Running head: The influence of grammatical gender in Polish, French and Dutch

(1)

The influence of grammatical gender: a study of Polish, French and Dutch speakers.

Maria Kościelniak S4102789 Ludovic Germaneau S3826791

MA thesis, Departments of European Linguistics and Frisian Language and Culture, Faculty of Arts, Rijksuniversiteit Groningen

Thesis supervisor: Dr. B. Hollebrandse Second reader: Dr. H. Loerts

24/06/2020

(2)

Contents

Abstract ... 5

The influence of grammatical gender: a study of Polish, French and Dutch speakers. ... 6

1. Background ... 7

1.1 Linguistic relativity ... 7

1.2 Gender in the linguistic relativity studies ... 8

1.3 Natural objects/ artefacts ... 10

1.4 The similarity and gender and the sex and gender hypothesis ...11

1.5 Polish... 13 1.6 Dutch ... 15 1.7 French ... 17 1.8 Statement of purpose... 18 2. Method ... 20 2.1 Research Participants ... 20 2.2 Materials ... 21 2.3 Procedures ... 23

2.4 Design and Analyses ... 24

Polish and French ... 24

Dutch ... 24

(3)

Polish and French ... 25

Polish, French and Dutch ... 26

3. Results ... 26

3.1 Polish participants ... 26

Ascribing masculine/feminine features in congruence with grammatical gender .... 26

Answers congruent with the grammatical gender in natural objects and artefacts ... 28

Ascribing masculine features to artefacts and feminine features to natural objects . 30 Influence of other factors ... 32

3.2 French participants ... 33

Answers congruent with the grammatical gender in natural objects and artefacts ... 35

3.3 Comparison of Polish and French results ... 40

Ascribing masculine features to artefacts and feminine features to natural objects . 41 3.4 Dutch participants ... 43

Identifiable pattern for ascribing masculine/feminine features ... 43

(4)

4. Discussion ... 49

4.1 Polish participants ... 49

4.2 French participants ... 50

4.3 Comparison of Polish and French results ... 52

4.4 Dutch participants ... 52

4.5 Summary of the findings ... 53

4.6 Similarity and gender and sex and gender hypothesis ... 55

4.7 Limitations of the study ... 56

5. Conclusion ... 57

References ... 60

(5)

Abstract

Linguistic relativity theory assumes language to influence the way that people perceive the world. So, what is the influence of grammatical gender on perception? This study intended to investigate this question for three different languages: Polish, French, and Dutch. Studies suggest that speakers of languages with grammatical genders tend to perceive objects with features in congruence with their grammatical gender. They also suggest that, in these languages, there is a supra-language gender categorisation of objects depending on whether they are natural objects or man-made artefacts. This research intended to replicate Izabella Haertlé’s study on French and Polish speakers and added a natural gendered language, Dutch, to the comparison. Two tasks (voice choice and ascribing stereotypical adjectives) were constructed in the form of an online questionnaire. The results of these experiments show that grammatical gender influences French and Polish speakers, both languages with grammatical gender, while Dutch speakers do not seem to ascribe genders in any recognisable pattern, confirming that the perception of nouns is influenced by grammatical gender. However, the results do not seem to support the hypothesis that there is supra-language gender categorisation in gendered languages as feminine features were not more often ascribed to natural objects and masculine features were not more often ascribed to man-made artefacts.

Keywords: Linguistic Relativity, Sapir-Whorf hypothesis, grammatical gender, French,

Dutch, Polish, natural gender language, voice choice task, stereotypically masculine/feminine features, natural objects/ artefacts

(6)

The influence of grammatical gender: a study of Polish, French and Dutch speakers.

The weak version of the linguistic relativity theory assumes language to influence the way that people perceive the world (Brown & Lenneberg, 1954). One of the focuses of linguistic relativity studies is the examination of gender, during which the researchers try to find the answer to questions such as: What is the influence of grammatical gender on perception? How do two people, speaking two different languages perceive the same object that in one language has feminine and in the other masculine grammatical gender? How do different gender systems affect people’s perception? Is there one, supra-language categorisation that makes people perceive objects in a certain way?

This study intends to investigate three different languages closely: Polish- a Slavic language with three grammatical genders, French- a Romance language with two grammatical genders and Dutch- a Germanic language with natural two-way gender system. Our research questions are as follows:

RQ1: Do Polish/French speakers ascribe masculine/feminine features to presented objects in congruence with the grammatical gender of their spoken language?

RQ2: Do Polish/French/Dutch speakers ascribe feminine features to natural objects and masculine features to artefacts?

RQ3: Is there an identifiable pattern in which speakers of Dutch, a natural gender language, ascribe masculine/feminine features to presented objects?

In order to answer the research questions two tasks, in a form of an online questionnaire were constructed: first assigning male/female voice to presented objects, second ascribing three adjectives from a list that best describe displayed nouns.

(7)

The paper is constructed as follows: the background section provides a literature review on linguistic relativity theory and studies on the influence of grammatical gender, linking it to research questions and hypothesis for this study; description of the participants, materials, procedures, design and analyses are introduced in the methods, followed by results where outcomes, tables and graphs are presented, subsequently, the discussion proposes the interpretation of the outcomes and finally conclusions including the summary of the results and statement of possible directions for further research.

1. Background 1.1 Linguistic relativity

Linguistic relativity theory, also named Sapir-Whorf hypothesis, claims that the way people perceive the world is affected by the language they speak. Brown & Lenneberg (1954) distinguished a weak (language influences the way people think) and a strong (language determines the way people think) version of this hypothesis. The latter was widely rejected by linguists, but the linguistic relativity concept itself inspired broader research in the field. The impact of language, related to the weak version of the hypothesis has been found on the perception of space (Boroditsky & Gaby, 2010), time (Nuñez, Cooperrider, Doan & Wassmann, 2012), quantities (Everett & Madora 2012), colour (Berlin & Kay, 1969), and object and substances (Lucy, 1996), among other domains. For instance, Bylund & Athanasopoulos (2017) have discovered that Spanish-Swedish bilinguals categorise time differently depending on the language context; the same can be said for Mandarin-English speakers (Boroditsky, 1999). Similarly, the way people perceive colours depends on the language and the culture attached to it (Athanasopoulos, 2009). Furthermore, languages can have an effect on human gaze behaviour as shown by Goller, Lee, Ansorge & Choi (2017) in a comparative study on Korean and German speakers; on attention and

(8)

information processing (Rhode, Voyer, & Gleibs, 2016); and even on the perception and memory of events (Boroditsky, Ham, & Ramscar, 2002).

1.2 Gender in the linguistic relativity studies

The perception of gender has also been shown to be influenced by language as Everett (2013, p. 245) states in his book “gendered language can prime or differentially induce certain kinds of thought”. For instance, people exposed to grammatical gender categories are, in certain circumstances, susceptible to consider that some nonlinguistic units have features of the gender that they are assigned to (Everett, 2013). Prewitt-Freilino et al. (2012) distinguished three language groups on the basis of their gender system: grammatical gender languages which usually belong to Germanic (German), Romance (French), Slavic (Polish), Semitic (Hebrew) and Indo-Aryan (Hindi) language families; natural gender languages for instance West-Germanic (English) and Northern Germanic (Scandinavian) languages and genderless languages such as Turkic (Turkish), Bantu (Swahili), Uralic (Finish), Iranian (Persian) and Sinitic (Chinese). In case of gendered languages, the nouns have feminine, masculine, or sometimes neuter gender. The same gender markers as the nouns they refer to are visible in the dependent forms, for instance pronouns and adjectives. In natural gender languages the majority of the nouns does not have grammatical gender and gender is differentiated by pronouns. In genderless languages, the noun system does not distinguish grammatical gender at all.

Linguistic relativity researchers examining the influence of grammatical gender use different tasks to test their hypothesis. Samuel, Cole & Eacott (2019) describe the most common ones used: properties judgment task, where participants are asked to invent adjectives describing inanimate objects and later a different group is asked to choose male or female voice to those objects; voice choice task in which participants assign male/female voice to presented objects; sex

(9)

assignment task where participants are asked to assign a sex or a name to an object; EAST- extrinsic affective Simon task in which participants respond to words under two conditions with two keys,

in the first one each key corresponds to one colour, in the second condition each key corresponds to one sex; priming task where participants are asked to determine the gender of a word after seeing either the words “man” and “woman”, symbols for male/female, pictograms of man or woman or a definite article indicating a gender; similarity task in which participants rate the similarity between images showing males/females and items; object-name memory association task where participants are shown pairs of objects and names given to them (congruently with the grammatical gender of these objects or not, e.g. “lamp Kate”) and after a while when shows the objects alone they have to provide a name that was assigned earlier. Overall, the biggest effect of grammatical gender was found in voice choice task and the lowest in object-name memory task.

Some studies have found an effect of grammatical gender only in nouns depicting animate objects and no effect in nouns depicting inanimate objects. In some cases, the grammatical gender does not seem to reach the semantic content of nouns depicting inanimate objects. Samuel, Cole & Eacott (2019) in their semantic review of linguistic relativity in the context of grammatical gender analysed the results of 43 studies ranging from 1990 to 2018. They have found the overall higher effect of grammatical gender in studies that used animate rather than inanimate objects in their experiments.

Sera et al. (2002) in task of assigning male and female voices to inanimate objects, showed that the effect of grammatical gender is mostly visible in languages that have two grammatical genders (in this case monolingual French and Spanish children and adults) as opposed to languages that have three grammatical genders (in this study monolingual German children and adults). This was again confirmed by Samuel, Cole & Eacott (2019) in their overview of linguistic relativity

(10)

studies as they found that the effect of grammatical gender was present more often in studies on languages with two, rather than three grammatical genders. However, recent study of Haertlé (2017) on Polish (that also has three grammatical genders) and French, demonstrated a clear impact of grammatical gender in both languages. Participants’ tasks in this study included assigning male and female voice to objects that were supposed to be actors in an upcoming animated movie and also ascribing three adjectives from a list to presented nouns. Moreover, a study by Rączaszek-Leonardi (2010) on Polish and Italian indicated a stronger effect in Polish than in Italian, that has two grammatical genders. In this experiment a group of participants was asked to assign adjectives to presented objects (either in a form of photographs or pictures) followed by a different group deciding whether the adjectives are more masculine or feminine.

The difference between the effect of grammatical gender on perception for children and for adults participants has been proven, namely in case of children weak or no effects have been observed. Sera et al. (2002) have proven that at the age eight the impact of grammatical gender on the notion of objects becomes evident. Studies by Bassetti (2007) and Nicoladis & Foursha-Stevenson (2012) have supported these findings. It may be the experience and familiarity with grammatical gender that allows biological sex connotations with items to appear (Samuel, Cole & Eacott, 2019).

1.3 Natural objects/ artefacts

The opposite view from linguistic relativity is universalist perspective, which claims that thought is independent of language (Pinker, 1994; Lucy, 2016). There is a hypothesis independent of language that assumes that there is gender classification where natural objects are perceived as feminine and artefacts as masculine (Mullen, 1990). It comes from the traditional distribution of roles in society: females are seen as those who give birth and nurse therefore the connotation with

(11)

natural objects (e.g. tree would be perceived as feminine), while males make and work with tools, hence the association with artefacts (e.g. table would be considered masculine) (Haertlé, 2017). Studies of Mullen (1990) with children and Sera, Berge & Castillo Pintado (1994) with children and adults confirmed that. However, in more recent, abovementioned study by Haertlé (2017) and another by Bassetti (2014) with young adult monolinguals, second language learners, and early bilinguals no such effect was found.

1.4 The similarity and gender and the sex and gender hypothesis

Vigliocco et al. (2005) assume that, if we can find an effect of grammatical gender on perception or meaning, it must be a consequence or by-product of language acquisition mechanisms. Maciuszek et al. (2019) give the following example to illustrate this learning process: “When learning the word sun, a person from Germany (die Sonne) can associate it with stereotypically feminine features, while someone from Spain (el sol) may focus on features stereotypically related to men”. This phenomenon is further explained by Boroditsky et al. (2003) where they state that these stereotypically gendered features may become more important in the mental representation (or perception) of a given object because of this need to refer to their nouns in a gendered way. In their paper “Grammatical Gender Effects on Cognition: Implications for Language Learning and Language Use”, Vigliocco et al. (2005) consider two new, alternative mechanisms by which grammatical gender might affect cognition. These are the similarity and gender hypothesis and the sex and gender hypothesis. These two hypotheses are not mutually exclusive and explain ways the influence of grammatical gender could occur during language acquisition.

In the similarity and gender hypothesis, Vigliocco et al. (2005) theorise that these effects occur during a learning mechanism used by children to tie the linguistic input to aspects of meaning

(12)

(Fisher, 1994; Fisher & Gleitman, 2002; Landau & Gleitman, 1985; Landauer & Dumais, 1997). They assume that words with similar syntactic and morphological properties usually have a similar meaning and that nouns with the same grammatical gender are used in the same linguistic context. These phenomena can be seen at the syntactic level because in sentences, words which have the same gender require the same gender agreement with determiners, adjectives, and pronouns. As the authors state, this hypothesis expects that these effects will occur both in languages with only two grammatical genders and in languages with more than two. Furthermore, it is expected that in languages with a large number of grammatical gender markers, a grammatical gender effect will occur in words depicting both animate and inanimate objects (Maciuszek et al., 2019).

On the other hand, the sex and gender hypothesis states that the effects of grammatical gender could be based on establishing associations between gender of nouns and sex (Vigliocco et al., 2005). It is theorised that, as a gendered language is learned, the learner will notice a relationship between gender and sex for nouns referring to humans. This relationship will then be generalised and applied in the same way to other nouns for which there is no direct correspondence. As the authors state: “Words of the same gender would be more similar among themselves than words of different gender by virtue of sharing male- or female-like semantic properties.” This hypothesis also predicts that this effect will be stronger in languages such as the Romance languages because of their high degree of transparent relationships between the gender of nouns referring to humans and the sex of the referents which facilitates the generalisation; and weaker or non-existent in languages such as German with multiple genders because the relationship between sex and gender of human referents will be more difficult to establish for the learner. This last prediction was later confirmed by Sera et al. (2002) and Vigliocco et al. (2005) where no grammatical gender effects were found in German. The original version of sex and gender

(13)

hypothesis assumes that the effect of grammatical gender will appear only for sexuated nouns (Vigliocco et al., 2005). However, in the less strict version of this hypothesis the effect of grammatical gender will encompass all words.

1.5 Polish

Polish stems from the West-Slavic language family and is a grammatical gender language with three categories in singular form: masculine, feminine, and neuter; and two additional genders in plural form: masculine-personal and non-masculine-personal. There is a considerable amount of gender markers encompassing morphological information accordant with the grammatical gender of the noun, for instance verbs, adjectives, numerals, personal pronouns, demonstrative pronouns (Maciuszek, Polak & Świątkowska, 2019). The distribution of the genders is rather arbitrary (Swan, 2015), e.g. Ziemia (‘Earth’) is feminine, księżyc (‘moon’) is masculine and słońce (‘sun’) is neuter, malina (‘raspberry’) is feminine, kalafior (‘cauliflower’) is masculine and jabłko (‘apple’) is neuter. Sometimes neuter gender is assigned to nouns that have defined biological sex, for instance dziecko (‘child'); szczenię (‘puppy’) or dziewczę (‘young girl’). Moreover, some nouns that depict men may have feminine grammatical gender and vice versa: jego dostojność (‘his majesty’ where ‘majesty’ is feminine) and podlotek (‘teenage girl’ is masculine) (Maciuszek, Polak & Świątkowska, 2019).

(14)

Table 1

Retrieved from "Grammatical gender influences semantic categorization and implicit cognition in polish." by J. Maciuszek, M. Polak, N. Świątkowska. 2019, Frontiers in Psychology, 10, 2208-2208.

Although Polish is a gendered language with three grammatical genders, various studies have shown the strong effect of grammatical gender on perception which questions conclusions by Sera et al. (2002) that the effect of grammatical gender is more visible in languages with two, rather than three grammatical genders. Above-mentioned outcomes of studies by Rączaszek-Leonardi (2010), and Haertlé (2017) stand in line with a more recent study by Maciuszek, Polak & Świątkowska (2019). In the latter research, three experiments were conducted: triadic similarity judgments where in case of animals the perceived similarity of words showed the effect of grammatical gender, but in case of abstract concepts or inanimate objects it did not; modified implicit association test which demonstrated the impact of grammatical gender on reaction times and the number of categorisation errors; and assigning male/female voices to animals and inanimate objects where the impact of grammatical gender on both animals and inanimate objects was present.

(15)

1.6 Dutch

Dutch, one of the West-Germanic languages, is a natural gender language with two genders: common and neuter. Definite article de is assigned to the common nouns (de hond- ‘the dog’), while the definite article het precedes neuter nouns (het ei- ‘the egg’) and also all diminutive forms (het hondje- ‘the doggy’). De serves also as a determiner for all plural forms, regardless of the gender of the noun (de honden-‘the dogs’; de eieren- ‘the eggs’). Hüning et al. (2006) state that Dutch gender system finds itself in between German, that has three genera (der, die, das) and English one with one genus (the). Furthermore, Audring (2006) addresses the occurrence of gender loss that is happening Dutch, where gender system is evolving towards a semantic system, similar to what is apparent in the case of English. Gender is marked on various components such as definite and demonstrative articles, adjectives, and relative pronouns (Brouwer, Sprenger & Unsworth, 2017). Pronouns that refer to inanimate nouns do not differentiate between masculine and feminine nouns- hij is used for both masculine and feminine nouns, het for neuter nouns (De Vogelaer, 2006). The adjectives in the attributive position normally end with a schwa, apart from situations when the noun is indefinite, neuter and in a singular form (Loerts, Wieling & Schmid, 2013). The amount of common and neuter nouns is not divided equally, as about 75% of all nouns in Dutch have common gender, which results from the fact that common gender fused obsolete masculine and feminine genders (Van Berkum, 1996).

(16)

Table 2

Gender Definite article Adjectives in definite

NPs

Adjectives in indefinite NPs

Single

Common de appel de rode appel een rode appel

Neuter het huis het rode huis een rood huis

English equivalent

The apple/house The red apple/house A red apple/house

Plural

Common de appels de rode appels –

Neuter de huizen de rode huizen –

English equivalent

The

apples/houses

The red apples/houses –

Retrieved from "Neuter is not common in dutch: Eye movements reveal asymmetrical gender processing." by H. Loerts, M. Wieling, M. Schmid. 2013, Journal of Psycholinguistic Research, 42(6), 551-570.

The effect of gender in Dutch has not been explored much, as, to the authors’ knowledge, there is only one research that includes Dutch in tested languages examining linguistic relativity in the context of gender. In a study done by Vandewynckel (2008) Dutch was assigned to an unsystematic language category, along with English; the second category, gender loaded languages

(17)

which are morphologically rich languages with various gender markers, included French and Arabic. The classification of Dutch came from that although it has a gender system (common and neuter), it does not differentiate between feminine and masculine nouns. In a study by Vandewynckel (2008), the participants’ task, as seen in earlier experiments, was to assign male/female voice to animals and inanimate objects. As can be predicted Dutch and English speakers’ choices were in majority conceptually motivated rather than grammatically motivated. Because Dutch is a natural gender language, like English, which has been researched more deeply, it can be hypothesised that some of the findings in English may also apply to Dutch. Outcomes of studies by Flaherty (2001) on Spanish and English, Ramos & Roberson (2011) on Portuguese and English, Belacchi & Cubelli (2011) on Italian and English have all shown no influence of gender on the responses of English participants.

1.7 French

French is a Romance language from the Indo-European family, a descendent from the Vulgar Latin language it inherited its gender system with only masculine and feminine being present. This means that every noun in the French language is either masculine or feminine. The definite articles are “le” for masculine and “la” for feminine; the indefinite articles are “un” and “une” for masculine and feminine respectively. Some words are feminine even though it can describe a male, eg. une personne (a person) and vice-versa, un professeur (a teacher) but the distribution is not as arbitrary as it may seem as it is estimated that a noun’s gender can be guessed with 80% accuracy by its ending. Furthermore, some nouns are only marked in gender by their article, eg. un catholique, une catholique (a catholic man, a catholic woman). Adjectives must agree in number and gender with the noun it modifies which means that grammatical gender has an extensive effect on both the written and oral form of adjectives.

(18)

Table 3: Example of French grammatical gender

Retrived from “French grammar”, 2020 (https://en.wikipedia.org/wiki/French_grammar)

Many studies have shown that grammatical gender has an influence on noun perception in French, for example, Haertlé (2017) showed that French speakers are very susceptible to pair stereotypically gendered adjectives in congruence with the grammatical gender of the presented nouns. Furthermore, studies by Sera et al. (2002) and Vandewynckel (2008) showed that French speakers would assign masculine or feminine voices to objects presented as pictures in accordance with their grammatical gender.

1.8 Statement of purpose

The current research aims to contribute to existing studies on the weak version of the linguistic relativity theory, focusing on the influence of grammatical gender on perception. The languages tested in this research-Polish, Dutch and French stem from three different language families (Slavic, Germanic and Romance). Polish and French are grammatical gender languages, with three and two grammatical genders respectively, while Dutch is a natural gender language with two-way gender system: common and neuter. Previous findings on Dutch are fairly limited and this combination of languages has not yet been tested, therefore the outcomes could expand upon the work of Rączaszek-Leonardi (2010), Haertlé (2017), Maciuszek, Polak & Świątkowska,

(19)

(2019) and Vandewynckel (2008). The goal of this experiment was to find the answers to the following research questions:

RQ1: Do Polish/French speakers ascribe masculine/feminine features to presented objects in congruence with the grammatical gender of their spoken language?

RQ2: Do Polish/French/Dutch speakers ascribe feminine features to natural objects and masculine features to artefacts?

RQ3: Is there an identifiable pattern in which speakers of Dutch, a natural gender language, ascribe masculine/feminine features to presented objects?

The obtained results will also be analysed and discussed in regard to similarity and gender and sex

and gender hypothesis.

It is hypothesised that Polish and French speakers will ascribe masculine/feminine features in congruence with the grammatical gender of their spoken language. Dutch speakers, on the other hand, are expected to assign masculine/feminine features arbitrarily to presented objects. As observed in the recent studies, it is presumed that Polish/French/Dutch speakers won’t ascribe feminine features to natural objects and masculine features to artefacts.

The two tasks of this study are constructed following the research done by Haertlé (2017) who based her experiments on work by Sera, Berge, & del Castillo Pintado (1994) and Boroditsky & Phillips (2003). The first task is assigning male/female voices to presented objects, that will be characters in an animated movie. The objects are presented as pictures and have the opposite grammatical gender in Polish and French (e.g. table has masculine gender in Polish and feminine in French). The division of common and neuter nouns in Dutch is 75% common and 25% of neuter nouns so that the natural distribution present in the language is preserved. In the second task, participants choose three adjectives from a list that according to them best characterises each noun.

(20)

This experiment is carried out in the form of an online questionnaire, which allowed for a greater amount of data collection.

2. Method

This study was based on research by Haertlé (2017) and investigated the influence of grammatical gender on the perception of words in three different languages: Polish, French, and Dutch. The experiment was constructed to see if Polish and French speakers ascribed masculine and feminine features in congruence with the grammatical gender of the respective languages; whether Dutch speakers ascribed masculine and feminine features in a certain way; and if Polish, French and Dutch speakers perceived artefacts as masculine and natural objects as feminine.

2.1 Research Participants

The participants selected for this study were 30 Polish, 30 French and 30 Dutch speakers. In each language group, the gender was fairly balanced, the mean level of education was a bachelor’s degree and mean age was approximately 30 years old. Participants with more than one native language had to be excluded from this experiment for consistency purposes. Table 4 presents the overview of Polish, French and Dutch participants. No compensation was given for participation in this experiment.

Table 4: The overview of Polish, French and Dutch participants

Number of speakers

Gender Mean age Mean level of

education

Polish speakers 30 12 male/ 18 female 28 Bachelor’s degree

French speakers 30 13 male/ 17 female 38.3 Bachelor’s degree Dutch speakers 30 11 male/ 19 female 29.23 Bachelor’s degree

(21)

2.2 Materials

Task 1

First, a list of 20 nouns describing animate and inanimate object was created. It was based on Haertlé’s list of words but there were a few changes made. Because this experiment examined not two but three languages (Polish, French and Dutch) the list of nouns had to also be adapted for the Dutch language. A list was created so that in Polish and French nouns would have the opposite grammatical gender and so that, in Dutch, 25% of the words were neuter (het-woorden) and 75% were common nouns (de-woorden) to best reflect the natural repartition present in the Dutch language. The nouns with neuter gender in Polish were excluded from this experiment. Secondly, pictures representing animate and inanimate objects were selected using a free website https://www.flaticon.com. We used this particular site because of another study on linguistic relativity by Maciuszek, Polak & Świątkowska (2019) that found that since colour can be associated with gender (e.g. blue for boys, pink for girls) (Cunningham & Macrae, 2011), the pictures we selected needed to be the most basic black and white images so that the colour would not interfere with the answers. By using pictures instead of words, the cues of grammatical gender were effectively removed so the assignation of masculine or feminine features to objects in a conceptual way was allowed.

(22)

Figure 1. Black and white pictures showing the images used and the way they are presented in the questionnaire

Task 2

For the second experiment, a list of 12 nouns describing animate and inanimate object was created. Again, it was based on Haertlé’s list of words with a few changes made and nouns with neuter gender in Polish were not used in this experiment. A list was created so that in Polish and French the nous have the opposite grammatical gender and in Dutch 25% of the words were neuter (het-woorden) and 75% were common nouns (de-woorden) to best reflect the natural division present in the Dutch language. Next, a list of 24 adjectives used from Haertlé’s experiment was translated into three languages. Heartlé created a list of the adjectives so that they would represent stereotypically masculine and feminine characteristics. It was based on descriptions in Bem (1993) and Kuczyńska (1992). In Polish, nouns do not have articles and the ending usually cues the grammatical gender. The adjectives that are masculine usually end with an y and feminine with an

a so there each adjective was put in two forms (miły/a- nice) so that the adjectives would not cue

the answer. In French, the nouns were presented with matching articles. The decision to place them with the nouns came from the fact that in Polish, the ending of the noun gives away the gender and this experiment does not test people’s knowledge about the article as it is assumed that native speakers are aware of that. The adjectives in French have been presented with masculine and feminine options, usually by adding an e at the end (froid/e-cold), except the cases where the adjective does not inflect (fragile-fragile). In the case of Dutch, the nouns were also presented with articles for the same reasons as in French. All of the adjectives were presented with an e at the end (sterke-strong) to match nouns with definite articles.

Three online surveys in Polish, French and Dutch including both tasks were constructed using the platform Qualtrics. The instructions, as well as the tasks, were written in Polish, French

(23)

and Dutch, respectively. The surveys were spread around using social media. There was no time limit on the tasks.

2.3 Procedures

Each participant received a link pointing to a questionnaire which was split into three parts. At the beginning of the questionnaire a short test was presented:

Thank you very much for participating in this survey. This questionnaire is not a test and there are no "right" or "wrong" answers. We are interested in your personal opinion. The results of this survey will be used only for research purposes so please give your answers sincerely to ensure the success of this project. Thank you very much for your help!

Next, the part with information about the participants was shown with three categories: gender- they could choose between male/female/other; age- participant had to fill in the age himself; the level of education- options to choose from were primary/secondary/vocational education/bachelor/master/PhD.

Subsequently, instruction to the first task was presented:

We want to film an animated movie where different characters will play. In a moment you will see a number of pictures showing animate and inanimate objects that are going to play in our movie. We need to decide which voice should each character have - male or female. We ask you to choose by each picture if a particular object should, in your opinion, have a masculine or feminine voice.

Then, participants were shown 20 pictures with male/female options to choose from. The images were shown to the participants in a random order to minimize the potential confounding variables effect.

(24)

For each noun match three adjectives from a list, that in your opinion best describe it.

Twelve nouns had to be matched with three out of 24 adjectives. Again, the nouns were displayed in a different order to each participant. The questionnaire was constructed in a way that an answer for each field was required and that, after completing a task, the participants could not go back to it and change their answer. There was no time limit for the tasks and the average time spent on filling in the questionnaire was 10 minutes.

2.4 Design and Analyses

Processing the data was the same for Polish and French, however, Dutch required different steps. The section below explains how the data were coded and analyzed in all three languages.

Polish and French

First, each answer that was congruent with the grammatical gender of the language was given 1 point and answer that was not congruent with the grammatical gender was given 0 points. They were later summed up for each participant as well as for the whole language group. This way responses for Experiment 1 and Experiment 2 were coded. Subsequently, the statistical association between ascribed voice and grammatical gender (for Experiment 1) and the statistical association between ascribed stereotypical adjectives and grammatical gender (for Experiment 2) were tested. To see whether features congruent with the grammatical gender were more often assigned to artefacts or to natural objects the sum of answers congruent with the grammatical gender for artefacts and natural objects were calculated. Then, the difference between congruent answers for artefacts and congruent answers for natural objects were statistically compared.

Dutch

The coding for Dutch data looked slightly different because Dutch does not have grammatical gender. Answers in which masculine features were ascribed to common and neuter

(25)

nouns were coded as M and answers in which feminine features was ascribed to common and neuter nouns were coded as F. This way responses for Experiment 1 and Experiment 2 were processed. The total number of answers where masculine feature was ascribed to common and neuter nouns as well as the number of answers where feminine features were ascribed to comon and neuter were calculated. Later, the statistical association between word type (common/neuter) and ascribed voice (for Experiment 1) and the statistical association between word type (common/neuter) and ascribed stereotypical adjectives (for Experiment 2) were tested.

Polish, French and Dutch

Next, each answer in which masculine features were ascribed to artefacts and feminine features were ascribed to natural objects was given 1 point and each answer in which masculine features were ascribed to natural objects and feminine features were ascribed to artefacts was given 0 points. Then, the points were summed up for each participant and for the whole language group. This data coding was done again for Experiment 1 and Experiment 2. After that, the statistical association between ascribed voice and word type (natural objects/artefacts) (Experiment 1) and the statistical association between ascribed stereotypical adjectives and word type (natural objects/artefacts) (Experiment 2) were tested.

Polish and French

Later, two language groups (Polish and French) were statistically compared to see whether one group’s answers to both experiments were more congruent with grammatical gender than the other. Afterwards, the two language groups were statistically compared to see if one group ascribed masculine features to artefacts and feminine features to natural objects more often than the other language group in both Experiment 1 and 2.

(26)

Polish, French and Dutch

Finally, the possible influence of other factors on responses was examined. Age, gender, and level of education were statistically tested to see if they had an effect on the outcomes from both experiments in both ways of coding (grammatical gender and artefacts/natural objects).

3. Results

In this section, we present the results for each language separately. First, the results of our Polish participants are introduced, followed by the French speakers’ outcomes. We then compare the Polish and French results, and lastly, we present our findings for the Dutch speakers group.

3.1 Polish participants

Ascribing masculine/feminine features in congruence with grammatical gender

To answer our first research question, whether Polish speakers ascribe masculine or feminine features to presented objects in congruence with their grammatical gender, two experiments were constructed: the first experiment asked the participants to ascribe a masculine or feminine voice to a presented picture. The second experiment asked the participants to ascribe 3 stereotypically masculine or feminine adjectives to a presented noun.

Experiment 1

A chi-square analysis revealed that the association between ascribed voice and grammatical gender of the noun was significant, χ2 (1, N=30) = 60.81, p < 0.001. Masculine voice was chosen more often when grammatical gender of the noun was masculine, while the feminine voice was

(27)

chosen more often when grammatical gender of the noun was feminine. This effect was medium

φ=0.32 and is visualized in Figure 2.

Figure 2: Bar Plot of the distribution of masculine (MaleVoice) and feminine voices (FemVoice) for nouns with

masculine (GGMasc) and feminine grammatical gender (GGFem).

Experiment 2

A chi-square analysis revealed that the association between stereotypical adjectives and grammatical gender of the noun was significant, χ2 (1, N=30) = 12.66, p < 0.001. Stereotypically masculine adjectives were chosen more often when grammatical gender of the noun was masculine and stereotypically feminine adjectives were chosen more often when grammatical gender of the noun was also masculine. This effect was small φ=0.11 and is visualised in Figure 3.

(28)

Figure 3: Bar Plot of the distribution of stereotypically masculine (MaleAdj) and feminine adjectives (FemAdj) for

nouns with masculine (GGMasc) and feminine grammatical gender (GGFem)

Answers congruent with the grammatical gender in natural objects and artefacts

Answers of Polish speakers congruent with the grammatical gender were compared between congruent responses for artefacts and congruent responses for natural objects in both experiments.

Experiment 1

A two-sample t-test showed no significant difference in scores of voices ascribed congruently with the grammatical gender for nouns depicting natural objects and artefacts. This difference is presented in Figure 4.

(29)

Figure 4: Boxplot showing the dispersion in results for nouns depicting artefacts and natural objects

Experiment 2

A two-sample t-test showed no significant difference in scores of stereotypical adjectives ascribed congruently with the grammatical gender for nouns depicting natural objects and artefacts. This difference can be observed in Figure 5.

(30)

Ascribing masculine features to artefacts and feminine features to natural objects

To answer our second research question, whether Polish speakers ascribe feminine features to natural objects and masculine features to artefacts, the same two experiments from 5.1.1 were constructed: the first experiment asked the participants to ascribe a masculine or feminine voice to a presented picture. The second experiment asked the participants to ascribe three stereotypically masculine or feminine adjectives to a presented noun.

Experiment 1

A chi-square analysis revealed that the association between word type (natural objects/artefacts) and ascribed voice was significant, χ2 (1, N=30) = 18.19, p < 0.001. Masculine voice was chosen more often when nouns depicted artefacts, whilst feminine voice was chosen

(31)

more often when nouns depicted natural objects. This effect was small φ=0.17 and has been visualised in Figure 6.

Figure 6: Bar Plot of the distribution of masculine (MaleVoice) and feminine voices (FemVoice) for nouns depicting

artefacts and natural objects.

Experiment 2

A chi-square analysis revealed that the was no significant association between the word type (natural objects/artefacts) and ascribed stereotypical adjectives. This is visualised in Figure 7.

(32)

nouns depicting artefacts and natural objects.

Influence of other factors

Age, gender, and level of education were tested to see if they had any effect on the results of Polish speakers.

Age

A Kendall rank correlation showed no significant relationship between age and responses of Polish participants.

Gender

In most cases, two sample t-tests showed no significant difference between the results of Polish men and women. Only in Experiment 1 (ascribing voices) RQ2 (natural objects/artefacts) a Wilcoxon rank sum test revealed that men (M=12.5, SD= 1.78) ascribed masculine voice to

(33)

artefacts and feminine voice to natural objects more often than women (M=11.22, SD=1.11). This difference was significant W=155.5, p=0.0397. The effect was large, δ =1 and has been visualised in Figure 8.

Figure 8: Boxplot showing the dispersion in results for Men (1) and Women (2)

Level of education

A Kendall rank correlation showed no significant relationship between level of education and responses of Polish participants.

3.2 French participants

To answer our first research question, whether French speakers ascribe masculine or feminine features to presented objects in congruence with their grammatical gender, two experiments were constructed: the first experiment asked the participants to ascribe a masculine

(34)

or feminine voice to a presented picture. The second experiment asked the participants to ascribe 3 stereotypically masculine or feminine adjectives to a presented noun.

Experiment 1

A chi-square analysis revealed that the association between ascribed voice and grammatical gender of the noun was significant, χ2 (1, N=30) = 48.50, p < 0.001. Masculine voice was more often chosen when grammatical gender of the noun was masculine, while feminine voice was chosen more often when noun had feminine grammatical gender. This effect was medium φ=0.28 and has been visualised in Figure 9.

Figure 9: Bar Plot of the distribution of masculine (MaleVoice) and feminine voices (FemVoice) for nouns with

(35)

Experiment 2

A chi-square analysis revealed that the association between stereotypical adjectives and grammatical gender of the noun was significant, χ2 (1, N=30) = 13.42, p < 0.001. Stereotypically masculine adjectives were more often assigned to nouns with feminine grammatical gender and stereotypically feminine nouns were more often assigned to nouns with masculine grammatical gender. This effect was small φ=0.11 and has been visualised in Figure 10.

nouns with masculine (GGMasc) and feminine grammatical gender (GGFem).

Answers congruent with the grammatical gender in natural objects and artefacts

The results of French speakers congruent with the grammatical gender were compared between congruent responses for artefacts and congruent responses for natural objects in both experiments.

(36)

Experiment 1

A two-sample t-test showed no significant difference in scores of voices ascribed congruently with the grammatical gender for nouns depicting natural objects and artefacts. This difference is presented in Figure 11.

Experiment 2

A two-sample t-test showed no significant difference in scores of stereotypical adjectives ascribed congruently with the grammatical gender for nouns depicting natural objects and artefacts. This difference is shown in Figure 12.

(37)

To answer our second research question, whether French speakers ascribe feminine features to natural objects and masculine features to artefacts, the same two experiments were constructed: the first experiment asked the participants to ascribe a masculine or feminine voice to a presented picture. The second experiment asked the participants to ascribe 3 stereotypically masculine or feminine adjectives to a presented noun.

Experiment 1

A chi-square analysis revealed that there was no significant association between ascribed voice and word type (natural objects/ artefacts). This is presented in Figure 13.

(38)

Experiment 2

A chi-square analysis revealed that the association between stereotypical adjectives and word type (natural objects/ artefacts) was significant, χ2 (1, N=30) = 7.22, p=0.007. Stereotypically masculine adjectives were chosen more often when nouns depicted artefacts and stereotypically feminine adjectives were chosen equally often when nouns depicted natural objects and artefacts. This effect was small φ=0.08 and has been visualised in Figure 14.

(39)

Age, gender, and level of education were tested to see if they had any effect on the results of French speakers.

Age

A Kendall rank correlation showed no significant relationship between age and responses of French participants.

Gender

Two sample t-tests showed no significant difference between the results of French men and women.

(40)

A Kendall rank correlation showed no significant relationship between level of education and responses of French participants.

3.3 Comparison of Polish and French results

The results of Polish and French participants have been compared to see if outcomes of any language group were more congruent with grammatical gender of their spoken language.

Experiment 1

Table 5: Descriptive statistics for Polish and French speakers in Experiment 1 RQ1

Mean SD Min Max

Polish speakers 14.03 3.36 4 19

French speakers 12.87 2.76 7 18

The answers of Polish speakers were more congruent with the grammatical gender of the nouns than the answers of French speakers (see Table 5), however a two-sample t-test showed that this difference was not significant.

Experiment 2

Mean SD Min Max

(41)

Polish participants (M=19.93, SD= 3.5) ascribed stereotypical masculine/feminine adjectives in congruence with the grammatical gender more often than the French participants (M=16, SD=2.88) as illustrated in Table 6. A Wilcoxon rank sum test revealed this difference to be significant W=747, p< 0.001. The effect was large, δ =0.83 and has been visualised in Figure 15.

Figure 15: Boxplot showing the dispersion in results for Polish (1) and French (2) participants

The results of Polish and French participants have been compared to see if either of these groups ascribed masculine features to artefacts and feminine features to natural objects more often.

(42)

Experiment 1

Mean SD Min Max

A Wilcoxon rank sum test revealed that Polish participants (M=11.73, SD= 1.53) ascribed masculine voices to artefacts and feminine voices to natural objects more often than French participants (M=10.57, SD=2.43) as visualised in Table 7. This difference was significant W=581.5,

p= 0.04948. The effect was large, δ =1 and has been visualised in Figure 16.

(43)

Experiment 2

Mean SD Min Max

French speakers ascribed stereotypically masculine adjectives to artefacts and stereotypically feminine adjectives to natural objects more often than Polish speakers as visualized in Table 8. Nevertheless, a Wilcoxon rank sum test showed that this difference was not significant.

3.4 Dutch participants

Identifiable pattern for ascribing masculine/feminine features

To answer our third research question, whether there is an identifiable pattern in which speakers of Dutch, a natural gender language, ascribe masculine and feminine features to presented objects, two experiments were constructed: the first experiment asked the participants to ascribe a masculine or feminine voice to a presented picture. The second experiment asked the participants to ascribe 3 stereotypically masculine or feminine adjectives to a presented noun.

Experiment 1

A chi-square analysis revealed no significant association between the word type (de/het) and ascribed voice. This is shown in Figure 17.

(44)

Figure 17: Bar Plot showing the use of masculine and feminine voice for common (de) and neuter (het) nouns

Dutch speakers ascribed masculine voice to common nouns (de) more often than to neuter nouns (het) and feminine voice to common nouns more often than to neuter nouns (see Table 9). This represents the effect of frequency that was replicated in this study. Masculine voice was ascribed to almost 50% of all common nouns and feminine voice to little more than 50% of all common nouns. Half of all neuter nouns had masculine voice ascribed and the other half feminine voice.

Table 9: Results for Dutch participants’ ascribing voice to common (de) and neuter (het) nouns

Men Woman

De 223 227

(45)

Experiment 2

A chi-square analysis revealed no significant association between the word type (de/het) and ascribed stereotypical adjectives. This is shown in Figure 18.

Figure 18: Bar Plot showing the use of stereotypically masculine and feminine adjectives for common (de) and neuter

(het) nouns

Dutch participants ascribed stereotypically masculine adjectives more often to common nouns than to neuter nouns and stereotypically feminine adjectives to common nouns more often than to neuter nouns (see Table 10). Stereotypically masculine adjectives were ascribed to 49% of all common nouns, and stereotypically feminine adjectives were ascribed to 51% of all common nouns. 46% of all neuter nouns had stereotypically masculine adjectives and 54% stereotypically feminine adjectives assigned to them.

(46)

Table 10: Results for Dutch participants’ ascribing stereotypical adjectives for common (de) and neuter (het) nouns

Men Woman

De 394 416

Het 146 124

To answer our second research question, whether Dutch speakers ascribe feminine features to natural objects and masculine features to artefacts, the same two experiments were constructed: the first experiment asked the participants to ascribe a masculine or feminine voice to a presented picture. The second experiment asked the participants to ascribe 3 stereotypically masculine or feminine adjectives to a presented noun.

Experiment 1

A chi-square analysis revealed that the association between natural objects/artefacts and ascribed voice was significant, χ2 (1, N=30) = 12.91, p < 0.001. Masculine voice was chosen more often when nouns depicted artefacts, while feminine voice was chosen more often when nouns depicted natural objects. This effect was small φ=0.15 and has been visualised in Figure 19.

(47)

Experiment 2

A chi-square analysis revealed that the association between natural objects/artefacts and stereotypical adjectives was significant, χ2 (1, N=30) = 5.93, p=0.0149197. Stereotypically masculine adjectives were chosen more often when nouns depicted natural objects, while stereotypically feminine adjectives were chosen more often when nouns depicted artefacts. This effect was small φ=0.07 and has been visualised in Figure 20.

(48)

Age, gender, and level of education were tested to see if they had any effect on the results regarding second research question (artefacts/ natural objects).

Age

A Kendall rank correlation revealed that there was no significant correlation between age and results of Dutch participants.

Gender

A two-sample t-test showed that there was no significant difference in the results of Dutch men and women.

(49)

A Kendall rank correlation revealed no significant correlation between level of education and results of Dutch participants.

4. Discussion

In this section, we discuss our results for each language separately. First, the Polish outcomes are discussed, followed by French participants’ findings, a comparison of Polish and French, the interpretation of our Dutch results, the summary of all results, the outcomes discussed in the context of similarity and gender and sex and gender hypothesis and the limitations of this study.

4.1 Polish participants

The outcomes regarding the first research question, whether Polish speakers ascribe masculine/feminine features to presented objects in congruence with the grammatical gender of their spoken language, were not as clear as expected. In the first experiment, it was obvious that grammatical gender of the noun guided the answers of Polish participants, namely, a masculine voice was chosen more often when noun had masculine grammatical gender and feminine voice was chosen more often when noun had feminine grammatical gender. This confirmed the hypothesis that grammatical gender would influence the answers of Polish participants and it stands in line with previous findings by Rączaszek-Leonardi (2010), Haertlé (2017) and Maciuszek, Polak & Świątkowska, (2019). However, in the second experiment stereotypically masculine adjectives were ascribed more often to nouns with masculine grammatical gender and stereotypically feminine adjectives also to nouns with masculine grammatical gender. This did not confirm the hypothesis that grammatical gender would guide the answers of Polish participants

(50)

and differs from the outcomes of the aforementioned studies, therefore no irrefutable conclusions can be made.

In both experiments there was no difference between answers congruent with the grammatical gender for nouns depicting natural objects and artefacts.

Results concerning the second research question, whether Polish speakers ascribe feminine features to natural objects and masculine features to artefacts, were, yet again, not univocal. The first experiment showed that masculine voice was chosen more often when nouns depicted artefacts and feminine voice when nouns depicted natural objects. This finding did not stand in line with Haertlé’s (2017) and Bessetti’s (2014) results. Nevertheless, the second experiment revealed no significant association between ascribed adjectives and nouns depicting natural objects/artefacts, which coincides with the previous studies.

An additional analysis showed no visible effect of age, gender, and level of education on the results. In most cases, no trend was observed, supposedly because of the homogeneity of the age group and uneven distribution of education levels. It was not expected to find an effect of age as previous studies have only found the difference between adults and children below 8 years old (Sera et al., 2002) and this study did not include children in the group of participants. Gender and level of education were also not expected to have an impact on the results because linguistic relativity studies in the context of grammatical gender do not focus heavily on those factors.

4.2 French participants

The outcomes from both experiments concerning the first research question, whether French speakers ascribe masculine/feminine features to presented objects in congruence with the grammatical gender of their spoken language, were contradictory. Findings from the first experiment confirmed the hypothesis that grammatical gender would guide the answers of French

(51)

speakers as a masculine voice was chosen more often when nouns had masculine grammatical gender and feminine voice was chosen more often when nouns had feminine grammatical gender. This outcome stood in line with findings by Haertlé (2017) and Sera et al. (1994). Nonetheless, the second experiment did not confirm these results, stereotypically masculine adjectives were chosen more often when grammatical gender of the noun was feminine and stereotypically feminine adjectives were chosen more often when grammatical gender of the noun was masculine. The outcomes did not coincide with the results of the previously mentioned studies.

Both experiments revealed no difference between answers congruent with the grammatical gender for nouns depicting natural objects and artefacts.

Results with respect to the second research question -Do French speakers ascribe feminine features to natural objects and masculine features to artefacts?- refuted the assumption that masculine features are more often ascribed to artefacts and feminine features are more often ascribed to natural objects. The first experiment did not reveal any significant association between ascribed voice and artefacts/natural objects which stands in line with Haertlé’s (2017) and Bessetti’s (2014) findings. The second experiment showed a small association between ascribed adjectives and artefacts/natural objects, but it still did not confirm the hypothesis. Stereotypically masculine adjectives were more often ascribed to nouns that refer to artefacts, but stereotypically feminine adjectives were equally often ascribed to nouns depicting artefacts and nouns depicting natural objects.

As presented in the case of Polish, there was no effect of age, gender and level of education observed on the results. In each test, the differences between the groups or correlations were not significant, which was most probably caused by the homogeneity of the age group and unequal distribution of education levels.

(52)

4.3 Comparison of Polish and French results

In both experiments, the answers of Polish speakers were guided by grammatical gender more often than in the case of French speakers, however only the second experiment showed that this difference was significant. This finding coincides with the outcomes found by Haertlé (2017) and yet again overthrows the thesis by Sera et al. (2002) that states that the effect of grammatical gender is mostly visible in languages with two rather than three grammatical genders.

For artefacts/natural objects in Experiment 1 Polish speakers ascribed masculine voice to nouns that depicted artefacts and feminine features to nouns that depicted natural objects statistically more often than French speakers did. Nevertheless, in the second experiment it was exactly the opposite, but not statistically significant. Hence, no clear conclusions about the difference between the two language groups can be made.

4.4 Dutch participants

Results regarding the third research question, whether there is an identifiable pattern in which speakers of Dutch, a natural gender language, ascribe masculine/feminine features to presented objects, in both experiments were consistent. In both cases, masculine and feminine features were ascribed more to common than to neuter nouns. Approximately 50% of all common nouns had masculine features ascribed to them and the other half feminine features. The same distribution was observed in neuter nouns, namely, around 50% of all neuter nouns had masculine features ascribed to them and the other half feminine features. The statistical associations between ascribed voice and word type (common/neuter) and between ascribed stereotypical adjectives were not significant. This proved that natural gender did not cue the answers of Dutch participants and coincides with the previous findings by Vandewynckel (2008). In the case of these experiments common and neuter nouns were equally viewed as feminine and masculine. This finding stands in

(53)

line with the weak version of the linguistic relativity theory (Brown & Lenneberg, 1954), which states that the spoken language influences people’s perception. In this case, the speakers of Dutch, a neutral gender language, with no grammatical gender to guide participants’ answers, perceive common and neuter nouns as equally masculine and feminine, while speakers of gendered languages (e.g. Polish and French) view nouns with masculine grammatical gender as masculine and nouns with feminine grammatical gender as feminine.

Outcomes concerning the second research question, whether Dutch speakers ascribe feminine features to natural objects and masculine features to artefacts, were contradictory. In the first experiment masculine voice was more often ascribed to artefacts and feminine voice to natural objects. This did not stand in line with Haertlé’s (2017) and Bessetti’s (2014) findings. However, a second experiment revealed that stereotypically masculine adjectives were more often assigned to natural objects and stereotypically feminine adjectives to artefacts. Therefore, the assumption that nouns depicting natural objects are viewed as feminine and artefacts as masculine can be rejected. Dutch- a natural gender language, showed significant but contradictory effects in both experiments, while Polish and French- grammatical gender languages each revealed one significant and one non-significant effect for different experiments.

As in the case of Polish and French, no effect of age, gender, and level of education on the results was observed. Every test showed no significant correlation or difference, presumably because of the aforementioned reasons.

4.5 Summary of the findings

This study was based on Haertlé’s (2017) experiment, however, important changes were made, one of them being the addition of the Dutch language to the comparison. It was expected that the influence of grammatical gender would be visible in the answers of Polish and French participants