Hitherto, there has been no complete morphological account of CFs explaining their origin and measuring their productivity in English, or their profitability in the coinage of new words. Relevant studies in the past focused on one or the other type but did not provide a thorough description of the whole category, nor did they provide a theoretical framework where all types could be accommodated in spite of their different characteristics. This work has shown that CFs are neither affixes nor compound constituents, but a necessary morphological category deserving attention and fine-grained qualitative and quantitative investigation.

In particular, the neoclassical, abbreviated, and secreted CFs analysed in Chapters 4–6 and the splinters discussed in Chapter 7 as potential CFs confirm the transitional nature of the formations obtained through them. The present study has substantiated that CFs are neither prototypical of derivation nor of compounding, but are peripheral representatives of both subcomponents of word-formation. In other words, they exhibit features of affixes as well as compound constituents, but do not share all the characteristics of either of them. Their peculiarity is that they can combine with either bound or free morphs, and they can even combine with one another. Hence, they represent an independent morphological category within word-formation, although a heterogeneous one, with its own locus within the morphological ecosystem of modern English.

The corpus-based quantitative analysis conducted here has shed some more light on their morphological productivity and better emphasised the demarcation between CFs and either compounding or derivation. The main finding is that CFs cover a broad spectrum of word-formation processes that range from compounding (the neoclassical category) to shortening (the abbreviated category), and can even involve a reinterpretation and level of abstraction that are typical of affixation (the secreted category). The origin and diachronic evolution of CFs also involve the blending process, in all its various patterns combining an initial and a final splinter, or a splinter with a full word. Therefore, splinters can be viewed as CFs at their birth. However, unlike blends, which display different formal patterns, CFs are more regular and predictable, both formally and semantically.

Chapter 4 has specifically demonstrated that neoclassical CFs constitute the most ancient category: historically, most of the forms analysed belong to the mid-twentieth century and are attested earlier than abbreviated and secreted forms. Their application is mainly in specialised sectors, ranging from palaeontology (ichno-) to linguistics (lexico-), from (bio)chemistry (un-, polydeoxy-, polyribo-, -mer, -ogen, -valent) to physiology (synapto-), and medicine (-penia). Many such forms are used in the names of units of measurement (e.g. exa-, nano-, peta-, tera-, yocto-/yotta-, zepto-/zetta-), nano- being the most productive CF among them.

From a morphological standpoint, neoclassical CFs have been considered compounds in the literature, owing to their origin from neoclassical words. To support this claim my analysis has verified that they often come from Greek or, less commonly, from Latin words and display much more lexical density than affixes. They chiefly combine with one another (e.g. leuco- + -penia, sarco- + -penia, poly- + -mer), as typically happens with compound constituents, but can also be added to full words (e.g. lexico- + grammar, nano- + dust, high + -valent), and even to numerical bases (e.g. 23 + -valent, 7 + -valent). On the one hand, like compounds, neoclassical CF combinations accept linking elements, such as an -o- connective (e.g. nan-o-watt, lexic-o-logy), and the CF can be stressed (nánowatt, nánodevice). Like other compounds, neoclassical combinations can accept solid or hyphenated spelling (e.g. tera(-)byte, tera(-)flop, tera(-)hertz), yet spelling as one word is the preferred option. On the other hand, neoclassical CFs are predominantly initial (15), even if they can also occupy a final position (4), but their position is fixed, unlike that of compound constituents. Like affixes, they do not occur as free morphemes and their meaning is abstract and generalised.

Chapter 5 has instead validated the hypothesis that the category of abbreviated CFs is more affix-like due to their bounded status and invariable position, either initial, like prefixes (dino-), or final, like suffixes (-lect). Since they are obtained by shortening, they mainly correspond to word portions and are not found alone. One may object that some of them also have an independent status as clippings of full words. This is the case of Brit-¹, dino-, and cyber-, which, as diachronically investigated, have different histories. Since Brit and dino are attested much earlier than the respective CFs (Brit was used about thirty years before Brit-¹ and dino was attested fifty years before dino-), we could claim that their use in combinations is closer to the composition process occurring in compounds, which are made up of two autonomous words. On the other hand, cyber is obtained by conversion from the ICF cyber-; in other words, it has become independent because of the productive and repeated use of the CF in combinations. These are, however, rare cases of CFs corresponding to free morphemes, as in general CFs are bound and non-independent (see (a) and (j) in Table 2.1).

Two of the abbreviated CFs analysed – i.e. atto- and femto- – are very close to the neoclassical type, first, because they are borrowed from foreign (Danish/Norwegian) words (i.e. att(en), femt(en)) and, second, because they accept linking elements (att-o-gram, att-o-second, femt-o-gram, femt-o-second). Their meaning having to do with units of measurement also reminds us of other neoclassical CFs, such as nano- or tera-. However, their origin is not Greek or Latin; hence, I decided that ‘abbreviated’ was a better label for them. Yet, we could claim that they belong to the borderline between the two categories. Moreover, like neoclassical CFs, abbreviated CFs are predominantly initial (15), with only a third (7) occupying a final position.

The meaning of abbreviated CFs is generalisable, but it entirely relies on the meaning of the full form from which the CFs are shortened (digi- represents ‘digital’, dino- stands for ‘dinosaur’), rather than involving the same abstraction as affixes. Some abbreviated CFs are ambiguous between two meanings (e.g. econo- stands for both ‘economic’ and ‘economical’), while others are formally identical but represent two different separate meanings: e-¹ developed first as an abbreviated form standing for electric (e-bike, e-scooter) but, more recently, e-² has substituted electronic in several formations (e-book, e-friend), becoming even more productive than the previous form.

As for the application of abbreviated CFs, some of them are specific to some sectors: e.g. cyber-, digi-, and e-² belong to information technology or technical fields; econo- pertains to economics; atto- and femto- belong to scientific fields; lamino- is predominantly medical; nega- and petro-¹ are mainly used in the context of energy and energy sources; -lect is used in linguistics and dialectology; and -olol, -onium, and -ylidene appear in the contexts of pharmacology and chemistry. Others, by contrast, may adapt to the context and vary their application depending on the element (either word, word part, or CF) with which they combine: e.g. dino- may be used in palaeontology and geology in words such as Dinosphere or Dinoseum, but loses its specialised flavour in other words which denote the world of dinosaurs (dinoverse, dinoworld, dino-fan, dino-phile, dino-obsessed), and it may even become jocular when used in the language of videogames (dino-war, dino-alien, dino-action). The transitional character of dino- is confirmed by its application to a variety of bound (-phile, -sphere, -verse) and unbound morphemes (action, war, world). The combination dinoseum could even be analysed as a typical AD blend, i.e. dino(saur) + (mu)seum, yet the productivity of dino- shows that its development from splinter to CF has already occurred.

The most interesting category has turned out to be the secreted CFs analysed in Chapter 6. Secreted CFs are attention-grabbing from many viewpoints. First, they are the most numerous among the three categories explored, and, unlike the others, FCFs (15) overtake ICFs (12). Since in English the right position is generally occupied by the syntactic and semantic head in compounds, but also by suffixes, which are class-changing elements within affixation, FCFs generally determine the word class and carry the semantic weight of the combinations. Their modifiers, indeed, are often replaced by a co-hyponym in paradigmatic substitution (see paradigmatic morphology in § 2.2.3 and surface analogy vs analogy via schema in § 2.4.8). However, FCFs bear higher semantic weight than suffixes: in the morphological resegmentation process (e.g. econom-ics → eco-nomics), the CF -nomics is charged with more semantic density than the suffix -ics. Hence, reanalysis has the effect of placing more semantic weight on the second element (i.e. the FCF), which is closer to a compound constituent than to an affix.

Second, there are many splinters that are potential secreted forms, and the fact that their development is especially concentrated in the last decades makes us suppose that they represent an expanding (open class) category. Like secreted CFs, splinters are more commonly final than initial forms. Among final splinters, I have indeed identified cases that are potential CFs, not only for their repeated use in new formations but also for the secretion process that intervenes in their shift from mere abbreviated form to reinterpreted form: e.g. -ercise (§ 7.1.7), -lish (§ 7.1.10), and -(t)arian (§ 7.1.11) have acquired a more specific meaning than their original forms exercise, English, and vegetarian, discarding some semantic elements and keeping others. More specifically, -ercise is used for ‘physical (or non-physical but often strenuous) activity involving —’, -lish indicates ‘a language variety combining elements of English with another language’, and -(t)arian denotes ‘someone whose diet is restricted to —’.

Third, from the semantic viewpoint, it is also interesting to note that many secreted forms have undergone a specification rather than a generalisation process as compared to the full words from which they are created. As a consequence, their use remains more focused on specific domains and semantic areas: e.g. agit-, alterna-, Brit-², and -adelic mainly apply to music, literature, and arts; digi-² is ambiguous between music and technology; hover- pertains to the vehicles domain; m- is related to mobile electronic media and devices; maxi- and midi- are mostly used in fashion; and -nomics naturally pertains to the politico-economic sector.

Some secreted form combinations originate in the language of the news or magazines: e.g. words ending in -ati/-erati are often used in magazines (glitterati, glamourati), -gate and -nomics are sometimes used in the news to refer to scandals (Camillagate, Sexgate) or foreign policy (Trumponomics, Brexinomics), whereas -orama may be used in magazines or advertising to refer to stores’ names, such as Burgerama or Bookarama. Still other formations are jocular or humorous new coinages ending in -gram (nastygram, strip-o-gram), -(i)stan (Absurdistan, Facebookistan), -rific (stuporific, Twitterific), or -zilla (sisterzilla, Trumpzilla). Hence, pragmatic factors encouraging the profitability of secreted CFs range from the attraction of readers’ attention to jocularity, punning, and humour.

By contrast, -verse is more versatile and applies to different contexts, from literature (Bronteverse), films (Potterverse), and TV series (Whoniverse), to music (Bowieverse), politics (Trump-verse), the Internet (Google-verse, Twitter-verse), and the media (media-verse, webverse), as well as to more general domains (foodiverse).

The same source word, robot, has even provided two different secreted forms: robo- is attested in a number of formations with the sense ‘a/an — resembling a robot, especially in being resilient, emotionless, or futuristic’, while -bot² is used in information technology to form nouns denoting ‘a type of automated program or (internet) software, especially one which searches out information’. The two secreted CFs also differ from the abbreviated CF -bot¹ found in a number of nouns denoting ‘a type of robot or automated device’. The same happens with the splinters docu- and -umentary, both coming from (and standing for) documentary.

From the formal viewpoint, it is noteworthy that some of these CFs are also used as free morphemes. For instance, maxi and midi are obtained by conversion from the secreted forms and are also used as a noun or an adjective, the former referring to ‘a maxi-skirt or maxi-dress’ and the latter for something ‘of an intermediate length or size’. I have also found cases, such as must-, -babble, and -think, in which the CF is shortened from a compound word base. These are borderline with affixoids, because they are free-standing words, but, unlike affixoids, they are obtained by shortening and reinterpreting the full words must-have, psychobabble, and doublethink. However, they are closer to compounds than to affixations. By contrast, -gate happens to correspond to an existing word gate, but it has lost connection with it completely. Indeed, Watergate is the name of the building in Washington DC that was burglarised in 1972 by people connected with the Republican administration. Hence, it is considered a compound only etymologically (from water + gate ‘channel for water’), but not synchronically.

Secreted CFs are also fascinating because they are added to an array of different bases, as observed in Section 6.1 for Franken- and -zilla, while -orama has even extended its use: initially, it combined with initial elements of ancient Greek origin (georama), but it was later also added to English initial elements (Motorama), and recently to a range of categories, including nouns (Fun-o-rama), verbs (spend-o-rama), personal names (Elvis-a-rama), and even abbreviations (CD-rama).

In -(o)rama and -(o)nomics, the linking element fluctuates between -a- and -o- (vote-a-rama/vote-o-rama, bowl-a-rama/bowl-o-rama, Trumpanomics/Trumponomics), rarely -e- (Fruiterama, Trumpenomics); hyphenated and solid spelling also alternate (Food-a-rama/Foodarama, strip-o-rama/striporama, Roger-nomics/Rogernomics).

Lastly, the most remarkable aspect of all is the analogical nature of these combinations, mainly formed after series. Consider, for instance, all formations ending in -nomics. The series of model words consists of formations whose first part is the name of a US president (i.e. Reaganomics, Clintonomics, Bushonomics, Carternomics). Thus, analogical formations are first formed by substituting the base with the name of other presidents (Obamanomics, Trumponomics). However, this series has also extended to other famous, mainly political leaders, not necessarily American (Gorbanomics, Putinomics, Thatchernomics, Maonomics, Gandhinomics). The most recent formations, finally, do not have a personal name as their base. Many new words, often occasionalisms, are related to the Coronavirus 2019 pandemic: Coronavirunomics, Corona-nomics, pandenomics (from pandemic), Lockonomics (from lockdown referring to ‘mass quarantines and stay-at-home orders during the COVID-19 pandemic’), and vaccinomics (from vaccine). By contrast, in relation to the Brexit deal establishing the withdrawal of the United Kingdom from the European Union, words such as Brexinomics and Boris-onomics have been coined. This means that the productivity of -nomics has encouraged profitability not only in terms of novel words, but also in terms of new bases, with a consequent meaning extension from ‘the economic policies of a US president’ to ‘the economic policies of a political leader, or related to a worldwide event or a country’s deal, etc.’. This also suggests, or confirms, that CFs rely highly on paradigmatic morphology (see Bauer et al. Reference Bauer, Lieber and Plag2013) and that the semantic factor encouraging the formation of CF combinations is the possibility of paradigmatic substitution of the first element with a synonym, a co-hyponym, or a semantically related word.

Semantically, my analysis has confirmed that ‘there is a cline from most semantically contentful to least semantically contentful bound morphemes’ (Bauer et al. Reference Bauer, Lieber and Plag2013: 486; see also Bauer Reference Bauer1998; Prćić Reference Prćić2005, Reference Prćić2008; Amiot & Dal Reference Amiot, Dal, Booij, Fradin, Guevara, Scalise and Ralli2007). At the more contentful end we have neoclassical CFs, which are close to independent morphemes in meaning (i.e. like compound constituents, hence the label ‘neoclassical compounds’ for their formations). Neoclassical CFs such as lexico- ‘lexical and —’ or -valent ‘to be worth’ have what Bauer (Reference Bauer1983: 215) refers to as ‘a higher density of lexical information’ than affixes. At the less contentful end we have what we call affixes. Yet, between the two extremes we can classify other bound morphemes, such as abbreviated (e.g. porta- ‘portable’, -tainment ‘entertainment’) and secreted CFs (Franken- ‘genetically modified —’, -nomics ‘the economic policies of a political leader’), which display a high density of lexical information and semantic weight. In the case of secreted CFs, semantic weight is increased thanks to morphological resegmentation or reanalysis: e.g. econom-ics is reanalysed as eco- + -nomics, with a formal shift from affix to CF, and a consequent higher density of lexical information.

From a quantitative viewpoint, the analysis of type frequency has shown that, among neoclassical CFs, nano- is the most frequent ICF, and -ogen is the most frequent FCF. Within the category of abbreviated CFs, initial cyber- represents the most frequent CF, followed by e-², while -bot¹ is the most productive final form. Among secreted forms, the ICFs Franken-, must-, and robo- all have a high type frequency, while -gate is the most frequent FCF. Lastly, even if splinters are not as frequent as established CFs, some of them are more stable, especially initial docu- and net-, and final -umentary and -zine.

From a theoretical standpoint, the transitional character of CFs has been amply confirmed by corpus data, showing that CFs create formations that range on a continuum from compound-like to affix-like. They can indeed be situated at the interface between two subcomponents of morphology, as some are non-prototypical representatives of compounding and others of derivation:

• - Neoclassical CFs, such as nano- and -ogen, are non-prototypical representatives of compounding because (1) they are allomorphic variants of classical (Greek or Latin) words and hence correspond to free morphemes; and (2) like compound constituents, they can combine with full words (nanotechnology) or even with one another (hydrogen). Yet, they also exhibit properties of affixation, being typically bound and having a fixed initial (nano-) or final position (-ogen) (cf. rare exceptions like morph- vs -morph).

• - Abbreviated CFs, such as e- and -bot, are non-prototypical representatives of affixation since (1) they are abbreviated from words (electric/electronic, robot), thus having higher lexical density than affixes, and (2) they are added initially or finally to full words (e-car, e-commerce, chatbot). However, they can also combine with one another (digi-bot) and rarely also correspond to a free morph (Brit-, cyber-), thus being transitional with compounding.

• - Secreted CFs, such as alterna- and -nomics, are also non-prototypical of affixation because (1) they are bound: in particular, they are abbreviated from words (alternative, economics) and added to bases (alterna-funk, Clintonomics); and (2) they involve abstraction due to secretion (‘a form or genre of popular music considered unorthodox or outside of the mainstream’, ‘the economic policies of —’). However, reanalysis (alternat-ive vs alterna-tive, econom-ics vs eco-nomics) confers higher lexical density or semantic weight on CFs, thus situating them at the interface with compounding.

The dynamism of CFs also lies within each of the above categories, which displays properties of either compounding or affixation depending on the bases to which CFs are added. Thus, for some CFs (e.g. dino-) either a compounding analysis or a derivation analysis is appropriate, and often a superposition of both. Moreover, some CFs are transitional between abbreviated and secreted (e.g. Brit-¹ vs Brit-², -bot¹ vs -bot²), showing their progressive diachronic development. Nevertheless, diachronic evolution is primarily evident in the shift from blend splinter to either abbreviated (e.g. docu-) or secreted CF (e.g. -(t)arian). These considerations confirm that CFs consist of an open class of elements destined to expand, to include novel splinters, and to form new series and new words to be integrated in dictionaries and in the English lexicon at large.