The fast-over-slow asymmetry: speed adverbs in Estonian

2.1 The notion of speed

speed can be defined as a “rate of movement” understood as “how fast something moves” and/or the “rate at which something moves or happens”.^[1] In linguistic studies, speed is mostly understood as a rate (e.g. Slobin et al. 2014) but also as a temporal concept (Plungian and Rakhilina 2013) or as a concept of change (Koev 2017). If seen as a rate, speed falls under the category of manner (Cardini 2008); and if seen as a temporal concept, speed can be seen as duration (Plungian and Rakhilina 2013). In most approaches, however, speed is differentiated from duration and time (Givón 1970; Hallonsten Halling 2018), and this approach is also followed in the current study. Thus, the study reports on Estonian speed adverbs expressing rate (even if the adverbs may also have a duration reading in some contexts).

The differentiation between rate and duration in the conceptualisation of speed highlights another important facet of it – namely, it seems impossible to conceptualise speed without motion. Even though all kinds of processes can be characterised by speed, not only physical motion, we can argue that motion speed is the source domain for expressing diverse target domains, similar to the time is space metaphor. In other words, the conceptualisation of any dynamic event is grounded in the conceptualisation of motion event. As a result, whenever we talk about the expression of speed in language, the prototype is the expression of motion speed, with the extensions to other domains, such as events unfolding in time or actions performed over time. Without the motion component (i.e. spatial component), only duration remains, which is a purely temporal notion.

Similar arguments have also been put forward in other studies. For example, Lahlou (2023) shows that in Arabic, the term ‘speed’ is conceptualised through two related conceptual metaphors, namely speed of action is speed of motion and speed of progress is speed of motion to a destination, both grounded in the more general metaphor change is motion (see also Lakoff et al. 1991). The metaphor speed of action is speed of motion has also been noted in the study of music language by Wu and Liu (2024) and in the analysis of Cantonese speed-expressing slang by Wong (2020).

Finally, all speed adverbs in our data can, in principle, be used to describe physical motion, providing further rationale for not distinguishing motion speed from the speed of other processes unfolding over time. Therefore, when discussing speed in language, we understand speed as motion speed (i.e. rate), while acknowledging its extensions to other, non-physical and non-motion domains. For the same reason, we also provide an overview of speed in the context of motion event studies (see Section 2.3).

2.2 Polar antonyms

Polar antonyms are gradable antonyms that “denote relative values along a single dimension, like length or weight, prototypically measured in conventional units” (Cruse 2006: 129–130). The English adjectives fast and slow are examples of typical polar antonyms. As polar antonyms, speed adjectives (and adverbs) are subject to certain characteristics. First, they “are typically evaluatively neutral, and objectively descriptive” (Cruse 1986: 208). This means that events described by such words can be measured (e.g. as kilometres per hour). Second, one of the two antonyms is more basic than the other. That is, one end of the scale of polar antonyms is more dominant than the other end. The dominant end can also be taken as the unmarked category and the other end as the marked one (Al-Kajela 2018; Ingram et al. 2016). Cruse (1986: 208–209) describes this characteristic of polar antonyms in terms of impartialness: “Only one member of a pair yields a normal how-question /…/, and this question is impartial”. For instance, it would be more neutral to ask How fast is it? than How slow is it? In addition, “a single scale underlies a pair of polar antonyms” (Cruse 1986: 210), and this scale is for the dominant member of the word pair (e.g. fast). Thus, as Cruse (1986: 210) notes, it is more natural to say fastness than slowness as the underlying scale is driven by the concept of fast, not by slow. Al-Kajela (2018) further claims that in terms of the unmarked (i.e. dominant) and marked (i.e. non-dominant) category member of the antonym pair, “the former is more frequent and neutral than the latter” (Al-Kajela 2018: 115). Cross-linguistically, however, gradable antonyms vary in that even though some cross-linguistic regularities regarding impartialness can be detected, the committed behaviour of the lexical items is also very much context-dependent (Cruse 1992).

A more general inherent feature of (polar) antonyms is that they can simultaneously be presented on two scales: on “an absolute scale, which covers all possible values of the scaled property from zero to infinity, and relative scale, which is movable relative to the absolute scale” (Cruse 1986: 205). That is, the speed of an object can vary from 0 m/s (= standing) to infinity (faster than light). Even though speed can be measured, in language, and particularly so at the word level, speed can be best treated on a relative scale: fast refers to faster motion than slow. Cruse (1986: 205) illustrates the two scales with the following figure in which the pivotal region refers to the neutral centre of the scale (see Figure 1).

Figure 1:

Speed as an illustration of polar antonyms. The figure is taken from Cruse (1986: 205).

This neutral centre, or pivotal region, can be captured as “some kind of implicit or explicit standard or norm /…/ against which judgements are made and in the light of which qualities are attributed” (Singleton 2016: 73). Often, however, there seem to be no lexical items that could be taken to convey the norm itself (Pajusalu 2009: 143). Even though there are very general items that could represent the pivotal region (e.g. moderately, normally), their semantics does not include the features of this specific domain. This applies also to speed in that there seem to be no lexical items conveying medium speed (apart from the general terms that do not incorporate speed features). Thus, it is an open question whether the pivotal region of the scale is realised in a specialised lexical inventory, or whether there is a gap in the scale between the items (e.g. between the items of fastness and slowness).

Polar antonyms are typically not discussed in the context of their whole semantic domain because other linguistic realisations of the same domain, and their relation to the clearest instance of an antonym pair, tend to lie behind the main scopes of the studies. Because antonyms can reflect the more general asymmetries in language and in cognition, we now turn to the studies that have addressed lexical means for speed – the focus of the current study – from a broader angle.

2.3 Asymmetries in the expression of speed

There are several ways in language to express speed. For instance, depending on the language, speed can be expressed by verbs (e.g. hurry), nouns (e.g. rush), adverbs (e.g. quickly) and adjectives (e.g. quick) (see also papers in Dixon and Aikhenvald 2004). In addition, speed can be conveyed by suprasegmental categories, such as speech rate, and even by phonemic iconicity (Zhao and Wu 2025). We can call lexemes that incorporate speed in their semantics as instances of speed lexicon. Speed adverbs – a specific class of speed lexicon and, simultaneously, a specific class of manner adverbs – are the focus of the current paper. However, speed adverbs are considerably less often addressed in the literature than speed adjectives, and far less frequently than (manner of) motion verbs that also lexicalise speed as a principal dimension. Thus, addressing speed, we start with motion verbs and then move to adverbs and adjectives as word classes that are closely linked in language as adverbs are often derived from adjectives (e.g. slow > slowly in English) or even share the same lexical form (e.g. fast as both an adverb and adjective in English).

As evidenced in studies of motion verbs, speed is highly pertinent to motion (Ikegami 1969; Malt et al. 2008; Slobin et al. 2014; Vulchanova and Martinez 2013). It is one of the core dimensions of manner (e.g. Cardini 2008; Kopecka 2010; Narasimhan 2003; Snell-Hornby 1983; Stosic 2019), introducing asymmetries in linguistic encoding (Ikegami 1969; Snell-Hornby 1983). For example, Ikegami (1969) suggests that there are more fast motion verbs than slow motion verbs in English, whereas slow motion verbs are often semantically more complex in that the slow motion is a consequence of laborious motion, as in the verb crawl. Similarly, Aurnague (2011, 2024), in his classification of semantic dimensions that license French verbs’ telicity reading in context, discusses speed in relation to fast verbs (e.g. ‘run’), and force in relation to slow verbs (e.g. ‘crawl’). Snell-Hornby (1983), in analysing German and English verbs, also treats fastness as a primary and slowness as a secondary dimension of motion verbs. At the same time, motion verbs, when rated for how fast or slow the motion they encode is form a continuum from the slowest to fastest verbs, as evidenced by studies on English (Zhao and Wu 2025) and Estonian (Taremaa 2017). Such rating studies do not seem to confirm the early observations of imbalanced sets of fast versus slow verbs, but both studies indicate that fast verbs receive more extreme ratings than slow verbs.

As for adjectives and adverbs, the seminal study of Dixon (1982; see also Dixon 2004a) links word classes with their characteristic semantic types and lists speed as one of the peripheral semantic types of adjectives. Peripheral means that adjectives of speed are not found in the lexicon of all languages, but rather only in those with medium- or large-sized adjective classes. Studies of diverse languages (see papers in Dixon and Aikhenvald 2004) confirm this suggestion by showing that even if speed is expressed by adjectives in some languages, there are languages in which instead of adjectives, speed is realised in verbs (Dixon 2004b) or adverbs (England 2004; Levy 2004). A recent study (Hallonsten Halling 2018) focusing on adverbs instead of adjectives further shows that for the word class of adverbs, speed can be taken as the core, not peripheral semantic type. It examines 60 typologically diverse languages, focusing on simple adverbs (in Hallonsten Halling’s approach, simple adverbs are mono-morphemic words). Of the languages that have simple adverbs (41 out of 60), all express speed with adverbs.

If speed adverbs or adjectives occur in a language, their lexicon can vary greatly, with some languages having only one item of speed per word class, whereas others have a numerous and diverse set of speed words. For instance, Hallonsten Halling (2018) shows that, based on her typological study of 60 languages, even though the majority (i.e. 41) of the study’s languages have simple speed adverbs, several of them have only one or two speed adverbs (it should be noted, however, that she only examined mono-morphemic speed adverbs). As another example, Nkami (spoken in Ghana) has only two speed adjectives, one for fast and one for slow speed (Asante 2021).

If a language has several items of speed adjectives/adverbs (which is typical, for instance, in European languages), a specific asymmetry may occur, as evidenced by the studies conducted in the domain of motion descriptions. That is, there is converging evidence for the asymmetry between fast and slow expressions in language, also labelled as the fast-over-slow asymmetry (Taremaa and Kopecka 2023a). As one realisation of this asymmetry, fast motion tends to be expressed more frequently and diversely than slow motion (Hallonsten Halling 2018; Plungian and Rakhilina 2013; Taremaa and Kopecka 2023a, 2023b). For instance, in Russian, there are over fifteen adjectives for fastness and only one main adjective for slowness (Plungian and Rakhilina 2013). Plungian and Rakhilina (2013) suggest the same holds true for other Slavic languages as well. In Estonian motion descriptions, fast motion is explicitly expressed by manner modifiers (realising in various morphosyntactic patterns) almost five times more frequently than slow motion (Taremaa and Kopecka 2023a). Also, it does not seem to be accidental that Schäfer (2023), in his study on ly-derivational adverbs of English, includes in his analysis 10 adjectives of fast motion (e.g. brisk, hasty) and only one adjective of slow motion (slow). Furthermore, as for processing in the sentential context, fast adverbs are read more quickly than slow adverbs in English (Stites et al. 2013), which is in accord with the finding that objects entailing faster motion tend to be named faster than those that imply slower motion (Ben-Haim et al. 2015). This suggests that fastness in general receives predominant processing over slowness, resulting in the asymmetries of adverbs (also in terms of antonyms).

2.4 Adverbs in Estonian

Estonian is a morphology-rich Finno-Ugric language with high manner saliency (Taremaa et al. 2024). One way of expressing manner (including speed as one dimension of it) is through adverbs. However, it should be noted that manner can be expressed not only by adverbs, and conversely, not all adverbs express manner. Adverbs in Estonian are defined as invariable words that, in the sentential context, function as adverbials, and, typically, express space, time, manner, state or quantity (Erelt et al. 1995: 23–26; Veismann and Erelt 2017: 417–421). Speed adverbs are subsumed under manner adverbs with no formal features to distinguish them from the rest of the (manner) adverbs. Instead, they share the same morphological characteristics with other adverbs.

Whereas a small number of Estonian adverbs have a simple morphological structure (e.g. tasa ‘quietly, slowly’) with no overt morphological marking, most Estonian adverbs are derived from adjectives or nouns by different derivational suffixes (e.g. aeglase+lt ‘slowly’ < aeglane ‘slow’ + -lt). The most common and most productive adverbial suffix is -lt; other suffixes with limited productivity are -sti, -misi, -ti/-di, and -mini (Kasik 2015: 383, 393). In addition, there are adverbs that are lexicalised case-inflected nominal words (e.g. hoo+ga ‘with a momentum’ as hoog ‘momentum’ + -ga as the comitative marker). As for complex, multi-stem adverbs, there are compounds consisting of fixed compounds, such as tasa+pisi [quietly+tiny.nom] ‘gradually, slowly’, and similes, such as välk+kiiresti [lightning.nom+fast] ‘lightning fast’. The other major class is reduplications (e.g. ruttu-ruttu ‘quickly-quickly’). All adverbs, regardless of their structure and semantics, can in principle be reduplicated in Estonian, and it is therefore a clear mechanism for meaning intensification (Erelt 2008; see also Dressler and Barbaresi 1994: 510–524). Furthermore, Erelt (2008: 271) suggests that “While väga ‘very’ and other augmentatives express a high degree of intensity, reduplication expresses ultimate intensity.”

Adverbs falling in the borderline of compounding and derivation are those composed of an intensifier as the first component and a speed adverb as the second component (e.g. hüper+kiiresti ‘hyper fast’) (see also Kiik 2021). Such intensifiers have sometimes been called prefixes (Hoeksema 2012), but more frequently prefix-like elements, semi-prefixes or prefixoids (Norde and Van Goethem 2018) or initial combining forms (ICFs; Prćić 2005). Depending on the take on the initial forms, the resulting complex words can be understood as compounds, prefixed words or something in between (see also Booij 2005).

For the purposes of the current study, the exact classification of such complex words is irrelevant as the main focus is on whether or not the meaning of a speed adverb is intensified by the initial component which, in essence, can attach to both slow and fast adverbs (cf., mega+kiiresti ‘mega fast’ and mega+aeglaselt ‘mega slowly’, üli+kiiresti ‘extra fast’ and üli+aeglaselt ‘extra slowly’, elu+kiiresti [life.nom+fast] ‘at full speed’ and elu+aeglaselt [life.nom+slowly] ‘very slowly’). Assessing intensification, we need to differentiate such word structures from fixed compounds and similes, as in the latter case the combinability of the first and second compound is much more restricted.

Simile compounds (e.g. välk+kiirelt [lightning.nom+fast] ‘lightning fast; very fast’, tigu+aeglaselt [snail.nom+slowly] ‘very slowly’) can also be seen as being based on the intensification mechanism but are not analysed here as such. This is because the combinability of the first component (e.g. välk ‘lightning’, tigu ‘snail’) with adverbs is much more restricted than for prefixoids. Not only can they not be added to the adverb expressing the opposite speed (e.g. *välk+aeglaselt ‘lightning slowly’, *tigu+kiirelt ‘snail fast’), but they cannot be easily combined with all adverbs of the same speed category either (cf. välk+kiirelt ‘lightning fast’ and ?välk+kähku ‘lightning’ + ‘quickly’). Prefixoids, on the other hand, can be attached to all kinds of speed adverbs (e.g. mega+kiirelt ‘mega quickly’, mega+kähku ‘mega quickly’, mega+aeglaselt ‘mega slowly’). Thus, for reasons of clarity, and to differentiate principally different word formation mechanisms, complex words with an intensifier as a prefixoid (e.g. mega- ‘mega’, hüper- ‘hyper’) are not termed as compounds but as ‘adverbs with prefixoids’ (for the sake of convenience, and very loosely, we will also use the term ‘prefixation’ to denote the mechanism of attaching prefixoids to adverbs). The term ‘compounds’ is used for fixed compounds (e.g. aega+mööda ‘slowly, gradually’) and similes (e.g. välk+kiirelt ‘lightning fast’).

3.1 Data collection

In the data collection process, two main difficulties had to be overcome: semantic and formal. The semantic challenge captures difficulties with the adverbs that could be positioned on the periphery of speed adverbs (or perhaps even be considered non-speed adverbs). For example, tasakesi in Estonian means ‘quietly’, but it may also imply slow motion without expressing any sound at all.

The formal challenge relates to difficulties in establishing the adverbs as a word class, as the boundaries of the adverb class in Estonian are fuzzy (Paulsen 2018). While in most cases, adverbs can be easily detected based on derivational suffixes or unchangeable word stems, certain case-inflected words are in the process of lexicalisation from case-inflected nouns to adverbs in Estonian. In addition, with regard to complex adverbs consisting of more than one word stem, it may be difficult to decide upon the bounds of an adverb (simple adverb vs. adverb with an intensifier as a separate word) due to the variation in whether the concept is written as one word or not. For example, super+kiiresti ‘superfast’ is almost as equally frequently written as two words (super kiiresti) or as one word (superkiiresti).^[2]

To overcome the difficulties and to collect speed adverbs, we took a multimethodological approach and combined the dictionary lookup method, corpus analysis and a small-scaled rating task. The full procedure is described in Appendix A. Briefly summarised, we started with a small list of speed adverbs (N = 38) found in an earlier study by Taremaa and Kopecka (2023a) and then used these adverbs in dictionary and corpus searches to gather as many speed adverbs as possible (e.g. by establishing synonyms and compounds). The consulted dictionaries include the main monolingual dictionaries of Estonian, Dictionary of Synonyms (Õim 2007), The Dictionary of Estonian (Sõnaveeb 2019) and The Defining Dictionary of Estonian (EKSS; Langemets et al. 2009). The corpus that was consulted is the Estonian National Corpus 2021 (ENC 2021, size 2.4 billion words; Koppel and Kallas 2022). To reduce researcher bias, adverbs were also rated for their speed reading by four native speakers of Estonian, all trained in linguistics. When rating for speed, the task of the speakers was to evaluate whether an adverb encodes speed, with the possible responses ‘yes’, ‘no’, ‘maybe’ and ‘unknown word’. When at least one speaker marked the adverb as expressing speed, it was included in the preliminary list of speed adverbs.

This resulted in 419 speed adverbs. The next step was to limit the set of adverbs to those that incorporate speed as a salient feature in their semantics. For this, we consulted two main dictionaries of Estonian which provide definitions for words: The Dictionary of Estonian (Sõnaveeb 2019) and The Defining Dictionary of Estonian (EKSS; Langemets et al. 2009). If an adverb was defined in at least one of the dictionaries through speed-related notions, it was included in the study. If it was not defined through speed or it was absent in the dictionaries, it was excluded. Regarding some adverbialised comitative or adessive forms that did not occur as separate entries, the meaning of the corresponding verb or noun was considered in establishing the adverbs semantics. If a reduplication or an adverb with a prefixoid was absent in the dictionary, the meaning definition of the second component was taken into account. Applying these criteria, 369 adverbs remained,^[3] and this is the data the current study reports on.

As for the speed content of the adverbs, the resulting list contains both prototypical speed adverbs (e.g. kiiresti ‘fast, quickly’, aeglaselt ‘slowly’), which we can refer to as primary speed adverbs, and also less prototypical ones where speed is just one (and possibly backgrounded) semantic feature, which we can refer to as secondary speed adverbs (e.g. tormiliselt ‘stormily, fast’, tasa ‘quietly, slowly’). In essence, the set of speed adverbs we have established and their further possible division into typical and less typical speed adverbs nicely illustrates the cline of speed the adverbs lexicalise. This cline is conceptual, corresponding to what Slobin et al. (2014: 702) have pointed out as follows: “Rather than representing discrete, compositional, Aristotelian categories, language use reveals conceptual continua”. However, differentiating between the primary and secondary speed adverbs and establishing the cuts on the speed cline for adverbs warrants a different type of study. Therefore, we focus on the opposition between fast and slow adverbs without further dividing them into additional semantic classes.

3.2 Data annotation

The adverbs in our data are annotated for the following variables (see Table 1). The main variables are Adverb, AdverbSpeed, AdverbComplexity and ComplexityType. The variable Adverb represents the individual adverbs in our list of speed adverbs (e.g. kiiresti ‘fast, quickly’, üli+kiiresti ‘extra fast’). AdverbComplexity shows whether the adverb is a simple or complex one. A simple adverb is a one-stem adverb, i.e. an un-prefixed, non-compound and non-reduplicated adverb, where it is considered a simple adverb regardless of the presence of any possible derivational suffixes. For instance, non-derived adverbs such as ruttu ‘quickly’ and derived adverbs such as kiiresti ‘fast, quickly’ (< kiire + the derivational suffix -sti) are both analysed as simple adverbs in this study. A complex adverb, as captured by ComplexityType, is either a compound (e.g. välk+kiiresti ‘lightning fast’), an adverb with a prefixoid (e.g. üli+kiiresti ‘extra fast’) or a reduplicated word (e.g. ruttu-ruttu ‘quickly-quickly’). In other words, a complex adverb consists of multiple stems. For modelling purposes, additional variables PrefixedOrNot and ReduplicatedOrNot are included to assess complexity type in more detail.

To address intensification mechanisms, the additional variables FirstComp and LastComp are coded to address which lexical information is used to intensify the meaning of fast versus slow adverbs, and how. FirstComp represents lexical items that are used as the first component (e.g. üli- ‘ekstra-’) of complex adverbs (e.g. üli+kiiresti ‘ekstra fast’). LastComp is the item used as the last component (e.g. -kiiresti ‘-fast’) of a complex adverb. In addition, for background knowledge, DerivSuffix is coded to assess if an adverb is a derivative, and if it is, which derivational suffixes it has.

As for semantics, AdverbSpeed divides the adverbs into fast (e.g. kiiresti ‘fast, quickly’) and slow adverbs (e.g. aeglaselt ‘slowly’). To assess background information, Descriptiveness shows whether the adverb mimicks via sound symbolism auditory or visible aspects of a process (Kasik 2015: 77) – or not (e.g. vudinal ‘running (with small steps), pattering’, viuh ‘with a whizz, quickly’). This variable also includes onomatopoeic and momentaneous adverbs (e.g. viuh ‘with a whizz, quickly’, suts ‘with a pop, quickly’).

Frequency information of the adverbs is captured by estimated absolute and relative frequencies as EstimAbsFreq and EstimRelFreq (paired with log10-transformed frequencies, Log10freq), as attested in the ENC 2021 (accessed through SketchEngine; Kilgarriff et al. 2014). In calculating the frequencies, any possible orthographic deviations or misspellings of a word are not taken into account. For instance, rutturuttu and ruttu-ruttu ‘quickly-quickly’ are analysed as the same lemma, as are üli+kiiresti and yli+kiiresti ‘extra fast’. To address the potential homonymy of adverbs and adjectives when inflectional and derivational suffixes overlap (e.g. -lt in aeglase+lt, which can be an adverb meaning ‘slowly’, as in jooksi-n aeglase+lt [run-pst1sg slowly] ‘I ran slowly’, or an ablative-inflected adjective form, as in aeglase-lt auto-lt [aeglane-abl car-abl] ‘from a slow car’), the raw frequencies were adjusted based on an additional corpus analysis. With all such adverbs (N = 99), 200 sentences (if available) were randomly taken from the ENC 2021. Then, the sentences containing an adverb (and not an overlapping word form of some other word class) were counted and, based on this, an estimated frequency of each adverb was calculated. Thus, the frequencies of the adverbs in our data represent the frequencies of the adverb forms only.

Nevertheless, this frequency measure does not show how often an adverb is used to express speed-related information. This is a concern for polysemous words, which can be used in a broad range of semantic functions. For instance, tasa ‘quietly, slowly’ can mean slow motion, but it can also express the absence of noise or that something is done without anyone noticing it (EKSS; Langemets et al. 2009). Our frequency analysis fails to account for the uses of an adverb in which there are no associations with speed, and, therefore, the analysis can be taken merely as a rough estimation regarding the possible differences between the use of slow versus fast adverbs.

Data analysis and visualisation were performed in R (R Core Team 2023), using also packages sjPlot (Lüdecke 2021) and ggplot2 (Wickham et al. 2021). Data and code are available on the OSF (https://osf.io/dhpc7).

3.3 Results of the pre-study: data overview

This study is based on 369 adverbs conveying the semantic feature of speed. The majority of speed adverbs are semantically complex. For instance, the descriptive adverb vuhinal ‘swooshing, fast’ expresses high speed, sound and intensity; non-descriptive tulisi+jalu ‘hotfoot, fast’ expresses high speed and intensity, but in addition to that, it specifies body movements (‘on foot’; the second component of the word, -jalu, is a historical instructive case form of the word jalg ‘foot, leg’). This semantic complexity of speed adverbs is similar to manner complexity of motion verbs (for which speed is one of many dimensions motion verbs can express) (e.g. Cardini 2008), also noted in earlier studies of manner as a concept in general (e.g. Stosic 2020). In addition, a number of adverbs in our data are descriptive (N = 59), mostly onomatopoeic (e.g. robinal ‘rustling, fast’), and many of those are also momentaneous (N = 40; e.g. lipsti ‘with a plop, quickly’).

The morphological structure of speed adverbs in Estonian can be described from the perspective of (i) compounding, prefixation and reduplication (variable ComplexityType; i.e. adverbs analysed as complex adverbs in our study) and (ii) derivational suffixes (variable DerivSuffix). Recall that derivational adverbs can occur as either simple or complex adverbs in our analysis, meaning that the defining feature of complexity is the number of stems, not morphemes. Simple adverbs are either monomorphemic or consist of a stem and a derivational suffix, whereas complex adverbs have multiple stems (including prefixoids).

Of the 369 adverbs, 123 are simple adverbs (e.g. kiiresti ‘fast, quickly’, aeglaselt ‘slowly’) and 246 complex adverbs (see Table 2). Complex adverbs are either reduplications (N = 40; e.g. ruttu-ruttu ‘quickly-quickly’, aeglaselt-aeglaselt ‘slowly-slowly’), adverbs modified by prefixoids (N = 140; e.g. mega+kiiresti ‘mega fast’, üli+aeglaselt ‘extra slowly’), or compound words (N = 66; e.g. välk+kiirelt [lightning.nom+fast] ‘lightning fast’, aega+mööda [time.prt+along] ‘slowly, gradually’). Compound words consist of fixed compounds (N = 43; e.g. tulisi+jalu [hot.adv+foot.adv] ‘hotfoot, fast’, aega+mööda [time.prt+along] ‘slowly, gradually’) and words which are based on the mechanism of comparison, i.e. similes (N = 23). In these simile compounds, the second component is a speed adverb and the first component expresses an entity with a characteristic motion speed (e.g. välk+kiirelt [lightning.nom+fast] ‘lightning fast’, tigu+aeglaselt [snail.nom+slowly] lit. ‘snail slowly’) or environment which dictates the possible motion speed (e.g. muda+aeglaselt [mud.nom+slowly] lit. ‘mud slowly’).

The most typical and frequent speed adverbs (e.g. kiiresti ‘fast, quickly’, aeglaselt ‘slowly’) are also the ones that are often combined with a diverse set of prefixoids (e.g. üli+kiiresti ‘extra fast’, super+kiiresti ‘superfast’, elu+kiiresti lit. ‘life fast’) and are frequently reduplicated (e.g. kiiresti-kiiresti ‘quickly-quickly’, aeglaselt-aeglaselt ‘slowly-slowly’).

Of the total of 369 adverbs (both simple and complex), 108 are non-derived (including fully lexicalised partitive, inessive and instructive case forms) and 261 derivational adverbs. Of the 123 simple adverbs, 28 (22.8 %) are non-derived and 95 (77.2 %) are derived adverbs. Of the 246 complex adverbs, 80 (32.5 %) are non-derived and 166 (67.5 %) are derived adverbs. Seven unique derivational suffixes occur (exceptionally, we include the comitative case ending -ga here as comitative-inflected nouns are in the process of lexicalising into adverbs; e.g. hoo+ga ‘with a momentum’). The most frequent by far are lt-adverbs (N = 144; kiire+lt ‘fast, quickly’), followed by sti-adverbs (N = 58; e.g. kiire+sti ‘fast, quickly’).^[4]

4.1 Hypothesis 1: type and token frequencies of fast versus slow adverbs

The hypothesis regarding the number of unique speed adverbs (i.e. type frequencies) is clearly confirmed, as can be seen in Figure 2. The total list of 369 adverbs has 105 (85.4 %) fast versus 18 (14.6 %) slow simple adverbs, and 193 (78.5 %) fast versus 53 (21.5 %) slow complex adverbs. In total, there are 298 (81 %) fast versus 71 (19 %) slow adverbs.

Figure 2:

The distribution of fast and slow adverbs across complexity types. Simple adverbs consist of one stem (and a derivational suffix); complex adverbs consist of multiple stems (including prefixoids).

The ten most frequent speed adverbs are presented in Table 3. Two fast adverbs, kiiresti and kiirelt ‘fast, quickly’, are the most frequent. They are followed by the slow adverb vaikselt ‘quietly, slowly’, which can be considered a secondary speed adverb. This is because it can denote only quietness (without encoding speed) or describe slow motion (without encoding quietness). The neutral primary slow adverb, aeglaselt ‘slowly’, ranks fifth according to their frequencies.

Regarding token frequencies (as attested in the ENC 2021), the total sum of all adverbs’ frequencies across fast versus slow adverbs is also different, but here fast outperforms slow only across simple adverbs (1,015,137 fast vs. 301,527 slow adverbs), whereas the opposite occurs for complex adverbs (28,233 fast vs. 91,054 slow adverbs) (see Figure 3).

Figure 3:

Summed estimated absolute frequencies (token frequencies) of speed adverbs across complexity types. Simple adverbs consist of one stem (and a derivational suffix); complex adverbs consist of multiple stems (including prefixoids).

The summed frequencies depend on the number of lexemes available. Therefore, by having more fast adverbs in the language, it is not surprising that the summed frequencies of fast adverbs are considerably higher than that of slow adverbs. Figure 4 indicates that there are high-frequency adverbs both in slow and fast groups, but the numerosity differs (extensively more fast than slow adverbs). In addition, simple adverbs (in blue) seem to be more frequent than complex adverbs (in yellow). This is also reflected in the top 10 adverbs in terms of frequencies, as shown in Table 3.

Figure 4:

Frequency distribution of fast and slow adverbs as simple (in blue) or complex adverbs (in yellow) across complexity types. Simple adverbs consist of one stem (and a derivational suffix); complex adverbs consist of multiple stems (including prefixoids).

Because the data is highly unbalanced, with fast adverbs comprising the majority of the data (which is a significant result in itself), modelling the entire dataset would not be reliable. Therefore, and to capture the most important speed adverbs in the language, we fitted a regression model to a subset of data consisting of the 60 most frequent speed adverbs (30 fast and 30 slow; see Appendix B). Only simple adverbs (N = 42) and compounds (N = 18) were included to ensure data consistency. We used a binary logistic regression model to analyse the relationship between adverb semantic type (AdverbSpeed; fast vs. slow) and the adverb frequency (Log10freq; as attested in the ENC 2021), controlling for adverb complexity (AdverbComplexity). The model was statistically significant (χ ²(2) = 15.37, p < 0.001) and moderate (C = 0.75);^[5] see also Appendix C. Consistent with the hypothesis, there is a reliable effect of adverb frequency on adverb speed (β = 0.96, SE = 0.4, z = 2.7, p = 0.008), with fast adverbs having a higher likelihood of being of higher frequency than slow adverbs (see Figure 5). Adverb complexity was marginally significant (β = –1.08, SE = 0.7, z = –1.6, p = 0.1), indicating that complex adverbs are more likely to occur among the most frequent slow adverbs than among fast adverbs. This shows the limited number of simple slow adverbs that are available for the speakers as compared to fast adverbs.

Figure 5:

Log10-frequencies of slow and fast adverbs (left panel; horizontal lines indicate median, square symbols indicate mean values) and predicted probability of fast adverbs as a function of log10-frequencies (right panel). The data subset includes the 30 most frequent and typical fast and 30 slow adverbs.

These results suggest that even though there are some slow adverbs that are particularly frequent, similarly to the most frequent fast adverbs, fastness is predominant in Estonian and can be expressed by a variety of adverbs. Slowness is much less frequently expressed and by fewer lexemes. Thus, the fast-over-slow hypothesis regarding type and token frequencies is clearly confirmed.

4.2 Hypothesis 2: meaning intensification of fast versus slow adverbs

We hypothesised that because fastness is more intense than slowness, the meaning of fast adverbs is more likely to be intensified than that of slow adverbs (Hypothesis 2). As a proxy for measuring meaning intensification, we use intensifying prefixoids and reduplications and count the occurrences of an adverb being (i) preceded by a prefixoid or (ii) reduplicated.

The proportions are presented in Figure 6. Contrary to our expectations, there are no significant differences in proportions between fast and slow adverbs in terms of being preceded by a prefixoid. The number of fast adverbs with prefixoids is much larger than that of slow adverbs (115 vs. 25), but this is due to the large number of fast adverbs. The proportion of adverbs with prefixoids is close to 40 % for both fast and slow adverbs. As for reduplication, even smaller differences occur.

Figure 6:

The proportion of adverbs with prefixoids (left panel) and reduplicated adverbs (right panel) across fast and slow adverbs.

To confirm these observations statistically, we fitted a binary logistic regression with adverb speed (fast vs. slow) as the dependent variable and the presence of a prefixoid and reduplication as predictors, controlling for adverb frequency (log10 frequency). The model was not statistically significant (χ ²(3) = 1.86, p = 0.6); see also Appendix C. Thus, there are no significant effects of meaning intensification through adding prefixoids or through reduplication.

Finally, not all adverbs lend themselves equally often to prefixation. In our data, only roughly 18 % of fast and 19 % of slow simple/compound adverbs occur with prefixoids (see Table 4). The choice of prefixoids is more varied across slow adverbs than fast adverbs, at least considering the ratio between the number of unique adverbs and the number of unique prefixoids (1.9 for fast and 3.6 for slow adverbs), even though the total number of unique prefixoids is larger for fast adverbs compared to slow adverbs (40 vs. 18 unique prefixoids).

These results indicate that contrary to our expectation, speed adverbs are not sensitive to intensification in terms of prefixation and reduplication. In other words, Hypothesis 2 is not confirmed. The proportion of adverbs being used with prefixoids is similar across fast and slow adverbs, but the array of possible prefixoids is broader for fast than for slow adverbs because simple fast adverbs are more numerous than slow adverbs. Slow adverbs (if preceded by a prefixoid), in turn, are more likely to co-occur with different prefixoids because on average, each slow adverb occurs with 3.6 unique prefixoids, whereas every fast adverb occurs with 1.9 unique prefixoids. However, it should be noted that these results depend on, firstly, what is considered a speed adverb, and, secondly, how similes and adverbs with prefixoids are differentiated from one another. We shall discuss this more thoroughly in the next section.