11.1 Introduction
Lexical derivation is taken here to be word formation. We assume a lexicon of morphemes and rules for combining them into words. Morphemes combine into words similarly as words combine into sentences. The sentence is formed by phrase-structure rules that expand S(entence) into the phrasal categories NP (noun phrase), VP (verb phrase), AP (adjective phrase), PP (prepositional phrase), and a functional category Inflection Phrase which governs the verb’s inflection. Phrase-structure rules further expand phrasal categories into the lexical categories N(oun), V(erb), A(djective), and P(reposition). The sentence acquires phonetic substance when lexical categories are lexicalized, that is, paired with morphemes from the lexicon. Word formation comes into play when lexical categories instead of being lexicalized are expanded by word-structure rules into their constituent categories and these are lexicalized.
But the lexicon may also be said to contain words, and then lexical derivation becomes a matter of relating some words to others. If the lexicon contains only the basic Noun, Adjective, and Verb, rules are needed for attaching suffixes to them.Footnote 1 The suffixes are -N (noun suffix), -A (adjective suffix), and -V (verb suffix). The rule N → A -N derives the deadjectival noun N[A -N]; the rule A → N -A derives the denominal adjective A[N -A].
If the lexicon contains both base words and derived words, word-formation rules have the task of relating the two. If the lexicon contains both a noun N and a denominal adjective A[N -A], it is redundant if it lists the form and meaning of N in both entries. To reduce redundancy, in the A[N -A] entry we list only the meaning of -A (e.g. ‘pertaining to N’) and let the meaning of N in A[N -A] be supplied by the reference to N.Footnote 2
A word approach has word-formation rules operating in the lexicon. For the noun jadovitost‘ ‘toxicity’,Footnote 3 for example, the noun /äd/ ‘poison’ is joined with the adjective suffix /ovit/ and with the noun suffix /ost/. The word so assembled is available for insertion into the sentence under N. With a morpheme approach morphemes combine in the sentence. An N in the sentence is expanded to N[ A -N], the A of which is expanded to A[N -A]. The derived N[A[N -A] -N] structure is lexicalized as N[ A[ N[/äd/] -A[/ovit/]] -N[/ost/]]. No lexical insertion is called for because the noun has been formed already in the sentence.
11.2 Morphemes and Phonemes
11.2.1 Morphemes
The morphemes in the lexicon are abstracted from words as they occur in sentences, and in that sense they are abstract. But they are never totally abstract, lacking phonetic substance. Observing the traditional definition of morpheme as a pairing of sound and meaning, we reject non-phonetic, null morphemes.Footnote 4 Alleged null morphemes are better accounted for in other ways. The null endings occurring in many nom. sg. and gen. pl. noun forms are fleeting vowels (Section 11.2.2), which are deleted by a phonological rule. Null derivational suffixes are not needed as heads of derived words. Rather than X → [Y -X] with a null X suffix, we prefer zero derivation, the recategorization X → Y that results in X[ Y], a Y morpheme with X syntax.
11.2.2 Phonemes
Morphemes may be said to consist of phonemes. But morphemes take various forms in various sentence environments and so do their constituents.Footnote 5 Accounting for this allomorphy is the task of phonology. Phonological rules turn A, a morpheme segment represented by a set of phonetic features, into B, a segment with a slightly different set of phonetic features, in environment C, also represented by a set of phonetic features. We observe Reference Postal and DinneenPostal’s (1966: 56) Naturalness Condition, that the relation between the lexical representations of morphemes and their phonetic representations be a natural one, both stated in phonetic features. Hence the ubiquitous fleeting vowels (jers), for example in son ‘sleep’ (gen.sg sna), are not, as sometimes represented, abstract non-phonetic /#/s which are turned into vowels by morphophonemic rules. They are real vowels, sets of phonetic features. Reference Halle and MatushanskyHalle & Matushansky (2006) describe them as high and lax, that is, produced without advanced tongue root. They represent them with small cap i’s, barred for the back jer. Here they are represented with the Cyrillic letters /ь/ (non-back) and /ъ/ (back). Rejecting non-phonetic /#/ is consistent with rejecting non-phonetic (null) morphemes. Representing morphemes in the lexicon calls for a larger set of phonemes than representing their surface form. In addition to /i e a o u/, there need to be a tense mid front /ě/, a low front /ä/, and a high back unrounded /y/.
11.3 Suffixation
The principal rule for combining morphemes into words is suffixation. It has the form X → Y -X and expands X into X[ Y -X], a Y root (N, A, or V) and a -X suffix (-N, -A, or -V). The suffix heads the word and determines its category and inflection. X → Y -X is recursive: it may apply to Y and also to its constituents. For the verb bezdel‘ničat’ ‘be idle’, V expands to V[N], the noun of which, bezdel‘nik ‘idle person’, contains the A[ PP -A] bezdel‘nyj ‘without business’, which contains the prepositional phrase bez del ‘without business’. X → Y -X rules provide for the alternation of suffixal categories but do not say which suffix occurs in a given slot, why the noun containing /pust/ ‘empty’ is pustota, not *pustost‘, and why the adjective containing pustota is pustotnyj, not *pustotskij. With the lexicon containing only morphemes, it is with morphemes that information about the suffixal makeup of the word must be associated. It is a feature of /pust/ that it selects /ot/, not /ost/, and also a feature of /pust/ that what follows /ot/ is /ьn/, not /ьsk/.
The occurrence of an adjective with a productive suffix may block its occurrence with a more productive suffix. The productive suffix /ot/ in polnota ‘fullness’, slepota ‘blindness’, and čistota ‘purity’ blocks these adjectives from occurring with the more productive /ost/Footnote 6 (no *polnost‘, *slepost‘, *čistost‘). But when the /ot/ noun is reified, as in kislota ‘acid’ and ostrota ‘witticism’, /ost/ occurs for the abstract non-reified meaning kislost‘ ‘sourness’ and ostrost‘ ‘sharpness’. In Polish the reified wspólnota ‘commonwealth’ with /ot/ leaves /ost/ available for abstract wspólność ‘joint ownership’.
With the lexicon containing only morphemes, the meaning of a root–suffix combination, if idiomatic, must be encoded with the root. Idiomaticity varies inversely with the productivity of the suffix: the more productive the suffix, the less likely its combination with a root will be idiomatic. Productive /ost/ and /ot/ mostly just recategorize adjectives as nouns, but the non-productive suffix /j/ forms N[A -N[/j/]] nouns with unpredictable meanings. It combines with gustoj ‘thick’ to form gušča ‘dregs’, with gryzt‘ ‘gnaw’ to form gryža ‘hernia’, and with pustoj ‘empty’ to form pušča ‘dense forest’. In Polish /j/ combines with tłusty ‘fat’ to form tłuszcz ‘grease’, with gęsty ‘thick’ to form gąszcz ‘thick undergrowth’, and with suchy ‘dry’ to form susz ‘dried fruit’.
Roots combine with suffixes for six word types: deadjectival nouns (N[A -N]) like BCS punoća ‘fullness’, Cze. blbost ‘stupidity’,Footnote 7 and Pol. ślepota ‘blindness’; deverbal nouns (N[ V -N ]) like BCS berba ‘picking’, Cze. učitel ‘teacher’, and Pol. badacz ‘researcher’; denominal adjectives (A[ N -A]) like BCS sunčan ‘sunny’, Cze. rohový ‘corner’, and Pol. sławny ‘famous’; deverbal adjectives (A[ V -A]) like osedlyj ‘settled’, Cze. plynulý ‘continuous’, and Pol. zamieszkały ‘resident’;Footnote 8 denominal verbs (V[ N -V]) like BCS kartati se ‘play cards’, Cze. stylizovat ‘stylize’, and Pol. standardyzować ‘standardize’; and deadjectival verbs (V[ A -V]) like BCS gladnjeti ‘become hungry’, Cze. vulgarizovat ‘vulgarize’, and Pol. popularyzować ‘popularize’.Footnote 9
11.4 Same-Category Suffixation
There are also branching rules of the form X → X -X, chiefly N[ N -N ] for denominal nouns. With root and suffix of the same category, it may be unclear which is the head morpheme, whether it is a matter of -N characterized by its relationship to N or of N modified by -N. OCS rybarjь ‘fisherman’ is clearly right-headed, a person, -arj-, with a relationship to fish, ryb-, while OCS rybica is left-headed, denotes a fish, ryb-, qualified as small, -ic-. This reading is supported by the syntactic roles of the suffixes: /arj/ determines the morphology of the noun to which it is suffixed; diminutive /ic/ does not. It lacks gender of its own and simply passes on the gender of the root. This is true also of bratec, which is masculine like brat ‘brother’, sestrica, which is feminine like sestra ‘sister’, and vinco, which is neuter like vino ‘wine’. Similarly in Polish, kawałek is masculine like kawał ‘piece’, śrubka is feminine like śruba ‘screw’, and piwko is neuter like piwo ‘beer’. In these left-headed nouns the right-hand part has a weak claim to morpheme status and may be better regarded as a theme (see Section 11.8).
A suffix with its own gender has a better claim to head morpheme status. Take nouns with the neuter suffix /ent/ which denote the young of the species. Next to OCS osьlъ ‘donkey’ there is suffixed osьlę, gen. osьlęte, next to Cze. husa ‘goose’ there is the /ent/ form house, gen. housete, and next to Pol. kot ‘cat’ there is the /ent/ form kocię, gen. kocięcia. Note that the gender contrast of husa and kot is neutralized by neuter /ent/. This suffix with orël ‘eagle’ shows up in plural orljata, gen. orljat. In the singular, /ent/ loses the /t/ and selects the singulative suffix /ъk/: orlënok, orlënka. The morphological reshaping of the noun due to /ent/ supports the right-headed meaning ‘the young of the N species’ rather than the left-headed reading ‘an N that is young’. In South Slavic, /ent/ is preceded by /ъk/, for example in Bul. momče, plur. momčeta ‘boy’ and in BCS čobanče, gen. čobančeta ‘shepherd’.
Another morphologically dominant suffix is the neuter non-count suffix /ij/. It occurs with nouns denoting inanimate objects, for example prut‘ë ‘branches’, OCS dǫbije ‘trees’, Cze. dříví ‘trees’, and Pol. kwiecie ‘flowers’. Suffixed to nouns denoting persons, for example in bab‘ë ‘women’ and starič‘ë ‘old men’, it is pejorative.
Also morphologically dominant is the singulative /in/ which occurs with plural animate nouns like graždane ‘citizens’ (sg. graždanin) and Pol. chrześcianie ‘Christians’ (sg. chrześcianin).
On formal and semantic grounds the abstract suffix /ьstv/ ‘state of’ should be considered the head in gosudarstvo ‘state’, Cze. kněztvo ‘principality’, and Pol. państwo ‘state’.
On the other hand, clearly left-headed are deadjectival adjectives like Pol. słabawy ‘rather weak’ and ciężkawy ‘rather heavy’, and diminutives like tolstjuščij (cf. tolstyj ‘fat’), Cze. tichoučký (cf. tichý ‘quiet’), and Pol. prędziutki (cf. prędki ‘fast’). The root here is the head constituent and the suffix a modifier.
11.5 Suffixless Recategorization
In addition to recategorizations via branching rules of the form X → Y -X, there are also the non-branching recategorizations X → Y, which generate X[ Y ] words having Y form and X syntax. These rules generate deverbal nouns (N[ V ]) like lož’ ‘lie’, Cze. skok ‘jump’, and Pol. ślizg ‘slide’; deadjectival nouns (N [ A]) like šir’ ‘breadth’, Cze. hloub ‘depth’, and Pol. dal ‘distance’; denominal verbs (V[ N]) like obrazovat’ ‘form’, Cze. sloužit ‘serve’, Pol. mieścić ‘place’;Footnote 10 and deadjectival verbs (V[ A]) like belit’ ‘whiten’, Cze. černit ‘blacken’, and Pol. dziczeć ‘grow wild’. Suffixless denominal adjectives (A[ N]) are few in number: perhaps Rus. rjaboj ‘speckled’ and zolotoj ‘golden’ (unless its relationship to zoloto ‘gold’ is different and zoloto is N[[A/zolot/]]). Deverbal adjectives (A[ V]) include živoj ‘living’ and plëvyj ‘worthless’ and Pol. luby ‘pleasant’.
In the absence of a suffix, the morphology of a recategorized word is determined by its syntax. The deadjectival nouns šir’ ‘breadth’, glub’ ‘depth’, and tiš’ ‘stillness’, etc. owe their i-declension forms to their N[ A] structure. The thematic /i/ of belit’ ‘whiten’ reflects its transitivity; compare intransitive belet‘ ‘show white’ with thematic /ě/.
The gender and declension class of suprug ‘husband’ and supruga ‘wife’ and of kum ‘godfather’ and kuma ‘godmother’ are not a matter of word formation, but of syntax. The phrase-structure rule that expands NP to N subcategorizes it as + animate or − animate (to account for agent/instrument nouns like istrebitel’ ‘destroyer’ with contrasting accusative forms reflecting animacy) and subcategorizes animate nouns as + feminine or − feminine.
11.6 Fused Suffixes
Viewing words as structured strings of sound–meaning pairings (morphemes), we expect each to make its contribution to the meaning of the word. This expectation is met in the derivational family pustoj, pustota, pustotnyj, pustotnost’, which share the adjective /pust/ ‘empty’ and where each recategorization is meaningful, as pustotnyj differs in meaning from pustoj and pustotnost’ differs from pustota. It is also met in babničat’1 ‘womanize’, the verbal recategorization of the noun babnik ‘womanizer’ (one, /ik/, who relates, /ьn/, to women, /bab/, in a certain way). But there is also babničat’2, which means simply ‘behave like a woman’ or ‘be a midwife’. Here /ik/ is semantically empty and /ьn/ has no function. Semantically, babničat’2 is a verbalized V[ N[ /bab/ ] ], which could be realized as *babet’ like vdovet’ ‘be a widow’. But it includes /ьn/ and /ik/. These two morphemes in losing meaning have lost morpheme status and have fused into the suffix /ьnik/, which occurs also in lakejničat’ ‘be a lackey’ and koketničat’ ’be a coquette’ (note the non-occurence of *lakejnik and *koketnik). So while babničat’1 is a suffixless derivative of babnik, babničat’2 is derived from /bab/ with the fused suffix /ьnik/.
11.7 Bound Roots
The rule N → N -N accounts for the nouns ispanec ’Spanish man’ and ispanka ‘Spanish woman’, except that ‘Spain’ is not *Ispan, but Ispanija. The Russian Academy Grammar (RG) holds that a word can only be derived from (motivated by) another word. Reference AronoffAronoff (1985: 22) agrees, stating that word-formation rules operate only on words, not on morphemes, and that rules “can only derive meaningful words from meaningful bases.” The bound morpheme /ispan/ although not a word is clearly meaningful. We could relax the word requirement and accept bound morphemes in the lexicon. More problematical are nemec ‘German man’ and nemka ‘German woman’. They appear to be structured N[ N[/něm/] -N[/ьc/]] and N[ N[/něm/] -N[/ъk/]]. But /něm/, a bound morpheme like /ispan/, is not immediately meaningful. Yet nemec and nemka are as meaningful as ispanec and ispanka. Nemec denotes a resident of Germany just as ispanec denotes a resident of Spain. The meaning of /něm/ here is ‘Germany’. which it derives from the meaning of its derivatives.Footnote 11 Maintaining the word criterion, RG derives ispanec from Ispanija via the truncation of /ij/. More questionably, RG derives nemka from N[ N[/něm/] -N[/ьc/]] via the truncation of /ьc/.
Jagodica ‘buttock’ and Pol. piwnica ‘cellar’ appear to contain the noun roots /ägod/ ‘berry’ and /piv/ ‘beer’. To understand jagodica as the part of the body that is smooth and round like a berry is simile; to understand piwnica as a place where beer is stored or served is metonymy. Simile and metonymy are common semantic operations in everyday language use. These operations must be active in jagodica and piwnica; otherwise these nouns have no structure.
11.8 Morphemes versus Themes
While most of the word constituents discussed in this chapter – prefixes, roots, suffixes, and endings – are morphemes, not all word constituents are morphemes. Some are themes, phonemes introduced as buffers between morphemes by readjustment rules. Their introduction is often morphologically motivated. When a consonant-final prefix precedes a consonant-initial root or a consonant-initial suffix follows a consonant-final root, the resulting consonant cluster may undergo a sound change that obscures the word’s morphological structure. For example, the prefixes /ot/ ‘from’ and /ob/ ‘around’ before a consonant-initial root conditioned cluster simplification, so that in Old Russian V[P[ /ot/ ] V[/xoditi / ] ] ‘go away’ and V[ P[/ob/ ] V[/xoditi / ] ] ‘go around’ fell together as oxoditi. A readjustment rule introduced /ъ/ between prefix and verb to keep these verbs distinct. The noun /kamen/ ‘stone’ in Old Church Slavonic was followed by the adjective suffix /n/, resulting in kaměnъ with an altered root. But thematic /ь/ was introduced between root and suffix, and the more transparent kamenьnъ resulted. ‘You eat’ in Old Russian was structured V[ V[/ěd/] E[/te/]] and realized as ěste with an obscured root. But this form has since been thematized to V[ V[/ěd/] /i/ E[/te/]] for present-day edite. It is thematization that made this form more transparent, not being restructured with a present-tense suffix.
It is widely held that prefixed imperfective verb forms in Russian and other Slavic languages are derived from their perfective counterparts by means of an imperfective suffix, impfv. spasat’ ‘save’ from pfv. spasti with a suffix /a/ and impfv. zapisyvat’ ‘write down’ from pfv. zapisat’ with a suffix /yva/. But no imperfective suffix relates impfv. prinosit’ ‘bring’ to pfv. prinesti. These forms differ only in their thematization. It is likewise by their thematization that spasat’ differs from spasti and zapisyvat‘ differs from zapisat’.Footnote 12
A string of phonemes can be a suffix in some words and not in others. Russian has a productive class of deverbal adjectives like gibkij ‘flexible’, padkij ‘susceptible’, and vërtkij ‘nimble’, structured A[ V -A ] with a verb root and the adjective suffix /ъk/. Russian also has a smaller, unproductive set of /ъk/ adjectives like gladkij ‘smooth’, nizkij ‘low’, and uzkij ‘narrow’. They do not have X[Y -X ] structure because the roots are adjectives and what follows is not a categorizing suffix. Note that the two classes of adjectives differ in their comparative forms: A[V -A ] adjectives include the /ъk/ suffix (gibče, vërtče), but gladkij, nizkij, and uzkij do not (glaže, niže, uže). RG ascribes the absence of /ъk/ in these forms to truncation, but the phonology assumed in this chapter does not countenance the deletion of phoneme sequences from a word form. What needs to be accounted for is the presence of /ъk/ in gladkij and nizkij, not its absence in glaže and niže. The /ъk/ in gladkij and nizkij is thematic; it is not a suffix.Footnote 13
11.9 Prepositions and Prefixes
Prepositions and prefixes are a single lexical category P, but they differ syntactically. Prepositions are introduced into the sentence with the expansion of the phrasal category PP to P NP, whereas prefixes are introduced with the expansion of X to P X. Here P is an adjunct, not a head determining the word’s category and inflection: N[ P N ] is a noun, A[ P A ] is an adjective, and V[ P V ] is a verb. Suffixes categorize the word while contributing little to its meaning; prefixes contribute to the word’s meaning while not categorizing it.
11.9.1 Prefixed Nouns
Some prefixed nouns, for example antisemit ‘anti-Semite’, are structured simply N[ P N ]. But when the prefix is followed by a suffixed noun, we have the familiar structural ambiguity of a three-morpheme sequence: is it [[A B] C] or [A [B C]]? Antikommunizm ‘anticommunism’ means opposition to communism, so the noun is structured [A [B C]]. But antidarvinizm ‘anti-Darwinism’ could be understood as opposition to Darwin, rather than to his theories, so it could be structured [[A B] C].
11.9.2 Prefixed Adjectives
Prefixed adjectives include raskrasivyj ‘very beautiful’, Cze. přihloupý ‘somewhat stupid’, and Pol. prześliczny ‘very beautiful’. They are structured A[P A ]. In Russian, /bez/ ‘without’ as a word constituent is mostly a preposition, for example in dephrasal bezvodnyj ‘arid’ and bespravnyj ‘lawless’. But besčelovečnyj means ‘inhuman’, not ‘lacking a human being’, so it is structured A[ P A[N -A ]] with /bez/ a prefix. Russian has a homophonous pair of adjectives differing in structure: sverxsročnyj ‘extra-term’ (sc. service) is dephrasal A[ PP -A ], whereas sverxsročnyj ‘very urgent’ is a prefixed adjective, A[ P A[ N -A ] ]. Structured like the former is sverx‘’estestvennyj ‘supernatural’; structured like the latter is sverxčuvstvitel’nyj ‘highly sensitive’.
11.9.3 Prefixed Verbs
The expansion V → P V may be lexicalized as V[ P[ /za/ ] V[ /pis/ ] ],Footnote 14 where /za/ changes the meaning of the verb from ‘write’ to ‘write down’ and makes it acceptable in a perfective sentence. In an imperfective sentence the prefix conditions extended thematization to zapis-yva-t’. In a prefixed V[P V ] both P and V can be expanded, P to P[ P P ] and V to V[ P V ]. The former expansion, to V[P[P P] V], is seen in perezarjadit’ ~ perezarjažat’ ‘recharge’ and Pol. odpowiedzieć ~ odpowiadać ‘answer’. Here P[ P P ] has the same aspectual function as P in zarjadit’ ~ zarjažat’ ‘charge’ and powiedzieć ~ powiadać ‘tell’. The latter expansion, to V[P V[P V]], is seen in pererasprašivat’ ‘question’ (many people) and Pol. pozamykać ‘close’ (many things). In these perfective-only verbs pere- and po- have the same aspectual function as po- in posidet’ ‘sit a while’ and Pol. pochodzić ‘walk a little’.
11.9.4 Recategorized Prepositional Phrases
A prepositional phrase can be recategorized as a word. Dephrasal nouns include podol ‘hem’, Cze. nádoba ‘vessel’, and Pol. bezdech ‘asthma’. Dephrasal adjectives include beznogij ‘one-legged’, Cze. bezbřehý ‘boundless’, and Pol. bezwłosy ‘hairless’. BCS obešumiti ‘deforest’ shows the phrase PP[ P[ /bez/ ] N[ /šum/] ] recategorized as a verb. Similar in structure is besslavit’ ‘dishonor’. But rasslavit’ ‘praise to the skies’ is a prefixed denominal verb, structured V[ P[ /raz/ ] V[ N[ /slav/ ]]].
The verb in a deverbal noun may be prefixed, thus N[ V[ P V ] ]. In zapas ‘supply’ this structure is supported by the aspеct pair zapasti ~ zapasat’ ‘stock, supply’. Likewise deverbal is Cze. postřik ‘spraying’, which shares V[ P V ] with postřikat ‘spray’. However, privetstvovat’ ’’greet’ does not contain the verb V[ P[ /pri/ ] V[ /vět/ ] ], but the noun N[ P[ /pri/ ] V[ /vět/ ] ]. Likewise in Polish, pokłonić się ‘bow’ contains the noun N[ P[/po/ ] V[ /klon/ ] ], not the verb V[ P[/po/ ] V[ /klon/ ] ], as there is no aspect pair pokłonić się ~ *pokłaniač się.
11.10 Summary
Lexical derivation is presented here as word formation. Words are formed in the sentence by word-structure rules that expand root categories like N and A to derived structures like N[A -N] and A[N –A], followed by the lexicalization of the root and suffix categories with morphemes from the lexicon. Differing from the morpheme approach to morphology is the word approach, which assumes a lexicon of words and favors processes over items. The advantage of processes is that they can deal with phenomena like ablaut, infixation, morpheme subtraction, and zero morphemes. Zero morphemes are touched on in Section 11.2.1. Ablaut and infixation may be dealt with in the phonology. Morphological subtraction is mentioned in Section 11.8. Grammars claim that subtraction is how some morphologically simple words are derived from morphologically complex ones, for example the noun intellektual ‘an intellectual’ from the adjective intellektual’nyj ‘intellectual’. The adjective could well have been known to Russian speakers before the noun, in which case the subtraction of the adjective suffix was, by backformation, a diachronic change. Synchronically, the noun is a simple N while the adjective is a derived A[N -A].
A central problem of morphology is overproductivity: it generates more morpheme combinations than actually occur in sentences. It is suggested above that suffix distribution can be coded in the root morpheme, that, for example, what Russian speakers know about the adjective /pust/ ‘empty’ is that it selects the noun suffix /ot/, not /ost/, and that the combination [N[A/pust] N[/ot/]] selects the adjective suffix /ьn/, not /ьsk/. But this suffixation information may be too much of a load to place on a root.
Several writers (e.g. Reference HalleHalle 1973) have suggested that in addition to a lexicon of morphemes there must also be a dictionary of actually occurring words. Morpheme combinations not in the dictionary would be blocked from occurring in a sentence. A dictionary of words would be one solution to the overproductivity caused by context-free morpheme combining.
