Exploring Ways to Intensify Verbs in Swedish

2.1 Intensification as Expression of Degree

As Sapir points out, “[e]very quantifiable, whether existent (say house) or occurrent (say run) or quality of existent (say red) or quality of occurrent (say gracefully), is intrinsically gradable” (Sapir 1944, 94). Likewise, Kennedy and McNally (2005) emphasize that gradability applies not only to adjectives but also to other grammatical categories. Paradis (2008), similarly, argues that degree is a pervasive phenomenon in language, possible to apply to most meanings in a motivated context.

Paradis (2008) divides degree modifiers into two types: totality modifiers (either maximizers, e.g., completely, or approximators, e.g., almost), associated with a definite boundary, versus scalar modifiers (boosters, e.g., very, moderators, e.g., quite, or diminishers, e.g., a little), which are unbounded and situated on a scale (2008, 321). Still, since degree modifiers determine the interpretation of the expressions they modify, they all permit a scalar reading of an otherwise nongradable meaning. A developmental path for degree meanings would thus proceed “from a bounded to an unbounded configuration” (Paradis 2008, 339).

Rainer (2015, 1340) holds intensification, as ‘expression of degree’, to be a linguistic universal, tending to be expressed analytically more often than via word-formation patterns in the languages of the world. As a linguistic term, then, intensification comprises “any kind of degree” (Rainer 2015, 1349) and “shades off imperceptibly into quantification, augmentation and diminution, qualification, and emphasis” (Rainer 2015, 1340). In addition, the pragmatic function of intensifiers makes them even harder to categorize (Rainer 2015, 1346–1347). Note however that Masini and Micheli (2020), on the contrary, highlight the importance of fleshing out intensification further by suggesting approximation as one functional domain, encompassing such notions as attenuation, vagueness or deintensification.

Notwithstanding, in line with Paradis’ (2008) classification of degree modifiers, and with intensification as expressing degree (Rainer 2015), the four native adjectives included in this study would be of two types: hel- ‘whole’ and halv- ‘half’ being totality modifiers, associated with a bounded scale, with hel- acting as a maximizer and halv- as an approximator; stor- ‘big’ and små- ‘small’ being scalar modifiers, associated with an unbounded scale, and with stor- acting as a booster and små- as a diminisher.

2.2 Basic Facts About the Swedish Word-formation Patterns in Focus

The constructions under investigation are briefly touched upon in reference grammars or handbooks on Swedish word-formation. Små-, stor-, and halv- are classified as adjectival compound constituents, combined with a verbal base in some of the listed examples provided by Söderbergh (1968, 12), Thorell (1984, 40–41), and Josefsson (2009, 101). Söderbergh (1968, 12, f.n. 3) claims, however, that their function is that of an adverb, and she lists hel- as an adverb constituent in adjectival compounds (1968, 11). Thorell (1984, 40–41), instead, classifies hel- as an augmentative prefix with the meaning mycket ‘much’, perhaps leaning on Teleman (1970, 44) who remarks the tendency of adjectives, that often occur as compound constituents, to receive simplified or altered meanings, thus making them similar to affixes, such as augmentative hel-. In Lundbladh (2002), hel-, mega-, and stor- are classified as productive evaluative or intensifying prefix-like constituents used in combination with adjectives or nouns.

Neoclassical prefixes are mentioned to varying degrees and without much specification. Söderbergh (1968, 47, 59) mentions kvasi- ‘quasi-’ and pseudo- as pejorative. Thorell (1984, 56–57) lists the most common neoclassical prefixes along with a meaning indication, such as makro- and mega- ‘big’ and semi- ‘half’, among which mini- ‘little, smallest’, mikro- ‘little, small’, and max(i)- ‘big, biggest’ are said to be productive, whereas Teleman (1970, 53) instead mentions kvasi-, pseudo-, and mini- as productive, and Kotcheva (2016, 2575), the intensifying hyper- and super- as productive.

As for present participles, their nature is a topic of contentious debate in Swedish grammar. Present participles are classified either as adjectives derived from verbs (Hultman 2003, 40, 79; Kotcheva 2016, 2573), as verb forms (Dahl 2003, 83; Thurén 2008) or, alternatively, as a word class of their own (Teleman et al. 1999, 582–583).

Morphologically, present participles are built by the verb stem of a simple, derived or compounded verb, plus the suffix -ande, or less frequently -ende, if the stem ends on an accentuated vowel (Teleman et al. 1999, 613–614; Hultman 2003, 155), such as om[fatt-ande] ‘encompassing’ or fram[kall-ande] ‘provoking’.^[2] Present participles, in their turn, can be prefixed or compounded (as the constructions in this study), with the first constituent expressing different semantic roles, such as theme, result, location or grade/manner. Some of the compounded present participles are synthetic, that is, no corresponding, compounded verb exists, for example hel[omfattande] ‘whole-encompassing’ (*helomfatta) or ångest[framkallande] ‘anxiety-provoking’ (*ångestframkalla) (Teleman et al. 1999, 616–617). In contrast to most adjectives in Swedish, which are marked for agreement, present participles are invariable.

Syntactically, present participles behave as adjectives. They can thus occur as bound or free complements (subject or object predicatives), attributes, and adverbials (Teleman et al. 1999, 617).

Semantically, present participles tend to carry an active meaning, similar to the verbs on which they are based (e.g., “ett gråtande barn = ett barn som gråter”, Hultman 2003, 156, i.e., ‘a crying child = a child who cries’), but sometimes their meaning has drifted, so to have a lexicalized adjectival doublet (e.g., spännande ‘exciting’, originally from spänna ’tighten, span’). Teleman et al. (1999, 583; see also Lundquist 2008, 57, 202) make a distinction between verbal and adjectival participles but remark that the distinction is gradual, and that it can be more or less apparent to different speakers.^[3] Overall, present participle tend to be formed by verbs denoting unbounded (atelic) processes and have a generic meaning, similar to many adjectives (Teleman et al. 1999, 610–611). According to Thurén (2008, 172–173), who adopts a distributed morphology framework, participles in -ande are verbs showing both imperfective/progressive aspect (via the aspectual suffix) and argument structure (e.g., Jag hörde Stina, sjungande nationalsången ‘I heard Stina, singing the national anthem’, Thurén 2008, 175).

In this study, we opt for an analysis of present participles as infinite verb forms, which syntactically behave as adjectives do. Our main motivation for this view is that they are fully productive. Yet, as Haspelmath (2024) argues, from a constructional perspective of morphology, the distinction between inflection (present participles as verb forms) and derivation (present participles as verb-derived adjectives) might be a mere terminological invention. Haspelmath (2024) admits, however, that inflection and derivation are useful comparative concepts for describing linguistic information and thus need to be defined in a clear-cut way, instead of being said to lie on a continuum – in his view, participles would on semantic grounds belong to derivation. Nevertheless, and most importantly for our purpose, irrespective of whether present participles are inflected or derived, they do contain a base in the form of a verb, which permits evaluation.

2.3 Theoretical Framework

Our study situates within a lexeme-based approach to word-formation, compatible with construction morphology (Booij 2010). Hence, words and word-formation processes can be seen as constructions, associating form and meaning and accounted for by more or less abstract schemas, interconnected via inheritance. In this framework (cf. Battefeld et al. 2018; Masini and Audring 2019, 370), affixes are not independent entities but part of the words they belong to, as in an evaluative, prefixoid construction of for example halvsjunga ‘half-sing’ in (1a) (also matching with Grandi’s 2009 two criteria for defining an evaluative construction). A compound analysis involving similar items is given in (1b).

Seeing that the four Swedish adjectives in our study have been classified differently (as also illustrated in the example above), we find it motivated to penetrate a bit deeper into the classification of the first element, that is, as a prefixoid (Sigurd 1983; Lundbladh 2002) or as a compound constituent (Söderbergh 1968; Thorell 1984, Josefsson 2009). In this respect, Van Goethem (2010) advances five parameters of the grammaticalization process of compound constituents gradually shifting into affixoids: the phonological parameter (change of stress pattern); the morphological parameter (loss of grammatical inflection); the semantic parameter (semantic specialization into a new or more restricted meaning); and two distributional parameters, productivity (increase) and paradigmatic alternation (complementary distribution or competition between affixoids and existing affixes). Of these five parameters, the phonological and the morphological ones are of no use for the present study: whether the adjectives appear as compound constituents or prefixoids would not result in a stress pattern change, and loss of inflection would apply in both cases (similar to Dutch adjectives, cf. Van Goethem 2010). Hence, only the three latter parameters will be taken into consideration later on, when we return to the nature of the first constituent and the type of construction, as being either a derivative (1a) or a compound (1b).

4.1 The Verbal Base

We start with some words about the verbs taking part in the different patterns.^[6] Intransitive verbs with animate/human referents predominate in the data. They tend to refer to activities expressing human movements or motions (4a-b), behavior or manner (4c-d), and sound-making (4e-f).

Inanimate referents often refer to natural forces, such as water (5a), fire (5b), or weather phenomena (5c). Verbs that are specified for internal arguments, such as cover, win, or sell (something) often leave the internal arguments unexpressed in the construction types (6) (cf. Teleman et al. 1999, 610–613).

Regarding the morphological complexity of the verbal base, there seem to be differences between the text genres. In the fiction data, the verbal base is exclusively simple and of Germanic origin. In the dictionary data, there are a few complex verb bases (7a-b), whereas the news and social media data contain more complex bases, such as compounded verb bases (7c), several bases of Romance origin (7d-e), and some borrowings from English (7f) (this spreading to include verbal bases of different kinds and origins may thus be taken as evidence of the patterns being entrenched and synchronically productive).

Among the neoclassical prefixes, mega- is the most frequently attested one as for type, although only occurring in the news and social media data. The types with mega- include several compounded bases and sometimes contain hyphens (8a-b).

4.2 Evaluative Meanings of the Different Patterns

We now turn to our attempt to analyze the data as for whether the different constructions express evaluative meanings. According to our analysis, shown in Table 2, they do not behave the same on this point.

As regards the four native adjectives, hel- has most often a non-evaluative meaning, which is not the case with halv-, små-, and stor-. Two examples of non-evaluative meanings for hel- are given in (9a-b), where the one in (9a) is, as previously mentioned, the most frequently occurring type as per tokens in our data.

There are also a few non-evaluative meanings for halv-, små-, and stor-, where it is not the verbal process per se that is being evaluated, as in (10a-c).

Turning instead to the four adjectives with evaluative meanings, we find many different examples involving halv-, små-, and stor-, and also a few with hel- (11a-d).

In addition, some of these items have acquired specialized, lexicalized evaluative meanings (12a-b), which sometimes also cooccur along with the “literal” evaluative meaning (12c-d).

If we look at the neoclassical prefixes, which are, as said, more scarcely represented in the data, they seem to behave a bit differently when it comes to evaluative meanings. Mega- and mini- are represented by most attested types in the data, all of them being evaluative (13a-b). Mikro- and makro- appear in fewer and non-evaluative types (13c-d). Semi- occurs in few types, only one is evaluative (13e) and, in fact, competing with halv- in this example. Max- is not attested in an evaluative construction.

In sum, among the neoclassical prefixes, only mini- and mega- show some productivity in evaluative constructions with verbal bases.

4.3 Competing Constructions

We now turn to the possible competition in our data, such as among the native adjectives as well as between the native adjectives and the neoclassical prefixes, which is summed up in an interim conclusion, before turning to competition between morphological constructions and syntactic phrases.

We begin by looking at competition involving the constructions with the four adjectives. According to Dressler and Merlini Barbaresi (1994), the pragmatic function of evaluative morphology is evidenced by the fact that complex words with a pragmatic function do not block one another. In this respect, we see that our dictionary data from SAOB contain many examples, where synonyms involving other evaluative constituents are given in the meaning explanations. For example, the meaning of stor- is often explained by the use of other intensifying modifiers (14a-b). We also find competition between halv- and små-, with små- given as a synonym of halv- (14c-d). In addition, we find meaning explanations where the verb base instead has been replaced by a synonym verb (14e-f).

Moreover, in the meaning explanations of constructions with halv-, we find paraphrases with nästan ‘almost’ and till hälften ‘to the half’ or mitt emellan ‘in between’, thus indicating an approximative value of halv- (15a-d) (cf. Masini and Micheli 2020). Besides, note also the indication of a progressive meaning of the complex verb in (15d), given the construction with a posture verb plus an infinitive (sitta och V ‘sit and V’).

Moving on, we cannot claim to have found much competition between the native adjectives, on the one hand, and the neoclassical prefixes, on the other. In our opinion, constructions with native adjectives constitute the entrenched patterns for expressing evaluative meanings of verbs, preferably including verbs of Germanic origin but also spreading to involve those of Romance origin (cf. Söderbergh 1968, 41, f.n. 2 who remarks that neoclassical affixes rarely penetrate the Germanic derivational system in Swedish, but if so, prefixes more often do than suffixes). For instance, stor- and små- versus mega- and mini- could rather be said to be in complementary distribution, seeing that mega- is more intensifying/boosting than stor-, and mini- is more diminishing than små-. With increase of use, semi- could, however, become a competitor of half-, considering that they express similar meanings. Independent support for these assumptions is given by some attestations in the news and social media data, where native and neoclassical prefixes are combined (16a-c).^[8]

To arrive at an interim conclusion, let us first give a short account of Masini and Micheli (2020) who identify five closely related values conveyed by Italian simil-expressions, namely fakeness (e.g., similpelle ‘fake leather’), imitation (e.g., similvita ‘virtual life’), resemblance (e.g., simil-Amazon ‘Amazon-style’), vagueness (e.g., simil marrone ‘sort of brown’), and kin-categorization (e.g., simil-influenzale ‘parainfluenzal’). They (2020) conclude that the polysemy of simil- gives rise to a continuum with some of the values being more on the categorizing lexeme creation side, such as kin-categorization, whereas others belong to the approximating evaluation side, such as vagueness. While the constructions in our data lend themselves to be analyzed according to a similar kind of model, the present study suggests that the border between constructions with halv- as an approximator and those with små- as a diminisher is fuzzy. Therefore, we rather tend to agree with Rainer (2015, 1341) who notes that with adjectival bases – or present participles with their adjectival function – the border between diminutive versus approximative/attenuative patterns as well as between augmentative and intensive patterns often becomes blurred. Furthermore, although hel- and halv- are totality modifiers, positioned on a bounded scale, and stor- and små- are scalar and open-ended, our data indicate that all four modifiers associate with an unbounded scale when occurring in evaluative constructions, in line with Paradis’ (2008) view.

As for the competition between the morphological constructions and syntactic phrases, we find some competition in the dictionary data from SAOB, considering that the meaning explanations sometimes include syntactic paraphrases of the morphological constructions (17a-b) (cf. also 15a-d).

If we now turn to the corpus data, shown in Table 3, they suggest that there are differences among the morphological patterns in relation to their correspondences to syntactic phrases. For små-, there are actually fewer attestations of syntactic phrases with smått, both per type and token. In contrast, for halv- and stor-, there are more syntactic than morphological constructions as per types but less syntactic phrases as per tokens (except for stort- in the fiction data). For hel-, however, there is an advantage for syntactic phrases over word-formation both for types and tokens (despite the fact that, heltäckande, the type with most tokens in our corpus-data, is a morphological construction).

Although our corpus data are not flawless (cf. the method and data section), we can still conclude that also the syntactic phrases are apt at expressing evaluative meaning and thus constitute competing options, especially for hel-, much more often occurring in syntactic phrases than in morphological constructions (18a-c).

To sum up, except for the adjective hel-, which is much more prone to be used as a free adverbial to intensify verbs, the data do not provide evidence for any blocking effects between morphological and syntactic constructions. Blocking effects would, according to Aronoff (e.g., 2018), be a consequence of competition, which would, in the course of time, give rise to different distributional patterns. Rather, our data align more with Lieber’s (2012, 2115–2116) suggestion that affixes that cover similar semantics (such as evaluative constructions can do) can result in mismatches between form and meaning, so that one affix pairs with several meanings or several affixes occupy the same semantic space.

4.4 Compound Constituents or Affixoids?

Before concluding this study, we come back to Van Goethem’s (2010) parameters of grammaticalization regarding the gradual transition of compound constituents into affixoids, namely semantic specialization, productivity, and paradigmatic alternation, which have relevance for the constructions under study.

Concerning the semantic specialization parameter, being a strong motivation for grammaticalization, the adjectives in their bound or free uses and combined with verbs predominantly express evaluative meanings (albeit to a lesser degree for hel-). In this sense, they could be argued to have a function similar to that of diminutive and augmentative constructions in other languages. For the adjectives små- and stor- with their scalar and open-ended meaning (cf. Paradis 2008), the semantic specialization may be said to turn from size to intensification – considering also that verb events are not easily quantified for size. For halv- and hel-, as quantifying concepts and expressions of degree on a closed scale, it is even easier to argue for a shift in meaning from a literal quantifying meaning to a more approximative or intensifying one when combining with verbs (cf. Rainer 2015; Masini and Micheli 2020).

The productivity parameter gives the strongest support for claiming that the adjectives are transitioning into prefixoids in their bound uses, given that they, except perhaps for hel-, occur in a non-neglectable number of both established and novel forms in our data.

If we turn to the paradigmatic alternation parameter, there is arguably some competition between små- and halv- in their evaluative meanings, as well as between stor- and other native intensifying modifiers (cf. 14a-b). In addition, hel- and halv- both belong to an additional second paradigm, next to the evaluative one, in which the adjectives keep their literal, quantifying meaning (e.g., half-paying). For hel-, this second paradigm seems to be more productive. However, we found little competition between the adjectives and the neoclassical prefixes. Compared to halv-, semi- is much less frequently attested in the data. Likewise, evaluative constructions with små- and stor- do not seem to compete with those with mini- and mega-, seeing that they express different degrees of intensification. Paradigmatic alternation between the adjectives and neoclassical prefixes is thus not entirely ruled out (and could eventually become more productive). Still, the native options could be argued to be much stronger entrenched, to the detriment of neoclassical prefixes, at least when it comes to being combined with verbs.

To sum up, all three parameters for the grammaticalization cline from compound constituents onto true affixes are fulfilled to some or a high extent, so to speak in favor of the adjectives in their bound uses, at least små-, stor-, and halv-, to be classified as prefixoids.