Cognitive Lexical
Cristiano Broccias (Genova)
[email protected]
PART I (Introduction)
‘Traditional’ lexical semantics
Cognitive semantics (prototypes, conceptualisation,
metaphors, conceptual spaces and frames)
Lexical classes and cognitive abilities
Simultaneity constructions (while, as)
Some (apparently) very simple questions involving
What is a cat?
What is beauty?
What is a red pen?
Is the Pope a bachelor?
How do we distinguish between
mosquito net and butterfly net?
lexical semantics
 sentence semantics
 text/discourse semantics
Two underlying assumptions:
 it is possible to identify lexical items
 it is possible to ‘isolate’ lexical meanings
Some basic notions:
- homonymy (e.g. bank)
- polysemy (e.g. mouse)
- monosemy:
There’s some fruit in the bowl.
There’s a crack in the bowl.
Major approaches:
- Structural semantics
- Semantic features
- Cognitive semantics
More recently also:
Interaction between
constructions and lexical items
Some naive conceptions about meaning
the meaning of an utterance consists of the sum of the
meaning of its parts (the building block metaphor):
red pen
mosquito net
butterfly net
Referential theory of meaning:
a word means what it refers to (e.g. we may
point to a cat to understand cat)
Some problems:
- abstract concepts (e.g. beauty)
- Hesperus and Phosphorus (different intensions
or senses but same extension or meaning, i.e.
Venus), the British Prime Minister (different
extensions but same intension)
Ogden and Richard’s (1923) semiotic triangle
sense (Sinn)
e.g. word
The systemic (i.e. NETWORK) nature of meaning
Words enter into various sense relationships with one another:
deictic verbs
‘Va bene. Ti porto il libro domani.’
‘Ok. I’ll bring the book tomorrow.’
‘Va bene. Porto il libro alla biblioteca domani.’
‘Ok. I’ll take the book back to the library tomorrow.’
vision verbs (semantic field of vision)
longer duration
shorter duration
by chance
We’ll come/go  back to networks later!
willful act
look (stare, gawk)
Structual semantics (see Lyons)
Three major types of relationship:
 hyponymy
 oppositeness
Synonymy (same denotation)
Context dependency:
pedigree  animals
ancestry/genealogy/lineage [ˈlɪn i‿ɪdʒ]  human beings
descent  both
The {peel/skin} of the orange is thick.
The girl’s {skin/*peel} is sunburned.
Many synomyms differ in respect to their
Register, social and geographical variation
What do you call this?
toilet (BrE)
lavatory (BrE), lav (informal)
WC (BrE, used especially on signs in public places)
the gents and the ladies (BrE, used for public conveniences)
loo (BrE informal)
bath/rest/washroom (AmE, cf. Italian ‘bagno’) = BrE toilet
john (AmE informal)
Hyponymy (i.e. category membership)
bass [bæs]
salmon [ˈsæm ən]
chinook [(ˌ)tʃ ɪ ˈnuːk] spring coho [ˈkəʊ həʊ]
sockeye [ˈsɒk aɪ]
It may be problematic to identify the superordinate terms:
brother & sister < sibling (formal)
uncle & aunt < ?
cow & bull < cow/cattle (collective)/bovine (technical)
human being & animal < animal (vs. vegetable, mineral)
Semantic networks
e.g. natural kind terms
But there are various problems with this model (apart
from the obvious fact that not all information is easily
represented in hierarchical form):
(1) A cow is an animal.
(2) A cow is a mammal.
Reaction time is faster in (1) than in (2) even though
‘animal’ is higher in the hierarchy than ‘mammal’!
(3) A pine is a church.
(4) A pine is flower.
Reaction time is faster in (3) than in (4) even though they
are both equally untrue (relatedness effect).
(5) A robin is a bird.
(6) A penguin is a bird.
Reaction time is faster in (5) than in (6)
even though both involve one semantic
link (prototypicality effect).
- Complementarity
- Antonymy
- Converseness
either X or Y, not both – non gradable concepts
single vs. married
dead vs. alive
legal vs. illegal
asleep vs. awake
true vs. false
male vs. female
pregnant vs. not pregnant
on vs. off
pass vs. fail
However we can sometimes think of
intermediate cases:
divorced (cf. single vs. married)
hermaphrodite (cf. male vs. female)
gradable concepts (e.g. scalar adjectives)
big vs. small
high vs. low
small vs. large
wet vs. dry
hot – warm – lukewarm – cool – cold
The reference value is context dependent:
A small elephant is a large animal.
A large mouse is a small animal.
A warm beer and a cold coffee may be the
same temperature.
Context dependency:
young  animate beings
new  inanimate objects
old  both
bitter  beer
sour  fruit
sweet  both
With scalar pairs, one is usually
How old are you?
How tall are you?
Context dependency:
in summer: How hot is it?
in winter: How cold is it?
relational opposites
verbs of transfer:
buy/sell, lend/borrow, give/receive
More examples
kinship terms and professional relationships:
husband/wife, brother/sister
teacher/student, employer/employee, host/guest, lawyer/client
time and space:
in front of/behind, outside/inside, north of/south of
Apparent cases of converseness:
What was often referred to as context
before can be related to what is also
traditionally called the syntagmatic axis:
Semantic features
(decompositional theories)
Semantic features are assumed to be
universal, part of our cognitive system.
Attempts have been made at reducing the
number of features to a few semantic
primitives, see e.g. Wierzbicka’s work.
But there are various problems with these
For example, there are categories which
do not have any obvious defining features
that are common to all their members, e.g.
Wittgenstein’s (1958) game example
(game is a category based on family
Further, many categories have fuzzy
boundaries. For many people it is unclear
whether a tomato is a fruit or a vegetable,
or both.
Cup, vase or bowl?
Semantic features or primitives might not
have linguistic counterparts (i.e. they might
be non-verbal).
Still, it seems likely that we (at least sometimes)
represent the meanings of words as
combinations of semantic features.
For example, we remember better sentences
“Pat sold the wand to Harry”
“Pat gave the wand to Harry”
Sell is more ‘complex’ than give.
Cogntive semantics
1970s as a reaction against truthconditional semantics
 research on prototypes (Rosch)
“best example” of a category: e.g. blackbird vs. penguin
for the category ‘bird’. But notice that the prototype may
be abstract.
Category membership is culture-dependent:
More on prototypes
not necessarily incompatible with
feature theories
fuzzy boundaries
family resemblance
Levels of categorization (e.g. furniture vs. chair vs. armchair)
In general, the closer an item is to the prototype, the easier we
process it. Further, basic level categories are easier to learn and
Prepositions (Lakoff 1987, Sandra and Rice 1995, Tyler &
Evans 2001)
reconciliation between
monosemists, homonymists and
Two major points
1)words as ‘access points’ (i.e. meanig as construal)
(scaffolding metaphor vs. building block metaphor)
mosquito net vs. butterfly net
As we build houses by using scaffolding so we
‘build’ complex meanings by using words
2)continuum between lexical items and
constructions (cf. kick the bucket)
Construction type
Complex and (mostly) schematic
Traditional name
Complex and (mostly) specific
Complex but bound
Atomic and schematic
Atomic and specific
word class
noun verb noun (i.e. transitive construction),
adjective noun (i.e. noun phrase)
I love you,
black cat
verb, adjective, noun, pronoun
love, black, cat, I, you
NB. On this view, plural –s is actually a
schematic noun:
semantic pole
phonological pole
Grammar and lexicon in Cognitive Grammar
(cf. Langacker 2008)
Different senses (lexical meanings)?
He sliced1 the bread.
Pat sliced2 the carrots into the salad.
Pat sliced3 Chris a piece of pie.
Pat sliced4 and diced his way to stardom.
Pat sliced5 the box open.
An alternative, constructionist solution
(which minimizes lexical polysemy):
one lexical meaning vs. various different constructions
He sliced the bread.
Pat sliced the carrots into the salad.
(caused motion)
Pat sliced Chris a piece of pie.
Pat sliced and diced his way to stardom.
(way- construction)
Pat sliced the box open.
Construction Grammar (Goldberg 1995, 2006)
Sem CAUSE-MOVE < cause goal theme >
Sem CAUSE-MOVE < cause goal theme >
< slicer
sliced >
But is this sharp distinction always possible?
Some evidence (from ‘resultative’ constructions):
She named the baby *(Sally).
He cut the bread (thin).
He painted the door (red).
He wiped the table (clean).
He talked himself hoarse.
He ran his sneakers threadbare.
She ate herself healthy.
Incidentally, RCs are probably motivated by the metaphor
event (E)
change (C)
(Sarah) (kissed) KEITH (the anxiety) (away from) (Keith)
Figure 25.
The FCS for Sarah kissed the anxiety away from Keith
Important assumptions in cognitive
1) The embodied cognition thesis
2) Meaning as conceptualization
1) The relation between conceptual structure
and the external world (embodiment)
a. She’s in love.
b. She’s slowly getting into shape.
c. She fell into depression.
The CONTAINER image schema is projected
onto the abstract conceptual domain of STATES.
(metaphorical mapping)
More on metaphors
The ‘fundamental roots of language are
figurative’ (Carter 2004);
metaphors are everywhere;
metaphors are systematic and culture-specific;
to stress the fact that metaphors are not
just literary devices but are pervasive, the
term conceptual metaphor is now used;
metaphors can be described as mappings
from a source domain to a target domain.
TIME IS MONEY (culture-specific)
source domain: money
target domain: time
How do you spend your time?
You’re wasting my time.
These are conventionalised (also known
as ‘dead’) metaphors: we are not
consciously aware of the metaphorical
source domain: war
target domain: argument
She attacked every point in my argument.
She tried to buttress her argument.
He withdrew his offensive remarks.
I hit back at his criticisms.
We have set out to prove that our theory is
Should we move on to the next point?
We have arrived at a disturbing conclusion.
If you don’t support your argument with
solid facts, its whole structure will collapse.
He showed her argument to be without
I’m tired of your empty arguments.
Your argument doesn’t have much
That argument has holes in it.
I’m feeling up.
That gives me a lift.
I’m down/low.
My spirits sank.
Of course, UP is not always positive and
DOWN negative:
He’s screwed up.
Depth of understanding.
Conventionalised metaphors involving
upper body parts:
John is a pig.
lean mapping  highlighting specific
aspects of the target concept
rich mapping  supplying a tangible
conceptual structure for abstract target
The mountains are sleeping.
Metaphor (personification): mountains
(target) are human beings (source)
but an alternative possibility is:
‘the mountains’ stands for ‘the people
living in the mountains’
The latter is an example of metonymy,
which, like metaphor, is not necessarily a
poetic device:
All hands on deck
used to say that everyone is needed to
help in a particular situation
With only half an hour to get everything
ready, it was all hands on deck.
Is all hands on deck just a metonymic
Metonymies can also be described as mappings
from a source to a target but they involve one
cognitive domain while metaphors involve two.
Metonymies typically have a referential function
(e.g. ‘the White House’ stands for or gives
mental access to ‘the President of the US’);
but they may have a highlighting function, as in
I’m all ears. (Remember that metaphors can also
have a highlighting function; hence, some
researchers claim that metaphor and metonymy
should be seen as a cline of cognitive
Metaphor and metonymy have been used to
investigate the conceptual structure of emotions
(sadness, anger, disgust/hate, fear,
joy/happiness, desire/love).
Metaphors are also routinely used in science
(e.g. when we speak of a ‘computer virus’) and
politics (e.g. when we say that ‘a country is ill’).
Metaphors are also used in linguistics:
e.g. complex expressions (‘red pen’) are usually
analysed in terms of the BUILDING BLOCK
But the BUILDING BLOCK metaphor is
hardly correct, cf. ice cream, newspaper,
An alternative metaphor for complex
expressions is the SCAFFOLDING
METAPHOR (the constituents are merely
the scaffolding for the construction job at
Back to meaning as conceptualisation
2a) ‘linguistic units’ as conceptualization
Morphemes, words (open-class and closed
class), constructions (e.g. active vs. passive) all
have meaning and refer to concepts in the mind
(vs. objectivism).
However, such concepts relate to our interaction
with the external world (vs. subjectivism), cf.
bachelor ‘unmarried adult male’.
Such concepts may be difficult to define (vs.
dictionary view).
2b) encyclopedic meaming
Words as ‘points of access’
‘Watch out jane, your husband’s a right
(a) The child is safe.
(b) The beach is safe.
2c) meaning construction as
The dynamic nature of meaning
construction has been explored in
Fauconnier and Turner’s Conceptual
Blending Theory (e.g. 2002).
Benetton family fancies a quick bite at Little Chef.
“Italy's super-rich Benetton family has
made approach to buy Little Chef, the
chain of roadside restaurants”
“Mental spaces are small conceptual packets
constructed as we think and talk, for purposes
of local understanding and action. … [They]
operate in working memory but are built up
partly by activating structures available from
long-term memory.” (Fauconnier and Turner
2002: 40, 102)
input space 1
(‘buying space’)
AGENT: Benetton
PROCESS: want to buy
PATIENT: Little Chef
input space 2
(‘eating space’)
EXPERIENCER: hungry person
PROCESS: fancy
input space 2
(‘eating space’)
input space 1
(‘buying space’)
EXPERIENCER: hungry person
PROCESS: fancy
AGENT: Benetton
PROCESS: want to buy
PATIENT: Little Chef
AGENT/EXPERIENCER: Benetton (hungry person)
PROCESS: fancy (want to buy)
PATIENT: Little Chef (food)
blended space/blend
input space 1
(‘buying space’)
input space 2
(‘eating space’)
cross-space mappings
AGENT: Benetton
PROCESS: want to buy
PATIENT: Little Chef
EXPERIENCER: hungry person
PROCESS: fancy
AGENT/EXPERIENCER: Benetton (hungry person)
PROCESS: fancy (want to buy)
PATIENT: Little Chef (food)
blended space/blend
What triggers input space 2 (the ‘eating
The name Little Chef.
Conceptual integration is quite a feat!
It is very difficult to spell out the various
conceptual operations we are so good at
performing quickly and unconsciously!
The blending approach differs from the
metaphor/metonymy approach in that the former
uses mental spaces constructed during online
processing while the latter operates with stored
cognitive models (e.g. BUYING A COMPANY IS
That is, mental spaces are context-dependent.
Metaphor/metonymy involves unidirectional
mappings from source to target, while blending
(typically) involves mappings from two input
spaces to the blended space.
Cross-space mappings involve so-called
vital relations, e.g. identity.
Vital relations include: change, identity,
time, space, cause-effect, part-whole,
representation, role, analogy, disanalogy,
property, similarity, category, intentionality,
The clipper ship Northern Light sailed in
1853 from San Francisco to Boston in 76
days, 8 hours. That time was still the
fastest on record in 1993, when a modern
catamaran, Great American II, set out on
the same course. A few days before the
catamaran reached Boston, observers
were able to say: at this point Great
American II is 4.5 days ahead of Northern
Blending, unlike metaphor theory,
underlines the importance of
context dependent, on-line
conceptualization (mental spaces are
different from cognitive models)
open-endedness (cf. red pen)
Blending is everywhere:
WASP (acronym with prop word)
jokes and riddles
What did the beach say when the tide
came in? Long time no sea.
Input 1:
THEME: sea/tide
GOAL: beach
Input 2:
THEME: person A
PATIENT: person B
AGENT: person A
‘Long time no see’
conceptualisation (dynamic, context dependent,
dependent on points of access)
continuum nature of linguistic units (seen as
Cognitive abilities and lexical items
- cognitive linguistics (vs. generative linguistics)
- cognitive abilities (esp. scanning) and grammar
- word classes (esp. verbs and prepositions)
- no need for scanning?
Cognitive linguistics vs. Chomskyan linguistics
Chomskyan approach:
- “Language Faculty” independent of general cognitive abilities;
- modular (i.e. separate modules as e.g. in computer science);
- language as a lexicon (i.e. a store of “words”) + a grammar (i.e.
rules to combine “words”);
- minimalist (e.g. reduce stored forms to a minimum, as in the case
of “regular” plurals like house  houses, which can be captured by
Cognitive linguistics (in particular Langacker’s Cognitive Grammar):
- language is not independent of general cognition;
- language is not necessarily a (separate) module;
- language is a “a structured inventory of conventional linguistic
units” (e.g. it includes strings which can be derived by rule such as I
love you, What are your doing tonight?, etc.  importance of
- redundancy is part and parcel of the linguistic system (cf. rule/list
- language is embodied (I take this as meaning: “you can speak a
human language iff you are human”).
Figure/ground organization:
Tom is near John.
John is near Tom.
The bike is near the house.
d. ?? The house is near the bike.
Sequential scanning (SEQ)
summary scanning (SUM):
- watching a ball fall (SEQ), as in a film
- looking at various positions of the ball at the
same time (SUM), as in a multiple exposure
build-up phase
final Gestalt
SUM vs. SEQ used in grammar as well.
Problem: how do you define word classes
(i.e. how do you distinguish e.g. verbs
from nouns)?
Two approaches:
 distributional
 meaning-based
I _______ chocolate.
(love, hate, adore, dislike, etc.  verbs)
My hair is very __________.
(long, short, etc.  adjectives)
Meaning-based (i.e. semantic or notional
cf. what you are (were?) usually taught at
school, e.g. “a noun refers to a thing or a
Obvious problem: beauty, love, happiness,
However, Cognitive Grammar takes the
notional approach to word classes
E.g. it defines a noun as a thing, which is a
technical term for a “set of interconnected
entities” (cf. team)
The Cognitive Grammar analysis of word classes
a. thing
b. relation
c. process
Schematic representations of things (a), relations (b) and processes (b)
Problem: how to distinguish notionally
between e.g. the verb enter and the
preposition into?
Intuitively, they are pretty similar (i.e.
something ends up in a place).
Langacker’s solution: we should appeal to
SUM vs. SEQ.
The rationale here is that differences in form must
always imply differences in meaning (i.e. words
belonging to different classes must be represented
differently in terms of our cognitive abilities).
I call this semantic atomism: every form (in a
construction) has meaning.
SUM vs. SEQ also used to distinguish between:
bare infinitives: She saw the ship sink. (SEQ)
to-infinitives: To eat chocolate is good for your health. (SUM)
-ing forms: She likes eating chocolate. (SUM)
Intuitively, the distinction is sometimes
problematic (see e.g. Duffley 2005):
The woman strolling down the beach is my
I found my little brother tearing my photo album
to pieces in my bedroom.
(We intuitively play the events of strolling and
tearing as motion pictures.)
Langacker himself is aware of the
somewhat speculative nature of his
analysis (1987: 235-254), see also (1999:
Nonetheless, Langacker (1987) defends
his analysis by claiming that SUM and
SEQ are needed in order to achieve
theory-internal coherence:
A hard-nosed linguist will doubtless ask for evidence to support
these claims. How can one prove that the conception of a process
(hence the meaning of every verb) requires sequential scanning
[…]? The request for justification is certainly legitimate, but we must
take some care that the form of the request does not embody
methodologically unreasonable expectations. In particular, one
cannot reasonably expect or demand the existence of direct
empirical evidence that bears on this question alone considered in
isolation from the overall descriptive context in which the analysis of
processes is embedded [emphasis ours]: I can no more substantiate
the claim that verbs imply sequential scanning—directly, and without
regard to how the total descriptive system meshes together
[emphasis ours]—than the proponent of a more fashionable model
can prove that movement rules leave traces without explicating the
function of these constructs as part of a much larger theoretical and
descriptive framework. The absence of direct and conclusive
empirical support is unfortunate, but no linguistic theory can provide
such motivation for all its constructs taken individually. (Langacker
1987: 253)
This position cannot be accepted: all
linguistically relevant cognitive abilities
postulated by Cognitive Grammar must be
supported by (direct or indirect)
independent evidence, or at the very least
be in principle amenable to experimental
So far, no psycholinguistic evidence has
been provided which confirms the
existence of SUM and SEQ.
E.g. Matlock’s research (2004, 2005) only
shows that mentally simulated motion is
involved in fictive motion processing (e.g.
The path rises quickly near the top.)
But the question of how we actually do this
hasn’t been answered yet.
The postulation of SUM and SEQ may be an
instance of the post hoc propter hoc fallacy:
SEQ implies that an element X can be inflected
but we know that X involves SEQ because X can
be inflected.
And what about languages that have verbs not
inflected for tense? (Remember that
Langacker’s characterization is meant to be
The issue of psychological plausibility should
be taken seriously.
I may very well have been being followed.
In Langacker’s analysis, this sentence involves
cyclical applications of SUM and SEQ:
(have (PERF4 (be1 (-ing (be2 (PERF3 (V)))))))
But if much in grammar is accessed as a
unit (linguistic holism), there is no need to
go through the “generative” procedure
illustrated before.
Further, what we know about language
comprehension and production casts
doubt on this analysis.
Let’s suppose Langacker’s analysis captures
(have (PERF4 (be1 (-ing (be2 (PERF3 (V)))))))
How can followed (PERF3 + V) be scanned
summarily if we haven’t processed any of the
preceding material?
Followed could be a simple past (and simple
past forms are taken to be scanned
The experimental evidence we have (see
Barsalou 1992 and Anderson 1995 for an
overview) converges towards the so-called
immediacy of interpretation:
we assign syntactic/semantic interpretations to
words as they come in.
This implies that the syncretism problem
(followed as a past participle/simple past) won’t
arise (i.e. when we reach being, we expect a
participle, not a simple past form).
What about language production?
It doesn’t seem to the case that speakers
start out at the lowest level of
constituency, and then work their way up,
step by step, in the tree or hierarchy.
(E.g. the passive schema may be
activated relatively early on.)
We conclude that Langacker’s analysis is
unlikely to be correct.
The postulation of SUM and SEQ may actually
blur the distinction between:
language as an object of investigation on the
part of the professional linguist and
language as a cognitive representation in the
speaker’s mind (see e.g. Sandra and Rice 1995
and Croft 1998).
We thus fully subscribe to Taylor’s (2002) view:
“As was the case with vowels and consonants,
there is an important sense in which the
categories of adjective and noun (and indeed
the other word classes) must be understood with
respect to the constructional schemas in which
they occur (Croft 1999). This is not to deny the
possibility of entertaining constructionindependent characterizations of the wordclasses, in terms of the nature of the concepts
that the words designate, for example (Chapter
9). Ultimately, however, a word class emerges
as a function of its role within a constructional
schema.” (Taylor 2002: 563)
The desire to see all linguistic elements as
meaningful (semantic atomism) and the
recognition of entrenchment (i.e. the view
that, because of repetition, much in
language is accessed automatically, i.e.
holism) constitutes a potentially
problematic duality in Cognitive Grammar.
On the one hand, Cognitive Grammar is a
semiotic model (see also Taylor 2003b)
where all elements are said to be
On the other, grammar is viewed as
emergent: it emerges out of concrete
forms which an individual is exposed to
and can manipulate. ( usage-based
model, cf. construction grammars, see e.g.
Goldberg 1995, 2006)
The usage-based perspective doesn’t
require maximum parcelling of meaning
(i.e. semantic atomism).
But if we don’t accept SUM and SEQ, how
can we distinguish between e.g. enter and
We should recognise the centrality of
distributional facts.
But the fact that enter and into are
distributionally different doesn’t mean that
they are identical semantically (even
without recourse to SUM and SEQ):
the cinema
a. disintegrated
b. integrated
Schematic representation of She walked into the cinema
The semantic poles of into and enter
Support for this analysis comes e.g. from
varieties where prepositions (not only into)
can be left unexpressed:
a. I need in the house.
b. “And you want into his knickers,” he
added a little laugh to put Gerry at ease.
(BNC BN1 1071)
Some concluding remarks:
SUM and SEQ needed to achieve internal coherence within a
lexicalist, semantics-driven theory;
but we lack experimental support; further, the evidence we have
doesn’t seem to support SUM and SEQ (at least as they are used in
Cognitive Grammar);
in order to develop a truly cognitive grammar, all allegedly
linguistically relevant cognitive abilities must be amenable to
experimental verification;
grammar as a semantics-driven model and grammar as a usagebased (corpus) model can coexist provided that lexical semantics is
grounded in constructions;
it is also conceivable that some structures
can’t be assigned a well-defined meaning
on their own:
“[s]peakers do not necessarily make the
relevant generalizations, even if clever
linguists can” (Croft 1998: 168)
The lexical meaning of simultaneity
Analysis of simultaneity as and while-clauses:
as and while are represented differently in our
“mental lexicon”, i.e. are associated with
different simultaneity constructional schemas;
different types of simultaneity clauses can be
recognised (they define a network);
Explicit coding of simultaneity
Various explicit devices can be used to code
simultaneity, i.e. total or partial temporal overlap,
between two events:
a. An armed robber was mugged of his loot as he
made his getaway. (BNC)
b. She said that the pain was a little better after the
pethidine she had been given and she was able to
rest quietly while she waited to be taken to theatre.
c. When he was in the airforce he flew Tornado jets.
Simultaneity (or temporal) as and while-clauses
are often compatible with additional semantic
a. She kept her head down as she spotted the
newsmen. (BNC) [causality]
b. Schools in the north tend to be better
equipped, while those in the south are
relatively poor. (BNC) [contrast]
Very little research on simultaneity.
Dynamic (multiphase) and stative (monophase) events
Morris (1996):
temporality vs. pure causality depends on
multiphase event (see (3)) vs. monophase event (see (4))
(3) As she grew older, …
a. As you are here…
b. As you know…
c. As he wore a red sweater…
In fact, as-clauses do occur with “monophase” events:
(5) The wind whips round us as we stand on the seafront. (Morrall 2003:
(6) He says it in a whisper, with his eyes upon her, as she sits at the
window bent over her work. (Waters 2002: 237)
(7) The company commander then moves in as Iman lies wounded and
helpless. (The Guardian, 24.11.2004, p.2)
(8) The bottle of Sylvaner from the cellar was cool and sweet. It reminded
him even more of Heidi. […] Her slow smile as she watched him. The
quivering strength of her grip as she held him to her. (Millar 2004: 197).
(9) … a day after eight blinging pieces of jewellery were snatched from his
bedroom as he slept with his wife, Sharon, in their Buckinghamshire
mansion. (The Guardian, 24.11.2004, p.3)
(10) My pager went off as I was on the train on Nov. 3.
( transplant/cst-nws-liverone26.html)
Admittedly, though, examples with the copular verb be
usually seem to be of a dynamic nature:
(11) When items are arranged in this way, most of the 1s
will appear as a peak at the bottom of the scale and
there will be a gradual decrease in frequency as the
attributes are less and less possible in human
performance. (Hatch and Lazaraton 1991: 204)
(12) That made me pause as I was halfway across the
building’s front plaza. (Connelly 2003: 80)
(13) As I was crouched, preparing myself for a quick raid
on the locker, a series of waves got me thinking. (Martel
2002: 169)
A genuine counterexample?
(14) As he was in the hospital, though, his
family, all the survivors from Sete, learned
that it was the Pirahãs who had attacked
them … (Everett 2008: 147)
“monophase” events are compatible with a temporal
reading (contra Morris 1996, Silva 1991);
still, truly stative be examples seem very, very rare;
the availability of causality (alongside temporality) is
(15) An embarrassment of produce becomes available to
Caroline as she walks towards The Mother’s Finest […].
(Faber 2003: 22)
(16) ‘Could it be William’s?’ she says as they walk up
the Rackham path together. (Faber 2003: 187)
If both as and while-clauses refer to temporary
configurations (hence, the impossibility of (17b)=(4b)),
then why do we have the contrast in (17a)?
a. {*As/While} you are here…
b. {*As/*While} you know…
Analysis of the first 443 pages (out of a total of 833) of
Faber’s novel The Crimson Petal and the White:
- 255 as-clauses vs. 64 while-clauses;
- while-clauses occur in contexts where either a
(relatively) long action is evoked or states/properties,
expressed through the verb be (or a modal verb), are
BNC data: BNCWeb, old version (
241 randomly selected instances of as=CJS from
imaginative (written)
241 randomly selected instances of while=CJS from
imaginative (written)
241 randomly selected instances of as=CJS from leisure
241 instances of while=CJS from leisure (spoken)
[NB 241=total number of instances of while=CJS in
leisure (spoken)]
while-clauses (imaginative): 178 examples
change verbs account for only about 20% of the data (change of
position verbs account for about 18% of the data). Only about 14%
of change-of-position verbs have the same (or part-whole) subject
as the main clause;
even considering –ing cases (i.e. while V-ing), only about 21% of the
while-examples have the same (or part-whole) subject as the main
as-clauses (imaginative): 100 examples
only one example out of 100 with be (in a pseudo-progressive
construction), vs. 19.5% for while (also, note that 3 negative whileclauses were found);
change verbs account for 72% of the data (change of position verbs
alone account for 62% of the data). Almost 50% of change-ofposition verbs have the same (or part-whole) subject as the main
more than half of the as-examples (54%) have the same (or partwhole) subject as the main clause.
while-clauses (leisure): 131 examples
in the spoken language, the use of the verb be is much more
frequent than in the written language (47.7% vs. 19.5%). By
contrast, the use of change verbs is approximately constant (20.3%
spoken vs. 20.7% written);
the percentage of same subject cases is higher than in the written
data, amounting to around 34% (also including 3 –ing cases).
as-clauses (leisure): 27 examples
in the spoken language, change verbs account for about 89% of the
data (24 tokens out of 27). One third of them are change-of-state
verbs and most change-of-place verbs (11 out of 18) are instances
of go;
the percentage of same subject cases is higher than in the written
data, amounting to around 63%.
As-clauses often involve change verbs (especially in the
spoken language; but remember that stative verbs are
also possible).
While-clauses do not show a strong preference for
change verbs. They seem to evoke more stable/static
configurations (especially in the spoken language).
As-clauses show a stronger preference for subject
identity (i.e. the degree of semantic integration between
the as-event and the main event is stronger in asclauses, see also Silva 1991 on this point).
Different lexical entries
Simultaneity while can occur with be, modals,
and negated verbs:
Instead, he eats his sausage {while/*as} it’s still
‘Because I must do something {while/*as} I still
can. […]’ (=(20))
Fat lot of use I’d be to any girl {while/*as} I
don’t have a job. (BNC: FRR 572)
While and as are associated with two
different constructional schemas:
Temporal while: [while NP VP]
Temporal as: [as NP VPchange]
That is, temporal as is more constructiondependent than while.
You could also say that, in lexical semantics
terms, while evokes temporality (i.e.
“susceptibility to change” in the sense of
Williams 2002) on its own (cf. also the noun
while) because, not relying on any specific verb
types, its temporal interpretation can be
‘detached’ from the construction in which it
Temporal as, by contrast, needs a temporal
‘exponent’ by way of the VP it occurs with.
In other words, temporality cannot be retrieved
from the verb be, modals, and negated VPs if as
is used.
But what about stative verbs (e.g. verbs of
posture) occurring with as?
Sit, stand, and lie, for example, have a high
degree of susceptibility to change (when they
apply to animate referents).
As-clauses construe path events.
as-path event
main clause event(s)
Figure 1
The motion analogue of the conceptual notion of path
easily explains the emergence of the causal meaning for
as-clauses (i.e. our perception of objects and events is
made possible by motion itself).
The fact that the notion of path is not intrinsic to whileclauses accounts for their more “static” character and the
lack of purely causal while. While-clauses can be
compared to the perception of external reality in the
absence of motion (e.g. when we look out of a window).
Further, immobility enhances the potential for an
adversative construal. Hence, the contrastive meaning of
many while-clauses.
Schematic variation in simultaneity as-clauses
A. The temporal expanse of the main clause and that of
the as-clause are usually either comparable or the
temporal expanse of the as-clause contains that of the
main clause:
(29) As she unfolded the pages this time, looking for the
picture of Harriet Shakespeare with her son, Jinny’s
hands were trembling.
(30) Once, as they were walking down St Martin’s Lane
together […] she caught a glimpse of their rippling
reflection in a shop window. (Heller 2003: 118)
B. In some cases (e.g. news reports, especially headlines), both
events are construed punctually:
(31) Five resign as police chiefs promise action agaisnt [sic] racism.
(The Wrap, 23.10.03)
(32) Tim Yeo became the latest senior Tory to rule himself out for the
leadership today as party heavyweights gave their support to
Michael Howard, who is expected to announce his candidacy this
afternoon. (The Wrap, 30.10.03)
(33) Praise for management as postal voters reject strike (The
Wrap, 18.09.03)
(34) Among the broadsheets, only the Independent chooses to lead
with something other than the Hutton inquiry, […] “Washington
suffered a double blow in its plans for Iraq yesterday as France and
Germany balked at proposals for an international force, […]”. (The
Wrap, 05.09.03)
C. Sometimes the temporal expanse of the main
clause is larger than that of the as-clause, which
can be either punctual or extended:
(35) The Telegraph highlights a row over the
“Mission accomplished” banner which hung
behind George Bush on May 1 as he declared
victory from the USS Abraham Lincoln. (The
Wrap, 30.10.03)
(36) The fog hung low on a brisk January dawn
in 2001, as several dozen police agents silently
rolled into position in the rugged hills around
Mezzojuso, a sleepy town 40km south of
Palermo. (Time Magazine, 2004, no.36, p.50)
temporal extension of main clause event (e.g. resign)
is equal to
temporal extension of as-clause event (e.g. promise);
see (31)-(34) (e.g. Five resign as police promise action)
temporal extension of main clause event (e.g. hang)
is greater than
temporal extension of as-clause event (e.g. declare);
see (35) (e.g. A banner hung behind him as he declared
temporal extension of main clause event (e.g. hang)
is larger than
temporal extension of as-clause event (e.g. move);
see (36) (e.g. The fog hung low as the police rolled into
Figure 2
The schemas in Figure 2 can be analysed as extensions
of Figure 1 obtained via the principle of family
The schema in Figure 2a arises from the compression of
the path arrow of Figure 1 into a single time point. [N.B.
The temporal equivalence between main clause and asclause is construed. In “objective” time, the event of e.g.
promising is antecedent to that of resigning (as well as
the cause for the latter).]
Figure 2b involves temporal compression and figureground reversal: The backgrounding function
prototypically assigned to the as-clause is carried out by
the main clause. Figure 2c only involves figure-ground
As and while-clauses are not identical despite what is usually reported in
dictionaries (see entry below from Longman Dictionary of Contemporary
English, CD-Version; observe that the examples are all path events!):
4 while or when
I saw Peter as I was getting off the bus.
As time passed, things seemed to get worse.
Just as the two men were leaving, a message arrived.
As and while-clauses form a simultaneity network:
as-clauses code path events. Unlike while-clauses, they are not compatible
with stative be, modals or negated VPs because temporality could not
otherwise be retrieved;
while-clauses are more stative than as-clauses: change verbs are not
peculiar to them. By considering while as a default temporal subordinator,
we can motivate its wider use compared to as;
at least two more types of as-clause (see Figure 2a and Figures 2b-c) have
been recognised (alongside the prototype in Figure 1), depending on the
relation between the construed temporal expansions of the as and main
clause events.

Slide 1