Tải bản đầy đủ - 0trang
3 Choices, features and cross-classification
Figure 3.8 Sex as a choice between ‘male’ and ‘female’
and so on. We can now consider how to build these choices into a network, with
the help of a new primitive relation called ‘or’.
Think of sex (aka ‘gender’, a term that I prefer to keep for grammar), one
of the most important choices that we make when classifying people. Sex contrasts ‘male’ and ‘female’ and we assume that everyone must have either male
or female sex, and nobody can have both. The question is how to include this
information in network structure.
The first step is to recognize that the sex called ‘male’ is different
from the type of person we call ‘male’. A male person has the sex ‘male’, which
isn’t a person but a property of a person. Similarly, an old person has the property
‘old age’; but old age isn’t itself a person.
What then is the sex ‘male’ or the age ‘old’? It’s a concept, but a very abstract one compared with, say, ‘person’. It probably doesn’t have any properties
of its own, and its main job in our minds is to help us to organize our ideas into
contrasting sets of alternatives. Even more abstract is the relation ‘sex’ or ‘age’,
which links a person to one of these concepts. To anticipate the discussion of
such things in language (7.3), we can call sex and age a FEATURE. A feature is
a kind of relational concept whose value is one of these abstract concepts, shown
in Figure 3.8 as the male and female symbols. The diamond arrows are explained
A notation for choice setsâ•‡ nnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnn
The second step in understanding features such as sex and age is
to look at the way in which the alternatives are organized so that we know, for
example, that ‘male’ is a possible value for ‘sex’, but not for ‘age’. In each case,
the alternatives are defined either by a list of members (e.g. ‘male’, ‘female’) or
by a description of the typical member (‘a measure of time’); in more technical
terms, they’re defined by a SET, a notion that you may have met in mathematics.
an i n t r o d u c t io n t o wo r d g r a mm ar
In Section 2.4.3 I called this a CHOICE SET, a set from which only one
member may be chosen (or, if you prefer, a collection of ‘opposites’). Each choice
set is itself a concept, with its own network node which we can represent in a diagram simply as a dot, such as the dot in the top right-hand corner of Figure 3.8.
Its members are the competing alternatives such as ‘male’ and ‘female’, so if we
know that ‘male’ belongs to such a set, we can find out by consulting the set node
that the only alternative is ‘female’.
The relation between ‘male’ or ‘female’ and its choice set is different from
any other relation we’ve considered so far, and seems to be another in our small
set of primitive relations. The obvious short name for this link is OR. If two or
more concepts have ‘or’ links to the same node, then they must be competing
In terms of notation, the ‘or’ relation is shown as an arrow with a diamond at
its base. (If you want a mnemonic, think of the diamonds that are sometimes used
in flow charts to show decision points because they have a number of alternative
attachment points.) You can see this notation in Figure 3.8.
The benefits of features, and their limitationsâ•‡ nnnnnnnnnnnnnnn
Like any other relational concept, a feature can be used even when we
don’t know its valueÂ€– when, for example, we want to know the value, as when
we ask What’s the sex of your baby? This is the unspecified use of ‘sex’ shown at
the bottom of Figure 3.8 by the arrow linking ‘person’ to an unspecified member
of the choice set, which must of course be either ‘male’ or ‘female’.
Features are important in generalizing about similarities or differences. For
example, if we can talk about the feature ‘colour’, then we can say that my shoes
both have the same colour; and then we can generalize to all shoes, saying that
one’s shoes should always have the same colour. It would be easy to say that
they’re both brown, or both black, but to make the generalization we have to be
able to separate the colour from the shoe and treat it as something that the shoe
‘has’ (comparable with its sole and heel).
This is what we do whenever we match two things ‘with respect to’ (or ‘in
terms of’) some abstract feature such as size, colour or whateverÂ€– a very commonplace and basic mental operation, but one which requires some high-level
mental apparatus. Continuing with shoes, they’re also supposed to match for
size, so ‘size’ is another feature; but they must be opposite for whatever we call
the left/right contrastÂ€– ‘leftness’? ‘rightness’? This is certainly a feature because
we use it for comparisons, but it’s one that has no name.
It’s very tempting at this point to think that features are so useful that we can use
them instead of taxonomies. (Unfortunately, this is a temptation that many linguists
haven’t been able to resistÂ€– Wikipedia:Â€‘Feature (linguistics)’.) For example, if
‘person’ has the feature ‘sex’, contrasting values called ‘male’ and ‘female’, we
might be tempted to dispense with the categories with the same names, ‘male’
and ‘female’. But this would be a mistake, because most subclasses aren’t neatly
Figure 3.9 Man, boy, woman and girl defined
organized in terms of features. What’s the ‘opposite’ of ‘shoe’, or ‘dog’? It’s far
too easy to think of other examples like these where feature structures just don’t
seem to be relevant.
Ordinary classification is very much simpler than features and it’s important
to keep the two ideas separate. However, one of the attractions of features is that
they provide very clear evidence for one of the general characteristics of ordinary
Take the categories ‘male’, ‘female’, ‘adult’ and ‘child’, which provide a very
basic cross-classification of people in which sex cuts across age to define four
categories:Â€‘male adult’, ‘female adult’, ‘male child’ and ‘female child’Â€– in other
words, ‘man’, ‘woman’, ‘boy’ and ‘girl’. The choices are shown, using the notation introduced above, in Figure 3.9, together with the four categories that the
choices allow. The main point is that the analysis does not allow combinations
like ‘man–woman’ or ‘man–boy’.
Advanced:Â€Part II, Chapter 7.3:Â€Morpho-syntactic features, agreement and unrealized
Examples of relational taxonomies
Kinship deals with relations between members of a familyÂ€– a central
part of the I-society discussed in Section 2.6. We’ve already considered a very
an i n t r o d u c t io n t o wo r d g r a mm ar
small set of kinship concepts including the relation between my mother and me
(Figure 3.5 and Figure 3.6), but we can now develop these ideas a little.
Take the Simpsons, for example. (In case you don’t know who they are, you’ll
find them well documented in Wikipedia.) The family has five members (plus a
cat and a dog which, interestingly, Wikipedia lists as family members, but which
I shall ignore for present purposes):
Homer, the father;
Marge, the mother;
Bart, the ten-year old son;
Lisa, the eight-year old daughter;
Maggie, the baby.
There are more distant relatives, but this little nuclear family will give us plenty
to talk about. Our analysis, therefore, recognizes five Simpson entities:Â€Homer,
Marge, Bart, Lisa and Maggie, who could easily be classified for sex and maturity in the taxonomy of Figure 3.9.
What we’re concerned with here is not how to classify them as individuals, but
rather how to classify their relationships to one another. We need relations such
as ‘father’, ‘mother’ and ‘parent’, but we also need to be able to talk about how
they fit togetherÂ€– about the relations among the relations, a very abstract idea
indeed but (I claim) one that’s central to all our thinking.
What’s needed is a two-level taxonomy of relations in which sex is ignored at
the higher level but recognized at the lower; so, for example, ‘parent’ divides into
‘mother’ and ‘father’ and ‘child’ into ‘daughter’ and ‘son’. For the other higherlevel relations, we have to use somewhat more rarified terms:Â€‘spouse’ and ‘sibling’, but although the term sibling isn’t part of ordinary English for most of us,
there can be little doubt that everyone recognizes the concept. We happen not to
have an ordinary name for it, but German does:Â€Geschwister, which is a perfectly
ordinary word with much the same stylistic feel as our parent. The taxonomy is
shown in Figure 3.10.
And now for the complicated part:Â€combining these two taxonomies. If there
are five entities and everyone is related to everyone else, then there are 5 × 5 =
25 related pairs; but to make it even more complicated, every pair involves two
different relations depending on who you take as the ‘argument’ (in the sense
defined in Section 3.2 above). For example, think of Homer and Marge:Â€ he’s
her husband, but she’s also his wife. This gives no fewer than 50 relations to be
defined just in this tiny nuclear family.
The point is not, however, that family relations are too complicated to understand. On the contrary, in everyday life we have no difficulty at all in coping with
them. The point is that even our most ordinary cognitive abilities are impressive.
Any theory of cognition must recognize these abilities (and especially so if it’s
laying the grounds for a theory of the most complex cognitive ability of all,
Figure 3.10 A taxonomy of family relations
Figure 3.11 How three of the Simpsons are related
In practical terms, of course, it’s hard to diagram 50 relations. I’ll just diagram a few of them and leave the rest to your imagination. Figure 3.11 shows
just three of the Simpsons:Â€Homer, Marge and Bart. The main point of this diagram is to show how their relations can be classified via the relation taxonomy in
FigureÂ€3.10. As I explained in connection with Figure 3.6 in Section 3.2 above,
all the relations should strictly speaking be shown in the same way as ‘son’, with
an isA link to the general category; but this would have made the diagram even
more complicated, so I cheated by adding labels directly to the relation arcs.
As I explained earlier, this rather complicated diagram merely lays out for
inspection a tiny fragment of what you and I know already and understand without any difficulty at all. Indeed, any two-year-old can recognize basic family
relations and their implications in terms of who sleeps with who, who cuddles
who (and how), who looks after who and so on. We return to these ideas in
Section 8.7.5, where they illustrate the link between language and culture.
an i n t r o d u c t io n t o wo r d g r a mm ar
Summary of this subsection:
We have a rich repertoire of relational concepts for distinguishing kinship relations which involves a taxonomy and pairs of reciprocal relations (such as child–parent).
We apply these relations in a rich cognitive network for the members of
our family whose complexity is comparable with the network we need
Interpersonal relationsâ•‡ nnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnn
I-society provides our next example as well. I make no apology for
this, because social relations provide an important foundation for the even more
complicated relations found in language; but in any case, a lot of linguistic choices
are sensitive to the social relations between the speaker and hearer. For example,
I have a number of names that people choose according to how they see their
relation to me:Â€I’m Dick to my wife and friends, Dad to my daughters and either
Professor Hudson or just Prof to my dentist. Such linguistic choices involve what
sociologists call INTERPERSONAL RELATIONS, the relations between two
people who interact in some way. (Wikipedia:Â€‘Interpersonal relationship’.)
A particularly important analysis of interpersonal relations was proposed
by the psychologist Roger Brown. It recognizes two contrasts:Â€ POWER and
SOLIDARITY. According to this analysis, your relation to someone else has a
‘vertical’ dimension of power in which you’re superior, equal or subordinate to
the other person, and a ‘horizontal’ dimension of solidarity, ranging from distant
strangers to close intimates. (Wikipedia:Â€‘Power (communication)’ and ‘Social
In principle, these two dimensions cross-classify one another, giving six logically possible combinations of superior, equal and subordinate with intimate and
stranger. Among intimates, your child (or cat) is a subordinate, your friend is
an equal and your mother is a superior; and among strangers, a child is a subordinate, another student is an equal and your boss or professor is a superior.
But the six combinations don’t all have the same status when we think of their
consequences for behaviour. At least in modern western society, we tend to treat
equals and inferiors in much the same way, in contrast with the respectful behaviour we reserve for superiors; and we have many ways of expressing intimacy
but very few for showing distance other than the absence of intimate behaviour
This polarization emerges very clearly in language, where we often find just
two options, one for intimate non-superiors and the other for superior non-
use title and/or
use given name
use title (e.g. Mum)
Figure 3.12 Four interactive relations and their default behaviours
intimates; for example, French speakers use tu when speaking to the first and
vous to the second, and English speakers use given names to the first (e.g. Dick)
but titles and family names (e.g. Professor Hudson or Sir) to the second. These
two combinations are the ones where behaviour patterns cluster most clearly, in
contrast with other combinations where we may be quite uncertain how to behave. For instance, what do you call your teacher who you’ve known for years?
It seems, then, that interpersonal relations are organized round just four relational concepts. Simply interacting with someone establishes a basic interpersonal
relation which we may call OTHER, and which is linked to certain behaviour
patterns such as eye contact which show that we are ready to communicate with
them. (Wikipedia:Â€‘Social interaction’.)
In addition, there are two special kinds of ‘other’:Â€ INTIMATE and
SUPERIOR, each of which carries various implications for behaviour. These
two can combine in the prototypical INTIMATE SUPERIOR, one’s parents
and other senior family members, a relationship with its own special linguistic
signal:Â€terms such as Mum or Auntie, which are used instead of the usual first
names demanded for intimates. This analysis is shown in Figure 3.12, which for
simplicity shows the relevant behaviours as properties.
This analysis of interpersonal relations shows the benefit of organizing relations in a taxonomy. The analysis only recognizes two kinds of relations, in contrast with the six kinds defined by three degrees of power and two of solidarity,
and thereby explains why some power–solidarity combinations are more clearly
defined than others.
Moreover, by organizing the categories in this way we allow multiple default
inheritance to apply. This allows just the right generalizations, with ‘eye contact’ applying to all relations, whereas ‘use given name’ applies by default to
an i n t r o d u c t io n t o wo r d g r a mm ar
typical intimates, but not to intimate superiors. This is why Bart Simpson calls
his mother Mom rather than the default Marge, although he applies other default
intimate behaviours such as kissing.
Summary of this subsection:
Interpersonal relations are the relations between people who happen to
be interacting at a given time.
These relations can be analysed in terms of two independent dimensions:
power (superior, equal or inferior) and solidarity (intimate or distant).
Although the dimensions are independent, they interact in defining behaviour in terms of just two ‘clear cases’:Â€‘intimate’ and ‘superior’.
Space and timê•‡ nnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnn
For a very different kind of relationship, we turn to relations of space
and time, for which we typically use prepositions such as in, behind, before and
during as in (1) to (4).
The ball is in the box.
The ball is behind the box.
It rained before the party.
It rained during the party.
In each of these examples, the position of one thing (in space or in time) is
defined in relation to another; for example, the ball is located relative to the box,
and not the other way round. The first two examples would be good answers
to:Â€Where is the ball?, but not to:Â€Where is the box?
Objectively speaking, the box and the ball may be the same size and in
other respects equal, but these sentences assume a particular perspective in which
their relation is unequal. Psychologists describe the box as the ‘background’,
or simply ‘ground’, and the ball as the ‘figure’, and almost every introductory
psychology textbook includes the picture in Figure 3.13 to make the point that
what you seeÂ€– in our terms, your percept (3.1)Â€– depends on how you divide the
visual input into a figure and a background. In this example, if you take the white
part as the background, then you see two faces (i.e. these constitute your figure);
but if the black part is your background, then your figure is a vase. And of course
you can switch between the two percepts, but you can’t have it both ways at the
same time. (Wikipedia:Â€‘Figure-ground (perception)’.)
Figure 3.13 Figure or ground?
Although the term ‘background’ seems reasonable when talking about pictures, it’s less helpful when we’re talking about balls and boxes, where we don’t
think of the box as in any sense the ‘background’ to the ball; indeed, the background in a picture isn’t an object but just the part of the picture that’s left over
when we remove the figure.
A much better term is LANDMARK, introduced by the linguist Ronald
Langacker, who recognizes that it means much the same as the psychologists’
‘background’ (Langacker 1987:Â€233). Landmarks are fixed points that we use for
navigating, and are always identifiable objects such as church towers or trees.
This comparison is exactly right for balls and boxes, or rain and parties, where
we use the box or the party as a fixed point from which the ball or rain takes its
‘position’ in either space or time. (We shall see below that there is probably no
need for the term ‘figure’.)
The theoretical claim implied by this discussion is that when we think about
where a thing is or when an event happens, we have to think in terms of landmarks; and similarly, when we’re planning our own behaviour such as deciding
where to put things, we plan in terms of landmarks. Consequently, the first step
in saying where something is or should be is to find a suitable landmark.
The Best Landmark Principle
To be suitable, a landmark must combine two qualities:Â€ ‘prominence’ and nearness. These qualities combine to define the BEST LANDMARK
PRINCIPLE:Â€ the best landmark is the one that offers the best balance of
an i n t r o d u c t io n t o wo r d g r a mm ar
Figure 3.14 Landmarks tend to be local
prominence and nearness. Prominence means being easy to find; if you want to
tell me where my socks are, there’s no point in telling me they’re next to my shirt
if I don’t already know where this is, and in general we choose landmarks that
are either already known or easy to find. (Jumping ahead to the ideas of Section
4.2, a good landmark is one that’s easy to activate mentally.)
Moreover, an object’s landmark should always be easier to find than the object itself, otherwise it’s not much help as a clue to the object’s whereabouts.
Consequently, we typically use larger and more prominent objects as landmarks
for smaller and less prominent ones. But size and prominence aren’t the only
things that count, and in syntax we shall see that a very small word such as is can
act as a landmark for much more prominent ones on the basis of more abstract
Prominence is balanced against nearness, since a nearby landmark may be
more helpful than a more prominent but distant one. Take my lost socks again. If
all that counted was prominence, then we might choose the entire house rather
than my shirt. But however easy the house may be to find, it’s too remote:Â€the area
‘in the house’ takes much longer to search than the area that’s ‘by the shirt’.
Our desire to strike the best balance between these two qualities helps to
explain a very general fact about landmarks that are linked in a chain so that
some landmark A is the landmark for B, which then serves as the landmark for
C. What about the relation between C and A?
Suppose you tell me that my shirt is to the left of the chair, while my socks are
to the right of my shirt; where do you think my socks are in relation to the chair?
In principle, they could be either to the right or the left of the chair, but most of
us would assume that they’re on the same side of the chair as their landmark, the
shirt. In short, we assume the spatial pattern shown in ‘map’ (a) of Figure 3.14
rather than the one in (b).
Why do we make this assumption? The Best Landmark Principle offers an
explanation. If A is the best landmark for B, then in principle it could also have
been used for C, but since it wasn’t, B must have the advantage of being nearer
to C. Putting it more generally, we assume that any object is nearer to its own
landmark than it is to the landmark of its landmark.
The Best Landmark Principle doesn’t only guide us in interpreting what other
people tell us, but it also guides our own behaviour. If we’re thinking about where
to put socks and shirts, we know it’s important to remember where they are and
therefore follow the Best Landmark Principle of locating things ‘where we can
find them’, which means in a memorable relation to a memorable landmark. That’s
why we have furniture such as cupboards that are always in the same place.
Different ways of relating to a landmark
Having found a landmark, of course, we then have to choose a relation to
that landmark. Is the ball in the box, or behind it or on it or…? Did it rain before, after
or during the party? And so on. This involves another part of the grand taxonomy of
relations, the part for relations in space and time. Let’s focus on time, because this is
the dimension most relevant to language (or at least to spoken language).
The basic temporal relations are, of course, ‘before’ and ‘after’, though there
are others (until, since, during, at, in, for). If we say that Thursday comes before
Friday, we’re taking Friday as the landmark for Thursday. We often use the words
before and after to link events, but even then we’re really thinking of the times
of those events; if we say that it rained before the party, we’re taking the party’s
time as the landmark for the rain’s time.
One general question is exactly how these finely classified relations mesh with
the basic ‘landmark’ relation. If the party is the landmark for the rain, how should
we add the information that the rain was before the party rather than after it? One
possible answer is that we’re dealing with two separate relations:Â€the ‘landmark’
relation which identifies the landmark, and the temporal relation which distinguishes ‘before’ and ‘after’.
But since the temporal relation always involves the landmark, a much easier
analysis recognizes ‘before’ and ‘after’ as special cases of ‘landmark’Â€– ‘before
landmark’ and ‘after landmark’, as it were. For instance, if the rain was before
the party, then the party is not only the rain’s landmark, but more precisely it’s the
rain’s ‘before landmark’. These isA links are shown in Figure 3.15, alongside an
analysis showing that the rain’s relation to the party isA ‘before’, which in turn isA
‘landmark’. Notice that ‘before’ and ‘after’ are competing alternatives, as you can
see from the ‘or’ relations that link them both to the same choice-set node (3.3).
One reason for this expedition into the vast territory of spatial and temporal
relations is simply to show how the theory applies to something other than social
relations; but these relations also have a special relevance for linguistics. For
one thing, they’re needed in the semantic analysis of many wordsÂ€ – not only
the prepositions discussed above, but also verbs such as follow and adjectives
like previous, and they’re even found right in the heart of grammar, in the tenses
(where a past-tense verb refers to an event that happened before now).
Above all, we need these relations in the analysis of word order in Section 7.4,
where the ‘before’ and ‘after’ relations fix the order of words in a sentence. We
shall find that these relations are precisely as expected in a sequence of events
(words):Â€one word always takes its position from another and stays as close as possible to it. Landmarks and the Best Landmark Principle are exactly what we need.
Advanced:Â€Part II, Chapter 7.4:Â€Default word order