Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）73−92 73
Affixation and the Notion Head of a Word ：
The Relationship between Prefixation a皿d Subcategorization＊
Takayasu NAMIKI
Depaγ伽eη oブEπgzεsん Fac読y oブE伽ca蜘η
乃α7a短翫加εγsめ
L
当ﾐo，310，Japaη
（Received October 15，1981）
Abstract
Evidence based on affixation in Enghsh and Japanese is presented for Williams s
（1981）asymmetric system of word formation and against I一ieber s（1980）symmetric system． Several English prefixes（bo秘η置e7−，∫η一，ε呪4γ一， and碗3りare discussed as marked examples which increase subcategorization possibilities of the stems to which they are attacLed or of the prefixed words as a whole． Especially it is argued that the negative prefixes constitute examples which require the computation of both the meaning of pre一
?奄??刀@and that of stems． Feature percolation mechanisms are tentatively proposed to account for the correlation of the prefixes and stems．
§1 1ntroduction
Significant progress has been made｛n the study of word formation in gene．
rative grammar in the past ten years． For example， a number of important pro一 posals have been made in Halle（1973）， Siegel（1974，1978）， Aronoff（1976），
Allen（1978a，1978b）， Roeper and S iege1（1978）， Carlson and Roeper（1980），
Lieber（1980）， Williams（1981）， and Selkirk（forthcoming）． In particular，
Williams（1981）and Lieber（1980）discuss the important notion of head of a word and general ways of assigning a head to a word．
In section 21examineWilliams s（1981）discussion of head of a word and Righthand Head Rule， and Lieber s（1980）critique of the Righthand Head Rule
．
≠獅п@the alternative she proposes． After reviewing arguments of Williams and Lieber， I adduce evidence in favor of Williams sanalysis in section 3． Further一 more， my own proposal is made on the distinction between Core Word Fbrmation and Periphery Word Formation with respect to a lexical head． In section 4 the interaction between word formation and syntax is discussed． Specifically a prefix is dealt with which affects the subcategorization of the stem to which it
74 Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
is attached． In order to cover such prefixes I make the distinction between Core Word Formation and Periphery Word Fbrmation again with respect to feature percolation mechanisms． I argue for the existence of facts which require the
−computation of both the． meaning of a prefix and that of a stem． Section 5 is a summa「y・
§2 Review of Williams（1981）and Lieber（1980）
2．1．Williams（1981）1）
Pointing out the three facts to be given（1）， Edwin Williams gives the definition of head as in（2）， and sets』up the Righthand Head Rule as in（3）．
一
i1）a．It is generally the case that a suffix determines the category of a word of which it is a part．（Williams 1981，p．248）
e．g、 X−ism−N X−ize→V X−ish＿A V。ion→N X−fy→V
V．．er→N
A引ess→N
b．Prefixes do not in general determine the category of the words they attach to；rather， the category of a word of the form pγeβκXis deter・
mined・by the category of X．（Loc． c拡）
e．9． counter十revolutionN is a N coullter一トsinkv is a V
coullter十productiveA is an A
c．This difference between p艶fixes and suffixes is exactly what we would expect given the notion of head @ asuffix will be headr in the structure（13a）but X will be head in a structure like（13b）．（Loc．
cπ．）
（13）a． b．
w s働茄κ prefix X
（2）Definition of head
If both X and the head of X are eligible members of category C， then X∈C…≡head of X∈C．（乃ゴd．， p．247）
（3）The Righthand Head Rule2）
In morphology， we define the head of a morphologically complex word to be the righthand member of that word．（泌ゴ｛義， p．248）
Williams illustrates the heads by italicizihg them in（4）， and regards both−∫oη
一
T丘a e∂払ca琵oπas heads ofγeε（1払caだ073 in（5）．
Namiki：Affixation and the Notion Head of a Word 75
（4）a． b．
o ■ o
撃獅rtrUCt 診oη re LηS置7°錫Cオ
（5）
re e4㏄aお。π
Taking these into consideration， he says， We must extend category membership to suffixes；that is， we must say that一おんbelongs to the category A，一たe to the category V， and一ゴoηto the category N， （乃ゴ6．．， p．249）showing the derivation of coηs置γ㏄翻oηandγecoη5孟γ㏄孟as in（6）．
（6）a． N b． V
constructV ゴoπN re coη5オ物c V
Williams also points out that a similar situation holds for most compounds 一the righthand member determines the category of the whole， as showp in（7）．
（7）a． A． b， N 酵 c． V
6ffp wん甜eA dryA （オocゐN barN 診e記v
Needless to say， Williams notices the existence of systematic exceptions to RHR：one is the prefix eπ一（as in enrage， endear， ennoble， encase）， which has a lefthand head；the other is exocentric compounds hke pπ5h zゆandγ槻（10り〃物 which have no heads at all．
To summarize， Williams proposes an＆3ッ観耀孟γゴc system in word formation in that he sets up RHR， that is， he defines the head of a morphologically complex word to be theγゴgん言んa掘member of that word．
2．2 Lieber（1980＞
Showing the three sets of counterexamples to Williams s RHR to be given b610w， Lieber argues for a 5ッ魏観ε伽弛system in morphology in that she allows alefthand head which is on a par with a righthand head．
First， Lieber says that there exist category・changing prefixes， that is lefthand heads， in German in addition to eπ一in English． She adducesッe処and ゐe．as such， as shown in（8）．
（8）a．HolzN−verholzenv； FreundN−befreundenv b．jungA−verj廿ngenv； ruhigA−−beruhigenv 冒 c．laufenv−verlaufen v； nehmenv−benehmenv d．langsamAdv−verlangsamenv
76 Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
Second， she states that there exist languages with lefthand compounds， al一 though they are predicted not to exist by the RHR． On the basis of a）leiKicalized meaning， b）the fact that they consist of only two stems， c）characteristic stress contour，and d）inseparabili‡yof the elements by modifiers， she gives two kinds of lefthand compounds in Vietnamese in（9）．
（9）a． nha thuong
establishment be wounded＝hospital b． lam ruδng
do， make rice field＝engage in farming
Third， heber shows that a transparent suffix exists， contrary to the predic一 tions of the RHR． She claims that a suffix diminutive s泥十yin Spanish is transparent in that it adds neither category features nor any gender features of its own． Observe the following：
（10） grandesito ・quite large 二A madresita mother−dimin． ＝N
To sum up， Lieber has presented three sorts of phenomena which all argue for the existence⑪f lefthand heads， and therefore for symmetrical structure．
（Lieber 1980， p．102）She presents a symmetric system as an alternative to Williams sasymmetric system． It embraces：
（11）a．the permanent lexicon including lexical entries， etc．
b．asingle context−free rewrite rule giving unlabeled structure
c．insertion of morphemes into this structure subject to subcategoriza・
曹 9 ・
狽撃nn reStrlCt10nS
d．node labeling conventions．
First， Lieber spermanent lexicon includes all lexical elements， whether they are words， stems， or affixes， as those having lexical entries． The purpose of a lexical entry is to specify all information about a terminal element which is unpredictable and idiosyncr興tic to that element． Such information consists
of：
（12）a．the category and conjugation or declension class of an item3）
b．phonological representation
■ ・
メDsemantlc representatlon d．subcategorization4）
e．diacritics
f．insertion frames5）
Namiki：Affixation and the Notion Head of a Word 77
Sample lexical entries are given in（13）：
（13）a．PREFIX：in− （phonological representation）
唐?高≠獅狽撃メ@representatlon：negatlve cat・g・ry／・ubcat・g・ri・ati・n・［A［A
insertion frame：（whatever insertion frames for As
look like）
diacritics：Level I
b．SUFFIX：−ize （phonological representation）
引 ， ．
唐?高≠獅狽撃メ@representatlon：causatlve
・at・g・ry／・ub・at・g・rizati・n・］Nコv insertion frame：NP．＿（NP）
diacritics：Level H
。．STEMS6）、 run（ph。n。1。gical represent。ti。n）
． り
唐?高≠獅狽撃メ@representatlon：．，．
categ°「y：V［＿］V
insertion frame：NP （NP）
diacritics：［−Latinate］
Second， Lieber posits a single context・free rewrite rule which will gener．
ate駕η1a6c1α1 binary branching tree structures． （乃認．， p．82）It will give us a variety of branching structures like the following：
（14）a． b． c． d．
Third， terminal morphemes are inserted into these tree structures sublect to their subcategorization restrictions， as illustrated in（15）：
（1与＞a． b．
haPP詞A ness］N
standard］N ize］v 竃］v
Finally， Lieber establishes a set of Feature Percolation Conventions：
（16）a．Convention I：all features of a stem morpheme， including category features， percolate to the first non−branching node dominating that morpheme．
b．Convention H：all features of an affix morpheme， including category
78 BulL Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
features， percolate to the first branching node domi・
nating that morpheme．
c．Convention 1皿：If a branching node fails to obtainfeaturesby Conven一 tion H，features from the next lowest labeled node are automatically percolated up to the unlabeled branching node．7）
d．Convelltion W：In compound words inEnglish features from the right・
hand stem are percolated up to the branching node dom一 inating the stems．
Conventions I・皿are universal and concern affixation， while Convention IV is，
Lieber claims， specific to English compounding． The following tree structures are cited to illustrate the apPlication of these conventions．8）
（17）a。 b． N
⑪
A A
①
l IhapPy ness hapPy ness
十A 十N
・・@ N ⑪ 乱 ＋N ⑰
N 十A 十N
戟@ ① l I①
counter weight black board
十N 十A 十N i i
The gist of Lieber s symmetricsystem lies inher statement that derived words as a whole characteristically adopt all feature values of their o窃εγ観05言mor一 phemes ．9）（1ゐ∫di．， p．84， italics mine TN）It does not matter whether such morphemes are righthand or lefthand． She also combines the category labeling and feature percolating mechanisms， effecting a significant simplification．
§3 Examination of Liebeピs and Williams s Proposals
The essential difference between Lieber sthesis and Williams s is that Lie一 ber argues for symmetry between prefixes and suffixes in the ability to change acategory of the whole word， while Wjlliams supports asymmetry． The relevant question here is：does the symmetry in the feature percolability between pre一 fixes and suffixes really exist？In order to answer this question， English and Japanese data are investigated in sections 3．1．and 3．2． respectively．
Namikl：Affixation and the Notion Head of a Word 79
3．1． Data in Present−Day English
Alarge number of living affixes are picked up from Jespersen（1942）and Marchand（1969）and are examined as to what categories they can attach to， and whether they are category−changing orηot and， if so， what categorieslthey yield．
Out of seventy three prefixes only six are category−changing．10） They are as follows：
（18） a．a− N （afire）
@ A
@ V （agl itter）
A−→Adv （aloud）
b．6e− N （behead）
一 V
A （becalm）
c．de・ N−一一今V （defrost）
d．d∫5− N （disbar）
@ V
@ A （disable）
e．eπ一 N （encage）
D V
@ A （enrich）
f．o％言一 N （outdistance）
@ V
@ A （outsmart）
Clearly these prefixes are marked．
In contrast， the malority of prefixes are non−category−changing。 Five of them
（6ε㌧dε㌧dis−，εη㌧and out一）coincide with the category−changing prefixes． （See Appendix．）Out of ninety eight suffixes， sixty six are category−changing and forty five are non−category−changing． Thirteen suffixes belong to both classes：
一acツ，−age，−do飢，一擢，一んood，一ゴsん，一お祝，一ゴs孟，一赫e，−4y，−s痂P，−so観ε， and一ツ．（See ApPen・
dix．）
3．2。 Data in Present−Day Japanese
Alarge number of living affixes are collected from Kanno（1964）， Nomura
（1977），and Nagashima（1980）， and are investigated from the same point of view as in English．11）
一一尼
nut。f。ixty eight prefixes・nly three are cleariyとateg・ry・changing．12）Th・y are：
80 Bull． Fac． Educ．，Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
（19）a．4aゴー N−A13） （daikibて）na， cf．＊daikibo）
（large） （large−SCale）
b。8んo− N−A （shokibona， cf．＊shokibo）
（small） （small−scale）
c．勧一 N＿A （hukisokulla， cf．＊hukisoku）
（non−） （irregular）
It is obvious that these prefixes belong to marked《−xalllples and that the majority of prefixes are non−category−changing．（See Appendix．）
Out of 110 suffixes， forty three belong t⑪ the category−changing class and sixty eight belong to the non−category−changlllg c量ass。（Only一んe belongs to both classes． See Appendix．）Both types of sllfflx｛｝s are numer⑪us．
3．3．Comparison of prefixes獄nd suffixes in Engiish and Japanese
Comparing the facts descrlわed｛n section 3．2．， we can state significant gen・
eralizations h⑪lding of bて）th Eng五ish and Japanese as follows：
（20）a．Prefixes：Category−challghlg prefixes al・e marked and Ilon−category一 changingて♪nes are unmarked in both English and Japanese．
b．Suffixes二Both categol・y−challging suffixes andnon−category−changing ones are numerous in both English and Japanese．
It is very unlikely that the facts stated above are accidental． If Lieber s sym一 metric system is correct， it can be expected that there should be many Ihore cat・
egory−changing Pγe1乞κes， like category−cllanging sufflxes． Ill fact， there exist only six such preflxes in English alld only three（or at most five）such prefixes in Japanese． Under Lieber sprinciples it is very difficult if not impossible to explain naturally why such an asymmetry between prefixes and suffixes does exist and， furthermore， why this asymmetry exists not only in English but also in Japanese．
Next， let us consider， in particular， unmarked non−category−changing pre一 fixes in English。 There are seventy two such prefixes， as shown in Appen一 dix． The prefix co駕厩eγ一amollg them can be seen as categoryless， that is， it can be attached to any lexical category． So it is compatible with Lieber s assump一 tion that derived words characteristically adopt all feature values of their out一 ermost morphemes． There are twenty two prefixes that can attach to only one category（N， A， or V）． They also are compatiblewith her assumption in question：
since the stems to which they can attach have the same category as the prefixes，
either the stem or the prefix can be responsible for the category of the whole word．
However， forty nine prefixes（about 70％of all）can be added to two or three
Namiki：Affxation and the Notion Head of a Wofd 81
categories， and they cannot be regarded as categoryless， as c側πオeγ一． They are as follows：
（21）Non−category−changing prefixes
a．those which can attach to threeρategories
N，A， V ＝cかc％観一， co−，（1∫3−，んツpeγ一，珈陀吃観ゴs・，ηoη一， posか，
14 Pγε㌧γe孟γo㌦5駕δ一，5叩εγ㌧and孟γaη8−（13）
N，A， Adv＝aη孟ゴー（1） ． b．those which can attach to two categories
N，A ＝aうa撹εうa祝孟o・，δゴ・， cηP置o㌧dあhc煽一，んツpo4η一，
碗プγa ， ぎ7拡γaう9πe言a−，7ηぎ（Z−，鵬0η0−， 盟配而一， πeo−，
pa1・・、 paη㍉pa・a−， P・・ど、 P・1ツー， P・・−2，仰一3，
35 Pse認o㌧seηzぎうs∠y7z−，置γゴ扁，配孟ra、鋭ηゴー， andッ♂ce−（30）
A，V ＝s叩γa・（1）
N，V ＝coη孟γa−，1bγε一， and魏a1−（3）
N，Adv ＝πo−（1） ・ As far as these examples are concerned， it is natural to think that it is not pre一
fixes but stems毛hat determine the category of the whole words， that is， that are
heads of the whole words． For these prefixes do not change the category of 辱
狽??@whole words when they attach to stems：for example， co−can be added to N
（co−a刎んoγ）， A（co−aのace初）， or V（coopeγa孟e）without changing the category of the whole．，
If we follow Lieber s Feature Percolation Conventions， we cannot get the correct information that， for instance， coopeγa孟εis a V and that a伽06∫ogγapんッ is a N． For 1）co−and a掘o−have the subcategorization features showing that the former can attach to N， A， or V and the latter to N or A；2）Convention H apPlies to coopeアa孟e and aπ診06ゴ09γapゐツ， letting the subcategorization features at issue percolate；and 3）Conventi6n皿cannot apply to let the features［十V］of opeγaオe and［十N］ofゐ ogγapんッpercolate because of the prior application of
曹
bonvention n（cf．（16c））． Observe the following：
（22）a．
［＋＿［N，［A，［v］
Conv． H Conv．皿 ．
十V
@ Ico operate
Conv．1
［十 ［N，鼠，隙］ ［十V］
幽
82 Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
b・ ［＋一一［N，鼠］
Conv．皿 Conv． H
Iauto biography十N
Conv．1
［十＿．一［N，鼠］ ［十N］
To sum up， the class of prefixes in（21）constitutes crucial counterevidence to Lieber s symmetric theory．
Ihave examined the validity of one of Lieberもcounterarguments toWilliams s RHR and shown its inadequacy． The material has concerned many prefixes（and suffixes）in English． Three other counterargumellts of Lieber s（i． e． category一 challging prefixes in German， lefthand compounds in Vietnamese， and the trans・
parent suffix in Spanish）remain to be discussed． I would like to point out their common weaklless in constl加tillg arguments against the RHR． In all the putative arguments it is I1⑪t demonstrated that the examples in question be．
long to unmarked cases hl the language at stake，14）and so they are inconclusive at best．
In conclusioll， it is a significant generalization that there is asymmetry be一 tween prefixes alld suffixes in unmarked cases despite Lieber s critique．
3．4． My ProposaI
In spite of its essential correctness Williams s system also is inadequate in that it does not explicltly distinguish between marked cases and unmarked ones in dealillg with affixation and compounding。 Therefore， I would like to make a proposal on markedness in morphology．
It is well knov》n that Chomsky and Lasnik（1977，p．430）claim that universaI grammar incorporates a theory of core grammar and that systems that fall within core grammar constitute the unmarked case∵ Following their idea，1 want to suggest here that we divide word formation into Core Word Formation and Periphery Word Formation and that the righthand head is involved in the former and the lefthand head is involved in the latter， as shown in（23）．
（23） ／Core Word Formation−一一Righthand Head Word Formation
Periphery Word Formation Lefthand Head
The examples of lefthand heads are a−，δε一， de−，（旛一， eπ一， and o初一in English，ッeγ一and
6e−in German， lefthand compounds in Vietnamese， and the transparent suffix in Spanish．
Namiki：Affixation and tLe Notion Head of a Word 83
§4 1nteraction between Word Formation and Syntax：Prefixation and Subcategorization
The effect of prefixation on the subcategorization of the stems has recently been investigated in Aronoff（1976）， Roeper and Siegel（1978）， and Carlson and Roeper（1980）．15）The gist of the claims made in these studies is that置んεa（昭ゴー 置∫・η・ブPγ・βκe・孟・W・・d・（especially veγbs）伽∫漉・sπ6ca言e8碗za孟 ・πP・諭ゐ灘εS oプ孟んε望eγ680γ抗ewん01e oγd3 加％伽za7んεd ca5εs． Consider the following
（Carlson and Roeper 1980）：
（24）a． Bob told Eddie to leave．
a！D＊Bob retold Eddie to leave，
b・ JohρIikes to eat grapes in school．
b1． ＊John dislikes to eat grapes in school．16）
c． John set them free．
cノ．＊iohn mis−set them free．
d． Ibelieve him a fool．
d ．＊Idisbelieve him a fool．
Iwant to discuss a marked prefix in English whose characteristic deserves attention． Observe the following：
（25）a．The data in（2．30）represent one type of potential co％η♂eγexample オothe claim that permanent aspect may be used to define compounds：
（Judith N． Levi．1978． TんεSッηオaκa記Se？ηaη言ゴcs oプCo観pleκ！Voη乙ゴ・
ηals． Academic Press． p．43．）（Italics mine TN）
b．Although not all rules for complex verb formation are necessarily compatible， there is clear coπη護eγ・evidenceオo this claim， namely， the existence of verbs with two productive prefixes（e．g．7e（海8eη〃aηcん∫se，
γe｛1 se犯ga8e，騙sγe∂ηa1ッ2ε，撹η4er一γεe磁owed， etc．）（from the text of Carlson and Roeper（1980， p．133））（Italics mine TN）
These are familiar expressions found in papers in science， especiaUy in linguis・
tics． But孟o NP seems unable to be chosen as a complement if co観言eγ一is not attached to e劣a7πp♂e， e頭1e㏄e， and aγg㍑皿eπ言， as shown below．
（26）a．Next， let us examine arguments ＊to Brame s Base Hypothesis．
幽≠№≠撃獅唐煤@ 「
b．Next， let us examine counterarguments to Brame s Base
？against Hypothesis．
In other words， this is a case where prefixation 1セcdls orゴ㏄γea3εs subcategori・
zation possibilities， contrary to the examples in（24）．17）In order to describe the
84 Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
behavior of c側η孟eγ一and孟o．．．， at least the information from the prefix co襯孟εr一 is necessary．18）Therefore we should allow certainfeatures of a prefix to be per一 colated in 7πaヅんed cases， otherwise it would be very hard if not impossible to explain the occurrenee o臼o NP in such instances as in（25）and（26b）．
There is a similar phellomenon with respect to negative prefixes． It is well kll吐》wn that the addition of negative prefixes changes possible PP complements．
C｛）n8ider thεfonowin菖：
（27）Adjectives
a．deperldent on 。。． a・。どπdependent ＊on … o！
b．diff・・ent f−．．． b 謝iff…ent／＊fr・m…
1 孟o
｛二。・espectiveφ c ．ε・respective・ブ…
師at量e「1t ﾚith｝ 6 ♂°加patient番ith °
1bγ
（28）Verbs
a．aml＿ wlth＿ a ．｛1ど3arm＿ ＊with ．．．
oブ
b．burderL．． with＿ b1． dおburden．．． ＊with ＿
oブ
c．corlnect＿with＿ c・． dl 5connect．．． ＊with … ルo観 d．enchant．．． w ith＿ dノ．（1おenchant＿ ＊with ＿
oプ
It is Ilot correct to say that Jη一，珈っandゴγ一〇f themselves cause the change of possible PP c⑪mplements in（27）． For these negative prefixes do not have close connect｛on with any of the three PP s（i．e． o！NP，置o NP， and lbゲNP）． Need一 less to say， the stems alone cannot determine what type of PP the whole pre一 flxed adjectives take． It follows from these grounds thatわo孟んオんeεがoγ盟a琵oηoブ 孟ん。s置。・η8 a磁オゐ・孟・価e p・・加・s磁ec・∬aが・d・c漉曲a吻Pε・！PP諺ん・pγe一 伽edaのed加e5孟aκe．
As f⑪r the examples of verbs in（28）the relation between the prefix痂s・and ζhe PP complements oプNP andブγo？πNP is more obvious． But the in一 f⑪rmation of the stems also is involved， because it is not the case that all verbs with漉s−take oプNP orかo魏NP． For example， d∫s1漉εdoes not take oブNP or ハro観NP as subcategorized complements． This seems to have a bearing on the fact that 1∫んe itself does not take subcategorized PP complements． Whether the unprefixed verb takes幅言んNP as such a complement seems crucial．
Namiki：Affixation and the Notion Head of a Word 85
To summarize， at least the information of a prefix isnecessary to cover facts on co駕窺eγ一3 the information of both prefixes and stems is nβcessary to account for facts on negative prefixes（in・， im−， ir扁 in（27）and（充s−in（28））．
Let us turn， finally， to the mechanisms used in accounting for the correlation of prefixes and stems． First， I would like to divide word formation into Core Word Formation and Pbriphery Word Formation again with respect to feature percolation， as shown below．
（29）Feature Percolation
Core Word Formation through lexical heads Word Formation
Periphery Word Formation
through both lexical heads and prefixes Secondly， I suggest that features of prefixes be classified into two types（cf．
Convention H）：those which are percolated（marked）and those which are not
（unmarked）． An example of the former is one which specifies that co％π渉eγ一takes 置oNP after the stem， which may be formulated as ［＋ Nto NP］． An example of the latter is one which specifies what category a prefix can attach to， which may be formulated as，・ay，［＋［N，［A］．
Thirdly， I agree with Williams and Lieber aboutthe necessity of some feature percolation mechanism．19）How features of a prefix and a stem percolate is schematically shown below， though explicit formulation of the conventions is
1・ft・pen．2°）
（30）a． 十N
［十 NtoNP］
Core Word Formation Periphery 十N
Wordeormation
藍COUnter ・argument
［＋＿Nt・NP］ ［＋N］噛
［＋＿［N，［A，［V，鼠d。］
b． 十A
、 、 ！ ＼ 十A 、
^ 1 ＼
！ in dependent 11 ［十＿［N，［A］ ［一十A］ ，、、［十negativ6］ ［十 〇n NP］ 1
86 Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
It is clear that the information thatピ舷epe磁ε配takes oプNP as a complement cannot be obtained from the prefixどπ一alone or from the adjective depe煽eπ置alone．
Rather， it seems that the lnformation under consideration is gotten somehow by means of tlle computation of the meaning of the negative prefixごη一and of the in。
．
?盾窒高≠狽奄盾氏@that（1epe縦1e窺takes oηNP as a complement and that oブis， in a sense，
・PP・sit・t・・π一i・meaning。2L
§5 Summary
Tw⑪alternative systems in morphology， tわat is， Williams s asymmetric sys・
tem and Lieber s symmetric system were discussed in section 2． The evidence for Williams sbased on affixati ）n in English and Japanese was given and a dis一 tinction was made between Core Word Formatlon and Periphery Word Forma一 tion with respect to lexical heads in section 3． In section 4 several prefixes in English were discussed as examples that change subcategorizational possibilities of stems or whole prefixed words． In particular， the negative prefixes were claimed to be examples that necessitate the computation of the properties of both prefixes and stems． Tentative mechanisms of feature percolation were present．
ed to account for theα，rrelatlon of the prefixes and stems．
Notes
＊） Thisarticle is based on a paper presented at the regular meeting of the Tokyo Ellghsh I」inguistic Circle held at Sophia UIliversi㌻y， Tokyo， on September 19，1981．1 want to express my gratitude to Akira Ota， Masaru Kalita， Toshiて）Nakao， Yukio Otsu，
Masaru Nakamura， and Takao Yagi for invaluable comments and suggestions． Thanks are also due to Edward Quackenbush， Richard Krause， and Robert Boys， who have willingly and patlently served as informants， and especially to Edward Quackenbush for his sugges一 tions for improvements of style． Needless to say， any remaining inadequacies are my own．
1） Though Williams spaper was published in 1981，量t was written in 1979． Thus h量spaper preceded Lieber s 1980 dissertation，
2） It seems that this should be called the Righthand Head Co御eη置ゴoηor Pγ加c pZe rather than R掘e．
3） Since this information is irrelevant here， an account of it is omitted． See Lieber
（1980，section 1。1）for detalls．
唱
S） Subcategorization frames indicate the category of items to which they can attach．
For example， the prefix郡η一in English can attach to adjectives of all sorts， and will there一 fore have a lexical entry with subcategorization frame［A＿［A． In Lieber s system affixes differ from non−affix morphemes only in that they have frames indicating the category of
items to which they attach．
5） Insertion frames are syntactic frames into which lexical term丘nals can be inserted．
For example，孟んγow is specified for a two place argument structure． For details， see 1」ieber（1980， p．65）．
6） By stems Lleber means free morphemes ．
Namiki：Affixation and the Notion Head of a Word 87
7） This convention applies to affixes such as co槻オeγ一． Since co祝π陀γ一can attach to N，
V，and A（cf．（1b））， Lieber assumes that it lacks category features entirely． In fact， eo膨η言eγ一 can also attach to Adv， as in co槻孟eγclocん幅se， as shown Iater．
8） Lieber suse of 十 is hot consistent量n her examples．
9） There is some contradiction in Lieber s assumptions． She postulates the single context−free rewrite rule generating unlabeled branching tree structures（cf．（14a−d）and
（15a−b））． On the other hand， she mentions the first non−branching node dominating that
（i．e． a stem TN）morpheme in（16a）and sets up tree structures such as（17a−c）． But it is not poss量ble to generate such structures as（17a−c）according to the context−free rule in question． The rule can， as it st母nds， generate a tree structure such as（15a）but not one
such as（17a）． It is impossible to determine which is the outermost morpheme in（15a）．
Therefore， some modification of her assumptions is necessary to remedy this defect．
10） 0刎・was brought to my attention by Toshio Nakao． Cf． also Carlson and Roeper
（1980，p．131）， Nagashima（1980， p．265）， and Kunihiro（1980， p．291）． The existence 、
一
Zfan． example which derives V from A（1．e． O鋤tsma7°t）was pointed out to me by Edward Quackenbush．
11） This survey is not as detailed as孟n English． Unlike English， there are， to my ρ
knowledge， no exhaustive studies of word formation for Japanese．
12） Two other negatixe prefixesん凄一and祝鯵are cited as such in Nagashima（1980， p．
262）．But the examples he adduces（ん々°oshihina and mukankeina）are inappropriate， because 、 they can be regarded as adjectives derived from prefixed nouns（んむosんゴゐゴ，7π拡a殖ea Cf．
also Kanno（1964）and Nomura（1977）．
13） In fact the outputs are slightly different from adjectives in inflection though Similar in fUnCtiOn．
14） The lefthand compounds in Vietnamese are not unmarked in natural languageS in general， even though they may be in Vietnamese．
15） Previous studies are Ross（1975）and， according to Aronoff（1976）， Williams
（1973）．
16） This is mentioned in Hornby（1975， p． v）too．
17） Carlson and Roeper（1980， p．125， p．131）also note the cases where prefixation feeds subcategorization possibilities， but their examples are limited to verbs．
18） It seems that the fact that co駕η置eγas a free morpheme has a subcategor量zed置o NP complement has a bearing on the phenomenon in question．
19） Williams shead percolation mechanism is inadequate in that it allows features to percolate only through heads． Cf． Williams（1981，p．254）in this regard．
20） The reason why it should be considered that the feature［十＿Nto NP］of co槻置eγ一 is percolated concerns the problem of how the relative position is defined in general of an element with complements and the complements themselves． Cf． Gu6ron（1980， p．642）in this regard．
21） The possibility of the computation of the meaning of both elements was suggested ・
to me by Masaru Kajita， in a discussion about the interesting behavior of coωη言eγ一． It turns out later that this idea seems to hold more appropriately of the negative prefixes discussed above．
Carlson and Roeper（1980， p．126）note this independently： lt is a combination of the propert三es of the prefix and verb wh童ch determ量nes subcategorization． But they do not deal
88 BulL Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
with c側刎eγ一，adjectives with negative prefixes， or the fact that the｛痂一attached to verbs typically takes oズNP or唱ルo？πNP．
Appendix
【English】
（＊lt・1ics mean th・t th・affi・i・que・ti・n i・used・・b。th。，at。g。ry．changi。g。ffix and a n。n category−chang童ng aff量x．）
1．category−changing Prefixes
（A）those which can attach to three categories （N，V， and A）＝a・（1）
6 （B）th・se whi・h can att・・h…w・ca・・g・ries（N and A）一ゐ・誠・・晩and磁［V−
@ formin9］（4）
（C）those which attach to only one category（N）＝dε一［V−forming］（1）
皿．non。category−changing Prefixes
（A）those which can attach to four categories（N， A， V， Adv）＝counter−（1）
（B）those which can attach to three categories
N，A， V＝c童rcum−， co−，協5−， hyper−， inter・， mis−， non−， post㍉pre・， retro−，
14 sub・， super−， and trahs噂（13）
N，A， Adv；anti−（1）
72 （C）those which can attach to two categories
N，A ＝a−， ante・， auto−， bi−， crypto−， di−， hemi−， hypo−， in−， infra・， intra−，
meta−， mid・， mono−， multi・， neo・， paleo−， pan−， para−， peri−， poly−，
35 …2・・P・・3㍉pseud・一・semi−・・y臣，・・i・，・lt・a−， uni−， and・ice−（3・）
A，V ＝supra−（1）
N，V ＝contra−， fore㍉and ma豆一（3）
N，Adv＝no−（1）
（D）those which attach to only one category
N＝ana−， archi−， demi−， epi−， ex−， micro−， per−， pro 1−， proto−， step−， sur−，
A＝cis−， extra−， preter−， and un・（4）
V．＝6e−，｛∫eっeηっo％君一， re，， and un2−（6）
皿．category・changing suffixes
（A）those which can attach to four categories（N， V， P， Num）＝一 sh［A・formin副（1）
（B）those which can attach to two categories
A，V ＝−age，−ance（−ence），−ancy（−ency），−ard，−do㎜・，・（e）ry，−ice，−ment，
and−th〔N−formin9］（9）
66 N・V ＝−abl・ヂative・一ん！・−ive・−1・（−el）ヂ1…ド・・m・・and一ツ［A・f・rmi・g
N，A 鵠一ify，．（i）ty， and・ize［V−forming and N−formin創（3）
N，P ＝−ward（s）［Adv・forming］（1）
V．P −−i・g1［N・f・rmi。創（1）
A，Part＝−ness［N．formin9］（1）
（C）those which attach to only one category
Namiki：Affixation and the Notion Head of a Word 89
A−forming ＝．（i）ac，．al，一（i）an，−ary，−ate 1，−ed，．en1，−ern，．ese，
to N −esque，−foldヂic，−ile，−ineヂ舜e，−1ッ， and−ous（17）
V・f・rmi・g＝−at。1（1）
N−for㎡1ing ＝−acツ，−cy，一んood，−ing，一♂5飢，一♂5置，−let，−ling，−s配P，−ster，
一ton， and−（1）tude（12）42 t°A V−f。rmi。g＿−en・（、）
Adv−forming＝．ly（1）
N−forming ＝−al，−ant（−ent），−ation，・ee，−er， and−ure（6）t・Vく A−forming ＝−ory（1）
tく）P＝−most［A−and Adv−formin9］（1）
to Num＝−fold［A−and Adv．formin9］and−some2［A−forming］（2）
工V．non麟category齢changing suffixes
（A）those which attach to only one category
N 二君cy，−adeドageドarianドate 1ドate2ドcy，一ゴ07π， −een， −eryう 一ese，
一ess，−et，−ett・，−1冠，−h・・dヂi，一蚤ana，−i・ian，−i・（−y）…na，−i。。，−i。g1，
一ぎs肌，一♂3∫，一琵e，・let，−ling，−ness，・ock，一（e）re1， 一（e）ry， −3んど1ρ， −ster，
45 43 −teria， and−trix（36）
A ＝一♂5ん，−1ッヂmost，−so祝e， and−）1（5）
V ＝．1。1（1）
Adv＝−ly（1）
（B）others
一 Q〈．er（which attaches to N and nominal．eer（which attaches to N and［A N］）留P°unds）（1）
【Japanese】
（＊Underlining ind重cates that the affix in quest五〇n is used as both a category−changing affix and a non−category−changing affix．
＊＊Admeans a sort of adjective which differs from ordinary adjectives in inflection． 噛
Cf． note 13．）
1．category−changing Prefixes
those which attach to only one category
N＝4aあsんo−， hu［A−forming］（cf． note 13）
II．non・category−changing Prefixes ま
@ （A）those which can attach to three categories（N， A， Adv）＝真．（1）
（B）those which can attach to two categories 一
68 N，A ＝お一，ご一，き一，ど一（4）
6 A，V ＝け一（1）
A，Adv＝こ一（1）
＼（C）those whiφattach to only one category
こ L．仁う むお だい はっ うい
N＝小一，小一，大一，大一，非一，不一，無一，初一，初一，新一，旧一，現一，
前一，元一，名一，重一，軽一，全一，主一，準一，別一，稀一，同一，対一，
逆一，超一，未一，既一，有一，反一，過一，被一，急一，再一，最一，両一，61 す なま
片一，複一，総一，各一，諸一，素一，生一，故一，み一 （45）
A＝か一，いち一（2）
V＝相一，ず（ば）一，おし㍉うち㍉かき㍉きり㍉さし㍉とり一，
90 Bull． Fac． Educ．， Ibaraki Univ．（Hum．＆Soc． Sci．）1982（31）
つき一，つけ一，とび一，たち一，ひき一，まい一（14）
皿1．category−changing suffixes．
（A）those which can attach to two categories馳
転
S34ｭ：に部fl盤需。盤1調 n創（2）
（B）those which attach to only one category
A−forming ＝一的，一らしい，一い，一がましい（4）
V−forming ＝一めく，一る，一ばる，一ばむ，一じみる，一する，
toN 一する（一じる，一ずる）（6）
Ad−forming＝一げ，一やか，一だらけ，一風，一用（5）
Adv−forming＝一に，一上，一中，一ずつ，一ごと，一ながら（6）
／N−forming ＝一さ，一み（2）
Ad−forming＝一らか，一だ，一です（3）
．／Nf⊂，rming 謝一ぶり，一かた （2） ！り ／to Vくく一一A・forming 二・しい （1） ＼ Adv−f〔，rming＝一っっ，一たけ（一だけ） （2）
・・N・m〈N−forming ＝一次，一代，一回（3̀dv−forming＝一回 （1））
IV． non−category−changing suffixes
（A）there are no suffixes that can attach to more than one category
（B）those which attach to on監y one category
か にん じん
N ＝一家，一士，一員，一者，一手，一人，一人，一師，一婦，一役，一番，
一族，一団，一派，一屋，一店，一料，一費，一賃，一さま，一さん，一君，
一ちゃん，一たち，一ら，一ども，一がた，一嬢，一殿，一度，一気，
一権，一性，一化，一質，一状，一論，一率， 一分，，一物，
68@ 一系，一業，一け，一剤，一だて，一下，一こみ，二串う，二革う．（49）
@ ＼v−．こむ，．あがるドあげるド、、るドいれる，一おとすド肋る，
一かける，一かねる，一きる，一たてる，一だす，一とおす，一うる，
一すぎる，一すごす，一つく，一っぶす（18）
Adv＝一と （1）
References
Allen， Margaret．1978a． The Morphology of Negative Prefixes in English．
NEL S 8， pp．1−11．
．1978b． Moη）ん010gゴcal l翫ve5泥gaオゴoηs． Unpublished Ph． D． disser・
tation， University of Connecticut．
Namiki：Affixation and the Notion Head of a Word 91
Aronoff， Mark．1976．Wo併d Fb襯a孟∫oη∫πGeηeγa伽e Gγa伽ηaγ． Linguistic Inquiry Monograph 1，MIT Press， Cambridge， Massachusetts．
Carlson， Greg， and Thomas Roeper．1980． Morphology and Subcategorization：
Case and the Unmarked Complex Verb． Le短cal Gγa伽ηaγ， ed． by Teun Hoekstra， Harry van der Hul合t， and Michael Moortgat， Foris，
Dordrecht， pp．123−64．
Chomsky， Noam， and Howard Lasnik．1977． Filters and Control． L∫㎎伽∫ゴc
噛㎎π∫賀y，8，pp．425−504．
Gu6ron， Jacqueline．1980． On the Syntax and Semantics of PP Extraposition．
L診㎎駕翻c地厩η，11，pp．637−78．
Halle， Morris．1973． Prolegomena to a Theory of Word Formation． Lゴ㎎窃s孟ゴc 1㎎郡ゴγッ，4，pp．3脚16．
Hornby， A． S．1975． G伽de o Paπeγηs a掘σ6agε加Eηg1ゴ5ん． Second Edition．
Oxford University Press， London．
Jespersen， Otto．1942．！1 ModeγηE㎎1ゴsんGγa観況aγoηH∫s♂oγゴcal Pγ珈c∫ples． Part
VI．George Allen＆Unwin， London．
Kanno， Hiroshi．1964． Settogo， Setsubigo（＝Prefixes and Suffixes）． Kogo一 6駕ηpoηo Mo捌aゴ診eη． Koza Gendaigo 6， Meiji Shoin， Tokyo， pp．231・47．
Kunihiro， Tetsuya．1980． Henla Hosetsu（＝ Supplement by the Editor）． 覧
0π5e∫言oκε泥aゴ． Nich i eigohikakukoza 1，Taishukan， Tokyo， pp．287・96．
Lieber， Rochelle．1980．0η跳e Oゲgaηゴ2a孟∫oηoプ♂んe Le娚coη． Unpublished Ph．
D．dissertation， MIT．
Mafchand， Hans．1969． TんεCaJcgoγゴεs aηd 7汐peεoプPγeseπ置一DaッE㎎1∫3んWoπ」一 Fbr7πa言ゴoη． Second Edition． C． H． Beck， MUnchen．
Nagashima， Yoshio．1980． Gokosei no Hikaku（＝Contrasts in Word For・
mation）． 0η5eπo Ke泥a ． Nichieigohikakukoza 1，Taishukan， Tokyo， pp．
235・85．
Nomura， Masaaki．1977． Zogoho（＝Word Formation）． Goゴ置o l㎞∫． Iwanami・
koza Nihongo 9， Iwanami Shoten， Tokyo， pp．245−84．
Roeper， Thomas， and Muffy Siegel．1978． A Lexical Transformation for Verbal Compounds． LJ㎎ω∫ε琵c地厩η，9pp．199・260．
Ross， John Robert．1975． Wording up． Unpublished paper．
Selkirk， Elizabeth． forthcoming． Pんoη0108ッaηd Sッη置aκ， Tんe Rela言ゴoπ6e伽εeη Soω磁aηd S碗c加γe． MIT Press， Camdridge， Massachusetts．
Siegel， Dorothy．1974． Top∫c5♂πEπgljsんMoη）ん010＆y． Ph． D． dissertation， MIT．
Published by Garland in 1979．
．1978． The Adjacency Condition and the Theory of Morphology．
NEL S 8， pp．189・97．
Williams， Edwin．1973． γe andゐacん． Unpublished paper， MIT．
．1981． On the Notions Lexically Related and Head of a Word ．
g2 Bull． Fac． Educ．， Ibaraki Univ，（Hum．＆Soc． Sci．）1982（31）
Lゴηg祝ご8だc1㎎駕ぎγッ，12， pp．245・74．
、 ．
