Schizophrenia is a severe and chronic mental disorder that affects 20 million people across the world (James et al., 2018). There are three types of main symptoms: positive symptoms (such as delusions, hallucinations, or disorganized speech); negative symptoms (affective flattening, alogia, or avolition) and cognitive symptoms (dysfunctions in working memory or processing speed, deficits in reasoning and abstract thinking, among others). Positive symptoms are manifested in acute episodes of the illness, while negative and cognitive symptoms are present during inter-episode phases, and they represent a significant factor in the limitation of the quality of life of patients. People with schizophrenia may present a high degree of social, occupational and academic dysfunction throughout their life (Ventura et al., 2009; Halverson et al., 2019).

The cognitive symptoms of the illness include deficits related to detection, processing and the use of social information, which enables integration and communication with other people. These tasks are part of Social Cognition (SC), and SC deficits are significantly related to deteriorating functionality in schizophrenia (Fett et al., 2011; Pinkham, 2014).

One of the bases of SC, which enables correct social interaction, is emotional processing. Emotional processing alterations do not improve with antipsychotic treatment (Penn et al., 2009), and they are related to the severity of the illness (Irani et al., 2012). Facial emotion processing deficits have been widely documented (Kohler et al., 2010; Gao et al., 2021). In studies related to other sensory channels, such as the auditory pathway, results are similar to those of facial recognition, although they are collected less frequently (Lin et al., 2018). In the auditory channel, emotional recognition is carried out through tone of voice, also called affective prosody.

Affective prosody is an important issue in the field of schizophrenia because it helps to clarify meanings and resolve ambiguity in human speech through the auditory channel, when no other channel is available (e.g., a phone call) (Paulmann and Pell, 2011). Moreover, prosodic recognition is integrated as part of the multisensory channels needed for emotion communication (composed of facial and linguistic expressions and paralinguistic inputs), and it is essential for interpersonal relations. If any of these channels fail to integrate emotional signals, interpersonal conflicts could arise (Lin et al., 2020).

Deficits in prosodic recognition are associated with negative and cognitive symptoms (Thomas et al., 2017). Basic auditory skills and auditory emotion processing are impaired in schizophrenia patients with cognitive disturbances (Kraus et al., 2019). The presence of negative symptomatology has been associated with worsening prosodic recognition (Leitman et al., 2005; Castagna et al., 2013). One of the hypotheses considered by Castagna et al. (2013) is that both the cognitive state and the negative symptoms that are altered in schizophrenia could bias the evaluation and discrimination of the auditory stimuli from the environment, making them less competitive, and therefore, more deficient than in non-pathological conditions.

Moreover, it has been shown that deficits in prosodic perception have a crucial effect on global functionality (Leitman et al., 2005, 2007). These prosodic deficits also have an impact on social skills, resulting in alterations, misunderstandings, and inappropriate social responses (Kee et al., 2003; Pinkham and Penn, 2006). Primary auditory processing has been directly related to social functioning, independently of cognition (Medalia et al., 2019).

Due to the impact of emotional processing on cognitive and social dysfunction, it is vitally important to study its treatment. Several research avenues have been developed in the treatment of prosodic recognition. Pharmacology, the use of oxytocin and neurostimulation are some of the possibilities under development, although sufficient evidence is still unavailable (Tan et al., 2018). Currently, the most effective treatment is cognitive rehabilitation (Horan and Green, 2019). According to the Cognitive Remediation Experts Workshop (Florence, Italy, April 2010), cognitive rehabilitation is defined as an intervention based on behavioral training, aimed at improving cognitive processes with the aim of achieving a durable and generalized improvement. The results of these objectives in the studies conducted are also varied. The generalized improvements observed after applying an intervention program, with regards to untrained cognitive functions, psychosocial functioning and symptomatology, have been verified in certain studies (McGurk et al., 2007; Kurtz and Richardson, 2012), although reviews point to a lack of attention to this objective (Fiszdon and Reddy, 2012). As for the durability of post-intervention effects, their short-term benefits have been demonstrated (Wykes et al., 2011; Fiszdon and Reddy, 2012; Revell et al., 2015), although research on the maintenance of these improvements in the long term continues to be limited.

Furthermore, among the cognitive rehabilitation programs published, the conditions that must be met to achieve greater effectiveness are not clear. While certain reviews support the use of computerized methods compared to other types (Byrne et al., 2015), others maintain that the type of program does not necessarily entail differences in the results (Morin and Franck, 2017). Likewise, there does not appear to be significant association between a specific intervention based on a particular cognitive function and a global intervention based on several cognitive aspects when designing an intervention program (Grant et al., 2017). Studies agree that the number of tools is scarce and the methodological quality of studies is modest (Wykes et al., 2011; Grant et al., 2017). It is necessary to further examine the design of cognitive rehabilitation tools, by conducting studies with greater methodological rigor, making it possible to compare tools and draw valid conclusions.

At present, social cognition rehabilitation programs are continuously emerging. Emotion recognition is the subdomain that presents a greater number of developed tools for rehabilitation (Fiszdon and Reddy, 2012). One of the main effective SC programs in schizophrenia is called Training Affect Recognition (TAR) (Frommann et al., 2003), which focuses on facial expression recognition of basic emotions. Another program reviewed, called Social Cognition Interaction Training (SCIT) (Penn et al., 2005; Combs et al., 2007), includes various treatment modalities and is based on emotional perception, theory of mind and social perception. The final type of intervention, named Cognitive Enhancement Treatment (CET) (Eack et al., 2007), combines aspects of SC with neurocognition training and includes a prosodic rehabilitation module.

Despite this, literature on prosodic recognition rehabilitation remains scarce. Certain programs have been designed, such as Cognitive Pragmatic Treatment (Bosco et al., 2016), which is administered based on a group therapy model, in 20 half-hour sessions. Training is based on linguistics, paralinguistics and extra-linguistics, theory of mind and other cognitive functions. The results of this study did not provide affective prosodic measurements, and the sample was too small (n = 17). In another type of program created by Bechi et al. (2018), based on auditory and visual training and comprising of 8 h long sessions, the results showed an improvement in auditory abilities with training and an extended benefit in the auditory channel upon completion of the visual training. Finally, SocialVille (Nahum et al., 2014) is another relevant program in prosodic recognition rehabilitation. This program combines 40 working sessions of SC exercises, including auditory perception of basic emotion training, visual affect perception, social cue perception, theory of mind, self-referential style and empathy. The results showed an improvement in SC scales and functioning (Miley et al., 2020).

In this context, the web platform www.e-motionaltraining.com was developed, which includes different rehabilitation games in SC and social skills for schizophrenia and other mental illnesses (Vázquez-Campo et al., 2016; Maroño Souto et al., 2018). A specific intervention was created on this web platform, focused on affective prosody recognition, called Voices. The Voices program consists of eight sessions of varying levels in the form of a game with 15 different trials. In each trial, sentences of speech with neutral lexical content (e.g., “I brought what you asked for”) are reproduced in different emotional valences (e.g., happy, angry, surprised…). These sentences are recorded by professional actors and actresses and uploaded to the program in an mp3 file. For each trial, two, three or four response options with different emotions are shown, with only one being the correct answer. Once the participant has chosen their answer, the program returns feedback regarding the correct/incorrect response. Trial difficulty increases gradually over the course of the sessions. A randomized, multicenter clinical trial was conducted, comparing the intervention with Voices in clinically stable patients with schizophrenia or schizoaffective disorder, compared to patients with the same diagnosis who regularly attended their routine treatment at psychosocial rehabilitation centers. The result of the prosodic recognition scale applied to evaluate the participants pre- and post-intervention was favorable for the Voices program, compared to conventional rehabilitation (p < 0.05). The participants following the Voices treatment filled out satisfaction surveys on the program, in which 80% of users considered the tool to be easy to use and entertaining (Lado-Codesido et al., 2019). Some of the limitations of the study include the scarcity of similar studies, hindering comparison between intervention programs, the minimal number of effective sessions and the lack of evaluation of other cognitive measures to ascertain the generalized effect of rehabilitation.

Following the initial results of the Voices program, the tool was optimized and adapted, creating the Voices 2 program. New sentences were added, recorded by new voluntary actors and actresses, equipping the tool with greater diversity. A reference glossary was also created to understand certain complex emotional terms (called “emotional glossary”), which can be used by users independently. The study methodology was improved, adding an active control group and a more comprehensive post-intervention evaluation, in order to observe the durability of improvements in the short term. The Voices 2 tool, like all the other programs on the platform www.e-motionaltraining.com, was developed in Spanish.

The main goal of this study was to assess improvements in prosodic recognition with a new version of a training tool called Voices 2, through a prosodic recognition scale, in clinically stable patients with schizophrenia or schizoaffective disorder. The second aim was to observe the maintenance of the benefits in a short period after intervention (1 month is proposed). The final objective was to evaluate the usability of Voices 2 tasks from the participant’s point of view.

A randomized, single-blind, multicenter clinical trial was conducted with 44 outpatients diagnosed with schizophrenia or schizoaffective disorder. The patients were recruited from four psychosocial rehabilitation centers in A Coruña, Madrid and Guipúzcoa. All patients attended the centers as outpatients. The selected patients were randomized by a computer-generated randomization list. The allocation sequence was randomly assigned and concealed from the research team. Three of the recruited patients were lost to follow-up, 2 in the control group (one of them refused to continue, the other moved house), and 1 in the intervention group (who refused to continue), also called the Voices group (Control n = 19, Voices n = 22). Therefore, the final sample was composed of 41 patients (Figure 1). The attrition rate of the intervention was 6.7%.

A total of 44 participants were recruited and met the selected criteria. A single blinded, randomized assignation was carried out to select participants for the control group or for the Voices group. Three participants dropped out of the study. Finally, the statistical analysis was performed over a sample of 41 participants. All the descriptive analyses of the main variables of the sample are described in Table 2.

In this randomized clinical study, a prosodic recognition program called Voices 2 has been tested in patients with schizophrenia or schizoaffective disorder. A pre- and post-intervention assessment was conducted with the prosodic recognition scale RMV-SV and, in the case of clinical state, with the PANSS scale in each group. When comparing post-intervention RMV-SV between groups, no significant differences were observed. Despite this fact, the results confirmed significant differences in the post-intervention RMV-SV scale compared to the baseline values in the Voices group. In the control group, whose training entailed an auditory program that was not related to emotions, these results were not observed. The scores on this prosodic recognition scale also confirmed the significant differences observed 1 month after finalizing intervention in the Voices group, which were not observed in the control group. There were no significant differences in the clinical state of patients pre- and post-intervention in any of the subgroups. Users of the Voices 2 program considered the tool as globally attractive and efficient. These results suggest that Voices 2 could be useful as a specific intervention program for the rehabilitation of emotional prosody in patients with schizophrenia or schizoaffective disorder.

The results obtained in the scores of the RMV-SV are similar to the first version of the Voices study (Lado-Codesido et al., 2019). In the current study, it is worth noting that the improvements designed for Voices 2 did not achieve better results compared to the first version of Voices (e.g., the emotional glossary, or the addition of new trials to the game). As explained above, the emotional glossary was applied in two of the four centers and, given that it was used autonomously, its impact was not strictly measured. Although participants found this tool appropriate, further research is needed to clarify its contribution. Likewise, the addition of new trials in Voices 2 did not represent a greater improvement in results, but this variability could help avoid the repetition of phrases, thus reducing the memorization bias. The addition of an active control group in the present study implies increased methodological quality and gives the results greater rigor (Kurtz et al., 2016).

These findings are also in line with similar rehabilitation programs. For example, the SocialVille program, which we consider to be the most comparable with Voices 2, is based on PC games and includes specific auditory training and an active control group (Rose et al., 2015). Although significant results in the SC measures were not obtained in the initial trial (Nahum et al., 2014), when the study was repeated with a larger sample (n = 55), favorable results were observed for SC scales, including a prosodic identification scale (Nahum et al., 2021). At present, programs that include prosodic recognition measures, which would enable us to compare our results with other programs, continue to be extremely limited (Fiszdon and Reddy, 2012; Grant et al., 2017; Tan et al., 2018).

In this study, Voices 2 shows the durability of the effects achieved 1 month after intervention, with results that are consistent with other rehabilitation programs. For example, SocialVille reported positive results in prosodic recognition after 6 months (Miley et al., 2020). In the review carried out by Fiszdon and Reddy (2012), 50 studies on SC treatments were reported. Six of these studies evaluate the short-term maintenance of effects (between 1 day to 6 weeks), with favorable results. Similar results can also be observed in other meta-analyses, with a low number of studies that include this variable (Grynszpan et al., 2011; Wykes et al., 2011). Other studies evaluate maintenance in the longer term (one and 2 years), such as the cognitive rehabilitation program Cognitive Enhancement Therapy (Hogarty et al., 2004; Eack et al., 2011). In a recent meta-analysis carried out by Nijman et al. (2020), it is concluded that improvements are maintained in the follow-up of studies evaluating this variable, but with lower values compared to post-treatment. These conclusions are consistent with the results of Voices 2. Given the significance of the durability of these improvements, as the main objective of cognitive rehabilitation, it is crucial to continue to investigate the effects in the long term, as well as the underlying neoplastic changes brought about by cognitive rehabilitation.

The scores of the satisfaction survey on Voices 2 are also similar to those of the first version of the program, confirming the perception of this tool as accessible and attractive. Other rehabilitation programs also include this type of subjective user evaluation (Nahum et al., 2014; Rus-Calafell et al., 2014; Palumbo et al., 2017). The positive evaluation of the tool, individual motivation, the subjective perception of improvements and feedback given to users during rehabilitation could all positively contribute to treatment response and program adherence (Fiszdon et al., 2020). These points were considered when designing the Voices 2 program.

With regards to program adherence, it is worth noting the low dropout rate of 6.7%, compared to other studies reviewed, with rates of between 23–37% (Fernandez-Gonzalo et al., 2015; Fisher et al., 2017; Grant et al., 2017; Nahum et al., 2021). This could be due to the use of Voices 2 at the regular care centers of participants, who take part in more extensive psychiatric rehabilitation. Furthermore, Voices 2 seeks to represent situations that are similar to the real life of patients. These two factors could be key for treatment adherence (McGurk et al., 2007, 2012; Wykes et al., 2011; Parker et al., 2013). The duration of Voices 2 (approximately 4 weeks, compared to 8–16 weeks for the aforementioned programs), could also be associated with the differences in dropout rates.

Finally, it is worth noting the innovative creation of the emotional glossary as an exploratory tool, with the use of NSM for the first time in a rehabilitation program. In the same way as the work carried out in autism, we find this approach to be very interesting, as a new avenue for the lexical understanding of emotional terms in populations that present deficits in emotional processing. Furthermore, such tools facilitate the autonomous use of the Voices 2 program. Despite the fact that it has not been possible to use the glossary at all centers, users defined it as appropriate (Med = 4). After this first experience with NSM, we feel it would be interesting to expand on and further the study of the emotional glossary for future versions of the program.

Certain limitations have been observed in this study. It is necessary to give cognitive rehabilitation studies sufficient methodological quality to enable comparison between studies. According to the criteria of the Clinical Trial Assessment Measure (CTAM), a larger clinical sample would be necessary. The small effect size is also a limitation. Moreover, repeating the RMV-SV scale on three different occasions could represent a learning bias, although this is not in line with the differences observed between subgroups. The repeated use of the RMV-SV is related to the lack of similar measures in Spanish, as well as its ease of application, without the need for prior training. In addition, the use of actors and actresses for play-acting vocal emotions could be considered an artificial way of expressing emotions (Jürgens et al., 2011), and it does not represent natural conditions. Furthermore, following the objectives of cognitive rehabilitation, the inclusion of measures of other cognitive or psychosocial functioning aspects in this study could make it possible to clarify whether the effects of Voices 2 are generalized, an objective that has not been investigated in this study.

Voices 2, the on-line prosodic recognition rehabilitation program, could be effective in improving prosodic recognition in patients with schizophrenia or schizoaffective disorder, as well as being described as attractive and efficient from the user’s perspective. The durability of the improvements observed is maintained 1 month after intervention with Voices 2. These results support the use of Voices 2 and promote the progress of aspects of psychiatric rehabilitation that still need to be developed. This new tool could provide benefits in terms of the interpersonal communication of patients with schizophrenia, and changes in prosodic recognition in the short term. Further studies could be conducted to continue to examine knowledge of prosodic rehabilitation, improving the quality of studies on emotion recognition intervention and promoting the personalization of rehabilitation treatment for schizophrenia.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://doi.org/10.17605/OSF.IO/G95C4.

The studies involving human participants were reviewed and approved by Clinical Research Ethics Committee of Galicia, Registration code: 2019/530, Euskadi, Registration code: PI2019192. The patients/participants provided their written informed consent to participate in this study.

AG-C and ML-C created the Voices 2 program, selected the participants, applied the intervention, extracted data, and supervised the study. AG-C, ML-C, and LM designed the study. ML-C, ML, RR, LM, JO, and MM participated in patient selection and obtained and extracted data. ML-C, AG-C, and RM wrote and revised the manuscript. All authors have had full access to the study data, have personally reviewed the manuscript and have given their final approval to the version attached.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.