1. Introduction
Bilingualism is a worldwide phenomenon across countries, social classes and age groups, due to heterogeneous factors such as the interaction between languages within a country, political and economic issues or personal reasons (Hülmbauer & Seidlhofer, Reference Hülmbauer, Seidlhofer, Berthoud, Grin and Lüdi2013; Jenkins, Reference Jenkins2015; Piller & Takahashi, Reference Piller and Takahashi2011). The question of the influence of bilingualism on cognitive processing is a much-debated issue wherein the bilingual advantage theory and the bilingual disadvantage theory have been in contraposition for several years. Research on the interaction between multilingual proficiency, and neural language mechanisms, spanning usage, structural organization and cognitive outcomes (De Groot, Reference De Groot2011) has produced diverse and sometimes contradictory findings (Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004, Reference Bialystok, Craik and Luk2008; Costa et al., Reference Costa, Hernández and Sebastián-Gallés2008; Finkbeiner et al., Reference Finkbeiner, Gollan and Caramazza2006; Francis, Reference Francis2005; Gollan et al., Reference Gollan, Montoya, Cera and Sandoval2008; Ivanova & Costa, Reference Ivanova and Costa2008; Kaushanskaya & Marian, Reference Kaushanskaya and Marian2007; Pelham & Abrams, Reference Pelham and Abrams2014). Thus, the exploration of bilingualism continues to challenge and reshape our understanding of human cognition, contrasting cognitive decline, increasing executive functions (EFs) and attesting that bilinguals are not two monolinguals in one person (Grosjean, Reference Grosjean1989; Grosjean & Li, Reference Grosjean and Li2013).
1.1. Theoretical background
1.1.1. Bilingualism versus second-language acquisition
A fundamental distinction in the study of multilingualism lies between bilingualism and second-language acquisition (SLA). SLA refers to the process through which people learn a language other than their native language in various contexts, including in educational settings, in immersed environments or through self-directed study. It involves several stages and is influenced by factors such as age, motivation, exposure and individual differences in learning styles.
Bilingualism, on the other hand, is the ability to use two languages proficiently and is sometimes not chosen but acquired in childhood due to the parents’ linguistic background, and it is concerned with ability and fluency in both languages, rather than just the acquisition process. It can manifest in various forms, such as simultaneous bilingualism (learning two languages from an early age) or sequential bilingualism (learning a second language [L2] after the first language [L1]): bilingualism refers, thus, to the ability to use two languages competently. It encompasses both the capacity to communicate in multiple languages, and cognitive, social and cultural implications of using more than one language, through the strengthening of cognitive skills, e.g., improving problem-solving abilities. In essence, SLA is primarily about the learning journey of acquiring a L2, while bilingualism focuses on the proficiency and use of two languages in everyday life (Grosjean & Li, Reference Grosjean and Li2013).
In the bilingual brain, the two languages are jointly activatable even when only one of them is used (Hernandez, Reference Hernandez2013; Hernandez et al., Reference Hernandez, Bates and Avila1996; Kaushanskaya & Marian, Reference Kaushanskaya and Marian2007; Marian et al., Reference Marian, Blumenfeld and Kaushanskaya2007; Marian & Spivey, Reference Marian and Spivey2003): this double constant activation requires an adaptive mind capable of selecting and avoiding interference from the unwanted language to guarantee efficiency in communication in the target language. The importance of the selective attention processes and inhibitory control has become a central point in the analysis of bilingual effects on cognition, which seem to spread throughout life, even in old age (Fan et al., Reference Fan, McCandliss, Sommer, Raz and Posner2002; Hernandez, Reference Hernandez2013; Hernandez et al., Reference Hernandez, Bates and Avila1996). EFs, or the set of cognitive processes such as inhibition, working memory, selective attention and cognitive flexibility, are in fact necessary for the cognitive control responsible for multiple functions. Enhanced EFs in bilinguals range from alternation between languages to the need to suppress one of them in parallel activation (Bialystok, Reference Bialystok2009; Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004, Reference Bialystok, Poarch, Luo and Craik2014). This continuous cognitive juggling not only strengthens executive control but also shapes how bilinguals structure and interpret information beyond linguistic competence. Moreover, their efficiency can be influenced by the constant use of multiple languages, with a simultaneous activation of the two languages at various levels and situations (Green & Abutalebi, Reference Green and Abutalebi2013). Therefore, mastering two languages with similar proficiency (= being bilingual) means possessing and controlling two different types of categorizations. In this study, the participants are qualified as bilinguals, using fluently two languages in daily life. They have acquired Italian either from birth or from an early age and have experienced long stays in immersion in Italy. Often, they have also studied Italian at a university level.
1.1.2. Language and color perception
Languages encode stimuli differently, but universal cognitive principles shape meaning (Cancho & Solé, Reference Cancho and Solé2003; Piantadosi et al., Reference Piantadosi, Tily and Gibson2011): semantic variation arises from communicative needs (Levy & Jaeger, Reference Levy and Jaeger2006), cultural context (Roberson et al., Reference Roberson, Davidoff, Davies and Shapiro2005) and lexical evolution (Mylonas et al., Reference Mylonas, Caparos and Davidoff2022). Recent studies on color naming highlight “informativeness” as a key factor in cross-linguistic variation (Gibson et al., Reference Gibson, Futrell, Jara-Ettinger, Mahowald, Bergen, Ratnasingam, Gibson, Piantadosi and Conway2017; Regier et al., Reference Regier, Kay and Khetarpal2007, Reference Regier, Kemp, Kay, MacWhinney and O’Grady2015), suggesting color categories emerge from optimizing linguistic and perceptual constraints (Piantadosi et al., Reference Piantadosi, Tily and Gibson2011; Regier et al., Reference Regier, Kemp, Kay, MacWhinney and O’Grady2015).
Color categories are anchored by focal colors, defined as perceptually salient, monolexemic terms predictive of object recognition (Berlin & Kay, Reference Berlin and Kay1969). Heider (Reference Heider1971) and Hering (Reference Hering1878, Reference Hering1964) reinforced the universality of these prototypes, rooted in perceptual salience and neurophysiology. Jameson and D’Andrade (Reference Jameson and D’Andrade1997), and later Abbott et al. (Reference Abbott, Griffiths and Regier2015), linked focal colors to both perceptual irregularities and linguistic constraints, explaining why unrelated languages often share focal points despite differing category boundaries.
Language affects perception: when two colors are labeled by the same term, speakers are more likely to judge them similar and confuse them in memory tasks, while distinct labels enhance discrimination (Kay & Kempton, Reference Kay and Kempton1984). This raises key questions for bilinguals: Does perception shift depending on lexical distinctions in L1 versus L2? Does a more fine-grained L2 enhance a shift?
In bilinguals, color categorization reflects a blend of lexical systems. Factors such as L2 proficiency and immersion impact category stability (Athanasopoulos et al., Reference Athanasopoulos, Damjanovic, Krajciova and Sasaki2011). Studies show bilinguals may adopt L2 focal colors and exhibit less stable boundaries, suggesting flexible, dynamic perceptual systems shaped by experience (Alvarado & Jameson, Reference Alvarado and Jameson2002; Andrews, Reference Andrews1994; Sayim et al., Reference Sayim, Jameson, Alvarado and Szeszel2005). This adaptability underscores the complex interplay between language, perception and cognition in bilingual individuals.
1.1.3. Studies on two blues
The color blue, conventionally and prototypically associated with the sea and the sky, is considered extremely salient in several cultures so that languages have sometimes developed a specific chromatic vocabulary to convey the multiple meanings (Uusküla, Reference Uusküla2014). In fact, a variety of basic color terms have evolved to denote the many blue tones connected to the corresponding color category.
Numerous languages from the Mediterranean region, the Uralic family and the Slavic region have attested this refinement; for example, Italian, Russian, Turkish, Greek, Maltese, Catalan, Polish, Ukrainian, Belarusian, Lithuanian and Udmurt have two distinct words for blue: one for the dark hue and one for the light hue (Athanasopoulos, Reference Athanasopoulos2009; Borg, Reference Borg2011; Coventry et al., Reference Coventry, Mitsakis, Davies, Jover, Androulaki and Gômez-Pestaña2006; Davies et al., Reference Davies, Corbett and Margalef1995; Davies & Corbett, Reference Davies and Corbett1994; Giacalone Ramat, Reference Giacalone Ramat1967; Grossmann, Reference Grossmann1988; Özgen & Davies, Reference Özgen and Davies1998; Paggetti et al., Reference Paggetti, Bartoli and Menegaz2011; Paramei, Reference Paramei2005; Paramei et al., Reference Paramei, D’Orsi and Menegaz2014; Putzu & Ignazio, Reference Putzu and Ignazio2000; Rätsep, Reference Rätsep2011, Reference Rätsep2012; Sagaspe et al., Reference Sagaspe, Sanchez-Ortuno, Charles, Taillard, Valtat, Bioulac and Philip2006; Sandford, Reference Sandford2011; Sinkeviciute et al., Reference Sinkeviciute, Mayor, Vulchanova and Kartushina2024; Thierry et al., Reference Thierry, Athanasopoulos, Wiggett, Dering and Kuipers2009). Japanese is also included in this group of languages (Kuriki et al., Reference Kuriki, Lange, Muto, Brown, Fukuda, Tokunaga, Lindsey, Uchikawa and Shioiri2017).
In this field of research, a special place is occupied by studies on the relationship between different blue categorizations and multiple languages subsisting in the same individual, like in bilinguals (Sinkeviciute et al., Reference Sinkeviciute, Mayor, Vulchanova and Kartushina2024).
For instance, Athanasopoulos’ work (Athanasopoulos, Reference Athanasopoulos2009) investigating the blueFootnote 1 categorization in Greek–English bilinguals revealed that proficient bilinguals produced a shift of the Greek ble (dark blue) focus toward the focus of blue of English monolinguals in the lightness dimension. At the same time, the Greek ghalazio (light blue) focal color deviated from the English blue focus since the distance between the two Greek blues was maintained, keeping the polarization on the lightness value constant.
Meanwhile, Paramei et al. (Reference Paramei, D’Orsi and Menegaz2016) conducted studies on the Italian blues comparing Italian monolinguals, English monolinguals and Italian–English bilinguals living in Liverpool, tested in the two languages. They used color-naming tasks to explore the focal colors of the Italian words blu, azzurro and celeste and for English blue and light blue, which led to the detection of a lightness shift and a hue shift. The lightness shift was observed for the blue foci of bilingual speakers, which were perceived as darker than the Italian blu foci, whereas azzurro foci were similar to the English blue foci and celeste foci corresponded to the light blue of English monolingual speakers. The authors explained the bilinguals’ semantic shift of blue toward blu as a consequence of phonological and orthographic cross-linguistic analogies, which might facilitate the concept mediation from L1 to L2 since the access to L1 meaning is easier and more direct than that of the L2 (Kroll et al., Reference Kroll, Hell, Tokowicz and Green2010). Therefore, for bilinguals, the English blue was globally darker than the English monolinguals’ blue. The hue shift observed mainly concerns the bilinguals’ azzurro category, which revealed both a shift toward the English monolinguals’ blue, with a more purplish tint, and a deviation from the azzurro foci of Italian monolinguals.
Other studies (Paggetti et al., Reference Paggetti, Bartoli and Menegaz2011; Paggetti & Menegaz, Reference Paggetti, Menegaz and Kutulakos2012, Reference Paggetti and Menegaz2013, Reference Paggetti and Menegaz2015; Sandford, Reference Sandford2012; Stroop, Reference Stroop1935) provided additional results in favor of the twelfth BCT Azzurro using a large set of experimental tasks. Paggetti and Menegaz (Reference Paggetti, Menegaz and Kutulakos2012, Reference Paggetti and Menegaz2015) have, for example, examined both the perceptual and linguistic aspects of the blue region perception in Italian whose results supported the theory that there exists a color category for light blue. This category is labeled by a lexeme that has all the features of a basic color term: azzurro.
All these findings attested the presence of a linguistic impact on color perception and categorization in bilinguals, with lexical labels altering their perceptual representations, in combination with additional factors, like the exposure to the double linguistic environment, which can determine the functioning of the bilingual integrated mental lexicon. Moreover, they brought additional evidence to the hypothesis that Italian has more than one basic color term (BCT) (Berlin & Kay, Reference Berlin and Kay1969) for the blue color category, which is lexically and perceptually distinguished in two separate chromatic categories defined on lightness.
1.2. The present study
The primary aim of this study is to investigate how learning and using two typologically related languages with different lexical color partitions (Italian and French) affects perceptual color categorization in proficient bilinguals. Specifically, we aim to determine whether L2 lexical distinctions (Italian blu versus azzurro) reshape the perceptual categories established by the L1 (French), even in non-linguistic perceptual tasks. By doing so, we assess the extent to which language influences low-level perceptual processing and test the weak linguistic relativity hypothesis in a bilingual context. This would indicate that conceptual representation is not fixed by the L1, but pre-existing L1 properties can change with the acquisition of new L2 specific features, producing a cognitive reorganization. We investigated French–Italian bilinguals to address the limited availability of research on the bilingual brain’s role in color perception and to examine the differences in color categorization within two typologically similar languages coexisting in the same individual. This approach provides critical insights into how linguistic convergence shapes perceptual and cognitive processing in bilingualism.
In our case, if L1 (French) color cognitive representation is affected by the L2 (Italian), French–Italian proficient bilinguals will utilize the Italian blue linguistic partitioning (blu category versus azzurro category) of color space even in a perceptual task. Therefore, the bilingual mind adopts the categorical differentiations of the language owning the lexical distinctions, even though it is their L2, leading to a facilitation effect in the discrimination of blue cross-category stimuli compared to blue within-category stimuli.
One of the main objectives of the present study is to provide stronger empirical support for the weak relativity theory, exploring the possibility of additional basic color categories (BCCs) beyond the 11 identified by Berlin and Kay. We investigate how language shapes color perception, even in typologically similar languages.
A key focus is demonstrating the existence of a 12th BCT in Italian (azzurro), distinguishing light blue as a separate category, as seen in Russian, Japanese, Maltese, Greek and Catalan. By comparing French and Italian blue semantics, we contribute to the understanding of categorical perception (CP) in color processing (Goldstone & Hendrickson, Reference Goldstone and Hendrickson2010; Harnad, Reference Harnad1987).
Furthermore, this research sheds light on bilingual linguistic and cognitive adaptation, particularly in the debate on the bilingual cognitive advantage. By examining lexical and conceptual evolution in bilinguals, we aim to reveal how mastering multiple languages shapes cognitive processing and perceptual categorization, with a special focus on color perception.
We also want to demonstrate how the partitioning of a single language’s chromatic space is affected by the corresponding lexical categorization even in languages of the same linguistic family with a parallel semantic historical development. As evidence that language can actively participate in extra-linguistic tasks, particularly perceptual ones, this phenomenon would seem to be most evident in the manner that French and Italian share their blue color spectrum.
This study, thus, globally investigates whether the acquisition and proficient use of Italian, with its two distinct lexical categories for blue, modulate French–Italian bilinguals’ perceptual discrimination of color in accordance with Italian categorical distinctions even during purely perceptual tasks; to what extent L2 lexical structure influences L1 color categorization as reflected by shifts in category boundaries and reaction times (RTs); how bilinguals’ performance compares to that of French and Italian monolinguals under varying cognitive load conditions; and whether the presence of finer-grained L2 categories induces measurable cognitive restructuring of color perception.
1.2.1. Hypotheses: enumerated and explained
-
1. L1 color prototypes’ shift based on L2
Following previous studies (Erwin et al., Reference Erwin, Lerner, Wilson and Wilson1961; Caskey-Sirmons & Hickerson, Reference Caskey-Sirmons and Hickerson1977; Zollinger, Reference Zollinger1988; Jameson & Alvarado, Reference Jameson and Alvarado2003; Sayim et al., Reference Sayim, Jameson, Alvarado and Szeszel2005), we expect that in bilingual speakers highly proficient in their L2 (Italian), the L2 causes a shift in the way color prototypes are used and defined in the L1 color words (French) due to the presence of a lexical distinction. Therefore, bilinguals have an idiosyncratic functioning, and they cannot be considered as two monolinguals in one person.
-
2. Lexical distinction always dominates the categorization used
Following the previous hypothesis, we expect that people mastering two linguistic and conceptual systems adopt the categorical differentiations of the language owning the lexical distinctions, even though it is their less dominant language (L2). French–Italian proficient bilinguals would, thus, use the Italian conceptual system, where there exists a linguistic distinction between two perceptual blue hues, even in exclusively perceptual tasks where language is not directly involved. These results would be in contrast with those of a recent study (Sinkeviciute et al., Reference Sinkeviciute, Mayor, Vulchanova and Kartushina2024) in Lithuanian–Norwegian bilinguals, which attested that color perception depended on language context.
-
3. Category advantage with a facilitatory effect
Categories can help identify cross-category color stimuli compared to within-category color stimuli, but cognitive factors, such as mnemonic and attentional allocations (Witzel & Gegenfurtner, Reference Witzel and Gegenfurtner2018), may also play a key role in the perceptual processing of color differences through a facilitatory effect. We anticipate that the two Italian BCTs for the two distinct hues determine the presence of a category advantage in the perceptual discriminations of colors belonging to distinct lexical categories for both Italian monolingual speakers and French–Italian bilingual speakers. This category advantage would be absent in French speakers’ color discrimination because they lack the linguistic distinction.
2. Methods
We conducted a speeded color discrimination task, conceptually replicating the design of Winawer and his colleagues (Winawer et al., Reference Winawer, Witthoft, Frank, Wu, Wade and Boroditsky2007) with some modifications.
This experimental protocol objectively investigates whether and how language influences color discrimination: by integrating RT measures and verbal distractors, and through simultaneous stimulus presentation, the design isolates the online effects of language on perception, minimizing memory demands and avoiding subjective judgments.
We assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration (https://www.wma.net/what-we-do/medical-ethics/declaration-of-helsinki/doh-oct2008/) of 1975, as revised in 2008.
2.1. Participants
We tested three types of observers: Italian monolingual speakers, French monolingual speakers and French–Italian proficient bilingual speakers (N = 66). They were all adults and had normal vision or corrected-to-normal vision and no color blindness: the Ishihara Pseudoisochromatic Plates (Clark, Reference Clark1924) and the barrage test (Diller et al., Reference Diller, Weinberg, Gordon, Goodkin, Gerstman and Ben-Yishay1974), used for the detection of attentional disorders, attested that each participant had the score higher than the set threshold in both tests. Before the experiment, participants completed a questionnaire concerning their linguistic history with information about their personal and professional experience, inspired by Li et al. (Reference Li, Zhang, Yu and Zhao2020). Bilingual participants also completed an additional survey about their linguistic habits and uses of their two languages (French and Italian): we used Birdsong’s model (Birdsong et al., Reference Birdsong, Gertken and Amengual2012). Then, they passed a linguistic placement test conceived by the AIL institution (Accademia Italiana di Lingua) to define their linguistic proficiency in Italian: this test is based on the AIL official syllabus for foreign speakers, containing 76 grammatical and lexical questions becoming more and more complex. The final score corresponds to one of the linguistic levels established by the CEFRFootnote 2.
In line with Li et al. (Reference Li, Zhang, Yu and Zhao2020) and Birdsong et al. (Reference Birdsong, Gertken and Amengual2012), we, thus, adopted a functional–sociolinguistic definition of “monolingual speaker” rather than a purely categorical one. In contemporary contexts, truly monolingual individuals, or those with no exposure to other languages, are extremely rare. Therefore, in this study, we defined “monolingual speakers” as individuals whose daily linguistic practices are overwhelmingly dominated by a single language that constitutes their primary and native linguistic environment that is acquired from birth and used across all major life domains (home, education, work, social interactions). Occasional passive exposure to foreign languages (e.g., through media or school instruction) does not disqualify participants as monolinguals, provided that such exposure does not result in functional bilingualism or influence perceptual categorization. This operational definition is crucial because in a globalized world, language contact is widespread and strict monolingualism is increasingly rare, affecting how control groups should be conceptualized and categorized in bilingualism research. This approach aligns with that of Li et al. (Reference Li, Zhang, Yu and Zhao2020), who emphasize language dominance and use patterns over categorical labels, and that of Birdsong et al. (Reference Birdsong, Gertken and Amengual2012), who propose profiling linguistic experience and proficiency to accurately characterize participants. Defining “monolingual” in this way ensures more precise group comparisons and acknowledges the spectrum of language experience rather than treating it as binary.
In this work, Italian monolingual speakers (N = 22) all had Italian as their L1, and they all were born in Italy and lived in Rome or its surroundings when they took part to the study. However, they all came from different regions of central Italy: this choice has been made to eliminate as much as possible the additional differences in the light blue lexicon (azzurro versus celeste) existing in several Italian regions (Del Viva et al., Reference Del Viva, Castellotti and Paramei2023; Paramei et al., Reference Paramei, Griber and Mylonas2018). In the survey, they completed a specific section on the three blues in Italian (azzurro versus celeste versus blu) whose results confirmed the assumptions of Paramei et al. (Reference Paramei, D’Orsi and Menegaz2016) on the respective status of the three blue words in Italian in the central regions.
The L1 of French monolingual speakers (N = 23) was French; except for two of them, all were born in France and lived in Paris or its surroundings when they participated in the experiment.
For the French–Italian bilingual speakers group (N = 21), we had a more heterogeneous group composed by Italian proficient speakers: the linguistic test they passed showed that they all had a C1–C2 level, based on the CEFR classification. Therefore, despite some differences in their linguistic experiences, all were considered French–Italian advanced bilinguals: they all lived in France when they took part in the experiment, but 15 of them had lived for more than 3 consecutive months in Italy previously and 13 of them for more than one year. Moreover, 14 of them were considered early bilinguals since they grew up in French–Italian mixed families, using the two languages in their daily life in different contexts. The other 7 were defined as late bilinguals since their language acquisition was more formal (Cenoz, Reference Cenoz2003, Reference Cenoz2009; Marinova-Todd et al., Reference Marinova-Todd, Marshall and Snow2000; Singleton & Ryan, Reference Singleton and Ryan2004; Singleton, Reference Singleton1989) and started after the age of 7, which is the age of acquisition limit to be considered a native speaker (Athanasopoulos, Reference Athanasopoulos2006, Reference Athanasopoulos2007; Athanasopoulos & Kasai, Reference Athanasopoulos and Kasai2008; Boroditsky, Reference Boroditsky2001; Cook et al., Reference Cook, Bassetti, Kasai, Sasaki and Takahashi2006). The mean age of their L2 onset (Italian) was 4.1.
They also affirmed being all fluent in English, learned at school, and 19 of them master another foreign language, which, for most of them, was Spanish or German but with a beginner-like level of proficiency.
In Table 1, we report some of the main characteristics of our participants’ survey, notably the average age, the gender, the educational level, the number of foreign languages spoken and the age of acquisition of their L2 or third language (L3).
Participants’ main characteristics: for each group of speakers, we compared the average age, the gender balance, the highest educational level, the number of foreign languages spoken and the age of acquisition of their L2 (L3 for bilinguals)
2.2. Material
All the stimuli were presented using version 2.0 of E-Prime software, Psychology Software Tools, Pittsburgh, PA (Schneider et al., Reference Schneider, Eschman and Zuccolotto2002), and the experiment lasted ≅ 30 minutes.
The monitor used for running both the experiments in France was a 24" ACER XF240H LCD display: the spatial resolution was 2160 x 1440 pixel, the refresh rate was 60 Hz, and the color depth was 8 bits per channel. The monitor used in Italy was a 23.8" Full View Full HD HUAWEI AD80HW Display: the spatial resolution was 1080p Full HD Pixels (2160 x 1440 pixel), the active signal resolution was 1920 x 1080 pixel, the refresh rate was 59,940 Hz, and the color depth was 8 bits per channel. The two monitors have been previously calibrated using the RGB values color calibration wizard on the database www.easyrgb.com; color rendering and gamma were also corrected using a X-Rite i1Display Pro colorimeter (SKU X-Rite i1D3 - i1D3DC + OEM) in the high-resolution mode using the display calibration software DisplayCAL2, powered by ArgyllCMS.
Maximum luminance was fixed (80 cd/m2), and the CIE 1931 chromaticity coordinates and luminance for the monitors’ primaries were R = 0.614, 0.356, 27.3; G = 0.286, 0.600, 60.1; B = 0.146,0.070, 9.4. The white point CIE-L*ab coordinates were L* = 100, a* = −0.002, b* = −0.014 (Yxy 100 0.3127 0.329; average ΔE*00 = 1.2; maximum ΔE*00 = 2.42), and all the stimuli were presented on a uniform grey background whose CIE-L*ab coordinates were L* = 77.43, a* = 0.01, b* = −0.01, which was metameric to D65 with a luminance of Y = 35 cd/m2 (RGB grey balance average absolute weighted ΔC ‘00 was calibrated at 0.25).
Twenty color patches ranging from light blue (azzurro) to dark blue (blu) have been created with Photoshop: Yxy color coordinates of the two extremes were 84, 0.214, 0.255 for stimulus 1 (the lightest blue) and 5.3, 0.154, 0.09 for stimulus 20 (the darkest blue), based on Winawer’s previous study. We generated the other 18 stimuli through an algorithm ensuring a perceptual linear equidistance between all the color patches with ΔE = 4.25 for adjacent colors, which differed in lightness and chromaticity (Figure 1). In the definition of our blue stimuli, we considered the results obtained by Paggetti et al. (Reference Paggetti, Menegaz and Paramei2016), who conducted two color-naming experiments where consensus, consistency, focal colors and centroids in Italian were analyzed, and we compared them with Winawer’s stimuli to ensure the consistency of color categories.
The 20 patches used as stimuli from the lightest blue (stimulus 1) to the darkest blue (stimulus 20).
Each patch was a square of 2.5 cm per side; three squares were presented at the same time on the screen in a triad: the one on the top was screen-centered and located at 3.1° above the bottom pair, which was 5.2° left and right from the y cartesian axis.
The speeded discrimination task was run in three different conditions: a simple task (no interference condition) and two double tasks, one with verbal interferences and one with spatial interferences. For the verbal interference condition, the words that need to be rehearsed were taken from the sets reported in Table 2: for the bilingual group, a mixed set, proportionally equally sorted, from the Italian and the French sets was created.
Verbal interference sets of stimuli for the three languages: 11 color words for French and Italian, and 22 color words, made up of French and Italian terms, for bilingual speakers. English translations are reported in parentheses
For the spatial interference condition, a 4x4 square grid composed by four random black squares, appeared on the screen. They were sorted by a set of 14 different grids differing one another in the location of only one black square.
Before the beginning of the experiment, observers had a practice block: we created additional grids and color words used for the interference conditions and 20 new color patches belonging to the green category, used partly as controls. The Yxy coordinates for the two extreme patches were: 151.532, 0.36037, 0.4788 for stimulus 1 (the lightest green) and 3.515, 0.22957, 0.41464 for stimulus 20 (the darkest green). Like for blue stimuli, green stimuli were perceptually equidistant one to the other, keeping constant the distance condition: two patches were considered near when they were two steps apart in the continuum of 20, whereas they were considered far when they were four or more steps apart in the continuum of 20.
2.3. Procedure
The experiments have been conducted between December 2021 and June 2022 in France (Paris) and Italy (Rome): each group of participants has been tested in their home country. Bilingual speakers participated in the experiments in France for their dominant language (French). In both cases, the room was quiet and darkened; participants seated at 60 cm from the screen, subtending 2° of visual angle.
All the groups received the instructions in their mother tongue; bilinguals were randomly divided into two subgroups: half of them received instructions in French (N = 10) and the other half in Italian (N = 11). They all signed a consent form for data collection (GDPR) and were recompensed for their participation with 15 Euros.
The experimental design was the same for the three groups, except for bilinguals whose verbal interference block was composed by both French and Italian words.
Participants were shown three color squares in a triad (Figure 2a): one of the two squares on the bottom (the match and the distractor) had the same color of the square on the top (the target). The task was totally perceptual: participants were asked to say which of the bottom squares was perceptually physically identical to the top square. The answer was manual by pressing a key on the keyboard: Z for left responses and N for right responses.
Trial event of the French speeded color discrimination task with all the three possible interference condition blocks (no interference, verbal interference and spatial interference).
Each color appeared equally often on the left and right in the triad and the same number of times as match, distractor and target. Each participant of each group was submitted to three blocks: one for each condition (no interference, verbal interference and spatial interference) made up of 136 trials.
The baseline block, or no interference block, consisted of a simple color discrimination task: each triad appeared at the center of the screen and rested until participant answered. If no answer was given, the stimulus disappeared after 2000 ms. Then, a fixation cross appeared for 1000 ms and the next triad appeared in the middle of the screen.
In the verbal interference block, the participant was shown a single color word, presented for 2000 ms, to rehearse, loudly or silently as they preferred, during the color discrimination task. Words’ stimuli were written in the participants native language, except for bilinguals whose verbal interference block was made of mixed stimuli in French and Italian: French and Italian color words were randomly sorted by the set of 22 color words, in equal proportion (Table 2). After the completion of eight trials, the recall was tested: a color word was presented on the screen, and the participant was asked to say whether it was the same color word shown just before or not.
The spatial interference block had the same design as the verbal interference block, but the color words were replaced by a 4x4 square grid. After the eight trials, a two-choice test was given: a new grid was presented, and the participant had to say whether it was the same grid shown before or not. The unmatched grids differed in the location of only one shaded square.
In the two interference blocks, 17 interference stimuli were used in each block and the participant had to press the SPACE bar when the second interference stimulus was the same as the previous one; no answer was required if the second stimulus was different. In the two interference blocks, 15% of the interference stimuli matched, for an average of 2,5 matched stimuli per participant; for the spatial interference, 14 different grids were created: six grids matched three times in each participants’ group and eight grids matched four times in each participants’ group. The distribution of the matched grids was random and counterbalanced between participants. In the Italian and French verbal interference blocks, each color word matched four or five times in each participants’ group: even in this case, the distribution was counterbalanced between participants and color words’ matches were random. Otherwise, in the bilingual verbal interference blocks, all the color words matched two times in the bilinguals’ group, except for six words (three Italian color words and three French color words) which matched four times; the distribution and the number of the match interference stimuli were counterbalanced between participants.
The order of the three blocks was also counterbalanced between participants in each group to avoid any block bias.
Figure 2 reports an example of an experimental trial showing the three experimental conditions.
At the end of the experiment, each participant completed a post-experiment to define his/her specific blue color boundary. The same 20 color patches covering the blue color space, from light blue (azzurro) to dark blue (blu), used in the speeded color discrimination task were presented twice, one by one, in random order. Participants were asked to classify each color manually, pressing a button on the keyboard, saying whether it belonged to dark or light blue category, for French speakers, and whether it was part of azzurro or blu category, for Italian and bilingual speakers. Each stimulus remained on the screen for a maximum of 2000 ms, followed by a fixation cross for 500 ms before the emergence of a new color stimulus (Figure 2c). Subjects had to answer as quickly and perceptually instinctively as possible.
3. Results
We analyzed two main effects: the color category effect and the distance effect. The color category effect is determined by the impact of the color category (light blue patches versus dark blue patches) on the performance based on the individual chromatic category boundary of each participant.
The main purpose was to examine whether there exists a correlation between the two color categories and the perceptual chromatic distance in the three experimental blocks.
The distance effect was based on the near/far condition analysis: two patches were considered near when they were two steps apart in the continuum of 20 with ΔE = 8.50, and they were considered far when they were four steps apart in the continuum of 20 with ΔE = 17. The color comparison was identified like a near-color comparison, when the distractor (the non-matching color square) was very similar to the other two or a far-color comparison, when the distractor appeared more different to the other two squares. Thus, color discriminations can be either easier, when the target and the distractor color squares were perceptually dissimilar, or harder, when the target and the distractor were perceptually closer.
This aspect allowed us to uncover how linguistic categories shape perception, revealing that their influence was stronger in more challenging perceptual tasks, such as finer color discriminations, than in simpler ones.
We conducted repeated measures ANOVA for RTs and accuracy for within-group and between-groups comparisons of the three groups of speakers tested, for both the color category and the distance effects. In all the analyses, we removed for each participant all the incorrect responses: for the French monolingual group, we excluded 5.5% of total answers; for Italian monolinguals, we excluded 4.2% of answers; and for the bilingual group, we excluded 4% of results.
3.1. Between-groups comparison
In the three-way ANOVA (I3 X C2 X D2) between-subjects comparison, the variable group was the between-subjects factor. The three repeated measures factors were: interference (I), made up of three levels (no interference block, verbal interference block and spatial interference block), color category (C) with two levels (dark blue – blu/bleu; and light blue – azzurro /bleu) and distance (D) made of two levels (near and far).
RT analyses showed several significant effects, whereas accuracy analyses were not statistically salient: we obtained a color category effect (F (1, 61) = 72.267, MSE = 751321.273, p < .001, η 2 = 0.029) and a distance effect (F (1, 61) = 343.536, MSE = 41930000, p < .001, η 2 = 0.161). Moreover, we detected an effect of group (F (2, 61) = 3.389, MSE = 211959.861, p = 0.040, η 2 = 0.055) and two interactions: a double color category X group interaction (F (2, 61) = 32.244, MSE = 10396.533, p < .001, η 2 = 0.026) and a triple interference X distance X group interaction (F (4, 122) = 3.952, MSE = 4036.917, p = 0.005, η 2 = 0.002).
We then performed several Bonferroni post hoc tests.
The color category effect distributions (pholm < .001) displayed that patches of the BLU category (MEAN = 821.007 ms, SD = 129.814 ms) were discriminated faster than AZZURRO category stimuli (MEAN = 876.170 ms, SD = 151.020 ms): MD = −62.618 ms, whereas the distance effect distributions (pholm < .001) showed that latencies for far stimuli (MEAN = 772.133 ms, SD = 123.151 ms) were 147.927 ms faster than latencies for near stimuli (MEAN = 920.016 ms, SD = 157.988 ms).
The Bonferroni post hoc for the group effect revealed that there was a significant effect in the comparison between bilinguals and French monolingual speakers (p = 0.047): the former was generally 101.925 ms slower than the latter.
Deeper analyses of the two interactions reported the following results: for the group X color category Interaction, we discovered that bilinguals and French monolingual speakers were faster (pholm < .001, MD = −98.662 ms for bilinguals and pholm < .001, MD = −109.139 ms for French monolinguals) in the discrimination of BLU stimuli compared to AZZURRO stimuli. Bilingual speakers were faster than French monolingual speakers in the discriminations of AZZURRO patches (pholm < .001, MD = −205.825 ms), but were 140.155 ms slower (pholm = 0.017) compared to Italian monolingual speakers’ discrimination. These results attested the presence of the AZZURRO category in the Italian lexicon, which affected the blue perception of Italian monolingual speakers and partially that of French–Italian bilinguals.
Furthermore, the RT distributional analyses of the triple interference X distance X group interaction showed that in some specific experimental conditions, group was a factor of distinction. A global trend was detected in far stimuli discriminations in the no interference block (p = 0.034), whereas in the verbal interference block, near stimuli discriminations (p = 0.024) and far stimuli discriminations (p = 0.05) depended on group. Nevertheless, distance was the only factor exercising a significant effect on the two other factors: far and near stimuli were always statistically differently discriminated in the three participant groups in the three experimental interference conditions (p < .001). This trend was similar for the three groups of speakers. Figures 3a and 4a report RTs and accuracy for the distance effect, respectively, whereas Figures 3b and 4b illustrate RTs and accuracy for the color category effect for the three groups, without any distinction of condition.
(a) Mean RTs (in ms) of the color category effect (dark blue versus light blue) with standard error (±1 SE) bars for the three groups (French, Italian and bilingual speakers). (b) Mean RTs (in ms) of the distance effect (near versus far stimuli) with standard error (±1 SE) bars for the three groups (French monolinguals, Italian monolinguals and bilingual speakers).
(a) Mean accuracy (%) of the color category effect (dark blue versus light blue) with standard error (±1 SE) bars for the three groups (French monolinguals, Italian monolinguals and bilingual speakers). (b) Mean accuracy (%) of the distance effect (near versus far stimuli) with standard error (±1 SE) bars for the three groups (French monolinguals, Italian monolinguals and bilingual speakers).
This between-subjects comparison analysis has allowed us to explore the presence of a possible interaction between the factors we analyzed and the three groups we tested. The more salient effect was a global slowness of bilingual speakers compared to the two groups of monolinguals in almost all the conditions: they probably adopted the color categorization of their L2 (Italian), but their performances were not exactly the same as the Italian monolingual ones. In Italian, the color category identified by the color word azzurro is strongly defined and stable, with an impact on perceptual discriminations between blue hues with different lightness degrees.
Therefore, despite some similarities between bilinguals and French monolingual speakers for BLU category stimuli, bilinguals’ color discrimination deviated from their L1 conceptual system, adopting the perceptual categorization of their L2 lexicon. However, a universal trend across the three groups was maintained in perceptually distant and close patches, independently of the BLU and AZZURRO categories: far colors pairs were always discriminated faster and more accurately than near colors pairs. Discrimination was, thus, facilitated by perceptual distance.
3.2. Within-group comparison
For the French–Italian bilingual group, the three-way ANOVA (I3 X C2 X D2) performed for RT showed both a color category effect (F (1, 17) = 42.678, MSE = 12504.124, p < .001, η 2 = 0.148) and a distance effect (F (1, 17) = 142.76, MSE = 8707.322, p < .001, η 2 = 0.346). Moreover, a marginal interaction between interference and distance was detected (F (2, 34) = 2.954, MSE = 4329.848, p = 0.066, η 2 = 0.007): responses for far stimuli were globally more rapid than for near stimuli in all the experimental blocks, without any distinction of category (BLU vs AZZURRO).
Bonferroni post hoc tests have been conducted for the distributions of the two effects: the color category effect (pholm < .001) displayed that BLU stimuli (MEAN = 830.432 ms, SD = 166.94 ms) were discriminated 99.410 ms faster than AZZURRO stimuli (MEAN = 923.237 ms, SD = 191.164 ms). On the other hand, for the distance effect (pholm < .001), we obtained similar results than for the two other groups: far stimuli (MEAN = 804.003 ms, SD = 161.939 ms) were discriminated 151.722 ms faster than near stimuli (MEAN = 949.062 ms, SD = 195.065 ms).
ANOVA results for accuracy did not show any statistically significant effect for both the color category effect and the distance effect: the two color hues were processed with similar degrees of accuracy from bilinguals in the three experimental blocks (BLU patches: MEAN = 83.1%, SD = 2.1; AZZURRO patches: MEAN = 80.76%, SD = 1.83). However, they were more accurate for far stimuli than for near stimuli (pholm < .001): far stimuli: MEAN = 69.29% (SD = 1.72); near stimuli: MEAN = 87.62% (SD = 1.63).
The most relevant result we obtained for the bilingual group concerned the dark blue patches that were discriminated more rapidly than light blue patches. This effect was probably due to the BLU focal color of bilinguals, which simultaneously overlaps the blue focal color of their L1 (French) and that of their L2 (Italian): ΔE*00 = 1.1. However, there was no distinction between the interference blocks, which means that the color of the patches was probably processed without any online direct influence of language on perception.
The two other within-group analyses concerned French monolingual speakers and Italian monolingual speakers, performing several three-way ANOVA (I3 X C2 X D2) for RTs and accuracy.
Results for RTs of French monolingual speakers showed that the two effects were statistically significant: for the color category effect F (1, 21) = 55.013, MSE = 14290.244, p < .001, η 2 = 0.193, and for the distance effect F (1, 21) = 79.121, MSE = 13343.252, p < .001, η 2 = 0.259.
We explored the distribution of the color category effect (pholm < .001) for RTs through Bonferroni post hoc tests, which revealed that dark blue patches (MEAN = 762.136 ms, SD = 143.89 ms) were discriminated 109.139 ms faster than light blue patches (MEAN = 859.272 ms, SD = 179.263 ms). On the other hand, the Bonferroni post hoc analysis for the distance effect revealed that responses for far stimuli (MEAN = 743.741 ms, SD = 140.503 ms) were 126.475 ms faster than those for near stimuli (MEAN = 870.216 ms, SD = 180.676 ms), without any distinction of category (dark blue versus light blue hues).
However, ANOVA for accuracy only revealed a color category effect (F (1, 21) = 11.260, MSE = 88.914, p = 0.003, η 2 = 0.143), wherein Bonferroni post hoc tests displayed more accurate responses (pholm = 0.003) of about 5.508% for dark blue patches (MEAN = 84.126%, SD = 9.423%) compared to light blue patches (MEAN = 78.618%, SD = 12.287%).
The within-subjects comparison for French monolingual speakers has shown that dark blue patches were better and faster discriminated than light blue patches without any distinction of experimental block (with or without interferences). This effect was probably due to the blue focal color, which, in French, corresponds to a darker hue in the blue region of the color spectrum and which is highly stable. Moreover, as we expected, there exists a facilitatory effect for far stimuli, easier to detect, compared to near stimuli, since their perceptual differences are finer.
For Italian monolingual speakers, the three-way ANOVA we conducted for RTs demonstrated an effect of color category (F (1, 21) = 4.628, MSE = 5674.206, p = 0.043, η 2 = 0.007), an effect of distance (F (1, 21) = 127.539, MSE = 14603.876, p < .001, η 2 = 0.514) and a significant triple interaction between interference, color category and distance (F (2, 42) = 3.886, MSE = 2214.67, p = 0.028, η 2 = 0.005).
Bonferroni post hoc tests conducted to explore distributions showed the following results: for the color category effect (pholm = 0.043), we discovered that, contrary to French monolingual and bilingual speakers, dark blue stimuli (MEAN = 842.146 ms, SD = 119.921 ms), labeled with the color word blu, were discriminated 19.946 ms slower than light blue stimuli (MEAN = 816.604 ms, SD = 115.243 ms), named by the color word azzurro. However, for the distance effect (pholm < .001), far stimuli (MEAN = 744.057 ms, SD = 94.698 ms) were again discriminated 167.99 ms faster than near stimuli (MEAN = 912.047 ms, SD = 143.605 ms).
Lastly, to analyze the RT distributions of the interaction between interference, color category and distance, we performed additional Bonferroni post hoc tests, combined with simple main effects. Results showed that the color category factor had an impact especially on the discrimination of far trials in the no interference block (F (1) = 7.887, MSE = 13050.801, p = 0.011) and in near trials in the spatial interference block (F (1) = 5.210, MSE = 20948.982, p = 0.033). In fact, in the no interference block, responses for dark blue far stimuli, belonging to the Italian BLU category (MEAN = 772.994 ms, SD = 150.1 ms), were 34.445 ms slower than for light blue far stimuli, corresponding to the Italian AZZURRO category (MEAN = 738.549 ms, SD = 136.216 ms). Latencies in the spatial interference block were 43.64 ms slower for dark blue near stimuli (MEAN = 934.981 ms, SD = 179.291 ms) than for light blue near stimuli (MEAN = 891.342 ms, SD = 145.375 ms).
These results revealed a trend partially different from French monolingual speakers where light blue patches were discriminated faster than dark blue patches, corroborating the hypothesis that in Italian there exists the AZZURRO color category. Like the BLU category, the AZZURRO category is considered salient, stable and basic. Moreover, the color category had a concrete impact especially on trials of the no interference block and in trials of the spatial interference block, which are the two blocks supposed to be mostly impacted by language processing, through distinct lexical color categories. In the verbal interference block, the advantage is, in fact, expected to be disrupted because of the double presence (online and offline) of linguistic processes.
Like French monolingual speakers, the repeated-measures ANOVA for accuracy reported only a color category effect (F (1, 21) = 12.841, MSE = 27.631, p = 0.002, η 2 = 0.153) but with an inversed trend. Descriptive statistics and Bonferroni post hoc tests (pholm = 0.002) revealed that Italian speakers were 3.279% more accurate for patches belonging to the AZZURRO color category (MEAN = 94.209%, SD = 2.613%) compared to patches belonging to the BLU color category (MEAN = 90.930%, SD = 4.039%). This slight discrepancy attests that all the patches were globally processed in the same way: a real distinction is absent due to the presence of two color categories simplifying both the discriminations.
At the end of the experiment, we run a post-experiment to define the blue categories boundary: for each subject, we identified the transition point based on their patches’ categorization (dark blue or light blue), representing their individual color boundary. If the transition was ambiguous or fell between two stimuli, we considered the slower RT to disambiguate the boundary since colors that are closest to the boundary tend to be categorized more slowly in simple classification tasks (Bornstein & Korda, Reference Bornstein and Korda1984). In some cases, disambiguation was particularly hard to solve since the transition point was not sharp enough: we used the two alternative forced choice (2AFC) paradigm to detect the color boundary, defining the threshold with the stimuli categorized to 50% (Hanley & Roberson, Reference Hanley and Roberson2011; Roberson et al., Reference Roberson, Hanley and Pak2009; Witzel & Gegenfurtner, Reference Witzel and Gegenfurtner2013). The locations of the BLU/AZZURRO category boundary in Italian and bilingual speakers and that of light blue/dark blue boundary for French speakers were similar: the average color boundary fell between patch 11 and patch 12 for Italian monolingual speakers (MEAN = 11.74, SD = 0,07), between patch 12 and patch 13 for French monolingual speakers (MEAN = 12.56, SD = 0,10) and between patch 11 and patch 12 for bilingual speakers (MEAN = 11.23, SD = 0,09).
4. Discussion
In this study, we considered lexical distinctions as one of the main factors responsible for the restructure of the bilingual cognitive framework in color perception and categorization (Cook, Reference Cook2014, Reference Cook2002, Reference Cook2003; Green, Reference Green1998; Malik-Moraleda et al., Reference Malik-Moraleda, Mahowald, Conway and Gibson2023; Matusevych et al., Reference Matusevych, Beekhuizen and Stevenson2018; Pavlenko, Reference Pavlenko1999, Reference Pavlenko2005; Pavlenko et al., Reference Pavlenko, Jarvis, Melnyk and Sorokina2017). The key issue focused on how language shapes perception in non-linguistic domains, specifically among individuals proficient in two distinct linguistic systems with different ways of categorizing the blue area of the color spectrum. Their distinct lexical segmentations to processing visual information directly influenced how these categories were represented across several cognitive domains, highlighting the effect of bilingualism (Bialystok, Reference Bialystok2017; De Groot, Reference De Groot2011; Green & Abutalebi, Reference Green and Abutalebi2013; Grosjean, Reference Grosjean1989; Houwer, Reference Houwer2009; Kroll & Bialystok, Reference Kroll and Bialystok2013; Slobin, Reference Slobin1996).
We found that proficient French–Italian bilingual speakers’ perceptual behavior in color discrimination was influenced by both their L1 and L2, resulting in an in-between idiosyncratic setting. Color discrimination of dark blue and light blue hues, belonging to two distinct categories in Italian but not in French, reflects French monolinguals’ performance for some aspects (e.g., BLU hues are globally discriminated faster than AZZURRO hues), but are similar to Italian monolinguals’ performance for others (e.g., AZZURRO hues are discriminated faster than BLU hues compared to French monolinguals). At the same time, their responses are overall slower than the two monolingual groups due to inhibition effects and cognitive conflict.
Moreover, far stimuli were discriminated faster than near stimuli from the three groups, revealing that perceptual distance has a facilitatory effect, notably when the visual stimuli belong to different categories.
Our study builds on previous research whose results have demonstrated that, in bilingual speakers, L2 might influence the use and the color prototypes of L1 color words through semantic shifts. Bilinguals’ use of L1 color terms can, in fact, be different from the corresponding monolingual groups (Erwin et al., Reference Erwin, Lerner, Wilson and Wilson1961): they can shift L1 color terms’ prototypes toward L2 prototypes (Caskey-Sirmons & Hickerson, Reference Caskey-Sirmons and Hickerson1977) or they can even modify the color-naming behavior of their L1 based on the distinctions made in their L2 when their systems diverge, pushed by informative purposes (Jameson & Alvarado, Reference Jameson and Alvarado2003).
Globally, bilingual linguistic and cognitive behaviors do not vary with the language system used; they are rather influenced by each language categorizations, identified by specific lexical labels, which, in turns, impact various aspects of the perceptual process (Ameel et al., Reference Ameel, Storms, Malt and Sloman2005; Malt & Sloman, Reference Malt and Sloman2003; Roberson et al., Reference Roberson, Davidoff, Davies and Shapiro2005).
This plasticity typical of the bilingual mind is determined by a well-trained inhibitory control: bilinguals must constantly perform a set of perceptual tasks where they choose between distinct systems with different categorizations, causing the emergence of a linguistic–cognitive conflict (Bialystok et al., Reference Bialystok, Poarch, Luo and Craik2014; Bialystok & Martin, Reference Bialystok and Martin2004). When one system possesses a finer discrimination compared to the other, it cannot be ignored nor suppressed, even in extra-linguistic tasks. Therefore, better interference suppression, connected to an inhibitory control advantage and faster RT in both tasks requiring and not requiring interference suppression, is one of the main bilingualism advantages.
One possible explanation of our results can be linked to enhanced inhibition processes.
As several studies have already shown, various cognitive processes benefit from the bilingual experience, notably inhibition, attention (Costa et al., Reference Costa, Hernández and Sebastián-Gallés2008; Fan et al., Reference Fan, McCandliss, Sommer, Raz and Posner2002; Grundy et al., Reference Grundy, Chung-Fat-Yim, Friesen, Mak and Bialystok2017) and executive control (Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004, Reference Bialystok, Craik and Luk2008; Fan et al., Reference Fan, Gu, Guise, Liu, Fossella, Wang and Posner2009; Marzecová et al., Reference Marzecová, Asanowicz, Krivá and Wodniecka2013). In our study, the main cognitive ability that benefits bilingualism is inhibition, where bilingual speakers must constantly suppress one language to operate in the other. This control operates at two levels (Green, Reference Green1998): globally, when selecting one language over another, and locally, when inhibiting non-target lexical items. These inhibitory demands extend beyond language to general cognitive tasks (Bialystok, Reference Bialystok2001, Reference Bialystok2009), mirroring our findings.
Our French–Italian bilinguals behaved similarly to French monolinguals, showing faster and more accurate discrimination of BLU category stimuli than AZZURRO stimuli. This suggests that the shared basic status of dark blue in both French (BLEU) and Italian (BLU) results in a stable, dominant category in the bilingual mind. The overlapping focal colors reinforce perceptual saliency.
However, bilinguals also outperformed French monolinguals on AZZURRO stimuli, indicating influence from the Italian-specific lexical distinction. Still, Italian monolinguals were fastest on AZZURRO, highlighting the saliency of the 12th Italian basic color term (Paggetti et al., Reference Paggetti, Bartoli and Menegaz2011; Valdegamberi et al., Reference Valdegamberi, Paggetti and Menegaz2015). In Italian, BLU and AZZURRO form distinct, well-defined categories based on lightness, supported by strong boundaries and shared naming norms. However, in some Italian regions where celeste is more common, boundaries may shift further: future studies should replicate this paradigm regionally to explore intra-language variation.
Our findings contrast with those of Sinkeviciute et al. (Reference Sinkeviciute, Mayor, Vulchanova and Kartushina2024), who tested Lithuanian–Norwegian bilinguals and found only a temporary categorical effect with verbal interference in Lithuanian (which also distinguishes blues lexically). Three key differences may explain this: first, our bilinguals had Italian (with two blue terms) as their L2, whereas in Sinkeviciute et al. (Reference Sinkeviciute, Mayor, Vulchanova and Kartushina2024), it was the L1. This reversal is crucial for understanding cross-linguistic transfer and whether L1 or L2 exerts stronger perceptual influence: in fact, our results suggest that even a less dominant L2 can restructure L1 categorization when it offers finer distinctions, hinting at lasting perceptual effects even in non-verbal tasks. This opens promising lines of inquiry into how acquisition order shapes category learning and the neural dynamics of language switching and control (Green & Abutalebi, Reference Green and Abutalebi2013; Kroll & Bialystok, Reference Kroll and Bialystok2013).
Second, the interference manipulation differed. While Sinkeviciute et al. (Reference Sinkeviciute, Mayor, Vulchanova and Kartushina2024) used number strings (like Winawer et al., Reference Winawer, Witthoft, Frank, Wu, Wade and Boroditsky2007), our participants rehearsed color words during the task. These two forms of interference differ in the linguistic mechanisms they engage (Nedergaard et al., Reference Nedergaard, Wallentin and Lupyan2023). Number rehearsal taxes general verbal memory but does not activate semantic color categories; color word rehearsal, by contrast, directly stimulates the lexical-semantic system tied to color processing. This likely increases top-down linguistic modulation and creates competition at the semantic level, which may suppress perceptual category effects due to interference from the same representational reservoir. Thus, we argue that color word interference better captures language’s role in perception than number string tasks. Third, in our study, we compared two typologically similar languages, whereas Sinkeviciute and colleagues compared two typologically distant languages.
We also introduced a mixed-language interference block (French and Italian), simulating real-life bilingualism where both languages are co-activated: this design enhances ecological validity, reflecting continuous cross-language interaction in lexical access (Grosjean, Reference Grosjean2001; Kroll & Bialystok, Reference Kroll and Bialystok2013). It allowed us to probe how bilinguals manage dynamic competition and inhibition when both languages are simultaneously triggered, suggesting that in color word rehearsal, a stronger top-down influence of language on perception can potentially limit the CP effects because of conflicting access to the same semantic reservoir.
Another general trend we observed is that fine perceptual discriminations (i.e., between similar hues) were generally slower than broader ones across all blocks (no interference, spatial, verbal), due to the closer perceptual distance. However, Italian monolinguals inverted this pattern under verbal interference: near stimuli were identified more quickly. This may reflect the active involvement of Italian color terms (blu versus azzurro) during the task. Their advantage for light blue stimuli was present only in the no interference and spatial interference blocks but disappeared in the verbal interference block, where the rehearsed color words competed for semantic access.
The absence of this effect in bilinguals supports the idea that their perceptual system is not a sum of two monolingual systems, but a hybrid, uniquely structured by dynamic language interaction. Still, group heterogeneity (i.e., gender balance or type of bilingualism) may have influenced our outcomes, representing a limitation. Future studies should use finer-grained participant categorization and employ both naming and perceptual tasks to better define bilingual blue category boundaries. This would allow deeper exploration of how lexical distinctions manifest perceptually in bilinguals versus monolinguals. In fact, a crucial methodological issue in studies comparing monolinguals and bilinguals concerns the operational definition of “monolingual speaker.” In contemporary contexts where true linguistic isolation is increasingly uncommon, even minimal foreign language exposure among individuals categorized as “monolingual” can introduce subtle cross-linguistic influences into their perceptual categorization. For example, in our study, French monolinguals, although typically showing stable categorization around the darker blue focal color bleu, may exhibit slight shifts toward lighter hues due to exposure to other languages or cultural contexts, while Italian monolinguals, despite demonstrating clear lexical distinctions between blu and azzurro, may display variability linked to regional differences (e.g., the use of celeste) or contact with English, which uses a single term “blue.” Such heterogeneity within monolingual baselines can blur contrasts with bilingual participants, diminishing observable differences in RTs and category boundary placements and ultimately leading to an underestimation of the real impact that bilingual lexical structures have on perceptual organization. Defining and ensuring the stability and homogeneity of monolingual baselines is therefore essential for accurately interpreting the magnitude of cross-linguistic effects. To achieve this goal, future research should incorporate detailed measures of language exposure and dominance, allowing for better statistical control of these variables and more precise comparisons across groups.
Meanwhile, an additional methodological consideration concerns the heterogeneity within the French–Italian bilingual group, which included both early bilinguals, raised in mixed-language families, and late bilinguals, who acquired Italian through formal education and periods of immersion. This variability could potentially influence individual differences in lexical activation patterns, perceptual categorization and inhibitory control, thereby introducing noise into group-level analyses. For instance, early bilinguals might exhibit more integrated and automatic use of Italian lexical categories such as blu and azzurro, whereas late bilinguals may rely more on learned distinctions or display incomplete category restructuring. Similarly, differences in the length and intensity of immersion experiences could shape the degree of perceptual adaptation to Italian color categories. Nevertheless, the reliability of our findings is supported by several factors: all bilingual participants demonstrated advanced proficiency (C1–C2) in Italian, shared similar levels of daily language use and showed consistent categorical shifts toward Italian lexical distinctions in their RTs and boundary placements. These converging behavioral patterns across a heterogeneous group indicate that the observed effects are robust and not limited to a specific bilingual profile, thereby reinforcing the validity and generalizability of our findings.
To conclude, our study represents an additional proof that bilingualism is a complex, multidynamic and multidimensional reality. Giving a fixed definition is almost an impossible task. Each individual mastering two or more languages has a specific linguistic history depending on the modalities, the contexts, the routes taken to learn all his/her languages and more other factors, such as age, environment and social groups, in the use of the simultaneous linguistic systems. Thus, bilingualism is not a categorical variable (Luk & Bialystok, Reference Luk and Bialystok2013) definable by theoretical frameworks, although the secular still unsolved debate between the monolingual (Bloomfield, Reference Bloomfield1984; Kent & Bolling, Reference Kent and Bolling1934; Thiery, Reference Thiery, Gerver and Sinaiko1978) and the holistic views (Deprez, Reference Deprez1994; Kohl et al., Reference Kohl, Beauquier-Maccotta, Bourgeois, Clouard, Donde, Mosser, Pinot, Rittori, Vaivre-Douret, Golse and Robel2008; Lüdi & Py, Reference Lüdi and Py2002) tries to define its fixed criteria of functioning. Additional examples of the experimental design and further graphs of statistical analysis can be find in the Supplementary Material.
5. Conclusion
The primary focus of this research is to explore the profound impact of language categories on color perception. We believe our findings offer a significant contribution to multiple research domains, spanning from color vision to cognitive linguistics, especially in the study of monolingual and bilingual cognition. If our results confirm the hypothesis that bilinguals are not merely two monolinguals in one mind (Grosjean, Reference Grosjean1989), but rather a unique blend with characteristics from both monolingual speakers and individual-specific traits, we acknowledge that some of the most pressing questions remain unresolved.
Future studies are critical to fully unravel the influence of language on color perception, the role of color semantics and the broader debate surrounding the bilingual cognitive advantage. Therefore, additional comprehensive research across diverse fields is essential to arrive at more definitive conclusions. Refining bilingual groups, with respect to proficiency levels, or balancing gender participation in experiments, for example, could help eliminate biases. Furthermore, identifying specific blue color foci among subjects and reproducing similar studies in others are crucial steps for future research.
Expanding the scope to include additional speaker groups, neurophysiological analyses and further exploration of color perception phenomena will deepen our understanding of how cross-linguistic differences shape perception – particularly in the blue spectrum – alongside new investigations into other parts of the color spectrum.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1366728925100710.
Data availability statement
The data that support the findings of this study, as well as the experimental design, are available under request to the corresponding author.
Acknowledgements
We would like to express our gratitude to each participant in the experiment, without whom this study would not have been feasible.
Competing interests
Both authors declare that they have no competing interests.
Open access
