How Has Poets’ Reading Style Changed? A Phonetic Analysis of the Effects of Historical Phases and Gender on 20th Century Spanish Poetry Reading

Abstract

Poetry reading remains a largely underexplored area in phonetic research. While previous studies have highlighted its potential and challenges, experimental research in the Spanish context is still limited. This study aims to examine the evolution of Spanish poetry reading over time, focusing on its main prosodic features. Applying the VIP-VSP phonetic model to 40 poetry recordings, we analyzed the organizational and prosodic indices that characterize poetry reading. Mean speech rate, plenus (the ratio of speaking time to pausing), and pitch span emerged as key parameters for capturing change. The results identified two distinct historical phases—first and second radio-television—showing significant effects on speech rate, plenus, and pitch span: speech rate and pitch span increased over time, while plenus decreased. Gender also played a key role, with female voices exhibiting significantly higher values in both pitch span and plenus. Variability and recurring strategies were observed within and across authors. This study confirms that poetry reading has evolved along a ‘stylistic-chronological’ trajectory, while also reflecting gender-based distinctions. These findings underscore the need for interdisciplinary analytical approaches and diversified classification groupings to fully capture the complexity of this mode of speech.

1. Introduction

The vocal dimension is crucial in poetry—not only in oral poetry but also in the performance of the written tradition—as an integral part of this literary genre, its transmission, and its place within the broader cultural system. This type of speech represents a valuable cultural heritage that must be preserved, promoted, and studied, and it plays a significant role in both linguistic and interdisciplinary research. However, despite the growing number of poetry vocal archives and the increasing scholarly attention to the phonic dimension of poetry, much remains to be done in the field of experimental phonetic research on poetry reading. This study aims to provide some insights into the evolution of Spanish poetic reading in the last century through the methodology of experimental phonetics.

1.1. Brief Overview of Phonetic Studies on Poetry Reading

Looking back at the origins of phonetic studies on poetry reading, we find their beginnings in the early 20th century, in two French phonetics laboratories. As Mustazza (2019) explains, phonetic interest in poetry recordings dates back to 1913—the same year in which four major Futurist works and manifestos were published. On the one hand, Abbé Jean-Pierre Rousselot, while studying vocal duration in regional French, described Ezra Pound’s voice to the poet himself using a kymograph. On the other hand, Ferdinand Brunot recorded and preserved Guillaume Apollinaire’s voice in collaboration with a recording company, ensuring its inclusion in Les Archives de la Parole [The Archives of the Word], specifically in the Les Voix Célèbres [Famous Voices] section. Founded by Brunot in 1911, this archive marked the first attempt to map French dialects and also signaled the beginning of phonetic and phonological analyses of poetry.

Throughout the 20th century, poetry vocal archives—whether developed for research or for public dissemination—played a key role in preserving and promoting poetry, opening new avenues for research. Among the earliest and most significant efforts in preserving and studying poets’ voices is the work on oral poetry conducted by Milman Parry in the early 1930s, and later continued by Albert Lord (Elmer, 2013). Their studies would go on to influence the theories of Ong (1967) and Zumthor (1983). Vocal poetry archives are thus a valuable resource not only for their cultural and literary significance (Lang, 2019), but also as essential tools for phonetic and interdisciplinary research.

Roman Jakobson (1966) was among the first to recognize the value of poetry reading, proposing a descriptive framework for oral approaches to poetic texts. Later, Iván Fónagy (1983) further emphasized this perspective. However, it was not until the 1970s that experimental phonetics and interdisciplinary research on poetry recordings began to develop more systematically.

Earlier, linguistic interest in the prosody of verse had already emerged in the early 20th century, initially focusing on rhythm. This soon expanded to include comparisons with music, intonation, and the identification of characteristic patterns. Following Sievers’ (1912) early contributions, phonological studies on rhythm were carried out by Servien (1930) and Grammont (1913, 1933). Drawing on concepts common to music, poetry, and phonology—such as hierarchical groupings and metrical grids—studies by Kruckenberg et al. (1991) emerged in the 1990s. A more music-focused comparison is found in the works of Nord et al. (1990), Fant et al. (1991), Lerdahl and Halle (1991), Lerdahl (2003), Pamies Bertrán (1999, 2010), and Fabb and Halle (2008, 2012).

Studies on intonation, particularly regarding disambiguation in poetry, began as early as the 1950s with Chatman (1956, 1966) and Loesch (1965, 1966), followed by the prosodic analyses of Funkhouser (1979), Cauldwell and Schourup (1988), Brazil (1992), and Cauldwell (1994). Its central role in free verse, as well as in detecting frequent patterns, was initially proposed by Mukafovsky (1933) and later demonstrated by Taglicht (1971), Crystal (1975), and Fónagy (1983, 2000), who also identified specific pauses and mismatches between verbal and metrical stress.

Further investigations into verse recitation patterns were conducted by Byers (1977, 1979, 1980) and Barney (1998, 1999), and later by Tsur (1997a, 1997b, 1997c), whose work—extended within a cognitive poetics framework—was further developed by Gafni and Tsur (2017) and Tsur and Gafni (2022). Byers (1979) provided an instrumental analysis of the poetic voice, identifying a “formula for poetic intonation” characterized by the following elements: slow speech rate, short tone units, frequent pauses, relatively equal-length units, low average pitch, narrow pitch range, simple falling melodies, and nuclei. Barney confirmed and refined this “formula” through comparisons with radio and liturgical speech, highlighting additional features such as semantically unexpected pauses and intonational echo patterns. Related to the particular use of pitch, Rumsey (2015) contributed to the theory of “downdrift” in French, describing a gradual lowering of pitch range.

Recent interdisciplinary research combining phonetics, poetry and digital humanities has significantly advanced the field across several linguistic contexts (e.g., Schirru, 2002; Wagner, 2012; MacArthur, 2016; MacArthur et al., 2018; Barbosa, 2022, 2023; Colonna, 2022; Baumann et al., 2023; Colonna & Romano, 2023a, 2023c; Wagner & Betz, 2023).

Following the notion of a recurring “formula” in poetic speech, MacArthur (2016) introduced the idea of a “monotonous incantation” style in academic poetry performance, consistent with theoretical models such as Cohen (1966). MacArthur et al. (2018, p. 61) defined “poet voice” as being marked by slow pitch speed, slow pitch acceleration, narrow pitch range, low rhythmic complexity, and slow speaking rate. However, their quantitative study of 100 poets born before and after 1960 revealed stylistic variation that challenges the idea of a singular “poet voice”. The analysis pointed to generational, gender-based, and cultural differences—for example, a less expressive style among younger female poets as a socio-cultural marker, or the use of creaky voice to emulate male patterns. Furthermore, comparisons between poets’ readings and conversational speech showed additional distinctions between speech types.

In Romance languages, Schirru (2002) offered the first phonetic descriptions of Italian poets’ reading styles. Colonna (2022) developed the first phonetic history and analysis of 20th-century Italian poetry reading, identifying both inter- and intra-author variation as well as diachronic trends. A three-phase classification—covering the first and second radio-television eras, followed by the digital era—revealed an increased speech rate, longer pauses, and a reduced pitch span over time in the transition from the first to the second era. This qualitative model, based on a set of indices for poetry reading description (see Colonna & Romano, 2023c), has since been applied to English (Colonna & Romano, 2023b), French (Colonna, in press), and Spanish (Colonna et al., 2024), the latter also serving as a starting point for this study. A comparative approach involving multiple voices reading the same text (Colonna, 2017; Colonna et al., 2019) enabled the identification of groupings based on various criteria. A recent gender-based comparison (Colonna & Román Montes de Oca, 2024b) revealed notable convergences, with pitch span emerging as the main point of divergence.

Comparative studies between prose and poetry have also highlighted distinct prosodic features. Wagner (2012) examined this from a phonetic–perceptual perspective in German, identifying distinct metrical groupings. Similarly, Colonna and Romano (2023a) applied perceptual phonetics to Italian, showing how poetry is distinguishable from prose depending on the reading style (actor, speaker, poet). Studies involving actors and performers have further enriched this field. Wagner and Betz (2023) analyzed the strategies used by professional actors versus laypersons when reading poetry and prose. They found that actors used more varied pauses and prosodic boundaries and tended to lengthen verse endings. Likewise, Colonna (2017) demonstrated that poets, radio speakers, and actors approach poetry reading in distinct ways, detecting a tendency to lengthen interpausal units in the voices of actors and radio speakers.

Perceptual studies on pleasantness and well-being in poetry recitation among European and Brazilian Portuguese speakers (Barbosa, 2022, 2023) emphasized the importance of pauses and voice quality, showing notable differences between the two linguistic varieties.

Lastly, large-scale computational and prosodic analyses—such as those by Meyer-Sickendiek et al. (2017), Baumann et al. (2023), and Schauffler et al. (2022)—have significantly advanced the study of poetry reading and the textual–prosodic interface in the German linguistic area.

1.2. Aims, Research Questions, and Hypotheses

Despite early interest shown by Navarro Tomás in the early 20th century in preserving Spanish poetic speech—as evidenced by the establishment of the Archivo de la Palabra y las Canciones Populares [Archive of the Word and Popular Songs] in 1930, which included some poetry recordings—and in studying its metrical, phonological, and phonetic features (Navarro Tomás, 1973a, 1973b), Spanish poetic speech remains a largely underexplored area in phonetic research. A notable exception is the work of Colonna et al. (2024), as well as contributions that emphasize the importance of preserving and studying the poetic voice in Spanish-speaking literary contexts, such as Mistrorigo (2018) and Meza Valdez (2020). Instrumental phonetic analysis of poetic speech still requires broader development across various linguistic contexts, with particular attention to Romance languages, where focused studies are especially needed.

This study aims to contribute to filling this gap by conducting a phonetic analysis of Spanish poetic speech, with particular focus on the effects of historical period and gender on 20th-century poetry reading, following the methodological framework established for Italian poetic speech by Colonna (2022). The objective is to identify key prosodic features within a corpus of 40 original recordings and to outline a periodization that reflects major prosodic changes over time. In this way, we seek to trace the first phonetic history of Spanish poetry reading.

The research questions (RQs) and hypotheses (Hs) that guide the study are as follows:

RQ1: 

At the organizational level, which prosodic patterns in relation to the text are most frequent within the corpus of readings?

H1. 

Following Colonna (2022), we expect an alternation of four types of interpausal units, with a predominance of shorter (“minor”) units. We also hypothesize significant internal variation both within and between authors.

RQ2: 

Can the parameters identified in earlier studies (Colonna, 2022; Colonna et al., 2024)—such as speech rate, the ratio of speech to pause duration (plenus), and pitch span—serve to define a periodization of Spanish poetic reading over time?

H2. 

We hypothesize that these parameters will prove essential in identifying the thresholds of change.

RQ3: 

Is gender a determining factor in a diachronic-based classification?

H3. 

In line with MacArthur et al. (2018) and Colonna and Román Montes de Oca (2024b), gender diversity must also be considered a key factor in such classifications.

RQ4: 

How has speech rate changed over time?

H4. 

As found in Colonna (2022), which revealed a tendency toward an increase in speech rate over time, albeit within a framework of heterogeneous behaviors, we hypothesize a gradual increase.

RQ5: 

How has the plenus evolved?

H5. 

While an increase might be expected (Colonna, 2022), Colonna et al. (2024) observed a decrease in Spanish poetry reading. We hypothesize that our findings will support this latter trend.

RQ6: 

How has pitch span changed over time?

H6. 

We expect pitch span to be generally high, contradicting notions of monotony (MacArthur, 2016) or narrow pitch range (Byers, 1977), and a slight increase over time (Colonna, 2022). We also anticipate significant gender-related differences (see RQ3–H3), in line with MacArthur et al. (2018) and Colonna and Román Montes de Oca (2024b).

RQ7: 

At the intonational level, have any recurring patterns emerged, and can they be interpreted in relation to gender or historical change?

H7. 

We expect to observe a consistent use of declarative and poetic-declarative intonation (Colonna, 2022), possibly with gender-related differences.

RQ8: 

Is there a correspondence between the proposed periodization and either the recording dates or the poets’ birth years?

H8. 

We anticipate only partial correspondence. As suggested by the model adopted here, stylistic change does not necessarily follow a chronological path linked to authors’ lifespans or recording dates. Rather, the evolution of poetic reading unfolds along a diachronic trajectory shaped by individual stylistic choices. In this sense, we aim to outline an “invisible timeline” of stylistic change and to envision future trends in poetic performance.

2. Materials and Methods

2.1. Recordings

For this study, 40 recordings were selected, belonging to 15 Spanish poets from the 20th century, comprising 5 women and 10 men. The minority presence of female voices reflects a tradition that, until recently, has favored male voices—not only in publications but also in the preservation of recorded voices (see Merlo, 2010). However, this work also aims to highlight the female voices of 20th-century Spanish poetry.

Below, we present the authors included in the study, along with the number of recordings per poet indicated in parentheses. The women poets, listed alphabetically, are Clementina Arderiu (3), Ernestina de Champourcín (3), Carmen Conde (3), Ángela Figuera (2), and Gloria Fuertes (2). The men poets, also listed alphabetically, are Rafael Alberti (3), Vicente Aleixandre (3), Dámaso Alonso (3), Luis Cernuda (3), Jaime Gil de Biedma (3), Ángel González (3), Miguel Hernández (1), Juan Ramón Jiménez (3), Luis Rosales (2), and Pedro Salinas (3)1. All recordings are in Spanish, except for Arderiu, who is the only case in Catalan. Her inclusion is justified because she is an integral part of the Spanish poetic tradition, representing both the country’s multilingualism and female voices, and she is well integrated into the contemporary cultural panorama. For the type of analysis conducted here, we do not consider this linguistic difference to be a determining factor, as the prosodic measures examined capture overall reading behavior rather than fine-grained phonological contrasts, allowing for a unified analysis across the corpus.

The selection of these authors is based on their representation of 20th-century Spanish literature, their stylistic diversity, and the availability of recordings. All authors have at least two recordings, except for Juan Ramón Jiménez, for whom only one audio document has been recovered—the oldest in the dataset dating back to 1937.

All the recordings come from historical sources that have ensured their preservation over time. These sources include Voz Viva [Living Voice] of the Universidad Autónoma de México, the Library of Congress in Washington (Archive of Hispanic Literature on Tape, AHLOT), vinyl records (RCA), CDs (Visor), and various digital platforms. The corresponding listening links, where available, or source references are included in the Voices of Spanish Poets vocal archive2—the first dynamic archive dedicated to the phonetic study of Spanish poets’ voices. This platform features entries for all these recordings (linking to the original sources or mentioning the complete source for copyright reasons), along with detailed phonetic descriptions of each one. The complete list of poems considered, along with the source of each corresponding recording, is provided in Appendix A.

The regional origins of the poets span seven of Spain’s seventeen autonomous communities, offering a representative diatopic variety at the prosodic level and contributing to the general overview this study aims to provide. More precisely, the regions represented are indicated in Figure 1, alongside the respective authors and number of recordings: Andalucía (3), Comunidad de Madrid (5), Catalonia (2), Basque Country (2), Region of Murcia (1), Asturias (1), and Valencian Community (1). However, regional variation will not be specifically analyzed in this study.

The dating of the recordings is heterogeneous. A historical reconstruction of the materials (see also Colonna et al., 2024) revealed that some dates are exact (e.g., recordings from the AHLOT archive, such as Dámaso Alonso, 1953; Juan Ramón Jiménez, 1949; Pedro Salinas, 1950), while most are approximate. In contrast, the publication dates of vinyl records are certain (e.g., 1971, Diez poetisas españolas dicen su poesía amorosa [Ten Spanish women poets read their love poetry], published by RCA and including Clementina Arderiu, Carmen Conde, and Ernestina de Champourcín).

To provide a clearer temporal framework, recordings have been grouped into approximate decades based on their likely recording period; for instance, those estimated to have been made between the 1960s and 1970s are assigned to the 1960s. The overall time span of the recordings—calculated by the average per author—ranges from the 1930s to the 1990s, with the highest concentration in the 1960s (see Figure 2). The earliest recording is that of Hernández, followed by Jiménez, while the most recent ones include Fuertes, González, and Rosales, with the majority clustering in the intermediate period. In terms of the poets’ birth years, they range from 1881 to 1929 (mean [M] = 1904, standard deviation [SD] = 12.9 years).

Finally, the 40 recordings correspond to 40 poems, of which 27 are in free verse and 13 follow a metrical pattern. The latter include all the readings by Aleixandre, Cernuda, and Hernández, as well as some by Arderiu, de Champourcín, Gil de Biedma, Rosales, and Jiménez.

2.2. Annotation and Data Extraction

The recordings were manually annotated using PRAAT software (Boersma & Weenink, 1992–2025), aligning the textual axis (TA) with the prosodic axis (PA). The TA includes the Verse Tier—VS (Level 4), while the PA comprises three tiers, organized bottom-up (Levels 3–1): poetic utterance (EP—enunciato poetico, the largest utterance unit), prosodic curve (CP—curva prosodica, or interpausal unit), and rhythmic word (PR—parola ritmica, or stress unit) (see Colonna, 2022).

Additionally, after a tiers check (Colonna & Román Montes de Oca, 2024d), six further prosodic tiers were added via a script (Colonna & Román Montes de Oca, 2024a) and annotated manually. These tiers (Levels 5–10), arranged top-down, include intensity marking (INT: through three intensity peaks per CP), number of syllables per CP (Nsil), marking of fragmented style when present (INTERR: especially identified by unusually long silences within CPs), presence of repeated intonations (SIN: categorized based on rhetorical figures also found in Baroque music—Synonymia for similar tones and Palilogy for identical tones), declarative intonation (DECLA: marking the toneme, encompassing both traditionally declarative and poetic–declarative intonations, characterized by a partially falling boundary, ending with final mid-tone), and possible focalization (FOC: on vowel nuclei). Pauses were classified into categories ranging from <pb> (<0.3 s) to <pll> (>1 s). These indices and the model are defined in Colonna (2022). An example of the annotation system is provided in Figure 3.

Following pitch verification and correction where necessary, the annotation system enabled the semi-automatic extraction of 20 key parameters for poetic analysis using the VIP-VSP-Radar Data Extraction for Poem Phonetic Analysis PRAAT Script (Colonna & Román Montes de Oca, 2024c). The automatically extracted data were manually verified, and a selection of parameters was subsequently considered for analysis.

2.3. Data Description

For this study, following a perceptual overview of the recordings, we focused on a selected subset of 10 parameters related to prosodic organization in relation to the text (PO), as well as prosodic features, including stylistic markers (PRO).

At the PO level, four indices were used to identify and compare CP types in relation to the poetic line (Colonna, 2022): CPs matching the length of the line (line-curve), CPs covering only a portion of a line (hemi-line curve), CPs spanning two lines (inter-line curve), and CPs covering two or more full lines (bi/poly-line curve).At the PRO level, we primarily considered (1) mean speech rate (calculated as the number of syllables per second, excluding pauses); (2) plenus, the ratio between total speech duration and total pause duration; and (3) pitch span (in semitones—ST, calculated as the range between minimum and maximum pitch for each CP). We also compared average CP speech rate and CP duration to explore possible correlations and both declarative and poetic declarative intonation patterns (/Da//).

We described both PO and PRO levels through descriptive statistics and visualizations generated with R-Studio (R Core Team, 2023) and JAMOVI (Jamovi Project, 2024).

At the PO level, we also carried out an exploratory analysis by applying a hierarchical clustering approach implemented in R-Studio in order to investigate possible groupings of recordings based on prosodic organization strategies. Distances between observations were measured using the Euclidean norm, and clustering was performed using Ward’s method to minimize within-cluster variance3.

After the general description, we adopted a comparative approach. Following a perceptual preliminary division, we used the PRO indices of mean speech rate, plenus, and pitch span to divide the corpus into two main groups corresponding to two historical phases. Building on the terminology and findings of Colonna (2022) and Colonna et al. (2024), who worked on radio and television poetic sources, we identified two “Radio-Television (RTV) eras”—an early phase (1rtv) and a later phase (2rtv)—to trace the evolution of poetry reading over time. The choice of this term and the associated periodization is grounded in the effectiveness of the division proposed in those works, where the term was coined primarily according to the type of medium and its cultural significance, thus contributing to the first phonetic history of poetry reading.

More precisely, the distinction between the first and second RTV eras throughout the 20th century, with their differing modes of reading, lies in the different technological, social, and cultural contexts of radio and television, which transformed the very modes of communication and reception in general, as well as the circulation of poetry.

The first RTV era was marked by a more traditional stylistic approach, generally characterized by declamation, formal delivery, and a strong influence of public oratory. This reflected an early stage in the diffusion of radio and television, including their use as media for the recording and transmission of poetry, when these media functioned as centralized, state-regulated platforms with limited distribution channels. By contrast, the second RTV era was generally associated with an increasingly informal style, visible within a shift from declamation to narration and experimentation with new strategies. This phase reflected a later stage, when radio and television had become consolidated mass media with broader accessibility, diversified programming, and new recording and broadcasting conditions, all of which fostered new cultural codes and styles. These shifts in communicative environments created a framework within which different approaches to poetry reading could emerge.

Additionally, we applied a second grouping approach based on gender, following MacArthur et al. (2018) and Colonna and Román Montes de Oca (2024b), to identify potential differences between male (M) and female (F) voices4. Specifically, we examined gender-based groupings in comparison with the RTV classification, analyzing differences in mean speech rate, plenus, CP duration and speech rate, and pitch span. In the latter, as well as in declarative intonation patterns, gender comparisons were also considered independently.

2.4. Statistical Analysis

After describing the dataset, we conducted inferential statistical analyses using JAMOVI software.

We first applied a chi-square test (Χ²) (Greenwood & Nikulin, 1996) to assess whether there was a significant relationship between the categorical variables of historical phases (RTV) and the author’s birth date, as well as between RTV and the recording date. The significance threshold was set at α ≤ 0.05.

The core of our statistical analysis focused on examining the effects of historical phase (RTV: 1rtv, 2rtv) and gender (F, M) on three dependent variables: speech rate, plenus, and pitch span. To this end, we performed three two-way ANOVAs, again using α ≤ 0.05 as the significance level. The Tukey adjustment was applied to correct for multiple comparisons.

Finally, we carried out a linear regression analysis (Fox & Weisberg, 2023; Lenth, 2023) to examine the predictive effect of historical phase (1rtv, 2rtv) and gender (F, M) on the three dependent variables. The goal was to estimate potential trends and forecast the behavior of these variables in a hypothetical third radio-TV phase (3rtv). Residuals were plotted using a Q–Q plot to assess the normality of the model5.

3. Results

3.1. PO and PRO Features: General Overview

The 40 recordings present a heterogeneous prosodic landscape, with considerable variation both within and across authors. From a perceptual standpoint, the recordings range from highly declamatory readings to those exhibiting a more spontaneous delivery, reflecting a stylistic evolution over time. The earliest recording in the dataset—that of Miguel Hernández—exemplifies a markedly distinct style: not only overtly declamatory but also reminiscent of a political rally speech, perhaps reflective of the original performance context and closely tied to his political engagement and experience during the Spanish Civil War. This reading is characterized by extensive use of repeated intonations (synonymias) and declarative patterns (over half of the CPs), combined with minimal tonal and register variation, and a narrow pitch span (8.9 ST), comparable only to that of Jiménez (ranging from 8.5 to 9.9 ST), the lowest values in the corpus. Conversely, the most recent recordings—such as those by Luis Rosales, Ángel González, and Gloria Fuertes—likely produced in professional studios or broadcasted on television programs present a markedly different communicative intent. Designed for a smaller yet heterogeneous audience, they display a more spontaneous and conversational approach to reading. These temporally distant examples delineate an initial perceptual division between an early and a later historical phase, grouping similar reading styles based on overall tone and the polarization of stylistic extremes.

To contextualize these data within a historical framework, we first present a description of the PO indexes to introduce them consistently, followed by an analysis of the PRO parameters to examine their characteristics and define potential groupings.

From a PO perspective, recurrent patterns predominantly favor line-curves and hemi-line curves as prosodic units (M = 0.379, SD = 0.242; M = 0.476, SD = 0.272, respectively). These values remain consistent regardless of whether the poem is metrical or in free verse (line-curves in metrical poems: M = 0.325, SD = 0.219; in free verse: M = 0.404, SD = 0.252; hemi-line curves in metrical poems: M = 0.462, SD = 0.312; in free verse: M = 0.483, SD = 0.256). Overall, a moderate degree of internal variation is observed across readings by the same author, preventing the identification of a uniform prosodic pattern within an individual poet’s recordings. Conversely, inter-line curves and bi-/poly-line curves are relatively rare.

Since the choice of CP types within an author’s recordings can generate distinct prosodic profiles, a hierarchical clustering analysis was applied to compare these readings and identify possible groupings. The resulting dendrogram is presented in Figure 4. Four major clusters were identified. The red cluster, which includes two of the three recordings by Arderiu, Alberti, Conde, and de Champourcín—as well as several single readings—exhibits a high degree of internal similarity, with very low distances (below 0.5). In contrast, the violet cluster shows the highest internal variance, with more scattered subgroups and greater distances, indicating more prosodically diverse readings.

The results confirm that strong stylistic consistency and uniformity are found only in selected cases, where at least two recordings by the same author cluster together (e.g., Conde, Jiménez, Aleixandre). In other cases, intra-author distances are greater, though still relatively low (e.g., Alberti, Gil de Biedma), while in some instances, recordings by the same poet are widely dispersed across different clusters (e.g., Fuertes, Champourcín). These findings suggest that while certain authors exhibit notable internal stylistic cohesion, others display greater variation, with prosodic strategies that align more closely with those of different poets rather than with their own corpus.

Nevertheless, the greatest distances—visible only at the topmost level of the dendrogram—indicate that, despite the differences, shared prosodic strategies enable the formation of coherent groupings. These results also support the decision not to use the PO level as a criterion for dividing the dataset into historical phases. However, they do offer an exploratory mapping of prosodic affinities across the recordings.

Moving on to the PRO indices, we considered the three parameters of mean speech rate, plenus, and pitch span to describe their characteristics and assess their evolution across time, supporting the perceptual framework. The descriptive statistics for these indices are as follows: mean speech rate (M = 4.75 syll./s, SD = 0.608, range = 3.50–6.03), plenus (M = 2.89 s, SD = 1.04, range = 1.37–6.13), and pitch span (M = 8.11 ST, SD = 4.05, max = 26.7). Notably, the highest degree of variation within the corpus is observed in the plenus index.

Regarding mean speech rate, some authors display strong internal consistency—for example, the high convergence resulting in nearly identical speech rates across different readings by de Champourcín, Gil de Biedma, and Arderiu. In contrast, a closer similarity is observed only between two out of three readings by Aleixandre, Salinas, and Figuera. Greater divergences are found in authors such as Fuertes, González, Alberti, and Cernuda, as illustrated in the histogram in Figure 5. Generally, the observed range reflects the variability between slower and faster readings that characterize an author’s style, although fluctuations may occur. Furthermore, varying degrees of intra-author variation are evident: the highest values are reached by Conde (Conde2), but other notable cases of high variability include Fuertes (Fuertes3), Alonso (Alonso1), González (Gonzalez1), and De Biedma (DeBiedma2). Conversely, the lowest levels of intra-author variation are found in the readings by de Champourcín (Champourcin3) and Salinas (Salinas1).

On the other hand, the use of plenus appears to be particularly divergent among poets and even within individual interpretations in this corpus. At the extremes of the histogram in Figure 6, we find the female voices of Fuertes (at the lowest recorded level) and de Champourcín (at the highest recorded level), both presenting two out of three recordings with these extreme values, highlighting a highly marked range contrast. Internal variation within an author can be observed in Fuertes, whose third reading deviates considerably. Conversely, cases of greater convergence within a single author (such as Alberti, de Champourcín, Gil de Biedma, and González) stand in contrast. However, the contrast between distinctly low and high values, despite the significant overall variation, mainly distinguishes between the styles of author groups.

Variation within the corpus is also evident for pitch span (M = 8.11; SD = 4.05), as shown in the histogram in Figure 7, where female and male readings have been marked in different colors: indeed, it is precisely within these groupings that the strongest affinities have been identified. Notably, the highest levels of the mean pitch span are found mainly among the female voices, particularly with the recordings by Fuertes (M = 15; SD = 5.16 in Fuertes3) and Figuera (M = 13.3; SD = 6.52, in Figuera2): these voices also reach the greatest levels of inner variation (M = 12.6; SD = 7.68 in Fuertes1), followed by Figuera2 and Arderiu1. This relevant variation and high mean pitch span is also associated with the consistent use of creaky voice, especially in mature voices, particularly involved in female recordings, combined with higher frequencies, ranging different tone levels within the same CP.

3.2. History of Spanish Poetry Reading: Radio-Television Eras and Gender

Building on the above-mentioned PRO features, we partitioned the corpus into two groups (historical phases, 1rtv and 2rtv) based on the convergence of three thresholds marking differences in the mean values of speech rate, plenus, and pitch span. The 1rtv group included the following seven authors: Rafael Alberti, Vicente Aleixandre, Dámaso Alonso, Clementina Arderiu, Ernestina de Champourcín, Miguel Hernández, and Juan Ramón Jiménez. The 2rtv group included the following eight authors: Luis Cernuda, Carmen Conde, Ángela Figuera, Gloria Fuertes, Jaime Gil de Biedma, Ángel González, Luis Rosales, and Pedro Salinas. Further comparisons involve gender.

Speech rate in 1rtv presents a median (Md) of 4.46 syll./s (range = 3.50–5.08), while in 2rtv it reaches 5.13 syll./s (range = 3.85–6.03), indicating a higher overall speech rate and a narrower distribution (see Figure 8a). Conversely, plenus shows the opposite trend: in 1rtv, the median is 3.37 s (range = 2.08–6.13), whereas in 2rtv, it is 2.29 s (range = 1.37–3.94), suggesting higher plenus in the 1rtv group (see Figure 8b), with greater standard deviation in 2rtv. These data reveal a pattern: an initial phase (1rtv) characterized by slower speech rate and higher plenus, contrasted with a second phase (2rtv) where the two parameters are reversed.

A gender-based comparison of the RTV phases for both indices reveals a greater divergence in speech rate within 1rtv: male voices exhibit higher values (Md = 4.47) compared to female voices (Md = 3.85). In contrast, this gender difference is less marked in 2rtv, with both groups showing generally higher values (Md = 5.46 for women vs. Md = 5.00 for men) (see Figure 9a). As for plenus, female voices show higher median values in both phases—1rtv (Md = 4.25 for women vs. 2.81 for men) and 2rtv (Md = 2.79 for women vs. 2.20 for men)—with the highest values observed in 1rtv (see Figure 9b). Greater variability is also visible in gender comparisons.

Analyzing pitch span features across the historical phases (Figure 10a), we observe a distributional shift between 1rtv and 2rtv (Md = 6.97 vs. Md = 8.15), though the difference is moderate. However, when introducing the gender variable (Figure 10b), the divergence becomes more evident, confirming the trends previously observed (see Figure 7): 1rtv shows Md = 8.66 for women vs. 6.44 for men; 2rtv shows Md = 9.29 for women vs. 7.33 for men. This comparison clearly shows that across both historical phases, female voices consistently exhibit higher median pitch span values and a consistent presence of high outliers (max = 26.4 in 1rtv; max = 26.7 in 2rtv).

To further examine the relationship between speech rate and duration, we compared CP speech rate and CP duration across all recordings, totaling 1063 CPs (M = 1.60 s, SD = 0.85, range = 0.2–5.63 s), grouped by historical phase and gender (Figure 11). Distinct trends emerge between the indices, supporting the hypothesis of Colonna et al. (2024). The highest density of CP duration (Figure 11a) is found at shorter values across both phases and genders, with one notable exception: female voices in 1rtv show a higher median duration (Md = 2.08) compared to male voices (Md = 1.36). This confirms that the CPs tend to be brief, aligning with previous PO findings regarding the prevalence of hemi-line curves and line-curve patterns. Nevertheless, notable outliers are present (1rtv: max = 5.24 by de Champourcín, 5.38 by Hernández; 2rtv: max = 5.63 by Figuera, 4.84 by Rosales).

Conversely, mean CP speech rate (Figure 11b) is generally lower in 1rtv (Md = 3.82 for female voices, Md = 4.66 for male voices) and higher in 2rtv (Md = 5.11 for female voices, Md = 5.10 for male voices). These results align with the general speech rate trends (Figure 9) while also revealing internal variation and different distributions. Again, single CPs show outliers (1rtv: max = 7.15 syll./s by Aleixandre, poem 2; 2rtv: max = 9.36 syll./s by Conde, poem 3, in correspondence with “Yo que no soy de nadie”; and max = 7.67 syll./s by Salinas).

Finally, the presence of declarative and poetic declarative intonations (/Da//) also proves relevant overall, although differences across phases and gender are less marked. The median value for 1rtv voices is slightly lower than that for 2rtv voices (Md = 0.464 vs. 0.512; see Figure 12a), while the gender-based comparison (Figure 12b) reveals differences in distribution and median values (Md = 0.520, SD = 0.168, range = 0.200–0.765 for women; Md = 0.484, SD = 0.138, range = 0.136–0.688 for men).

3.3. Periodization and Dates: A Stylistic–Chronological Framework

In light of the periodization proposed in Section 3.2, we statistically examined the relationship between the mean values of year of birth and the (approximate) recording year of the materials included in the two historical phases (RTV: 1rtv and 2rtv). The results of the chi-squared test show that the association between the historical phase (1rtv and 2rtv) and birth or recording date is not statistically significant (p = 0.2771 for recording date; p = 0.4135 for birth date). Therefore, periodization is not strictly determined by either the date of birth or the recording date.

Nonetheless, divergent trends emerge in the recording dates across the two phases, despite some overlap. In 1rtv, the median recording date falls in the 1950s, with a range from the 1930s to the 1960s. In contrast, the 2rtv recordings have a median date of 1975, spanning from the 1950s to the 1990s. Thus, while some overlap exists, recordings classified as 2rtv predominantly belong to the period from the 1970s to the 1990s, whereas 1rtv recordings are mostly situated between the 1930s and 1960s.

Exceptions within 2rtv that feature earlier recording dates—such as those by Salinas, Cernuda, and Conde—can be attributed to individual trajectories: the experience of exile in the first two cases and an international background in the third (see also Colonna et al., 2024). These elements suggest that personal stylistic factors may have played a role. A comparative analysis of recordings made before and after exile could offer valuable insights into the influence of displacement on reading style.

As for birth dates, the median year in 1rtv is 1898 (SD = 9.81; range = 1881–1910), whereas in 2rtv it is 1909 (SD = 12.7; range = 1891–1929), marking an approximate ten-year generational difference. However, 2rtv exhibits greater internal variability, and there are generational overlaps across both groups, which reinforces the idea that a purely chronological criterion is insufficient to explain the evolution of reading styles.

In conclusion, this chronological assessment supports the theory of a stylistic–chronological framework that distinguishes the two historical phases. While temporal separation is observable, it is not a wholly reliable criterion due to the overlap of readings from different stylistic periods within the same timeframe. Instead, stylistic and individual factors appear to play a decisive role in shaping the proposed groupings.

3.4. Effects of the Historical Phase and Gender on Poetry Reading

In light of the proposed groupings, we measured the effects of historical phase (RTV: 1rtv, 2rtv) and gender (M; F) on the indices of mean speech rate, plenus, and pitch span by applying a two-way ANOVA for each index.

The effect of historical phase was found to be significant in speech rate (p < 0.001; F = 53.202), as was the interaction between RTV and gender (p < 0.001; F = 17.696). Post hoc analysis confirmed that the comparison between historical phases 1rtv and 2rtv was statistically significant (p < 0.001). All pairwise comparisons involving historical phase and gender also yielded statistically significant results: female voices in 1rtv (M = 3.82; EE = 0.177) and 2rtv (M = 5.39; EE = 0.140), male voices in 1rtv (M = 4.52; EE = 0.110) and 2rtv (M = 4.94; EE = 0.106), as well as female and male voices in the same or different phases, except for female and male voices within 2rtv (see Figure 13).

These results indicate that speech rate is statistically different across historical phases and when comparing gender with them, while also showing that gender-based differences are reduced within 2rtv.

For plenus, the effects of both historical phase and gender were statistically significant (p < 0.001), and their interaction was also significant (p = 0.029). Post hoc analysis confirmed these findings (p < 0.001). Significant contrasts (p = 0.002) were observed between female (M = 4.63; EE = 0.351) and male voices (M = 2.98; EE = 0.218) in 1rtv, as well as between female voices in 1rtv and both female (M = 2.72; EE = 0.278) and male (M = 2.72; EE = 0.278) voices in 2rtv (p < 0.001), as visible in the estimated marginal means graph (Figure 14).

These results show that plenus significantly differs across phases and genders: when both factors are considered jointly, gender differences disappear only within 2rtv.

As for pitch span, a statistically significant effect was found for historical phase (p < 0.001) and gender (p = 0.005), but not for their interaction. Post hoc analysis confirmed significant differences (p < 0.001) between 1rtv (M = 8.32; EE = 0.208) and 2rtv (M = 9.0; EE = 0.176), as well as between female (M = 10.06; EE = 0.231) and male voices (M = 7.35; EE = 0.143), as shown in Figure 15. Additional post hoc comparisons of the interaction between historical phase and gender revealed significant contrasts (p < 0.001) between male and female voices in 1rtv, male voices of 1rtv and female voices of 2rtv, female voices of 1rtv and male voices of 2rtv and male and female voices in 2rtv.

Finally, we applied linear regression models to examine the predictive power of historical phase and gender on speech rate, plenus, and pitch span in view of a potential extrapolation to a hypothetical 3rtv phase.

For speech rate, the model yielded an R² of 0.416. The coefficient for historical phase (2rtv—1rtv) was 0.7799 syll./s (p < 0.001), indicating a significant increase in speech rate from 1rtv to 2rtv. The coefficient for gender (male—female voices) was 0.0321 syll./s, suggesting a slight increase for male voices, though this was not statistically significant at the 0.05 level.

For plenus, the model showed an R² of 0.338. The historical phase coefficient was −0.993 s (p = 0.002), indicating a significant reduction in plenus from 1rtv to 2rtv. The coefficient for gender was −1.010 s (p = 0.003), showing a statistically significant reduction in plenus for male voices compared to female voices.

For pitch span, the R² was 0.115. The historical phase coefficient was 0.725 ST (p = 0.003), indicating a significant increase in pitch span from 1rtv to 2rtv. The coefficient for gender was 2.727 ST (p < 0.001), confirming that female voices display a significantly wider pitch span than male voices.

In all three models, Q–Q plots of the residuals confirmed that the normality assumption was met.

On this basis, a hypothetical 3rtv phase would likely exhibit a higher speech rate, lower plenus, and wider pitch span—especially among female voices. The predicted decrease in plenus for male voices suggests a growing use of silence, while the broader pitch span among female voices points toward increased pitch modulation.

4. Discussion

This study has outlined the main phonetic features and historical evolution of 20th-century Spanish poetry reading. A framework for a phonetic history of poetry reading was established through a selection of prosodic parameters and factors such as historical phase and gender. The analysis highlighted significant changes over time, confirming perceptual and previously formulated hypotheses (Colonna, 2022). Additional aspects at the PO and PRO levels—such as organizational and intonation patterns—also revealed generally shared strategies, marked by internal variation.

The aim was to trace stylistic changes in poetic speech over time. As stated in H8, we hypothesized that these changes were not strictly chronological, i.e., not directly linked to birth dates or recording periods. Indeed, neither variable was statistically significant for the periodization into the historical phases of the first (1rtv) and second (2rtv) radio-television, based on Colonna’s (2022) model. However, a difference of about 10 years in median birth dates separates the two phases, though overlaps were observed (e.g., Salinas, Cernuda, Conde, and Figuera in 2rtv, despite birth dates compatible with 1rtv). A more marked separation was observed in the median recording dates, differing by approximately 25 years between phases. Yet again, some authors exemplify transitional traits in readings from the 1950s, suggesting a shift already underway. These cases illustrate how stylistic and individual factors influence the historical timeframe, especially where both thresholds (birth and recording date) support the hypothesis of an innovative reading style. Thus, although chronologically closer to 1rtv, their readings align stylistically with 2rtv (e.g., Salinas, Cernuda, and Conde)6.

In this regard, it is important to take into account the technological context in which a poet was born, as this may help explain variations in how they approached the microphone—depending, for instance, on whether it was already part of their daily experience. Microphones only began to be used more widely in commercial recording studios in the mid-1920s, coinciding in Spain with the emergence of radio broadcasting, officially regulated by the Real Orden [Royal Order] of 14 June 1924 during Primo de Rivera’s dictatorship. Although experimental transmissions had taken place earlier, it was only from 1924 onwards that the first private stations—such as Radio Barcelona, Radio España, and Radio Ibérica in Madrid—were established (López-Escobar & Faus-Belau, 1985). The role of the microphone evolved in parallel with the development of the media. During the Spanish Civil War and the early Franco regime, radio became a central tool of propaganda, which led to the widespread diffusion of microphones. Furthermore, it was only at the height of the Civil War that the first public radio station was created in Spain, whereas private broadcasters had previously held a monopoly (Bagant & Arboledas, 2011). Later, with the advent of television in 1956 and the growth of the recording industry, microphones became standard instruments in musical studios and dubbing, and by the 1970s, they also entered domestic contexts. This produced distinct media soundscapes for authors born before and after these technological milestones.

Most of the poets considered in this corpus were born before the advent of Spanish radio broadcasting, which meant that they encountered microphones and mediated sound only later—in early childhood, adolescence, or even adulthood—thus experiencing them as both a novelty and, in some sense, a new challenge. By contrast, poets such as Ángel González and Jaime Gil de Biedma grew up within the new media environment, where radio and microphone transmission were beginning to be, or had already become, part of everyday life. Belonging to what we classify as the 2rtv generation, we can suppose that their ears were exposed to this new stimulus from the outset, making the microphone a familiar and formative presence rather than a disruptive innovation7. This, in turn, may imply a different mode of reception and a consequent reconfiguration of their acoustic imagination—an interpretation that might also find possible empirical support in their plenus values, which, uniquely among all the readings, remain consistently low across the recordings, a result that could perhaps be linked, at least in part, to a distinct perception of silence and voice.

The proposed periodization also reflects the broader historical context: recordings from the 1rtv cluster around the Spanish Civil War and the early Franco regime, as do several 2rtv recordings. Compared to the Italian case (Colonna, 2022), where a stylistic shift occurred in the mid-1970s, the Spanish transition appears more gradual and begins earlier—already in the 1950s with authors like Salinas, Cernuda, and Conde, as mentioned before—but consolidates in the 1970s, coinciding with the decline of Francoism (culminating in 1975, the median year of 2rtv). This new reading style is more spontaneous and less declamatory, perceptible, even without detailed analysis, and supported by the data. It progressively supplants the earlier style. Notably, the Spanish corpus spans a broader time frame than the Italian one (extending about 30 years earlier) and shows a wider transition phase. The nature of the sources also differs: while radio-television dominates the Italian corpus, it accounts for a smaller share in the Spanish corpus, where archival and discographic recordings for preservation or commercial purposes are more common. Moreover, the Spanish corpus reflects a higher degree of internationalization, due in part to the exile or international activity of many poets. We believe that migration and cultural exchange may have influenced the reading style of several authors—a phenomenon also observed in the Italian context (e.g., Giuseppe Ungaretti and Amelia Rosselli), where international experiences contributed to a shift in poetic performance conventions (Colonna, 2022). Indeed, many early recordings were made abroad: Hernández in Paris (1937, during the Spanish Civil War); Jiménez (1949), Salinas (1950), and Alonso (1953) at the Library of Congress in Washington for the AHLOT, and Alberti and Cernuda at the radio of the Universidad Nacional Autónoma de México (UNAM) in the 1960s. These contextual, cultural, and personal elements likely played a role in shaping the evolution of poetic reading, alongside individual stylistic choices.

In light of the comparison between the Spanish corpus and the Italian reference study—which nonetheless provided the terminology adopted here, particularly the distinction between the first (1rtv) and second (2rtv) radio-television phases—a question arises: does the concept of radio-television remain appropriate even when the sources differ from those originally associated with radio and television? The influence of media as a force shaping speech and communication throughout the 20th century—and constantly evolving—cannot be overlooked when discussing poetic speech mediated by recording intended for a broad audience. This holds true not only for radio-television sources, but also for formats such as vinyl records or poetry sound archives. Within the field of sound studies, Dotto (2019) has shown how the introduction of the phonograph in the early 20th century transformed the semiographic orders of reality. The evolution of this and subsequent media has continued to shape both the structures of the world and their communicative transmission, including artistic expression. Considering this, we believe that this terminology continues to be relevant. It accounts for stylistic evolution within a broader classification framework informed by media development. Furthermore, just as Umberto Eco (1983, 1990) identified a semiotic shift from paleo- to neotelevision in the early 1980s, the phonetic histories of Italian and Spanish poetry readings both show transformations that began earlier and extended over several decades—transformations that can, in part, be compared to shifts in media evolution. Nonetheless, we remain open to introducing additional classification labels in the future, such as “Declamatory Phase” and “Spontaneous Phase” within the phonetic history of poetry reading, to address discrepancies in source typologies and better reflect stylistic criteria.

The proposed periodization was based on speech rate, plenus, and pitch span indices (H2), previously described through an overview of each reading, highlighting the variation across readings and authors, and subsequently analyzed using descriptive and inferential statistics. This confirmed the earlier perceptual classification. Statistical analysis revealed that all these indices were decisive for distinguishing historical phases. Furthermore, the interaction of the historical phases with gender appeared significant for speech rate and plenus. On the contrary, this was not the case with pitch span. In general, a gradual increase in speech rate validated our initial hypothesis (H4), consistent with trends observed in the 2rtv readings of Italian poets (Colonna, 2022). As for plenus, H5 was also confirmed: We observed a decline over time, indicating a divergence from Italian patterns, yet aligning with preliminary findings by Colonna et al. (2024). Specifically, in Spanish 2rtv readings, pauses tend to be longer relative to speech duration. With respect to pitch span, values generally increased in the 2rtv readings, confirming H6 and observations in Colonna (2022). Overall, the data revealed a broad tonal variation in poetic reading, contradicting claims of uniformly low pitch (Byers, 1977; Barney, 1998) or monotony (MacArthur, 2016). Instead, tonal diversity emerges, as previously observed (MacArthur et al., 2018; Colonna, 2022), and relevant inner variation in readings. Gender comparison also revealed interesting information, showing significant differences in plenus and pitch span, where higher values were reached by female voices. In the latter case, we observed higher variation, reaching the highest pitch span levels, in female voices: It aligns with the theories of MacArthur et al. (2018, p. 33) on creaky voice in women’s readings and of Colonna and Román Montes de Oca (2024b). These results agree with H3, showing significant differences in relation to gender and confirming that a gender-based comparison is essential when establishing categorizations in this type of speech.

Commenting on gender differences from a broader perspective can be highly informative. In this regard, possible reasons for the use of particular pitch strategies by women poets may lie in their intention to achieve greater expressivity, vocal warmth, and firmness. Reflections on the historical framework in which these features emerged can be particularly valuable. The markedly higher speech rate observed in women of the 2rtv, who substantially surpass male voices from a previous lower level, combined with a significant increase in the duration and presence of pauses compared to the 1rtv—reaching levels similar to men—may reflect a growing assertion of social space by female voices over time. This change combines accelerated speech with longer pauses, potentially as a strategy to convey authority. In this way, their discourse—balancing temporal and spatial dynamics through these strategies—can be interpreted as both a reflection and enactment of their contemporary historical moment, oriented toward increased pace and a departure from traditional male patterns.

The new use of silence, serving as a form of reverberation for both the word and attentive listening, and the emphasis on speed highlight the evolving role of women’s voices. This process of asserting authority likely has its roots in the final years of Francoism and the subsequent transition period, which marked a shift away from strongly patriarchal laws and norms and coincided with the rise of feminist movements in Spain. These phonetic changes should not be considered incidental, but rather as outcomes of this broader socio-historical framework, reflecting concurrent social and cultural transformations. Such changes are mirrored in the emancipatory and rights-focused activities of poets such as Carmen Conde and Gloria Fuertes and are also evident in the work of Ángela Figuera.

These findings partially resonate with studies on media speech in television journalism. Despite their differences, poetic and media speech share goals of longevity and public dissemination and often adopt similar strategies. Research on media speech confirms a trend toward faster speech rates in radio and television across Italian (Pettorino & Giannini, 1994; Giannini & Pettorino, 1999; Giannini, 2004), Hispanic (Rodero Antón, 2012), and Slavic contexts (Tivadar, 2017; Robin & Herring, 2019). However, regarding pauses, studies on radio silence by Balsebre (1994) and Rodero Antón (2003, 2012) indicate a general reduction over time. Specifically, Rodero Antón (2012) found a decrease in pause frequency and an increase in speech rate in both Italian and Spanish radio news—trends also reported for Italian poetic speech by Colonna (2022). In comparison, our findings of Spanish poetry readings suggest the same feature for speech rate and opposite features for the relationship between speech and pauses. This highlights poetic speech as a phenomenon that both aligns with and diverges from contemporary media speech, developing its own distinctive strategies.

Focusing on further prosodic aspects of this corpus, our analysis at the PO level confirmed H1: dominant organizational strategies include line-curves and hemi-line curves, while inter-line and bi/poly-line curves are less frequent. However, clustering analysis showed that within these dominant patterns, each author’s readings vary in their affinity with others—and even with their own. In several cases, a poet employed different strategies across separate readings. This suggests that at the PO level, reading styles connect authors across different periods and genders, supporting RQ1. The analysis of average CP duration revealed broad uniformity among authors, with expected variation across phases and gender, without a direct correlation with CP speech rate.

Concerning PRO indices, particularly intonational patterns like declarative and poetic declarative intonation, further interesting results emerged. The /Da// index relatively changed across historical phases and gender, partially confirming H7. The use of falling contours was previously noted by Byers (1977, 1979) and Barney (1998, 1999) in the Anglophone context, while the medium-descending pattern was identified by Colonna (2022) in Italophone readings. Our data confirm the widespread use of both declarative and poetic-declarative intonations as a defining feature of this type of speech, with gender-based differences supporting H3 and H7. A more detailed study of the intonational inventory in this corpus would be a valuable next step.

Within this general framework, provided by our “zoom-out” or distant-listening perspective (Clement, 2013), it is also useful to take a slightly more critical approach to the selected poets and their placement within the periodization proposed here. The corpus includes poets of the Generación del 36 [Generation of 1936] (Miguel Hernández); postwar poets of poesía social [social poetry] (Ángela Figuera, Gloria Fuertes); poets of poesía arraigada [rooted poetry] (Luis Rosales); predominant poets of the Generación del 27 [Generation of 1927] and related movements (Clementina Arderiu, Ernestina de Champourcín, Carmen Conde, Rafael Alberti, Vicente Aleixandre, Dámaso Alonso, Luis Cernuda, and Pedro Salinas); and poets from the previous generation (Juan Ramón Jiménez) and the Generación del 50 [Generation of 1950] (Jaime Gil de Biedma, Ángel González). These movements should not be considered as an exhaustive classification or as independent from one another, but rather as interacting elements and active participants in the broader Spanish cultural landscape.

As can be observed, belonging to the same movement does not necessarily imply the same distribution across the two RTV eras, as already shown by Colonna et al. (2024). Instead, similarities can be found between authors from different generations and movements, although in some cases the distribution of poetry readings does align with literary-historical classification. Furthermore, of these readings, 13 were of texts with a metrical scheme, while the others were in free verse, distributed with relative consistency among authors, except in some cases where readings of metrical and free-verse texts alternate for the same author (e.g., Jiménez, Rosales, de Champourcín). This also reflects the internal variation and coherence that we identified and that could be further studied.

Examples of stylistic divergence within this corpus, representing readings at opposite ends of the two radio-television eras, can be found in Hernández and Fuertes. Both engaged in forms of activism—Hernández in the context of the Civil War, and Fuertes in the struggle for gender equality and advocacy for civilian victims—demonstrating socially and politically engaged poetry of their respective times. In this selection, Hernández addresses the theme of a soldier husband through a declamatory and political–oratory prosodic approach, whereas Fuertes explores autobiographical themes and children’s writing through a more informal, emphatic style. This shows how, starting from texts with very different formal and thematic structures (metrical in Hernández’s case), and set in different literary, cultural, and technological contexts, poets developed markedly different yet equally effective reading strategies.

In this study, intra-authorial analysis was limited to the selected corpus and chosen parameters, showing possible variation and coherence within an author. A broader study, considering variation in a poet’s reading over time, would provide valuable insight and might help refine the historical periodization while better distinguishing individual style from broader cultural trends. In this regard, it would be highly desirable for such studies to be conducted through targeted archival work capable of “capturing” a poet’s reading at different stages of their life.

Looking ahead, the hypothesis of a potential third radio-television represents the first step for future research following the phonetic history of Spanish poetic reading. Linear regression analysis suggested that a hypothetical 3rtv might involve faster speech rate, lower plenus, and even greater pitch span, together with gender differences, like higher speech rate, lower plenus, and lower pitch span in male voices in comparison with female voices. Access to contemporary 21st-century recordings—such as the Voices of Spanish Poets corpus (Colonna, 2024)—would allow for a future verification and deeper comparative and gender-balanced study. In this respect, regarding the consideration of a hypothetical third RTV era, it is important to note that, as in previous cases, it would result from contemporary technological changes. Based on digital recordings, this phase would take into account the rise of the internet and the spread of shared speech on social media. Indeed, the proliferation of internet platforms can exert a significant influence on stylistic change, as it allows for broader dissemination and accessibility, as well as immediacy and ease of sharing.

We view these findings as a first step toward identifying a broader trend that supports our hypothesis of a transformation in Spanish poetry reading over time. However, this change must be understood in its internal complexity, acknowledging variation between individual readings and the heterogeneity of the corpus in terms of factors such as gender and other possible groupings (e.g., language, poem style). This framework provides an initial orientation for interpreting Spanish poetic speech as a phenomenon that both reflects and contributes to literary and cultural history.

5. Conclusions

This study has presented a phonetic framework for the reading of Spanish poetry over the past century, analyzing how poetic reading has evolved over time. We have described phonetic parameters that both unify and differentiate our corpus of recordings, highlighting variation and shared features among interpretations and authors.

Through the identification of key indices, we proposed a periodization into historical phases that outlines the first phonetic history of poetry reading in Spain, while also accounting for gender-based differences. Based on the analysis of speech rate, plenus, and pitch span indices, we identified two historical phases—termed the first and second radio-television periods—borrowing the terminology from Colonna (2022). This periodization follows what we define as a “stylistic-chronological” timeframe, acknowledging that the distinction between the two phases only partially aligns with the chronological dating of the recordings. Instead, a transitional chronological overlap justifies a stylistic classification that traces a shift from a declamatory to a more spontaneous approach to reading.

As expected, our analyses confirm that speech rate and pitch span have significantly increased over time, while plenus has significantly decreased. Additionally, gender comparison emerged as relevant—both independently and in combination with historical classification—across all three indices. In particular, gender differences were especially pronounced in plenus and pitch span, with female voices exhibiting significantly higher values and greater variation (for pitch span) than male voices. The interaction between gender and historical phases was also significant for pitch span, revealing distinctive features.

Other prosodic aspects were also examined to describe the defining characteristics of these readings. At the level of prosodic organization, the predominant strategies involve curves that align with portions of a line or the entire line, alongside individual variability and a general predominance of short mean CP durations. Moreover, the analysis of intonational behavior revealed a consistent presence of declarative and poetic-declarative intonation patterns.

Given the significance of the changes observed across historical phases and gender, we hypothesize that future studies on contemporary voices will confirm the trends identified here. We advocate for a more detailed and balanced analysis in subsequent research, particularly in light of the growing availability of sources, which may reveal further variation and new possibilities for grouping. These findings underscore the need for interdisciplinary analytical approaches and diversified, integrated classification criteria in order to fully capture the complexity of this mode of speech. This approach underscores the richness of Spanish poetic reading—still a largely unexplored field—and marks a first step toward a broader historical perspective. It provides a valuable lens through which to understand the literary phenomenon in its performative dimension and to trace the evolving nature of artistic speech within a wider cultural framework through the application of phonetic methodologies.