Abstract and Keywords
The origin and diffusion of two English suffixes are treated. The first is 3sg. ‐s, which is not of Scandinavian origin but resulted from generalization of ‐s from is to has to other irregular monosyllabic verbs, to which ‐s was restricted in Early Modern English, and finally to regular verbs. The second study is deadjectival ‐en, which expanded under Norse contact in northeast England. Novel English verbs were created by semantic analogy to the Nordic models, which imposed a set of constraints: monosyllabic base, trochaic foot structure, and root‐final obstruent. A brief comparison with the spread of other affixes reveals that derivational and inflectional formatives spread by lexical diffusion. General utility and extralinguistic factors determine the rapidity of the spread and the degree of productivity.
On most recent theoretical accounts, morphology is not autonomous, but interacts with at least three other domains: (i) the lexicon/culture, (ii) syntax, and (iii) phonology and perception.1
The first can take the form of special formatives, like classifiers or evidential markers. In terms of morphological change, one can find loss of culturally unimportant vocabulary items and categories, stability of culturally important morphemes, or expansion of the same to derivational and even inflectional status. Examples of the latter include the diminutive pejorative (‘lousy little—’ ) in Fula (Anderson 1985: 117; Miller 1993: 12) and development of the feminine gender in Indo‐European (Ch. 5*).
The second interface, that with syntax, is revealed not only in trade‐offs between morphological (synthetic) and syntactic (analytic) realization of the same structures, but more frequently in the expression of what several structures have in common or how they differ. Chapter 7* documents the changes from Latin to Romance in the coding of reflexive, anticausative, middle, and passive. In Early Latin, the last three had the same morphological form, but until very recently in Romance the first three have had the same morphosyntactic expression. Morphology can express some general fact, such as the unavailability of an argument for saturation in syntax. Formal renewal of one category, the passive, dissociated it from other structures with similar effects. There is a constant struggle between each syntactic structure having its own exponent, or having a single marker that captures what several structures have in common.
The interface with phonology and perception has been exemplified throughout this work. Many morphological changes by analogy and systematic repair have been discussed. Cliticization and phonological reduction can (p.103) create a mirror principle violation that is subject to repair (§6.13). Compounds, clitics, and affixes interact (Ch. 3*). And so on.
This chapter looks at several suffixes in which multiple factors play a role.
4.2 The origin and diffusion of English 3sg. ‐s
In Old English, a typical (strong) verb like ‘bear’ was conjugated in the present tense as follows: 1sg. be(o)ru, 2sg. biris ~ bir(e)s(t),2 3sg. birið (~ bir(e)ð), pl. berað. By Late Old English the singular was leveled to bere, berest, bereð (〉 beareth). Apart from Krug (2003: 20ff.), who assumes a leniting of /þ/ to /s/, Modern English 3sg. ‐s (bears) is generally attributed (in one way or another) to Scandinavianization. At most, the influence was very indirect, limited to the fact that the initial spread of ‐s to all forms of the paradigm took place in the north, where contact served as the catalyst for the initial restructurings. The ultimate outcome however resulted from a generalization of ‐s from is to has to other irregular monosyllabic verbs, to which ‐s was restricted in Early Modern English, and finally to regular verbs (cf. Miller 2002b).
4.2.1 The Scandinavian hypotheses
There have been at least three different ideas of the way in which ‐s is the result of Scandinavian influence. The first hypothesis is that the identity of the second‐ and third‐person singular in Old Norse prompted the same identity in northern English, with ‐s borrowed from Norse. Keller (1925) dates the Nordic change of ‐s/z to ‐r in c10.3 Thus, OIce em, ert, er (‘am, are, is’ ) at the time of the settlements was em, es, es.4 Keller cites barutz ‘he breaks, you break’ on the Björketorp stone [600–800] as evidence of the identity of 2/3sg. during the Viking period (cf. Kisbye 1982: 81). This 2/3sg. identity is ostensibly mirrored in the Lindisfarne and Rushworth Gospel glosses [c10], where ‐s variably occurs in the 3sg.; cf. (Lindisf.) loseð/loses ‘loses’ (Matt. 10.39); cymeð ‘comes’ (Matt. 24.50) beside cymes (Matt. 24.46); etc. (Berndt 1956: 106, 209).
The second Scandinavian hypothesis takes English 3sg. ‐s from ‐sk (§6.13.1), but mediopassive ‐sc was consistently written in Norse manuscripts until mid c13 (Ottósson 1992: 57ff., 117ff.). Moreover, ‐sk is directly represented in several (p.104) loans, e.g. bask = OIce baðask ‘to bathe (oneself)’ (de Vries 1977: 22; Kisbye 1982: 80), busk = OIce búask ‘prepare onself; get ready’ (see distribution map in Bator 2007).
According to the third Scandinavian hypothesis, ‐s was a phonetic adaptation of ‐þ under Nordic influence (cf. Jespersen 1942: 18). Kroch et al. (2000: 379) see the imposition of less marked ‐s for more marked ‐þ as a result of imperfect learning of English among the Scandinavians of the Danelaw. They acknowledge that Old Norse had /þ/, but argue that (i) replacement of a marked by an unmarked segment is frequent in adult language acquisition; (ii) Norse had /ð/ except word‐initially while English had /ð/ only word‐internally; (iii) in Norse, /s/ but not /þ/ could occur in final position; and (iv) verbal endings in Old English were weakly articulated and prone to misperception. The last point is especially misleading because the third‐person ending, as van Gelderen (2000b) shows, was very strong in earlier English, and even allowed for null subjects, in contrast to the first‐ and second‐person endings which were prone to reduction already in Old English and rarely allowed null subjects. The strength of the third person enabled ‐þ to resist very long against innovated ‐s.
All of the Scandinavian hypotheses are inconsistent with the distributional facts.
4.2.2 The distributional facts and paradigmatic generalizations
Holmqvist (1922) demonstrated that the identity of 2/3sg. was an illusion; that, in reality, 2sg. ‐s spread first to 2pl. (and 2pl. imperative), then to 3pl., and, last of all, 3sg. The statistical evidence has been reassessed in detail by Berndt (1956: 204), who cites the following percentages (rounded off) of ‐s per category in the Lindisfarne and Rushworth Gospels: 2sg. 89 percent, 3sg. 33.4 percent, 1pl. 66.2 percent, 2pl. 66.6 percent, 3pl. 45.1 percent. Stein (1986: 640) modifies the figures slightly. His percentages for ‐s (including the Durham Ritual) are 2pl. 66 percent, IMPVpl. 54 percent, 3pl. 42 percent, 1pl. 42 percent, 3sg. 31 percent, which strongly supports Holmqvist's generalization from 2sg. to 2pl.5
A generalization involving second‐person and imperative forms is not unusual. The bond among second‐person forms and imperatives is well known (cf. Kuryłowicz 1964: 240f.). A similar generalization (but from 2sg. (p.105) imperative to 2sg. to other forms of the paradigm) is documented for Río de la Plata Spanish by Moyna (1996: 26, 84): innovation of oxytonic imperatives, such as tené (standard ten) to tener ‘have’ , motivated 2sg. tenés (replacing tienes), then other paradigmatic forms.
To summarize, the comparative rarity of 3sg. ‐s is problematic for all three Scandinavian hypotheses. For hypothesis I, the identity of 2/3sg. in Old Norse is simply irrelevant to the Northumbrian spread of ‐s from 2sg. to 2pl. Hypothesis II is irrelevant to ‐s as any sort of person agreement marker, and ignores the fact that ‐sk is directly represented in English. Hypothesis III falsely predicts that the highest figures for ‐s should be in the 3sg. and fails to predict that the second person should play any role whatever, much less that ‐s should have spread from 2sg. to 2pl.
Finally, on any of the Scandinavian accounts, it is a complete accident that, after a long period of variation, ‐s ends up precisely where it is etymologically justified in is.
4.2.3 An account of the distribution
After a period of generalization, the restriction of ‐s to the third‐person singular was initially triggered by retention with is and has (3sg. hæfis ‘has’ and næfis ‘has not’ are frequent in the Gospel glosses). Then 3sg. ‐s spread to other monosyllabic bases, especially those ending in a dental, e.g. sitteþ (Matt. 19.28) beside sittes (Matt. 25.31+) ‘sits’ ; cueðas (frequent) ‘says’ ; and so on.
What happened between then and Middle English must be reconstructed because of the lack of northern texts during that period. However, the records of the London Grocers' Company in (1) provide overlooked distributional information.
a) Also he grauntez to al þe company euery ʒer
aftor þat he duellez in london to gyffe a pipe of red wyn
For the whilke benefet he hase gyffyne…. (LGrC 2.190) 
‘Also he grants to all the company every year after he dwells in London to give a pipe (cask) of red wine, for which benefit he has given….’
b) 1) the whiche ‘which’
2) who so brekythe ‘whoever breaks’ (LGrC 2.190) 
c) yaffe ‘gave’ (LGrC 2.193) 
A northeast midlander evidently wrote the entry in (1a). All three of the 3sg. forms (grauntez, duellez, hase) end in a sibilant; whilke has its northern/Danish form (Miller 2004c); and give begins with /g/ (like ON gefa ‘id.’ ). (p.106) Contrast the entries (1b/c) in a more southerly dialect. Consistently, in entry after entry, there is a correlation of ‐s with other northern features, and ‐þ with other southern features. This cannot be accidental.
Evidence from the orthoepists is cited in (2).
(2) Evidence from the orthoepists
Writing in 1586, William Bullokar allows the use of ‐S as a poetic contraction of ‐ETH. In the 1620s, Alexander Gill labels HAS as a northern variant of HATH, regarding the forms with the dental fricative (HATH and DOTH) as normal, but adds that in other verbs ‐ETH can shorten to ‐s or ‐z, or become ‐ez after a sibilant. Finally, in 1643 Richard Hodges notes that, although ‐(E)TH may appear in writing, it is commonly pronounced ‐s or ‐z in ordinary speech….Hodges, too, continues to use the unsyncopated suffix with the dental fricative in his transcriptions of liturgical speech in The English Primrose (1644). Variation is therefore clearly in evidence in the verbal suffix in the spoken registers of Londoners even in the middle of the seventeenth century.(Nevalainen and Raumolin‐Brunberg 2000: 238)
The variation among Londoners in c17 is a continuation of that in c15 in (1). Also important are Hodges' notice that many people said ‐s/‐z but wrote ‐(e)th, and Gill's notice that has is northern but ‐eth can ‘shorten’ to ‐s/‐z.
Nevalainen and Raumolin‐Brunberg (p. 242) document a drop of 3sg. ‐s in London from 30 percent in 1460 to 0 in 1500, followed by a sharp rise between 1540 and 1580. By 1660 ‐s reaches nearly 100 percent. The explosion of ‐(e)s in London between 1580 and 1619 coincides with loss of the post‐stem vowel in the third‐person singular, motivating the orthoepists to interpret syncopated ‐s as a colloquial contraction of ‐ETH (Nevalainen and Raumolin‐Brunberg, p. 244). Overall, the change to ‐s in the standard language substantially occurred between 1410 and 1680 (Raumolin‐Brunberg 2005: 45).
It is significant that all of the verbs in the LGrC entry (1a) for 1428 are monosyllabic. Something along the lines of generalization from is and has to other (irregular) monosyllabics of high frequency is confirmed by the distribution of 3sg. ‐s in Early Modern English, Period I [1500–1570] and Period II [1570–1640] (Ogura and Wang 1996). The highest figures for 3sg. ‐s in Period I in their corpus are found with have (6 examples) (vs 366 hath!), know (5), pray (4), go and think (3), make and keep (2), and give, take, stand, need, pay, tell, follow, answer, each with 1 token. In Period II, the figures for 3sg. ‐s on the frequent irregular verbs increase: say (84), have (34), come (34), make (21), go (16), think (11), etc. At the same time, ‐s on regular verbs failed to increase, or even decreased: pray (1), pay (1), answer (2), pass (2), etc.
(p.107) That there was no direct continuation from Scandinavian is clear from the general absence in the early period of 3sg. ‐s on verbs borrowed from or influenced by Scandinavian. In the corpus of Ogura and Wang (1996), apart from give (Period I: 1 example of ‐s; Period II: 9 examples) and take (I:1; II:8) other verbs with likely Scandinavian links (want, dwell, call, die, get, quicken, cast, raise) never take ‐s in Period I (but note duellez in the LGrC entry (1a) from 1428, prior to their Period I), and rarely admit it in II: get (6), want and dwell (5), call (1), the rest zero.
The crucial point by Hodges in (2) that many people were saying ‐s/‐z but writing ‐(e)th is ample testimony that textual statistics are of little value except as a general indication of the types of verbs that were more likely to show 3sg. ‐s. More crucial evidence is that of the London Grocers' Company records, in which 3sg. ‐s correlates with other northern features (whilk, give, etc.) while ‐(e)th correlates with other southern features (which, yaffe ‘gave’ , etc.). This confirms that ‐s did in fact spread from the north. At the same time, however, the oldest distribution shows that ‐s had nothing to do with Scandinavian, but spread from 2sg. to 2pl. and 2pl. imperative, then to the rest of the plural, and, last of all, to 3sg. The most frequent forms with 3sg. ‐s were is and has (hæfis ‘has’ , næfis ‘doesn't have’ ), from which ‐s diffused first to other irregular monosyllabics (especially those ending in a dental), then to regular monosyllabics, the status attested in the London Grocers' Company records, then, finally, over the course of several centuries, to all verbs. The change, not surprisingly, was far more advanced than is reflected in the canons of writing.
In summary, the cognitive interconnections among irregular monosyllabics facilitated the diffusion of ‐s in the 3sg. from is and has, while the English‐specific strength of the third person retarded the diffusion (except in the north) to regular and polysyllabic verbs.
The replacement of ‐þ by ‐s had an interesting syntactic consequence. In Old English, third‐person subjects were more frequently null than first‐ or second‐person subjects. Van Gelderen (2010: ch. 2) suggests that the third person could have i‐phi features (§3*.8.1) and therefore not require checking (cf. §5.11.1). While there was apparently no difference in the use of null subjects with either 3sg. form ‐þ or ‐s in Northumbrian (van Gelderen 2000b), the victory of ‐s in London ca.1660 correlates with the final loss of null subjects (van Gelderen 2000a) and, shortly after that, the loss of V‐to‐I raising (Lightfoot 1999: 161–7; Miller 2002a: 277, 379). This implies that ‐s changed its (p.108) phi feature to uF. On the anomaly of Agr marking only in the third person, see §6.12.
4.3 English deadjectival ‐en
The main function of English ‐en is to create deadjectival verbs, that is, verbs derived from adjectives, like flatten from flat.6 This was in part a new category in English. Old English had about a dozen ‐n‐ verbs, many not deadjectival. The category expanded under Norse contact in northeast England, first in Northumbrian. Subsequently, about a dozen deadjectival ‐en verbs were transferred from or influenced by Nordic. Novel English verbs were created by semantic analogy to the loanverbs.7 The models ensured maintenance of the constraints imposed by the loanverbs: monosyllabic base, trochaic foot structure, and root‐final obstruent (stop or fricative).
Deadjectival verbs typically have a causative : unaccusative/inchoative alternation. Compare Embick's (2004b: 365f.) derivation of the metal flattened from the structure of ‘the metal became flat’ (3a) and I flattened the metal from that of ‘I made the metal become flat’ (3b). For his AG[ENT] I have substituted CAUS (§6*.6). Since ‐en occurs in both the unaccusative and causative alternants, Embick takes it to be the exponent of the fientive projection (traditional BECOME operator §§6*.5.1, 6*.6). The derivation must also contain a Degree Phrase (§§1*.10, 6*.2) because, as noted by Ramchand (2008: §2.1.2), a gap can widen but does not necessarily become wide. The DegP in flatten me wafer‐thin (3c) indicates the extent to which the final result state is realized.
In this Distributed Morphology type of account, the radical sign indicates the root being verbalized. Derivation from a root predicts that states other than adjectives can potentially take ‐en; cf. out → outen  ‘put out’ , i.e. ‘cause x to get‐to‐be [in a state of being] OUT’ ; E.Angl. uppen  ‘bring up’ . Technically, then, the formation is DERADICAL (a term used extensively in Miller 2006), but since adjectives are the primary target of this discussion, the less precise deadjectival will continue to be used.
Degree expressions in deadjectivals have been incorrectly analyzed as secondary predicates by Rappaport Hovav and Levin (2001), van Kemenade and Los (2003: 86f.), and others, but they fit Hale and Keyser's Degree projection (2005: 24f.), as in (3c), in which wafer‐thin is a degree expression analogous to ‘exceedingly flat’ . In other words, (3c) is tantamount to ‘(they) cause me to (p.110) become exceedingly/wafer‐thin flat’ . The DegreeP, then, indicates the extent to which the final result state is realized and in some languages is obligatorily non‐null.8
4.3.2 History of ‐en
Germanic had a fientive in *‐nō‐ ~ *‐na‐ (IE *‐n(e)h 2‐), e.g. Goth. (ga)waknan = OIce vakna = OE wæcnan WAKEN (intr.), which was productive only in Gothic and Old Norse (Ringe 2006: 176ff., 259f.). This originally had nothing to do with the denominal type wundian ‘to wound’ (wund ‘a wound’ ), deadjectival gladian ‘to glad’ (⇒ gladden), etc. (see Dalton‐Puffer et al. 1992–3). Standardly, ‐en‐ian was created by reanalysis due to referral of verbs like (4c, e) directly to the underlying adjective (Raith 1931; cf. Bloomfield 1933: 414; Koziol 1937: 186; Dalton‐Puffer et al. 1992–3). This was indeed a contributing factor but could not have been the sole impetus. Old English had a little over a dozen ‐n‐ verbs, a few of which appear in (4). Many are not deadjectival.
(4) Some Old English verbs in ‐n‐
a) crīstnian ‘Christianize’ [ca.890], ‘baptize’ [ca.1180 Orm], CHRISTEN [ca.1450] (derived from OE crīsten ‘Christian’ ; cf. crīst CHRIST)
b) fæʒ(e)nian ‘rejoice’ (from fæʒen ‘glad, rejoicing, FAIN’ )
c) fæstnian ‘make fast, FASTEN’ (〈 Gmc. *fastinōjan‐ HGE 95) (fæst FAST)
d) glisnian [ca.1000] GLISTEN (replaces OE glisian †GLISE ‘to glitter, shine’ )
e) op(e)nian ‘to open’ (open OPE; the adj. open is backformed from the verb)
f) tācnian ‘symbolize, represent’ (from tācen ‘symbol, sign, TOKEN’ )(p.111)
The spread of ‐en was motivated by Norse contact, as suggested in the past (e.g. Björkman 1900–2: 15ff.; Miller 2004c), but neverbefore demonstrated. The expansion first occurred in Northumbrian (well known for its early Nordic features, e.g. infinitives in ‐a); cf. (5) where, however, only (5a) is deadjectival.
(5) Northumbrian verbs in ‐n‐ [c10]
a) berhtnia BRIGHTEN (from berht BRIGHT)
b) lācnia ‘heal’ (from lǣce LEECH)
c) lysna ‘listen’ (ME lystnen replaces LIST V 〈 OE hlystan ‘hear; listen’ ; with the northern by‐form lysna compare Swed. lyssna ‘id.’ )
d) þrēatnia ‘oppress; THREATEN’ (replaces THREAT V 〈 OE þrēatian ‘urge, press on’ ) (from OE þrēat ‘throng’ )
Despite Marchand's assertion (1969: 271) that “we know nothing about the motives which led to the extension of the suffix,” Scandinavian influence is unmistakable. All of the Norse‐influenced ‐en verbs in English since c12/13 are deadjectival; see (6–7).
(6) Deadjectival verbs entering English in c12–c14
a) harden [ca.1180 Orm]; cf. OIce harðna ‘to harden’ (replaces HARD V 〈 OE heardian ‘id.’ , derived from heard [OE] HARD)
b) gladden [a1300] ‘be glad, rejoice’ ; cf. OIce glaðna ‘be gladdened’ (replaces GLAD V 〈 OE gladian ‘id.’ , from glæd [OE] GLAD)
c) quicken [a1300]; cf. OIce kvikna ‘to come to life; revive’ (replaces QUICK V 〈 OE quiken [ca.1175] ‘id.’ , from cwic [OE] ‘living; QUICK’ )
d) whiten [a1300]; cf. OIce hvítna ‘to become white’ (replaces WHITE V 〈 OE hwītian ‘id.’ , from hwīt [OE] WHITE)
e) liken ; cf. Swed. likna, Dan. ligne ‘id.’
f) Dial. slocken [a1325 Cursor Mundi] ‘quench, extinguish’ (〈 ME sloknen; cf. OIce slokna ‘to expire, go out’ , from slokinn ‘extinguished’ )
g) meeken [?ca.1380] (cf. OSwed. miukna, Swed. mjukna ‘become tractable’ )
h) loosen [a1382 Wyclif] (cf. OIce losna ‘get loose; loosen’ )
(7) More recent ‐en verbs with Scandinavian links (more in Bator 2006: 299)
a) weaken [1350 1x, then 1530+] (cf. Swed. vekna ‘to become weak’ )
b) thicken [ca.1425] (cf. OIce þykkna ‘to make or become thick(er)’ )
c) slacken  (cf. OIce slakna ‘to slacken, become slack’ )
d) redden  (cf. ME rudnen [a1225] ‘become red’ = OIce roðna ‘id.’ )(p.112)
As the deadjectival class began to spread under Nordic influence, many new verbs were created on native or nativized English bases; cf. (8), chronologically listed in modern spelling. More recent examples can be found below.
(8) Deadjectival verbs on native or nativized English bases
a) sicken [ca.1180 Orm] (gradually replaces SICK V [a1150] ‘id.’ ) (sick [OE])
b) blacken [a1300] (replaces BLACK V 〈 OE [a1225] ‘id.’ ) (black [OE])
c) darken [a1300] (replaces DARK V [LOE] ‘id.’ ) (dark [OE])
d) lighten 2 [a1300] (replaces LIGHT V 〈 OE līhtan ‘id.’ ) (light [OE] ‘bright’ )
e) lessen [ca.1350] (replaces LESS V [a1225] ‘id.’ ) (less [OE])
f) lighten 1 [a1375] (vs LIGHT V 〈 OE līhtan ‘id.’ ) (light [OE] ‘not heavy’ )
g) soften  (replaces SOFT V [a1200] ‘id.’ ) (soft [OE])
h) sharpen [ca.1450] (replaces SHARP V 〈 OE scyrpan ‘id.’ (sharp [OE])
i) worsen [ca.1450] (replaces WORSE V 〈 OE wyrsian ‘id.’ ) (worse [OE])
j) chasten  (chaste [a1225] 〈 OF chaste  ‘id.’ )
k) straighten  (replaces STRAIGHT V  ‘id.’ ) (straight [ca.1350])
l) moisten  (replaces MOIST V [ca.1325] (OF moiste  MOIST)
m) fatten  (replaces FAT V 〈 OE fǣttian ‘id.’ ) (fat [OE])
n) cheapen  (cf. OE cēapian ‘to barter’ 〉 CHEAP V) (cheap [OE])
o) widen  (replaces WIDE V 〈 OE wīdian ‘id.’ ) (wide [OE])
p) roughen  (rough [OE])
q) toughen  (tough [OE])
r) deafen  (replaces DEAVE V 〈 OE dēafian ‘id.’ ) (deaf [OE])
s) deepen [a1605] (replaces deep V 〈 OE dīepan ‘id.’ ) (dēop [OE] DEEP)
t) flatten  (flat [ca.1320] from ON; cf. OIce flat‐r ‘flat, level’ )
u) dampen [ca.1630] (damp , etymology unknown)
v) deaden  (replaces DEAD V 〈 OE dēadian ‘id.’ ) (dead [OE])
w) plumpen  (plump  prob. 〈 MDu plomp ‘blunt, stupid’ )
x) tighten  (tight [ca.1435] 〈 thight [ca.1375] from ON)
y) brisken  (brisk ; cf. brusque and Welsh brysg ‘brisk of foot’ )
z) smarten  (smart [OE] ‘causing pain, sharp’ ;  ‘clever, capable’ )
In terms of natural classes (§4.2), it is easy to see how harden (6a) paved the way for soften (8g), and how whiten (6d), blacken (8b), darken (8c), and lighten 2 (8d) mutually reinforced one another, especially given brighten (5a). Fatten (8m) paved the way for plumpen (8w), slacken (7c) for tighten (8x), widen (8o) for deepen (8s), etc. With deaden (8v), cf. enliven  and liven . Roughly synonymous and antonymical pairs were created around the same time or slightly later.
Also contributing to the spread of ‐en deadjectivals was the fact that phonological change had rendered most of the original verbs in (4–8) (p.113) identical with the corresponding adjectives. Nevertheless, the fact remains that, without the Nordic impetus for the initial spread of ‐en deadjectivals, the homophony would have been tolerated, as it was for a long time. Observe the numerous adjectives that undergo conversion and do not make ‐en verbs, e.g. (to) warm, clean, dirty, wet, dry, wise (up), clear, etc. (see Bauer et al. 2009).
4.3.3 Non‐deadjectival ‐en
Although the derivation in (3) predicts that bases other than adjectives should admit ‐en, most nouns that take ‐en are fully categorized, as in (9), and not simple roots, rendering the derivation of these more complex.
(9) Sample denominal formations
a) strengthen [ca.1450] takes the place of *strongen ‘make strong(er)’
b) lengthen  replaces longen [Chaucer] ‘make long(er)’
c) hearten  ‘embolden’
d) heighten  takes the place of *highen ‘make high(er)’
The examples in (9) are motivated largely because corresponding ‐en deadjectivals (e.g. *highen, *strongen) are prohibited by phonological constraints (see §4*.3.4 below). Where a deadjectival can be formed, there is no reason for a denominal, e.g. broaden renders breadthen [2x: 1809, 1884] vacuous; widen preempts *widthen.
Since the analogical models under (8) maintained a preference for deadjectivals, other ‐en formations, except for the type in (9), were not very successful, e.g. backen , fainten [1612–15] ‘cause to faint’ , glassen  / glazen , shapen  ‘impart a shape to; shape’ , shiften , nighten  ‘become night’ , fruiten  ‘make fruitful’ , but piecen  ‘piece together; repair’ persists regionally.
Several formations are ambiguous as to whether they are extended verb forms or derived from nouns:
(10) Ambiguous formations
a) hasten  (extended from haste V [a1300] or built on haste N [a1300] from French)
b) frighten  (extension of fright V (OE fyrhtan ‘id.’ ) or built on fright N)
In summary, the analogical models favored deadjectival formations, but did not exclude derivatives from other states, like outen. The more complex denominals were most successful when phonological constraints prohibited the expected deadjectival.
(p.114) 4.3.4 Constraints on ‐en affixation
Among the heavy constraints on ‐en, a feature match with stative adjectives of a non‐latinate variety is required, which can be lexically stipulated. This excludes adjectives like lax, grave from making verbs like *laxen or *graven. Verbs in ‐st from French adjectives are among the only non‐Germanic roots that admit ‐en.
Affixation of ‐en in competition with French loanverbs was largely unsuccessful. For instance, richen  could not compete with the well‐entrenched borrowing enrich [Wyclif] (Fr. enrichir ‘id.’ ), and largen  had little success against enlarge [ca.1380] (Fr. enlarger ‘id.’ ). On the other hand, freshen  was built on Eng. fresh from Fr. fraîche, fem. of frais ‘id.’ Since the French verb is fraîchir ‘to freshen’ and no *enfresh was borrowed, freshen had no competition and remained. The prefix en‐ was extended beyond its Romance domain, resulting in the replacement of obsolete bolden  (from bold [OE]) by embolden  (more examples in Jespersen 1942: 368).
Verbs created before c19 drop t after f or s: glisten, hasten, moisten, soften (cf. Marchand 1969: 273). Swiften  is exceptional in keeping /t/. The suffix attaches to monosyllabic bases with highly restrictive phonological properties, illustrated in (11).
(11) Constraints on ‐en affixation
Group (i) consists of monosyllabic roots ending in an obstruent (whiten, redden). The sonority (§6.7) of root‐final consonants must be at least as low as a fricative: toughen, lessen, heighten, worsen.
Group (ii) shows that non‐trochaic feet are excluded, hence liven  but not *aliven. Apart from enliven (see below) and awaken, which is a survival from Old English onwæcnan, the only exception known to me is quieten , attested in the OED as recently as 2003. Many speakers reject this (p.115) form but Laurie Bauer informs me (p.c.) that for him it is disyllabic “with smoothing of the triphthong.”9
Constraint (iii) prohibits roots ending in a sonorant: *firmen, *greenen, *tannen, *blu(e)en. Exceptions like dullen , dimmen [1828–30] are exceedingly rare and attested one time each in the OED.10 Constructs like *easyen, *yellowen are predicted to be bad because they violate both the non‐trochaic foot constraint and the root‐final sonorant constraint. Particle roots also obey these constraints: outen, regional uppen, but *innen, *downen, *overen, *underen.
Group (iv) shows that ‐en does not attach to bases ending in a nasal plus voiced consonant: *orangen, *strongen, *wrongen. Longen [Chaucer] was replaced by lengthen (§4*.3.3). For the voicing constraint, contrast strengthen, dampen, pinken .
4.3.5 Deadjectival ‐en: constraint evolution and maintenance
The most remarkable fact of the thousand‐year expansion of ‐en is that the constraints in (11) have very seldom been violated. The non‐trochaic foot constraint was irrelevant to Old English, but came about partly by phonological reductions in Middle English and partly by Nordic influence. A glance at the Norse verbs in (6–7) reveals that not one violates the trochaic foot requirement: harðna, glaðna, kvikna, hvítna, likna, slokna, mjukna, losna, þykkna, slakna, roðna. The non‐trochaic foot constraint was very rarely violated since it was introduced, and significantly, speakers who use quieten make it trochaic. Given the optimality of trochaic feet (see Ch. 9), especially in a language of the Germanic stress‐type, this is probably not accidental. Nevertheless, the constraint is arbitrary because it is not obeyed by other suffixes, each of which observes different constraints, e.g. ‐ation does not attach to iambic bases (*distùrbátion) unless the stress can shift, as in ìnspirátion (LSDE 98); ‐ify allows up to two unstressed syllables (solídifỳ, *beaútifulifỳ, but ?beaútifúllifỳ does not seem as bad) and excludes identical consecutive onsets, as in *stiffify, *toughify (LSDE 247). And so on.
(p.116) Helena Halmari (p.c.) suggests that the sonority constraint in (11) is grounded in salience. Adjacent segments identical in sonority lose salience (see §6.7.2). This is supported by constraint (iv) that prohibits a string [nasal consonant–voiced consonant–(reduced vowel)–nasal consonant], all of which are very close in sonority. Earlier English lacked the sonority constraint, at least for the denominals; cf. OE leornian, ME lernen LEARN; OE war(e)nian, ME warnen WARN.11 Again, a look at the Norse verbs that influenced the development of the English deadjectivals reveals that all of them are built on roots ending in an obstruent (stop or fricative).
Although the constraints in (11) may be natural in some sense, they remain arbitrary (because each English suffix observes a different set of constraints) and must therefore be historical accidents. The accident in this case was that the constraints were imposed by the loanverbs and constructs influenced by their Norse counterparts.
The spread of ‐en amounts to (partial) formal renewal of a category the language once had. In Old English, denominals and deadjectivals were both characterized by ‐ian: denom. wundian 〉 (to) wound; deadj. gladian 〉 (to) glad (⇒ gladden), hwītian 〉 (to) white (⇒ whiten), etc. The reflexes of the older formations continued to exist for many centuries, usually in competition with innovated ‐en forms, as the examples in (8) above illustrate. That is, conversion and ‐en affixation (plus various other affixes: Kjellmer 2001; Miller 2006) have long been and remain in competition for adjectival bases (Dalton‐Puffer et al. 1992–3; Bauer 2006; Bauer et al. 2009).
Occasionally, there has been a split such that the causative/inchoative function takes ‐en, relegating the unsuffixed form to other uses; cf. stiffen [?a1500] ‘make obstinate’ ,  ‘make rigid’ ,  ‘make stiff’ vs stiff  †‘grow strong’ ,  †‘make stiff, stiffen’ ;  ‘cheat’ (cf. §4.10). Similarly, to rough is not the same as roughen (cf. Jespersen 1942: 357), to smart is not the same as smarten, light ≠ lighten, and so on. Jespersen (1942: 357) mentions that loose ‘set free’ is not the same as loosen ‘make loose(r)’ , slack/slake is not the same as slacken, etc. Bauer et al. (2009) cite e.g. whiten ≠ white (out), madden ≠ mad ‘to exasperate’ (supposedly US colloquial, but everyone consulted rejected it), (en)liven ≠ live, etc.
The expansion of ‐en deadjectivals was by lexical diffusion (§3.3), literally one word at a time. The spread was at least initially by lexical‐semantic analogies. The presence of one word increased the probability of another in (p.117) the same semantic field, including antonyms. Since most of the ‐en formations were created by analogy with an existing word, it is clear how the constraints in (11) evolved. Potential neologisms by their analogical nature did not violate the essential structural properties of existing constructs. One might also argue that, since the constraints are to some extent natural, once imposed, they would tend to be maintained. In any case, recent neologisms like crispen  and embiggen [ca.2000] (Kjellmer 2001: 154) illustrate that deadjectivals in ‐en continue to be derived within the constraints in (11). This is broadly consistent with Harris and Campbell's (second) generalization (1995: 102) that diffusional changes create natural classes. Perhaps more to the point, diffusion creates and/or maintains natural classes, and the data presented here have shown why that should be the case. As elegantly worded by De Smet (2008: 85), “diffusional change will be sensitive to the constraints characteristic of the linguistic sub‐system in which it operates.”
Productivity even within ‐en's very limited domain is difficult to calculate. Some deadjectivals have not survived competition with their conversion counterparts, e.g. righten [a1340] and slighten  have fallen into disuse while similar lighten, straighten, whiten remain. There is something peculiar about milden , olden , encolden [1x 1627], and colden , for which it is tempting to posit a constraint against *‐ld‐n‐. The constraint cannot be old (although none of the Norse prototype verbs have a base ending in ‐lð‐) because of (em)bolden, but a lack of salience of ‐ld‐(e)n may help explain why en‐ came to be obligatorily prefixed. Another potential case for non‐salience is blunten [1x 1615], the only deadjectival with ‐nt‐n‐.
Blithen is attested twice (1824 and 1830) and smoothen  is moribund (last occurrence in the OED: 1945), but these are the only examples with [ð] and therefore anomalous (Helena Halmari, p.c.). A (recent?) constraint against bases ending in a voiced continuant might also explain the oddity of braven [1x 1865] (but note enliven  and liven 12) and the absence of deadjectivals like *wisen ‘make wise’ .
Apart from lessen, loosen, worsen, and ‐st bases, [s] seems disfavored; cf. rare or dialectal scarcen , hoarsen , coarsen , fiercen , closen , grossen [1x 1899]. Except for gross (out), with a different meaning, these do not undergo conversion either.
In sum, there has been a gradual increase of ‐en deadjectivals over time but not without considerable retrenchment. Bauer (2001: 141) mentions the possibility of he gruffened his voice, but adds (p. 142) that ‐en is not fully (p.118) productive. In fact, the maximum productivity of ‐en was during the sixteenth century, with some eighteen (non‐Scandinavian‐influenced) deadjectival types, and only scarcen and non‐trochaic †endarken do not survive in everyday use.13
Of the seventeen or so attested types in the seventeenth century, approximately half remain in standard use.14
The eighteenth century attests a little over a handful of new types, three of which (madden, broaden, tighten) remain in common use.15
The nineteenth century witnessed a major resurgence, but four of the twenty‐three words are attested only one time, and two (dimmen, dullen) violate the sonority constraint.16 Judgments on the remainder of the nineteenth‐century innovations are very mixed. To mention two extremes, for me (from Eastern Pennsylvania), only smarten, liven remain in everyday use, but Bernadette Russo (from Brooklyn, NY) informs me that she uses coarsen, tauten, smarten, harshen, steepen, braven, tarten, liven, and neaten. Still others do not react negatively to the newer formations but never use them. Recent neologisms like crispen , embiggen [ca.2000], and Bauer's gruffen are accepted in varying degrees, embiggen generally evoking the most negative reaction except among those who encountered it on an episode of The Simpsons . Gruffen is generally ranked the best of the three by speakers consulted.
As to analysis, the derivation in (3) is subject at least to the phonological constraints in (11). But how? It is clear that occasional violations of the constraints are permitted. Generally only adjectival and certain other monosyllabic roots ending in an obstruent undergo (3). Occasional other roots bear (p.119) a lexical mark for admitting (3). Additionally, there are many lexical exceptions of roots that meet the criteria for (3) but do not undergo it.
To summarize, it appears that ‐en is past its period of productivity from the E‐language point of view. The nineteenth‐century examples and occasional modern coinages indicate that ‐en affixation remains AVAILABLE (able to be used at any time: Bauer 2001: 205), but only for some speakers, for novel I‐language creations. Most striking is how few of the recent constructs make it to standard E‐language usage. This suggests that productivity, which is all about POTENTIAL (Bauer 2001: 41), can be relevant to only a small number of individual grammars (I‐language).
Availability can be restricted by PROFITABILITY, or pragmatic utility (Bauer 2001: 42f., 143). Although some ‐en formations would be of limited real‐world use, many would be highly useful, which should favor ‐en formations. One inhibiting factor is the strength of phonological constraints, which can be very persistent, like Spanish *#sC‐ (§1.11, end). Another is competition with conversion and the other affixes that derive causative and inchoative verbs. Beyond that, the erratic acceptance of novel forms suggests that the motivation for this suffix in the PLD (primary linguistic data) is insufficiently salient. Unless this trend reverses itself, and adult speakers begin using more ‐en derivatives to increase its salience in the PLD—an unlikely prospect at present—the suffix will never attain full productivity, even in its very limited domain, and will continue to shrink. Some speakers, myself included, have difficulty accepting novel ‐en constructs for the simple reason that the suffix has ceased to be available and accepted types are lexically calcified.
The two morphological changes discussed in this chapter have both proceeded by lexical diffusion (§3.3). Of the two, the development of ‐en is perhaps more to be expected since that involves traditional derivational morphology. What is interesting, however, is that the inflectional morpheme ‐s had the same kind of development, although inflection is typically considered to be more automatic. In the latter case, there was always some inflectional marker present, be it ‐th or ‐s. It thus differed from the expansion of ‐en which was replacing no affix at all, and ‐s eventually did become automatic. It is doubtful whether ‐en will ever be predictable even within its very restrictive domain (cf. Bauer et al. 2009).
The early lexical diffusion of every major English suffix of Latin origin can be followed in Miller (2006), where all lemmata are dated for their first occurrence in English. For instance, ‐able spread one word at a time initially, (p.120) but quickly reached a critical mass and became productive (Miller 1997, 2006: 227–32). At the outset, ‐able was borrowed from French in words like acceptable, damnable, excusable, inestimable, malleable, notable, portable, tolerable, treatable, etc. The suffix was advantaged by the fact that English had no easy way to express the same semantic content.17 Consequently, it spread quickly to native roots: believable, (un)healable, (un)seeable, (un)speakable, sellable, stretchable, tellable, unknowable, etc.—all prior to 1400. Between 1450 and 1500 another fifteen are attested, and before 1500 nearly another twenty. The result is that ‐able became productive so quickly that by 1800 one finds constructs like get‐at‐able , laugh‐at‐able , etc. (cf. Jespersen 1942: 400f., who cites un‐keep‐off‐able, un‐do‐without‐able, and several others). In this case, a privileged morpheme had nothing inhibiting its lexical diffusion or full productivity. The borrowing of able [a1338] facilitated this development (Jespersen 1942: 399).
Third‐person ‐s eventually escaped from all of its quirky phonological and morphological conditioning (roots ending in dentals, irregular monosyllabics, etc.). Given that different verb classes can take different inflectional morphemes, this was not a necessary outcome, but it simplified the verbal system—apart from the anomaly of having the 3sg. alone overtly characterized. Perhaps that was part of the conditioning. In conservative (e.g. religious) discourse, ‐eth correlates with 2sg. ‐est. By contrast, ‐s correlates with complete absence of any other person marking. It is easy to imagine sociolinguistic polarization of urban, secular ‐s from rural, conservative ‐eth. It was not until ‐s became the prestige London form ca.1660 that it rapidly became victorious.
It seems then that various sociolinguistic and grammatical factors played a role in the success of ‐s while ‐en has never had anything but analogy and lexical diffusion driving it since it was first imposed by Scandinavians learning English in northeast England. The suffix ‐able was also imposed on English (by French speakers in the south) but had the advantage of being the only way to express the capacity of a passive event being fulfilled. More simply, in its main productive function, it denotes a possible passive event. Since that could be expressed on hundreds of borrowed verbs (accept, damn, excuse, etc.), the spread to native verbs was well motivated. In the case of ‐en, English had many simple deadjectival verbs converted from adjectives that could express the causative/unaccusative alternation (clear, warm, slow, dirty, brown, wet, (p.121) etc.), not to mention other affixes with the same derivational property, so the motivation for this specific morpheme was not as compelling.
There is good evidence that morphological formatives initially spread by lexical diffusion. Factors that accelerate the spread include advantage to the grammar, i.e. general utility, and social prestige, which played a role in the spread of ‐able by the intelligentsia. An inhibiting factor would be lack of social prestige. Sociolinguistic and grammatical neutrality, as in the case of ‐en, allow for nothing major to happen over long periods of time.
(1) I am indebted to Laurie Bauer and Helena Halmari for extensive comments on this chapter.
(2) The final ‐t is from ðū ‘thou’ (Brunner 1965: 271). For the parallel change in Old High German and Modern Bavarian, see Fuß (2005: 162ff.). The Tatian, for instance, attests ‐s, ‐s thu, ‐st(t)u, ‐st.
(4) Assimilation of 2sg. ‐z and 3sg. ‐þ to liquids and nasals entailed the identity of 2sg. and 3sg., generalized to the present indicative of all verbs in West Norse, but to all tenses and moods in Old Swedish (Kuryłowicz 1960 : 81–3).
(5) Keller (1925: 86) finds this generalization to the 2pl. remarkable. Since pre‐Old English had a single form for the entire plural, the very presence of a differentiated 2pl. is likely due to Scandinavian influence. While the diffusion of ‐s occurred via the second person and imperative forms, it spread quickly to the entire plural, thus reinforcing the lack of differentiation in the plural.
(6) This is an elaborated version of a talk given at Sam Houston State University (Miller 2009b). Thanks to Helena Halmari, Rob Adams, and members of the audience for helpful comments and suggestions.
(7) A reader objects to the long interval between the period of Scandinavian settlers in northeast England (who assimilated to the English during c11/12) and the later occurrence of putative loanwords. Because of the language mixture in the area, loanwords and hybrid formations enter the general vocabulary, many first recorded centuries later (see §5.11.3 and Bator 2006, 2007). Loanverb is thus shorthand for a form produced by convergence. As stated by De Caluwé‐Dor (1979: 680), “Many…words have no clear etymology because they cannot possibly be the survival of the corresponding OE words; but at the same time they are not plain Scandinavian borrowings either. This shows how thorough the assimilation has been.” I assume, then, for instance, that Eng. (to) †hard and Scand. harðna mutually influenced one another to yield harden.
In French, resultatives supposedly behave like depictive adjuncts (Legendre 1997), on which see Cormack and Smith (1999). This is because French has no true resultatives (ii‐a) but only pseudo‐resultatives (ii‐b) (Levinson 2006). The lack of concord in (ii‐b) renders it interpretable (Charles Brasart, p.c.).
‘I tied the laces of my shoes very tight’
(9) In southern US English, disyllabic quieten is also common, as Rob Adams and several others inform me. As a disyllabic, of course, it does not violate the trochaic foot requirement. There remains the potential problem that ‐en should not attach to a disyllabic base like quiet. If it did, forms like *viv(i)den, [izyən] for *easyen, etc. should in principle be good as well. At least in southern US English, there is no problem because quiet is monosyllabic. Thanks to Rob Adams also for suggesting to me an advanced Google Book Search which turned up 1800 citations of quieten.
(10) Dullen, however, is quite common in southern US English, as Samuel Unger informs me. It is clearly motivated by sharpen, given his context of scouts bringing him their “dullened knives to sharpen.” Such analogical formations may eventually relax (some of) the constraints. An advanced Google Book Search of dullen turned up 890 tokens.
(11) On the surface, German wärmen ‘to warm’ (cf. warm ‘warm’ ) does not observe the sonority constraint either, but it is a different kind of formation; cf. Goth. warmjan ‘to warm’ 〈 Gmc. *warm‐jan‐ (HGE 449), presumably with a reflex of the Indo‐European causative suffix *‐éye/o‐ (IEL 173).
(12) Enliven is not parallel to embolden because the en‐ of enliven has aspectual properties like the up of liven up; note the conspicuous absence of *enliven up.
(13) English deadjectivals [c16]: sadden [a1500], stiffen [?a1500], shorten , straiten , chasten , bolden  (and embolden ), straighten , moisten , fatten , sweeten , ripen , cheapen , widen , roughen , toughen , scarcen , †endarken , deafen .
(14) English deadjectivals [c17]: milden , deepen [a1605], slighten , †embrighten , greaten [1x a1375; 1614+], blunten [1x 1615], encolden [1x 1627], flatten , dampen [ca.1630], enliven , smoothen , swiften , biggen , justen , deaden , plumpen , freshen .
(15) English deadjectivals [c18]: olden , madden , broaden , tighten , hoarsen , quieten , brisken .
(16) English deadjectivals [c19]: coarsen , meeten  ‘make meet or fit’ , louden , tauten [a1814], smarten , harshen , blithen , dimmen [1x 1828–30], fiercen , dullen [1x 1832] (but see §4*.3.4), stouten , largen , steepen , closen , colden , braven [1x 1865], richen , laten , tarten , liven , pinken , neaten , grossen [1x 1899]. Jespersen (1942: 355) also cites palen as a nonce word from this century, but it does not appear in the OED.
(17) There were of course various semantic values for ‐able derivatives both in borrowings and in native formations (details in Miller 1997, 2006), e.g. understandable was active in Wyclif; the passive occurs first ca.1450. For simplicity, this section reviews the spread of only the passive use of ‐able.