, comes
from unconscious processes, which result in some entities being more likely to be
replicated in a given situation — more likely to come to one’s mind. For instance the
more fit — the more economical and likely to be replicated — is the entity, which is easier to
remember (retain in memory), and with a shorter recall time. And easier to remember and
with a shorter recall will be a piece of information, which is more relevant to us, emotionally
tinted, interconnected with many other pieces of information, mentioned as the first or most
recent item in a series of items, one that is more often repeated, needed, resembling something
known, yet distinct enough so as not to be confused with another entity. A more economical
solution will also be one saving memory space, easier to code and decode (perceive mentally,
categorize), easier to comprehend, and so on.
Note that memorizing, coding, decoding, perceiving, comprehending etc., — all are
involuntary cognitive processes. People do not weigh consciously how optimal any of them is.
Such cognitive processes operate on unconscious level, they interact, “add up”, and inform the
interlocutor of the final verdict only by bringing to his attention most strongly the solution that
is the fittest in a given situation1.
And, just as once you have seen something you, cannot “unsee” it, and it will impact your
future acts of visual perception; or just as by listening to music and musical training, you can
improve your musical hearing, the same way, every linguistic utterance pronounced in a
certain social setting affects, usually only slightly, the brain architecture of the interlocutors. Or
at least, it modifies the distribution of chemical and electric excitations, and thus the strength
of synapses of the most active neurons. Consequently, every utterance, heard or spoken,
impacts our future linguistic choices and associations, i.e., our idiolect and the map of the ease
of the accessibility of our knowledge of the world, including that of the language spoken.
Summing up, the optimization of language is importantly determined by the
characteristics of involuntary cognitive processes responding in a simple way to stimuli from
local situations and by the history of our verbal interactions, including the contingencies
involved in every event. These factors together lead to the changes in dominant electric traffic
routes and the brain architecture of the interlocutors, manifesting themselves in the increased
likelihood of the fittest entities’ coming to mind during a given type verbal interactions. That
increase in the likelihood of interlocutors’ making fit choices in the future, results in selfregulation of their idiolects. When the idiolects of individual members of a given language
community self-organize and self-regulate, so does the community language, understood as a
statistical average of the idiolects functioning in that linguistic community.
*
1
To know more how this happens, we need to know more about the mechanisms of relevant cognitive processes opportunistically
underlying verbal communication system, because their functioning determines the criteria of fitness of linguistic entities. These protolinguistic processes, you might say, include anything from simple data processing, e.g., (sensory stimuli, affective signals, motor commands),
categorization, (e.g., object recognition, emotion classification, action selection), perception, attention, joint attention, forming associations,
interlocutors’ intentions reading (mentalizing system), encoding in sensory/working memory, memorizing in and retrieving from long time
memory, all the way to processing mental states (e.g., mental imagery, emotion experience, conscious goals, planning, reasoning), and social
cognition. To shed some light on how such cognitive processes conspire to make people go for fit solutions when speaking and thus optimize
language, an appendix adumbrating the major characteristics of the brain and its functioning has been included at the end of this paper.
3
4
2
The mechanisms of “languaging”
A theory of language constructed within an empirical framework, consistent with the
hypothesis that languages self-organize and self-regulate as a result of continual adaptation in
analogy to biological development and evolution2 must account for two major mechanisms.

A replicating (predictive, categorizing) mechanism that accounts for regular and
mutated occurrences of linguistic entities during verbal interactions – as a result of the
process correlating symbol(s) with information in a given social situation (pragmeme).
In other words, it is the mechanism accounting for categorization of items, on one hand,
and for the interpretation of utterances (made of anything from single words uttered in a
specific situation to larger linguistic constructions), on the other. In a nutshell, the
replicating mechanism is a “coding”/”decoding”3 (interpreting) mechanism.

An adaptive mechanism that accounts for the selection of the fittest language
entities over time – i.e., of the fittest correlations of content with form, and/or of
patterns of form or of content. By “selecting” the fittest, we mean here adjusting
current brain architecture and preferred chemo-electric activities in the brain the way
that the chances of future replications of these entities grow.
As discussed more extensively in the introduction, the fitness of a linguistic entity depends
on its characteristics contributing to minimizing effort involved in verbal communication in
view of socio-cognitive capabilities of humans and their communicative needs. Examples of
such characteristics include: how often a given entity is useful during verbal interactions (the
frequency of the occurrence of a given entity), how similar it is to commonly used items,
whether it is sufficiently different from other items so as not to cause confusion, the degree of
the emotional load the given item carries and its relevance for the interlocutor. All such factors
are conducive to remembering the given entity better and to shortening its recall time. The
similarity of a given entity to other items in language, additionally, limits the required memory
space, which also optimizes language.
2.1 The joined functioning of Replicating and Adaptive mechanisms
The two mechanisms function in a loop. The replicating mechanism gets input in part
from the non-verbal information concerning the given communicative situation, from the
2
and consistent with the state of the art knowledge of the brain structure and functioning, (adumbrated in the
appendix)
3
I use the terms “decoding” and “coding” to mean selecting a best matching item in the expectation field, and
choosing a word or larger unit whose semantic (encoded) meaning is good enough to single out the desired
item among the items in the expectation field.
5
current electro-chemical state of the brain, (i.e., what the interlocutor is thinking about right
now) and in part from more stable elements of the brain architecture (that contains, among
others, semantic and syntactic information). Given that data along with the degree of their
accessibility, the replicating mechanism encodes the given content into an utterance, or
decodes (interprets) the given utterance.
The adaptive (selecting and remembering the fittest) mechanism operates on the current
input provided by replicating (predictive, categorizing) mechanism generating symbol(s)meaning4 pairings and the resultant symbol-symbol and content-content pairings. The electrochemical activity, reflecting the current pairing, adjusts the brain architecture and the dominant
brain traffic routs, in the way that makes the future replication of the fittest pairings more
likely. In a nutshell, the adaptive mechanism optimizes future coding/decoding choices –
influencing the functioning of the replicating mechanism.
The Adaptive mechanism
The operation of the adaptive mechanism stands in a close analogy to Darwinian selection
of the fittest, except that selection takes place not by gene transmission during reproduction but
by the adjustment in the plastic area of the brain during speaking and listening. The respective
changes in the plastic area of the brain corresponding to the degree of fittest of the linguistic
entity used, increase the likelihood of its future replication. The factors influencing the fitness
have already been adumbrated.
The replicating mechanism – a predictive coding/decoding, mechanism
The functioning of the mechanism replicating linguistic entities, either identical with
ones that are already present in one’s language, or mutated (novel) ones, needs a more
extensive explanation. As proposed by Zielińska in Journal of Pragmatics in 2007 replicating
(coding/decoding) mechanism categorizes and interprets linguistic input by creating models
predicting what is likely to be said next (the expectation field) and using encoded linguistic
input of the words heard to select between the options generated by the model, the item
matching best the encoded meaning of the word(s) used for selection. It functions somewhat
similar to the recent predictive coding models of perception – by selecting from the field of
expectations of the interlocutor the option that shares the most similarity with the meaning of
the word used for selection. 5 More exactly, this predictive-selecting mechanism, operates in
the following way.
2.2 Establishing the expectation field
At any point of a verbal exchange, while decoding a symbol/word, or another
morpho-syntactic unit (un-, -able, -ful, etc.), the brain brings to mind the expectations
as to what may be talked about, said next, along with the information of relative
probability of potential options. We say that the brain establishes an expectation field.
4
Or more exactly symbol(s)- sense - reference
Given the speed of categorization, replicating mechanism must be a rapid mechanism, relying on redirecting
electrical and chemical activity of the brain. See more in the appendix.
5
6
Predictions likely to come to one’s mind, i.e., options constituting the
expectation field, may be exclusively non-verbal — take the form of episodic memory,
have visual and/or aural format - or be semantic in nature. So this can be solely what
we “see and hear” with our mind’s eyes. Yet, options in the expectation field may also
have words likely to express related content in particular situations in addition to that
content. An example of verbal constructions particularly likely to come to mind
accompanying potential meanings are the so-called Situation Bound Utterances, that is,
utterances such as you are all set, or can I help you, which are elements strongly
defining a relevant communicative situation (pragmeme).
2.3 “Decoding” - selecting the meaning (sense and/or reference) of word(s)
Now, on hearing a successive word (or some other language entity), the brain of
the listener, in addition to generating the expectation field, simultaneously brings to
mind (activates) the encoded representation(s) of that word –the averaged or most
recent representation(s) referred to with the symbol being interpreted, (i.e., its current
semantic (declarative) content located primarily in LTM-Wernicke’ region of the
Brain).
Next, the listener uses that semantic (i.e. encoded) content of the word(s) being
interpreted to select in the field of one’s expectations the item (reference or sense of a
thing, process, event, etc.) that
a) matches exactly the semantic content of the word(s) being interpreted or
b) resembles the semantic content of the word(s) being interpreted better than
any other item in the expectation field does.
(c) On rare occasions in adult speech, but commonly during language
acquisition in children, a word or pattern novel to the listener, (i.e., lacking a correlated
encoding), can be assigned the most conspicuous item in the expectation field, for
which the listener had no label as yet. Psychologists call that last mechanism ‘fast
mapping’.
2.4 “Coding”
As far as (en)coding is concerned, the speaker looks for words and patterns that
match well the elements of that message, i.e., the match is established according to the
Aristotelian definition of the meaning of each word, or larger language unit. He may
also, however, use words or patterns, whose encoded meanings suffice to single out the
intended elements of the message, in the envisaged expectation field of the listener.
2.5 “Sending feedback” to the adaptive mechanism
If the words just interpreted make sense in the given communicative situation,
either literally, or selectively, two things happen. First, the resultant correlation of a
form and meaning, as reflected by the relevant electric excitations, adjusts current expectation
field, in which the successive word(s) will be decoded. Second, by affecting physical
connections between neurons or the strength of respective synapses, the new correlation
7
affects the predictive model of the encoded meanings of the words, and possibly larger
constructs just interpreted, and indirectly, future expectation fields. In other words we
might say that the error information between encoded and interpreted meaning is sent
up to the Long Term memory LTM (i.e., predicted vs selected word content)
Summing up
The Replicating Mechanism
accounts for speech productions and interpretation - the occurrence of correlations of words and
meanings (word forms with sense and/or reference) and of word patterns, as a result of
categorization. It allows for the recurrence of already existing correlations, which predominate, but
also the occurrence of novel ones – mutations. i.e., language change. Mutations occur due to the
following mechanism:

During verbal interaction the brain generates expectations concerning the potential purpose, content
of the intercourse and, possibly, verbal ways of achieving it – an expectation field

The encoded content of the words and larger chunks of language being interpreted is used to select in
the expectation field the content meant by finding the item(s) best matching the encoded content of
the words used for selection.
Since the expectation field can be very different in different situations, so can be the outcome of the selection
of the best matching option. Such one-time outcome of a given selection is not tantamount, however, to the
change in the encoded meaning of the linguistic entity used during that selection. The encoded meaning,
typically, is the result of the uses of a given linguistic entity over time, and it depends on the impact the
Adaptive Mechanism has on each of the resultant correlations.
Adaptive (“remembering the fittest”) Mechanism
accounts for storing in long term memory the fittest of the language entities that occurred as a
result of Replicating Mechanism
Adaptive mechanism “remembers” the fittest outcomes from Replicating
Mechanism – by storing them in Long Term Memory due to physical changes in the
8
brain. A fit item is remembered well and has a short recall time, thus comes to mind
easily on similar occasions, constituting the relevant expectation field.
The fitness (the efficiency) of the item considered, depends on the cognitive
benefits vs. costs its use and storage incurs in relation to these incurred by alternative
items. The degree of fitness of a linguistic item reflects the chance of its future
replication through replicating mechanism (RM).
Both the “replicating”, (coding/decoding, categorizing/interpreting), and the adaptive mechanism opportunistically
employ for “languaging” proto-linguistic cognitive functions. See the appendix: The brain for the linguist.
9
3
The expectation field
The process of communication modelled by the selection of the item best matching the
encoded meaning of the linguistic entity used for selection, among items in the expectation
field (via Replicating Mechanism), can be illustrated by the following analogy. The cartoon
presented below, coming from The New Yorker, shows the failure of communication process
between a policeman knowing this neighborhood well, and a tourist not familiar with the area
he is visiting. The maps in their respective thought bubbles may serve as an analogy for their
respective expectation fields -the predictions of what is going to be talked about as formed in
their respective brains.
The arrows correspond to the interpretations (selection made in actual expectation
fields) of successive words, which serve to generate successive, adjusted expectation fields
needed to select the interpretation of the next word uttered. The more the respective successive
expectation fields of the interlocutors differ, the more their respective interpretations of
successive words do, (in analogy to the differences in the positions and direction of respective
arrows).
3.1 Perception and the non-verbal part of the expectation field
10
The expectation field we use when communicating is based in part on the history of
our perception of the environments met during daily interactions as well as that of the
worlds built verbally we have come across. In a nutshell, the expectation field includes the
memory of the history of the surroundings perceived by the interlocutor physically or
mentally.
By conducting a simple experiment, we can observe that our expectations due to our
experience affect even our current perception of the surroundings, which constitutes a part
of the expectation field we use when communicating verbally. To carry out the
experiment, look at the stairs in the picture below and note which side A or B is closer to
you. Now, turn the page upside down, watching the picture all the time. When the picture
is upside down, note again which of the sides A and B is closer to you, now. The upside
down version of picture 4 is shown in picture 5.
Fig. 4
Fig 5
11
In picture 4 figure A, which is perceived as a side of the stairway, is closer to the viewer, than figure
B perceived as a wall beyond the flight of stairs. In figure 5 depicting picture 4 turned upside down,
it is figure B perceived as a side of the stairway that seems to be closer to the viewer.
The reason why we see what we see, is because our brain selects for our conscious view from the
options offered by the memorized history of our experience with the environment. Since the brain
tries to select from what it is familiar with, and in our daily lives we come across stairs at the
bottom of a staircase and not at its ceiling, whether we look at the picture the right side up or turned
upside down, the brain classifies what it sees by looking for the likely solution the picture matches
best
Similarly, when decoding an utterance, we let its encoded content select from what we expect is
likely to come. Yet, in case of language, our expectation field is richer than that formed in the case
of visual perception alone, because it results also from the history of our mental experience,
including that acquired verbally.
3.2 The information structure of the (verbal) expectation field
When we say that some utterance, usually a sentence uttered, communicates a message,
we tacitly assume that it encodes the comment and the topic i.e., tells us {what it says} and
[about what]. This division of the meaning of a sentence uttered in a given situation into its
topic and its comment is called the information structure of that utterance (i.e., of the sentence
uttered in a given situation).
Note, however, that what we take for the meaning of a sentence, need not encode which
part of the information is the topic and which is the comment, and consequently need not state
what the message is. A given grammatically unambiguous sentence, with different information
structures assigned (i.e. different topic comment divisions), may communicate different
respective messages- it may have different meanings. Therefore, the expectation field must
treat topics and comments as separate pieces of information and interlocutors must model
predictions concerning each of them separately.
Let me illustrate why the same sentence with unambiguous encoded meaning can
reflect different messages based on differently assigned topics and comments. Let me start
with assessing the message to the sentence: Dog owners must carry dogs on this elevator6.
Think of the meaning of that sentence as an answer of the following questions. 1. Which
elevator must dog owners who carry their dogs use? 2. Can dog owners use this elevator with
dogs standing next to them? 3. Can a dog owner use this elevator when not accompanied by
their dog? 4. What is required of dog owners who want to use this elevator? 5. Can dog owners
use stairs when carrying their dogs? When answering each of these questions, the message
conveyed by the same sentences mentioned is different. In each case both the topic, contained
in the question, i.e., what question the sentence answers [what it is about] and the comment
{what the sentence says about the topic} are different.
The sentence mentioned spoken with an appropriate echoic intonation (when repeating
someone else’s words ironically because of getting into trouble after trusting these words) can
also be used to state that “you tricked me into believing that dog owners must carry dogs on
this elevator” This time the whole sentence is a comment on the topic [what you tricked me to
believe into]- whichever way you chose to interpret the comment. Thus, the message may have
the topic designated non-verbally and the content of the sentence may all serve as the
6
This example comes from Eva Hajcova, I believe.
12
comment. Apparently, with new topics, sentences designate new comments and thus, new
messages.
The same sentence when interpreted in relation to different topics may also have
opposite truth values. Compare, for instance, the interpretation of the sentence English is
spoken in Burma when used to answer the question 1. “Where is English spoken?” and the
question 2. “What language is spoken in Burma”, respectively. As Eva Hajcova noted the
statement “English is spoken in Burma” is false as the answer to question 1, but true as the
answer to question 2.
As the last example of the role of the information structure of the expectation field in
establishing the information conveyed with a given sentence, consider different preferential
readings of the following two sentences. Wielu [many] ludzi [people] trenuje [practice] niewiele [few]
dyscyplin sportowych [sport disciplines]. „Many people practice few sports.” and Niewiele
dyscyplin sportowych trenuje wielu ludzi. “Few sports are practiced by many people”. I shall
discuss the Polish translational equivalents, because in Polish these two sentences differ in
word order only - there is no passive voice required in the second sentence in the pair.
In the first sentence, beginning with the phrase wielu ludzi (many people), the
expectation field takes wielu ludzi (many people) to be the topic of the sentence. Most
sentences start with topics – informing what the sentence is about. Since the words wielu ludzi
(many people), pronounced without a specific expectation field, refer to a big group of people,
and people in a big group of people, most likely, do a different thing each, stating that they
practice few sports, likely means that each of them practice different sports - a small, and in
many cases, a different subset of sport disciplines out there. And “practicing few sports” is the
comment of the sentence discussed - a piece of information about individual representatives of
“many people”. When, in turn, we start the sentence with the phrase few sports – lack of
specific expectation field brings to mind as the sentence topic, “a small group of sports”. So
next when we talk about the number of people practicing them, which is the comment in this
sentence, we talk about the number of people who practice the same, small group of sports.
Although an equivalent pair of English sentences involves passive voice in one of the
sentences whose Polish equivalents were discussed, it is not the use of the passive voice that
causes the different readings of such pairs of sentences. That difference is caused by the
change in the word order accompanying the transition from the active to passive voice in
English, and thus by the resultant change in respective expectation fields during the
interpretation of the same words.
3.3 Information structure reflected in the pre-grammatical skills of toddlers
acquiring language
The examples just discussed illustrate the fact that the expectation field of a message
needs to be informationally structured. It must include separately the information about the
likely topic and comment. This observation is consistent with the results of psychological
experiments concerning communicative processes in infants and toddlers. It has been
established that children as young as 9 months start indicating objects with gestures, which
shows that they are capable of joined attention -focusing on a topic. At 12 month they start to
indicate the object of joined attention verbally, i.e., point out the topic of interest with first
words. At 18 months, they start using wholophrases as comments on topic differentiated
situationally, usually by gesture. It is at 24 months that they start communicating with
13
sentences - linguistically indicating both a topic and a comment. At that time, these two-word
utterances are not syntactically related yet.
So little children are already aware of shared knowledge and are capable of
differentiating between what is new for the speaker, i.e., old knowledge [the topic commented
on], and what is said about it, i.e., what is new knowledge {the comment} before they develop
a grammatical language (i.e., language with utterances bound by syntactic markings and
relations.) These mind reading skills, present already in pre-syntactic communication in
humans, must certainly be reflected also in the fields of expectations. With the arrival of
grammar (syntactic markings of relations between words), the information structure became
reflected in part in the grammars of languages, yet as illustrated above, only in part.
3.4 Jokes - exploiting dynamic expectation fields
The replication mechanism advocated, based on predictive categorization, allows us to
account for the functioning of the majority of jokes. Most jokes rely on the fact that the same
linguistic entity may have different meanings in different expectation fields and that the
expectation field changes as we progress with the interpretation. Consider for instance the
following joke. Q: Why can't you trust an atom? A: They make up everything. The first sentence,
as the topic of further discussion, triggers the expectation field, considering as the topic of the
following sentence, the reason why atoms are not trustworthy. Along with the words: make
up everything– the expectation field transitions to having as the topic the behavior of atoms –
[what do atoms do so that you cannot trust them? = how do atoms present everything?]. At this
point the interpretation of the verb make up as an active verb is lie, and the final comment is
{[Atoms concerning everything they say]make up} = Atoms lie concerning everything they say.
Since, however, we do not attribute such behavior to atoms, the brain having
considered this interpretation, keeps searching for another one. The topic of an alternative
interpretation that comes to one’s mind is [what do atoms make up?], where the sense of the
verb made up is to ‘constitute’. Therefore, [the topic] is [atoms make up . . .] and{the
comment} is that it is {everything}. In other words, {[atoms make up=constitute]
everything}
3.5 The reason why the expectation field helps us to communicate
The crux of the theory advocated follows from two observations. The first observation
is that the expectation field, reflecting what we think about at a given moment in a given
situation, concerns a minuscule bit of our knowledge and, consequently, words used selectively
are not required to differential between all potential entities one is ever capable of talking
about. Instead, we need to differentiate merely between few, potentially likely options.
The second observation is that the world we talk about is highly structured. Therefore,
we do not need to describe its every element, as if pixel by pixel. Providing information about
the position of few of such pixels suffices to single out the item at stake. If we lived in a totally
unpredictable world, where a person could have any number of heads, or grow any number of
legs, or continuously change into anything else (be ¼ of a person and 2/3 of a tree), we would
have to describe that world pixel by pixel, in a continuous space, you might say. In that
alternate world, our language could encode meaning only by enumerating all characteristics of
every item talked about.
14
Yet, our world is highly structured. And the sets of the features co-occurring in an item
are sparse and come in limited combinations. In a given situation of verbal interaction, the
number of possibilities of what may happen next in the discourse is drastically limited, too.
Because of that, we can, for instance, identify a given item in a discourse by providing enough
information to differentiate it from others without necessarily describing it in full. Even less
information is needed to single out a given item, when the number of options to choose from is
small, which is precisely what the expectation field ensures.
15
4
“Decoding” by selecting entities in the expectation field with the help
of pictorials, string of words in no-syntactic languages and pockets of
non-syntactic constructions in syntactic languages.
“Decoding” - selecting among a set of items the item best matching a given pattern; or
reversely “encoding” -assigning an item to that one of a set of patterns that matches it best
“encoding” - in other words predictive categorizing7 - is one of the most fundamental human
cognitive skills. It underlies our ability to categorize. Once, it let us differentiate between pleasant events
or entities and unpleasant ones thus, between potentially dangerous items and benign ones,
laying grounds for our perception. At first fundamental for survival, next it opened the door
for verbal communication. Without seeing the world in terms of instances of a number of
categories, if we treated every dog as a separate item, and every time it barked as a separate
activity, first we could not learn from experience, second, we could not say anything new communicate what the other person did not already know.
When talking about “decoding” and “encoding”, it cannot be emphasized enough that
these are not the same processes as the corresponding decoding and encoding. Traditionally,
decoding and encoding means relying on the Aristotelian definition either to state what
properties the entity decoded has (these are the properties enumerated by the category
definition), or to attest that a given item has properties included in the Aristotelian definition
of objects in a given category, thus encoding that entity or concept. Decoding and encoding
are symmetrical, both depend on comparing the properties and relations present in the
definition of the category with those of the entity being categorized, which does not depend
on context. “Decoding” and “encoding” on the other hand, strongly depend on the situation:
the categories available at the given moment and the verbal entities available to the
interlocutor at the moment.
To illustrate the process of verbal “decoding” and “encoding”, let me use the title of a
popular Covid-19 vlog run by dr John Campbell. One of these vlogs discussing the stoppage
of Oxford-Zeneca vaccine clinical trials is entitled News, vaccine hiccup. Without any
context, the phrase is absolutely meaningless. Yet, knowing the current situation, the interest
and knowledge of his viewers, Dr Campbell used the phrase that easily differentiated between
the topics related to these discussed recently on his vlog, which included the up-dates on the
Covid-19 vaccine development. The listeners of this vlog, in turn, easily identified (“decoded“
in the expectation field), which aspect of the information provided by that vlog the phrase
7
In fact, the absolute Aristotelian coding and decoding can function only in mathematics. Since in nature, we do not have God’s
eye view of reality, using the Aristotelian definition of a category, requires first identifying the properties and relations defining that given
group of category members by physically differentiating these properties from other viable properties, which is done, as psychologists teach
us, by predictive coding, which operates very much the way “coding” does.
16
selected – putting clinical trials of the Oxford Covid-19 vaccine on hold because of serious
complications in one of the subjects participating in it.
Whether the label hiccup used in the phrase vaccine hiccup will receive a new encoded
meaning (that of a stoppage of some process), or whether this utterance will turn out to be only
one time usage of the word hiccup departing from its regular use, will depend on the
functioning of the adaptive “remembering” mechanism.
In today’s society, the ability to assign an entity to one in the given set of categories
(“to encode”, categorize), and reversely, to “decode”, by selecting from the options in the
expectation field one matching a given category better than the other entities in that
expectation field do, is fundamental in every walk of life. Doctors considering the choice of a
treatment, for instance, ponder which of the known diseases the symptoms presented fit better
- is it a flu or Covid-19?; AI scientists commonly categorize medical images by assigning
them to the best fitting of possible outcomes. In everyday life, we constantly rely on
“decoding” by selecting from options that one, which matches a given sign, often one seen for
the first time, better than any other option in the expectation field does. We do that when
using anything from kitchen appliances to smartphones and all other technologically advanced
devices, which rely on pictograms or single words as short-cuts for the instructions of their
use.
4.1 Decoding by selecting the option the most similar to a given pictogram in the
expectation field
People are really good at selecting among potential options the option most similar to “pictogram X”
containing minimalistic information, even when faced with novel pictograms. To illustrate the point,
look at the picture below
Figure 3
When printed out of any context, most likely you do not know how to interpret this picture.
Yet, if you find this picture drawn on a ceramic stove top, then, knowing what ceramic stoves
do, you will immediately guess that the two top big circles, the size of an average bottom of a
pot, mark the right and left hotplates, respectively. The little ellipsis at the top of the picture,
one with right half colored black, and the other ellipsis with its left half colored black, each
with plus and minus next to it, mark the touch buttons that operate the left and right plate,
respectively. The expectations as to the potential functions of the elements of a given picture,
let us use the little pictorial information there is to communicate a complex message. Similarly,
we can communicate with the help of highly limited amount of information actually contained
17
in the given pictogram, be that one used by smartphone apps, tablets, or other technological
devices, even when we see them for the first time, if only we are aware of the function of the
given device or software application – have a right expectation field available to us.
Psychological experiments on interpreting pointing gestures in children have shown
that already toddlers are capable of using limited information to differentiate between options.
These little subjects have been found to use the same pointing gesture to refer to different
actions in different situations8. Later on they learn to use words encodingly to comment on
topics selected by gesture. No other primate is known to be capable of communicating with
novel, previously unseen and thus untaught signs. This pre-linguistic ability that lets us, in
particular, communicate fluently through apps using icons that we have never seen before, but
which contain the minimal amount of information sufficient to single out the action meant
among few possible actions9 turns out to be central to using and developing language.
4.2
Decoding by selection from options in pre-syntactic languages
Ever more scientists believe that syntactic languages have evolved from pre-syntactic
ones, which operated very much the way we can use pictorial information to communicate in
specific settings, as illustrated above. In line with the hypothesis of a pre-syntactic languages
giving origin to modern languages Ray Jackendoff1 & Eva Wittenberg2 write: “Such a
language would still be capable of conveying certain semantic relations through word order—
for instance by stipulating that agents should precede patients10. However, many other
semantic relations would have to be based on pragmatics and discourse context. We find
evidence of linear grammar [symbolic language without syntax- DZ] in a wide range of
linguistic phenomena: pidgins, stages of late second language acquisition, home signs, village
sign languages, language comprehension (even in fully syntactic languages), aphasia, and
specific language impairment. We also find a full-blown language, Riau Indonesian, whose
grammar is arguably close to a pure linear grammar. In addition, when subjects are asked to
convey information through nonlinguistic gestures, their gestures make use of semantically
based principles of linear ordering. Finally, some pockets of English grammar, notably
compounds, can be characterized in terms of linear grammar. We conclude that linear
grammar is a plausible evolutionary precursor of modern fully syntactic grammar, one that is
still active in the human mind.”
What Jackendoff calls a linear grammar language is a language, that forms messages
whose structure (e.g., what is the patient what is the agent, what is the comment, what is the
topic; which element is modified and which one modifies, etc.) is not marked by syntax (word
order or flection). To structure a messages using a pre-syntactic (linear grammar) language,
one needs to rely heavily on selecting from expected options. But even in syntactic languages,
as discussed in detail in part 5, to communicate verbally, to interpret words, phrases and
sentences, speakers still rely on selecting from likely options.
8
Grown-ups do it too.. For instance, recently, I looked at my snoring dog in the way catching the attention of my husband, and pronounced
with a smile, he he, to which my husband responded he he. The bottom line of the conversation could be interpreted selectively in that situation
in the following way: “Look, our dachshund is snoring in such a funny way”. The response: “I agree, the way our dog is snoring is funny.)
Obviously selected meaning is not as precise as encoded one, yet sufficiently precise to convey basic functional facts: by saying he he and
looking at my dog in a way that caught my husband’s attention, I communicated that my dog’s snoring is funny.
9
Thank-you blinking
John hit a cat agent patient
10
18
4.3
“Decoding” English compounds –a pocket of a non-syntactic language
Compound nouns are nouns, consisting of two or more words not related
syntactically, such as coronavacation, rainfall or music box, or bus stop, which mean
significantly more than the sum of the meanings of their components. Compounds function somewhat
the way “two points select the whole, specific line among lines”. Two nouns, as if, point out
“the line” and not merely “its two points”. These points, as if, select from viable options - the
expectation field. For instance, the words bus and stop select in the expectation field
generated by our knowledge of the world a bus stop as a place where buses can stop for
passengers to get on and off the bus.
In English, there are some common semantic patterns relating the meanings of
components to the meaning of the compound as it is in the case of whisky-soda, which is
whisky and water mixed together, or a poet doctor is a doctor who is also a poet. Bernd Heine
and Tania Kuteva give interesting examples of other systematic ways of combining nouns to
mean more than the sum of the meaning from other languages. “In “Mordvin t'et'a.t-ava.t
(father.PL-mother.PL) 'parents', Tibetan srab-mthug (thin-thick) 'density'; Khalkha xaluun
xüjten (heat cold) 'temperature' In the Trans-New Guinea language Sentani, where pigs and
dogs are prominent domestic animals, the word for 'animal' is obo-joku (pig-dog), which is
not restricted to these two kinds of animals. Similarly, the compound KNIFE-FORK in
American Sign Language (ASL) means 'silverware', that is, it is not restricted to knives and
forks. … Additive compounds in American Sign Language (Wälchli, 2005: 19-20) APPLEORANGE-BANANA-ETC. 'fruit' BEANS-CARROTS-CORN-ETC. 'vegetable' RINGBRACELET-NECKLACE-EARRINGS-ETC. 'jewelry'.”
But there are many more compounds with unique relations between the meanings of the
components of the compound and that of the compound itself. An apple tree, for instance, is a
tree that grows apples. But a queen bee is a bee with the social position in the beehive
comparable to that of a queen in a monarchy. A mountain foot, in turn, is the part of the
mountain that is at its bottom, the way the foot is the bottom part of a person. These are all
very different relations, yet the meanings of respective compounds are easy to identify,
knowing what the world looks like. Consider for instance Coronavacation, (coronavirus +
vacation), I came across that word for the first time, when the Polish Minister of Education
addressed school age children during the lockdown caused by Coronavirus stating: “The
lockdown is not coronavacation”. Naturally, everyone could understand this new compound
derived from the meanings of words Corona and virus in a very unique way.
There are also numerous other non-syntactic combinations of different other word
classes, not only nouns, whose meaning is sub-sective and unpredictable from the meanings
of the respective components. A drive through coronavirus gene test, for example, clearly
means more than the sum of the meanings of its components. It is a test checking for
coronavirus RNT, which is administered at special locations accessible by car without having
to step out of it. A must have, in turn, is a combination of two verbs, which is a noun, This
combination of verbs must be interpreted subsectively, by selecting from options from the
field of expectations and it means “a thing you must have”. An egg head, in turn, is neither a
head, nor an egg, but a highly academic person, which again requires selectin from expected
options, and results in a complex interpretation.
Finally Kuteva illustrated the issue discussed with fruit names in the home sign
community of Nicaragua (Morford, 2002: 333) PEEL EAT 'banana' RUB-ON-SHIRT EAT
'apple' SLICE-OFF-TOP-WITH-MACHETE EAT 'pineapple. Not surprisingly, sign
languages rely on selection from communicative fields, too.
19
All these examples show that words put together non-syntactically can also convey a
complex content, normally believed to be reserved for syntactic languages. The interpretation
of such a word string can be best accounted for by stating that it has been selected from
alternatives likely in a given situation, whose meaning includes the meanings of the words
used for selection.
4.4
“Decoding” the meaning of word blends
An interesting case of selecting meaning in the expectation field are word blends. Interpreting
a word blend requires, first, recognizing in it parts of the words they constitute so as to use the
meanings of these original words to select some entity from the relevant expectation field. .
For instance, let us look at the meanings of the words brunch (breakfast+ lunch),). Brunch is
made up of parts of the words breakfast and lunch to indicate an in-between meal – a large
late breakfast you might say. It can be understood for the first time in a narrow expectation
field consisting of meals. Otherwise, it could be short for many other word pairs and have
numerous meanings. For instance, brunch could stand for bread+ lunch and mean “bread
typically eaten for lunch” and many other things and events.
My favorite Polish blend is przywitul, which combines two words przywitaj (to
welcome) and przytul (“to hug”), and means “to welcome someone with a hug”. (Honestly, I
came across that word only once - I saw it printed on a sticker in an elevator on a university
campus.) What’s so special about this blend is that both words start with the same two first
syllables and even one phoneme from the third one: przy+wi+t- so this blend sounds
particularly appropriate because of the high degree of similarity to the original words. In the
case of this blend, the component words przywitaj and przytul truly stick out from this blend.
They also generate a clear expectation field of actions related to meeting someone, allowing
easy selection in it.
5
20
“Decoding” with single words
5.1 “Decoding” the meaning of single words by selecting in the expectation field
Selective use of language is possible with units of every level of language
organization, including single words of every word class. The word classes best known
for such uses are pronouns and adjectives. Consider, for instance, the pronoun He. Since it
may refer to half a population of humans and bigger mammals, it might seem pretty useless.
No one ever needs to say something about that group of entities. Yet, it is an extremely
frequently used word. It is useful, because it is typically used selectively in narrow
expectation fields. Numerous communicative situations generate the expectation field
consisting of one option having the property “male”+ ”singular”. For instance, when saying
the words he’s coming after hearing footsteps in the front yard at 11 p.m. when only one
family member, a teenager Tony, isn’t back home yet, we will expect that the person
approaching the door at that time will be Tony. Since Tony is male and “single”, the pronoun
he will be good enough to identify that person. When using pronouns selectively, one
capitalizes on the fact that the field of options to select from is very limited. In such a field,
even a very restricted semantic content of a pronoun (e.g., singular +male) suffices to single
out the object meant to be referred to.
Another category of words that have also been widely acknowledged to “select”
content are adjectives. Consider, for instance, the meaning of the adjective “big”. Note, that
despite the fact that the encoded meaning of big implies the size much bigger than that labeled
tiny, a big Wuhan virus refers to an object muchly much smaller than the size of even the
tiniest of tiny galaxies. This is so because the selected (interpreted) meaning of a big x is
typically used in the expectation field of items of the same category and thus effectively
means “big in comparison to the available sizes of the category generated by the expectation
field”. Similarly, the phrase a hot day must be interpreted in the expectation field of similar
days. Therefore, it usually means that the temperature during the given day is hot in relation
to these typical for a given season and the geographical location of the speaker, and possibly
even hot for average temperatures this particular year.
The need for selective interpretation of adjectives is even more striking when
evaluating adjectives such as good, (or nice). Consider the characteristics of a good
researcher, a good doctor, a good child, a good dog, a good picture, good health, etc. There is
not much, if anything, that the meanings of good and nice share when used in the phrases
mentioned. The adjectives nice and good assess quality of the items they modify by marking a
point on the scale of the qualities of these items as assumed by the respective expectation field
generated during a given social encounter. So to understand the meaning of good in the
phrase a good doctor, for instance, we rely on our expectation field informing us, first,
what words are used to assess doctors in similar social situations –e.g., in popular
press {outstanding, superb, very good, good o-key, average, bad, incompetent} or in
informal conversations {good, not so good, bad} and second, what is the scale of the
quality talked about. Next we find on the scale of quality considered, the place
21
corresponding to the position of the adjective good in relation to adjectives used in
similar situations.
Just as it is with adjectives, interpreting adverbs also typically requires selection in the
expectation field. When a child, a horse, or a racing dog run fast, they are going at
very different speeds, depending on the scale of speeds possible for each of them.
Actually, interpreting verbs requires selection, too. Instead of postulating that when
saying that a toddler is running, we use a different sense of the word run than when
saying a horse is running, or an ant is running, it makes more sense to say that the
verb run requires selection between the ways of movements of a given entity available
in the relevant expectation field. This is done by finding on the expectation field scale
the value that corresponds to the position of the meaning of the word used in
comparison to other words that could be used to assess the quality being evaluated.
To illustrate the fact that interpreting prepositions also requires selection from an
expectation field options, let us compare the meanings of in used in the following phrases: a
cat in a hat, a cat in a bag, a cat in London, a cat in a window, a cat in a crown, a cat in a
tree. Interpreting the preposition in in each of these phrases requires selection between
possible geometrical arrangements likely to come to the interlocutors minds between the cat
and a hat, a bag, London, a window, a crown and a tree, respectively, (i.e., establishing
relevant expectation fields), and for each pair (expectation field) correlation viable
geometrical relations with those that have verbal labels for. Just the way we did it for
adjectives, adverbs, and verbs. When interpreting the preposition in, we decide, for instance,
which position of a cat in relation to the tree will count for the cat to be in a tree, over a tree,
under a tree, next to the tree, given possible arrangement of cats and trees. Because of the
selection in the expectation field, an object A on object B may turn out to actually be under B,
e.g., a fly on a ceiling.
We encode the fly directly under the ceiling and attached to it with the preposition on
– a fly on the ceiling – because it resembles more animals on a table, desk or a chair, than
things labeled as under ceiling such as a lamp. These animals keep their legs on the surface of
a table, chair, windowsill, a road, or anything else, and cover with their body only a small area
of that surface. These is especially appropriate if we compare alternative geometrical relations
and alternative words expressing possible geometric relations. Now, we may use, for instance,
the preposition under in the phrase a fly under the ceiling, to mark situations, when the fly is
not touching the ceiling and is the right side up as in The fly is flying under the ceiling.
Note, that it is not merely the matter of geometrical arrangements between items that
decides on the choice of the preposition, or any other word. The whole relevant expectation field
including also social perspective can matter, too. Not surprisingly, the same physical arrangement
of elements of the same pair of items may need to be expressed with a different preposition in
different social situations, i.e., in different expectation fields. Consider, for instance, the
relation between the position of a cat at Gatwick Airport and that of London. Can we say that
the cat is in London when it is at Gatwick? Yes, this sentence will be true in the expectation
field generated, for example, by a phone conversation about the international travel of the cat
from Rome to London. A happy cat owner may be talking to his friend over the phone and say
that his cat has just returned from a medical treatment in Rome and is already in London. Yet,
stating that the cat is in London, when it actually is at Gatwick, will be false in a different
situation, for instance, when a cat owner, who at the moment is at his residence located
22
downtown London, talks about picking his cat up with his roommate not aware of cats
international travels. The cat owner could say “I am going to pick up my cat” to which the
roommate could say “where is the cat?” In that situation, the cat’s owner reply “The cat is in
London”, would be false.
23
6
”decoding” with syntactic languages
6.1 Selecting with phrases and sentences
It turns out that even when interpreting syntactically combined words, most of seemingly
regular constructions require to have their meaning selected from options in an expectation
field. Let us look, for instance, at the phrases red eyes and blue eyes, which select not only the
relevant shades of the colors marked by the respective adjectives, but also some specific
region of the eye, which is different when modified by the adjective red and blue,
respectively. Moreover, just as it was illustrated with the use of the preposition in, the phrase
a red eye means different physical characteristics in different expectation fields. For instance,
in the field generated by a situation of a photographer speaking about photography, and in that
generated by a situation of a doctor talking about a sick patient. In the former case, the phrase
red eye means an eye of a photographed person in a picture with a red dot in it, which was
caused be a common shooting mistake with technologically less advanced cameras. In the
latter situation, the phrase red eyes means eyes with reddish whites caused by a sickness or
tiredness.
And these are not only nouns modified by adjectives, but virtually all syntactically
bound word classes that require selection in an expectation field to interpret them and whose
meaning depends on the relevant expectation field. For example, the request “zmień nogę”
uttered by a riding instructor in Polish (lit. “change leg”- pick up the correct lead) has a
different meaning depending on whether the rider is cantering or trotting at the moment. In
the former case, the rider must make the horse start moving differently, placing forward the
opposite leg first (change the lead). In the latter case, the rider himself is supposed to change
the way he is moving in respect to the movement of the horse’s leg – sit down when the horse
places down the opposite leg.
In fact, even sentences, or even texts, often need to be interpreted selectively in a field
of expectation, which we usually do without even realizing that. For instance, once I said to a
friend: Antek zorganizowal nam wycieczke do Augustowa, which I used to comment on the fact that I
had to cover 1200 km during one day to pick up my son from a holiday resort, after he got sick the day
after the arrival to an IT camp located 600km away from home, and after asking him not to go to that
camp because of having a minor cold when leaving for the camp. Note that the verb phrase organized a
trip not only selects as the sentence topic: our trip to Augustów to rescue a sick son, which had us
extremely annoyed, but simultaneously comments on the selected topic with its encoded content.
Calling an emergency rescue trip that we were upset about, an organized trip, sort of holidays, is meant
as an irony, of course. I call such double (selective and encoded) use of a single linguistic entity compact communication. As it was in this case, compact communication often requires some further
inferencing based on the selected and encoded contents – respective topics and comments.
As expected, the same sentence used selectively, whether metaphorically or not, may
have a different sense when interpreted in a different expectation field. For instance, consider
the sentence "His words cut deeper than a knife." used in two successive situations. First,
when talking about feelings, for instance, as preceded by the sentence “I hate him.” When
talking about emotions, after hearing mean words, we may experience unpleasant emotions,
just as we do when experiencing physical pain. This situation will generate the expectation
24
field in which the topic of that sentence concerns negative emotions. In this situation the
sentence His words cut deeper than a knife will serve to establish a point on a scale of feeling
unpleasant by comparing the degree of these unpleasant feelings caused by words to the
acuteness of physical pain when cutting one’s flesh with a knife on the scale formed by the
acuteness of all possible pains. In the case of the same sentence used in the context of an
academic dispute, however, the utterance his words cut deeper than a knife may also generate
the expectation field concerning the precision of expression. In that latter situation, the same
sentence his words cut deeper than a knife could refer to the degree of precision of thinking in
natural languages.
A good example of texts’ being used selectively, in different expectation fields, are
parables in the Bible. It is because their encoded content selects in the current reality of
listeners, which changing over centuries that its content can be relevant for successive
generations over two thousand years.
At the other end of complexity, meaningful word parts happen to be used selectively,
too. For instance, frankowicz in Polish (lit. franc+er frank+owicz) is a person who lived in
Poland in 90s, did not qualify for a credit he wanted in Pl zloties, was offered a credit in Swiss
francs, and when the value of the franc in relation to pl zloty went sharply up, that person
went bankrupt. Since there was a considerable group of such people, and their situation
became highly publicized and the ethics of such bank loan offers questioned, these people got
a name and were referred to as frankowicze (lit. francers). Obviously, no language looked at
as an abstract system can have an inbuilt mechanism that could provide such an interpretation.
6.2 Extreme selection in the expectation field
Some interpretations of syntactically bound words can be rather extreme. The sentence
dust furniture and the noun phrase świnka morska (Polish for a guinea pig, which in a literal
translation reads: “marine” pig), are, in turn, example of subsective usage, in which actual
meaning contradicts encoded meaning of respective entities. Świnka morska is a rodent, not a
pig, and is not made of sea the way a sea breeze is. This discrepancy can easily be accounted
for with the theory advocated. The guinea pig has been brought to Poland by sea, and it
resembles a pig. Since no other animal fits that description, the words sea and pig suffice to
select the best fitting animal among animals and their stories.
Similarly, the words dust and furniture in the sentence dust furniture select among
activities done to pieces of furniture. Since the expectation field generated by the words dust
and furniture does not include the possibility of putting dust on pieces of furniture
deliberately, thus, although the encoded meaning of dust in dust furniture contradicts its
selected meaning, the encoded meaning, nonetheless, is enough to select the actual meaning
in the relevant field of expectations.
6.3 What and whi(t)ch
To sum up this section, we may notice that words can be used both descriptively –
enumerating the properties of some referent or a category - and selectively, i.e. by pointing out
in the expectation field the referent or sense that matches best the encoded meaning of the
words used for “decoding”. When saying that a word can serve to describe some object by
attesting the match of its properties with the properties of the representation correlated with
that object, we mean it can be used to answer the question “what is it like?” When proposing
that a word can also serve to identify an item in the field of expectations, say which one in the
field of expectations it is, we mean it can serve to answer the question, which one out of the
items in the field of expectations is the most like the encoded representation of that given
25
word. In this latter case, the word used for selection says what the given item is like only
indirectly. Although there are prototypically “whichy” words like pronouns (he), and
prototypically “whaty” words like nouns, practically every word category can be used both
ways. Letting words act “wi(t)chy” limits the size of vocabulary needed without
compromising on achievable precision of communication. It makes language flexible and
accommodating, allows language acquisition and contributes to self-organization. The
capability to use language the witchy and whaty way is the result of the cooperation of the
replicating and adaptive mechanisms. Witchy use of language gives life to language.
26