ELIA 6, 2005-6, pp. 13-39
Es udios de
lingüís ica inglesa aplicada
FORMULAIC LANGUAGE: FIXED AND VARIED
No be Schmi
Uni e si y o No ingham
Fo mulaic language has been shown o be an impo an componen o
language usage. This pape summa izes e idence o his impo ance and
hen goes on o ocus upon wo key cha ac e is ics o o mulaic language:
ixedness and a iabili y. Fo mulaic language is usually concep ualized as
being basically ixed, bu examples a e gi en o illus a e ha in many cases
o mulaic language con ains a conside able amoun o a ia ion. The deg ee
and ype o a ia ion depends on which kind o o mulaic language is being
add essed: idiom, a iable exp ession, o lexical bundle. Idioms, which a e
supposedly ixed, show he g ea es amoun o a ia ion, while a iable
exp essions and lexical bundles seem o con ain much mo e s able ixed
co es. I is sugges ed ha a iable exp essions (and pe haps lexical bundles)
may be s o ed in he mind as indi idual uni s, because he e a e ela i ely
ew ins ances o s o e. Con e sely, idioms may in ol e so many a ian s ha
only he canonical o m may be s o ed as a empla e, om which unca ed
and no el o ms can be ecognized. Teaching implica ions o hese di e en
o ms o s o age a e conside ed.
Key wo ds: o mulaic language, mul iwo d uni s, language p ocessing,
lexical s o age
14 N. Schmi
ELIA 6, 2005-6, pp. 13-39
1. The Impo ance o Fo mulaic Language
One o he cu en ‘ho ’ a eas in applied linguis ics is he s udy o o mulaic
language. I is becoming inc easingly clea ha i is an impo an elemen o
language lea ning and use, in ways ou lined o e he yea s by Pawley and
Syde (1983), Na inge and Deca ico (1992), Moon, (1997), W ay (2002),
and Schmi and Ca e (2004), among o he s. The main easons why we
should be in e es ed in o mulaic language a e summa ized as ollows:
● No mal discou se, bo h w i en and spoken, con ains la ge pe cen ages o
o mulaic language. E man and Wa en (2000) calcula ed ha 52-58% o he
language hey analyzed was o mulaic, and Fos e (2001) came up wi h a
igu e o 32% using di e en p ocedu es and c i e ia.
● I much discou se is made up o o mulaic language, hen his implies ha
p o icien language use s know a la ge numbe o o mulaic exp essions.
Pawley and Syde (1983: 213) sugges ha he numbe o "sen ence-leng h
exp essions amilia o he o dina y, ma u e English speake p obably
amoun s, a leas , o se e al hund eds o housands.” Jackendo (1995)
concludes om a small co pus s udy o spoken language in a TV quiz show
ha people may know a leas as many o mulaic sequences as single wo ds.
I mus be said howe e , ha he e is li le ha d esea ch ye o ei he suppo
o e u e hese asse ions.
● Fo mulaic language is no a homogeneous phenomena, bu a he qui e
a ied. Schmi and Ca e (2004) illus a e his di e si y:
… o mulaic sequences can be long (You can lead a ho se o wa e , bu you
can' make him d ink) o sho (Oh no!), o any hing in be ween. They a e
commonly used o di e en pu poses. They can be used o exp ess a
message o idea (The ea ly bi d ge s he wo m = do no p oc as ina e),
unc ions ([I'm] jus looking [ hanks] = declining an o e o assis ance om
a shopkeepe ), social solida i y (I know wha you mean = ag eeing wi h an
in e locu o ), and o ansac speci ic in o ma ion in a p ecise and
unde s andable way (Wind 28 a 7 = in a ia ion language his o mula is
used o s a e ha he wind is 7 kno s pe hou om 280 deg ees). They
ealize many o he pu poses as well, as o mulaic sequences can be used o
Fo mulaic language: Fixed and a ied 15
ELIA 6, 2005-6, pp. 13-39
mos hings socie y equi es o communica ion h ough language. These
sequences can be o ally ixed (Ladies and Gen lemen) o ha e a numbe o
'slo s' which can be illed wi h app op ia e wo ds o s ings o wo ds
(_[someone/ hing, usually wi h au ho i y]_ made i plain ha _[some hing
as ye un ealized was in ended o desi ed]_). (p. 3)
● Mo eo e , o mulaic language is used o ealize a numbe o di e en
unc ions in language use, including:
Func ional use. The e a e ecu ing si ua ions in he social wo ld ha equi e
ce ain esponses om people. These a e o en desc ibed as unc ions, and
include such (speech) ac s as apologizing, making eques s, gi ing
di ec ions, and complaining. These unc ions ypically ha e
con en ionalized language a ached o hem, such as I'm ( e y) so y o hea
abou ____ o exp ess sympa hy and I'd be happy/glad o _______ o
comply wi h a eques (Na inge and DeCa ico, 1992). Because membe s
o a speech communi y know hese exp essions, hey se e a quick and
eliable way o achie e he ela ed speech ac .
Social in e ac ion (pha ic communion). People commonly engage in 'ligh '
con e sa ion o pleasu e o o pass he ime o day, whe e he pu pose is no
eally in o ma ion exchange o o ge someone o do some hing. Ra he , he
pu pose is social solida i y, and people ely on non- h ea ening ph ases o
keep he con e sa ion lowing, including commen s abou he wea he (Nice
wea he oday; Cold isn’ i ), ag eeing wi h you in e locu o (Oh, I see wha
you mean; OK, I’ e go i ), p o iding backchannels and posi i e eedback o
ano he speake (Did you eally?; How in e es ing). Resea ch has shown ha
such ph ases a e a key elemen o in o mal spoken discou se.
Discou se o ganiza ion. Fo mulaic ph ases a e a common way o signpos
he o ganiza ion o bo h w i en (in o he wo ds, in conclusion) and spoken
discou se (on he o he hand, as I was saying).
P ecise in o ma ion ans e . Technical ocabula y a e wo ds which ha e a
single and p ecise meaning in a pa icula ield (scalpel is a speci ic ype o
kni e used in medicine). Bu his phenomena is no es ic ed o indi idual
wo ds. Indeed, ields o en ha e ex ended ph aseology o ansac
16 N. Schmi
ELIA 6, 2005-6, pp. 13-39
in o ma ion in a way which minimizes any possible misunde s anding. Fo
example, in a ia ion language, he ph ase Taxi in o posi ion and hold clea ly
and concisely con eys he ins uc ions o mo e on o he unway and p epa e
o depa u e, bu o wai o inal clea ance o akeo .
● The use o o mulaic language helps p o icien speake s be luen . Pawley
and Syde (1983) sugges na i e-speake s ha e cogni i e limi a ions in how
quickly hey can p ocess language, bu hey a e also able o p oduce
language seemingly beyond hese limi a ions. They look a he
psycholinguisi c li e a u e and conclude ha na i e speake s a e unable o
p ocess a clause o mo e han 8-10 wo ds a a ime. When speaking, hey
will speed up and become luen du ing hese clauses, bu will hen slow
down o e en pause a he end o hese clauses. P esumably hese pauses
pe mi he speake o o mula e he nex clause. Speake s seldom pause in
he middle o a clause. Toge he , his e idence sugges s ha speake s a e
unable o compose mo e han abou 8-10 wo ds a a ime.
On he o he hand, na i e-speake s can luen ly say mul i-clause
u e ances. Conside he ollowing examples:
1) You don’ wan o belie e e e y hing you hea .
2) I jus goes o show, you can’ be oo ca e ul.
3) You can lead a ho se o wa e , bu you can’ make him d ink.
They ha e inc easingly mo e wo ds, and Example 3 is clea ly
beyond he limi o 8-10 wo ds. Ye na i e speake s can say hem all wi hou
hesi a ion. Pawley and Syde sugges ha hese examples can be luen ly
p oduced because hey a e ac ually al eady memo ized, ha is, as
p e ab ica ed ph ases which a e s o ed as single wholes and a e, as such,
ins an ly a ailable o use wi hou he cogni i e load o ha ing o assemble
hem on-line as one speaks. Pawley and Syde sugges ha he mind uses i s
as memo y o s o e hese p e ab ica ed ph ases in o de o compensa e o
a limi ed wo king memo y (and he capaci y o compose no el language on-
line).
Fo mulaic language: Fixed and a ied 17
ELIA 6, 2005-6, pp. 13-39
O e all, hese poin s illus a e ha o mulaic language is in insically
connec ed wi h unc ional, luen , communica i e language use.
2. Mul i-Wo d Uni s: How Fixed?
I belie e he abo e summa y shows ha o mulaic sequences a e an
impo an elemen o language. I his is accep ed, hen as language
specialis s we need o know how hese sequences beha e. A conside able
amoun o wo k has been done on o mulaic language ( he bes su ey is
W ay, 2002), bu his esea ch has ended o be widely sca e ed ac oss a
numbe o ields (child L1 acquisi ion, psychology, co pus linguis ics), and
o en been o pe iphe al in e es o he esea che s (e.g., in child L1
acquisi ion, he ocus is he newly u e ed wo d s ings hemsel es, no he
ac ha hey a e o mulaic in na u e). This di usion is illus a ed by he
wide a ie y o e minology W ay (2002, p. 9) ound o he a ious so s o
o mulaic language:
chunks o mulaic speech mul iwo d uni s
colloca ions o mulas p e ab ica ed ou ines
con en ionalized o ms holoph ases eady-made u e ances.
To u he ou knowledge o o mulaic language, we can ei he look a hei
psycholinguis ic aspec s, o example, how o mulaic language is p ocessed
and acqui ed (see Schmi 2004 o his pe spec i e), o we can look a wha
migh be called he linguis ic aspec s, e.g., he o ms which a e used and
hei cha ac e is ics. I is his second app oach I would like o pu sue in his
pape .
Moon (1997) is one o he schola s who has looked a he o ms o
mul iwo d uni s (MWU) (he p e e ed e m) in de ail.1 She ocuses on h ee
ea u es o MWUs:
1. Ins i u ionaliza ion he deg ee o which a mul iwo d i em is
con en ionalized in he language

18 N. Schmi
ELIA 6, 2005-6, pp. 13-39
2. Fixedness he deg ee o which a mul iwo d i em is
ozen as a sequence o wo ds
3. Non-composi ionali y he deg ee o which a mul iwo d i em canno
be in e p e ed on a wo d-by-wo d basis, bu
has a specialized uni a y meaning.
(Moon, 1997, p. 44)
O hese ea u es, pe haps ixedness is he mos impo an , as he whole idea
behind MWUs is ha a leas some o hei componen s a e ixed, allowing
hem o be memo ized and used as wholes, a he han being newly c ea ed
o each use. On i s inspec ion, one migh assume ha all MWUs a e
comple ely ixed, bu his is no he case. O cou se some a e, and idioms a e
usually ci ed as examples. Fo ins ance, co pus e idence shows ha once in
a blue moon occu s almos exclusi ely as ha exac ph ase, and no as
a ia ions such as * wice in a blue moon, *once in a yellow moon, o *once
in a blue ime. In o he wo ds, i we wan o use an idiom o exp ess he
no ion ‘some hing which occu s e y in equen ly’, we can only use he
in ac idiom once in a blue moon, no some a ia ion. Ano he example o a
o ally ixed MWU is he wa ning ph ase Wa ch Ou ! I is ins an ly
ecognizable, p ecisely because i is ixed, and li le p ocessing should be
equi ed o unde s and i . We could shou some hing like Wa ch he ca
coming behind you!, bu i milliseconds coun , hen a sho e , mo e
con en ionalized wa ning is likely o be mos e ec i e.
Howe e , many MWUs a e no comple ely ixed, and in ac allow o a
su p ising amoun o lexibili y. Moon (1997, p. 53) illus a es his:
no ouch someone/some hing wi h a ba gepole (B i ish s. Ame ican
no ouch someone/some hing wi h a en oo pole English)
bu n you boa s ( a ying a lexical componen )
bu n you b idges
Fo mulaic language: Fixed and a ied 19
ELIA 6, 2005-6, pp. 13-39
cos an a m and a leg ( e b a ia ion)
pay an a m and a leg
spend an a m and a leg
cha ge an a m and a leg
e e y cloud has a sil e lining ( unca ion)
sil e lining
b eak he ice ( ans o ma ion)
ice-b eake
ice-b eaking
In ac , i seems ha once a MWU becomes well-known in a speech
communi y, i can be c ea i ely adap ed and s ill be comp ehensible. I is
wo h expanding on Moon’s example o unca ion. The well-known idiom
E e y cloud has a sil e lining occu s in he 100-million wo d co pus 7
imes. Bu he ph ase sil e lining occu s in he co pus 75 imes ep esen ing
he same meaning. Clea ly people p e e o use a sho ened e sion o he
idiom in p ac ice, o en in highly c ea i e ways:
-This p o ed o be much mo e han a sea ch o he p o e bial sil e lining.
-[…] may con ain a sil e lining o he consume s.
- e o med shopaholics almos always speak o a sil e lining o he cloud
which hung o e hei li es (and bank accoun s).
Beyond idioms, i is no su p ising ha many o he ypes o MWUs also
con ain a ia ion, o i is an ad an age in much o language use o allow
mo e lexibili y o meaning. Fo example, i we wish o exp ess he no ion
ha some ac i i y o achie emen is unusual, unexpec ed, o excep ional,
hen we can use ph ases like Diane hinks no hing o unning 5 miles be o e
b eak as o He hinks no hing o d i ing 100 miles pe hou on he eeway.
20 N. Schmi
ELIA 6, 2005-6, pp. 13-39
The unde lying s uc u e o hese sen ences is ‘_____ hinks no hing o
_____’, which allows he lexibili y o exp ess he ‘unexpec ed’ no ion in
many di e en si ua ions.
In his pape I will expand upon his idea o a ia ion in o mulaic
language, pa ially because he mo e I wo k wi h o mulaic language, he
mo e a ia ion I ind. Bu mo e impo an ly, I hink his a ia ion may ha e
impo an implica ions o how we heo ize ha o mulaic language is s o ed
and p ocessed. This in u n has implica ions o how we may bes go abou
eaching MWUs. I will a emp o d aw ou hese implica ions in he inal
wo sec ions o he pape .
3. Va ia ion in Fo mulaic Language
3.1. G amma ical and Lexical Va ia ion
I will s a he explo a ion o a ia ion in o mulaic sequences wi h he
no ion o g amma ical a ia ion. I ook an idiom which on he ace o i
appea s ixed: s and shoulde o shoulde . In all o he co pus analyses in
his pape , I e e ed o he Longman Co pus, a 100-million wo d co pus
based p ima ily on he B i ish Na ional Co pus. I ound ha he expec ed
‘canonical’1 o m s and shoulde o shoulde occu ed 11 imes, such as he
ollowing example:
-[…] whe e he g izzled he oes inally s and shoulde o shoulde .
As one migh expec , any ime a ph ase has a e b in i , ha e b is likely o
change i s in lec ion acco ding o he ense. The e o e, we ind cases o
simple p esen , pas , and con inuous o ms (numbe o ins ances in he
co pus a e in pa en heses):
1 The canonical o m is he mos s anda d o m, and hus he one someone is mos likely o
know. Wi h idioms, he canonical o m is likely o be he ull idiom, a he han a unca ed
a ian .
Fo mulaic language: Fixed and a ied 21
ELIA 6, 2005-6, pp. 13-39
-While F ance s ands shoulde o shoulde wi h Ge many […] (1)
-Now he ees we e enced wi h a med men s anding shoulde o shoulde .
(9)
-[…] o we s ood shoulde o shoulde wi h he Omanis in hei s uggle.
(15)
In addi ion o his kind o g amma ical a ia ion, he e a e also cases o
di e en wo d choices which do no change he meaning o any g ea deg ee:
-He and I ough shoulde o shoulde agains appeasemen . (3)
-[…] as hey wo ked shoulde o shoulde in a school bus-size labo a o y.
(2)
These a e only a ew examples, bu he poin is ha co pus e idence
clea ly shows ha
a) o mulaic sequences o en ha e a ia ion in ense
b) o mulaic sequences o en ha e a ia ion in lexical choice o one
o mo e cons i uen s.
This a ia ion is ubiqui ous, bu i would be a mis ake o hink ha i
a ec s all ypes o o mulaic language equally. One o he p oblems wi h
mos esea ch and discussion in o o mulaic language is ha i is ea ed as
one homogenous se . In eali y, he e seem o be a numbe o di e en kinds
o MWU, and each ca ego y is likely o used and e en p ocessed in
somewha di e en ways. I would no be su p ising ha a ia ion would
a ec each ype di e en ly. To illus a e his, I will examine h ee di e en
ypes o o mulaic language which ha e a ying deg ees o ixedness:
idioms, a iable exp essions, and lexical bundles.2
2 In his pape , I shall use he e ms o mulaic language, mul iwo d uni (MWU), and chunk
in e changeably as co e e ms o any kind o language ha is o mulaic in na u e. Howe e ,
28 N. Schmi
ELIA 6, 2005-6, pp. 13-39
When I sea ched o he s ing wi h a wildca d in he place o no ed, he e
we e 1091 cases, which means ha in almos exac ly hal o he cases he e
was a subs i u e wo d o no ed. Some o he mos common subs i u es a e:
-I should be emembe ed ha i is impossible o an icipa e e e y mino
de ail. (137)
-I should be emphasized ha he e was no c i icism o he o he olun ee s.
(68)
-I should be s essed ha income is only one ac o in de e mining
consump ion. (56)
-I should be clea ha he a i ude o wo ke s is de e mined by many o ces.
(39)
-I should be ecognized ha many o he p oblems ha ace us […] (38)
The e we e many o he wo ds occu ing in he wildca d slo ,
including said, added, ob ious, ecalled, and appa en . In his lexical
bundle, he e is a g ea deal o a ia ion in his ‘con en wo d’ slo , al hough
in abou hal he cases, no ed is used, making i a and away he mos
ypical o m o exp essing he no ion o ‘highligh ing’ inhe en in his
bundle.
I nex used a wildca d o check he a ia ion in he modal o his bundle i X
be no ed ha . Should ook his slo in 546 cases; a numbe o di e en
modals made up he o he 160 cases.
-i may be no ed ha (54)
-i mus be no ed ha (49)
-i will be no ed ha (37)
-i migh be no ed ha (12)
-i can be no ed ha (5)

Fo mulaic language: Fixed and a ied 29
ELIA 6, 2005-6, pp. 13-39
Howe e he o m o be seems s able; when sea ching o i should X no ed
ha , in he 546 ins ances, only he o m be was used.
In sum, i seems ha lexical bundles a e no always ixed in he
sense ha hey a e he only o m which can impa a ce ain meaning. E en
he e he e is a ia ion, al hough his p obably depends a g ea deal on he
indi idual bundle. I a bundle has a modal e b, i is likely o allow o he
modals; also, some bundles seem o allow a ia ion in con en wo ds while
o he s do no . Fo example, in he 3-wo d lexical bundle I wan o, wan can
easily be eplaced by wish o like, bu in he bundle he numbe o , i is
di icul o hink o any con en wo d ha could eplace numbe and mean
he same hing (amoun and deg ee would change he meaning somewha ).
Howe e , he bo om line is ha lexical bundles do con ain a ia ion.
4. How Do P o icien Speake s S o e and P ocess he Va ia ion in
Fo mulaic Language?
4.1. Recognizing Idioms
F om he abo e discussion we ha e seen ha he e is a ia ion in mul i-wo d
uni s. Al hough only a ew examples we e gi en, ha ac ha a ia ion was
so easy o ind sugges s ha many MWUs con ain such a ia ion. Mo e
esea ch is needed o es ablish his, bu I suspec ha a ia ion is he no m,
wi h mos o pe haps e en all MWUs con aining a ia ion o some kind. I
his is he case, hen i aises in e es ing ques ions abou how MWUs wi h
a ia ion a e s o ed and p ocessed by he mind. The ‘holis ic s o age’ heo y,
whe e each MWU is s o ed as an indi idual memo ized chunk, is he
commonly espoused iew. This app oach seems o make good sense as long
as he MWUs a e in ac , unchangeable wholes, bu uns in o p oblems i
a ia ion is inse ed. Fo example, how a e no el a ia ions o he canonical
o m ecognized? We know ha p o icien use s a e c ea i e wi h MWUs,
and once a MWU is es ablished in a speech communi y, i is o en unca ed
o he o de o he main componen s swi ched a ound. To illus a e his, I
30 N. Schmi
ELIA 6, 2005-6, pp. 13-39
ha e con i ed some new e sions o well-known idioms, which do no
occu a all in he Longman co pus, and as a as I know, a e comple ely
unique. Ne e heless, al hough you almos ce ainly ha e no seen hese
o ms be o e, I am ai ly ce ain ha you will be able o unde s and hem
well enough o answe he ques ions.
-I ha ed aking he subways in Japan. The sa dine-like ain ca s always
made me sick.
Wha exac ly is he complain abou he Japanese ains?
-Was ing ime in mee ings d i es me c azy. The wo s a e he bush-bea e s.
-I would be be e i hey we e expelled immedia ely.
How do he bush bea e s was e ime in a way ha d i es he speake c azy?
Sa dine-like e e s o he idiom packed in like sa dines, and hus he
complain is ha he subways a e a oo c owded. Bush-bea e s is ela ed o
bea a ound he bush, which means ha he speake dislikes he way ha
such people alk in ci cles and ne e ge o he poin . I you we e able o
make hese connec ions and ca ch he meaning, hen you we e able o
in e p e hese comple ely no el MWUs, e en hough you ha e ne e seen
hese o ms be o e, and e en hough he o ms hemsel es a e qui e
dissimila o he unde lying canonical o m. The p oblem his p esen s o
holis ic s o age is ob ious; hese o ms we e no p e iously s o ed in you
mind, ye you we e in all likelihood able o in e p e hem. I ollows ha
some o he p ocess is necessa y o help he mind make his ype o
connec ion.
In conside ing how his p ocess migh ope a e, i is use ul o know
jus how minimal he o mal connec ions need o be be ween a no el o m
and he canonical o m, o o s a e he issue ano he way, jus how li le o a
MWU needs o be gi en in o de o in e p e i . We can explo e his in he
ollowing con i ed example in ol ing an idiom.
Fo mulaic language: Fixed and a ied 31
ELIA 6, 2005-6, pp. 13-39
My iends and I wen ou dancing las Sa u day nigh . We i s wen o he
XENON club, and we e ha ing a g ea ime. Bu one o my iends almos
go in o a igh wi h a g oup o e y big guys and we jus managed o escape
__ ___ ____ __ ___ _____. A e ha we jus el like going home and
ha ing a quie pizza.
Can you ill in he blanks and unde s and he comple e meaning? I is highly
unlikely. Bu le us see wha happens when we ill in pa s o he idiom. Fi s
le us inse some o he unc ion wo ds. Can you make ou he idiom now?
-Bu one o my iends almos go in o a igh wi h a g oup o e y big guys
and we jus managed o escape __ he ____ o ou _____. …
When I ha e ied his in o mally wi h s uden s and con e ence a endees,
hey we e la gely unable o ecognize he idiom. Howe e , many mo e
ecognized i when he i s unc ion wo d was added.
-Bu one o my iends almos go in o a igh wi h a g oup o e y big guys
and we jus managed o escape by he ____ o ou _____. …
So i seems ha knowledge o idioms can be s ong enough ha i can be
ecognized wi hou any con en wo ds ac ually being gi en. Bu i con en
wo ds a e gi en, hen he idiom appea s o be much easie o ecognize.
Would you be able o ecognize i wi h only he ollowing con en wo d
p esen ?
-Bu one o my iends almos go in o a igh wi h a g oup o e y big guys
and we jus managed o escape __ ___ ____ __ ___ ee h. […]
How ecognizable would he MWU be wi h he i s con en wo d gi en?
-Bu one o my iends almos go in o a igh wi h a g oup o e y big guys
and we jus managed o escape __ ___ skin __ ___ _____.
In his example, i seems ha con en wo ds a e he key p omp s o
ecogni ion, as one migh expec . Con en wo ds ca y e e en ial
in o ma ion, and a e much less equen han unc ion wo ds. Howe e , i an
32 N. Schmi
ELIA 6, 2005-6, pp. 13-39
idiom s a s wi h a numbe o unc ion wo ds, an idiom may well be
ecognized be o e a con en wo d appea s. This sugges s he in ui i e idea
ha wo ds a he beginning o he idiom a e ela i ely mo e impo an o
ecogni ion, in he same way ha he coho model posi s ha he i s le e s
o a wo d a e impo an o i s ecogni ion (Ma slen-Wilson and Tyle ,
1980). I didn’ look a he possibili y he e, bu ecogni ion may well be bes
p omp ed by a ce ain combina ion o con en wo ds, in his case skin and
ee h. The e is no way o judge he ela i e impo ance o hese di e en
p omp ac o s a he momen , bu once we ha e a clea e idea o exac ly
wha p omp s he ecogni ion o MWUs, we will be much close o
unde s anding how hey a e p ocessed.
4.2. Recognizing Va iable Exp essions
As we ha e seen be o e, a ious kinds o MWU beha e di e en ly when i
comes o a ia ion. This sugges s ha i would be a mis ake o ea all kinds
o MWU he same. Le us y he same exe cise as abo e wi h a iable
exp essions, o explo e wha he minimal elemen s o ecogni ion a e o his
ype o MWU. In his a iable exp ession, he a iable slo is illed wi h my
skills.
I ook my pilo ’s lying examina ion yes e day. The es pilo was e y
igo ous and eally ___ my skills __ ___ ____. Bu luckily my ins uc o
p epa ed me well and I managed o pass.
Nobody I ha e ied his wi h has been able o igu e ou he exp ession om
only om con ex , and e y ew can ecognize he exp ession wi h only
unc ion wo ds inse ed.
-The es pilo was e y igo ous and eally ___ my skills o he ____.
Fo mulaic language: Fixed and a ied 33
ELIA 6, 2005-6, pp. 13-39
Likewise, gi ing he i s wo d, he delexicalized e b pu , helped ew
people.
-The es pilo was e y igo ous and eally pu my skills __ ___ ____.
P o iding he con en wo d in he exp ession, es , did no seem o help mos
people ei he .
- The es pilo was e y igo ous and eally ___ my skills __ ___ es .
O e all, i seems ha he e is no single elemen o he a iable exp ession
pu _____ o he es which eliably allows ecogni ion by i sel . Ra he , o
mos people, all o he ixed elemen s o he exp ession need o be p esen
o ecogni ion. This is cong uen wi h ea lie discussion ha he ixed
elemen s a e necessa y o p o ide he sca old o he slo , wi h he
implica ion ha hese ixed elemen s mus be p esen o ecogni ion. This is
highligh ed e en u he by equency igu es. The idiom by he skin o [my,
his, he e c.] ee h only occu s 32 imes in he Longman Co pus, bu could
be ecognized in con ex wi h only ce ain elemen s being p esen .
Con e sely, pu _____ o he es occu ed a ound 270 imes, ye is i ually
un ecognizable wi hou all o i s ixed elemen s being p esen .
4.3. Recognizing Lexical Bundles
Now we will y he same hing wi h a hi d ype o MWU: lexical bundles.
So, i he Consume P ice Index climbs 5 pe cen , hei bosses may be
obliged o gi e hem 5 pe cen aises o main ain hei s anda d o li ing.
Though many people ocus on ___ _______ __ ___ CPI su ey, he u h is
ha i 's a sloppy sys em.
You almos ce ainly canno guess his bundle om he con ex , so le us ill
in he unc ion wo ds.
-Though many people ocus on he _______ o he CPI su ey, he u h is
ha i 's a sloppy sys em. …

34 N. Schmi
ELIA 6, 2005-6, pp. 13-39
-You a e p obably s ill ha ing ouble ecognizing he bundle, so le us inse
he con en wo d.
-Though many people ocus on ___ esul s __ ___ CPI su ey, he u h is
ha i 's a sloppy sys em.
Wi h his con en wo d, you can p obably ge he gis o he meaning in his
con ex , bu may well no be able ecognize he bundle he esul s o he. In
his case, al hough he whole bundle is basically ixed, i is di icul o
ecognize he bundle wi h elemen s missing. This is simila o he beha io
o a iable exp essions, which also needed i s ixed elemen s p esen o be
ecognized.
5. Implica ions o Va ia ion o he P ocessing o MWUs
We ha e seen ha a ious kinds o MWU ha e di e en deg ees and ypes
o a ia ion. Bu wha does his ell us abou how hey a e s o ed and
p ocessed? I seems o me ha he di e ence in a ia ion implies ha he e
a e di e ences in s o age and p ocessing as well. Fi s , le us discuss idioms.
They possess e y conside able a ia ion, and ye people a e able o p ocess
no el a ian s. I each a ian was s o ed indi idually in i s own igh , hese
ac s could be explained. This is a possible explana ion, bu wi h so many
a ian s possible, i would be unpa simonious o speake s o ha e each
a ian s o ed indi idually. Besides, speake s a e able o ecognize no el
a ian s, which would no ye be s o ed. I each a ian is no s o ed
indi idually, hen how a e hey p ocessed?
A possibili y I would like o sugges is ha he comple e canonical
sequence is s o ed and is used as a kind o empla e o exempla . In his
empla e, he key componen s would be he con en wo ds, pa icula ly lowe
equency ones which a e less likely o combine widely. Judging om my
e y explo a o y examples, he empla e can be accessed ia one o mo e o
he con en wo ds, bu i is no usually necessa y o all o he hem o be
ecognized o access o occu . In con as o con en wo ds, unc ion wo ds
a e less use ul in accessing he empla e. My sense is ha mos empla es
Fo mulaic language: Fixed and a ied 35
ELIA 6, 2005-6, pp. 13-39
ha e a ‘co e colloca ion’ (usually made up o con en wo ds) which eliably
leads o access o he empla e. D awing on he examples I ha e used abo e,
he co e colloca ions would be:
- packed / sa dines (li e ally packed like sa dines in a box)
- sc ape / bo om / ba el (he’d had o sc ape he ba el a bi .)
(we’ e down o he bo om o he ba el)
Va iable exp essions ha e di e en cha ac e is ics. They con ain
ixed elemen s ha a e ela i ely s able, and so he e a e ela i ely ew
a ian s. O cou se he slo (s) can be illed wi h a wide ange o possibili ies,
bu i we assume ha MWUs wi h some ixed elemen s and some open
elemen s can be s o ed as an indi idual chunk, hen he limi ed numbe o
a ian s may well be s o ed as sepa a e o ms. In o he wo ds, i we assume
ha he ollowing exp essions
a minu e ago
a hou ago
a day ago
a week ago
a mon h ago
a yea ago
a e all ac ually s o ed as he single a iable exp ession
a ( ime pe iod) ago,
hen he e is only one o m o s o e, as he ixed elemen s a and ago canno
be changed. E en in a iable exp essions which do con ain some a ia ion, i
is usually no g ea , and he ew addi ional o ms would no pose an one ous
bu den o memo y. Va ia ion which does exis is o en in he ense/modal
cons i uen s (s and shoulde o shoulde / s ood shoulde o shoulde ).
Pe haps each a ia ion in hese cons i uen s en ails a sepa a e o m o
acqui e, o i migh be ha hese cons i uen s beha e mo e like a
g amma ical slo in he sequence, in which case he e would be only one
o m o be s o ed ( [in lec ion o s and] shoulde o shoulde ).
36 N. Schmi
ELIA 6, 2005-6, pp. 13-39
Lexical bundles a e simila o a iable exp essions in ha hey a e
ela i ely ixed, and so by analogy may also be s o ed indi idually.
Howe e , he e is a leas one key di e ence. Va iable exp essions ha e a
close connec ion wi h meaning and unc ional language use. Lexical bundles
on he o he hand ha e been iden i ied by co pus s a is ics and o en ha e
less ob ious ela ionships wi h any pa icula meaning o language unc ion.
Fo example, you know wha and he ac ha occu equen ly as pa o
language, bu do no seem o ealize any unique meaning o unc ion in hei
own igh . Ra he , hey a e building blocks which come o gain meaning
once combined wi h o he wo ds o lexical bundles. Gi en his lack o
dedica ed meaning, i is ques ionable whe he hey a e ac ually s o ed in a
o mulaic manne a all. In ac , he e is some p elimina y e idence ha a
leas some o hem a e no s o ed holis ically (Schmi , G andage, and
Adolphs, 2004).
I should be no ed ha some o he suppo o he abo e discussion
de i es om he ill-in- he-blank asks illus a ed in his pape . O cou se, in
he eal wo ld, people see o hea he en i e MWU, ha is, wo ds a e no
no mally blanked ou . So jus because people can’ ecognize he MWUs
wi h blanks he e, he p ocesses may di e when hey can see o hea he
comple e ph ase. Howe e , I would a gue ha he ill-in- he-blank asks can
be use ul in illumina ing he unde lying p ocesses, e en i hey do no
exac ly mi o hem. In addi ion, some componen s o a MWU may be
mishea d o mis ead because o deg ada ion o he speech s eam o page. In
hese cases, he expe ience would no be a o o he one eplica ed in he
asks.
In sum, some ypes o o mulaic language may well be s o ed
holis ically as indi idual chunks because he e is li le a ia ion, and hus
he e a e ew di e en o ms o hold in memo y. Va iable exp essions and
pe haps lexical bundles may all in o his ca ego y. On he o he hand, o he
ypes o o mulaic language, e.g., idioms, con ain a g ea deal o a ia ion,
and i seems less ob ious ha a la ge numbe o a ian s would be
indi idually s o ed. In his case, people who know he canonical o m may
use i as a empla e o exempla ; a leas his is a heo e ical explana ion
Fo mulaic language: Fixed and a ied 37
ELIA 6, 2005-6, pp. 13-39
wo h pu suing. The main poin is ha di e en kinds o o mulaic language
may in ol e di e en kinds o s o age and p ocessing.
6. Teaching Implica ions o Va ia ion in Fo mulaic Language
I di e en ypes o o mulaic language a e s o ed and p ocessed di e en ly
in he mind, hen his means we migh ha e o ake di e en app oaches in
he eaching and lea ning o he a ious ypes. I a iable exp essions a e
s o ed indi idually because o hei ela i ely high ixedness, and i he e a e
ew a ian s, hen hese o ms may be wo h eaching. This is because
a iable exp essions a e closely connec ed wi h he exp ession o pa icula
meanings o unc ions, and so end o be ecu en and use ul. I only one o
a ew a ian s need o be add essed, his is a manageable lea ning bu den
conside ing he communica i e bene i s ecei ed.
Lexical bundles may o may no be s o ed holis ically in he mind,
bu hei use as ‘building blocks’ o language (i.e., o en no ealizing
meaning con en in hei own igh ) sugges s ha hey may be less amenable
o explici eaching, simply because he o m-meaning ela ionship is less
anspa en . Howe e , some o he lexical bundles iden i ied by Bibe e al.
(1999), en ail a clea meaning (e.g., I don’ know, I wan o, on he o he
hand), and hese cases may well be wo h eaching.
Idioms a e highly con en ionalized MWUs, so much so ha hey
ha e he cha ac e is ic o non-composi ionali y. Fu he mo e, he mo e
con en ionalized a MWU is, he g ea e he chance ha i may be c ea i ely
manipula ed in language use and s ill be in e p e able. This is because
speake s know i so well ha hey can ‘play’ wi h i in c ea i e ways, and
hei in e locu o s know i so well ha hey can in e p e he c ea i e o ms.
We ha e seen jus his wi h idioms. I ha e sugges ed ha well-known, bu
a iable, MWUs may be s o ed and p ocessed as empla es which can be
ecognized om only some cons i uen s. I his is indeed so, hen o his
ype o o mulaic sequence, we migh be bes eaching he comple e
‘canonical’ o m, and hen gi ing examples o how i can be unca ed and
manipula ed.