How fast is fast and how slow is slow in mental simulation? Two rating studies on Estonian speed adverbs

1 Introduction

Research into processing the speed of visual motion suggests an inherent difference and asymmetry in the domain of speed. This, in turn, can be reflected in language. Studies in experimental psychology have shown that neural activations in response to fast vs. slow motion differ in the affected brain regions and pathways (Yan et al. 2023) and that activation strength is enhanced for fast over slow motion (Wang et al. 2003). In addition, reaction times to fast motion are shorter compared to slow motion (Hülsdünker et al. 2019; Stolz et al. 2023). In relation to language, recent psycholinguistic research has provided evidence that speed-induced language can influence behaviour and attention allocation. For instance, English adverbs of fast motion are read faster in sentential context than slow adverbs (Stites et al. 2013), naming vehicles of fast motion (e.g., aeroplane) and assigning speed ratings to them takes less time than naming or evaluating vehicles of slow motion as evidenced by a study in Hebrew (Ben-Haim et al. 2015; for similar results, see also Irie et al. 2021; Pan et al. 2024; Von Sobbe et al. 2021), and fast motion verbs can guide one’s attention faster or in larger looking times to the endpoint of a described motion (Lindsay et al. 2013). Furthermore, when describing slow motion, people tend to speak more slowly than when describing fast motion (Perlman et al. 2015). Linguistic studies have further shown that the expression of speed in language is asymmetric, in that fast motion can be expressed more often and by a larger inventory than slow motion (Plungian and Rakhilina 2013; Taremaa et al. Under Review).

These findings of psychology (Hülsdünker et al. 2019; Stolz et al. 2023; Wang et al. 2003; Yan et al. 2023) and linguistics (Ben-Haim et al. 2015; Irie et al. 2021; Lindsay et al. 2013; Pan et al. 2024; Perlman et al. 2015; see also Stites et al. 2013; van Dam et al. 2017; Von Sobbe et al. 2021) suggest that fast motion receives predominant processing over slow motion (i.e., fastness is associated with enhanced processing, as reflected in neural activation, processing speed, and attention allocation), which is also reflected in language use. In addition, speed-induced language seems to evoke mental simulation, with expressions of fastness processed differently from expressions of slowness, in response to actual physical experiences of fast vs. slow motion. This, in turn, has been interpreted as evidence for grounded cognition, whereby language evokes sensorimotor patterns comparable to the described action (e.g., Lindsay et al. 2013; Speed and Vigliocco 2014; see also Barsalou 2008; Gibbs 2006). However, it is an open question exactly how the predominance of fastness manifests itself at the language level. Consequently, the current study aims to elaborate whether word semantics reflects the predominance of fast motion over slow motion, using Estonian as a test case due to its high manner salience (Taremaa et al. 2024), manifested in the large inventory of speed lexicon. In particular, the study focuses on speed adverbs and aims to assess the semantics of speed adverbs and mental simulation from two angles:

1. How does speed information (fast vs. slow) modulate speakers’ understanding of the relative speed of speed adverbs?

2. To what extent do meaning intensification mechanisms reinforce the meaning of fast vs. slow adverbs?

The hypotheses are grounded in the idea that fastness is more intense than slowness, meaning that fastness is more foregrounded in perception and cognition, receives predominant processing, and, in essence, this can be reflected in subjective intensity measures (see also Zhou et al. 2024). Consequently, fast adverbs should embed more intensity than slow adverbs, which should be reflected in the attributed meaning and speaker behaviour.

To address the research questions, two rating experiments were conducted. Experiment 1 collected Estonian native speakers’ ratings for simple speed adverbs (i.e., adverbs having one stem and an optional derivation suffix; e.g., kiiresti ‘fast’, aeglaselt ‘slowly’). Experiment 2 collected ratings for complex speed adverbs where the meaning of the main adverb was reinforced by word-formation mechanisms such as adding intensifying prefixoids (e.g., üli+kiiresti ‘extra fast’), reduplicating the adverb (e.g., ruttu-ruttu ‘quickly-quickly’), or forming simile compounds (e.g., välk+kiirelt ‘lightning fast’).

2 Experiment 1: simple adverbs

The aim of Experiment 1 was to assess whether mental simulation occurs in response to simple speed adverbs. To that end, ratings for simple speed adverbs for their implied speed were collected. Simple adverbs are defined as monomorphemic (e.g., ruttu ‘quickly’) or consisting of one stem and a derivational suffix (e.g., kiire+sti ‘fast’).

2.2 Results

The standardized ratings of the adverbs can be found in Figure 1. Adverbs pre-classified as slow have all received low ratings that are distinct from those of adverbs pre-classified as fast. Although distinct from each other, the two groups are relatively homogeneous. For instance, the slowest adverb among the slow group, jõuetult ‘strengthlessly’ (M = −1.88, SD = 0.35) is only slightly slower than the fastest slow adverb, ruttamatult ‘without hurry’ (M = −1.27, SD = 0.38).^[1] Nevertheless, the independent two-tailed t-test (with Welch’s correction to account for heterogeneity of variance) shows that the difference is statistically significant (t(280.21) = −14.09, p < 0.001). The same applies to fast adverbs, with kimaga ‘speeding’ being the fastest (M = 0.95, SD = 0.34) and jõudsasti ‘thrivingly’ being the slowest of the fast adverbs (M = 0.01, SD = 0.46), with the difference between them also being statistically significant (t(254.1) = 19.87, p < 0.001). This indicates that adverbs in both groups form a cline, with a distinct speediness reading at both ends of the cline.

Figure 1:

Standardized ratings of the adverbs in Experiment 1 ordered by means (simple adverbs).

To assess whether fast adverbs are significantly further away from the centre of the scale than slow adverbs (as hypothesized), the transformed data (ratings for slow adverbs transformed) was used, and a linear mixed-effects model was fitted using the R package “lme4” (Bates et al. 2015). The response variable was the transformed rating. The predictor was the adverb speed (fast vs. slow) as the fixed effect. Participant was added as a random intercept, with a by-participant random slope for adverb speed (fast vs. slow). The model was statistically significant, χ²(1) = 47.26, p < 0.001 (see Appendix B for the full model output). Figure 2 indicates that contrary to the hypothesis, the ratings for slow adverbs tended to be more spread out (further away from the middle of the scale) than those of fast adverbs, χ²(1) = 54.69, p < 0.001. In other words, slow adverbs have more extreme speed values than fast adverbs, which have less extreme values.

Figure 2:

Predicted probabilities of the transformed speed ratings of the adverbs as a function of adverb speed (fast vs. slow).

3 Experiment 2: complex adverbs

The aim of Experiment 2 was to collect ratings for complex adverbs in order to establish whether, and to what extent, meaning-intensifying word-formation mechanisms influence the conceptualization of speed adverbs. The hypothesis suggests that, compared to slow adverbs, the meaning of fast adverbs is more strongly intensified by the addition of prefixoids (e.g., mega+kiirelt ‘mega fast’, mega+aeglaselt ‘mega slowly’), reduplication (e.g., kiirelt-kiirelt ‘quickly-quickly’, aeglaselt-aeglaselt ‘slowly-slowly’), and simile compounding (e.g., välk+kiirelt ‘lightning fast’, tigu+aeglaselt ‘lit. snail-slowly’). In other words, extreme fastness is perceived as faster than extreme slowness is slow.

3.2 Results

The standardized ratings of the adverbs are presented in Figure 3. The figure is symmetrical, with continuums displayed for slow adverbs on the left and fast adverbs on the right. The slowest adverb is mega+aegluubis ‘in mega slow motion’ (M = −1.28, SD = 0.26), while the fastest of the slow adverbs is rahulikult-rahulikult ‘calmly-calmly’ (M = −0.6, SD = 0.27). The difference between their ratings is significant, as indicated by a two-sided independent t-test (t(193.75) = −18.06, p < 0.001). The fastest of the fast adverbs is välk+kiirelt ‘lightning fast’ (M = 1.21, SD = 0.24), while the slowest of the fast adverbs is kribinal-krabinal ‘quickly, scuttling’ (M = 0.53, SD = 0.34). The difference between the means of their ratings is statistically significant (t(151.34) = 15.79, p < 0.001).

Figure 3:

Standardized ratings of the adverbs in Experiment 2 (complex adverbs).

When compared to Experiment 1, the standardized ratings of complex adverbs in Experiment 2 are higher for both slow and fast adverbs (see Figure 4). For slow adverbs, the mean rating is 0.63 points higher in Experiment 2 (Experiment 2: M = −0.97, SD = 0.33; cf. Experiment 1: M = −1.59, SD = 0.38). An independent two-tailed t-test confirms that this difference is statistically significant (t(5,684) = −71.48, p < 0.001). For fast adverbs, the mean rating is 0.41 points higher (Experiment 2: M = 0.91, SD = 0.3; cf. Experiment 1: M = 0.51, SD = 0.43). Again, an independent two-tailed t-test indicates that this difference is statistically significant (t(11,632) = −62.86, p < 0.001). This suggests that meaning intensification enhances the speediness of fast adverbs. However, rather than intensifying the sense of slowness of slow adverbs, the impact appears to reverse direction, with complex slow adverbs being rated as implying faster activities than simple slow adverbs.

Figure 4:

Comparison of ratings in Experiment 1 (simple adverbs) and Experiment 2 (complex adverbs) across slow and fast adverbs.

Across complexity types, the ratings for fast and slow adverbs in Experiment 2 are closest to each other for reduplications and furthest apart for adverbs with prefixoids and for similes (see Figure 5). In other words, the speed of reduplications (e.g., ruttu-ruttu ‘quickly-quickly’) is rated as less extreme than that of simile compounds (e.g., välk+kiirelt ‘lightning fast’) or adverbs with prefixoids (e.g., ultra+kiirelt ‘ultra fast’).

Figure 5:

Standardized ratings of adverbs across three complexity types (adverbs with prefixoids, simile compounds, and reduplications) and adverb speed (fast vs. slow).

To test Hypothesis 2 and establish whether fast complex adverbs receive higher ratings (in relation to 5 as the centre of the scale) than slow complex adverbs, a linear mixed-effects model was fitted to predict the transformed rating (i.e., ratings measured as the deviance from 5) with adverb speed (fast vs. slow) and complexity type as fixed effects, and participant as a random intercept, along with a by-participant random slope for adverb speed. The results can be found in Figure 6. The model significantly predicted ratings, χ²(3) = 1,033.6, p < 0.001 (see also Appendix B). However, contrary to the hypothesis, there was no main effect of adverb speed (fast vs. slow; χ²(1) = 0.001, p = 0.972). Instead, the rating varied as a function of adverb complexity, with reduplications receiving significantly less extreme ratings than adverbs with prefixoids and simile compounds, χ²(2) = 1,101.06, p < 0.001. This confirms the observation made above based on Figure 5.

Figure 6:

Predicted probabilities of the transformed ratings of adverbs (ratings measured as the deviance from 5).

The inspection of the ratings in relation to prefixoids occurring with both fast and slow adverbs further shows that some prefixoids trigger more extreme ratings than others (see Figure 7). The strongest intensification is triggered by the reduplicated prefixoid üliüli- ‘super-duper’ (mean of −1.27 for slow and 1.17 for fast adverbs), surm- ‘death’ (mean of −1.21 for slow and 1.02 for fast adverbs), ulme- ‘sci-fi’ (mean of −1.19 for slow and 1.1 for fast adverbs), and ultra- ‘ultra’ (mean of −1.15 for slow and 1.14 for fast adverbs). It is worth noting that surm- ‘death’ as a prefixoid is more left-skewed than the other prefixoids, meaning that this prefixoid seems to slow the meaning of both slow and fast adverbs. Moreover, the ratings for surm- ‘death’ adverbs are more extreme (mean −1.21 for slow and 1.02 for fast adverbs) than for those with the prefixoid elu- ‘life’ (mean −0.94 for slow and 0.87 for fast adverbs).

Figure 7:

Standardized ratings of complex adverbs across prefixoids.

As for reduplications, the repetition of the adverb three times (triplications; ruttu-ruttu-ruttu, kiirelt-kiirelt-kiirelt ‘quickly-quickly-quickly’; aeglaselt-aeglaselt-aeglaselt ‘slowly-slowly-slowly’) and the reduplication of a compound (i.e., aegluubis-aegluubis ‘in slow motion’) enhance the meaning of speed more strongly than the two-time repetitions (e.g., kiirelt-kiirelt ‘quickly-quickly’, aeglaselt-aeglaselt ‘slowly-slowly’; see Figure 8).

Figure 8:

Standardized ratings of reduplications of adverbs.

4 Experiments 1 and 2: response time analysis as a follow-up analysis

To further examine whether adverb speed has behavioural underpinnings, we conducted an exploratory analysis of response times. In particular, we assessed whether fast adverbs trigger faster responses than slow adverbs. To do this, we combined the two datasets and fitted a linear mixed-effects model to predict response time as a function of adverb type (fast vs. slow) and adverb rating (transformed ratings), controlling for adverb complexity (simple vs. complex; i.e., Experiment 1 vs. Experiment 2), adverb frequency, and adverb length in letters (range 4–29). To assess the possible impact of combining the two datasets (from Experiment 1 and Experiment 2), three pairwise interactions were added to the model, involving the key variables: adverb transformed rating, complexity (simple vs. complex), and adverb semantic category (fast vs. slow). Pairwise interactions were preferred over a three-way interaction for reasons of clarity, and because the more complex model did not provide a better fit, as indicated by the nearly identical Akaike information criterion (AIC) values. Participant was again included as a random effect, with a by-participant random slope for adverb speed (fast vs. slow). Response time was measured as the total time taken to evaluate the meaning of a word, that is, from the onset of the adverb’s presentation on the screen to the moment the participant pressed the button to proceed to the next page (i.e., the next adverb).

The results are presented in Figure 9 (see also Appendix B). The model was statistically significant, χ²(8) = 86,49, p < 0.001, and all three interactions significantly predicted response times. The interaction between adverb speed (fast vs. slow) and adverb transformed rating, χ²(1) = 5.24, p = 0.022, showed that fast adverbs receiving more extreme ratings were associated with shorter response times than fast adverbs rated closer to 5, the middle point of the scale; slow adverbs were insensitive to this effect. The interaction between adverb speed (fast vs. slow) and adverb complexity (i.e., whether the adverb originated from Experiment 1 [simple] or Experiment 2 [complex]), χ²(1) = 9.81, p = 0.002, indicated that fast adverbs elicited shorter response times particularly among complex adverbs (Experiment 2), although a similar trend was observed with simple adverbs (Experiment 1). The strongest interaction, between adverb transformed rating and complexity, χ²(1) = 26.21, p < 0.001, revealed that response times decreased as ratings increased for complex adverbs (Experiment 2), whereas simple adverbs (Experiment 1) showed no such sensitivity. In addition, there was a main effect of adverb frequency, χ²(1) = 7.81, p = 0.005, showing shorter response times for adverbs with higher frequencies. There was no effect of adverb length (p = 0.425). In other words, response times did not differ as a function of adverb length in letters.

Figure 9:

Predicted probabilities of response times.

Overall, the results indicate that not only adverb frequency, but also adverb speed (fast vs. slow), rating extremeness, and adverb complexity affect response times in that fast adverbs with more extreme ratings tend to trigger shorter response times, particularly when they are complex.

5 Discussion and conclusion

In this study of Estonian speed adverbs, we hypothesized that mental simulation would occur during the semantic processing of adverbs, and that this would be reflected in the speed ratings attributed to them. More specifically, since fastness is more intense than slowness (receives predominant processing in terms of processing speed and enhanced neural activation (Hülsdünker et al. 2019; Stolz et al. 2023; Wang et al. 2003; Yan et al. 2023) and could also evoke higher subjective intensity), we expected that ratings of fast adverbs would diverge more significantly from the neutral midpoint of the rating scale than those of slow adverbs (i.e., fast adverbs are more intense than slow adverbs).

However, contrary to this hypothesis, slow simple adverbs tended to receive more extreme values than fast simple adverbs (Experiment 1), and no difference was found for complex adverbs (Experiment 2). If mental simulation had occurred, we would have expected the opposite: fast adverbs would result in ratings that deviate further from the centre of the scale than slow adverbs, as fastness is more intense and allows one to cover longer distances in shorter periods of time. Indirect support for this comes from studies showing that displacement effects vary as a function of motion speed. That is, people tend to estimate the vanishing point of a moving object as being farther away from its actual vanishing point in response to fast motion (Actis-Grosso et al. 2008; Hubbard 1995). In addition, thought translates easily to behaviour when it comes to speed (Correia et al. 2022), hand movement speed increases in case of action-congruent linguistic manipulation (Fargier et al. 2012), and, in general, language can evoke motor patterns characteristic to the linguistically expressed action as “implicitly or explicitly, action involves meaning and intention” (García and Ibáñez 2016: 78).

As an exploratory follow-up study, we found that shorter response times were more likely to occur in response to (i) fast adverbs compared to slow adverbs, (ii) adverbs receiving more extreme ratings, and (iii) adverbs with higher frequencies. The finding that fast adverbs are rated more quickly than slow adverbs replicates results for the processing time of speed words in other languages, such as English (Stites et al. 2013), Japanese (Irie et al. 2021), and Mandarin (Pan et al. 2024). As such, it provides converging evidence for grounded cognition (Barsalou 2008; Gibbs 2006), suggesting that language triggers behavioural responses characteristic of processing the physical activity itself, as fast motion is generally processed faster than slow motion (e.g., Hülsdünker et al. 2019). The fact that high-frequency words are processed faster than low-frequency words is well established in the literature (Brysbaert et al. 2018). However, the finding that adverbs with more pronounced semantics in terms of describing extreme fastness or slowness are processed faster requires further research, as does the finding that adverb type (fast vs. slow) and rating extremeness influenced the response times particularly in the case of complex adverbs, whereas simple adverbs were rather insensitive to this impact. It may be that the meaning of such adverbs is clearer, making them easier and quicker to decide upon.

The two experiments enable us to evaluate the differences between simple and complex adverbs. The main conclusion we can draw from this comparison is that adverb complexity is associated with implied speed. Notably, simple adverbs in Experiment 1 were rated as significantly slower than complex adverbs in Experiment 2, even though the same rating scale was used in both experiments. This suggests that meaning intensification may be fast-biased in that within the context of intensification (as in Experiment 2), even slow adverbs acquire some traits of fastness or urgency that are otherwise absent in simple adverbs. Alternatively, complex adverbs, being longer than simple adverbs, may reflect the cognitive load effect of intensifiers reported by Bennett and Goodman (2018), whereby longer words require greater cognitive load and, as a result, are perceived as more intense.

Furthermore, among complex adverbs, reduplications tended to receive the least extreme ratings. This suggests that reduplication may be a less effective device for meaning intensification than simile compounding or adding prefixoids to adverbs. Nevertheless, extreme reduplications (e.g., reduplicated prefixoids such as üliüli- ‘super-duper’ and triplications such as kiiresti-kiiresti-kiiresti ‘quickly-quickly-quickly’) strongly enhance the meaning of an adverb. This novel finding regarding prefixoids vs. reduplications and their differences in intensity has not been addressed in the literature. The more modest intensifying function of reduplications compared to prefixoids may also be related to the diverse functions that reduplications can have in a language, such as approximation, diminution, and so on (Barbaresi and Dressler 2020). This may dilute the intensifying function of reduplications, whereas intensifying prefixoids, as the name suggests, are specialized for intensification, providing a clearer and stronger means for this purpose.

Taken together, the results regarding mental simulation revealed varied outcomes. Contrary to the hypotheses, fast adverbs do not seem to imply more extreme speed than slow adverbs, at least based on the rating studies. However, response times suggest evidence of mental simulation, as fast adverbs were processed more quickly than slow adverbs. Additionally, meaning intensification through word-formation mechanisms appears to be fastness-biased, with both slow and fast complex adverbs receiving significantly higher speed ratings than their simple counterparts.

As such, the results add to the growing body of research on mental simulation by highlighting behavioural patterns association with linguistic factors, and thus with contextual information, and by diversifying the methodological approaches applied in this domain (see also Ibáñez et al. 2023; Körner et al. 2023). They also advance linguistic knowledge on adverbs (see also Duplâtre and Modicom 2022; Hallonsten Halling 2018), and speed adverbs in particular (see also Taremaa et al. Under Review), word-level intensification mechanisms (see also Barbaresi and Dressler 2020; Dressler and Barbaresi 1994; Iacobini 2023), and potential asymmetries in how speed is expressed in language, as evidenced by studies of motion verbs (Ikegami 1969; Snell-Hornby 1983), adverbs/adjectives (Plungian and Rakhilina 2013; Schäfer 2023), clausal patterns of motion events (Taremaa and Kopecka 2023), speech rate (Perlman et al. 2015; Taremaa et al. 2024), and phonemic iconicity (Zhao and Wu 2025).

Future studies can use the speed ratings of the adverbs to investigate their clausal patterns in relation to the adverbs’ implied speed. More controlled lab-based experiments are also needed to validate the response time findings from the web-based experiments. In addition, an experiment that includes both simple and complex adverbs may be necessary to directly compare evaluations of simple adverbs with their modified, more intense complex forms. Cross-linguistic research is also needed, collecting speed ratings for speed-related lexemes in different languages, in order to validate the present findings and assess potential language- or culture-specific dimensions embedded in adverb meanings.

Overall, the current study contributes new empirical evidence to the study of mental simulation, word-level intensification, and the linguistic encoding of speed. By analysing both simple and complex speed adverbs and their effects on rating behaviour and response times, the study provides novel insights into asymmetries in speed expression. The findings also highlight the role of word formation in shaping semantic evaluation and processing speed, offering a valuable basis for future work at the intersection of psycholinguistics and lexical semantics.