The complexity and diversity of insults make them a challenging subject to analyse from multiple disciplinary perspectives. This is due to the fact that insults encompass social, psychological, legal, linguistic, and other aspects of human life. Over the years, and with particular reference to communication approaches, insults have been a significant area of interest from a pragmatic standpoint.

In the domain of speech impoliteness, various forms of pejorative language coexist, including insults, accusations, criticisms, threats, slurs, swear, and taboo words. Distinguishing between these speech acts can prove challenging, as their boundaries are often gradational and overlapping, and the theoretical frameworks used to delineate them exhibit considerable variability. Within the wide context of impolite speech acts, insults represent merely one of several negative acts aimed at demeaning the target. They represent a form of aggressive verbal communication, including a broad spectrum of expressions intended to cause emotional and psychological distress, and to damage the reputation of a given target (either an individual or a group), often leading to social conflict and physical confrontation. Insults are a manifestation of linguistic impoliteness or verbal aggression, as they always involve a high degree of social transgression and censorship (Moïse, 2006). The issue becomes more complex when it is recognised that insults can be considered as both a form of impolite speech and, in particular contexts, as a form of hate speech. This phenomenon occurs when insults are directed towards a specific social group on the basis of race, religion, sexual or political orientation or other identitarian factors. Such insults are frequently associated with expressions of prejudice, negative stereotyping and incitement to violence or hate (Culpeper, 2021; Guillén-Nieto, 2023). Consequently, insults can be distinguished by their level of social acceptance or violation, as well as by their degree of semantic offensiveness.

Over time, scholars have proposed various taxonomies of insults. For instance, Jucker & Taavitsainen (2000, p. 74) distinguished insults based on the following criteria: formal degree, context-dependence, semantics, speaker-attitude and reaction. Similarly, Mateo & Yus (2013) provided a detailed categorisation of insults, encompassing three key aspects: intention, interpretation, and reaction.

According to Culpeper (1996, p. 357), insults can be regarded as a form of positive impoliteness output strategies, with the purpose of threatening the positive face of the speaker. In the taxonomy of impoliteness formulae proposed by Culpeper (2011a, pp. 135-136; 2021) and van Olmen et al. (2023, pp. 24-25), insults can be categorised as relating to quality, relational and social identity face, highlighting their social nature.

The etymology of the word already suggests a negative content. The term derives from the Latin verb IN+SULTĀRE, which literally means "to jump on," and is metaphorically used to convey the act of attacking. Consequently, insults are often associated with words that allude to physical violence, such as strokes, punches, slaps, or bullets.

In the field of pragmatics, insults present a number of interesting aspects that have been the focal point of extensive research (see, inter alia, Jucker & Taavitsainen, 2000; Mateo & Yus, 2013; Alfonzetti, 2017). Studies have identified a number of subtypes of insults, including pure unfactual insults, calumnies, and factual put-downs (Feinberg, 1985).

From a more general pragmatic perspective, doing away with possible subtyping, insults can be considered as a hybrid phenomenon, combining the illocutionary force of a verdict and that of an expressive act (Jucker & Taavitsainen, 2000; Moïse, 2006; Alfonzetti, 2017). This hybrid nature is attributed to the fact that insults convey a negative evaluation of an individual or objects. Moreover, they are known to elicit a perlocutionary effect that exploits a speech face-attack to immediately produce a form of violence and denigrate the offended person. The act of insult is defined as a speech act in which both the speaker (the insulter) and the addressee (the insultee) assume actives roles. Insulting actions are typically deliberate and targeted towards a specific individual by the speaker ¹ . In contrast, the addressee of the insult is not merely the recipient of the negative assessment; they also play an active role in evaluating the offensiveness of the act by implementing various strategies, ranging from no reaction to the insult (silent and ignoring strategies) to counter-insult. This latter reaction frequently gives rise to a verbal escalation ² .

Furthermore, in the absence of a verb that has an explicit performative function (e.g. *I insult you), insults are typically expressed through the use of the so called pejorative language. This is defined as the use of lexemes with a negative semantic value, which are drawn from semantic fields that are often considered taboo (e.g. sex, bodily fluids, genitals, etc.). However, recent studies have shown that any word can be employed with a dysphemic value, when the speaker uses specific facial expressions, aggressive gestures, body postures, and certain prosodic and paralinguistic signals. In spoken language, a range of prosodic strategies, in conjunction with specific lexical choices, intensifying modifiers and morpho-syntactic structures, are used to increase the intensity of insults (Culpeper, 2011b). In contrast, in written and computer-mediated discourse (CMD), graphic devices, such as emojis, emoticons and the inappropriate use of capital letters, or punctuation marks, are frequently employed in conjunction with or as substitutes for text as symbols of affective states ³ .

The majority of studies on impolite speech acts have focused on pragmatic strategies, neglecting the analysis of prosody for many years. Due to their peculiar pragmatic nature, insults are never neutral speech acts, since their expressive and derogatory force is conveyed through both form and content. In spoken language, insults are characterised by a substantial alteration of all acoustic parameters. Nevertheless, there is a surprising lack of studies on the prosody of insults, probably due to the difficulty of collecting them in spontaneous contexts. However, extensive research on linguistic (im)politeness and on the acoustic correlates of emotions offers valuable insights.

In the field of the impoliteness theory, Culpeper (1996; 2011a; 2011b) and Culpeper et al. (2003) conducted pioneering research on impolite prosody, focusing on exploitative TV shows and quizzes or intercepted phone-calls. The researchers demonstrated the feasibility of employing impoliteness strategies, whether negative or positive, through the utilisation of prosody alone. They identified several strategies, the first of which is employed to hinder the linguistic process, as it serves to interrupt the interlocutor’s turn in a dialogue, conveying a sense of speech closure and dissociation. In terms of intonation, such utterances are characterised by high f ₀ onsets and falling intonation contours.

The second strategy involves invading the auditory space of the listener by increasing the volume and the pitch range of the voice. It is important to note that speakers are sensitive to the perception of appropriate levels of voice in everyday interactions. In interpersonal communication, the production of a speech that exceeds the necessary level of loudness, given the physical distance between interlocutors, is often perceived negatively and may be interpreted as an auditory invasion. The use of a high-pitched voice in communication can be considered an attempt to openly express one's emotional state to another individual during the course of an interaction. This strategy can be employed as standalone act of rudeness or, when combined with other strategies such as insults, as a means of increasing the offensiveness of an utterance ⁴ .

Nevertheless, Culpeper et al. (2003) posit an observation pertaining to the predictability of prosodic cues. They further note that the association between these indices and pragmatic meanings is frequently contradictory. For instance, in some languages, certain phonetic cues are employed to convey polite nuances, while in others, these cues are utilised to convey impolite meanings ⁵ . With regard to this aspect, research has shown that the pitch range is one of the most problematic acoustic parameters when comparing the prosodic cues of polite and impolite utterances (Brown & Prieto, 2017; Caballero et al., 2018), as well as distinguishing between genuine and mock impoliteness (McKinnon & Prieto, 2014; Andreeva et al., 2016). This can be attributed to the gradient and expressive nature of tonal excursion (both in level and span), as well as its direct involvement in the experience of emotions.

In some languages, a wider pitch range may be perceived as more polite to the listener, as is the case in Dutch, English (Chen et al., 2004), Canadian English (Caballero et al., 2018), and Japanese (Ohara, 2001). Conversely, a narrower pitch extension tends to convey rudeness (Uldall, 1960). In contrast to this trend, Stadler (2007) demonstrated that high pitch can convey aggressive attitudes in German and New Zealand English. Furthermore, Winter & Grawunder (2012) found a correlation between a narrow pitch range and the use of a polite register in Korean speakers. Finally, Andreeva et al. (2016) found no significant differences in pitch range when comparing genuine and mock utterances. The researchers employed both production and perception texts to compare four utterances with an offensive or supportive attitude produced in German and Polish. The two languages seem to select different prosodic indices, e.g. mean intensity is used to distinguish between the two attitudes in Polish, but not in German ⁶ . Conversely, the opposite is true for speech rate, with German speakers exhibiting a higher rate of speech in a supportive context. In contrast, the nuclear pitch accent in Polish serves a differentiating function, exhibiting a falling intonation in instances of genuine impoliteness and a rising intonation in instances of mock attitude. However, in German, no clear differentiation is exhibited.

In a similar way, duration and speech rate show conflicting trends. The use of polite speech is often slowed down by the inclusion of pauses and fillers, which serve to convey a greater sense of insecurity and the conveyance of politeness (Winter & Grawunder, 2012). In Japanese, polite speech is characterised by slower speech rate (Ofuka et al., 2000); however Álvarez et al. (2011) found for Venezuelan Spanish that impolite utterances showed greater temporal variability and longer syllables than polite or neutral ones. In parallel, a study focusing on the prosodic distinction between polite and impolite requests elicited by controlled tasks in Canadian English found slower speech rates and longer durations in rude utterances (Caballero et al., 2018). These results suggest a complex interpretation of linguistic impoliteness, in which the role of phonological and phonetic indices (local and global) is combined with both lexical and morpho-syntactic indices and information derived from the pragmatic context, leading to a dynamic scenario.

In the domain of hate speech studies, a novel research approach has been developed which analyses the correlation between the prosody of impolite speech acts and their level of social acceptability. In Niebuhr & Neitsch (2020), 28 participants were presented with offensive stimuli in German that were intended to convey irony or genuine insults towards the Holocaust. The participants were asked to assess the personal acceptability and degree of social violation of the stimuli. The results indicated that stimuli with lower acoustic indices, including mean f ₀ and speech rate, were evaluated as less acceptable and more socially censored.

The quality of the voice and the use of non-verbal cues, such as gestures, facial expressions, gaze and bodily movements, also constitute a significant factor in the performance of insults in spoken language. The research on multimodal aspects of impolite speech revealed the pivotal role of non-verbal cues, including gestural, physical proximity towards the addressee, hand and arm gestures, many of which have obscene connotations, and overt alterations of facial expression, such as frowning eyebrows (McKinnon & Prieto, 2014; Brown & Prieto, 2017). The gestural component is frequently accompanied by a hoarse, rough, and nasalised voice which is typically indicative of the expression of anger and aggression ⁷ .

As regards Italian language, Vinciguerra (2023b) presented the first experimental findings concerning the speech rate in insults. The insults, comprising personalised negative vocatives as the incipit of impolite sentences, exhibited no differences in speech rate when compared to unmarked utterances. Nevertheless, the analysis demonstrated that the epithet was produced by the speakers at a substantially slower rate than the “impolite coda”. The author postulated that the impolite utterances may have undergone an overall slowing down due to the insulting epithet being placed in the initial position.

The results indicated that insults have distinctive acoustic properties when compared to the utterances employed as control. The study revealed that the mean intensity of offensive sentences was higher (+6 dB) (see Figure 1) and statistically significant, as indicated by the one-way Anova [F(3, 180) = 6.02, p = .001] and Tukey post-hoc test. In detail, INS_male/CON_male (p = .037); INS_female/CON_female (p = .020); the comparisons between INS_male/INS_female: (p = .755) and CON_male/CONf_emale (p = .801) do not reach statistical difference.

The first hypothesis was supported by the results; the actors employed a louder volume of voice when uttering insults, in comparison to the control forms. In addition to being more intense, the insulting portion tends to be expressed with a slower speech rate (-1 syll/sec), with an average rate of 5 syll/sec, as shown in Figure 2.

The mean speech rate in insults was observed to be lower than that calculated for the control sentences, despite the high level of expressiveness and emotional value of the former. However, this difference was not found to be statistically significant [F(3, 180) = .856, p = .465]. Therefore, the second hypothesis was not confirmed.

In contrast with the initial hypothesis, the production of insults was not found to be significantly slower. However, it should be noted that there is a certain degree of variability in the way speakers perform these speech acts. While some speakers showed no notable difference between insults and control sentences, two female subjects (F ₁ and F ₃) employed contrasting strategies for insults. Specifically, subject F ₁ used a faster speaking rate (6.8 syll/sec), while subject F ₃ used a slower rate (4.5 syll/sec).

Despite the absence of statistical significance with regard to the difference in speech rate, a clear discrepancy was observed in the duration of the nuclear vowel in the intonation unit containing the insult when compared with that of the control sentences. In order to mitigate the potential for phonetic distortions, all vowel durations were normalised based on the speaking rate of each participant. Normalisation made it possible to reduce the amount of temporal conditioning between fast and slow speakers. As illustrated in Figure 3, the duration of the vowels on which the nuclear pitch accent falls underwent a process of lengthening in insults, resulting in a temporal prominence that was also discernible through perceptual feedback. The statistical comparisons conducted using One-way Anova yield significant results [F(3, 180) = 10.35, p = .000], as did comparisons between subgroups; INS_male/CON_male (p = .001); INS_female/CON_female (p = .000). The post-hoc analysis between INS_male/INS_female did not reach statistical results (p = .996), as well as CON_male/CON_female (.939); this suggest that the groups are homogeneous, regardless of the gender of the subjects.

The Pearson correlation coefficient indicates that there is a linear negative relationship between the speech rate of the sentences and the duration of nuclear stressed vowels. An increase in the duration of nuclear stressed vowels (NV) is associated with a decrease in the speech rate of sentences; female groups= r(45) = -.85, p=.000); male groups= r(45) = -.75, p=.000.

The data suggests that insults are generally associated with a temporal decrease in speaking rate, rather than an increase as originally postulated. This is supported by the observed behaviour of nuclear stressed vowels. In the control groups (both male and female groups), the normalised duration of the nuclear vowel shows a modest degree of dispersion, as evidenced by the relatively low height of the boxplot. In the control group, 50% of the values were observed to fall between 29 and 32 ms (normalised durations), whereas the time range for the insult group exhibited a wider distribution.

With regard to intonation, it was observed that insulting expressions were produced as distinct tonal units, separated from the rest of the utterance by a pause or a boundary tone. Insults were typically identified by an intonation contour that, after a high f ₀ onset, remained at a high level without reaching a true f ₀ minimum. The remainder of the statement showed a falling intonation pattern, which was consistent with the conveyed offensive meaning.

For each speaker, the mean and median fundamental frequency (f ₀) were higher in the insults than in the control sentences. This had an impact on both the pitch level and pitch span. The behaviour of the pitch range is unclear as it was unexpectedly lower (-5 ST) in the insulting sentences compared to the control group (Figure 4). This differs from the findings of Culpeper et al. (2003) and the hypothesis proposed in our research ¹² . Nevertheless, despite the presence of some degree of internal variability (±3 ST), a one-way ANOVA between groups yielded a statistically significant result [F(3, 180) = 31.62, p = .000], as did comparisons between subgroups; INS_male/CON_male (p = .000); INS_female/CON_female (p = .000). The post-hoc analysis between INS_male/INS_female did not reach statistical significance. (p = .095).

A review of the data indicates that the highest normalised duration of nuclear vowels is observed in the utterances where the highest pitch range values are also reached. In order to assess the relationship between these two parameters, a Pearson correlation coefficient was calculated between the duration of the nuclear vowel (NV) and the pitch range value in semitones within the same sentence. The participants were divided into two groups according to gender. The results demonstrated that in both groups, the two parameters exhibited a linear positive correlation, with an increase in one parameter (duration) resulting in a increase in the other (pitch range); female groups= r(45) = .44, p =.001); male groups= r(45) = .39, p =.004.

The data indicates that the pitch range in insults was narrower. This is due to the fact that, in many cases, the f ₀ minimum was not reached at the end of the sentence. The difference between insults and control sentences in terms of semitones is evident for all participants except for speaker F ₂ (see Figure 5).

To shed light on this unexpected outcome, an attempt was made to more accurately quantify the intonation variation of the sentences by employing the median f ₀ as the neutral frequency point to represent the pitch level. The distances in semitones between the median fundamental frequency (f ₀x̃) of each sentence and the maximum and minimum values of fundamental frequency (f _{0 max} and f _{0 min}) were calculated for each speaker. In the sample of insults, the median f ₀ is higher compared to that of the control sentences (M: p = .03; F: p = .04).

As shown in Figure 6, regardless of gender, the f _{0 x̃}/f _{0 min} distance remained closer to the reference point, graphically represented by the zero value, while the f _{0 x̃}/f _{0 max} distance was smaller. It should be noted that the average value of f _{0 min} is higher in the insults than in the control sentences (males: .005; females: .006). Diversely, the value of f _{0 max} may undergo a substantial increase, as this parameter determines the degree of expressiveness and prominence.

Thus, in insults, the pitch level is higher than in control speech, while the pitch span is reduced. The pitch range of sentences containing insults was found to be contracted, due to the failure to reach a minimum frequency value, resulting in a smaller range than that observed in the control speech. This trend aligns with the use of the vocative function conveyed by the nominal phrase employed during the construction of the target sentences.

The study has produced preliminary findings on the prosody of Italian insulting expressions, with significant trends being observed in the data. However, it is important to note that the results only pertain to linguistic forms elicited by professional theatre actors in predetermined contexts. Although the speech sample was produced with a high degree of naturalness and without inhibition, it cannot be considered representative of spontaneous speech due to its controlled nature. To address this limitation, a new corpus of insults produced by naive speakers in less controlled contexts is currently being collected. This will provide a basis for comparison with the existing data in future research.

The analysis showed that insults have a specific prosodic realisation that leads to the activation of numerous acoustic indices. The offensive nature of insults is conveyed not solely through lexical selections and the conveyance of a negative polarity evaluation, but also by means of phonetic components. It is noteworthy that insults share diverse aspects with highly activated negative emotions, such as hot anger, as they always contain an aggressive component achieved through a gradual increase in voice volume and tonal extension.

In their research, Culpeper (2011a; 2011b) and Culpeper et al. (2003) postulated a broadening of the pitch range of impolite sentences, but our findings did not fully support this. Instead, it was observed that the pitch range shifted upwards while remaining compressed, creating a perceptual effect of a dominant voice. Furthermore, the pitch span has been found to correlate with the duration of the nuclear vowels, with a longer duration corresponding to a wider pitch range.

The effect was further enhanced by the use of a scratchy voice quality, resulting in a verbal attack that sounded aggressive. This overpowered the auditory space of the addressee, who is denigrated and, at the same time, dominated. This is in line with the fact that all types of violent speech, including insults, along a semantic continuum, are overt strategies of denigration and clear face-attacks.

There is a belief that insults are a form of invasion into the space of the listener. The speaker adopts a centric-position, appearing dominant and tending to invade the personal and auditory space of the listener. This acoustic strategy is the result of several processes. The most noticeable of these strategies involves increasing the volume of the voice, making the insult perfectly audible and effectively offensive ¹³ . The maintenance of a high pitch level serves to emphasise insults and ensure the transmission of their hurtful content, a behaviour consistent with the expression of intense anger. The findings of the research indicate that the duration of nuclear vowels has also a significant impact on the production of insults.

It is our opinion that the invasion of auditory space occurs along two dimensions: vertical and horizontal. On the vertical axis, it results in an increase in voice volume and pitch level, a voice strategy which serves to ensure audibility and give prominence to the insult. On the horizontal axis, the insult causes the nuclear vowel to be lengthened, making it more prominent, not only in terms of loudness and f ₀ frequency, but also in terms of duration. The use of an insulting word has been shown to take up more time, thereby creating a specific effect of dominance and hostility. At the same time it denies the offended person the opportunity to take the turn in the dialogue, which can be seen as a form to prevent a reaction. Verification of this behaviour should be carried out in natural conversations.

There is still much work to be done in this area. One noteworthy finding from our analysis is that during the production of target sentences, the actors frequently hurled insults at an unknown listener, resulting in a performance that closely resembled a spontaneous interaction. However, the speaker occasionally opted to deliver the insult in a hushed tone, almost whispering it, which can be viewed as an alternative but equally offensive tactic. It seems reasonable to posit that this behaviour is responsible for a certain degree of cross-speakers variability, and thus represents a worthy subject for future investigation. This data appear to suggest that the average speech rate and intensity may be influenced by the speaker's position on the dominance and power continuum as well as the nature of the interaction. In this regard, a perception test is being administered to assess the speakers' degree of offensiveness and dominance.

Moreover, a comprehensive examination of prosodic indices in a less controlled speech sample, encompassing aspects such as voice quality, is essential to enhance our understanding of the role of parameters in conveying offensive meaning.