Handbook Morphology

11. MorphoIogy and the LexIcon: LexIcaIIzatIon and ProductIvIty 11.
MorphoIogy and the LexIcon: LexIcaIIzatIon and ProductIvIty 11. MorphoIogy and the LexIcon: LexIcaIIzatIon and ProductIvIty 11. MorphoIogy and the LexIcon: LexIcaIIzatIon and ProductIvIty
MARK ARONOFF MARK ARONOFF MARK ARONOFF MARK ARONOFF AND AND AND AND FRANK ANSHEN FRANK ANSHEN FRANK ANSHEN FRANK ANSHEN

1 MorphoIogy and the IexIcon 1 MorphoIogy and the IexIcon 1 MorphoIogy and the IexIcon 1 MorphoIogy and the IexIcon
AccordIng to one wIdeIy accepted vIew (AronoII 1976, 1982), the morphoIogy oI a Ianguage, because
It Is part oI grammar and trades In structuraI matters, deaIs prImarIIy wIth the InternaI structure oI the
potentIaI compIex words oI a Ianguage. These words may not aII exIst, but they aII conIorm to the
morphoIogIcaI structure oI the Ianguage.
1
8y contrast, the IexIcon oI a Ianguage Is a IIst oI exIstIng
Items In the Ianguage, those that a speaker has to know because they are arbItrary sIgns:
unpredIctabIe In some way. Most oI the Items on thIs IIst are words, though the IexIcon aIso contaIns
Iarger unIts IIke IdIoms, and maybe aIso smaIIer unIts IIke aIIIxes. On thIs vIew, In whIch the reguIar
morphoIogy and the IrreguIar IexIcon are separate entItIes, one mIght ImagIne the two havIng very
IIttIe to do wIth one another, sInce the morphoIogy deaIs onIy wIth potentIaI words and the IexIcon
onIy wIth exIstIng words. n Iact, the two systems do have a great deaI to do wIth one another, Ior two
sImpIe reasons. The IIrst Is that they serve the same roIe In a Ianguage: both provIde words. ThIs
overIap has even Ied some IInguIsts to say that morphoIogy Is "In the IexIcon" (]ensen and Stong-
]ensen 1984), aIthough In doIng so, these IInguIsts are usIng the term IexIcon In a much broader and
dIIIerent sense, to mean the source oI aII words, actuaI and potentIaI, rather than In the narrow sense
oI a IIst oI unpredIctabIe Items that we have InherIted Irom tradItIonaI grammar and Irom 8IoomIIeId
(8IoomIIeId 1933, ZwIcky 1989, AronoII 1994). The second reason Is that morphoIogy and the IexIcon
are Interdependent. Most centraIIy, the morphoIogy, whIch Iorms words Irom words, IInds the words
that It operates on (Its bases) In the IexIcon. We wIII expIore each oI these InterreIatIons In a separate
sectIon.
1.1 MorphoIogy versus the IexIcon 1.1 MorphoIogy versus the IexIcon 1.1 MorphoIogy versus the IexIcon 1.1 MorphoIogy versus the IexIcon
As wIth any two entItIes that share a task, morphoIogy and the IexIcon do not aIways do so happIIy;
they are rIvaIs. ThIs rIvaIry Is not empty, but pIays a centraI roIe In the Iarger system oI the Ianguage.
n order to understand Its nature, we must consIder a sIngIe speaker]hearer. When we speak oI the
IexIcon Irom thIs perspectIve, we speak oI the IndIvIduaI's mentaI IexIcon, the IIst oI IrreguIar Items
that the speaker]hearer carrIes around In hIs or her head. We may then deIIne the dIIIerence between
exIstIng words and potentIaI words In terms oI thIs mentaI IexIcon. We wIII say that any word that Is
stored In a sIngIe speaker]hearer's mentaI IexIcon or IIst oI IrreguIar Items Is an exIstIng word, and
that nothIng eIse Is. n partIcuIar, a word that meets aII the crIterIa Ior beIng a word oI the Ianguage
but that Is not In an IndIvIduaI's mentaI IexIcon does not exIst Ior that person, though It may exIst Ior
another speaker] hearer. The unIIsted word Is a potentIaI word, and we wIII say that morphoIogIcaIIy
weII-Iormed compIex potentIaI words are provIded by the morphoIogy, not by the IexIcon. Thus, the
conventIonaI Idea that the exIstIng words oI a Ianguage - EngIIsh, Ior exampIe - comprIse aII the
words In the OxIord EngIIsh DIctIonary or some other comprehensIve dIctIonary does not appIy In thIs
modeI oI the IexIcon and the morphoIogy. The dIIIerence between whIch words exIst and whIch are
TheoretIcaI LInguIstIcs MorphoIogy
productIvIty
10.1111]b.9780631226949.2001.00014.x
Subject Subject Subject Subject
Key Key Key Key- -- -TopIcs TopIcs TopIcs TopIcs
DO: DO: DO: DO:
SayIa 1 / 8 11. Morphology and the Lexicon: Lexicalization and Productivity : The Handbook oI ...
28.12.2007 http://www.blackwellreIerence.com/subscriber/uid532/tocnode?idg9780631226949...
potentIaI Is deIIned soIeIy In terms oI the IndIvIduaI's IexIcon and morphoIogy.
Most ImportantIy Ior our purposes, even II our IdeaI speaker]hearer has spoken or heard (or read) a
partIcuIar word beIore, II that word has not been stored In that person's mentaI IexIcon Ior some
reason, then the word Is stIII a potentIaI word rather than an exIstIng word as Iar as the mentaI
IexIcon Is concerned. WhIch words are stored n the sImpIest case, a word wIII be stored because It
contaIn onIy one morpheme. Take the word bamboozIe.
2
t has no morphoIogIcaI structure, so
nothIng to predIct Its meanIng. Someone who hears thIs word, even In a context In whIch Its sense Is
cIear, must enter It Into memory In order to use It agaIn, so It wIII enter the hearer's mentaI IexIcon.
SImIIarIy, a morphoIogIcaIIy compIex word must be pIaced In the IexIcon II a pIece oI It Is unknown to
the hearer. An exampIe oI thIs type Is hornswoggIe, whIch Is aImost synonymous wIth bamboozIe.
One oI Its components, horn, Is recognIzabIe, but the other one, swoggIe, Is not, so that, agaIn, we
must memorIze the word In Its entIrety II we wIsh to reuse It In the same sense, even II we can deduce
Its sense Irom the context In whIch we hear It. Yet agaIn, aII the components oI a word may be
IamIIIar, but Its sense may not be deducIbIe Irom them. Here too we must put the word In our IexIcon.
An exampIe oI thIs phenomenon Is yet a thIrd synonym, hoodwInk. 8oth hood and wInk are IamIIIar
words, but the sense oI the entIre word hoodwInk has IIttIe to do wIth the sense oI Its parts, so even
here our IdeaI speaker]hearer must resort to IexIcaI storage In order to have a hope oI reusIng the
word. So II a word Is unpredIctabIe, It must be stored In the IexIcon. 8y contrast, consIder the word
rIgIdIIIcatIon, encountered In a journaI artIcIe recentIy. The parts oI thIs word are readIIy apparent:
rIgId, IIy (whIch Iorms verbs Irom adjectIves and nouns), and atIon (whIch Iorms abstract nouns Irom
verbs), wIth IIIc beIng a contextuaI varIant oI IIy that appears reguIarIy beIore atIon. AIso transparent
Is Its meanIng, whIch can be paraphrased roughIy as 'the act or Iact or state oI makIng or becomIng
rIgId'. The paraphrase Is ambIguous, but whIch oI these senses Is meant In a partIcuIar Instance wIII
generaIIy be cIear Irom the context. SInce the actuaI sense oI the word does not dIverge Irom Its
predIcted sense, based on Its parts and Its morphoIogIcaI structure, there Is no need Ior thIs word to
be IIsted In the speaker]hearer's IexIcon, Ior the morphoIogIcaI component oI the speaker]hearer's
grammar Is abIe to process It entIreIy. The word wIII thereIore be spoken, heard, and most IIkeIy
dIscarded by aII partIes, perhaps to be created and dIscarded agaIn, but not stored, unIess It Is used
In some specIaI sense that Is not predIctabIe Irom the morphoIogy.
3

So Iar, the morphoIogy and the IexIcon do not Interact. The IIrst creates reguIar words, and the
second stores IrreguIar words. To see how they do Interact, we must Iook at a case where both the
IexIcon and the morphoIogy are In prIncIpIe capabIe oI beIng Invoked. We wIII begIn wIth a sImpIe
case, that oI the pIuraI oI a noun In EngIIsh. Some pIuraIs come Irom the IexIcon, and some Irom the
morphoIogy. The pIuraI wIII come Irom the IexIcon In case It Is IrreguIar and stored there on account
oI Its IrreguIarIty, IIke women or peopIe, and It wIII come Irom the morphoIogy In case It Is reguIar,
IIke dogs. 8ut now a questIon arIses. I a word has an IrreguIar pIuraI stored In the IexIcon, why does
It not aIso have a reguIar pIuraI, whIch comes Irom the morphoIogy n the case at hand, how does a
speaker know not to say womans Instead oI or as weII as women Or why doesn't the speaker
sometImes say one and sometImes the other SomethIng must be preventIng the morphoIogy Irom
producIng a reguIar pIuraI just In case an IrreguIar pIuraI Ior the same word exIsts In the IexIcon. The
same Is true oI IrreguIar past tenses oI verbs. A person who knows that the past tense oI go Is went (a
Iact that must be stored In the IexIcon) wIII not say goed, aIthough a young chIId or someone In the
earIy stages oI IearnIng EngIIsh as a second Ianguage mIght say goed, because the chIId or Iearner
hasn't yet Iearned the Iorm went. The IexIcon and the morphoIogy seem to Interact In assurIng that
onIy one Iorm wIII be used, but how Does the speaker]hearer somehow check the IexIcon to see II a
word Is there, and onIy resort to the morphoIogy II there Is none
A cIue to the rIght answer to thIs questIon has been known Ior centurIes: Ianguages tend to avoId
synonyms (though not aIways, as bamboozIe, hornswoggIe, and hoodwInk reveaI). n most cases, the
speaker wIII use a word Irom hIs or her IexIcon (women, went) rather than resort to the morphoIogy to
produce a new word wIth the same meanIng. ThIs phenomenon, "the nonoccurrence oI one Iorm due
to the sImpIe exIstence oI another" (AronoII 1976: 43), Is caIIed bIockIng, and Its eIIects can be seen
not onIy In InIIectIon, but aIso In derIvatIon, where a word IIke *IurIosIty (Iormed Irom IurIous) wIII be
bIocked by Iury, whIch aIready exIsts In a speaker's IexIcon. We can teII that bIockIng Is at work In
renderIng *IurIosIty unacceptabIe, because other words oI the same pattern are perIectIy acceptabIe,
when there Is no aIready exIstIng word to bIock them. Thus, curIosIty, whIch Is structuraIIy anaIogous
to *IurIosIty, Is perIectIy acceptabIe, because there Is no word *cury to bIock It. The eIIects oI
bIockIng are aIso IeIt In syntax, where an exIstIng word wIII sometImes bIock an entIre synonymous
phrase, as HoIIman (1982) IIrst noted. We do not, Ior exampIe, IInd thIs nIght used In standard
EngIIsh In a sense paraIIeI to thIs mornIng or thIs evenIng, because oI the exIstence oI synonymous
tonIght. As the exampIe shows, exact synonymy Is crucIaI, Ior the expressIon thIs nIght can In Iact be
used, so Iong as It Is not synonymous wIth tonIght (e.g. In the questIon "Why Is thIs nIght dIIIerent
Irom aII other nIghts")- There wIII aIso be no bIockIng wIthout synonymy, so we may IInd a paIr IIke
brethren and brothers, precIseIy because the Iormer reIers not to actuaI brothers but rather to IeIIow
members oI an organIzed group oI some sort. The most cogent account oI why bIockIng occurs Is
Horn's (1984, 1993), based on generaI prIncIpIes oI economy oI expressIon, to whIch we wIII return In
sectIon 2.3.
8ecause bIockIng Is a psychoIogIcaI phenomenon, It Is subject to the vagarIes oI the mInd: II a person
has temporarIIy Iorgotten the word Iame, then that person may In Iact use the word *Iamousness,
whIch Iame wouId otherwIse bIock. ThIs seemIng IaIIure oI bIockIng Is especIaIIy common In chIIdren,
who coIn new words quIte IreeIy, because theIr vocabuIary Is not as entrenched as that oI aduIts. An
artIcuIate chIId mIght use words IIke Iamousness and IIquIdIze In conversatIon wIthout hesItatIng.
8IockIng Is aIso subject to another psychoIogIcaI Iactor: IamIIIarIty or Its more easIIy measurabIe
counterpart, Irequency. n generaI, the more IrequentIy used an IrreguIar Iorm Is, both In absoIute
terms and compared to Its base, the more IIkeIy It Is to bIock the correspondIng reguIar Iorm (Anshen
and AronoII 1988, RaIner 1988). ThIs eIIect oI Irequency can be detected not onIy experImentaIIy
(PInker and PrInce 1991), but aIso In chIIdren's "overreguIarIzatIons," as 8ybee and SIobIn (1982) have
shown Ior IrreguIar verbs In EngIIsh. The eIIect oI Irequency can aIso be seen In morphoIogIcaI
reguIarIzatIon over tIme: In generaI, the more IrequentIy an IrreguIar Iorm Is used, the more resIstant
It wIII be to beIng repIaced by a reguIar Iorm, whIch Is to say, the more IIkeIy It Is to bIock the
correspondIng reguIar Iorm (Anshen and AronoII 1988). The most wIdeIy accepted modeIs oI bIockIng
take Irequency Into account by transIatIng It Into processIng speed. AccordIng to these modeIs
(MacWhInney 1975, Anshen and AronoII 1988, PInker and PrInce 1991), the search Ior the proper
word can be vIewed as a race between the mentaI IexIcon and the morphoIogy. 8oth operate
sImuItaneousIy, and the Iaster one wIns. I It Is true that the speed oI IexIcaI access Ior IndIvIduaI
stored IrreguIar words Is proportIonaI to the IogarIthm oI theIr Irequency, then the more Irequent an
IrreguIar word Is compared to Its base, the more IIkeIy It wIII be to bIock the morphoIogy. Note that
thIs generaI modeI does not InvoIve any dIrect InteractIon between the mentaI IexIcon and the
morphoIogy. The two components oI the Ianguage mechanIsm can thus be InsuIated Irom each other,
as Iar as thIs one phenomenon Is concerned. They Interact more cIoseIy In the actuaI operatIon oI the
morphoIogy, whIch we wIII now expIore.

TabIe 11.1 AII words oI the Iorm TabIe 11.1 AII words oI the Iorm TabIe 11.1 AII words oI the Iorm TabIe 11.1 AII words oI the Iorm XIdIIy XIdIIy XIdIIy XIdIIy and and and and XIdIIIcatIon XIdIIIcatIon XIdIIIcatIon XIdIIIcatIon In a very Iarge word IIst In a very Iarge word IIst In a very Iarge word IIst In a very Iarge word IIst
XIdIIy XIdIIy XIdIIy XIdIIy XIdIIIcatIon XIdIIIcatIon XIdIIIcatIon XIdIIIcatIon
acIdIIy acIdIIIcatIon
deacIdIIy deacIdIIIcatIon
reacIdIIy reacIdIIIcatIon
dIsacIdIIy *
rancIdIIy rancIdIIIcatIon
* IucIdIIIcatIon
rIgIdIIy *
* vaIIdIIIcatIon
soIIdIIy soIIdIIIcatIon
resoIIdIIy resoIIdIIIcatIon

1.2 MorphoIogy based on the IexIcon 1.2 MorphoIogy based on the IexIcon 1.2 MorphoIogy based on the IexIcon 1.2 MorphoIogy based on the IexIcon
MorphoIogIcaI patterns are abstract. ReturnIng to the word rIgIdIIIcatIon, we may vIew It sImpIy as an
InstantIatIon oI the abstract pattern [[[X]
A
IIy]
V
atIon]
N
. 8ut II we Iook more cIoseIy at attested words
that IIt thIs pattern, we see the eIIect oI the IexIcon. Let us narrow our gaze Irom the generaI pattern
just mentIoned to a sIIghtIy more partIcuIar one, In whIch the adjectIve Is oI the Iorm XId (e.g. acId,
IIvId, candId). I the pattern were compIeteIy Independent oI the IexIcon, then we mIght expect to IInd
that any adjectIve oI the Iorm XId couId serve as the base oI an attested word oI the Iorm XIdIIIcatIon.
We have access to a very Iarge EngIIsh word IIst, compIIed Irom eIghteen generaI and technIcaI
dIctIonarIes, contaInIng about 400,000 entrIes. Among these are approxImateIy 1,000 words oI the
Iorm XId. 8ut there are onIy 14 words oI the Iorm XIdIIy, whIch IndIcates that thIs partIcuIar ruIe Is not
very productIve. Thus, we do not IInd the IoIIowIng words In the IIst, though aII oI them have easIIy
constructIbIe senses: *rabIdIIy, *IIvIdIIy, *acrIdIIy, *stoIIdIIy. There are sImIarIy onIy 14 words oI the
Iorm XIdIIIcatIon. RemarkabIy, as tabIe 11.1 shows, In aII but two cases In each coIumn, the words oI
the two Iorms share a base. We cannot, oI course, extrapoIate dIrectIy Irom a Iarge dIctIonary to the
mentaI IexIcon, but dIctIonary data oI thIs type, whIch can be repeated Ior many patterns In whIch one
suIIIx Is added to another, suggest that the actuaI productIon oI morphoIogIcaIIy compIex words Is
done IargeIy by appIyIng morphoIogIcaI ruIes (addIng aIIIxes) to actuaIIy occurrIng base words that
are stored In a speaker's mentaI IexIcon.
Another IndIcatIon that morphoIogIcaI ruIes operate on words In the IexIcon Is the InherItance oI
IrreguIarIty. The most common type oI InherIted IrreguIarIty Is semantIc. CompIex words oIten have
conventIonaI senses that dIIIer sIIghtIy Irom theIr predIcted sense (see sectIon 2.2). ConsIder the word
ImmeasurabIe. ]udgIng by Its parts, It shouId mean 'that cannot be measured'. n actuaI use, It aImost
aIways means Very Iarge'. The adverb Iormed Irom It, ImmeasurabIy, thereIore means 'greatIy', as In
expressIons IIke " have beneIIted ImmeasurabIy Irom your assIstance." Another exampIe Is naturaIIze,
whIch shouId mean 'to make naturaI', but whIch has a number oI specIaIIzed senses, IncIudIng 'to
conIer the rIghts oI cItIzenshIp upon (an aIIen)' and 'to adapt (a pIant or anImaI) to a new
envIronment'. The noun naturaIIzatIon that Is derIved Irom thIs verb has the nomInaI derIvatIves oI
these two as senses, because It Is Iormed on the actuaI verb In the IexIcon wIth aII oI Its specIaIIzed
senses. Furthermore, because the most IamIIIar sense oI the verb Ior most peopIe Is 'to conIer the
rIghts oI cItIzenshIp upon (an aIIen)', the most saIIent sense oI the noun Ior most peopIe Is based on
thIs sense oI the verb. So, Ior exampIe, a search oI our unIversIty IIbrary's computerIzed cataIogue
under the key word naturaIIzatIon reveaIs a Iarge number oI books havIng to do wIth ImmIgratIon to
varIous countrIes and no other books (though we mIght expect a dIIIerent outcome at a schooI oI
hortIcuIture).
The InherItance oI the phonoIogIcaI IrreguIarItIes oI words In the IexIcon Is a IIttIe harder to detect,
IargeIy because we tend to be Iess aware oI them, but one exampIe that sprIngs to mInd Is the word
comIortabIe, whIch Ior many peopIe Is pronounced [kmIt(r)bI]. The adverb Is sImIIarIy pronounced
[kamIt(r)bII], InherItIng the phonoIogIcaI IrreguIarIty oI the adjectIve base. 8y contrast, aIthough
probabIy Is oIten pronounced [prabII], probabIe Is not pronounced *[prabI], showIng that a derIved
word (probabIy) may devIate phonoIogIcaIIy Irom Its base (probabIe) and acquIre Its own IexIcaI entry.
n concIusIon, we have seen that morphoIogy Is dIstInct Irom the IexIcon (at Ieast II by the word
IexIcon we mean a speaker]hearer's mentaI IexIcon oI unpredIctabIe Iorms), and that the morphoIogy
and the IexIcon are rIvaI sources oI words. The morphoIogy depends on the IexIcon, however,
humIdIIy humIdIIIcatIon
dehumIdIIydehumIdIIIcatIon
nIdIIy nIdIIIcatIon
renIdIIy renIdIIIcatIon
IapIdIIy IapIdIIIcatIon
IIuIdIIy IIuIdIIIcatIon
Inasmuch as the bases oI morphoIogIcaIIy compIex words are normaIIy IexIcaI entrIes.
2 MorphoIogIcaI productIvIty 2 MorphoIogIcaI productIvIty 2 MorphoIogIcaI productIvIty 2 MorphoIogIcaI productIvIty
2.1 QuantItatIve and quaIItatIve productIvIty 2.1 QuantItatIve and quaIItatIve productIvIty 2.1 QuantItatIve and quaIItatIve productIvIty 2.1 QuantItatIve and quaIItatIve productIvIty
MorphoIogIcaI productIvIty may be deIIned InIormaIIy as the extent to whIch a partIcuIar aIIIx Is IIkeIy
to be used In the productIon oI new words In the Ianguage. On thIs vIew, productIvIty Is a probabIIIstIc
contInuum that predIcts the use oI potentIaI words. At one end oI the contInuum are the dead or
compIeteIy unproductIve aIIIxes, whIch are not IIkeIy to be used at aII In coInIng new words. One
exampIe oI thIs Irom EngIIsh Is the nomInaI suIIIx -th (as In truth or growth), whIch has not been
used successIuIIy to Iorm a new word Ior 400 years, despIte vaIIant attempts at terms IIke cooIth
(whIch Is attested sporadIcaIIy, but whIch just never seems to be abIe to survIve Iong). At the other
end In EngIIsh are the productIve InIIectIonaI suIIIxes -ed, -Ing, and -s, whIch are added whenever
syntactIc condItIons are approprIate and there Is no IrreguIar Iorm aIready In exIstence to bIock them,
and hIghIy productIve derIvatIonaI suIIIxes IIke -ness and -atIon. n the mIddIe, we IInd the Iess
productIve derIvatIonaI suIIIxes IIke -Ity. Some IInguIsts treat morphoIogIcaI productIvIty as an
absoIute notIon - a pattern Is eIther productIve or unproductIve - but there Is a good deaI oI evIdence
Ior the exIstence and utIIIty oI IntermedIate cases, whIch we wIII revIew beIow, so we wIII assume In
thIs chapter that aIIIxes may dIIIer contInuousIy In productIvIty, rather than IaIIIng onIy Into the poIar
categorIes oI compIeteIy productIve and compIeteIy unproductIve, to whIch some IInguIsts have
restrIcted the dIscussIon.

TabIe 11.2 MorphoIogIcaIIy restrIcted de TabIe 11.2 MorphoIogIcaIIy restrIcted de TabIe 11.2 MorphoIogIcaIIy restrIcted de TabIe 11.2 MorphoIogIcaIIy restrIcted de- -- -adjectIvaI nomInaI suIIIxes adjectIvaI nomInaI suIIIxes adjectIvaI nomInaI suIIIxes adjectIvaI nomInaI suIIIxes

AsIde Irom quantItatIve consIderatIons, there are quaIItatIve morphoIogIcaI Iactors that are reIevant to
productIvIty. ThIs can best be seen by examInIng rIvaI aIIIxes, aIIIxes that are very sImIIar In theIr
semantIc and syntactIc condItIons. ConsIder the severaI suIIIxes that Iorm nouns Irom adjectIves In
EngIIsh. Most oI them are productIve onIy wIthIn a morphoIogIcaIIy restrIcted domaIn, as shown In
tabIe 11.2, whIch cIassIIIes the suIIIxes accordIng to the precedIng suIIIxes that they generaIIy occur
wIth. The suIIIx -ness Is dIIIerent Irom those exempIIIIed In tabIe 11.2. t occurs wIth a wIde varIety
oI base adjectIves patterns, IncIudIng monomorphemIc words and those oI the morphoIogIcaI types
IncIuded In tabIe 11.2: dryness, redness, wetness, IetIdness, proIIIIcness, venerabIeness, trIvIaIness,
obIIvIousness, recentness, pIeasantness. 8eIng unrestrIcted, -ness operates as the deIauIt aIIIx Ior
condItIoned suIIIx condItIoned suIIIx condItIoned suIIIx condItIoned suIIIx exampIe exampIe exampIe exampIe condItIonIng suIIIx condItIonIng suIIIx condItIonIng suIIIx condItIonIng suIIIx base word base word base word base word
-(c)e toIerance -ant toIerant
putrescence-ent putrescent
-(c)y buoyancy -ant buoyant
Iatency -ent Iatent
-Ity separabIIIty -abIe]-IbIe separabIe
IegIbIIIty IegIbIe
IegaIIty -aI]-ar IegaI
popuIarIty popuIar
toxIcIty -Ic toxIc
stupIdIty -Id stupId
agIIIty -IIe agIIe
IerocIty -ous IerocIous
IormIng de-adjectIvaI nouns, the aIIIx that Is normaIIy used when no addItIonaI morphoIogIcaI
condItIons on the base adjectIve hoId. LInguIsts have Iound that one oI the members oI any sImIIar
set oI rIvaI aIIIxes or operatIons wIII usuaIIy be the deIauIt, quaIItatIveIy unrestrIcted. ThIs quaIItatIve
dIIIerence Is usuaIIy mIrrored quantItatIveIy as weII: the quaIItatIveIy Ieast restrIcted operatIon among
a set oI rIvaIs wIII most oIten aIso be the quantItatIveIy most productIve, though cases oI the deIauIt
not beIng the most productIve have been Iound. CIahsen et aI. (1992), Ior exampIe, have shown that
the deIauIt pIuraI suIIIx In Cerman Is not the most productIve or most common. t Is aIso not aIways
true that the Ieast restrIcted member oI a set oI rIvaIs wIII be totaIIy unrestrIcted In Its dIstrIbutIon.
For exampIe, we have shown In earIIer work (Anshen and AronoII 1981, 1988) that -ness does not
attach at aII to words oI the Iorm XIbIe, despIte beIng the deIauIt Ior de-adjectIvaI nouns In EngIIsh.
How can quantItatIve productIvIty be measured 8aayen (1992) has deveIoped a number oI measures
that take advantage oI modern computatIonaI anaIysIs oI Iarge corpora In EngIIsh and other
Ianguages. FIrst oI aII, productIvIty Is reIated to growth (the rate at whIch new words In generaI are
beIng added to a Ianguage). The growth rate oI the vocabuIary oI any Ianguage Is estImated Irom a
Iarge corpus as the ratIo oI those words occurrIng onIy once In the corpus (hapax Iegomena,
henceIorth hapaxes) to the totaI number oI word tokens In the corpus. For rIvaI operatIons, a sImIIar
measure (the ratIo oI hapaxes Iormed by that operatIon to the totaI number oI tokens oI the same
morphoIogIcaI type In the corpus) can be used to compute reIatIve growth rates. The EngIIsh suIIIxes
-Ity and -ness, Ior exampIe, show growth rates oI 0.0007 and 0.0044 In 8aayen's caIcuIatIon (based
on a corpus oI 18,000,000 words), meanIng that -ness Is sIx tImes as productIve as -Ity, regardIess
oI any dIIIerence In the quaIItatIve morphoIogIcaI restrIctIons on the two. Another statIstIcaI measure
Is what 8aayen caIIs gIobaI productIvIty, whIch depends not onIy on the IIkeIIhood oI encounterIng
new words oI a gIven morphoIogIcaI type, but aIso on the number oI words oI that type that a speaker
aIready knows. 8y thIs measure, -ness Is not quIte three tImes more productIve than -Ity.
The ratIo oI hapaxes to tokens In a corpus Is cIearIy assocIated wIth a Iesser average Irequency oI
types; the Iower average Irequency oI -ness IormatIons compared to -Ity IormatIons has Iong been
noted (AronoII 1982). 8aayen's measures assume that the Iarger ratIo oI hapaxes Is the cause oI the
Iower average Irequency, rather than vIce versa. We can thread our way out oI thIs trap II we go back
to the orIgInaI concept oI productIvIty mentIoned above, the extent to whIch a gIven aIIIx Is used In
the productIon oI new words In the Ianguage. The data presented In tabIe 11.3 show that there are
cIose to twIce as many -ness words as -Ity words In the Ianguage, but It aIso shows that the
productIvIty oI -Ity as opposed to -ness has shown a steady Increase over tIme (wIth one exceptIon In
the IIIteenth century) untII the OED shows more -Ity words than -ness words coIned In the twentIeth
century. CIven the more restrIctIve envIronments Ior -Ity versus -ness, the tabIe shows that -Ity Is In
Iact synchronIcaIIy more productIve In EngIIsh than -ness, at Ieast judgIng Irom the dIctIonary data.
The dIctIonary method and the corpus-based method do not agree. A word oI cautIon Is In order,
though: dIctIonarIes are not aIways dependabIe IndIcators oI actuaI usage, sInce the entrIes In a
dIctIonary are seIectIve rather than IncIusIve, and sInce hapaxes are Iess IIkeIy to be seen as merItIng
dIctIonary entrIes. Counts based on actuaI Iarge corpora oI the sort that 8aayen empIoys are generaIIy
more reIIabIe, sInce they measure actuaI use, rather than beIng IIItered edItorIaIIy. Thus, It may be
that the -Ity words, though In Iact Iess productIveIy Iormed, are more IIkeIy to be IIsted In the
dIctIonary, because they are more memorabIe than the -ness words, a poInt that we wIII return to In
sectIon 2.3.

TabIe 11.3 The growth In productIvIty oI TabIe 11.3 The growth In productIvIty oI TabIe 11.3 The growth In productIvIty oI TabIe 11.3 The growth In productIvIty oI - -- -Ity Ity Ity Ity over the centurIes, based on entrIes In the over the centurIes, based on entrIes In the over the centurIes, based on entrIes In the over the centurIes, based on entrIes In the OED OED OED OED
century century century century - -- -ness ness ness ness - -- -Ity Ity Ity Ity percent percent percent percent - -- -ness ness ness ness
8th 4 0 100.00
9th 38 0 100.00
10th 67 0 100.00
11th 6 0 100.00
12th 21 0 100.00

2.2 Frequency and productIvIty 2.2 Frequency and productIvIty 2.2 Frequency and productIvIty 2.2 Frequency and productIvIty
Word Irequency, whIch we dIscussed above In reIatIon to bIockIng, Is aIso reIated to productIvIty: the
Iess productIve a morphoIogIcaI pattern Is, the more Irequent on average Its IndIvIduaI members wIII
be. 8ut Irequency Is aIso Important In the seIectIon oI bases: a Iess productIve aIIIx Is generaIIy Iound
attached to hIgher-Irequency base words than Is a more productIve aIIIx (AronoII 1982). ThIs makes
sense In terms oI what we know about the connectIon between Irequency and IexIcaI recognItIon:
words wIth hIgh-Irequency bases are more readIIy recognIzed than words wIth sImIIar Irequency but
Iow-Irequency bases (Laudanna and 8uranI 1985). I Iess productIve aIIIxes are at a dIsadvantage to
begIn wIth, then they seem to beneIIt Irom the boost provIded by a more Irequent base, aIthough the
exact psychoIInguIstIc mechanIsm behInd thIs pattern Is not yet cIear.
2.3 PragmatIcs and productIvIty 2.3 PragmatIcs and productIvIty 2.3 PragmatIcs and productIvIty 2.3 PragmatIcs and productIvIty
Some schoIars have InsIsted that the study oI morphoIogIcaI productIvIty shouId conIIne ItseII to the
study oI words that are produced unIntentIonaIIy (SchuItInk 1961). ThIs ruIes out entIreIy the study oI
unproductIve morphoIogy, whIch resembIes more margInaI Iorms oI word creatIon IIke the IormatIon
oI bIends (e.g. smog as a bIend oI smoke and Iog) or acronyms (e.g. Iaser Iormed Irom the InItIaI
Ietters oI the phrase LIght AmpIIIIcatIon by StImuIated EmIssIon oI Radar), In beIng more IIkeIy to be
IntentIonaI or notIced. However, Iess than IuIIy productIve morphoIogIcaI patterns are pervasIve In
Ianguage, and they seem to serve a IunctIon that arIses Irom theIr very unproductIvIty. When we
compare the set oI words Iormed by means oI a Iess productIve aIIIx to the set Iormed by a rIvaI aIIIx
In the same morphoIogIcaI envIronment, we generaIIy IInd that the meanIngs oI the Iess productIveIy
Iormed set are Iess predIctabIe, makIng the entIre set Iess coherent semantIcaIIy. ThIs dIIIerence In
coherence carrIes over to newIy coIned words: the meanIng oI a new word Iormed by means oI a Iess
productIve aIIIx wIII be Iess predIctabIe semantIcaIIy. For that reason, Iess productIve aIIIxes may
easIIy be used to coIn specIaI or narrowIy technIcaI terms (AronoII 1982). For exampIe, the word
specIaIIsm has come Into use quIte recentIy In 8rItIsh EngIIsh In the very restrIcted sense oI 'what a
(usuaIIy medIcaI) specIaIIst practIses'. SImIIarIy, we IInguIsts (and partIcIpants In other technIcaI IIeIds
IIke economIcs) use the technIcaI term productIvIty Instead oI the more productIveIy Iormed
productIveness, because the very unproductIveness oI the aIIIx aIIows It to be used In more
specIaIIzed senses. Horn (1984, 1993) expIaIns thIs use oI Iess productIve aIIIxes In terms oI the
InteractIon oI two CrIcean pragmatIc prIncIpIes. The prIncIpIe oI reIatIon (say no more than you must)
Ieads the speaker not to use the Iess productIve Iorm In most Instances, because the more productIve
Iorm Is more readIIy avaIIabIe; but the prIncIpIe oI quantIty (say as much as you can) wIII Interact wIth
that oI reIatIon, and Iead the speaker to use the Iess productIve aIIIx In order to make a specIaI poInt
or to caII attentIon to some aspect oI the word. What the speaker Is usuaIIy caIIIng attentIon to Is a
specIaI sense, whIch the Iess productIve aIIIx Is more IIkeIy to aIIow. Thus we IInd that morphoIogy
and pragmatIcs act together to enrIch Ianguage's expressIve potentIaI. n Horn's own words: "there Is
aIways (gIven the DIvIsIon oI Labor) a suIIIcIent reason, but It Is not aIways the same reason." Horn's
account aIso heIps us to understand why the productIvIty oI InIIectIonaI aIIIxes Is generaIIy more
poIarIzed: they are IIkeIy to be eIther compIeteIy productIve or compIeteIy unproductIve, and there are
13th 63 14 81.82
14th 225 72 75.76
15th 145 91 61.44
16th 610 193 75.97
17th 913 447 67.13
18th 308 196 61.11
19th 506 480 51.32
20th 152 166 47.80
totaI 3058 165964.83
8IbIIographIc DetaIIs 8IbIIographIc DetaIIs 8IbIIographIc DetaIIs 8IbIIographIc DetaIIs
The Handbook oI MorphoIogy The Handbook oI MorphoIogy The Handbook oI MorphoIogy The Handbook oI MorphoIogy
EdIted by: EdIted by: EdIted by: EdIted by: Andrew Spencer And ArnoId M. ZwIcky
eS8N: eS8N: eS8N: eS8N: 9780631226949
PrInt pubIIcatIon date: PrInt pubIIcatIon date: PrInt pubIIcatIon date: PrInt pubIIcatIon date: 2001

very Iew In-between cases resembIIng -Ity. n the case oI InIIectIon, whose roIe Is the reaIIzatIon oI
morphosyntactIc InIormatIon, whIch Is aIways composItIonaI, there Is nothIng Ior the speaker to caII
attentIon to, and hence Iess productIve morphoIogy has no roIe. OnIy productIve morphoIogy or
IexIcaIIzed Iorms wIII surIace.
1 For reasons oI sImpIIcIty, we wIII couch our dIscussIon oI morphoIogy In terms oI aIIIxes throughout thIs
paper. n Iact, none oI the Issues that we dIscuss here bears heavIIy on the Issue oI whether the morphoIogy
Is organIzed In terms oI aIIIxes or In terms oI operatIons.
2 We wIII use exampIes Irom EngIIsh Ior the most part, Ior ease oI exposItIon, and aIso because most oI the
work done on the topIcs covered here has deaIt wIth EngIIsh data.
3 There Is evIdence, though, that some reguIar words can gaIn entry Into the IexIcon sImpIy on grounds oI
IamIIIarIty or Irequency (Stemberger and MacWhInney 1986b, 1988).
CIte thIs artIcIe CIte thIs artIcIe CIte thIs artIcIe CIte thIs artIcIe
ARONOFF, MARK and FRANK ANSHEN. "MorphoIogy and the LexIcon: LexIcaIIzatIon and ProductIvIty." The
Handbook oI MorphoIogy. Spencer, Andrew and ArnoId M. ZwIcky (eds). 8IackweII PubIIshIng, 2001.
8IackweII ReIerence OnIIne. 28 December 2007 <http:]]www.bIackweIIreIerence.com]subscrIber]tocnode
Id=g9780631226949chunkg978063122694914>
18. DIachronIc MorphoIogy 18. DIachronIc MorphoIogy 18. DIachronIc MorphoIogy 18. DIachronIc MorphoIogy
8RAN D. ]OSEPH 8RAN D. ]OSEPH 8RAN D. ]OSEPH 8RAN D. ]OSEPH

The varIous chapters oI thIs Handbook have paInted a IaIrIy compIete pIcture oI what morphoIogy Is,
what constructs are needed In the morphoIogIcaI component oI a grammar, and how these constructs
Interact wIth one another and wIth other parts oI the grammar. For the most part, the perspectIve
taken on these questIons has been pureIy synchronIe; yet, as wIth aII aspects oI Ianguage (and Indeed
oI human InstItutIons In generaI), a dIachronIc perspectIve Is possIbIe as weII, IocusIng on what
happens to morphoIogy through tIme. Thus In thIs chapter, severaI questIons are addressed whIch are
dIachronIc In theIr Iocus:
What can change In the morphoIogIcaI component
What aspects oI the morphoIogy are stabIe
Where does morphoIogy come Irom
What trIggers change In the morphoIogy
s a generaI theory oI morphoIogIcaI change possIbIe
Moreover, through the answers gIven to these questIons, especIaIIy the IIrst two, severaI exampIes oI
varIous types oI morphoIogIcaI change are presented.
1 What can change What Is stabIe 1 What can change What Is stabIe 1 What can change What Is stabIe 1 What can change What Is stabIe
The easy answer here Is that just about everythIng dIscussed In the prevIous chapters as constItutIng
morphoIogy Is subject to change, especIaIIy so once one reaIIzes that reguIar sound change can aIter
the shape oI morphs wIthout concern Ior the eIIect oI such a change In pronuncIatIon on the
morphoIogIcaI system.
1
Thus, Ior exampIe, once-dIstInct case endIngs can IaII together by reguIar
sound change (as a type oI "syncretIsm"), as happened wIth the nomInatIve pIuraI, accusatIve pIuraI,
and genItIve sInguIar oI (most) consonant-stem nouns In SanskrIt.
2
StIII, morphoIogIcaI change goes
beyond change Induced mereIy by sound change, aIIectIng not just the actuaI reaIIzatIons oI
morphemes but aIso the categorIes Ior whIch these Iorms are exponents and the processes and
operatIons by whIch these Iorms are reaIIzed. Thus It Is possIbIe to IInd change In the Iorm taken by
the varIous types oI InIIectIonaI morphoIogy, such as markIngs Ior person, number, gender,
agreement, case, and the IIke, as weII as the addItIon or Ioss or other aIteratIon oI such categorIes and
the Iorms that express them; In the derIvatIonaI processes by whIch stems are created and modIIIed,
and In the degree oI productIvIty shown by these processes; In the morphoIogIcaI status (compound
member, cIItIc, aIIIx, etc.) oI partIcuIar eIements; In the overt or covert reIatIonshIps among
morphoIogIcaI eIements, and, more generaIIy, In the number and nature oI the entrIes Ior morphemes
and words In the IexIcon, etc. Some exampIes are provIded beIow.
3

For Instance, the category oI person In the verbaI system oI Creek has seen severaI changes In the
TheoretIcaI LInguIstIcs MorphoIogy
10.1111]b.9780631226949.2001.00021.x
Subject Subject Subject Subject
DO: DO: DO: DO:
SayIa 1 / 15 18. Diachronic Morphology : The Handbook oI Morphology : Blackwell ReIerence O...
Iorm assumed by specIIIc person (and number) endIngs. AncIent Creek aIIomorphy between -saI and
-aI Ior the 2SC.MEDOPASSVE.PRESENT endIng (generaIIy
4
dIstrIbuted as -saI aIter consonants, e.g.
perIect IndIcatIve tetrIp-saI 'you have (been) rubbed ((Ior) yourseII)', Irom trIb- 'rub', and -aI aIter
voweIs, e.g. present IndIcatIve tImI 'you honor (Ior) yourseII, contracted Irom ] tImae-aI], or II
'you are unIoosIng Ior yourseII, contracted Irom ] Ie-aI]) has been resoIved (and uItImateIy,
thereIore, reduced) through the contInuatIon oI a process begun In AncIent Creek (note voweI-stem
mIddIe Iorms IIke deIknu-saI 'you are showIng (Ior) yourseII aIready In CIassIcaI Creek) that resuIted,
vIa the extensIon oI one aIIomorph Into the domaIn oI the other, In the generaIIzatIon oI the
postconsonantaI Iorm Into aII posItIons In Modern Creek, gIvIng, Ior exampIe, tImase 'you honor
yourseII (as II Irom earIIer *tIma-saI). SImIIarIy, In some Modern Creek dIaIects, the endIng Ior
3PL.MEDOPASSVE.MPERFECTVE.PAST has Innovated a Iorm -ondustan Irom the -ondusan Iound
eIsewhere; the InvoIvement (vIa a type oI change oIten reIerred to as contamInatIon or bIendIng
5
) oI
the 1PL]2PL endIngs -mastan ] -sastan Is most IIkeIy responsIbIe Ior the InnovatIve Iorm, Inasmuch as
the InnovatIve Iorm shows the IntroductIon oI an otherwIse unexpected -t- at exactIy the same poInt
as In the 1PL]2PL endIngs. As a IInaI exampIe, Irom verbaI endIngs but a dIIIerent Ianguage group,
there Is the case oI the West CermanIc 2SC.ACTVE endIng; the InherIted endIng Irom proto-CermanIc
was *-Iz (as In CothIc -Is), yet It underwent the accretIon oI a marker -t, gIvIng Iorms such as OId
EngIIsh -est, OId HIgh Cerman -Ist, whIch Is wIdeIy heId to be a reIIex oI an encIItIc Iorm oI the
second-person pronoun pu
6
bound onto the end oI a verbaI Iorm (thus probabIy the resuIt oI
cIItIcIzatIon, on whIch see beIow).
A change In the reaIIzatIon oI number markIng aIone can be seen In the IamIIIar case oI the nomInaI
pIuraI marker ]-s] In EngIIsh, Ior It has been spreadIng at the expense oI other pIuraI markers Ior
centurIes. For Instance, the earIIer EngIIsh Iorm shoo-n, as a pIuraI oI 'shoe', wIth the pIuraI endIng -
n stIII Iound In oxen, has gIven way to shoe-s, wIth the most Irequent, and Indeed deIauIt, pIuraI
endIng -s; In thIs case, the marker has not passed Irom the Ianguage aItogether, as oxen shows, but
the domaIn oI a partIcuIar marker has come to be more and more restrIcted, and that oI another has
expanded. The "battIeground" here In the competItIon between morphemes Is constItuted by
partIcuIar IexIcaI Items and the markIngs they seIect Ior.
7

Somewhat paraIIeI to such changes In the Iorm oI endIngs themseIves are changes In eIIects
assocIated wIth the addItIon oI such endIngs. The aIIIxatIon oI the pIuraI marker ]-s] In EngIIsh
occasIons voIcIng oI a stem-IInaI IrIcatIve wIth a reIatIveIy smaII set oI nouns, aII InherIted Irom OId
EngIIsh: Ior exampIe, IoaI ([IowI]) ] Ioaves ([Iowv-z], house ([haws]) ] houses ([hawz-z]), oath
([ow) ] oaths ([ow-z]), though the deIauIt case now Is to have no such voIcIng, as IndIcated by the
Iact that nouns that have entered the Ianguage sInce the OId EngIIsh perIod do not partIcIpate In thIs
morphophonemIc voIcIng: Ior exampIe, cIass, gaII, gas, gauss, gross, guII, mass, oaI, puII, saIe, skIII.
Many nouns that do show thIs voIcIng are now IIuctuatIng In the pIuraI between pronuncIatIons wIth
and wIthout the voIcIng, so that [ows] Ior oaths, [(h)worIs] Ior wharves, and [hawsz] Ior houses can
be heard quIte IrequentIy.
8
t Is IIkeIy that the InnovatIve pronuncIatIons wIII eventuaIIy "wIn out,"
thereby extendIng the domaIn oI the deIauIt pIuraI markIng and essentIaIIy assImIIatIng thIs cIass oI
nouns to the now-reguIar cIass.
9

The creatIon oI new markers aIso represents a change. Thus, when the earIy CermanIc nomInaI suIIIx
*-es-, whIch orIgInaIIy was nothIng more than a stemIormIng eIement - that Is, an extensIon onto a
root to Iorm certaIn neuter noun stems, as IndIcated In the standard reconstructIon NOM.SC *Iamb-Iz
'Iamb' versus NOM.PL *Iamb-Iz-
10
- was reInterpreted, aIter sound changes eIImInated the IInaI
syIIabIe oI the sInguIar and pIuraI Iorms, as a marker oI the pIuraI, a change In the markIng oI
(certaIn) pIuraI nouns In CermanIc came about.
11
The uItImate Iorm oI thIs marker, -(e)r wIth the
trIggerIng oI umIaut In the root (e.g. OHC nomInatIve sInguIar Iamb ] nomInatIve pIuraI IembIr, NHC
Wort ] Wrt-er 'word]words'), reIIects the eIIects oI other sound changes and reInterpretatIons
InvoIvIng umIaut In the root trIggered by suIIIxatIon.
12

WIth regard to case markIngs, one can note that evIdence Irom unproductIve "reIIc" Iorms embedded
In IIxed phrases poInts to an archaIc proto-ndo-European InIIectIonaI marker *-s Ior the genItIve
sInguIar oI at Ieast some root nouns, whIch was then repIaced In varIous Ianguages Ior the same
nouns as *-es or *-os, aIIIxes whIch exIsted as aIIomorphIc varIants markIng genItIve sInguIar aIready
In proto-ndo-European, In use wIth dIIIerent cIasses oI nouns. For exampIe, the HIttIte Iorm nekuz
'oI evenIng' (phonetIcaIIy [nek
w
t-s]) In the IIxed phrase nekuz me ur 'tIme oI evenIng', wIth Its *-s
endIng, can be compared wIth Creek nukt-s, LatIn noct-Is, wIth the InnovatIve endIngs *-os]-es.
13

SImIIarIy, the genItIve endIng *-os (as above, wIth a varIant *-es), whIch can be InIerred Ior n-stem
nouns such as no-ma 'name' (wIth -ma Irom *-mn)
14
In pre-Creek based on the evIdence oI
SanskrIt nmn-as and LatIn nomIn-Is 'oI a name',
15
underwent a cycIe oI changes In hIstorIcaI Creek.
t was IIrst aItered through the accretIon oI a -t-, gIvIng -tos (e.g. onma-tos); aIthough the exact
source oI thIs -t- Is dIsputed, and aIthough It Is Iound uItImateIy In other cases, It seems to have
arIsen earIIest wIth the genItIve,
16
and so Its appearance perhaps shows some InIIuence Irom an
abIatIvaI adverbIaI suIIIx *-tos Iound In Iorms such as SanskrIt ta-tas 'then, Irom there' or LatIn caeII-
tus 'Irom heaven'. Whatever Its source, It at IIrst created a new genItIve sInguIar aIIomorph -tos; but
Iater, when thIs -t- was extended throughout the paradIgm, gIvIng Iorms such as the datIve sInguIar
onma-t-I (Ior expected *nomn-I; cI. SanskrIt IocatIve nmn-I), the -t- became a vIrtuaI stem
extensIon. At that poInt, one couId anaIyze noma as havIng been "reIexIcaIIzed" wIth a dIIIerent base
Iorm ]onomat-], thereby reconstItutIng the genItIve endIng agaIn as -os Ior thIs noun cIass.
Another reIatIveIy common type oI change In the reaIIzatIon oI case endIngs InvoIves the accretIon oI
what was orIgInaIIy a postposItIon onto a case suIIIx, creatIng a vIrtuaI new case Iorm. ThIs process
seems to have been the source oI varIous "secondary IocaI" cases In (OId) LIthuanIan (Stang 1966:
175-6, 228-32), such as the IIIatIve - Ior exampIe, gaIvn 'onto the head', Iormed Irom the accusatIve
pIus the postposItIon *n (wIth varIant Iorm *na) 'In' (probabIy connected wIth SIavIc na 'on') - and the
aIIatIve - Ior exampIe, gaIvspI 'to(ward) the head', Iormed Irom the genItIve pIus the postposItIon
*pIe (an encIItIc Iorm oI prIe 'at') - where InIIuence Irom neIghborIng (or substrate) 8aIto-FInnIc
Ianguages Is oIten suspected as provIdIng at Ieast a structuraI modeI.
17
SImIIar deveIopments seem to
underIIe the creatIon oI an InnovatIve IocatIve Iorm In Oscan and UmbrIan - Ior exampIe, Oscan hrtIn
'In the garden' (so 8uck 1928: 114), where a postposItIon en Is responsIbIe Ior the Iorm oI the
endIng,
18
and may be vIewed In progress In the aIternatIon between a IuII comItatIve postposItIon IIe
'wIth' In modern TurkIsh (e.g. Ahmet IIe 'wIth Ahmet', Fatma IIe 'wIth Fatma') and a bound suIIIx-IIke
eIement -(y)Ie (wIth harmonIc varIant -(y)Ia), e.g. AhmetIe, FatmayIa). t shouId be noted, however,
that though common, the deveIopment whIch these combInatIons apparentIy show, Irom noun-pIus-
Iree-postposItIon to noun-pIus-case-suIIIx, Is not unIdIrectIonaI; NevIs (1986), Ior Instance, has
demonstrated that In most dIaIects oI Saame (aIso known as LappIsh) an InherIted sequence oI aIIIxes
*-pta-k-ek]n markIng abessIve has become a cIItIc word (taga, wIth varIant haga), and more
specIIIcaIIy a stressIess postposItIon, whIIe In the EnontekI dIaIect, It has progressed Iurther to
become a noncIItIc adverb taga.
19

As the TurkIsh exampIe suggests, In LIthuanIan and Oscan, there most IIkeIy was a perIod oI
synchronIe varIatIon between aIternates beIore the uItImate generaIIzatIon oI a new case Iorm.
20

There can aIso be varIatIon oI a crossIInguIstIc sort here, In the sense that what Is ostensIbIy the same
deveIopment, wIth a postposItIon becomIng a bound eIement on a nomInaI, mIght not Iead to a new
case Iorm, II the overaII "cut" oI the Ianguage does not permIt the anaIysIs oI the new Iorm as a case-
marked nomInaI. For Instance, the specIaI IIrst- and second-person sInguIar pronomInaI Iorms In
SpanIsh, respectIveIy mIgo and tIgo, that occur wIth the preposItIon con 'wIth' and whIch derIve Irom
LatIn combInatIons oI a pronoun wIth an encIItIc postposItIon - Ior exampIe, m-cum 'me-wIth' -
couId be anaIyzed as obIIque case-marked pronouns. However, they are probabIy not to be anaIyzed
In that way, sInce there Is no other evIdence Ior such case markIng In the Ianguage, eIther wIth
pronouns other than these or wIth nouns; one couId just as easIIy, Ior Instance, treat the eIement -go
as part oI a(n admIttedIy restrIcted) bIpartItIte dIscontInuous "cIrcumposItIon" con.-go.
21

As exampIes InvoIvIng the creatIon oI new case Iorms show, InIIectIonaI categorIes - Ior exampIe,
AIIatIve In OId LIthuanIan - can be added to a Ianguage. ndeed, a typIcaI change InvoIvIng categorIes
Is the addItIon oI a whoIe new category and the exponents oI that category, though sometImes the
addItIon Is actuaIIy more a renewaI or reInIorcement oI a prevIousIy or aIready exIstIng category, as
wIth the LocatIve In Oscan. Loss oI categorIes, though, aIso occurs. For Instance, hIstorIcaI
documentatIon reveaIs cIearIy that the duaI was present as an InIIectIonaI category In the verbaI,
nomInaI, and pronomInaI systems oI earIy Creek (cI. the AncIent Creek endIng -methon noted above),
yet there are no traces oI the duaI In any system In Modern Creek; sImIIarIy, a duaI category Is
assumed Ior the proto-CermanIc verb based on Its occurrence In CothIc, and Is attested Ior the
personaI pronouns oI earIIer stages oI the CermanIc Ianguages (e.g. OId EngIIsh Ic '' ]w 'we]PL' ]
wIt 'we]DU'), yet such pronomInaI Iorms are not Iound In any oI the modern CermanIc Ianguages, and
verbaI duaI Iorms occur nowhere eIse among the oIder, or Indeed the more recent, CermanIc
Ianguages. Thus, as an InIIectIonaI category, one Ior whIch paradIgmatIc Iorms exIst or mIght be
expected to exIst, duaI number Is no Ionger present In Creek or CermanIc. SImIIarIy, there was a Ioss
oI a synthetIc perIect tense between AncIent Creek and Iate KoIne Creek, so that AncIent Iorms such
as IeIuka ' have untIed' became obsoIete reIatIveIy earIy on In the post-CIassIcaI perIod; compare the
mergIng oI perIect and sImpIe past tense Ior some speakers oI Modern EngIIsh, Ior whom DId you eat
yet Is as acceptabIe as Have you eaten yet ActuaIIy, though, the reconstItutIon (and thus addItIon) oI
the category "perIect" occurred In the medIevaI Creek perIod through the deveIopment oI a
perIphrastIc (anaIytIc) perIect tense wIth 'have' as an auxIIIary verb out oI an earIIer 'have' Iuture ]
condItIonaI tense.
22

n the case oI the Creek perIect, the medIevaI InnovatIon Ied to what was a new category, Ior there
had been a perIod oI severaI centurIes In post-CIassIcaI tImes when there was no dIstInct perIect
tense. n some Instances, though, It Is not so much the creatIon oI a new category as the renewaI oI
the category through new morphoIogIcaI expressIon. The Iuture In Creek provIdes a good exampIe,
Ior throughout Its hIstory, Creek has had a dIstInct Iuture tense, contrastIng IormaIIy and IunctIonaIIy
wIth a present tense and a past tense, but the expressIon oI the Iuture has been quIte dIIIerent at
dIIIerent stages: the synthetIc, suIIIxaI, monoIectIc Iuture In AncIent Creek (e.g. graps ' wIII wrIte')
gave way In post-CIassIcaI tImes to a varIety oI perIphrastIc Iutures wIth InIInItIves pIus auxIIIary
verbs, IIrst wIth 'have', Iater wIth 'want' (e.g. theI grapseIn, IIt. '-want to-wrIte'), In whIch the parts
maIntaIned some Independence (e.g. they couId be separated by adverbs or Inverted), but whIch In
turn have uItImateIy yIeIded a new synthetIc, monoIectIc Iuture Iormed wIth a bound InseparabIe
preIIxed marker (In Standard Modern Creek, a, as In a rapso ' wIII wrIte').
23

There can be change as weII In the content oI a category, whIch, whIIe In a sense a semantIc shIIt,
nonetheIess can have morphoIogIcaI consequences, In that the category comes to be reaIIzed on
eIements not orIgInaIIy In Its domaIn. For Instance, the SIavIc Ianguages have deveIoped a subcategory
oI "anImacy" wIthIn the set oI nomInaI gender dIstInctIons, marked IormaIIy by the use oI genItIve
Iorms where accusatIves occur Ior InanImates; In earIy stages oI SIavIc (as represented e.g. by the
earIIest Iayer oI OId Church SIavonIc), onIy certaIn types oI maIe humans (e.g. aduIts or Ireemen, as
opposed to chIIdren or sIaves) partIcIpated In such "anImacy" markIng; whIIe Iater on, a wIder range oI
nouns came to beIong to thIs subcategory (e.g. In RussIan, nouns Ior IemaIes show the anImate
decIensIonaI characterIstIc In the pIuraI, and In Serbo-CroatIan, an anImaI noun such as Iava 'IIon'
IoIIows the anImate pattern).
24

SImIIar to change In the content oI a category Is the possIbIIIty oI change In the IunctIon ] vaIue oI a
morpheme: morphoIogy InvoIves the paIrIng oI Iorm wIth meanIng, so It Is approprIate to note here
as weII Instances In whIch there Is change In the IunctIon oI a morpheme, even though that mIght be
better treated under the rubrIc oI semantIc change. For Instance, the deveIopment oI the Cerman
pIuraI marker -er dIscussed above cIearIy InvoIves a reassIgnment oI the IunctIon oI the suIIIx *-Iz-
( -er) Irom beIng a derIvatIonaI suIIIx servIng to create a partIcuIar stem cIass oI nouns to beIng an
InIIectIonaI marker oI pIuraI number. So aIso, the poIarIzatIon oI was]were aIIomorphy In some
dIaIects oI EngIIsh to correIate wIth a posItIve ] negatIve dIstInctIon, so that were Is more IIkeIy to
occur wIth -n't than Is was (TrudgIII 1990, SchIIIIng-Estes and WoIIram 1994), shows a
reInterpretatIon oI aIIomorphy that once sIgnaIed sInguIar versus pIuraI (or IndIcatIve versus
subjunctIve).
The changes IIIustrated so Iar have been IaIrIy concrete, In that they concern the phonoIogIcaI
reaIIzatIon oI morphoIogIcaI categorIes or the categorIes themseIves (whIch need some reaIIzatIon).
There can aIso be change oI a more abstract type, and a partIcuIarIy IruItIuI area to examIne Is the
matter oI IexIcaI reIatIons. The components oI grammar concerned wIth morphoIogy, whether a
separate morphoIogIcaI component or the IexIcon, reIIect the reIatIonshIps that exIst among Iorms oI
a Ianguage, whether through IexIcaI "IInkIng" ruIes, IexIcaI redundancy ruIes, or common underIyIng
Iorms. SIgnIIIcant changes can occur In the saIIence oI certaIn reIatIons, to the poInt where Iorms that
were cIearIy reIated at an earIIer stage oI the Ianguage are just as cIearIy perceIved by speakers at a
Iater stage not to be reIated. EtymoIogIcaI dIctIonarIes
25
provIde dozens oI exampIes InvoIvIng
separate IexIcaI Items that have Iost any trace oI a connectIon except Ior those speakers who have
secondarIIy acquIred knowIedge oI the reIatIonshIp: Ior exampIe, two and twIne, orIgInaIIy a 'doubIe
thread' (both Irom the earIIer root Ior 'two'), or yeIIow and gaII (both orIgInaIIy Irom a root Ior 'shIne',
but wIth dIIIerent orIgInaI vocaIIsm and dIIIerent suIIIxaI IormatIons),
26
to name just a Iew such sets
Irom EngIIsh. ThIs sItuatIon IrequentIy arIses wIth words that are transparent compounds at one stage
but Iose theIr obvIous composItIon. For Instance, the modern EngIIsh word sherIII derIves Irom an OId
EngIIsh compound scrgerIa, IIteraIIy the 'reeve' (gerIa) oI the 'shIre' (scr), but Is not obvIousIy
connected In any way wIth Modern EngIIsh shIre or reeve; nor Is Iord pIausIbIy connected
synchronIcaIIy wIth IoaI or ward, the modern contInuatIons oI Its OId EngIIsh components (hIIord,
IIteraIIy 'bread-guardIan', Irom hII 'bread' pIus weard 'guardIan'). n these cases, both sound
changes, whIch can obscure the once obvIous reIatIonshIp, as wIth I(-ord) and IoaI, and semantIc
changes, as wIth (I-)ord and ward (the Iatter no Ionger meanIng 'guardIan'), can pIay a roIe In
separatIng IexIcaI Items once reIated synchronIcaIIy.
27
And borderIIne cases provIde some dIIIIcuItIes
Ior anaIysIs; Ior Instance, are the semantIcaIIy stIII compatIbIe words two and tweIve to be
synchronIcaIIy reIated In Modern EngIIsh, and II so, does two derIve Irom a Iorm wIth an underIyIng
cIuster ]tw-] To a certaIn degree, the answers to such questIons wIII depend on meta-theoretIcaI
concerns, such as a decIsIon on the degree oI abstractness to be aIIowed In morphophonoIogIcaI
anaIyses (on whIch, see beIow).
n the Iace oI such exampIes oI change, It Is equaIIy Important to reIIect on what does not or cannot
change In the morphoIogy. To the extent that there are weII-estabIIshed prIncIpIes and constructs
that are taken to be part oI the basIc theoretIcaI Iramework Ior morphoIogy - Ior exampIe, LexIcaI
ntegrIty, MorphoIogy-Iree Syntax, dIsjunctIve orderIng Ior competIng morphoIogIcaI ruIes, and the
IIke - presumabIy these wIII not change; they are the theoretIcaI buIIdIng bIocks oI any account oI the
morphoIogIcaI component, and thus cannot change dIachronIcaIIy (though they can oI course be
aItered by IInguIsts In theIr descrIptIons]accounts II synchronIe or dIachronIc Iacts make It cIear, Ior
Instance, that syntax Is not morphoIogy-Iree, or the IIke).
Among these theoretIcaI buIIdIng bIocks are some that have a sIgnIIIcant Impact on dIachronIc
accounts oI morphoIogy, In partIcuIar those that aIIow Ior the determInatIon oI the borderIInes
between components oI grammar. That Is, It Is wIdeIy recognIzed that there Is InteractIon at Ieast
between morphoIogy and phonoIogy (wItness the term morphophonoIogy and the possIbIIIty oI
phonoIogIcaI constraInts on morphoIogIcaI ruIes) and between morphoIogy and syntax (wItness the
term morphosyntax). Thus It becomes approprIate to ask how we can teII when some phenomenon
crosses the border Irom "pure" phonoIogy Into morphoIogy, or vIce versa, or Irom "pure" syntax Into
morphoIogy. AIthough there Is a pureIy synchronIe questIon here oI how to characterIze a gIven
phenomenon In a gIven Ianguage Ior a gIven perIod oI tIme, the matter oI crossIng component
boundarIes Is aIso a dIachronIc Issue. I a once-phonoIogIcaI phenomenon comes to be condItIoned
compIeteIy morphoIogIcaIIy, and Is consIdered to be part oI the morphoIogIcaI component and not
the phonoIogIcaI component, then there has been a change In the grammar oI the Ianguage wIth
regard to that phenomenon; the surIace reaIIzatIon oI the Iorms may not change, but the grammatIcaI
apparatus underIyIng and producIng or IIcensIng those surIace Iorms has changed. Thus, when the
voweI IrontIng Induced by a IoIIowIng hIgh voweI (so-caIIed umIaut) In earIy Cerman came In Iater
stages oI the Ianguage, when the phonetIc motIvatIon Ior the IrontIng was obscured or absent on the
surIace, to be an eIIect assocIated wIth the addItIon oI certaIn suIIIxes (e.g. the dImInutIve -chen, the
noun pIuraI -e, etc.) or wIth the expressIon oI certaIn categorIes (e.g. pIuraIs oI certaIn nouns whIch
take no overt suIIIx, such as 8ruder 'brother', wIth pIuraI 8rder), one InterpretatIon Is that the
umIautIng process Is no Ionger phonoIogIcaI In nature, but rather Is a morphoIogIcaI process Invoked
by certaIn morphoIogIcaI categorIes.
28
SImIIarIy, at a stage when the expressIon oI IocatIves In (pre-)
Oscan was accompIIshed by a noun pIus a postposItIon, syntactIc ruIes that IIcense postposItIonaI
phrases were responsIbIe Ior the surIace Iorms; when the noun Iused wIth the postposItIonaI eIement
to such an extent that a vIrtuaI new case-marker was created, the responsIbIIIty Ior the uItImate
expressIon oI the IocatIve eIIectIveIy moved out oI the reaIm oI syntax and Into the morphoIogIcaI
component.
These exampIes and the reIevance oI theoretIcaI decIsIons separatIng components oI grammar poInt
to the need to recognIze the Impact that the theory oI grammar one adopts has on dIachronIc
anaIyses. For exampIe, permIttIng a degree oI abstractness In phonoIogIcaI anaIyses can oIten aIIow
Ior a descrIptIon that Is pureIy phonoIogIcaI rather than morphoIogIcaI In nature. UmIaut In Cerman,
Ior Instance, couId stIII be consIdered to be pureIy phonoIogIcaI II each suIIIx or category now
assocIated wIth umIaut oI a stem were represented underIyIngIy wIth a hIgh Iront voweI to act as the
trIggerIng segment; deIetIng that segment beIore It couId surIace wouId have to be consIdered to be
aIIowabIe abstractIon. SImIIarIy, the paIataIIzatIons oI stem-IInaI veIars In varIous SIavIc Ianguages
that accompany the attachment oI certaIn suIIIxes (e.g. RussIan adjectIvaI -nyj, as In vostoc-nyj
'eastern' Irom the noun vostok '(the) east') were once trIggered by a suIIIx-InItIaI short hIgh Iront
voweI (the "Iront yer") that was uItImateIy Iost In most posItIons In aII the Ianguages; thus a
synchronIe, pureIy phonoIogIcaI anaIysIs couId be constructed sImpIy by posItIng an abstract Iront yer
that trIggers the paIataIIzatIon and Is then deIeted.
29

2 Where does morphoIogy come Irom 2 Where does morphoIogy come Irom 2 Where does morphoIogy come Irom 2 Where does morphoIogy come Irom
The exampIes In sectIon 1 show that the prImary source oI morphoIogy Is materIaI that Is aIready
present In the Ianguage, through the medIatIon oI processes oI resegmentatIon and reInterpretatIon
appIIed In a varIety oI ways, as weII as by other processes oI change - Ior exampIe, sound changes -
that Iead to grammatIcaIIzatIon. n addItIon, morphoIogy may enter a Ianguage through varIous Iorms
oI Ianguage contact.
Thus, exampIes oI bIendIng or contamInatIon InvoIve preexIstIng materIaI, as In the case oI Creek
1DUAL.MEDOPASSVE endIng (see n. 5), where a "crossIng" oI the 1PL.MEDOPASSVE endIng -metha wIth
the 2DUAL.MEDOPASSVE endIng -sthon yIeIded -methon. n a paraIIeI IashIon, when a sequence oI
eIements Is resegmented - that Is, gIven a dIIIerent "parsIng" by speakers Irom what It prevIousIy or
orIgInaIIy had - materIaI aIready In the Ianguage Is gIven a new IIIe. The EngIIsh -ness suIIIx, Ior
Instance, derIves Irom a resegmentatIon oI a CermanIc abstract noun suIIIx *-assu- attached to n-
stem adjectIves, wIth subsequent spread to dIIIerent stem types; thus *ebn-assu- 'equaIIty' (stem:
*ebn- 'even, equaI') was treated as II It were *eb-nassu-, and Irom there *-nassu- couId spread, as In
OId EngIIsh ehtness 'persecutIon' (Irom the verb eht-an 'to pursue') or gdness 'goodness' (Irom the
adjectIve gd). The extreme productIvIty oI thIs new suIIIx In Modern EngIIsh, capabIe oI beIng added
to vIrtuaIIy any new adjectIve (e.g. gauche-ness, uptIght-ness) shows how Iar beyond Its orIgInaI
Iocus a Iorm can go, and aIso how the productIvIty oI a morpheme can change, sInce -ness orIgInaIIy
had a more IImIted use.
Other types oI reanaIysIs sImIIarIy draw on materIaI present at one stage oI a Ianguage In one Iorm
and transIorm It at a Iater stage. n many cases oI desyntactIcIzatIon, Ior Instance, where once-
syntactIc phrases are reInterpreted as word-IeveI unIts wIth aIIIxes that derIve Irom orIgInaI Iree
words or cIItIcs, as In the Oscan IocatIve dIscussed above, the same segmentaI materIaI Is InvoIved,
but wIth a dIIIerent grammatIcaI status. SometImes, though, such reanaIyses are accompanIed (or
even trIggered) by phonoIogIcaI reductIons, so that the resuIt Is just added segmentaI materIaI wIth
no cIear morphoIogIcaI vaIue; the -t oI OId EngIIsh wIt 'we two', Ior Instance, comes Irom a
phonoIogIcaIIy reguIar reductIon oI the stem Ior 'two' In an unstressed posItIon - that Is, Irom *we-
dwo - and sImIIar cases InvoIvIng oId compounds - Ior exampIe, sherIII and Iord - were noted above.
Moreover, when sound changes obscure the condItIonIng Iactors Ior a phonoIogIcaIIy Induced eIIect,
and a new morphoIogIcaI process arIses, as wIth umIaut In Cerman, agaIn what has occurred Is the
reanaIysIs oI aIready exIstIng materIaI, In thIs case the IrontIng oI a stem voweI that accompanIes the
addItIon oI an aIIIx; the new process Is then avaIIabIe to spread Into new contexts, havIng been Ireed
Irom a connectIon to a partIcuIar phonoIogIcaI trIgger.
SometImes semantIc shIIts are InvoIved In such reanaIyses. The weII-known exampIe oI the new suIIIx
-gate In EngIIsh Is a case In poInt. ThIs suIIIx orIgInated Irom the phrase Watergate aIIaIr (or scandaI
or the IIke), reIerrIng to the events In the aItermath oI a burgIary at the Watergate apartment compIex
that brought down the NIxon admInIstratIon In the earIy 1970s, through a truncatIon oI the phrase to
Watergate (e.g. NIxon resIgned because oI Watergate) and a reanaIysIs In whIch the -gate part was
treated as a suIIIx and not the compound member It orIgInaIIy was In the pIace-name Watergate. t
then spread, gIvIng coInages such as rangate (Ior a scandaI In the 1980s InvoIvIng seIIIng arms to
ran), Coobergate (Ior a scandaI aIIeged In 1979 to have InvoIved then-PresIdent Carter's peanut
warehouse), and numerous others.
30
What Is especIaIIy InterestIng about thIs reanaIysIs Is that In the
process oI -gate becomIng a suIIIx, there was a shIIt In Its meanIng, so that In X-gate, the suIIIx -
gate (but not the Iree word gate) ItseII came to mean 'a scandaI InvoIvIng X', an abbrevIatIon, as It
were, Ior 'a scandaI InvoIvIng X remInIscent oI the Watergate scandaI'.
Other processes sImIIar to these that create pIeces oI words produce as weII new IexIcaI Items, and
thus contrIbute to the morphoIogIcaI component, to the extent that It IncIudes the IexIcon. WIthout
goIng Into great detaII, one can note actIve processes oI word IormatIon such as compoundIng,
acronymIc coInage (e.g. cpu (pronounced [sIpIyu]) Ior centraI processIng unIt, ram ([rm]) Ior
random-access memory, rom ([ram]) Ior read-onIy memory), cIIppIng (e.g. dIs Irom (show) dIsrespect,
rad Irom radIcaI, prep Irom prepare and Irom preparatory, vet Irom veteran and Irom veterInarIan),
IexIcaI bIends (e.g. brunch Irom breakIast crossed wIth Iunch), phrasaI truncatIons (such as the source
oI the word street vIa a truncatIon, wIth a semantIc shIIt, oI LatIn vIa strta 'road (that has been)
paved' to sImpIy strta), and so on. t Is worth notIng here that whereas vIrtuaIIy any pIece oI a word,
even suIIIxes, can be "eIevated" to the status oI a Iree word vIa cIIppIng, InIIectIonaI morphemes seem
to be resIstant to such an "upgradIng"; thus aIthough Ism as a Iree word meanIng 'dIstInctIve doctrIne,
system, or theory' (AHD 1992) has been extracted out oI communIsm, socIaIIsm, etc., Instances In
whIch suIIIxes IIke EngIIsh -ed or -s become words Ior 'past' or 'many' or the IIke appear not to exIst.
One IInaI Ianguage-InternaI path Ior the deveIopment oI morphoIogy InvoIves Instances In whIch the
condItIons Ior an anaIysIs motIvatIng a sequence oI sounds as a morpheme arIse onIy somewhat
accIdentaIIy. n partIcuIar, II a sItuatIon occurs In whIch speakers can recognIze a reIatIon among
words, then whatever shared materIaI there Is among these words can be eIevated to morphemIc
status. ThIs process Is especIaIIy evIdent wIth phonesthemes, materIaI that shows vague assocIatIve
meanIngs that are oIten sensory based, such as the InItIaI sequence gI- In EngIIsh Ior 'brIghtIy
vIsIbIe', as In gIeam, gIItter, gIIsten, gIow, and the IIke. Some IInguIsts are hesItant to caII these
eIements morphemes, and terms IIke quasI-morpheme, submorphemIc unIt, and others have been
used on occasIon, even though by most deIInItIons, they IuIIIII the crIterIa Ior beIng IuII morphemes.
LeavIng asIde the synchronIe Issue they pose Ior anaIysIs, It Is cIear that they can come to have some
systematIc status In a grammar, Ior they can spread and be expIoIted In new words (e.g. gIItzy, whIch,
whether based on Cerman gIItzern 'to gIItter' or a bIend InvoIvIng rItzy, nonetheIess IIts Into the
group oI other "brIght" gI- words). A good exampIe oI thIs process Is aIIorded by the accumuIatIon oI
words In EngIIsh that end In -ag (earIIer [-ag], now [-g]) and have a generaI meanIng connotIng
'sIow, tIred, or tedIous actIon', specIIIcaIIy drag 'Iag behInd', Iag 'grow weary', IIag 'droop', and Iag
'straggIe', aII attested In MIddIe EngIIsh but oI varIous sources (some ScandInavIan borrowIngs, some
InherIted Irom earIIer stages oI EngIIsh); at the poInt at whIch Iour words wIth both a sImIIar meanIng
and a sImIIar Iorm were present In the Ianguage, by roughIy the thIrteenth century, an anaIysIs was
possIbIe oI thIs -ag as a (sub-) morphemIc eIement. That It had some reaIIty as such a unIt Is shown
by the Iact that these words "attracted" a semantIcaIIy reIated word wIth a dIIIerent Iorm Into theIr
"orbIt" wIth a concomItant change In Its Iorm; sag 'sInk, droop' In an earIy Iorm (sIxteenth century)
ended In -k, yet a perceIved assocIatIon wIth drag]Iag]IIag]Iag and the avaIIabIIIty oI -ag as a marker
oI that group brought It more In IIne wIth the other members, gIvIng uItImateIy sag.
The exampIe oI -gate above aIso shows Ianguage contact as a source oI new morphoIogy In a
Ianguage, Ior It has spread as a borrowed derIvatIonaI suIIIx Into Ianguages other than EngIIsh;
Schuhmacher 1989 has noted Its presence In Cerman, Kontra 1992 gIves severaI Instances oI -gate
Irom HungarIan, and ]oseph 1992 provIdes Creek and Serbo-CroatIan exampIes. Numerous exampIes
oI borrowed derIvatIonaI morphoIogy are to be Iound In the LatInate vocabuIary In EngIIsh, but It
shouId be noted aIso that InIIectIonaI morphoIogy can be borrowed. VarIous IoreIgn pIuraIs In EngIIsh,
such as crIterIa, schemata, aIumnae, IIIustrate thIs poInt, as do the occurrence oI TurkIsh pIuraI
endIngs In some (now oIten obsoIete) words In AIbanIan oI TurkIsh orIgIn - Ior exampIe, at-IIare
'Iathers', bej-Iere 'IandIords' (Newmark et aI. 1982: 143)
31
- and the verb paradIgms In the AIeut
dIaIect spoken on the IsIand oI Mednyj, whIch show RussIan person ] number endIngs added onto
natIve stems - Ior exampIe, uuIju ' sIt' ] uuI-It '(s)he sIts' (Thomason and KauIman 1988: 233-
8). AIthough It Is wIdeIy beIIeved that InIIectIonaI morphoIogy Is partIcuIarIy resIstant to borrowIng
and to beIng aIIected by Ianguage contact, Thomason and KauIman (1988) have shown that what Is
crucIaI Is the socIaI context In whIch the contact and borrowIng occur. Thus the Intense contact and
the degree oI bIIInguaIIsm needed to eIIect contact-Induced change InvoIvIng InIIectIonaI morphoIogy
sImpIy happen not to arIse very oIten, so that any rarIty oI such change Is not a IInguIstIc questIon per
se. Moreover, the spread oI derIvatIonaI morphoIogy across Ianguages may actuaIIy take pIace
through the spread oI whoIe words, whIch are then "parsed" In the borrowIng Ianguage; the -gate
suIIIx In Creek, Ior Instance, occurred IIrst In IabeIs Ior scandaIs that IoIIowed the EngIIsh names
dIrectIy (e.g. "rangate") beIore beIng used Ior Creek-InternaI scandaIs.
3 What trIggers change In the morphoIogy 3 What trIggers change In the morphoIogy 3 What trIggers change In the morphoIogy 3 What trIggers change In the morphoIogy
HIstorIcaI IInguIsts tend to dIvIde causes oI change Into those InternaI to the IInguIstIc system ItseII
and those that are externaI - that Is, due to Ianguage contact. The dIscussIon In sectIon 2 shows that
Ianguage contact Is Indeed one potentIaI cause oI morphoIogIcaI change, and that under the rIght
socIaI condItIons Ior the contact, vIrtuaIIy any morphoIogIcaI eIement (InIIectIonaI, derIvatIonaI,
bound, Iree, whatever) can be transIerred Irom one Ianguage to another. ExamInIng contact-Induced
morphoIogIcaI change then becomes more a matter - an Important one, to be sure - oI cataIoguIng
the changes and determInIng the socIoIInguIstIc mIIIeu In whIch the contact occurs.
32
There Is Iar
more to say, however, about InternaI Iorces trIggerIng change In the morphoIogy.
From a consIderatIon oI the exampIes above, It emerges that much morphoIogIcaI change InvoIves
"anaIogy," understood In a broad sense to take In any change due to the InIIuence oI one Iorm on
another.
33
ThIs process Is most evIdent In bIendIng or contamInatIon, where there Is mutuaI
InIIuence, wIth a part oI one Iorm and a part oI another combInIng; but It extends to other types oI
morphoIogIcaI change as weII.
For Instance, the spread oI -t- descrIbed above In the stem oI Creek neuter nouns In -ma InvoIved
the InIIuence oI the genItIve sInguIar Iorms, the orIgInaI Iocus oI the -t-, over other Iorms wIthIn the
paradIgm. Such paradIgm-InternaI anaIogy, oIten reIerred to as "IeveIIIng," Is quIte a common
phenomenon. An InterestIng exampIe, to be reexamIned beIow Irom a dIIIerent perspectIve, InvoIves
the reIntroductIon oI -w- Into the nomInatIve oI the adjectIve Ior 'smaII' In LatIn: In earIy LatIn, the
adjectIve had nomInatIve sInguIar parw-os and genItIve parw-, and paradIgmatIc aIIomorphy par-os
versus parw- resuIted when a sound change eIImInated -w- beIore a round voweI; paradIgm-InternaI
anaIogIcaI pressures Ied to the restoratIon oI the -w-, gIvIng uItImateIy the CIassIcaI LatIn Iorms
parvus ] parv.
AnaIogIcaI InIIuence among Iorms Is not restrIcted to those that are paradIgmatIcaIIy reIated. Two
eIements that mark the same category, but wIth dIIIerent seIectIonaI propertIes, can exert anaIogIcaI
pressures, IeadIng to the spread oI one at the expense oI another. ExampIes oI such anaIogIes IncIude
cases across Iorm cIasses where the eIements InvoIved are dIIIerent morphemes, as wIth the spread
oI the -s pIuraI In EngIIsh at the expense oI the -(e)n pIuraI, dIscussed In sectIon 1, as weII as cases In
whIch one condItIoned aIIomorphIc varIant extends Its domaIn over another, thereby destroyIng the
once-condItIoned aIternatIon, as wIth the spread oI the Creek 2SC.MEDOPASSVE endIng -SA, aIso
dIscussed above.
SImIIarIy, In cases oI IoIk etymoIogy, speakers reshape a word based on other Iorms that provIde what
they see as a semantIcaIIy (somewhat) motIvated parsIng Ior It; Ior exampIe, toIu Ior some speakers Is
[toIud], as II a compound wIth Iood, and crayIIsh, IIrst borrowed Irom French In the Iourteenth
century as crevIse, was remade as II contaInIng the Iexeme IIsh. n such cases, whIch are quIte
common wIth borrowIngs or words that are unIamIIIar Ior reasons such as obsoIescence, there Is
InIIuence Irom one Iorm beIng brought to bear on the shape oI another. More generaIIy, many cases
oI reanaIysIs]reInterpretatIon InvoIve some anaIogIcaI pressures, especIaIIy when the reanaIysIs Is
Induced by modeIs that exIst eIsewhere In the Ianguage; Ior Instance, when MIddIe EngIIsh pease, a
sInguIar noun meanIng 'pea', was reanaIyzed as a pIuraI, aIIowIng Ior the creatIon (by a process
known as "backIormatIon") oI a sInguIar pea, the InIIuence oI other pIuraIs oI the shape [. V-z]
pIayed a roIe.
Thus there Is a cognItIve dImensIon to (certaIn types oI) morphoIogIcaI change, In the sense that It
oIten InvoIves speakers actIveIy makIng connectIons among IInguIstIc Iorms and actIveIy reshapIng
theIr mentaI representatIons oI Iorms.
34
ndeed, anaIogy as a generaI mode oI thInkIng and reasonIng
has Iong been treated wIthIn the IIeId oI psychoIogy, and studIes by Esper (e.g. Esper 1925 and the
posthumous Esper 1973) were an earIy attempt to determIne the psychoIogIcaI basIs Ior anaIogIcaI
change In Ianguage.
35
More recentIy, anaIogIcaI change has been vIewed Irom the perspectIve oI a
theory oI sIgns; AnttIIa (1972), Ior Instance, has argued that the semIotIc prIncIpIe oI "one Iorm to
one meanIng" drIves most anaIogIcaI change, In that IeveIIngs, Iorm-cIass anaIogIes, IoIk etymoIogy,
and the IIke aII create a better IIt between Iorm and meanIng, whIIe proponents oI NaturaI
MorphoIogy
36
sImIIarIy work wIth the Importance oI degrees oI IconIcIty In the Iorm-meanIng
reIatIonshIp and, Ior exampIe, evaIuate changes In the markIng oI InIIectIonaI categorIes or
derIvatIonaI reIatIonshIps In terms oI how they Iead to a better IIt wIth unIversaI IconIc prIncIpIes. Even
the process oI grammatIcaIIzatIon has been gIven a cognItIve InterpretatIon; HeIne et aI. (1991:150),
Ior Instance, have argued that "underIyIng grammatIcaIIzatIon there Is a specIIIc cognItIve prIncIpIe
caIIed the 'prIncIpIe oI the expIoItatIon oI oId means Ior noveI IunctIons' by Werner and KapIan (1963:
403)," and they note that In many cases grammatIcaIIzatIon InvoIves metaphorIcaI extensIon Irom one
cognItIve domaIn - Ior exampIe, spatIaI reIatIons - to another - Ior exampIe, temporaI reIatIons (as
wIth behInd In EngIIsh).
37

MovIng away Irom these more cognItIve, IunctIonaI, and]or mentaIIstIc vIews oI what causes
morphoIogIcaI change, one can IInd varIous IormaI approaches to anaIogy. The most notabIe
38
Is the
generatIve approach, In whIch anaIogy Is nothIng more than changes In the ruIe system that generates
a gIven paradIgm. The LatIn case mentIoned above whereby a paradIgm oI parw-os ] parw- yIeIded
par-os ] parw- by sound change and IInaIIy parvus ] parv by paradIgm IeveIIng couId be seen as the
addItIon oI a ruIe oI w beIore round voweIs (the sound change) operatIng on an underIyIng Iorm
Ior the nomInatIve wIth the -w-, and then the Ioss oI that ruIe gIvIng the underIyIng stem-IInaI -w- a
chance to surIace once agaIn. What Is IeIt unexpIaIned In such an account Is why the ruIe wouId be
Iost at aII; earIy generatIve accounts (e.g. R. KIng 1969, KIparsky 1968) sImpIy gave a hIgher vaIue to
a grammar wIth Iewer ruIes or Ieatures In the ruIes (but then where, as Andersen (1973: 766) asked,
wouId added ruIes come Irom, and why wouId they even be added In the IIrst pIace) or unnaturaI ruIe
orderIngs, whereas Iater accounts (especIaIIy KIparsky 1971) gave hIgher vaIue to grammars that
generated paradIgm-InternaI reguIarIty, a condItIon that tacItIy admIts that the tradItIonaI reIIance on
the InIIuence oI reIated surIace Iorms had some vaIIdIty aIter aII. Another type oI generatIve
reInterpretatIon oI anaIogy Is that gIven by Anderson (1988a), who, as observed In Iootnotes 7 and 9,
sees anaIogIes such as the spread oI the EngIIsh -s pIuraI or the Ioss oI morphophonemIc voIcIng In
certaIn EngIIsh pIuraIs as beIng actuaIIy changes In the IexIcaIIy IdIosyncratIc specIIIcatIons Ior the
InIIectIonaI markIngs, derIvatIonaI processes, and the IIke seIected by partIcuIar IexIcaI Items.
FInaIIy, any dIscussIon oI causes must make reIerence to the Iact that, as Is the case wIth aII types oI
Ianguage change, the spread oI morphoIogIcaI InnovatIons Is subject to socIaI Iactors governIng the
evaIuatIon oI an InnovatIon by speakers and Its adoptIon by them. ndeed, II one takes the vIew that
true Ianguage change occurs onIy when an InnovatIon has spread throughout a speech communIty,
39

then the varIous processes descrIbed here onIy provIde a startIng poInt Ior a morphoIogIcaI
InnovatIon, but do not descrIbe uItImate morphoIogIcaI change In the Ianguages In questIon. The
presence oI synchronIe varIatIon In some oI the changes dIscussed above, as wIth the Ioss oI
morphophonemIc voIcIng In EngIIsh pIuraIs, shows how the opportunIty can arIse Ior nonIInguIstIc
Iactors to pIay a roIe In promotIng or quashIng an InnovatIon.
4 s a generaI theory oI morphoIogIcaI change possIbIe 4 s a generaI theory oI morphoIogIcaI change possIbIe 4 s a generaI theory oI morphoIogIcaI change possIbIe 4 s a generaI theory oI morphoIogIcaI change possIbIe
Over the years, there have been numerous attempts to deveIop a generaI theory oI morphoIogIcaI
change, and the approaches to the causes oI morphoIogIcaI change outIIned In the prevIous sectIon
actuaIIy represent some such attempts. To a greater or Iesser extent, there have been successes In
thIs regard. For Instance, the recognItIon oI a cognItIve dImensIon to anaIogy and to
grammatIcaIIzatIon has been sIgnIIIcant, as has the correspondIng understandIng oI the roIe oI
IconIcIty. The generatIve paradIgm has been embraced by many, but a Iew Iurther comments about It
are In order.
Most Important, as noted above, an account oI anaIogIcaI change In paradIgms that Is based on
changes In the ruIes by whIch the paradIgms are generated does not extend weII to anaIogIcaI
changes that cannot InvoIve any ruIes, such as bIends or contamInatIon. As Hock (1991: 256) poInts
out, a deveIopment such as MIddIe EngIIsh IemeIIe (a Ioan word Irom French) becomIng IemaIe by
contamInatIon wIth maIe does not InvoIve any generatIve ruIes; yet It stIII took pIace, and one wouId
be hard-pressed to account Ior the change In the vocaIIsm oI thIs word wIthout some reIerence to
pressure Irom the semantIcaIIy reIated maIe. SImIIarIy, the change dIscussed by AnttIIa (1972: 89), In
whIch the nomInatIve sInguIar oI the unIqueIy InIIected word Ior 'month' In the EIean dIaIect oI
AncIent Creek became mes (wIth genItIve mn-s, versus e.g. AttIc nomInatIve meIs), based on the
unIqueIy InIIected word Ior the god Zeus (nomInatIve Zes, genItIve Zn-s), couId not InvoIve any
generatIve phonoIogIcaI ruIes, sInce both words were the onIy members oI theIr respectIve
decIensIonaI cIasses, and thus were probabIy IIsted In the IexIcon rather than ruIe-governed In terms
oI theIr InIIectIon.
40
On the other hand, the semIotIc and cognItIve vIews oI anaIogy - Ior Instance,
InvokIng a one-Iorm-to-one-meanIng prIncIpIe - can provIde a motIvatIon not onIy Ior the putatIve
cases oI anaIogy as ruIe-change, but aIso Ior those that couId not InvoIve ruIe change.
41
Moreover,
cases oI bI-dIrectIonaI IeveIIng, as presented by TIersma 1978 wIth data Irom FrIsIan, In whIch some
paradIgms InvoIvIng a partIcuIar phonoIogIcaI ruIe are IeveIed as II the ruIe had been Iost, whIIe
others InvoIvIng the same ruIe are IeveIIed as II the ruIe had been generaIIzed, make It dIIIIcuIt to gIve
any predIctIve vaIue to a ruIe-based approach to anaIogy.
42
FInaIIy, the recognItIon oI paradIgm
unIIormIty as a part oI the evaIuatIon metrIc In KIparsky 1971 Is tantamount to recognIzIng anaIogy In
Its tradItIonaI sense. As AnttIIa (1972: 129, 131) puts It: "What ruIe changes aIways descrIbe, then, Is
the beIore-aIter reIatIonshIp. They gIve a mechanIsm Ior descrIptIon, not a hIstorIcaI expIanatIon.
RuIe change Is not a prImary change mechanIsm, but an eIIect."
ThIs Is not to say, however, that tradItIonaI anaIogy Is not wIthout some probIems. As has IrequentIy
been poInted out, It oIten seems unconstraIned, and there Is an eIement oI unpredIctabIIIty about It.
When wIII anaIogy occur What dIrectIon wIII IeveIIng take WhIch Iorms wIII serve as modeIs And so
on. n part to address thIs uncertaInty about the workIngs oI anaIogy, some schoIars have attempted
to IormuIate a set oI generaI tendencIes or reguIarItIes governIng anaIogy. The two most wIdeIy
dIscussed schemes are those oI KuryIowIcz (1945-9)
43
and Maczak (1958). A IuII dIscussIon oI these
proposaIs Is beyond the scope oI the present chapter,
44
but It Is generaIIy heId that KuryIowIcz's
"Iaws" are, as CoIIInge (1985: 252) cItIng AnttIIa (1977: 76-80) puts It, more "quaIItatIve and IormaI"
In nature, whereas Maczak's tendencIes are more "quantItatIve and probabIIIstIc." t can be noted
aIso that some oI theIr specIIIc proposaIs compIement one another, some are contradIctory, some are
tautoIogous and thus oI IIttIe vaIue, but some
45
- Ior exampIe, Maczak's second tendency ("root
aIternatIon Is more oIten aboIIshed than Introduced") and KuryIowIcz's IIrst "Iaw" ("a bIpartIte marker
tends to repIace an IsoIunctIonaI morpheme consIstIng oI onIy one oI these eIements") are vaIuabIe
tooIs In anaIyzIng anaIogIcaI changes, as they reIIect tensIons present In Ianguage In generaI:
respectIveIy the need to have redundancy Ior cIarIty and the desIre to eIImInate unnecessary or
unmotIvated redundancy. Moreover, KuryIowIcz's Iourth "Iaw" has, In the estImatIon oI Hock (1991:
230), proved to be "a very reIIabIe guIde to hIstorIcaI IInguIstIc research." ThIs "Iaw," whIch states that
an InnovatIve Iorm takes on the prImary IunctIon and that the oIder Iorm It repIaces, II It remaIns at
aII, does so onIy In a secondary IunctIon, can be exempIIIIed by the oIt-cIted case
46
oI EngIIsh
brethren; thIs Iorm, orIgInaIIy a pIuraI oI the kInshIp term brother, Is now reIegated to a restrIcted
IunctIon In the meanIng "IeIIow members oI a church" or the IIke, and, sIgnIIIcantIy, cannot be used In
the prImary sense oI brothers as a kInshIp term.
Other generaI tendencIes oI morphoIogIcaI change have been proposed and have proved quIte useIuI.
For Instance, there Is the Important observatIon by C. W. WatkIns 1962 that thIrd-person Iorms are
the major "pIvot" upon whIch new paradIgms are constItuted.
47
However, as wIth other proposed
prIncIpIes, "WatkIns' Law" Is aIso just a tendency; the change oI the 3PL past endIng In Modern Creek
to -ondustan dIscussed In sectIon 1, whIch shows the eIIects oI pressure Irom 1PL and 2PL endIngs
on the 3PL, mIght constItute a counterexampIe, Ior Instance.
n the end, It must be admItted that much morphoIogIcaI change InvoIves IexIcaIIy partIcuIar
deveIopments, and It Is sIgnIIIcant that even the spread oI anaIogIcaI changes seems to be tIed to
partIcuIar IexIcaI Items; thus, unIIke sound change, whIch generaIIy shows reguIarIty In that It appIIes
equaIIy to aII candIdates Ior the change that show the necessary phonetIc envIronment, morphoIogIcaI
change, especIaIIy anaIogIcaI change, Is sporadIc In Its propagatIon. Thus, as shown In sectIon 1, even
wIth the vast majorIty oI nouns In EngIIsh now showIng an InnovatIve -s pIuraI, a Iew Instances oI the
oIder -(e)n marker remaIn In oxen, chIIdren, and brethren.
Thus, It may weII be that Ior morphoIogIcaI change, a generaI theory - that Is, a predIctIve theory - Is
not even possIbIe, and that aII that can be done Is to cataIogue tendencIes, whIch, however vaIId they
may be, do not In any sense constItute InvIoIabIe predIctIons about what types oI changes wIII
necessarIIy occur In a gIven sItuatIon. n that sense, accounts oI morphoIogIcaI change are generaIIy
retrospectIve onIy, IookIng back over a change that has occurred and attemptIng to make sense oI It.
SayIa 10 / 15 18. Diachronic Morphology : The Handbook oI Morphology : Blackwell ReIerence ...
5 ConcIusIon 5 ConcIusIon 5 ConcIusIon 5 ConcIusIon
AIthough morphoIogIcaI change In generaI shows much that Is unpredIctabIe, the exampIes IIsted
hereIn provIde a good overaII vIew oI the types oI changes that are IIkeIy to be encountered In the
hIstorIes oI the Ianguages oI the worId, the causes underIyIng these changes, and the ways IInguIsts
have gone about expIaInIng the observed changes.
One IInaI observatIon on the extent oI the domaIn oI morphoIogIcaI change Is In order. Much
morphoIogIcaI change, as descrIbed here, InvoIves change In IexIcaI Items - In theIr Iorm, theIr
seIectIonaI propertIes, theIr reIatIons to other IexIcaI Items, and so on - and thIs Is aII the more so II
InIIectIonaI aIIIxes are IIsted In the IexIcon Instead oI beIng Introduced by morphoIogIcaI ruIes. t Is
generaIIy accepted that at Ieast certaIn types oI sound changes InvoIve Iexemeby-Iexeme spread (the
cases oI so-caIIed IexIcaI dIIIusIon - cI. Wang 1969 but especIaIIy Labov 1981, 1994), and It seems
that In some Instances, at Ieast, the Impetus Ior the spread oI a pronuncIatIon Into new IexIcaI Items
Is essentIaIIy anaIogIcaI In nature.
48
AIso, there are many so-caIIed IrreguIar sound changes - Ior
exampIe, metathesIs or dIssImIIatIon - that appIy onIy sporadIcaIIy, and thus end up beIng IexIcaIIy
partIcuIar rather than phonoIogIcaIIy generaI. Moreover, at Ieast certaIn types oI changes typIcaIIy
reIegated to the study oI syntactIc change, Ior Instance, changes In agreement patterns,
grammatIcaIIzatIon, movement Irom word to cIItIc to aIIIx, reductIon oI once-bIcIausaI structures to
monocIausaI,
49
and the IIke - that Is to say, much oI syntactIc change other than word order change -
uItImateIy InvoIves morphoIogy or at Ieast morpho-syntax In some way. Thus It Is possIbIe to argue
that much - perhaps most - Ianguage change has a morphoIogIcaI]morphoIexIcaI basIs, or at Ieast
has some morphoIogIcaI InvoIvement. Such a vIew wouId then provIde some dIachronIc justIIIcatIon
Ior the Importance oI morphoIogy In Ianguage In generaI, and thus Ior a morphoIogIcaI component In
the grammars oI partIcuIar Ianguages.
50

ACKNOWLEDCEMENTS ACKNOWLEDCEMENTS ACKNOWLEDCEMENTS ACKNOWLEDCEMENTS
grateIuIIy acknowIedge a IeIIowshIp Irom the AmerIcan CouncII oI Learned SocIetIes ]oInt CommIttee
on Eastern Europe and a sabbatIcaI Ieave Irom the CoIIege oI HumanItIes, The OhIo State UnIversIty,
both oI whIch enabIed me to produce the present pIece, an earIIer versIon oI whIch appeared In OhIo
State UnIversIty WorkIng Papers In LInguIstIcs, 46 (1995), 16-37. wouId IIke to thank Rex WaIIace,
NIgeI VIncent, and RIchard ]anda Ior heIpIuI comments on thIs chapter. owe a consIderabIe
InteIIectuaI debt to ]okI SchIndIer, who opened my eyes some twenty years ago to the wonders oI
dIachronIc morphoIogy and whose stImuIatIng Iectures provIded some oI the exampIes IncIuded
hereIn; dedIcate thIs work to hIs memory, and hope that It wIII serve as a IastIng monument to hIs
InIIuence In our IIeId.
1 ThIs statement conceaIs a Iarge controversy whIch cannot be dIscussed adequateIy here: vIz. whether
sound change Is a pureIy mechanIcaI phonetIc process that Is bIInd to the specIIIc morphemes and words It
operates on, and to theIr morphoIogIcaI composItIon, e.g. whether they are morphoIogIcaIIy compIex or
monomorphemIc. Thus, In prIncIpIe, one couId ImagIne that sound changes couId be morphoIogIcaIIy
condItIoned, and so couId IaII to appIy In, or couId appIy onIy to, certaIn categorIes or partIcuIar
morphemes. The evIdence, however, seems to Iavor vIewIng sound change as beIng onIy phonetIcaIIy
condItIoned In Its outcome at Ieast, wIth apparent cases oI nonphonetIc (so-caIIed grammatIcaI)
condItIonIng beIng the resuIt oI phonetIcaIIy condItIoned sound change IoIIowed by anaIogIcaI
(morphoIogIcaI) change. See Hock 1976 Ior some dIscussIon and reIevant IIterature.
2 These endIngs aII have the Iorm -as In SanskrIt, but, as comparIsons wIth other ndo-European Ianguages
show, they derIve Irom three dIIIerent sources (CEN.SC *-os, cI. Creek pod-s 'oI a Ioot'; NOM.PL *-es, cI.
Creek pd-es 'Ieet'; ACC.PL *-ns, cI. Creek pd-as 'Ieet').
3 These exampIes are drawn prImarIIy Irom the Ianguages know best and thus am best abIe to vouch Ior;
they thereIore have what mIght be perceIved as an ndo-European bIas. However, there Is every reason to
beIIeve that the same types oI exampIes are to be Iound In other Ianguages, and that the phenomena
IIIustrated here are not just ndo-European types oI changes. See e.g. 8IoomIIeId 1946: 18-20; AnttIIa
1972: 91, 97; Robertson 1975; Hock 1991: 200-2; and DaI 1990 Ior some exampIes Irom AIgonquIan,
EstonIan, Mayan, MaorI, and MandarIn ChInese, respectIveIy, to mentIon just a Iew weII-estabIIshed cases
Irom other Ianguage IamIIIes.
4 8ut see beIow regardIng Iorms IIke deIknusaI that dIsturb thIs otherwIse reguIar aIIomorphIc pattern.
5 The AncIent Creek InnovatIve 1DUAL.MEDOPASSVE endIng -methon, whIch IIIIed a gap In the paradIgm (note
the absence oI a 1DUAL.ACTVE Iorm) and seems to have arIsen as a bIend oI 1PL.MEDOPASSVE endIng -metha
wIth the 2DUAL.MEDOPASSVE endIng -sthon (note aIso the 2DUAL.ACTVE -ton), provIdes another exampIe oI a
change In a personaI endIng due to bIendIng] contamInatIon.
6 The encIItIc Iorm, occurrIng as It does wIth a stop, presumabIy reIIects a combInatory varIant oI pu aIter a
sIbIIant.
7 See Anderson 1988a Ior dIscussIon oI the spread oI the s-pIuraI In EngIIsh; he argues that the mechanIsm
Is one oI the eIImInatIon oI IexIcaIIy specIIIed IdIosyncrasIes and the emergence oI the deIauIt markIng; he
notes that thIs InterpretatIon Is consIstent wIth, and In Iact predIcted by, the prIncIpIe oI dIsjunctIve
orderIng Ior morphoIogIcaI ruIes. For a sImIIar exampIe Irom Cerman, where an -s markIng Ior pIuraI Is
spreadIng, see ]anda 1990.
8 For Instance, [ows] and [(h)worIs] are gIven In AHD 1992 as (InnovatIve) varIants; [hawsz], whIIe
common In CentraI OhIo at Ieast, has not yet been enshrIned In the dIctIonary.
9 As wIth the spread oI the s-pIuraI (see n. 7), thIs Ioss oI morphophonemIc voIcIng can be seen as the
removaI oI an IdIosyncratIc specIIIcatIon Irom the IexIcaI IIstIng oI each such noun. See aIso AnttIIa 1972:
126-7 Ior dIscussIon oI thIs exampIe and oI paraIIeI ones InvoIvIng consonant gradatIon Irom 8aItIc FInnIc.
t shouId be noted that occasIonaIIy the IdIosyncratIc markIng has spread to a noun not orIgInaIIy
undergoIng thIs process; e.g. dwarI orIgInaIIy had no overt pIuraI marker In OId EngIIsh, so that the varIant
pIuraI dwarves, aIongsIde the synchronIcaIIy more reguIar dwarIs, represents a spread oI the synchronIcaIIy
IrreguIar pattern.
10 See e.g. Prokosch 1938 Ior thIs reconstructIon.
11 The sItuatIon Is actuaIIy a bIt more compIIcated, as Is cIear Irom the Iact that earIy OHC had -Ir- In some
sInguIar Iorms, specIIIcaIIy the genItIve, the datIve, and the InstrumentaI; but as the suIIIx came to be
Interpreted pureIy as a marker oI number, as the nomInatIve Iorms wouId Iead a speaker to surmIse, It
dIsappeared Irom the sInguIar. StIII, SaImons (1994: 224-5), In hIs recent dIscussIon oI these Iacts, notes
varIabIIIty, In partIcuIar wIth regard to -Ir- Iess pIuraI Iorms, throughout the OHC perIod and dIaIect space,
and concIudes that -Ir- as markIng onIy pIuraIIty was not "IIrmIy estabIIshed In many dIaIects." See aIso
Anderson 1988a Ior an InterpretatIon In terms oI changes In IexIcaI specIIIcatIons.
12 Note aIso that sInce In earIIer stages oI CermanIc, Wort dId not have thIs pIuraI markIng (cI. OHC SC
wort ] PL wort), the extensIon oI thIs umIaut-pIus-(e)r pIuraI markIng Is a process paraIIeI to the exampIe
gIven oI the -s pIuraI In EngIIsh; see aIso nn. 7 and 11.
13 That thIs archaIc InIIectIon Is embedded In a IIxed phrase (IIkewIse VedIc SanskrIt dan 'house] CEN.SC',
Irom *dem-s, Iound In the IIxed phrase patIr dan 'master oI the house') Is not surprIsIng, Ior It shows the
retentIon oI an oIder pattern In what Is In essence a synchronIcaIIy unanaIyzabIe expressIon (IIke an IdIom).
From a methodoIogIcaI standpoInt In doIng hIstorIcaI morphoIogy and morphoIogIcaI reconstructIon, It Is
oIten useIuI to Iook to such expressIons Ior cIues as to earIIer patterns.
14 The reconstructIon oI the root Ior thIs word Is somewhat controversIaI, and onIy the stem suIIIx Is at
Issue here, so no attempt Is made to gIve a compIete reconstructIon.
15 The *-os]-es endIng In these Ianguages may ItseII be a Iate PE repIacement Ior an earIIer sImpIe *-s
endIng, based on such Iorms as the OId rIsh genItIve sInguIar anmae 'oI a name', where the endIng Is Irom
*-men-s (so Thurneysen 1970: 60); hence the specIIIcatIon "pre-Creek" Is used here Ior the endIng, sInce It
may not be the oIdest Iorm oI thIs InIIectIonaI endIng wIth thIs noun In PE.
16 A -t- extensIon Is Iound wIth severaI other nomInaI stem cIasses In Creek - e.g. the neuter -as- stems -
but It Is not Iound wIth aII members oI the cIass, and a Iew specIIIc nouns - e.g. kreas 'meat' - show It
earIIest In the genItIve sInguIar (4th century 8C), wIth spread to other case Iorms comIng much Iater. Even
wIth a noun IIke noma, whIch, as noted beIow, shows the extensIon oI the -t- Into other case Iorms, earIy
(HomerIc) Creek shows no (metrIcaI) trace oI the -t- In the datIve pIuraI (see ChantraIne 1973: 74-5, 82-3).
17 See Thomason and KauIman (1988: 242-3) Ior some dIscussIon oI the substratum hypothesIs, though
Stang (1966: 228-9) argues agaInst thIs vIew.
18 That thIs one-tIme postposItIon has become a true case endIng In Oscan Is shown by Its appearance on
an adjectIve, In apparent agreement wIth the noun It modIIIes; see 8uck 1928: 114 Ior thIs InterpretatIon.
ThIs InnovatIve Iorm presumabIy repIaced an InherIted IocatIve, stIII Iound to a IImIted extent In LatIn.
19 WIthIn the IIterature on grammatIcaIIzatIon (e.g. Traugott and HeIne (eds) 1991a, b; Hopper and Traugott
1993) there Is much dIscussIon oI the cIaIm that deveIopments In grammatIcaIIzatIon are subject to a
prIncIpIe oI unIdIrectIonaIIty, whereby movement supposedIy Is aIways Irom Iess grammatIcaI to more
grammatIcaI, wIth meanIngs aIways goIng Irom concrete to abstract; see ]oseph and ]anda 1988, CampbeII
1991, ]anda 1995, and ]oseph 1996a Ior dIscussIon oI some counterevIdence to thIs cIaIm.
20 Compare the sItuatIon wIth morphophonemIc voIcIng In EngIIsh pIuraIs, dIscussed above (and see n. 8),
and note the ongoIng varIatIon In the markIng oI past partIcIpIes In EngIIsh, wIth oIder -(e)n In some verbs
gIvIng way to the more wIdespread -ed (as In sewn ] sewed, shown ] showed, proven ] proved, etc.).
21 The SpanIsh exampIe suggests that changes In case-markIng systems are not restrIcted to the dIstant
past, though the IaIIure oI -go to spread to other pronouns (Indeed, It has retreated somewhat Irom wIder
use In oIder stages oI the Ianguage) or to be used wIth other preposItIons argues that It Is not reaIIy a
casemarkIng devIce. SImIIarIy, the InnovatIve use In certaIn varIetIes oI wrItten EngIIsh oI InwhIch, as In
ShoppIng Is a task InwhIch one shouId enjoy, has Ied some researchers - e.g. ]. R. SmIth 1981 and RIIey and
Parker 1986 - to anaIyze It as a new case Iorm oI the reIatIve pronoun, though Montgomery and 8aIIey
1991, In an extensIve study oI the use oI the Iorm, argue persuasIveIy agaInst that InterpretatIon.
NonetheIess, such exampIes provIde the opportunIty to wItness the Iate oI case-IIke Iorms that occur In a
restrIcted domaIn oI the grammar, and thus provIde some InsIghts Into the generaI processes by whIch such
Iorms can arIse and take hoId In a Ianguage.
22 Most IIkeIy, the path oI deveIopment was through the condItIonaI tense (past tense oI the Iuture) shIItIng
IIrst to a pIuperIect (compare the IIuctuatIon In Modern EngIIsh between a pIuperIect Iorm and what Is
IormaIIy a past tense oI the Iuture utIIIzIng the modaI wouId In II cIauses - e.g. I had onIy known = I
wouId have known), Irom whIch a present perIect and other perIect IormatIons couId have deveIoped. See
]oseph 1983: 62-4; 1996b, Ior some dIscussIon.
23 The exact path Irom theI grapseIn to a rapso Is a bIt convoIuted and IndIrect; see ]oseph 1983: 64-7;
1990: ch. 5; 1996a Ior dIscussIon and Iurther detaIIs. The onIy materIaI that can Intervene between a and
the verb In Modern Creek Is other bound eIements, In partIcuIar the weak object pronouns.
24 Even In OId Church SIavonIc, there was some varIabIIIty In category membershIp, and nouns Ior 'sIave',
'chIId', varIous anImaIs, etc. showed some IIuctuatIon between anImate and nonanImate InIIectIon; see Lunt
1974: 46 and MeIIIet 1897 Ior some dIscussIon. The descrIptIons In ComrIe and Corbett (eds) 1993 provIde
a useIuI overvIew oI the reaIIzatIon oI anImacy throughout the varIous SIavIc Ianguages. Thomason and
KauIman (1988: 249-50) suggest that thIs category may have deveIoped through a UraIIc substratum
shIItIng to SIavIc.
25 For EngIIsh, the AmerIcan herItage dIctIonary oI the EngIIsh Language, 3rd edn (1992), wIth Its "ndo-
European Roots AppendIx" by CaIvert WatkIns (see aIso C. W. WatkIns 1985), Is an exceIIent exampIe oI such
a resource.
26 yeIIow Is Irom OId EngIIsh geoIu, Irom proto-CermanIc *geIwaz; gaII Is Irom OId EngIIsh geaIIa, Irom
proto-CermanIc *gaIIn-.
27 Note aIso that words that are etymoIogIcaIIy unreIated can come to be perceIved by speakers at a Iater
stage as reIated, perhaps even merged Into dIIIerent meanIngs oI the same word: e.g. the body part ear and
ear as a desIgnatIon oI a pIece oI corn are etymoIogIcaIIy dIstInct (the Iormer Irom PE *ous- 'ear', the Iatter
Irom *ak- 'sharp'), but they are IeIt by many speakers to be dIIIerent meanIngs oI one poIysemous IexIcaI
Item.
28 See ]anda 1982, 1983, Ior a thorough dIscussIon oI the reIevant Iacts supportIng thIs anaIysIs oI Cerman
umIaut. The productIvIty oI umIaut does not In ItseII argue Ior It stIII beIng phonoIogIcaI; In that sense, the
Cerman sItuatIon Is now sImIIar In nature, though not In scope, to the very IImIted umIaut eIIects stIII
present In EngIIsh, e.g. In a Iew IrreguIar pIuraIs (man]men, Ioot]Ieet, etc.) and verbaI derIvatIves (drInk]
drench, etc.).
29 Thus there Is an Important InteractIon wIth sound change to note here, Ior sound change can obscure or
remove the condItIonIng eIements Ior a phonoIogIcaI process, thereby renderIng the process opaque Irom a
phonoIogIcaI standpoInt and makIng It more amenabIe to a morphoIogIcaIIy based anaIysIs. RecaII aIso that
sound change can pIay a roIe In the reductIon oI compounds to monomorphemIc words and oI phrasaI
unIts, such as noun pIus postposItIon, to monoIexemIc expressIons.
30 Many such -gate Iorms are documented In notes In AmerIcan Speech; see ]oseph 1992 Ior reIerences.
31 OI course, some oI these EngIIsh Iorms are susceptIbIe, seemIngIy more so than natIve pIuraIs, to
reanaIysIs as sInguIar; crIterIa Is quIte IrequentIy used as a sInguIar, and a pIuraI crIterIas can be heard as
weII. SImIIarIy, the AIbanIan pIuraIs In -IIare]-Iere show the natIve pIuraI suIIIx -e added to the TurkIsh -
Iar]Ier endIng, somewhat paraIIeI to Iorms IIke crIterIas.
32 The dIstInctIon drawn by Thomason and KauIman (1988) between borrowIng and Ianguage shIIt Is a
crucIaI one, wIth the Iatter sItuatIon beIng the contact vehIcIe Ior some oI the more "exotIc" morphoIogIcaI
changes. TheIr dIscussIon Is perhaps the most compIete enumeratIon oI the wIde range oI possIbIe contact-
Induced changes, IncIudIng those aIIectIng the morphoIogy. See aIso n. 24 above concernIng a
IanguageshIIt source Ior the IntroductIon oI the new anImacy subcategory In SIavIc.
33 See AnttIIa 1977 and AnttIIa and 8rewer 1977 Ior basIc dIscussIon and bIbIIography on anaIogy In
Ianguage change.
34 AnaIogy can aIso provIde dIrect evIdence Ior the exIstence oI the tIght reIatIons among members oI
cIusters oI Iorms that aIIow Ior an InIerence oI a (psychoIogIcaIIy) reaI category. For Instance, the Iact that
drag]Iag]IIag]Iag couId aIIect [sk] and draw It Into theIr orbIt as sag Is prIma IacIe evIdence oI the
strength oI the connectIons among these Iour words. SImIIarIy, the dIaIectaI extensIon oI the -th
nomInaIIzIng suIIIx, whIch shows IImIted productIvIty wIthIn the domaIn oI dImensIon adjectIves (cI.
wIde]wIdth, deep]depth, etc.) to hIgh, gIvIng [hayt] (thus aIso wIth some contamInatIon Irom heIght to
expIaIn the occurrence oI the -t-) can be seen as evIdence oI the subcategory wIthIn whIch the suIIIx Is
productIve.
35 Another perspectIve on the cognItIve dImensIon In anaIogy Is provIded by Andersen's IntroductIon oI the
roIe oI abductIve reasonIng In anaIogIcaI reanaIysIs, as dIscussed most notabIy In Andersen 1973, 1980.
36 EspecIaIIy the work by WoIIgang DressIer, WIIII MayerthaIer, WoIIgang WurzeI, and others; see e.g.
DressIer et aI. 1987, MayerthaIer 1981, WurzeI 1984. See aIso ShapIro 1990 (wIth reIerences), where a
somewhat dIIIerent vIew oI the roIe oI semIotIcs In Ianguage change, as appIIed to morphophonemIcs, can
be Iound.
37 OI course, not aII grammatIcaIIzatIon InvoIves morphoIogIcaI change, except InsoIar as It aIIects IexIcaI
Items. The papers In Traugott and HeIne (eds) 1991a, b contaIn numerous reIerences to the cognItIve
dImensIon oI grammatIcaIIzatIon; see aIso Hopper and Traugott 1993 Ior dIscussIon and reIerences.
38 See aIso the recent work by Skousen In whIch an expIIcIt and IormaI deIInItIon oI anaIogy Is used to
create a predIctIve modeI oI Ianguage structure; among the tests Ior thIs approach (In Skousen 1989: ch. 5)
Is Its appIIcatIon to hIstorIcaI drIIt In the IormatIon oI the FInnIsh past tense.
39 ThIs vIew has Iong been assocIated wIth WIIIIam Labov, and Is expressed most recentIy In Labov 1994:
45: "n IIne wIth the generaI approach to Ianguage as a property oI the speech communIty, wouId preIer to
avoId a Iocus on the IndIvIduaI, sInce the Ianguage has not In eIIect changed unIess the change Is accepted
as part oI the Ianguage by other speakers."
40 One couId say oI course that there has been a change In the morphoIogIcaI ruIes that Introduce the stem
varIants Ior 'month', but that stIII brIngs one no cIoser to understandIng why the change occurred. Once
'Zeus' and 'month' share the same patterns oI aIternatIon, then a generaIIzatIon over these two Iorms Is
possIbIe, aIIowIng Ior some sImpIIIIcatIon In the grammar. However, the change cannot have occurred just
to sImpIIIy the morphoIogIcaI ruIes Ior 'Zeus' somewhat by gIvIng them wIder appIIcabIIIty, sInce a greater
sImpIIIIcatIon wouId have arIsen had the stem aIternatIon Ior thIs noun been eIImInated aItogether (as It was
In some dIaIects that Innovated a nomInatIve Zn).
8IbIIographIc DetaIIs 8IbIIographIc DetaIIs 8IbIIographIc DetaIIs 8IbIIographIc DetaIIs
The The The The Handbook oI MorphoIogy Handbook oI MorphoIogy Handbook oI MorphoIogy Handbook oI MorphoIogy
EdIted by: EdIted by: EdIted by: EdIted by: Andrew Spencer And ArnoId M. ZwIcky
eS8N: eS8N: eS8N: eS8N: 9780631226949
PrInt pubIIcatIon PrInt pubIIcatIon PrInt pubIIcatIon PrInt pubIIcatIon date: date: date: date: 2001

41 Thus IemaIe makes more "sense," and thus Is a better IIt between Iorm and meanIng, when IormaIIy
paIred wIth Its antonym maIe; sImIIarIy, provIdIng a "partner" Ior the unIque stem aIternatIons oI 'Zeus'
makes the Zeu-]Zn- aIternatIon Iess IrreguIar, and thus more motIvated and easIer to deaI wIth Irom a
cognItIve standpoInt.
42 SImIIarIy, note Iorms such as dwarves In EngIIsh, mentIoned above In n. 9, that run counter to the
generaI IeveIIng out oI stem dIIIerences due to voIcIng oI IrIcatIves In the pIuraI.
43 See WInters 1995 Ior an EngIIsh transIatIon, wIth some commentary, oI thIs Important oIt-cIted yet
generaIIy IIttIe-read paper.
44 See VIncent 1974; CoIIInge 1985: 249-53; Hock 1991: ch. 10; and WInters 1995 Ior more detaIIed
dIscussIon and comparIson oI the two schemes.
45 The statements oI these prIncIpIes and theIr comparIson are taken Irom the IIIumInatIng account In Hock
1991: ch. 10.
46 See Robertson 1975 Ior an exampIe oI the Iourth Iaw Irom Mayan.
47 See CoIIInge 1985: 239-40 Ior dIscussIon and reIerences.
48 For exampIe, a possIbIe scenarIo Ior IexIcaIIy dIIIuse spread oI a sound change Is the IoIIowIng: II IexIcaI
Item X shows varIatIon In pronuncIatIon between X and X', and Item Y has some oI the same phonoIogIcaI
Ieatures as Item X, speakers may extend, anaIogIcaIIy usIng X as the modeI, the varIant pronuncIatIon X' to
Y, so that Y comes to show varIatIon between Y and Y'. I the competItIon Is uItImateIy resoIved In Iavor oI
X' and Y', the sound change wouId have been generaIIzed.
49 See e.g. DeLancey 1991, regardIng such cIause reductIon In Modern TIbetan (dIscussed In Hopper and
Traugott 1993: 198-201).
50 n ]oseph and ]anda 1988, the cIaIm Is advanced that grammars are "morphocentrIc," and the prevaIence
noted above In sectIon 1 oI dIachronIc movement Into morphoIogy (Irom syntax and Irom phonoIogy), as
opposed to the reIatIve rarIty oI movement out oI morphoIogy, Is taken as dIachronIc evIdence Ior the
centraIIty oI morphoIogy. ThIs cIaIm Is based on an assumptIon that Iacts Irom dIachrony can have
reIevance Ior the constructIon and evaIuatIon oI synchronIe grammars, and to the extent that It Is vaIId,
provIdes some support Ior treatIng such Iacts as Important.
CIte thIs artIcIe CIte thIs artIcIe CIte thIs artIcIe CIte thIs artIcIe
]OSEPH, 8RAN D. "DIachronIc MorphoIogy." The Handbook oI MorphoIogy. Spencer, Andrew and ArnoId M.
ZwIcky (eds). 8IackweII PubIIshIng, 2001. 8IackweII ReIerence OnIIne. 28 December 2007
<http:]]www.bIackweIIreIerence.com]subscrIber]tocnode
Id=g9780631226949chunkg978063122694921>

Handbook Morphology

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Handbook Morphology

Hochgeladen von

Copyright:

Verfügbare Formate

11. MorphoIogy and the LexIcon: LexIcaIIzatIon and ProductIvIty 11.

Das könnte Ihnen auch gefallen