4.1 Old English

In our OE data, -ish derivatives amount to a total of 3,076 tokens, which constitute all in all 194 types. As figure 1 shows, all bases are nominal. They include both proper names and common nouns, and the latter can be simplex as well as compound. The proper names primarily denote geographical areas or ethnic groups (demonyms and ethnonyms, respectively).

Figure 1. Distribution of OE -isc derivatives across types

The lion's share of types (i.e. 74 percent) is represented by ethnonymic and demonymic formations, based on the names of countries, areas, cities and the like, and in some cases, persons (englisc ‘English’, lundenisc ‘from London’, madianitisc ‘Midianite; descendant of Midian’). This indeed is the well-known kind of -ish derivation, widespread in the Germanic languages, representing the original use of the affix.

Most forms with more than a few attestations are found both as adjectives and in nominalized uses representing either a person or persons of a certain nationality or origin, or their language (se denisca flota ‘the Danish fleet’, se denisca ‘the Dane’, on denisc ‘in Danish’). The fact that 55 of such forms identified in the corpus are hapax legomena should not be overinterpreted in the earliest period: it is hardly surprising that englisc ‘English’, denisc ‘Danish’, grecisc ‘Greek’, iudesic ‘Jewish’ and Nazarenisc ‘of Nazareth’ are highly frequent, whereas e.g. amalechitisc ‘Amalecite’, ethiopisc ‘Ethiopian’ and nyceanisc ‘of Nycene’ are each recorded only once. This merely reflects the topics dealt with in the preserved records and has little to do with any characteristics of the language at a stage when -ish had little competition as a marker of geographical or ethnic origin.

Closely related to this kind of -ish derivation is the one based on proper nouns, which amounts to six types in total and which results in adjectives denoting either a family relation (herodiasc ‘Herodian, [daughter] of Herodias’, pontisc ‘Pontian, of the Pontius family’) or an origin relating to the person in question (arrianisc ‘Arian, adhering to the doctrine of Arius of Alexandria’, davidisc ‘by David’).

The category of greatest interest, in view of -ish uses at later stages of the language, are the forms derived from nominal bases. In our data, these represent 32 types/641 tokens. With 462 tokens, mennisc ‘human’ is by far the most frequent type. As shown by the consistently mutated form of the base (spelt menn-, occasionally mænn-), this is a prehistoric formation, and richly attested also in nominalized uses (‘man’, ‘people’). Interestingly enough, i-mutation can still be found in a wide array of -ish derivatives, and also with bases of Latin origin; see, e.g., milisc ‘honeyed, sweetened with honey’, derived from Latin mel(l) ‘honey’. As can be expected, the members of the nominal category are primarily based on simplex common nouns (e.g. ceorlisc ‘churlisc’ < ceorl ‘churl’, eotenisc ‘made by a giant’ < eoten ‘giant’), though compounds are also attested (e.g. dun-lendisc ‘hilly’ < dun-land ‘hilly country’, god-spellisc ‘evangelical’ < god-spell ‘gospel’).

As for the etymological origin of the base, we find that the suffix -ish, as expected, primarily attaches to Germanic roots, but also to roots of Latin origin (e.g. puerisc ‘boyish’ < Latin puer), or Graeco-Latin bases (e.g. deóflíc ‘devilish’ < Greek διάβολος via Latin diabolus and early Germanic). Several of the base forms from Latin appear to be relatively well-established loans in OE, such as gimm ‘jewel’ (the base for gimmisc ‘jewelled, set with gems’) from Lat. gemma, laur ‘laurel’ (> laurisc ‘of laurel’) from Lat. laurus, and possibly also cristalla (> cristallisc ‘of crystal’) from Lat. crystallum.

All in all, then, there is evidence that already at this stage, English -ish was less particular in its choice of partners than many of the current adjective-forming suffixes, which, as noted above, more faithfully select bases of either English (i.e. Germanic) or Romance origin. In this respect, as Campbell (Reference Campbell1959: 219) points out, -isc, along with the agentive suffix -ere and the infinitival ending -ian, is more prone than other affixes to attach to foreign bases.

In a nutshell, although the attested OE ‘inventory’ of -ish derivatives is dominated by ethnonymic bases (74 percent), the remaining 26 percent foreshadow the variance of -ish derivatives as witnessed by hybrid formations and -ish suffixation to morphologically complex bases.

4.2 Middle English

How did the word-formation pattern of -ish derivation change after OE times? Figure 2 offers various insights: adjectives established themselves as a fully fledged new word category available for -ish derivation, verbs also appeared on the scene, and concomitantly, the overall distributions changed considerably. It needs to be emphasized, though, that the statistics are to be taken with a pinch of salt: due to the nature of the database, the numbers are admittedly neither directly comparable to nor as reliable as the frequencies reported for OE or the post-ME eras. Therefore, in the following section on ME -ish derivatives, we will restrict our discussion to type frequencies as evidenced in the data at hand, which nonetheless allows us to draw conclusions as for the further development of -ish derivation.

Figure 2. Distribution of ME -ish derivatives across types

The data extraction revealed a total of 191 types, distributed over five different base categories: for nominal bases, we again differentiated between simplex and compound nouns, and as it is commonly argued that color adjectives are the entry route for adjectival bases in general (Marchand Reference Marchand1969: 305), we distinguished between color and common adjectives. As can be seen at first glance, the share of -ish types across word categories has changed enormously in the transition from OE to ME, with nominal bases now having overtaken the group of ethnonyms. From clear winner downgraded to runner-up, ethnonymic formations take second place in ME with 37 extracted types (19 percent), not too far ahead of the newcomer, i.e. the deadjectival type. The data for ethnonymic formations corroborate Dalton-Puffer's (Reference Dalton-Puffer1996) and Ciszek's (Reference Ciszek2012) observation that -ish loses ground in this domain as the nationality-denoting function is nowadays largely expressed by other suffixes that were added to the derivational system in ME, primarily -ian, -an and -ite. Nationality-denoting -ish, the only Germanic suffix used with ethnonyms (cf. Dixon Reference Dixon2014: 268), is by and large only used with those formations that have existed since OE but has ceased to be available for creating new ethnonymic adjectives.Footnote ¹⁵

Similarly, proper noun bases decline drastically in ME, with just two types attested in our data, namely Magdalenish and Pilatus Pontiuisce. These formations, which are probably remnants from OE Bible translations, bear a locational/relational sense (‘from Magdala’, ‘belonging to the Pontius family’) that would nowadays be absent with -ish forms derived from proper nouns. As shown below, proper noun bases disappeared completely for some time before they experienced a comeback in the nineteenth century.

A closer look at the 109 simplex noun bases reveals that this type of denominal -ish forms did indeed thrive in ME. In this respect, the data at hand do not confirm Dalton-Puffer's (Reference Dalton-Puffer1996: 173) intuition that with ‘common noun derivatives being rare in Old English … our data would indicate that the same was the case in Middle English.’ While a great deal of the OE ethnonymic formations did not survive into the ME period (e.g. affricanisc, bulgarisc and grecisc became obsolete), the vast majority of denominal ones did – and apart from that, a great number of new coinages can be found, with an observable extension of the nominal class. Whereas most OE formations denoted animate beings, ME formations go beyond persons (e.g. foolish, knavish, thievish) and animals (e.g. doggish, foxish, swinish) in that they also comprise a wide array of inanimate entities denoting substances, materials or shapes/appearances:

1. (16) When [blood] is o þe vesy it is mare lumpryssh & clumpryssh & cruddyssh & spottyssh … (a.1425) (MED)

   ‘When blood is in the urinary bladder, it is more lumprish & clumprish & cruddish & spottish …’

Interestingly enough, for many of the -ish derivatives we also find variant forms in -y/-i, e.g. cruddi vs cruddyssh in (16) or cloudi vs clowdyssh, which is indicative of a system in flux, with functionally equivalent suffixes competing. In this regard, it is also no surprise that very many attested ME -ish forms have later on been replaced by -ly derivatives, e.g. lifish vs lively, daiish vs daily, hevenish vs heavenly.

At the same time, we can observe lexicalization processes setting in, with -ish derivatives gaining an idiosyncratic meaning in some cases. A prime example is childish, which used to be perfectly interpretable as ‘childlike’ in the associative sense and which has now come to be negatively connotated as ‘not befitting maturity’ (OED); the pejorative sense has been attested since 1405. As for the claim that most denominal -ish formations are pejorative, which according to Marchand (Reference Marchand1969: 305) originates from OE nouns such as ceorlisc ‘churlish’ and hæþenisc ‘heathen’, there is indeed a conspicuous number of inherently negative bases (cf. fool, thief, knave). It remains doubtful, however, whether this suffices to claim that this ‘derogatory shade of meaning’ (Marchand Reference Marchand1969: 305) carries over to any person-denoting -ish derivatives (even though this might be appropriate for animal-denoting formations, as Malkiel (Reference Malkiel1977) claims in his investigation of the competition between -ish and -y with respect to zoonyms).

Moving on to deadjectival -ish formations, which make up the third largest group of base categories at 13 percent, we see an innovation that paves the way for the approximative sense ‘nearing, but not exactly X’ (Marchand Reference Marchand1969: 306), commonly associated with -ish formations derived from adjectives. The path of development that leads to the rise of this novel function is closely tied to the by then well-established associative sense of similitude; as Bauer et al. (Reference Bauer, Lieber and Plag2013: 313) put it,

It is often assumed that color adjectives are the forerunners in this development, dragging along all other kinds of adjectives (i.e. those that we subsumed under the label ‘common adjectives’) only later with a considerable delay. The claim is tentatively made in the OED entry for the suffix -ish, and reiterated in more concrete terms by Marchand (Reference Marchand1969), who specifies that ‘from its use with adjectives denoting color the suffix was extended to other adjectives with the same nuance of approximation (chiefly 16th century and later)’ (Reference Marchand1969: 306). This claim is definitely not corroborated by our data, which show that -ish formations derived from color adjectives (amounting to 11 types) were used at approximately the same time as others derived from common adjectives (26 types). The earliest formations based on a color adjective are whitish (1379), yelwish (1379) and reddish (1392), for those based on common adjectives, the earliest attestations are fattish (1369), palish (a.1398) and sourish (a.1398).

We might still object that the approximative sense is not as firmly established with common adjectives as it is with color adjectives. Indeed, there are some attestations where -ish seems to be pleonastic. For instance, sorwefullish is simply given as a variant for sorweful in the respective MED entry, and for adjectives such as palish and swolnish it might be argued that -ish rather serves to mark weakly entrenched adjectives more unequivocally as such, the first one being a new French loan, the second being an adjectivally used past participle. However, the following medical guideline from Liber uricrisiarum (a.1425) (Jasin Reference Jasin1983) can be taken as an early metalinguistic comment and evidence for the prevalence of the approximative sense with at least some adjectives such as thin or thick:

1. (17) For to wys how þis terme “thynnyssh” sall be takyn, undyrstand þat þare is difference betwen “thyn” and “thynnyssh”: “Thyn” […] is when þe uryn is fullyk thynne […] “thynnyssh”, when it is a party thyn or ellys menely thyn. (MED)

   ‘In order to know how this term “thinnish” shall be used, understand that there is a difference between “thin” and “thinnish”: “thin” is when the urine is fully thin […] “thinnish”, when it is partially thin or else slightly thin.’

Interestingly, nowadays such deadjectival formations are assumed to predominantly occur in journalistic prose and fiction (because they are perceived to be somewhat jocular); some of the first attestations, however, are found in medical texts.

Apart from deadjectival formations, deverbal ones emerge as another newcomer – albeit, with merely three attested types, a feeble one (and also one not destined to prosper in later periods). As Dixon (Reference Dixon2014: 293) remarks, -ish is nowadays represented ‘with a small number of Germanic verbs’ (e.g. snappish, ticklish); the earliest attestations of this type, however, are with Romance verbs, namely errish (< err), servish (< serve) and boudish (< from Old French bouder ‘sulk’ or ‘swell or protrude the lip’ (Godefroy Reference Godefroy1881: 349, s.v. bouder)). All of these early deverbal derivatives became obsolete and are not even recorded in the OED. The fact that -ish never quite succeeded in establishing itself with verbal bases is probably related to its competition with -y, which did indeed successfully conquer this domain (see Bauer et al. Reference Bauer, Lieber and Plag2013: 306).

To wrap up the ME state of affairs, -ish combines with an extended inventory of bases, now comprising ethnonymic bases, proper, simplex and compound nouns, common adjectives and color adjectives, as well as verbs. The share of -ish derivatives from ethnonymic bases and simplex nouns has changed dramatically, with the former recessing and the latter booming. Also, the diversification and spread of -ish derivatives from adjectival bases is considerable, with color adjectives and common adjectives both well represented. It does indeed seem as if -ish found ‘a new purpose in life’, as Dalton-Puffer (Reference Dalton-Puffer1996: 198) puts it, on its way to turning into ‘a popular Similitudinal suffix in Modern English times’.

4.3 Post-Middle English: from Early Modern English to Present-day English

This section outlines the development after the ME era, i.e. from the sixteenth century to PDE. Figure 3 displays the distribution of -ish derivatives over various base categories. We see that the picture becomes increasingly more varied as we approach the twentieth century, indicating that the pattern of -ish derivation is in full bloom. It needs to be noted that, at this point, we decided to eliminate any nationality-denoting -ish derivatives from the tally. The reason for this is basically twofold: on the one hand, the type–token ratios reported on in section 5 would not have been a meaningful rate if we had kept the large numbers of ethnonymic adjectives in; and on the other hand, these large numbers would not have indicated productivity in any way because, as elaborated in section 4.2, -ish ceased to be available for creating ethnonymic adjectives once it had lost out to its Romance rivals.

Figure 3. Distribution of -ish derivatives across types from the sixteenth to the twentieth century

Overall, there is a sharp increase in -ish types already as of EME, with a true explosion in types as of the twentieth century. Still, the base categories that have been well established since ME continue to play a significant role, with denominal formations constituting the most dominant class across the board, and deadjectival formations defending their second place. The nineteenth century is particularly interesting in that deadjectival formations are almost as well represented type-wise as denominal ones, when, for the first time, their shares differ from each other by a mere 12 percent. Denominal formations from complex nouns continue to be comparatively sparse, even though there is some new input from derived nouns as of the eighteenth century (lawyerish [ECF2]). Deverbal derivatives, which made but a feeble entrance in ME, never gained a strong footing, with just a small number of them becoming well established, such as ticklish (EEPF), snappish (ECF2) or peckish (BNC), whereas other novel deverbal formations such as severish < sever (NCF1) or stiflish < stifle (NCF2) did not survive the nineteenth century:

Apart from the well-established base categories, additional word classes start to allow for -ish suffixation, i.e. adverbs (uppish [EEPF], offish [NCF2], forever-ish [BNC]), quantifiers (more-ish [BNC])Footnote ¹⁶ and most conspicuously, numerals (fortyfivish [BNC]), which constitute an innovation of the twentieth century that serves to consolidate the approximative sense of -ish.

A word category that experiences a comeback as of the eighteenth century is proper nouns. After they ceased to be frequently used in ME, they completely disappear from the picture in EME, until they resurface in Late Modern English, first with just two occurrences in our corpora (Quakerish [NCF1], tartufish [ECF2]). Crucially, though, their usage is entirely different to the one in OE and ME, i.e. -ish derivatives from proper nouns are no longer relational or ethnonymic (cf. davidisc ‘by David’, Magdalenish ‘from Magdala’), but now exclusively express the associative sense of similitude. This can nicely be illustrated by means of Quakerish, an -ish form derived from the name of a religious group:

1. (18) The consequence of this was, that at a very early hour, […] all who considered themselves as belonging to that class, were seen arriving in their very becoming sad-coloured suits, with their smooth braided tresses, and Quakerish bonnets and caps. (1837, F. Trollope, The Vicar of Wrexhill [NCF2])

The group of people arriving in the small hours, however, do not belong to the Quakers, which becomes clear from an earlier passage; indeed, the word Quaker is not used elsewhere in the entire novel. In other words, the bonnets and caps only resemble those worn by members of the Quaker community, but they do not indicate a direct relation. After the first reappearances in the eighteenth century, proper noun bases become more frequent in the next two centuries, with the associative similarity sense firmly taking hold as the following examples from the twentieth century demonstrate:

1. (19)

   1. (a) a handsome chap in a sinister sort of way, Al Caponish with a dash of Dracula

   2. (b) a huge Swansea-style computer keeping a Big Brotherish eye on every car

Apart from the extension of base categories, there is also a new degree of complexity added to the picture, when -ish starts to attach to the first phrasal bases from the eighteenth century onwards ((20) is taken from the NCF2, (21) from the BNC):

1. (20)

   1. (a) Hale's was some temporary or fanciful [[fine-lady]_NP ish]_A indisposition …

   2. (b) for Miss Coe answered questions with an [[old-maid]_NP ish]_A scream

2. (21)

   1. (a) Aware that the dress had a fey, [[other-world]_NP ish]_A air …

   2. (b) It was still dark, [[middle-of-the-night]_NP ish]_A, but I'd scribble something …

In all of these cases, the scope of -ish extends over the entire phrase: in fine-ladyish, the host of the affix is not just the noun lady but the whole noun phrase fine lady. While -ish in (20)–(21) displays clitic-like characteristics, such as attaching to phrases, it still derives adjectives. A further step in the development of -ish is the clitic stage at which -ish combines with any kind of phrasal base without inducing category change ((22a) from the BNC, (22b) from Kuzmack Reference Kuzmack2007: 5f.):

1. (22)

   1. (a) and the [[forever]_AdvP -ish]_AdvP trickly sound of her high giggle

   2. (b) So, yeah, we're [[friends]_NP -ish]_NP.

For the purpose of compatibility with the literature discussed in section 6, we refer to both -ish in (20)–(21) and (22) as clitic. Cliticization may have paved the way for the development of ish as a free lexical item in the twentieth century:

1. (23) You must try to remember that some people are normal. Ish. (BNC)

One characteristic trademark of free lexical ish, which seems to abound especially in spoken registers, is that it may modify a previous conversational contribution, thus functioning like an epistemic marker or hedging device (see, for example, Traugott & Trousdale Reference Traugott and Trousdale2013: 236f.). We return to this in section 6.

All in all, the developments during the five centuries after the ME period contribute essentially to -ish obtaining the extraordinary characteristics commonly attributed to this suffix in PDE, due to a concomitant gain of new base categories and a semantic bifurcation of the associative sense vis-à-vis the approximative sense.

6.1 Inverse grammaticalization theory

In the first strand, the diachronic journey of -ish from derivational suffix to clitic and ultimately to free lexical item is taken to be indicative of an inverse grammaticalization process. Both Kuzmack (Reference Kuzmack2007) and Norde (Reference Norde2009) focus on the inverse grammaticalization cline that -ish has been moving along since OE, with Kuzmack (Reference Kuzmack2007) referring to the process as antigrammaticalization and Norde (Reference Norde2009) as degrammaticalization.Footnote ¹⁹ Labels aside, both Kuzmack (Reference Kuzmack2007) and Norde (Reference Norde2009) argue that the development of -ish from a derivational suffix to ultimately a free lexical item is accompanied by resemantification, i.e. an increase in semantic content, with the suffix developing from a purely formal adjectivizer with a nationality/ethnicity/origin-denoting sense (OE egiptisc ‘Egyptian’, cristallisce ‘made of crystal’) into a derivational suffix expressing association/comparison (OE hæðenisc ‘heathen’, ME shepishse ‘sheep-like’) and later also approximation/qualification (easyish ‘somewhat easy’, nowish ‘vaguely now’). The three main types may be labeled ish ₁, ish ₂, and ish ₃ ((24) adapted from Kuzmack Reference Kuzmack2007: 1):

1. (24) ish ₁  expressing nationality/ethnicity/origin, i.e. ‘of X origin, made of X’

   ish ₂  expressing association/comparison, i.e. ‘of the character of X, like X’

   ish ₃  expressing approximation/qualification, i.e. ‘somewhat X, vaguely X’

Our data corroborate Kuzmack's (Reference Kuzmack2007) claim that the three senses of -ish both overlap historically and persist into PDE, with ish ₁ and ish ₂ being attested since OE, the comparative sense of ish ₂ developing in ME, and ish ₃ being a later addition. The development of ish ₂ and ish ₃ coincides with a loosening of selectional restrictions. While ish ₁ is restricted to selecting nominal bases (25), ish ₂ and ish ₃ gradually allow for a categorially varied array of bases. According to Kuzmack (Reference Kuzmack2007: 4f.), ish ₂ derives adjectives from nouns and, as our data show, also from verbs (26), and ish ₃ derives adjectives from adjectives and adverbs (27). These observations are borne out with the reservation that the few de-adverbial -ish derivatives in our data, notably offish, uppish and foreverish, all instantiate ish ₃ (see (27)):

While ish ₁ ceases to be productive, ish ₂ and ish ₃ see an increase in productivity, with ish ₂ being more productive than ish ₃ (cf. Kuzmack Reference Kuzmack2007: 2; Traugott & Trousdale Reference Traugott and Trousdale2013: 235ff.; see below for discussion). Not only are their selectional restrictions loosened with respect to the category of the base, they are also loosened with respect to the format of the base: in addition to combining with X⁰ elements, ish ₂ and ish ₃, as of the nineteenth and twentieth centuries, respectively, also combine with phrasal elements. Kuzmack (Reference Kuzmack2007) analyzes occurrences of ish ₂ as in (28) and ish ₃ as in (29) as clitics:

Note, however, that ish ₂ in (30), despite taking scope over the phrases it combines with, shares with the derivational -ish the property of deriving adjectives, whereas ish ₃ in (31) does not affect the category of its host ((31a, b) from the BNC, (31c, d) from Kuzmack Reference Kuzmack2007: 5f.). Note also that, while the scope of -ish in (31a) is restricted to the adverb forever, the scope of -ish in (31b–d) could be either the Num(eral)P (7),Footnote ²⁰ the NP (friends) or the PP (on their way), respectively, or the containing constituent:

1. (30)

   1. (a) for Miss Coe answered questions with an old-maidish scream (NCF2)

   2. (b) The programme had some good out-of-the-wayish music: (BNC)

2. (31)

   1. (a) her back slender and white, and the forever-ish trickly sound of her high giggle

   2. (b) Probably arrive about 7-ish, if that's OK.

   3. (c) So, yeah, we're friends-ish.

   4. (d) Happier still, Jessica, Brian and Erik are [on their way]…ish…

According to Kuzmack (Reference Kuzmack2007), the low degree of selection of ish ₃ (in tandem with phonological strengthening) may have facilitated the later development of the clitic variant of ish ₃ as in (31) and ultimately the development of the free lexical item ish ₃ in (32) (adapted from Kuzmack Reference Kuzmack2007: 6; see also (23) above):

1. (32)

   1. (a) Can you swim well? Ish.

   2. (b) Is everyone excited? I am – ish.

All instances of the clitic variant of ish ₃ in our data involve adverbs and numerals as host categories (31a, b); the free lexical item ish ₃ occurs only once (see (23) above). The scarcity of these elements in our data may be due to the relatively early completion date of the BNC, 1993, rather than to accidental gaps. In other words, the development of both the clitic variant and the free lexical item may still only be incipient in the early 1990s. The variety examined, British English, may be a contributing factor, as may the fact that our investigation is limited to the written domain of the BNC, while ish ₃ is still primarily a feature of informal, spoken language (see also Plag et al. Reference Plag, Dalton-Puffer and Baayen1999: 220) and, of course, written representations of such style.

All in all, the trajectory of change depicted in inverse grammaticalization theory is corroborated by our empirical study.

6.2 Constructionalization theory

A second strand of theorizing concerned with the diachronic trajectory of -ish formations involves proponents of constructionalization theory. In this framework, diachronic processes of constructionalization are conceived of as resulting in new form–meaning pairings, with the rise of new meanings being concomitant with changes in syntax or morphology and ‘accompanied by changes in degree of schematicity, productivity, and compositionality’ (Traugott & Trousdale Reference Traugott and Trousdale2013: 22).Footnote ²¹ Taking the typology of historically overlapping and persisting senses of -ish in (24) as a point of departure, Traugott & Trousdale (Reference Traugott and Trousdale2013: 234) suggest subsuming semantically similar ish ₁ and ish ₂ under the general OE word-formation schema in (33), with two specific subschemas covering ish ₁ (33a) and ish ₂ (33b).

(33) is a descriptive formalization of both the selectional properties of affixal ish ₁ and ish ₂ and the interpretation of the derived adjective: both ish ₁ and ish ₂ select nouns with the noun denoting either an ethnic group or some kind of entity; the property the derived adjective expresses depends on the denotational properties of the noun selected:

1. (33) OE ish schema:  [[N_i.isc]A_j  ↔  [having character of sem_i]  property]_j]

   1. (a) Ethnic ish (ish ₁):

      [[N_i.isc]A_j↔ having character of ethnic group_i] property]_j]

   2. (b) Associative ish (ish ₂):

      [[N_i.isc]A_j↔ [having character of entity_i] property]_j]

According to Traugott & Trousdale (Reference Traugott and Trousdale2013: 234), the Ethnic ish subschema (33a) became recessive in late ME,Footnote ²² until it ultimately ceased to be productive. Our data corroborate the stipulated development of ish ₁, both for the narrow origin interpretation, i.e. ethnic group, as well as for the wider interpretation including provenance in general, e.g. OE aquinensisc ‘from the town Aquino’, and material origin/substance, e.g. OE cedrisc ‘of cedar’ or ME clayish ‘made of clay’. Surviving ish ₁-derivatives, e.g. English, Jewish, outlandish, are lexicalized with individual items displaying varying degrees of transparency, e.g. British (related to Britain) vs Cornish (related to Cornwall). Productive formations, such as Londonish (34), typically instantiate the Associative ish subschema (33b), which, after having laid somewhat dormant in OE and ME, gained ground and proliferated in EME:Footnote ²³

1. (34) He was strolling down the steep narrow street towards the sea, his hands deep in his pockets and his shirt open at the throat, very pale and Londonish, looking about him with the fond, proprietorial air of an Englishman returning to a favourite spot abroad. (BNC)

Following Marchand (Reference Marchand1969: 305), Traugott & Trousdale (Reference Traugott and Trousdale2013: 234) point out that many of the surviving early -ish derivatives of the type in (33b) have undergone pejoration. The semantic shift from ‘typical of N’ to ‘typical of and with the negative characteristics of N’, e.g. childish ‘childlike’ > ‘childish’, indicates constructional change, i.e. a change within an existing construction rather than the emergence of a new construction. For Traugott & Trousdale (Reference Traugott and Trousdale2013: 235) both the loosening of the selectional restrictions of Associative ish ₂ (33b) as well as the change from suffix to clitic, as in old-maidish (30), are instances of constructional change. Conversely, the development of approximative ish, as in 7-ish (31), gives rise to the formation of a new construction (35) and thus is an instance of constructionalization (see Traugott & Trousdale Reference Traugott and Trousdale2013: 234f.):

1. (35) Approximative ish schema

   [[A_i/N_i.isc]A_j   ↔   [having character like sem_i] property]_j ]

Since Approximative ish sees a loosening of its selectional restrictions as well as the rise of the clitic variants in (30) and (31), the Approximative ish schema (35), like the Associative ish subschema (33b), is subject to constructional change. Our data support Traugott & Trousdale's (Reference Traugott and Trousdale2013: 237) claim that both schemata have seen an increase in productivity (see section 5).

Traugott & Trousdale (Reference Traugott and Trousdale2013: 236f.) see the rise of the free lexical item ish (32) as the result of a partially grammatical constructionalization process, with neoanalysis affecting both the formal properties of ish ₃ (from clitic to free lexical item) and its meaning (from approximator to epistemic marker). Note that their postulate that the detachment of clitic ish ₃ and thus the rise of the free lexical item ish is accompanied by a semantic shift from approximator to epistemic marker is irreconcilable with Kuzmack's (Reference Kuzmack2007: 1, 8) claim that both ish ₂ and ish ₃ preserve their respective identities as ‘comparer’ and ‘qualifier’ across construction types. Despite relevant data being extremely scarce in the BNC, free lexical ish clearly has an epistemic flavor as it codes the speaker's epistemic stance towards ‘veracity of the item as a member of a particular set’ (Traugott & Trousdale Reference Traugott and Trousdale2013: 236; see also Oltra-Massuet Reference Oltra-Massuet2016).

Even though expressing epistemic stances brings free lexical ish in the vicinity of evaluative adverbs, ish is not a representative of adverbial categories (pace Kuzmack Reference Kuzmack2007: 2, 8; see also below). Traugott & Trousdale (Reference Traugott and Trousdale2013: 236) argue that free lexical ish does not hold membership in any lexical category (see also Norde Reference Norde2009: 225), but rather is a functional item in the vicinity of scaling degree elements.

As with the first strand of theorizing, our empirical study again provides substantial support for the descriptive formalization of the diachronic trajectory of -ish, as outlined in constructionalization theory.

6.3 Distributed Morphology

While Traugott & Trousdale (Reference Traugott and Trousdale2013: 236) consider free lexical ish a degree element syntactically, Oltra-Massuet's (Reference Oltra-Massuet2017) morphosyntactic account, the third strand of theorizing, rests on the assumption that all instances of affixal -ish Footnote ²⁴ represent degree operators semantically (see also Bochnak & Csipak Reference Bochnak and Csipak2014). The theoretical background for Oltra-Massuet's (Reference Oltra-Massuet2017) analysis is Distributed Morphology (see Marantz Reference Marantz1997, Reference Marantz2007), a syntactic model of morphology where roots are not specified for category until they are inserted into exoskeletal syntactic structures, so-called categorization structures, where an uncategorized root merges with a categorizing head. Essentially, in her analysis, ish _1, ish₂ and ish ₃ are spell-outs of one and the same underspecified root -ish that result from -ish being inserted into different category-defining functional head positions in clause structure, so-called phase heads, which are selected by a higher functional head endowed with the feature [approx(imator)].

Details aside, depending on the phase head that affixal -ish is inserted into, -ish derivatives from nominal and verbal bases receive one of three interpretations (see Oltra-Massuet Reference Oltra-Massuet2017: 65ff.):

1. (36)

   1. (a) manner → like N/similar or close to a (proto)typical N

   2. (b) central coincidence → have/with N

   3. (c) undefined → idiosyncratic interpretation

The interpretation in (36a) is typically found with root nouns denoting humans or animals (37a,b), proper names (37c) and nationality-denoting bases (37d) (data from BNC):Footnote ²⁵

1. (37)

   1. (a) He pursed his big mouth into such a babyish pout …

   2. (b) This, she thought with a sheepish giggle to herself, was ridiculous.

   3. (c) He was a handsome chap in a sinister sort of way, Al Caponish with a dash of

      Dracula and a smidgen of Rambo thrown in …

   4. (d) Later in the year, when we visit the area again, we heard that the Swedish Lapps had fenced off an area of the Dividal National Park, with over five kilometres  of wire.

Roots that are mass nouns typically trigger the interpretation in (36b):

1. (38)

   1. (a) I rose, broke my fast and slipped the landlord some pieces of silver which made his vinegarish face look more congenial and subservient. (BNC)

   2. (b) His temperature was high, almost feverish. (BNC)

The idiosyncratic interpretation (36c) is typically found with verbal roots (39), and the derivation process is synchronically unproductive (data from BNC):

1. (39)

   1. (a) He's the nearest I've ever seen him to snappish …

   2. (b) I don't want to know you're bound to feel a bit peckish when you wake up.

A property that is shared by all derivatives in (37)–(39) is that they are gradable adjectives and as such support degree modifiers, e.g. too, enough, slightly, a bit (data from BNC):

1. (40)

   1. (a) For one thing it was too babyish – Moses Arkwright would never do anything

      with so little risk attached to it …

   2. (b) she found her slightly feverish and put her to bed.

   3. (c) you're bound to feel a bit peckish when you wake up.

However, Oltra-Massuet (Reference Oltra-Massuet2017: 63ff.), essentially following Morris (Reference Morris1998), claims that -ish derivatives from adjectival bases, such as tallish, are non-gradable adjectives, and as such resist gradation and degree modification ((41) from Oltra-Massuet Reference Oltra-Massuet2017: 63):

1. (41)

   1. (a) *more tall-ish, *tall-ish-er

   2. (b) *very tall-ish, *extremely tall-ish

   3. (c) *John is too tall-ish to become a miner.

   4. (d) *I don't know how tall-ish Sue is.

Under her analysis, the strings in (41) are ruled out since the phase head that affixal -ish is inserted into is quantificational and categorizes ish as a degree operator on a par with degree expressions, such as comparative and superlative markers or very, too and how in (41) which Oltra-Massuet (Reference Oltra-Massuet2017: 63) takes to spell out the same head.

In her ‘descriptive typology of -ish’, Oltra-Massuet (Reference Oltra-Massuet2017: 57) lumps together -ish derivatives from adjectival bases with -ish derivatives from adverbial, numeral and phrasal bases, all incarnating ish ₃ and all claimed to be non-gradable adjectives. Unfortunately, she does not provide any data in support of this claim. In fact, the assertion – perpetuated from Morris (Reference Morris1998) – that deadjectival -ish forms ‘refuse most attempts at intensification’ (Morris Reference Morris1998: 210) defies the empirical evidence. Instances of degree modifiers with -ish derivatives from adjectives (42), and to a lesser extent also adverbs (43), quantifiers (44) and phrases (45) are well-attested across our corpora as of (Early) Modern English:

1. (42)

   1. (a) they thought that tasted a little Bitterish to the Palat (EEPF)

   2. (b) She had an Angelical Countenance, onely somewhat brownish by the Suns

      frequent kissing of it; (EEPF)

   3. (c) This has been rather smartish, Mr. Simple. (NCF1)

   4. (d) ‘Perhaps it was too prudish,’ she said repentantly. (NCF2)

2. (43) But then I was more uppish than I had ever been. (EEPF/ECF1)

3. (44) This stuff is very more-ish. (BNC)

4. (45)

   1. (a) Lord, child, don't be so precise and [old maid]ish. (ECF2/NCF1)

   2. (b) but Doleful, a very [cock-a-hoop]ish caller on his own account (NCF2)

Like Traugott & Trousdale (Reference Traugott and Trousdale2013: 236), Oltra-Massuet (Reference Oltra-Massuet2016) assumes that free lexical ish, which she refers to as propositional ish, is speaker-oriented (see also Bochnak & Csipak Reference Bochnak and Csipak2014: 435). Specifically, she argues that free lexical ish expresses the speaker's ‘lack of full commitment to the illocutionary force of the [assertive] speech act and [her] lack of full commitment to the proposition expressed’ (Oltra-Massuet Reference Oltra-Massuet2016: 312), which brings it into the vicinity of speaker-oriented adverbs, notably evaluative adverbs, such as fortunately or surprisingly, which she takes to be located in the specifier position of Sentient/Eval(uative)P (cf. Speas & Tenny Reference Speas and Tenny2003: 331ff.). The derivation of free lexical ish involves the insertion of the underspecified root into the head position of Sentient/EvalP, which is the complement of the head of SpeechActP ((46) adapted from Oltra-Massuet Reference Oltra-Massuet2016: 311):

1. (46) [_SpeechActP SpeechAct⁰ [_approx] [_{Sentient/EvalP} Sentient/Eval⁰ […]]]

In Sentient/Eval⁰, ish establishes a relation between the propositional complement and ‘some sentient mind that evaluates it’ (Oltra-Massuet Reference Oltra-Massuet2016: 311). Since ish is in the immediate scope of SpeechAct⁰[_approx], the propositional complement is evaluated as a weak assertion. Thus, free lexical ish as in (23) above, repeated for convenience in (47), reflects the speaker's lack of full commitment to the assertion that some people are normal:

1. (47) You must try to remember that some people are normal. Ish. (BNC)

Oltra-Massuet (Reference Oltra-Massuet2016, Reference Oltra-Massuet2017) aims at developing a unified morphosyntactic approach to both the affixal variants of ish and free lexical ish. Her analysis does not take into consideration the diachronic trajectory of ish, but rather presents synchronic (morpho-)syntactic snapshots of ish ₂ and ish ₃. By and large, her analysis is compatible with our empirical findings. Not tenable empirically, however, is Oltra-Massuet's (Reference Oltra-Massuet2017) claim that -ish derivatives from adjectives, adverbs, numerals and phrasal bases are non-gradable adjectives and thus incompatible with degree modifiers.

To conclude our empirically based reassessment of the theoretical accounts of -ish in sections 6.1 to 6.3, different theoretical credos and different emphases aside, both the diachronic trajectory of -ish from derivational suffix to free lexical item and the partial arcs that constitute the trajectory are surprisingly accurate and faithful to our empirical data. ‘Surprisingly accurate’ because none of the theoretical accounts discussed seems to have based their description and analysis on a large-scale empirical study.