lflfGUf$f0$ votume rr /72 Is?

e r7tz2s/221
In this issue
D. A , CRUS E Revercves RI CHA RD A T E S CO Bulkrngt morpholexcaj a act G E E RT . B O O IJ E Semantic regularities word formaton in R. M , W . DI X O N Corections ,ls and comments concerning Heath,s Dyir6al ergative?, MA-RTHA DANLy and WtLL|AM E. COOPER Sentence productjoni Closure versus Inrtation constituents of EIRLYS DAVIES E. Somerestrictions condttronal on rmperatives Short note W . L. M A G NUSS O N Chronostructural ofientation rool panern on development Review article BE RNA RD MR IE CO The languages Nicronesa of Book revews DE S M O ND D E R B YS H IR E C. Caib-_speakng lndans. Culturc, Societyand Lngudge, edited by E. B. Basso TH. R. VON STOCKERT Aphasie,by G. peuser SZABOZOLTAN St stk, by B. Sandig Research project Thesyntax, semntcs pragmatics the comparative and ol construction Publicatonsreceived Author index to Lngusticsvalume f7 (197g) l

967

'1003 1017 1039

' 1055

1057

1073 107 4 1076

10gl

1085 1093

ISSN @24-39K)

mouton publshers.the hague . pads .

new york

Semantic regularties in word formation

E. GEERT BO O I J

Abslracl the This paper discusses problem how to account Jbr regularities in the the interjritation of complex words. It will be shown th(lt in some cases detetmined by its nteipretaton of a complex word is not completely mor)hologcal structure, b is ;he resuh of an nteractonbetweenlnguistic 'progof structure and non-ltnguistic information. The consequences this processesforthe theory matic'approach to the semanticsofwotdformaton of morphology will be discussed. l. Introduction

ftir
I

One of the salient features of natual languages is that they show structural parallelisms betweenform and meaning on two levels,the level Therefore, an adequategrammar ofthe word and the level ofthc sentence. ofa languagemust specifyhow and how far (i) the syntactic structure of a sentence determines the interpretation of that sentence, and (ii) the morphological structure of a complex word determinesthc interprtation of that complex wod. I will call a representation of the structually determined semantic aspects of a linguistic expression (a sentence, a phrase or a word) the logical form of that expression.r The structures of complex words are specified by Word Formation Rutes(WFRs, cf. Aronoff, 1976)that createcomplex words from existing or possible words (cfl Booij, 1977; Roeper and Siegel, 1978)'z For xample, the following two WFRs can be assumedfor Dutch: (1) [xh ' [be * [xh]v (2) [x]a --+[[x]" +iteitl' where x stands for a sequenceof phonological segmentsand boundaries' Theserules create e.g. bedjk 'lo provide with a dike' lrom dijk'dike' and fom aad'absurd'. absurdtet'absurdity'
{985 $2.00 0024-3949/79l@17 The @ Mouton Publishrs, Hague 985 1001. 17 Linguistics (1979),

986

ceert E. Booij Semantic regulaities in word formation 987

assoclated with a boundarv svt

ld, {
,i).,

sutnxes-ror1, -menr and-, ru!,'"o

orthe word structure in which o""i,,1"r it [1"'ifr'1,-l],iiiillfss correrate.wFRs ,"l"";:ll#i "#:iX|l,t3iltt ""emeaning Sme out
by^ Jackendoff the English for

word. Note thar the claim tnat W.nscontrir," i"irr^#""""i.r, correctlyimpliesthar the bound " morpheme_has no :-r::1":::.0, meaning: onecannotsayanythingab.., ,f* 1:o:O^:1-o*, _."ni* ,f a"

can-beclrcumscibed as ,to provrde with , where is the meaning of.the base

meaning contributi;;;"ilffffi*il:

fiom Dutch. The structural parallelism br fom and.meaning of complex wordsrmplies that th" a compositionalrun"tion or-til Tomplex word is'-atleastin principle, -.uninn.jll*

.+';cr. uouoau,v soolj'f't;i;1:"*,J",?;i.::.lli,H# {",,,;'":: lwo.types boundaries means of by of data

:::,:,i[,#:ffi *l;

;'#t"i'EiJ'T::"1,1""fi j;";;"',i;,*. a,."''.r",* l:"1::.:'-'':l*.-*l ,, i...-il;.,'J,:1 1""#ii;i.il,",iiij,i1iT,,L,Jl1[,,,.;i

rne arrows. parr thesrrucrures are of mention.a rrr".igi,iiini)...*r. to formularion theWFRsembodies of rhewell_known Tli:]ntt hypothesis

rhe rrrand ,u:s ;:::la;'Ji;ji1 :Hl*,':*"T[','.? : il:1:t:;

oe77),b;;;J;;i,"ff;;","T?;:#?i";i:i,,%li*li,?l is the lexical functional structure of to hit.It 'NPl HIT NPr' that
this doesn01necessarilv mr

WFRs are some times called

speciliesthe relation betwten syntactic and logical form. NPt indicates the grammatical function of subject, and NPr indicates the grammatical function of object. The grammatical function of a constituent is defined configurationally, i.e. by its position in syntactic structure. Fo instance, the NP that is dominated diectly by the node S has the function of subject, and the NP that is dominated directly by the VP node, has the function of object. The syntactic context in which a verb can occur, is indicated by a NP] that feature, in the case of lril the feature [subcategoYization means that the verb i1 can occur before a sister-NP. Bresnan (19?8: 21) remarks that 'the syntactic contexts appear to be redundant-that is, predictable fom the functional structures'. However, she does not formulate such redundancy rules. Furthermore, Bresnan proposes that (optionally) transitive verbs such as to eat have two specificationsin the lexicon: (4) ear, Y (D l(ii) t-] NPI NP, EAT NP, (3Y) NP' EAT Y

correra,es. and impricario-ns-;;;i:'";;;5::'i:,"i#rJl?,"j,nj,.i's rhe

2. The smanticcorrelt of the agentive suffix _er

section 3, I * l 1t""* r"a"@lng

:J.i:f, il:i: :it;'.':,T.1*s,i:l,J*."':'**:*"$i".#i::il,.#

"niy".;;;r##;:':i:1,].ntt:
corelate ca', be accounted for' In

wrn, r'"""

650)Moreover, some .

That is, the Iogical object variable of ro eal can be bound by an existential quantifier. Consequently, the semantic representation of a sentencewith to eat wlll be well formed, although there is no specified direct object, because there are no free variables. Variables always need to be interpreted in order to get well-formed expressions(Tarski, 1964:77). This can be done by (i) assigningthe value ofa constant to that variable, or (ii) binding the variable by means of a quantifier, or (iii) binding the variable by means of a sentential function. Th interpretation of a sentence such as Jolz hts Mary with the structure [pohn]p [[hits]v [Maryhp]vpls is now deived as follows: first, we apply the configurational definitions of the grammatical functions to the NPs in this sentence: John : sublect,Mary: object, and assignindices to these NPs. Secondly, we substitute these indices into the approprlate argument positions in the lexical functional structure of the verb: (5) a. NP, HIT NP, NP,: John : i NPr; Mary : j b. John : i, Mary : j i H ITJ (5ii) is the interpretation of the sentenceJohn hits Mary.a) . The lexical functional structure of the verb o hit car' 6e seen as an abbreviation for:

y r"rran iiiai]r,h" l?l'..:-t-nrt' she assumes u".b u,,1, rhar ru"h a

.;.n"::, r,.,.un oii;;' i :JJ# il :lTJ::fi r l." 4).rhisn g, is""i,i ;'J J:",[l: hyporhesis worked,*
il;;;, i"ii"*,
Npl Npl HrT Np,

q*lmar it is often assumedrht the meaning of 3.*::":ilty" a sentenca ls the poduct of the intactior

i,#i:i,lji

(3) ht, v, t_

988 Geert E. Booij ( 6) x HI T y & x : N P r & y : N p , This reformulation of the lexical functional stuctre of lo lil is useful i.or the formulation of the meaning correlates of word. structures. 2.I

Semantcregularilies in word formation

989

That is, agent nouns such as koper and bewerker both have two logical forms:

(10) toper,N,
bewerker,N,

(i) x[x BUY y & y:NP,] (ii) x[] y [x BUY y]l

(i) xlx ADAPT Y & Y: NP.] (ii) x[] y [x ADAPT y]l

The togicatform of deverbalugenr nouns

Let us now, given the theoretical framework above, turn to a specifictype ofword constuction, the fomation of deverbal agent nouns in Dutch, rn order to illustrate how we can account fo the structurally detemined semantlc corrlate of a word construction. A representative list of deverbal agent nouns is given in (7): (7) lntransitive wrb: zwem (to swim) fiets (to cycle) Transtive ye: koop (to buy) bewerk (to adapt) vertel (ro tell) eet (to eat) zwemmer (swrmmer) fietser (cyclist) koper (buyer) bewerker (adapter) verteller (narrator) ete (eater)

According to Jackendofrs cross-categorial definitions of 'grammatical object'(: NPr) (Jackendoff, 1971 '7l-72), (van) dit hus and (ran) dt boek in (9) will qualify as the grammatical objects of koper and bewerker respectively. Apparently, the logical form (8) must be reformulated as follows: (8)' by x [sententialfunction expressed the verb and value assignmentof

vl
The-syntactic transitivity of these gent nouns can now be predicted by a redundancy rule: (l l) NP] la rule that is also valid for transitive verbs. the This analysisof agent nouns conectly expresses fact that thcre is na automatic, complete corespondencebetween a verb and its agent noun with respect to subcategorization featues, as would be expected rn a transformational account of this word formation process:the transfer of explicitly by (8)'. Note, transitivity from the verb to the noun is expressed for instance, that although Lhe verb zwem (to swim) can occur with a directional PP, the corresponding agent noun cannot: (12) a. b. Degenendie zwemmennaar Giethoorn... Those who swim to Giethooru... *De zwemmers naar Giethoorn. The swimmers to Giethoorn. [v : NPJ

The noun must be assumed to have the morphological structure [[x]u 1erlr. The meaning coelate of this word strutture ian U, lir"u._ scribed as'soneone who Vs,. That is, the logical form ofa deverbal agenr noun ls: (8) x [sentential function expressedby the verb]

O-nc have added the logical correlate ofa detemine to an we expression ol this type, such an expression is a well-fomed interpreted urg"*"n,, srncethe free vaiable x is bound by a sentential function. The logiJal form o,f e.g. zwemmeris 'x[x SWIM]'. The value assignmentof the vaiiable rn the functional stucture of the yerb, 'x: Npri is not transfered lo rne corresponding agent noun, becausex is already bound by the sentential function. The (optional) transitivity of verbs is transfered to their corresponding agent nouns, fo istance: (9) a. Jan is de koper (van dit hursl. John is the buyer (of this house). b. Jan is de bewerke (van dit boek). J o h n j s th e a d a p re r{ o f rh i s b o o k ;.

Of course, a transformationalist theoy of word formation could describe these differencesbetween verbs and agent nouns by adding conditions to the relevant transformation, but it does not give an explanation fo these differences.On the other hand, the lexicalist theory implies that nearly every correspondencemust be stated explicitly, and so we do not expect that, for instance,agent nouns can be followed by a directional PP. Only the intransitivity of agent nouns derived fom intransitive verbs is predicted, by rule (11): since there is no logical argument _t' present, rule (l l) cannot apply, and consequently the subcateorization feature is not assignedto these agent nouns. t-NPl

990 Geert E. Booij . The properties ofdeverbal

Semantic regul ities n word formation

991

j :1""'"1''#iT;;i:d:',:,i{"f,"i'J":,g:';;i::TT,i:x'"Ji:
3. Senanticindeteminacis in word constructions

$ll ,!),

::il';""fi:Tii1l: ;":";i,H*;J'.."::ffi 'H:llLiliT::'-":ff

The same semantic indetermi

jil.jfflil +ln:n;*_,:x;"filT:ttiliF.tli:j#fii. ff$:T,ii::[i:i:[*$m:i:d:l*.,",*:""1]!H'"ili,,ii: semanric erations b;**;;il:',:dT# jl;;x,rfi ;:^',*:^1"i.:;"ff:;:*; :.*ltiilll.;!uLl*:j,.J ::"*:r:i,1,'" n*1,:t:
sentences. Every sentence has a c

llfill*i"T#T"i;l|I

il?.-lll

assumed inrerprerarion that the ora

(13) a. a rural policeman : a policeman who exets his function in a rutal area b. a chemical engineer: an engineer who occupies himself with chemical processes : a violinist who plays the violin badl., c. a bad violinist : a father who takes care of his duties as a d. a good father father : a student who is eager to study e. an eager student f. a historical linguist: a linguist who studies the history of languages The head nouns in these NPs express a function, and the adjective somehow modifies this function. So the following regularity can be observed: (14) Given a noun phrase in which an adjective modifies the referenceof the head noun that expressesa function, the adjective must be interprcted as modifying somehow the function expressedby the noun. Note, however, that this rule does not predict the exact semantic interpretation of the NPs because the nature of lhe somehow ts not determined by rule. Miller (1978) calls such rules construal rules. He proposes the following construal rule fo NPs with evaluative adjectives such as good, excellent, bad and av)ful: modifres Give a noun phrase which a positive(negative) in evaluative adjective as that theentitydenoted by thenoun,thecombination to beconstrued meaning is (lesser) degree thoseproperties requiedfor the noun hasin greate than average (Miller, 1978:104). activity,useor appearance suchobjects. of the expectd

:*;f n:f:* iT;:ffi**h#il,lfiliT:T":t,Tl il: "';,"ffi ffii.:,:J::'J;J:T.'"1,:.jff [:'ifl,,,i; $1T,[XH;P;;;

'enemy',groenig 'greenish. fom gro

rs :;rruj,'*#i:,!T,J::^:;;; F, ffi,fr w:, :tr :^;'t::!:::'i;:#;n:?il#:{di:'"'1"""'?1l6:.""'J#:r

3.1. Compound.s

ji :t''(*,:!x:#J:{i:ir:{::j::::j{{'*iq;::,?f iTii j:; :;:t:1,:"j;;ll:::;;:;i:::::;:;:::tr:"i::!;"Yi^"",z#

properties, or .;;;:;; ihe fiiiffii:::T nungtl policeman is a policeman ';hJfJ;i'ii'::Ti',T:.$B:

;*:H:[!a*i[il*i:t#,#"":::xi,"iJ'"i#,:iir;t]":,::

t:!-iffi T"i".Hrulf i1""#:!.Til3lT.f ;5;;3:,'.":l:;;#; goena r

ctis'greesr,' h i,.'" fl illJ,ffi:.1;15i; ;:,lk;i ".,",':ift ,",'l.lLl"11l',i'l? ;l

It will be clear now, that such construal rules do not specify structurally determined maning corrlates,but regularities in the interaction between linguistic structure and non-linguistic knowledge, i.e. pragmatic regularities. In the next subsections,it will be shown that the sameapplies to several kinds of word constructions: the regularities in the interpretation of compounds, denominal adjectives and denominal verbs are also pragmatic ones, which has important consequences morphological theory. for

ilJ'i.,i.*T:.;iliifi:'',".#iT,.Trftl#_ *:

The logical form of compounds, i.e. of complex\ords with the structure [Y + XJx,, where X and Y are variables for lexical categories,can be circumscribed as:

_ ".r ,,

..

EOOT J

co,,pound derermi"., pkr or . i.,i,,,]-ll'j ^Tlt8"'v :f the ,econd r5)craimsrhar,n.,n."",""i"..l".,"Ti.rr".Jro,Jlinr...,T[::li*:*,i?

i$";:t*.::ili.effij:"l ;::fiH::::::i;

(15) Xr that has somerelat

Semantic regulaties in word formation

993

#J::i;":,,T'';"!,|;: rra r-o n,s

tl. '-t;"'tr'1,ffi l:iiiiltffi::lil,Xfi +r,l',,'ru* ;;::;;:1,'f Xl'.l::.,rfi i[ i*f:iT.,.J,ilH:ft

rm io s .. ; ;,,,;" j.1? ;tL';' ; j"T:":t, ;i; a, n ,::J e.a T"L b,y

ri*;:ttri#jilnr***;:itj"t,f ,t::,i,,:,."{.l;ff T #_T,il,1,i;:ffill';" ii::

de,e,abre

j,;:ri:JT,"l1?:ilJi{il:; jiili":iTi*fi*:lllj5l ..JH

co bined i ;;: il:il::tr'fl',TJ; ll'i" I are m in,u n"i "onr ",

transformations(cf Levi, l9?8: 139) A secondargument ot syntactic Le\r's is lhe following: in eglaities lhe of the a Moreover, theorywhichdenes existence fundamental freely and could not explain the fact that speakers of lcompounds] formation frequentlycreate novel [compounds]. . without having to provide explicit as iinguisticcreationslf therewereindeed little spontaneous fo denitios these have linguists as by expressed [compounds] some to theeiationships systematicity would situations in and readers nomal communicative both suggested, listeners with which to interpetsuchunfamilirforms (Levi' hai! no intenalresources 1978:53). This is, however, no valid argument for the claim that there is linguistic systematicity in the meaning relaton between the two parts of a compound. On ihe contrary, the productivity of compound formation can be process, ixplained by the great semanticversatility ofthis word formation so many meaning relations between two words can be expressedin since this way. Of course, a native speaker is normally very well able to but determi;e the meaning relationbetween the two parts ofa compound, knowledge that does not prove that this ability is due to /ig,,ric Thirdly, Levi's theory predicts that many, logically possible meaning imdations between the two parts of a compound are linguistically However, such impossible meanpossiblefor newly created compounds. ing relations may, according to Levi, be found in existing comporrnds that are no loriger productive: formed by means of processes . there is a small set of [compounds]. which seemto have been For example, fneejust to Nr'; this set includes formedon the pattern of 'N, which reaches (:'the upperpart of a balcony') That ad wt:tislcoat, breast-ail hip-boots, pants, in by maybe seen the impossibility contemporary ihis patternis now unpoductive hair fo suchds *calf skirf, ceili g vy, or shoulder Engiishof coiningnew forms samerelationship(Levi 1978:55). the express This argument has been refuted by Downing (19'/'1:82'7-828),in reply to some pre.niouspublications of Levi's. Downing convincingly shows that, altho;gh the meaning relations expressed by Levi's RDPs ae rathe frequeit in compounds, they do not exhaust the list of possible meaning relations, For example, a toe web is tnterpretedby some native speakersas 'web between toes' a cowtree as a 'tree that cows like to rub up against' , adan eggbrdas a'bird that stealsother birds' eggs' Downing therefore concludJ that the only constraints on compound formation are that, preferably, the meaning relation must be interpretable given a certain iontext or common knowledge of speaker and addressee, that the information provided by the first part of the compound should not be compltely redundant (as is the case in e g. wind-flag), and that for a

,,.,*,,m***"4rpit{,iii,,,,".+,l:;;,iffi li,iffi .tr1i,;:.,;f,f ;l:: ilj:,;:l: ll. l;Jj.i:

il11f.? :: H;J::ilJ Levi.srLri'otnn neracr with (in

j"ilT":r iglfLi:.["iir#..*il*:ff i *"li..{ ffi ll"i;i]i'', ff i":1}ii iJ: ijiJ: *..,",'l

iilT[Ti: T'i.',n,',,,.

.l;;", '.e,,"," .lT:::ff "';l;:;:;.ffi ;::,ith:i;.,fr, ".,,.*

Levi adducesthe followino,,,

tri.'i,'''.: iir.'';l?11iiffi *::J':i+ new m :ill' i'-',ff Hffi"J' co :'ill#:tt **1;; ifiJ

jliri"y.t'*,'":l',',1.m1t.,,"* :Uill;i;:lily.ti

ifi8ff X?tl;,.t:i:^. ;",.i,iil..j;;:, :i

tt+

ceerf E. Boo Semantic regularities in word formation 995

"i".,noi,,li,j t'" increase rto irr ," f.'il1'Jn:'li,;ili,,i:l5l'ritv

r.un.ro,",nrtionulLt ;;;;;:;:::'-19iinst,
Apart from the specific obiectio

j:*::l'l HHa*ii"?i;,:#ri,'l"di''it''?;;:;:fl j*ff1ffi r"r.y,+rlnni+ o".etricauyff lrxl#;bxli:xlll opposed in",o*",.,'. ilili'ffb,li *un,n"
lharil *ilj U. l*""ir.l]iil"-ay
(17) a. b. c.

i'e,i,{:i,"lT,:5Jff Hi'i:,:I.'[f,riff .,.,.".n. ['"'.:^i::;;::; ii.'iunrl'i,',,, ffll;:il:';l#"T3::i''3:fH:: uu,,on,

leltctously used. be anrl-rr," ,,kelihood

tli.r,

meaning ofits basenoun to the meaning of the head noun of the NP. For instance, the phrase gemeenteljke verorclening 'municipal byJaw' is semanticallyequivalent to the compound gemeenteverorderlng'municipal g bylaw' with the morphological structufe lfgemeentelN lverordennglyl. Some other example ae: [[bestuur] lijkel functie 'executivefunction' (bestuur: boad) 'professorial absent[[professor]"alelo verstrooidheid mindedness' l[politih onele]Aactie'police action'

*'' "::;lr ' ::,''';, nji:ii:{{lii,fifl"

Levi's theorv raised above. any

1,.'.::d:-:o**il,i-."",';i:ilffi :ii"Tii:,T[-J:il.:.J; i'J',:ffi g ;:**f* :r:l ff *tur'it#fi:':l'fi1Jlfiff if"iT:i: *.d;il::i#:.illi.il,:l;:: ,of i,;""'".''. #"1'J
here connrms arso rhe rouowing ",# oT3tl,i^Tr"?illl$,1"'"'o"d rire ;i:ili:1f,*Tff:';11il*:."iTil:nin ear invorves the inreacrion or
Carol and Tanenhaus(1975: 5

'1"":.!,:ff j:T:f ::' f,Jilil;,.j1',1',.^1'.?Tiji:l* :;j:i"1: ,-:'T

The semanticequivalenceofcompounds and theseNPs was also observed by Levi (1978) who uses this observation to defend the hypothesis that NPs such as those in (17) and the corresponding compounds have common underlying syntactic structures. However, in the preceding section we concluded that compounds should not be deived from s/rxtdcficstructurs. Therefore, we must assumethat reference-modifying adjectives sucb as bestuurlijk, professoraal and politoneel are derived by means of Word Formation Rules. The structure of such complex adjectivesdetermines the interyretation of NPs in the following way: (18) In a structure [(Det)[[Nr] suffrxlANrlp, ifA is interpreted referencemodifyingly, the NP must be interpreted as 'N2 that beas some relation R to N1'. So neither the interpretation ofthese types ofcomplex adjectives,nor that of the NPs in which they occur, is completely structurally determined. Their interpretation is, like that ofcompounds, the result ofan interaction betwen linguistic structure and nonlinguistic knowledge, since nonlinguistic knowledge determines the nature of relation R.

;::".T.'tilJ,iltf fi ,il#j':;l**iti;;t+iidJitH",,"
3.2. Denominal adjectives

il.-J ii:, :i,'j::: l:..:"'0.y :;,. :iil'..il,",l.o::il:".:]i:ji,:il :,1:

3.3. Denominal verbs Both Dutch and English have a very productive word formation process for the derivation of vebs fom nouns without adding phonological material, i.e. by means of so-called implicit transposition. The following Dutch examples illustrate this process:

(19) a.

j::"ffi #;j::l;J: ;H::3i1'J,.,';.:,J;i,1".#r*il."n:.;;m:

l$lll1'.tj.iil-*^ :'*':i-.ldenc;ll .inrernrerar ween wh ionber Nps r
I

N tafel'table' zout'salt' klei 'clay' water 'water,urine'

tafel 'to haYea meal' zout 'to salt' klei 'to play with clay' water 'to urinate'

996 Geert E. Booj Semantic regularities in word Jarmaton 997

(le) b.

, coined trythar no .;";;;'iliil;::1"r",,#the LT"T?:::H","r?,11iolll1: words recen l:,:g foinsrancefolowing are t ;XT;
N d i e e t' d i e t' bankier'baqker' lift lift' carnaval 'carnival, cche'ceche, lek 'leak' sloelbak'shovel-board' caravan .caravan, servtce'service' dieet 'to be on a diet, bankier 'to handie banking affairs, rrr|go by Iift. carnlval 'to celebrate carnival, creche 'to take care of a creche, leK to provjde secreinformation. sJoetbak'to play with a shovelooad caravan'to stay in a caravan'

assumc someplausiblcrclation between verb and its basenoun that is the in accordance with his general and situation-bound knowledge and the context in which the verb is used. Rose claims that with respect to denominal vebs four rgular basic relationships to their base nouns must be distinguished: (i) CAUSE N GO TO object (e.g.water the lawn) (iD CAUSE N COME FROM object (e.g.peel an orange) (iii) CAUSE object BE + LOC N (e.g.crate books) (iv) BE (LIKE) N TO object (e.9. mother a child) These semantic relations are very vaguely formulated, however, and therefore extralinguistic knowledge has to play a role anyway. Moreover, many denominal verbs are not adequately characteizedby one of these four relationships,e.g. to hammer, to bke, to buon, to campaign efc. (cf. Marchand, 1969:368-371for many other examples). The sameholds for Dutch denominal verbs, witness the vebs die,carnaval,cftche, sjoelbak and servcelisted in (19b). Apparently, the only regularity that can be observed is that the verb expresses an activity in which the entity mentioned the noun playsa role. But the exactnatureof this role is noL by determined by linguistic principles. The interpretation of such denominal verbs dependson (i) our systemsof knowledge and belief, and (ii) on the contextin which the verb is used.De Vries (1975:163),for instance, points out with respectto (i) that the difference betwcen rs,to live in a house a'].d,beikl'to chisel', derived from i 'house' and eiel .chisel, respectively,is causedby the fact that housesare not used as instruments, rn contrast to chisels,and with respectto (ii) that the verb punt in Ik punt eenpotlood'I sharpena pencil' differs in meaning from pant in Ik punt mijn haar'I remoye dead ends lrom my hair'. The data reported in Clark (1979),Clark and Clark (1979) and Karius (1977) also show that the interpretation of denominal verbs depends on geDeral,cultural knowledge, situation-bound knowledge and/or the context. Miller (1978: 104) suggeststhat certain regularities with respect to th meaningsofthese denominal verbs (or, as he puts it, 'nouns usedas verbs') do exist: 'When nouns of type M are used as verbs, the meaning of ,r M.r )., is to be construed as 'r; covers th surfaceof with M'. Type M nouns are, according to Miller (a subcategory ofl) the mass nouns. For instance, given the mass rroun painf, the yerb to pqint mans 'to cover with paint,. Miller admits, howver, that 'Without an independent critedon ro determine which nouns ae of type M, the rule is relatively weak' (Miller, 1978:104-105).However,it appearsthat the dennitionof type M nouns involves knowledge of the conventional use of nouns: .It is difficulr Lo characterizemost type M nouns without describing their use for covering

,{galn. the question turns completely derermineo by ttre ml

words^dejH .|l,;", * :::"*," X:lI :f*l*, and ClarkEnglish ( tark ( t978) :r:ew and Clark (

noun rhemeaniog and conrrburion rheword,..r"iXTi*Jf tnt base of

t denominar,".brn" ^l-ll,^ojevious ",o j* ".;;;;;;;;iltrJ::T,,j?,lJjffi"::iJ:T:f ::r::
In other goupsthe semantic relatiol srammaticallv denable. derive.agreaimany v;;;;;;;::1i: Thus,we "t of vrjouschanges, zero-element, te .";;;;;;;;;hl"t^ includinga bur il,means

1979J. whethe the meaning of the verb is

compounds. '"lx:"1;'::3] Bloomn",o

:i""':""'::1;'n'J:'l'

to this probrem possibre' are

section with respectto

hand. Rose (1973) craims that rhere is a restrictedset 3:":*^::!:l possr semantic bte of reiations betwei basenoun' and in support this claim h. ,dd,;;;';;; or ""d 'gument that Levi (1978) ,;;;;::.::* men;;s with cspecrto compounds:

:;;:r;x:;i:l",li:iii,in:i;iili!:'xr,;,,.:ffi Lil;tfi,r;tr

Oneobvious explanation thegene for aitenarives seveiil;i,;;: is ae ':':,:pretabilitv.ofinovatiors that the il;Tl n de ired formation ways. se!en miShtbe rwo or ways. o;..";j.;bi;';'r'.:-igrve ootrous: ir surely but noLbe innnirer) musr ;;;,*;;i"* "* i ;;;"j;:,:",' 3:j l:::'::-0,*"rion formarion\ is absud urrurn" u,". to ttut llll-1jlr,waysambiguous. even ", lnd it is gen nundred The "iri,0" "

dl:

of the generat rerprera rn bitiry

of new

can .,,,ui,"- simpry ft':ril':H:y:l,iil."iff Tt::ff'"ii

eeE Gee E. Booii

4
4. Conclusions

t hinB S(Mi l e r. 9 7 g : 1 0 5 1 . T h a t i s . rh eei s no pure/y l i ngui strc 01 lyp M nouns.Moreove, dehnrrron nol noun 'usedas a veb'is used in aciordance*t,h Mt;;;;i.:-l::v Tutt rule' witnessthe examples in (20b): 'trual (20) a. * verf.paint, verf'to Paint' olie ,oil, o l i e ' t o oi | l e e r.ta , teer 'to ta' yernis .varnish,

Semantcresulartesin wtrd fcrmalion 999

b k";;:";;*,., ;:::':,;'.:1.J:fl",", klei.'clay, .t"

melk .milk, koek_.cake, modder.mud,

flr;
rii"

fh,

mett ,t mi .,-'-' -'*, koek .to coagulate, modder .to fr""""iro*Ul", Although the nounsin (20b) are nouns' and could in princinle used be .to cove *,n"ririlrl",rr" . -ass verbs ao not have the meaning'tocoverwith i'b;;;rr:T:jponding thatthe mateia bv the noun, i' rerered to ;;::.::,*": 4' third- problem for Miller's '.,-;r;i construalruie is' that ;il;;ir;t]::'-19 bv a denominarverb can perrormed*i,h;.;;;;;;;:;';'r:Il'^'-'.'.l9 be ttntioned by its base For nslance. auntr".'rn-.". noun. on. ",ntttY. Ji,lnt b'l alsooil with syrup'paint wilh mud' tt" rrtt*rt",'" .""*'J'l^? -11'it; denominar for ue'bsshtuid be formurated .,;;";;i; ;r';::1:i.1"'" (21) When nouns are used as
This construal rule coves the ca me rnterpretation of denominar :::J:t::1,1t (20a), but aiso accounts fo uerbsdervedfrom non-mass nouns: (22) N knikte ,mable, knikke ,to plly marbles, s t aa l ' s l a v e '

H"i

pr"y,iii1i""u,

In this paper I discussedthe problem what kind of regularities can be observed in the interpretation of complex words. Fo cedain types of complex words, e.g. deverbal agentive nouns, a specificsemanticcorrlate of the morphological structure can be fomulated. But the interpretatlon of compounds, reference-modifyingdenominal adjectivesand dcnominal verbs is only partly determined by morphological structure: it is the result of an interaction between linguistic structure and nonJinguistic knowledge- Such an 'interactionist' approach to wod formation processes maks it possibleto formulate a restrictive theory of word formation. The enonous productivity of compounds, reference-modifying denominal adjectives and deuominal verbs is explained by the semantic vrsatility they possessdue to the absenceof specific semantic aorrelates. Relating these conclusions to the theoretical famework of the Extended Standad Theory as outlined in Chomsky and Lasnik (1977), we claim that the smantic interpretation of complex words is derived in two steps: rules such as (15) and (18) derive Semantic Interpretation-I (the logical form), and rules such as (21) derive Semantic Interpetation-Il from Semantic InteDretation-I. Receved12 October 1979 Revsedversionreceved 28 January 1980 Insttuut voor N eerlandstiek Universiteit van Amsterdam Herengracht 330-336 l016 CE Amsterdam The Netherlands

:[:";",'l;,ff;"ilf.i'i{";:.i",1; ff"',TJ;i";;;''$ffi

Notes
l. This is in accordancewith the use ofthis term ir the Extended Standard Theory: 't used the expressior /ogicdl/m to designatea level oflinguistic representationincorporaling all smantic properties that are strictly detemined by linguistic rules'(Chomsky, 1979: 145). This does not necessarilymean that the element to the right of the arrow is a word. ln languagessuch as Latin we will pcrhaps have to assumestmsas word forming elements. Nevertheless,such stems will aiso be assigneda syntactic category symbol. There is an alternative to the hypothesis of labelled bracketing. One could assume WFRS of the following type: ( l )' (2)' [x]N [be+x]v [x]a + [x + tten]N

.ro pr.y ;::1,y.'^j:::f.,, voerbar r""ii"ii koelie .lo

e ... oil.

formation processes.

characterisric activity associared li;'#:'l:.Tlies' w;,h thisis not thecase lhe nou^n-s i" tzool. tisred for ilr.".ri,*"iill: ,tl"tt applv.to verbs rhe in (20b). risred Moreover. lil',i::'-"?' ;;;;J;il: if ua knowledge u to..,n.'JJni.],.i; o, '.1:"9' notpplv. sit tion-bound rhe inrerpretation the denominil or ri;;i;' prasmatic ""b o;;pr.il;f; theinterpretaion ';;''r';:':l::'1"* resularities rn .tT:: :JI*:t r---' out
-vrus. they do not constrain these word

or..";;'.;;,;;"ur ,ur.(2r;r:;i,''o 'oir' if there.is some

2.

3.

These rules do not produce internal labelled brackets. Therefore, it remains to be investigated whether such internal labelled brackets are really necessary-Dell anci

tuUU

Geerf E. Booi

j,':i "rTl ;:.:, 1".::.i::;"3: 1,;,J:,',:.T'"1*n":J"T.J:i f1;l J 4 nmy on., ;i#il] :T:Tl#H:::::,,j+iji,*;:ili.;T:'fl opinj,,..,;.; l :..;y;, i"iff I,rr*:;l*l*:;:i,[:j;r*ili::{:"{i n:#.r.",'ffJr*i iHt,!li * m :li1:ij.]l,i.l,j::l.jli"ff ill;";l#,{i, l,1tH:j ;:.i:Ll
_ r n r e r , r e r h lh e p r o b le m wr o fh

,"""" . :trJ::l :l:".T ,h.i, "": ,,h;;pr:.;,i;,;: ;T-:lrn::j"JJJ:,1.::;1: j:i"" jii:,

/."rheserunr rions coutd be

q Se t(ir k t 97x r ugge, r s r bart he. descriplonof rhe Leamed Frenchhas ro .;; BcrngRureror ;;; l; :::::,ijar

Semantic rcgularities in word Jbrmation l00l

Jespersen, O. (1942). A Modern English Grammat on Hisrorical Ptitlcples Coper'aqenl Ejnar Munksgaard. Karius. I. (1977). fnstrumentalilat und denominale nullsuffigierte Verben ds Englischen ln: H- E. Breckle ud D. Kastovsky (eds), Perspektiw der lVo bildungsfarschuns,104 115. Bonn: Bouvier Velag Hermann Grundmann. Lees. R. (1960). The Gramar ol Engtish Nominalizations.Bloorringronr Indiana Univrist) Press. Levi, J. N. (19?8). The Sytax antl Semanticsof Complex Nomhab. New York: Academic Marchand, H. (1969). The Categories and'Dpes oJ Prcsent'Da)JEnglish Wot Fornolik Mnchen: verlag C. H. Beck. Mcys, W. J. (1975). Compouui Adjctivcs in English and the tdeal SpeakerJistenet Amsterdamr Noth Holland. Millr,6- A. (1978). Semantic rclations among words. In M. Halle, J. Bresnan and G. A. Miller (eds), t"8adt. feary an.l Psythological Reall], 60-l I8. Cambridge, Mass-: MIT Press. aaLen taalgebtuik. Reichling, A. (1967). fit Woortl Ee studie o trent de de grondslag 'an Zwoller Tjeenk Witlink. Roeper, Th. and Siegel, M. E. (1978). A lexical transformation for verbal compounds Lins istic Inqut 9 (2\, t99-26O. oJ Rose, J. H. (1973). Pricipled limiiations on productivity in denominal \/erbs.Foutlkttons Lansuase \0(4), 509-526. Siegel, D. (197?). The adjacency condition and the theory of morphology. (Unpubiished paper, Denver, Coiorado). Taski, A- (1964). Inleiding tot de lo+a (Dutch versio of I ro.luction to LoCi() Amsterdam: Nood-Hollandsche Uitgevers Maalschapp. Vries, J. W. de (1975).1- xicale morfologie van het vetkwooftl in Modetn Nedrrlrn^. Leide':': Universitaire Pers Leiden. Wasow, T- (1977). Transformations and the lcxicot. In P. w. Culicovef, T. wasow and A Akmajian (eds), rrndl SJntax, 321-360. New York: Aadcmic Press.

expressedaddins b*"0,, by su ,. ,.1.1_l',lll:i.::l.:l'd

References
Aronoff. M I I07). word Fo,,na on tn ca"".-,,--

u,::;lJ"o D^^'d;:";",,31';i'';:;,;,::,:;:f,,:::,:,i:;:":ii;::,"il:H:::
r.

lli tl 'i.

lr i,

Lrax ua, Sillli[.?, i,-'j,i]i1',:".1,:::','n uB.,]4 A stu4 ot od

Dordrechr: ,.l",ri.iiff"'"t' "Tll

Fonatian Generul,e in orammar.

tl;J,.I'iltl,#p

run'Itnn rhc or Letircn rnn,rorno,iona!.eea,urrvc in (,,omma,.

B r e s n a n , J .W .

C d r r o t J ,J . M a n d T a n e n h a u s, M. Chomsky, N

;;:;':;:;:' iiliT;)I f,, f;'j::.''."''";'; :;Ii:i:;i'f: i,l;{'fi;:;: il:-", ,;;;;1;^; i:#" T ,ff lji ;" ii; ;,Li,iili, :1,".i":::].fi 1.j,,T, "l,,if l,il;:i:
K. I tt9:,9l Laneuas an R

9 7 8 1 A r e a istjc r a n sfl r

.,I11

l:-

;ffi:;l

.,1.':-;.tn*1r:,

s. J. Keyserred.r,n",",, ijlii'"lli'.i"''arr) 8o'"erned vowerdternaron in c1nu,ior.. v""., t,i p....' ' ' "a:latmattoaatstu.ti?ru European ansua,et, t ^^'.-fi {re77). on rfie creatron ":i;lt" and use of Enstisbcomprnd nouns.tansuase 53, Iodor. J. D. gttt. Senanu, pressLtd. -"4 Jhco,, d Meaninsn Generciye6;.arr,_ 'kurtcs Hassocks: ,-Haryesrer t ,n, t. senantrc terprctuton '"fiiffi rn in Generatiye ' Grummat.cambridse j Mass. t ,'ttr). Morphorosicat and semantic resularies t he.lexicon in Laneuase l"!ll?$11 s (1s77) i.'I'ntax A stud! phrasesrrurrure. of lfLlt' " cambridgevo.r., ru,..

;.',il;i.'"ili.il1,L llll;i5:,:::..^:,I,e:: verbr a: ksuase B,, s516j i french fn

d con Linsui,!ic 8.42s. rct tnquirv "l,i;i,i""-.i;[,i,]ii,il:::,"" ;;; ;;:"1,];"iii;tll, ."**,*,1"',.1"i1 r"

".,t, """ti""i

Based convercaons on with Mi$ou sJi, ^esponsibiti|

i,',^".,::;r';;':