Iconic words are characterized by a sense of resemblance between their form and their meaning. The most common examples are onomatopoeias such as “cock-a-doodle-doo” or “woof-woof,” but recent research shows that speakers of various languages perceive a relationship between sound and meaning in various words of their language. This is the case for English words such as
Palavras icônicas são caracterizadas por uma relação de semelhança entre sua forma e seu significado. Os exemplos mais comuns são as onomatopeias como "cocoricó" ou "au-au", mas pesquisas recentes mostram que falantes de várias línguas percebem uma relação entre som e significado em diversas palavras de sua língua. Este é o caso de palavras do inglês como
Iconic words are characterized by a sense of resemblance between their form and their meaning. The most common examples are onomatopoeias such as “cock-a-doodle-doo” or “woof-woof,” but recent research shows that speakers of various languages perceive a relationship between sound and meaning in various words of their language. This is the case for English words such as
Iconic words are characterized by a sense of resemblance between their form and their meaning. In Portuguese, for example, words such as
There is now substantial evidence that iconicity is a fundamental feature of all human languages, both spoken and signed. This observation is supported by analysis showing that form-meaning correspondences are found in thousands of typologically diverse natural languages (Blasi et al., 2016; Winter et al., 2022), and studies indicating that certain phonemes are prevalent in the words used in languages from different families to denote concepts related to perceptual categories (Sidhu et al., 2021; Erben Johansson et al., 2020; Joo, 2020). As a result, the study of lexical iconicity and its role in linguistic structure – long overlooked as minor or anecdotal within language research – has been gaining prominence in experimental studies.
Beyond shaping linguistic structure, lexical iconicity also influences the cognitive aspects of language. There is empirical evidence that iconicity in the lexicon of oral languages plays a role in language acquisition (Perry et al., 2015, 2018, 2021; Ananias and Godoy, to appear; Imai et al. 2008), word learning and recall (Sidhu et al., 2023; Van Hoey et al., 2023; Dingemanse et al., 2016), and language emergence and evolution (Vinson et al., 2021; Perlman, Dale and Lupyan, 2015; Verhoef, Kirby and de Boer, 2016). These studies indicate that iconicity significantly influence the cognitive dimension of language by providing a way to link linguistic structure with speakers’ perceptual experience (Perniss, Thompson and Vigliocco, 2010).
The relationship between iconicity and language acquisition/emergence has been investigated for spoken languages by the use of corpora and experimental research (Perry et al., 2015, 2018; Massaro and Perlman, 2017; Imai et al., 2008
The authors frame their theory using triangle models of word recognition because they specify a route by which meaning is retrieved via phonology. According to these models, word recognition encompasses three elements: a word’s meaning, its orthography and its phonology. The word’s meaning would be accessed via two paths: directly from its orthography, and indirectly from its orthography via its phonology. The connections among these components are bidirectional, and the relative weighting of each pathway shifts according to task demands (Balota, Paul and Spieler, 1999). Supporting evidence comes from findings that phonological variables (e.g., spelling-sound regularity; Hino and Lupker, 1996) exert greater or lesser influence on word recognition depending on the degree to which the orthography-phonology pathway is engaged.
As Sidhu et al. (2020) emphasize, triangle models consider the mappings between phonology and semantics to be arbitrary and learned through experience. The authors’ hypothesis, on the other hand, is that some of these mappings might arise or occur naturally – or be easier to acquire – because of the intrinsic similarity between phonology and semantics. Iconic words with more natural (or direct) links between semantics and phonology would, then, be easier to process. This hypothesis was tested in two experiments.
The authors operationalized iconicity on a gradient scale, using the iconicity ratings from Perry et al. (2015) and Winter et al. (2017) to determine each word’s level of iconicity. In Experiment 1, a lexical decision task, participants were asked to judge if a string of letters was a word (“Is this letter string a word?”). In Experiment 2, a phonological lexical decision task, they were asked whether the string of letters sounded like a word (“Does this letter string SOUND like a word?”). By changing the question, Sidhu
Although compelling, evidence of the relationship between language processing and iconicity in the natural lexicon of spoken languages remains scarce. Unlike the study conducted by Sidhu et al. (2020), research reporting an iconicity effect have primarily focused on onomatopoeias and/or ideophones, and have typically relied on electrophysiological data (Meteyard et al., 2015; Peeters, 2016; Vigliocco et al., 2020; Lockwood and Tuomainen, 2015). More recent studies that relied primarily on behavioral measures have also shown that iconicity correlates positively with with word guessing accuracy (Bonnie, Dunn and Dingemanse, 2023; Van Hoey et al., 2023), but their goal was not to assess whether iconicity in one’s native language would facilitate language processing.
When it comes to the effect of iconicity on reaction times in a word-recognition lexical decision task, the available evidence is limited to data from English speakers in Sidhu et al. (2020). Because it is unknown whether the findings in this study generalize to languages other than English, we propose replicating their study with speakers of Brazilian Portuguese. Specifically, we aim to the test the author’s claim that “iconicity provides a benefit to visual word recognition in typical adults” (Sidhu et al., 2020, p. 179). In order to do so, we plan to replicate their non-degraded condition in Experiment 1 with a sample of Brazilian Portuguese words.
The decision to replicate only the non-degraded condition in Experiment 1 was based on the absence of task or condition effects in both experiments of the original study. The hypothesis that participants would rely more heavily on phonological cues in certain contexts – and that this would modulate iconicity effects on written word recognition – was not supported. The author’s final remarks on this topic are not conclusive, as they recognize their data do not support that iconicity effects necessarily arise from special links between phonology and semantics. They note that the main effect of iconicity observed in both experiments is consistent with Meteyard et al.’s (2015) speculation that iconic words may have additional connections between semantic representations and modality-specific features. Therefore, we choose to concentrate our efforts in replicating the main finding that reaction times in a classic lexical decision task paradigm could be influenced by the stimuli’s level of iconicity.
The choice of Brazilian Portuguese (BP) as the target language for replication also warrants some consideration. BP is a specific variety of Portuguese, a Romance language of the Indo-European language family. It differs from European Portuguese and varieties used in Portuguese-speaking African countries, and it is the language of more than 200 million inhabitants of Brazil. Experimental studies have shown that BP speakers display patterns similar to those of speakers of other languages with respect to sound symbolism, a phenomenon related to iconicity. BP speakers tend to use low vowels and longer words to name larger objects, and high vowels and shorter words to name smaller objects (Godoy et al., 2020). They are also sensitive to the bouba-kiki effect, associating segments such as /p, t, k/ with spiky objects and /m, l, b/ with rounded objects (Godoy et al., 2018; Silva and Bellini-Leite, 2020; Godoy and Ananias, 2022). As in English and Japanese (Kawahara et al., 2021), they also associate voiced obstruents to semantic categories such as ‘evil’ when asked to name fictional characters (Godoy et al., 2021). These studies have focused on the elicitation or judgement of pseudo-words, and did not look for iconic mappings in BP. Recent studies examining the ionicity of the BP lexicon have shown a negative correlation between subjective iconicity ratings and age of acquisition (Godoy, 2026; Ananias and Godoy, to appear), an effect already reported for other languages (Perry et al., 2015).
Similarly to other languages, BP has a set of lexical items that are believed to be etymologically derived from onomatopoeias (e.g.,
English also differs from BP in regarding the presence of systematic submorphemic sound-meaning mappings known as phonaestemes. The most well-known case of phonaestheme in English is
It remains an open question to what extent the presence of both non-arbitrary sound-meaning associations (such as phonaesthemes) and onomatopoeic-sounding nouns and verbs (e.g.,
Finally, it is worth mentioning that BP does use full or partial reduplication to create expressive vocabulary, the most productive of this process being the reduplication of verbs
This decision also aims at replicating the original study as faithfully as possible. Reduplication is also present in English, and words formed by this morphological process are also perceived as highly iconic in this language (cf. Winter et al. (2023) for the mean iconicity ratings of
Differences in the expressive vocabularies of English and BP may influence the extent to which speakers of each language are affected by iconicity during word recognition tasks. At the same time, social and cultural differences between Canada and Brazil – particularly within their academic communities, which were the sample populations in the original study and our replication – may also influence how language is processed. Psycholinguistic research involving typologically diverse languages and heterogeneous populations provide an opportunity to test experimental hypotheses in new contexts, and to gain insights into the universality of their findings. This is the case of the proposed replication: our main goal is to investigate whether the main finding in Sidhu et al. (2020) can be generalized to a different target language and its speakers.
We will run a word recognition lexical decision task to evaluate whether iconicity influences language processing in typical adults. Our replication of the original study aims to generalize its findings by running the same experiment with a sample of Brazilian Portuguese speakers and by using a sample of Brazilian Portuguese words. Each of these samples is described in detail below.
Sidhu et al. (2020) sampled English words from Perry et al.’s (2015) and Winter et al.’s (2017) iconicity norming studies. In these studies, participants were presented with a Likert scale that ranged from -5 (the word’s sound seemed to contradict its meaning) to 5 (the word was highly iconic). Rating a word as 0 would indicate that the sound/meaning relation was completely arbitrary. The lexical decision task to be replicated selected 120 words from these studies: 40 non-iconic words, with ratings ≥ -0.5 and ≤ 0.5; 40 onomatopoeic words and 40 non-onomatopoeic words, all of which were highly iconic, with ratings ≥ 2.5. Onomatopoeic words had a phonology intended to mimic a sound (e.g.,
In order to draw a sample of Portuguese words following the same criteria described in the original study, we will use preliminary data from an iconicity norming study in Brazilian Portuguese (Godoy, 2026). In this study, 7,000+ Brazilian Portuguese words were subjected to an iconicity rating task. Participants were shown 40 words and asked to rate the extent to which they believed each word sounded like its meaning. Instructions were adapted from Winter et al. (2023) and probed participants to think of iconicity as a continuum. Ratings ranged from 1 to 7, reflecting recent discussions in iconicity research suggesting that this scale more accurately captures perceptions of iconicity than the –5 to 5 range used in previous studies (Winter et al., 2023; Hinojosa et al., 2021).
The main difference between the iconicity norms from the BP study and those used by Sidhu et al. (2020) lies in the range used to measure iconicity. We propose using the samples’ mean and standard deviation as cut-off points to select the words to be included in the replication. In Sidhu et al. (2020), the non-iconic stimuli were selected from words with iconicity ratings at least 0.3 standard deviations below the sample mean
Finally, we will ensure that words in each group are matched on length, number of morphemes, log subtitle word frequency, orthographic Levenshtein distance, phonological Levenshtein distance, and mean bigram frequency. For that, we will use Alzahrani (2025) metrics for Brazilian Portuguese. In case this is not possible, we can resort to the P-PAL dataset (Soares et al., 2018) for orthographic and phonological Levenshtein distance, and mean bigram frequency. Sidhu et al. (2020) also matched the words in the three groups considering their age-of-acquisition, concreteness, and initial syllable rime phonological consistency. We may be able to control for the first two variables, as there are normative studies on age-of-acquisition and concreteness in Portuguese (e.g., Cameirão and Vicente, 2010; Ananias and Godoy, 2024; Soares
Additionally, 120 nonwords will be created. These nonwords will conform to Brazilian Portuguese phonotactics, and we will avoid creating pseudo-homophones with real words (e.g.,
The original study sampled a total of 80 participants (40 for the non-degraded condition) from the undergraduate student body of a Canadian university (60 female;
As stipulated by the national guidelines on research ethics involving human participants, no participant will receive any form of compensation for taking part in the study. The research plan must be approved by the Research Ethics Committee of the university where the experiment will be conducted.
Participants will take part in a lexical decision task in which they have to categorize a presented letter string as a word or nonword using the keyboard. For each trial, participants will see a fixation cross for 400ms, then a blank screen for 200ms, and then the target stimuli. Target stimuli will remain on the screen until participants respond to them. Response will trigger a 500ms blank screen, and then the next trial will be presented. If there is an error, they saw the word “Incorrect”, and hear a brief sound during the blank screen. Stimuli will be presented in random order in two consecutive blocks, with a break between them. In each block, participants will see an equal proportion of stimulus type. All these procedures were adapted from the original study.
Reaction times and response accuracy will be analyzed to replicate the original analyses by Sidhu et al. (2020). Before fitting statistical models, raw data will be processed.
Consistent with the original study, we will apply the following steps to clean our data for the analysis of reaction times:
1. All incorrect responses will be excluded;
2. Trials with reaction times less than 200msec or more than 3000msec will be removed;
3. Remaining trials that fall over 2.5 standard deviations from a participant’s mean will also be removed.
These steps will also be used in the analysis of accuracy data, except for step 1.
We will take a confirmatory approach and run linear mixed models to test the main experimental hypothesis that iconicity can affect reaction times. We will also rely on the same model reported in the original study: a linear mixed effect model with reaction times as the response variable, Iconicity score as a predictor, and length, log frequency and orthographic Levenshtein distance as co-variates. The model, summarized in (1), will also include (following the original study) random intercepts for participants and items. The statistical model used in Sidhu et al. (2020) did not include a random slope for iconicity per participant due to convergence issues, but we will add this random effect if the model converges. Although Iconicity will be treated as a factor for stimuli selection, it will be coded as a continuous variable in the analyses to match the original study. Therefore, we will use Iconicity scores from Godoy (2026).
(1) reaction time ~ iconicity + length + log frequency + orthographic Levenshtein distance + (1+iconicity|participant) + (1|word)
This initial analysis will corroborate or not the original finding we are trying to replicate. Our main interest here is the effect of iconicity on reaction times: if we find a negative and significant correlation between these two variables, one can say we were able to replicate the main finding of the original study. Failure to find iconicity effects in this model, or a positive, significant correlation between ionicity and reaction times, will both be interpreted as a failure to replicate the original finding.
Following the original study, we will use the packages
Following the original study’s analysis, we will also run a logistic mixed effects model with accuracy as the response variable and the same fixed and random variables as in the reaction time analysis. If we find that iconicity increases the chance of a correct response, we can say we successfully replicated the main finding of the study regarding the effect of iconicity on accuracy. Failure to find iconicity effects in this model, or a negative, significant effect between iconicity and accuracy (i.e., more iconic words lead to less accuracy), will both be interpreted as a failure to replicate the original finding.
The analysis carried by Sidhu et al. (2020) in their dataset follow the standards currently used by the academic community for the analysis of reaction time data, so we do not intend to run our data on other models to test the generalizability of their finding. Our main goal, as previously stated, is to test their finding against a new target language, not against other statistical tools. However, we plan to run additional analysis depending on the distribution of our data. Sidhu et al. (2020) do not mention whether their model violates linear model’s assumptions, specifically, multicollinearity, homoscedasticity and normality of the residuals. We will perform a residual analysis to assess the goodness of fit of the initial statistical model. In case residuals violate one of the aforementioned assumptions, we will take on the following steps to model our data.
- In case of
- In case of
- In case of
The
We will run an additional analysis to explore the interaction between iconicity and trial number on reaction times. This will be done to replicate an exploratory analysis described in the original study.
Apart from their main analysis, Sidhu et al. (2020) also examined whether the effect of iconicity on reaction times could also be observed in the English Lexicon Project dataset (ELP; Balota et al., 2007). They fitted a linear model similar to those described in Section 3.2, but without random effects. The analysis showed no effect of iconicity ratings on either reaction times or accuracy. The authors argue that a list context effect could explain why iconicity effects differ between the two datasets. In their dataset, but not in the ELP, two thirds of all words presented to participants (or 30% of all stimuli) were highly iconic. This could trigger participants to use iconicity in the experimental setting as a cue to solve the task.
They tested this hypothesis by adding trial number and its interaction with iconicity to the model described in section 3.2. Results showed a significant interaction between those factors, such that iconicity had a greater effect on reaction times for later trials. A detailed inspection of the data showed that iconicity had no effect on reaction times for the first 60 experimental trials, but the negative correlation between those two variables were evident for the last two third of trials. This corroborates the list context effect and indicates that participants may “learn” to use iconicity as a valuable cue during the task.
In case we successfully replicate the findings in Sidhu et al. (2020) with the main analyses described in section 3.2, we will run this analysis to test whether Brazilian participants are also prone to this interaction effect. In case we fail to replicate the main finding that iconicity negatively correlates with reaction time, this additional analysis may help us gain insight on the differences between participants. It may be the case that there is no iconicity effect at all for the Brazilian cohort, or it may be the case that an effect only emerges in the final trials of the experimental items, making it too subtle to be observed as a main effect in the dataset.
In case our main analysis finds a null result, we will run an additional analysis to explore some alternative hypothesis. Cnudde, Sidhu and Pexman (in preparation) reanalyzed the data of Sidhu et al.’s (2020) Experiment 2 to investigate the role of individual differences on reaction times, and found that participants with greater pseudohomophone accuracy showed larger iconic facilitation in the task. In a subsequent phonological lexical decision task experiment, they controlled the imitativeness of the iconic stimuli used: some iconic words were highly imitative of their meaning (e.g.,
In light of these findings, we may run the same model described in section 3.2, but this time dummy-coding Iconicity as a factor with three levels: non-iconic words, non-onomatopoeic iconic words and onomatopoeic iconic words. If imitativeness plays a greater role than iconicity itself in word recognition, we may observe faster reaction times for this latter group. The aim of this additional analysis is not to replicate the original study’s main findings, but rather to shed light on the causes of a possible null effect. Regardless of its outcome, the analysis will be described as exploratory and will not be used to claim that the replication was successful or unsuccessful. As previously stated, we will only run this analysis in case our main analysis finds a null result.
Data collection will only start after the approval of this registered report. We anticipate that data collection and statistical analysis will require approximately three months to complete. We also note that data collection cannot be conducted in January and February due to the academic recess (summer break) in Brazil. Drafting and submitting the final manuscript will take an additional month.
The data that support the findings of this study will be openly available in Open Science Framework at https://doi.org/10.17605/OSF.IO/D827B
Iconicity in the lexicon remains an understudied topic in Brazil and, more broadly, in Latin America. At the same time, there is a large international community dedicated to this topic, as attested by the international publication cited in this work and scientific conferences such as the
The authors declare no competing interests.
We also declare that no AI tools were used in the creation of this manuscript or in any aspect of the research reported in it.
This work was carried out with the support of the Coordination for the Improvement of Higher Education Personnel – Brazil (CAPES) – Financing Code 001, and the National Council for Scientific and Technological Development (CNPq, Brazil), grants 304393/2024-0 and 201419/2024-6.
ALZAHRANI, Alaa. Jiwar: A database and calculator for word neighborhood measures in 40 languages.
ANANIAS, Thayná C.; GODOY, Mahayana C. A iconicidade no Português e sua relação com idade de aquisição e classes de palavras [Iconicity in Portuguese and its relationship with age of acquisition and lexical category].
ANANIAS, Thayná C.; GODOY, Mahayana C. Normas de concretude para 521 palavras do português e sua relação com aquisição da linguagem.
BALOTA, David A.
BALOTA, David. A.; PAUL, Stephen; SPIELER, Daniel H. “Attentional control of lexical processing pathways during word recognition and reading”.
BATES, David
BLASI, Damián E.
BERGEN, Benjamin K. The psychological reality of phonaesthemes.
CAMEIRÃO, Manuela L.; VICENTE, Selene G. Age-of-acquisition norms for a set of 1,749 Portuguese words.
CNUDDE, Kelsey; SIDHU, David; PEXMAN, Penny. The role of phonology in iconicity effects: Evidence from individual differences.
Couto, Hildo Honório do. A reduplicação em português.
DINGEMANSE, Mark
ERBEN JOHANSSON, Niklas
GARCIA, Guilherme D.
GODOY, Mahayana C.; ANANIAS, Thayná C. The bouba-kiki effect in a production task.
GODOY, Mahayana C.
GODOY, Mahayana C.
GODOY, Mahayana C.
GONÇALVES, Carlos Alexandre; VIALLI, Luciana Albuquerque. “Reduplicação”.
HINO, Yasushi; LUPKER, Stephen. J. Effects of polysemy in lexical decision and naming: An alternative to lexical access accounts.
HINOJOSA, José A.
IMAI, M.
JOO, Ian. Phonosemantic biases found in Leipzig-Jakarta lists of 66 languages.
KWON, Nahyun; ROUND, Erich. Phonaesthemes in morphological theory.
LOCKWOOD, Gwilym; TUOMAINEN, Jyrki. Ideophones in Japanese modulate the P2 and late positive complex responses.
LÜDECKE, D
MASSARO, Dominic W.; PERLMAN, Marcus. Quantifying Iconicity’s Contribution during Language Acquisition: Implications for Vocabulary Learning.
METEYARD, Lotte.
NIELSEN, Alan; DINGEMANSE, Mark. Iconicity in word learning and beyond: A critical review.
PEETERS, David. Processing consequences of onomatopoeic iconicity in spoken language comprehension. 38th ANNUAL MEETING OF THE COGNITIVE SCIENCE SOCIETY.
PERLMAN, Marcus; DALE, Rick; LUPYAN, Gary. Iconicity can ground the creation of vocal symbols.
PERNISS, Pamela; THOMPSON, Robin; VIGLIOCCO, Gabriella. Iconicity as a General Property of Language: Evidence from Spoken and Signed Languages.
PERRY, Lynn K.
PERRY, Lynn K.; PERLMAN, Marcus; LUPYAN, Gary. Iconicity in English and Spanish and its relation to lexical category and age of acquisition.
PERRY, Lynn K.
R Core Team.
SANTOS, Leandro Borges dos
SIDHU, David; KHACHATOORIAN, Nareg; VIGLIOCCO, Gabriella. Effects of Iconicity in Recognition Memory.
SIDHU, David
SIDHU, David; VIGLIOCCO, Gabriella; PEXMAN, Penny. Effects of iconicity in lexical decision.
SILVA, Daniel M. R.; & Bellini-Leite, Samuel C. Cross-modal correspondences in sine wave: Speech versus nonspeech modes.
SOARES, Ana Paula
SOARES, Ana Paula
VAN HOEY, Thomas
VERHOEF, Tessa; KIRBY, Simon; DE BOER, Bart Iconicity and the emergence of combinatorial structure in language.
VIGLIOCCO, Gabriella
VILLALVA, Alina; GONÇALVES, Carlos Alexandre. “The phonology and morphology of Word-formation”.
VINSON, David P.
WINTER, Bodo
WINTER, Bodo
WINTER, Bodo
EDITOR 1: Miguel Oliveira Jr
ORCID: https://orcid.org/0000-0002-0866-0535
AFFILIATION: Universidade Federal de Alagoas, Alagoas, Brasil.
-
EDITOR 2: Mailce Borges Mota
ORCID: https://orcid.org/0000-0002-8674-2480
AFFILIATION: Universidade Federal de Santa Catarina, Santa Catarina, Brasil.
-
ASSESSMENT: This registered report proposes a timely investigation of iconicity effects in lexical decision using Brazilian Portuguese. The study addresses an important gap by extending research beyond English and WEIRD populations, using a well-designed replication framework with robust methods. The sample size planned, the approach to analysis, and the materials selected are appropriate for examining whether iconic form-meaning mappings facilitate word recognition in Brazilian Portuguese. Upon completion of the suggested revisions, this registered report will make a valuable contribution to our understanding of iconicity across languages and provide important data for the field of psycholinguistics.
REVIEWER 1: Thomas Van Hoey
AFFILIATION: Katholieke Universiteit Leuven, Lovaina, Bélgica.
-
REVIEWER 2: Zlatomira Ilchovska
ORCID: https://orcid.org/0000-0001-6682-9952
AFFILIATION: University of Birmingham, Birmingham, Inglaterra.
-
REVIEWER 1
2026-01-15 | 10:10 AM
The way the research in this registered report is conceived makes sense and addresses a gap in the understanding of iconicity in Brazilian Portuguese. The methods are not wholly original, but rather a novel combination of robust, well-understood methods and materials. This has the benefit that it is easy to draw up a convincing plan for conducting the research and also analyze the behavioral data that will be gathered (accuracy and reaction time). I see no leaps in logic that do not match the state of the art. Overall, this is a very well-designed study. Results may be of use to researchers interested in iconicity, as well as researchers focusing on Brazilian Portuguese.
In general, this registered report takes an important next step for studying the phenomenon of linguistic iconicity, by investigating its effects on lexical decision. The choice for Brazilian Portuguese seems opportune, as there is an emerging body of work on iconicity in this language. I found it an interesting notion that the authors explicitly include attention for the written word, on top of sound in the form-meaning mapping. This reminds me of a body of work on Chinese ideophones and iconicity where this is also highly relevant: speakers of a language will use whatever semiotic resources are available. In languages with a tradition of writing that thus also includes written words, not just sounds. In terms of the main experiment, I think the design is convincing. The proposed target sample (N=40 -> comparison with previous studies) is adequate, and the different analyses (reaction time vs. accuracy as dependent variables) make sense. I like the contingency plan and also the proposed timeline. Finally, the dissemination strategy makes a lot of sense too, especially by making the study available afterwards in Portuguese is a great idea (fn 8).
-
REVIEWER 2
2026-01-08 | 03:45 PM
Thank you for the opportunity to review and acquaint myself with this manuscript. The study plan will investigate iconicity in Brazilian Portuguese, replicating earlier research by Sidhu et al. (2020), in particular – Experiment 1 of their work. It will present Brazilian Portuguese speakers with a lexical decision task to measure the role of iconicity for language processing at the word level. Taken the limited evidence of various linguistic phenomena, including iconicity, in languages other than English and in non-WEIRD samples, this study will provide valuable findings and is for high significance for our understanding of semantic word processing. The replication is planned well and in line with the original research. There are only some minor comments to be addressed below.
Abstract
1. It would be good to mention a sentence or two of the expected findings here and the planned analyses. Introduction
1. “Their initial hypothesis, that more iconic words would lead to faster recognition, builds upon previous studies showing that an iconic link between form and meaning leads to electrophysiological differences during language processing (Peeters, 2016), more resistance to aphasia-related loss (Meteyard et al., 2015) and less accuracy in semantic decision tasks (Lupyan and Winter, 2018).” It would be good to clarify here in a bit more detail the exact direction of the relationship between iconicity and the findings of these studies (e.g., if higher iconicity is related to higher accuracy – as you previously suggest – or lower accuracy in semantic decision tasks). 2. Similarly to other languages, BP has a set of lexical items that that are believed to be etymologically derived from onomatopoeias (e.g., miar (“to meaow”), bufar (“to snort”), roncar (“to snore”)). Repetition of “that”. An extra letter in “to meow” here and in a few further instances. Methods
1. “The data collection for the iconicity norming study and its analysis is due to March 2026. However, at the moment we already have data from more 2,300 participants (all BP native speakers;” Missing words typos in “due to appear in March 2026” and “data from more than”.
2. “Trials with reaction times less than 200msec or 3000msec will be removed;” Should me “less that 200msec or more than 3000msec”.
3. In point 3.3 you talk about logistic mixed effects model but then mention correlations. Could you please specify what kind of additional correlational analyses you have selected to perform and what their motivation is, in addition to the mixed effects models?
4. In 3.5 the following again talks about correlations “A detailed inspection of the data showed that iconicity had no effect on reaction times for the first 60 experimental trials, but the negative correlation between those two variables were evident for the last two third of trials.”, but from my understanding, you have selected to run linear regression models. Could you please clarify and specify what model was selected in both this and the previous instance?
5. In 3.6 for the following “In light of these findings, we may run the same model described in section 3.2, but this time coding Iconicity as a factor with three levels: non-iconic words, non-onomatopoeic iconic words and onomatopoeic iconic words.”, it would be good to mention the contrast scheme that will be used to code the different levels, for completeness and clarity of any effects you may observe and their alignment with your hypotheses.
-
2026-02-06 | 10:34 AM
We are grateful to both reviewers for their careful reading of our manuscript and for their valuable suggestions to enhance its flow and readability. I apologize for taking 3 days longer than the deadline previously stipulated to submit the new version of the registered report, but I was dealing with personal issues last week.
We have incorporated nearly all suggestions and made partial adjustments in the few cases where our understanding differed somewhat from the reviewers' comments (largely due to our failure to be sufficiently clear or explicit in the first version of the manuscript).
Regarding Reviewer 1's comment 3, previous work on Pokemon names in BP found size sound symbolism that may have iconic roots (especially the /i/-/a/ distinction). We did not remove the mention of this and other studies on sound symbolism from the paper, but added that they focused mainly on the elicitation or judgement of pseudo-words, and did not investigate iconic mappings in the BP lexicon.
With respect to Reviewer 2's fourth comment, we are describing a model in which iconicity (predictor) and reaction times (response variable) are both numerical variables. The result of the regression analysis show a nagative correlation between these variables, mediated by trial number. In this case, the correlation is not a new statistical analysis. It is just the correlation tested within the linear regression model.
Finally, we noted that both reviewers were confused by the mention of the large-scale study with 7,000+ words in Portuguese (Godoy, 2026). This is a different study, from which we will sample the 120 stimuli for the lexical decision task. We may have caused this confusion by adding non-important methodological information about it in the text (number of participants, their demographics etc). We chose to omit these methodological details, which can be recovered from the iconicity study (Godoy, 2026). We hope this will improve the readability of the text.
Finally, we note that reviewers do not oppose – and one of them actually agrees with – our plan to submit the Final Report in English and Portuguese. As stated in the Registered Report, this decision aims at making our research (which in publicly funded) available to the international community and also to undergraduate students and other members of the national scientific community who are not fluent in English. If the Editorial Board agrees with that, we will prepare the Final Report with our results in both languages.