 So, welcome everybody. My name is Madana Seyfriedini-Poy. I'm the Director of the Endangered Languages Doctrine and Mutation Program for all of you. I see a lot of new faces that I don't know yet, but hopefully I will get to meet you over time. And it's my pleasure to host this seminar today and to introduce you to Roland Lander. Roland is a linguist who has come to work with us for two months from Griffith University in Australia. And he has worked on grammatical gender, on the influence of grammatical gender on cognition, a clouding effect. And I will put out a disclaimer right now that it is my fault that the talk has ten less slides on statistics. And he really, really is very unhappy about that. So anything when you see it hurts him because he can't talk more about the formula, it's my fault. So yes. So we're very happy to have him here. He is sitting in room 346 where also we are located, the Endangered Languages Archive, as well as the Endangered Languages Documentation Program right next to the Multilingual Lab, which is the student room where we hope to see you a lot because the room is for you guys. And if you want to talk to Roland and learn more about his work, please come and visit us in room 346. That's where he's placed until mid and November hopefully. He is here to explore possibilities on collaboration with our linguists, like for example the Crossroads Project or the Bantu Project, because those languages have from a European gender system really crazy categories. And when you learn about what he has done, you will understand why he's so excited about the work we are doing here and the understanding we have about languages that are usually not subject to this type of study. So I'm going to hand over to Roland. Thank you very much, Mandana, for that introduction. And thank you everyone for coming. Continuing on on your disclaimer, I actually kept those slides are here, so if you have any questions at the end, we can get, we can, we can talk about them after the presentation, they're in an appendix. So I'm going to talk about, as Mandana said, about the influence of grammatical gender on cognition, especially in conceptualization of objects. I'll give you a bit of a background, I'll keep that fairly short, talk about some of the previous studies that dealt with this issue. Then I'll talk a little bit about conceptualization, what it is, what are some of the issues that we need to think about when we when we study conceptualization. And then I'll focus mainly on my study that I carried out all about the aim, participants, the languages that I covered, which to some of you may be a bit boring, because this is kept within kind of the European realm. The design, it was a three-phase experiment, so it's fairly complex, but I'll make it simple. And I'll talk a little bit about the formal analysis. And the main part will be about the findings of what I actually found and some of the implications. So first of all, language and cognition, I mean, is there a connection between language and cognition? I mean, this is a question that's been pondered for literally millennia. And it was brought into sharper focus recently by Saper and Wolf, the Saper-Wolf hypothesis, many of you may have heard about that. And in fact, if there is a connection between language and cognition, if language influences cognition, the next question is, to what extent and how exactly does language influence cognition? But what I looked at specifically was grammatical gender and cognition, because it's a lot more specific and a lot sexier. Many studies have put out the proposition that there is a relationship between grammatical gender and conceptualization. Namely, it's a positive correlation that if something, if a language assigns, say, masculine to an object, the speakers of that language will conceptualize that object as more masculine than feminine. And vice versa, if it assigns feminine, they will conceptualize that object as more feminine. Now, some of the languages, as you know, like Spanish and German, assign a grammatical gender to each noun. So for example, in Spanish, this would be a desk, el escritorio, it's masculine. That's a table, and that's la mesa, it's feminine. So according to this hypothesis, Spanish speakers would conceptualize the desk slightly more masculine than the table, because of the grammatical gender that is assigned to it by Spanish. Now, this is not so clear cut, so there are, there is a lot of disagreement in the field about whether this is the case. And there are a number of studies, I only mentioned a few that argue against such a link. And there are also some studies that basically, as I put it, they sit on the fence. I kept the whole, when language affects cognition and when it does not, I mean, that's the title of the article. This is actually one of the best articles that you can read on the topic, except that I have to warn you that the conclusion, after the description of a series of really great experiments, the conclusion is the biggest fissure, because it says sometimes it does, sometimes it doesn't. Yeah, the end. But it's worth reading, just as long as you don't expect too much from the conclusion or any answers. So one study that I would like to have a look at in a little bit more detail, simply because my methodology was inspired by this study, is Borodis Kishmitin Phillips from 2003. They carried out a study. They claimed that people's ideas about objects can be influenced by aspects of grammar. They went a little bit further. They actually were talking about even ideas, design, architecture as well, that it's all basically influenced by grammatical gender. They tested German and Spanish speakers. They tested them in English. I have issues with that as well. So these were German and Spanish speaking participants, but they were all proficient speakers of English. They were all tested in the United States in English. I know. They were asked to generate adjectives to describe objects, which I'll tell you why I thought that was actually quite clever. But first I'll give you German and Spanish. Yes, in English. I mean, it is, look, I agree that it is more expedient, and that's where my agreement kind of ends with testing, you know, when you're looking for an effect in someone's native language and you test them in another language. I know. I had the same look on my face, honestly. But anyway, this was a fairly influential study, and it cited a lot for some reason. They were asked to, so participants were asked to generate adjectives about objects. And I'll give you one example. And this example was bridge. Bridge is grammatically masculine in Spanish, and it's grammatically feminine in German. So these participants were asked to describe a bridge with adjectives, the first adjectives that came to mind. And behold, Spanish speakers describe bridges as big, dangerous, long, strong, sturdy, towering. What do you think? Sorry? See, that's another very good question. We don't know how the stimuli were actually administered. No idea. I'd love to know. I hope so. This was, yeah. The study was never published in a book, and then it just went. Nevertheless, it took off like wildfire from that book chapter. And it's one of the studies that just pops up everywhere, including Stephen Fry. It's sexy. It is very sexy. It really is. It's a good story. It's an excellent story. I couldn't agree more. It really is a great story. But so what do you think Germans describe beautiful, elegant, fragile, peaceful, pretty, and slender? Yes. I mean, wow. I mean, this is really amazing, isn't it? So obviously, Boroditsky and her team, they concluded that there is a positive correlation between the grammatical, gender-related languages, science, and object, and its gender-related conceptualization. So meaning that, as you could just see, Spanish speakers, because it's grammatically masculine, will conceptualize it as more masculine and vice versa. I do have some butts. Many other studies, especially some recent studies, are suggesting that the interrelationship between grammatical features and conceptualization is a lot more subtle, and it's a lot more complex than a positive correlation, as described by Boroditsky and her team. And that the effects triggered by language-specific tasks may not necessarily mean, may not point to a direct influence on conceptualization. So what is conceptualization? Just a few words about that. This is conceptualization, right? Yes. What is it? It's a melon, yeah? Describe a watermelon, yeah. Describe a watermelon. Round and green. Round and green. Stripey. Juicy. Sweet. Heavy. Colorful, yeah? Anything else? Cool, yeah? Give me some adjectives, as in what is it like? Yeah? Okay, so we said, now what are from these, I mean, what can we actually see? Like, what are the perceptible attributes of this watermelon? What can we see? I mean, we can see that it's round. We can see that it's green. If we actually had a watermelon here, we could touch it. We could probably feel that it's smooth. If we lift it, we could feel that it's heavy. But we also said it's juicy, sweet. We could argue that it's delicious. How do we know that? We don't know that until we actually cut a watermelon. So how do we know that? We know that because we had previous interaction with a watermelon, right? I mean, most of us would have tasted our watermelon. So we know that when we cut it, it's going to be red. It's going to be juicy. When we taste it, it's going to be sweet, delicious, you know, all those things that we just mentioned. The only point I'm trying to make here about conceptualization is that what characteristics we assign to an object goes way beyond what we can actually perceive about the object. It also comes down to previous experience with the object, etc. It also comes down to environmental factors, idiosyncratic factors, as in how do I feel about a watermelon? I might hate watermelon. So I might not say that it's sweet and juicy. I might say it's yucky and disgusting, right? Cultural factors, etc., etc. So conceptualization is not as straightforward as it may seem from a round watermelon. Now, the big question is whether linguistic features such as grammatical gender play a part in how we conceptualize objects. And this is, and if it does, how and to what degree? This is what I set out to have a look at, to have a closer look at. So about this study, a few words. The aim of the study was to investigate the relationship between grammatical categories and cognition by looking at grammatical gender and conceptualization of objects. It was a fairly large-scale study. 1,290 participants took part in this study from 24 countries, which I'm very proud. And they were all native speakers of the target language. And I should also mention that all the experiments were kept in the target language. So it was none of that English business, other than for the English speakers. But while we're talking about the languages, the languages are sitting on a so-called gender-loading scale. Now, this gender-loading scale was originally proposed by Alexander Giora back in the 80s. Basically, it is a scale that is looking at languages according to the degree to which they necessitate speakers to be mindful of gender. And I will explain that a little bit more. We can probably argue about how they sit on here, but we can do that in the Q&A. Basically, we start with a zero gender-load language such as Hungarian. Hungarian does not have any grammatical gender. And it does not even have gender pronouns. So there is no he and she. Of course, there is a third person pronoun. It encodes that we're talking about a person, we're talking about a third person, but it does not include any information about the gender of that person, whether we're talking about a male or a female. And I should mention at this point that Hungarian, in this respect, is not unique. There are a number of languages that do not have any grammatical gender whatsoever. And not even a gendered pronominal system. All Finnell Greek languages, so Finnish, Hungarian, Estonian, all Turkic languages are non-gendered. They don't have any gender at all. Now, why I put, and this is why on this scale, English is sitting as a partial or low gender-loading language. Because when we put it next to a language like Hungarian, there is some gender in English. While English does not assign a grammatical gender to each noun, it does have a gendered pronominal system. So there is he and she. So if I say that Mandana stood up, tripped and hid himself, is that correct? No. What did I get wrong? My gender. If I can get the gender wrong, that means it features, I have to be mindful when I speak English, I have to be mindful that who I'm talking about is a male or a female. I cannot make this mistake in Hungarian, for example, or in a completely non-gendered language. You cannot make this mistake because there's no gendered pronominal system. So this is why. And now, Spanish, I put it at a full with two grammatical genders, masculine and feminine. Full just means that Spanish assigns a grammatical gender to each noun. So each noun, as I explained about the desk and the table, would have a grammatical gender. Hebrew was another language that also has two grammatical genders, masculine and feminine, but it permeates the language to a much greater degree than it does in Spanish. It features in the verb conjugation, it features in first person pronouns, it features a lot more than it does in Spanish, for example. And then I put German on top, and as I said, we can argue about that, which also has a full grammatical gender system, but with three grammatical genders, masculine and feminine, and neuter as well. And it does assign this agreement, et cetera. Now let's have a look at the stimuli that were used. This is, and I'll explain in a minute in a bit more detail, about what I did exactly, but I would like to show you the stimuli, first of all, that were used, because it was a fairly similar study to Boroditsky's that I asked my participants to describe objects. Except that to make sure that all the experiments were gauging the gender load associated with these objects accurately, I included some control items that were clearly, you know, the natural gender was clear. There were four feminine items, bride, ballerina, girl, woman, four masculine items, boy, sailor, man, boxer. These were mixed in with all the other. Now, this is where the trickery came, because I'm looking for objects that had opposite grammatical genders in some of these languages. So there were four objects that had opposite grammatical gender in German and Spanish. So like bridge, I already mentioned bridge before and bridge fork table star. Then there was a group of four that had opposite grammatical gender in German and Hebrew. Then there were four that had opposite grammatical gender in Spanish and Hebrew. And the last four was the same grammatical gender in all of them, two feminine ones, two masculine ones. So what participants had to do, they had to look at these objects. They were presented in different conditions, either text or image. The ones that had the text condition did not know about the image ones. The ones that had the images did not know about the text ones. And they were asked to come up with the first three adjectives that came to mind to describe these objects. And then the five most frequently occurring adjectives were collated. And just to give you an example, this was the stimulus fork, text condition, image condition. In English, these were the five most frequently occurring adjectives to describe a fork. Sharp, shiny, pointy, useful metallic. As you can see, there's not a huge difference between text and image condition in this case. They're basically the same adjectives in a slightly different order. Shiny, sharp, useful, metallic, pointy. Just to show off, I'll show you the Hebrew ones as well, that the five most frequently occurring adjectives were around shiny, sharp, useful, small, pretty. And for the image condition, sharp, useful, shiny, serrated, effective. Well, it is, I suppose, effective. So these were around just some examples of the five most frequently occurring adjectives. So in phase two, these adjectives were rated. And this was by a completely different cohort of participants who were not aware of the first phase. They were simply presented with lists of adjectives. And they did not know where these adjectives came from. So these, the adjectives collected in this first phase were used in the second phase as stimuli. Now, this is where some trickery had to be employed. They were told a bit of a story about a cartoon that we are making. And they were asked that an animated character in a cartoon described by a given adjective, should this character have a male voice or a female voice? Okay. So if a character is described as big, should it have a male voice or a female voice? It was a voice attribution task. And, you know, if a character is described as pretty, should it have a male voice or a female voice? And they were also given an opt out. They were given a can decide option which was in the instructions that was discovered. So it said, you know, if you really cannot make up your mind, put down can decide. So this is how these adjectives were rated. This is a quick example of bridge. From the image condition in English, these were the five most frequently occurring adjectives, long, strong, tall, big, sturdy. And you can see the gender ratings there as well, that there was fairly clear consensus on bridge in English. So male voice for long, for example, 49 participants said it should have a male voice, like, you know, someone described as long. 19 said female voice and seven couldn't decide. With sturdy, there was all 75 participants in this language in this phase agreed that some, like a character described as sturdy, should have a male voice in our cartoon. So overall, as you can see, this was rated as overwhelmingly more masculine when we relate these scores back to bridge. Bridge was conceptualized as more masculine than feminine. So in phase three, this phase was the complicated one. This was the one where we had to figure out exactly what was going on with all these scores, where we had to relate these scores that we got from the adjectives back to the original stimuli that generated them, to establish the gender-related conceptualization. And whether it has anything to do with the grammatical gender that is assigned to these objects in some of the languages. So if gender load scores were established for each adjective, then these gender load scores had to be related back to the original stimuli. And also the relative strength of the connection between these, so taking into account frequencies and weighting, et cetera, had to be established as well. A very quick word about the formal analysis, and I'm not going to use any equations, I promise. But one thing, because it is important, one thing that occurred to me as I was looking through the literature is that where a statistical analysis was even described or explained, they were mainly using an analysis of variants or a repeated measures analysis of variants. This is a statistical tool that you use for a continuous response, which is why this was one of the problems. I mean, gender association responses of the type used here are essentially binary and discrete. So, ANOVA is also a bit inefficient in isolating confounders, of which we have many. Keep in mind that we need to try to disentangle all kinds of effects, from cultural effects to media cinematic, remember the melon, and find what is actually due to the grammatical gender that is assigned to these objects. So, we used a logistic regression and I promise to keep it short because it is actually a bit like ANOVA except that it's developed for a binary response. So, it was a lot better suited for this kind of data. It is specifically designed for this kind of analysis, and it is a lot more effective in dealing with multiple confounders, as I said, of which we have many here. So, we could actually detect subtle effects a lot more effectively with this kind of analysis. Now, to give you one example of all the complexities that we had to deal with, one was we had to develop an adjective stimulus association measure, and why we had to do that is because adjectives obviously occurred for sometimes for multiple stimuli, okay, and I'll explain, I'll illustrate that in a minute. But if an adjective was uniquely associated with a stimulus, everything was clear, except that in most cases that was not the case. So, in which case the implications were a bit ambiguous. I'll show you what I mean. For example, here are from the Spanish cohort, there are two adjectives, onda or onda, means deep in Spanish, grande means big. Ondo was easy. It was associated with one stimulus, only with plate, okay. Plate generated deep in Spanish, and no other stimulus generated the same adjective. So, that was clear. These are the male voice, female voice, can't decide. So, that was easy to relate back to plate. However grande was associated with 11 of the stimuli from different conditions. So, obviously there were frequency issues, waiting issues that had to be taken into account here. And as you can see, it ranges from a frequency of 8 to 34. So, this would mean that these adjectives were relating back to the stimuli quite differently. So, the implications, this is what I meant that the implications were somewhat ambiguous. This had to be built into the model about which I'm not allowed to talk. But anyway, so this is just to give you an idea of, you know, some of the complexities that we had to deal with and why a stock standard analysis was just not going to cut it simply. So, we had to come up with a finer grain. Let's have a look at the findings. As I said, there were control and test stimuli. The control stimuli were simply there to test the instrument, to test the experiments that they could gauge gender-related conceptualization. A quick word about the scores, they were represented on a continuous scale. Anything positive means masculine. Anything negative means feminine. Okay? Sorry? Don't read too much into that. It's like, it had to be one or the other. So, let's have a quick look at the control stimuli and let's see if the instrument worked. I mean, I'm standing here talking about itself. Yes, it did. Sorry, spoiler. Here are the scores for ballerina. As you can see, consistently in all the languages, all the conditions, even though with these adjectives, they contextualize given to a whole different cohort than these scores related back to the original stimuli, it still came up consistently in all the conditions, all the languages as feminine. Boxer came up as masculine, which was great news. But actually, just to give you a further idea, all of them in all the languages, in all the conditions. So, we were fairly confident that this way of trying to gauge gender-related conceptualization worked because all the female controls were rated as feminine and all the male controls were rated as masculine. But the real juicy part is really going to be the test stimuli. And here's bridge. Now, to explain my buttons, which are a lot of fun, we're having a feminine side here, a masculine side here. These buttons are sitting where grammatical gender would suggest that they should be sitting. Okay, keep in mind the Borodicka hypothesis. Hungarian and English, sorry, different flags, sorry. Hungarian and English, they don't assign a grammatical gender, they're in the middle. Spanish and Hebrew, masculine. So, they're sitting on the masculine side. German, feminine. So, we would expect if Borodicka is right, this should stay somewhere here, those two should stay somewhere over there, and we have no idea what these will do. However, this is what it ended up looking like at the end. So, as you can see, they were actually pulling towards the masculine in all languages. Probably the one that was least is the one that really shouldn't have been. Because in the image condition in Hebrew, that was the one that was close to zero, so it was a little bit unclear. But all of them were rated as masculine. It's also interesting tonight is that Hungarian and English speakers were pretty sure about what they were doing, which was, I'll get back to that, because this gets juicier. Chair, I should take bits. I need to make some money here. What do you think happened with Chair? Yeah, but once again, it's remarkable how, regardless of grammatical gender, they were pulling on the one side, in this case on the masculine side, regardless of what grammatical gender. And again, English and Hungarian speakers are just great. They're very good at this, which was a bit of a surprise at first. Ring, like that. That's it. I tried not to, they were all black and white, very standard pictures. Masculine or feminine? It's feminine in all the languages. Table, masculine. But basically, the point I'm making here is that grammatical gender did not seem to be a predictor at all. It did not appear to play a part in this at all, regardless of the grammatical gender that is assigned to this. And I have the rest of them here. There's bridge. Now, there were obviously a few outliers. For example, with fork, the only condition in which it was rated a bit more feminine than masculine was in Hebrew, where it's grammatically masculine. So it was kind of going against grammatical gender. Table, we already saw that. Star. Again, it was a bit all over the place. But look at German. It's der Stearn. It's masculine in German. It was rated the most feminine in German. Again, the point I'm trying to make here is that it did not. Like, we did not see any evidence that grammatical gender was predicting the gender conceptualization here. Cloud. Same story. A few outliers. Absolutely, apparently, nothing to do with grammatical gender. Pistol was a pretty clear case. That it was all rated masculine. It was one of those really good cases, actually. Sorry, quick story about the pistol was. I mean, the adjectives are very consistent throughout four of the languages. Dangerous, deadly, loud, all that. Except in Hebrew, but the top two most frequently occurring adjectives were loaded and useful. Which I thought was cool. Ring. Same kind of thing. Moon. Once again. You see, it's feminine in Hebrew. It was rated not very strongly, but somewhat masculine. So once again, it's feminine in Spanish and it was rated masculine. So you're kind of getting my point, I hope, that we did not find any evidence for grammatical gender predicting the gender conceptualization of these objects. Hand, fish, now. So as I said, as you could expect, even after very careful analysis, there was no evidence that there was a positive correlation between grammatical gender and the gender conceptualization of an object. In fact, the effect turned out to be slightly paradoxical. I'm talking about a tiny effect here. So don't read too much into it. But it's not that it didn't reach statistical significance. It was actually slightly paradoxical, meaning that grammatically feminine objects were slightly more likely to be rated as masculine. As you can see, this number is positive here. And grammatically masculine objects were slightly more likely to be rated as feminine. So I was fairly convinced that this positive correlation effect was, we didn't have any evidence for it. However, I have lots of butts in, however. When we were looking at the data, something interesting was emerging from the data, which was, as you could see, remember me praising English and Hungarian speakers? How well they did with these tasks? Well, that's exactly what occurred to us, that it looked a little bit odd. So we decided to look into, as an additional, and this was really an exploratory measure at the time, to look into the absolute magnitude of the scores. And the magnitude here just means that how far away from zero they are. Because we thought that it would be informative of the confidence of the consensus. The further away a score is from zero, the clearer the consensus would appear to be. And if it's really just hovering around zero, then it indicates that the consensus is really not all that clear. So that's why we thought, since there's some information in these magnitudes, let's have a look at them, whether there's any pattern. Because it prompted a speculation that there may, when we looked at the magnitudes, there was a pattern, in fact. And it was suggesting that the complexity of gender systems may actually, to some degree, suppress the confidence with which participants assign gender characteristics to objects, which was interesting to say the least. So we thought, well, let's test it another way. Because this is really cool. So we looked at the can decide responses. And remember, I mentioned that participants were given a can decide option that was discouraged. I'll make a very quick point here about how participants were great in all languages because only 6% of them gave a can decide response overall. Except that, and for statistics, we need numbers. Because we had so many participants and so many responses, this was actually possible. Because even that 6% translated to 4,163 can decide responses overall. So we could actually have a look, we could do a proper analysis on whether there's any pattern of these can decide responses across the languages. And guess what? It was actually suggestive of the same thing as the magnitude test. It suggested pretty much the same thing. That the indecision increases with the complexity of a gender system, which was fascinating. To show you, these are the most beautiful five dots that I've ever seen in my entire life. They are gorgeous. So here, as you can see, we're going to decisive towards indecisive. And this is taking into account the magnitudes and the can decide responses. And this is the gorgeous pattern that emerges from that. And we start with Hungarian that has zero gender, then closely followed by English, which has some gender loading. And then we have a big gap. And then we go with the fully gendered language is Spanish, Hebrew, and German, most indecisive Germans, sorry. So this was really suggesting that there was something going on. So while we did not find any evidence for a positive correlation effect, this is really, this is basically suggestive of some other effect that maybe, maybe the complexity of a grammatical gender system does have an influence on the decisiveness with which gendered characteristics are assigned. Now, a few notes at the end. This was an effect that jumped out at us. So it was an unexpected finding, which means that these experiments were not designed with this in mind. Okay, so keep this in mind. Obviously, I would say that more studies and targeted research, I mean, this is offering a hypothesis that maybe we should be looking at grammatical gender in a different way. And we should be careful with what we're actually looking for, or what kind of effect it is that we're looking for. And that's why I would say that the approach and preconceptions that if we're looking for an effect, it is going to be a positive correlation. That preconception should probably be revisited and reexamined. And that's all I've got today. Do you have any questions? Thank you. Yes. The languages in terms of the functional load of natural gender within the gender system. My hunch would be that that figure in Spanish but in German where female nouns can have grammatical female gender are automatically masculine gender. My hunch would be that in Spanish the proportion of naturally female nouns that have grammatically female gender would be higher. And would that influence the results? That would be very interesting to look into, but no, we didn't look into that particular thing. But that would be, I agree, it would be very interesting to see whether that is the case. Yeah, we didn't check it. I know it's not a massive list of like typical items, so I understand. But so do you think our gender is not influencing people's attribution of adjectives? Well, what I'm saying, it's not a positive correlation. I'm actually, like these results suggest that they do influence to some degree, but it's not in a way that was proposed before. So it's not a positive correlation. It doesn't mean that if something is grammatically masculine, it will be conceptualized as more masculine. It could easily be, because I mean, admittedly, all these languages, I mean, one thing that I did which wasn't done much before is actually breaking out of the Indo-European mold. Because if you look at these studies, if you look at psycholinguistic studies about grammatical gender, I mean, it gets as exotic as Italian. Or, you know, language du jour, lately, is Slavic languages. But it's basically all Indo-European, same cultural sphere, et cetera. I did try to break out of that because Hungarian is not Indo-European and Hebrew is not Indo-European. So it was, you know, I was interested to see. But in terms of culture, I would say that they would, you know, I would put them in the same cultural sphere. So we don't, you know, it's not, it would be very interesting. I mean, why do you think I'm here? Being nice to people. Because it would be fascinating to have a look at that as an outside of this cultural sphere and with languages that have some really, really groovy noun classification systems. Yeah. Yeah, I mean, for all the other semantic aspects encoded in the non-class system, that they are the least best at focusing on natural gender, actually. Because, you know, they have all these non-classes that encode other semantic aspects. And it's normally only one pair of class or gender that has animals. Yeah. So there's a really, you know, marginal category in large systems. That would be nice. I think it would be fantastic. You could have a study with non-class languages, like reduced systems, like Wall-off or some bit of common languages, five, six, seven non-classes, and then, you know, to the other extreme like Atlantic, five, two. Yeah. Yeah. I mean, I couldn't agree more. I mean, that's why I flew 27 hours to be here. So it's, yeah, that's why, because I think it would be absolutely fascinating to do apply, you know, some sophisticated quantitative analyses to some of these languages and see whether we can find some patterns and what is actually going on. I'm sorry. I think that was a question. Yeah. And you could possibly argue that things that are supportive, like a table of rigid chairs, you know, kind of more masculine in your, not to do them equitably, to be a profession. No, because it has a lot, I mean, it has a lot to do with metaphorical features as well, as in, you know, like the shape of it, the size of it, the, you know, associations as in used by someone, like used by women more than, like, these associations. Yeah. So I'm wondering, yes, they actually, it was more about, they did some studies with made up genders. And like, I don't know if you heard about the use of even supertive categories. Same. It was another Boroditsky study and it was actually quite interesting because they came up with totally arbitrary as in just, they made up two gender classes. And they put in objects, like a violin, I think it was, that they grouped with either like, with some natural gender, you know, controls like a man or a giant or king or something. And when they were grouped with them with males, the claim was that it was described as, I think my favorite was overused, and which is apparently a masculine quality. Yeah, let's not think about that too hard. And basically this, the same kind of thing that the adjectives really differed, like based on what they were, whether they were grouped with men or with women. But yeah, with made up objects as in that would be, but that would be very difficult to actually like gauge whether language has anything to do with that. If you're working with made up objects. And I would say that you could probably do something and I would say that, you know, anything that was kind of roundish and small would be conceptualized as more feminine and whereas angular and, you know, would be more masculine, I'd say. But no, I'm not aware of such a study. Yeah. When you mentioned the Hebrew conception of pistol. Yeah. Once again, don't read too much into it. It could be interesting to compare Israeli Arab with, and I don't necessarily mean Palestinian, because they're also speakers of Arab that do not speak in English, but are integrated in the urban society of Israel, so culturally somehow connected to Europe compared to speakers of Arab in the regions that do not. Yeah, which would be, which I agree would be, would be interesting. I mean, there were some which I would call cultural effects. I mean, the other, the other interesting thing was that the star, for example, Hungarian speakers, one of the most frequently occurring adjectives was communist. So, you know, obviously that is like this, it's specific to certain events, like to, like a certain group of certain countries. So, yeah, I mean, these, these, these things occur. Sorry. It's masculine. Yes. For a couple of the examples you showed us, I'm sorry, I've forgotten which ones now, but there were a couple of the outliers. Yes. And there was quite a difference between the score you got with the text stimulus. Yes. Yes. Yes. Yes. Did you have a chance to investigate that further or is the score? I did actually, and that was one of the reasons why I had the two conditions in the first place, to see whether there's any difference between, there were some, some claims and they're not very well tested to date. Whether you see, you know, a text or if you see an image, that same question, whether the language that you speak, you know, it may affect the way you think for speaking, but it may not affect your conceptualization as such as how you see the world around you, so to speak. But what I found here in this study, there were no consistent differences. So whatever, like there were some, as you point out, there were some where, but the difference was, was quite great between the two conditions, but they didn't, there didn't seem to be any pattern to that. So they were really just outliers that sometimes it happens that it's conceptualized one way in one condition and the other in another. And overall, when you look at the overall results, it was actually remarkably consistent to be perfectly candid. I did not even expect anything, like anyone need is consistent. But, and also it was a limitation of this study that it's really, you know, for, to test that, I would have to give participants a purely non-linguistic task to test that, but because they had to generate adjectives here, that's a linguistic task. So it's not, you know, that was a probably a bit of a clumsy attempt to have a look at that, but we didn't find anything so it serves us right. But yeah, to answer your questions, we didn't find any, any pattern there, but there are some studies that, that look into that. I can think of a Ramos and Roberson study from 2010. I'm happy to give the, the, the reference that was looking at like how the stimuli are administered, whether that has any effect on the results. And they found that it, it appears to have some effect. Yes. But it does need further studies. Yes. How did you manage so many participants? How did I manage to meet? How did you get so many participants? This was an online study. Okay. This was online because that was, that's what I had to trade in and see the pilots were in person and it made them, it had made it very clear that if I won the numbers, it's just not going to work. It's going to take the rest of my life. So no, these were, the, I mean, there's a bit of a trading with the, with the online. I mean, what I didn't mention is that almost 200 participants responses were not taken into account because they didn't like, there was either something dodgy or they, they gave a native language, especially in Hebrew, that was an issue that I actually had quite, like a good number of respondents, Hebrew speakers, except that they were not native speakers. And it's actually quite common that they're proficient Hebrew speakers, but that's not their native language. But yeah, that's how. And it was utilized, like all kinds of networks. And I got ethical clearance to even use social media and everything. So it was not easy in Australia, ethical clearance, not easy. So I was very happy about that. Yes. Did you somehow control for which other languages people spoke? Because you could speak German and French. Yes, to some degree, we did. We see, ideally, I would love to have made it a condition that they have to be monolingual or control for the language. But these numbers, it was simply not, not, not feasible to control through. What we know is that most of the vast majority of the participants spoke other languages. I think there's a correlation between being part of an online study. Plus it was, I mean, when we talk about, you know, when we're talking about like German or Hungarian participants, I mean, in Hungary it's compulsory to, you know, by the time you finish primary school, you have to speak two languages. So it's, you know, it will be a bit of a mission to find monolinguals. But you would have to know what they speak. Yes. Yeah. Now, in this study, we only control for whether they do or don't. And yeah, the majority of them did speak some other language. Yes. But sorry, can I just say one more thing? That is like one of the reasons that the native language had to be the target language that was a condition. So we did control for that. And I really was looking for, you know, whether, like effects in the native language. So that's what I mentioned at the beginning that a lot of the studies actually used another language to test that, which I found a bit strange. But next time. Yes. Back to them about the loop test, like gender would be assigned to the figures or whatever. So do you happen to know anything about new words in languages and which gender is assigned to them and if there's any patterns or? It's very interesting because talking about new words, I mean, my favorite, this is just my observation that German appears to assign newta to most of the new words or borrowed words may not be the case. Sorry, correct me if I'm wrong. Yeah, this is, I don't have the data to bug this up. But interestingly, for, like Spanish is probably my favorite, where the two varieties, you know how there's Iberian Spanish, the Spanish spoken in Spain, and then there's Latin American Spanish. Computer in Spain is el ordenador, el masculine. In Latin America, same language, la computadora. It's the same thing, same language, but it's actually assigned to two. Yeah, but still, I mean, it's quite, I mean, talking about new, new words, I mean, yeah, yeah, very, I agree. Yeah, it will be very interesting to see. And they do not have the same gender. Yeah, it would be certainly a very interesting study to have a look at. In my dialect, we have a lot of feminine words, but in what they're speaking also, the more normalized version is basically masculine and phased out. Yeah, yeah, I know. There's a new study that's just come out this year about Norwegian, the two versions, the Buchmann and, you know, Skiere, comparing the grammatical general of that, like the two, how apparently they're quite different, because they were trying to see, you know, like without cultural differences, whether just the language produces differences, and their findings are actually very similar to mine. Yeah, that there's, there's, there's no, certainly no positive correlation. So it's interesting. Yes. So did you look at what countries the native speakers are from, like especially English and Spanish, since it's so, but those are just a widespread, and did you look at that as a factor? Well, I didn't really look at that as a factor. It was, it was checked, as in that was controlled for, that they had to be from a, from a country where the, the languages are, you know, a first language, and it's the native language, and it's the main language spoken. They were, I mean, obviously they were transfer English, you know, the majority of the participants came from Australia. For Spanish, the majority of them came from Spain, and then, you know, with Argentina and in Mexico, close second. But yeah, I mean, I've got all the, you know, I've got the numbers. No, we didn't look at any correlation, whether there's, whether that made a difference. Because with the other languages, I mean, with, you know, Hungarian, you know, all little Hungarians that kind of live in the one country. And with Germans as well, I think we had all up, all up two Austrians. Everyone was from Germany. So, yeah. As I said, we have the data. It's something that we can, we can pike a little bit later and have a look whether there's any correlation there. Yes. Just with a part of the showing people an object and asking them to assign a test, as an assignment agenda, like a bug test, is it possible to invent words and ask people to assign a agenda without giving them any meaning? There's a correlation between sounds of the words. Yes, which would be interesting. Well, not necessarily to deal with gender, but you heard of the kiki and buba kind of thing that, you know, how, how, you know, something sounds. Have you, have you not? It's about how like if you, oh, I don't have a, I don't have a thingy. But does anyone have a marker? Sorry. Sorry, but this has to do with, don't worry, I'll just, oh, thank you. Yeah, there was, I mean, this goes back to the 20s. And this was when a psychologist came up with this. And then there are many, many studies on this that, you know, when they ask participants, sorry, which one is kiki and which one is buba, which one do you think is kiki? Kiki, buba. The agreement on that appears is like almost, you know, universal, that this is like, this one is kiki here, and this one is buba. So there appear to be some correlation between the form of the word or the sound of the word and, you know, associations that we make with them. So to, yeah, yeah. With, with gender, yeah, with gender, I'm not aware of a study that's, that's looked into that. In fact, what I'm trying to, and I'm, you know, a tiny bit of criticism of my field, which I probably shouldn't do, is that as I said, that really, when I said that it gets as exotic as Italian, I wasn't really exaggerating or joking even. They're kind of getting a little bit samey, and looking at the same thing over and over and over again. But to be fair, there's really still no consensus on what's going on. That's why what I was trying to do is, first of all, compare more than two languages. That's why I went for five. So, because it's a bit more reliable to draw any kind of conclusion when you compare more. I tried to go for more participants and that sort of thing to get a clearer picture of what may be going on so we can move on to interesting stuff like that, for example. And to, you know, Atlantic, Bantu languages and all that, and have a look at noun classes and the correlations that they may offer up. Yes? I mean, moving to those languages, I mean, most speakers of these languages are at least bilingual. You know, for instance, for many Bantu languages spoken in Anglophone, East Africa, English, and one or several Bantu and other languages. And, you know, for the Atlantic languages, many of them are spoken in areas where Roman languages are the official languages, or Portuguese-based creole. So, you always would have multilingualism that involves different gender types. Yes, yes. So, just trying to imagine, you know, what would be an interesting hypothesis for such a study. So, would, would you expect all the languages to influence... Well, that would be very interesting to see. For that, we would have... Or would you, would you imagine different language conditions? So, you know, depending on which language... Well, it would be depending on the usage patterns of these languages, for example, as in, like, in what situations they use to, to, like, how much they are used, like, how, how much they're, like, I don't know, do they code switch between one and the other all the time or, or are there certain situations, like, I go to the market and use X and then I go home and use Y and then I go to school and then use Z. I mean, are there such... These would all have to be taken into account, but obviously that wouldn't, that wouldn't need an even finer grain analysis than this one. But everything's possible. So, one of the things is that, that can, that is part of the statistical measure, right, to see how the weighting of, so one of the issues of use, right, has something to do with frequency, right. So, if you use something a lot, if you have, if you speak a number of languages that have gender systems and you use that a lot and there should be more influence because that's your habitual way of, well, if you have something where you don't have that. So, it would be important in the design of the study to look at the balance of the two and find comparisons, right. And then in the study itself, when you can assess, for example, frequency of usage to a certain thing, then you can throw this in as a factor in the, in the statistical analysis. And that would be, would have to be done in these kind of contexts. Yeah. But that's, I mean, that can be controlled, I mean, it can be, it has to be controlled for it. That's the most important thing. Well, I mean, in a similar study, we, you know, Chelsea's, we ruled out that we can control for language frequency. So, you know, we have that now as a continuous variable because we cannot control for that, you know. But it's a continuous variable. Yeah. That's what I mean. That's, you have a continuous variable. So, you control for it. You have a counterpoint. Yeah. In the analysis. Yes. That's what you do. So, that's the most important thing. So, you need to assess that. Yes, take it. Yeah. That's the same for German. You have a pattern, and the pattern goes against exactly that. Yeah. German doesn't have the hazard, too, that it's morphological. There is a pattern that goes around with them. Yes. It goes back to the conversation. I think what would be interesting on the African languages is, in part, you're containing this study, because in the description. But the other thing, of course, which is interesting is to look at what seems like the semantic parameters encoded in these non-classes. Yes. So this, here, your semantics is very clear. You have gender, but that is independent language. But then you could go to the non-class systems and say, OK, what's in there? Gender, precisely, it's not in there, so that would make an interest. But it's also in terms of the tendency. Well, animacy, clearly, is in a very important picture. Size or shape often plays a role. Then there is attitudes, derogatory, like, you know, ugly, or not so nice, or beautiful. And that correlates with shape sometimes. So it's harder, I guess, to get the semantics filtered out. But in principle, I think, would be really, really exciting. Oh, yes. So look at those. And then you can go back to these languages, which don't have any, or very much less so in the system, and see whether you get an inverse correlation between what seems to be encoded and the semantics of it. So I think we talked about it. I think the language is really exciting for that, because the semantics are so different. But the grammar side is actually quite similar. So you have the morphological marking of the agreement, the non-classification system. So it would be really nice to look at that. I keep repeating myself that I couldn't agree more. I mean, that's why I'm here. Because I think it would be a very exciting thing to do, to have a look at these languages. And yeah, I mean, to, as I said, baby steps, I broke out of the Indo-European. So let's jump into it. Let's take it further and go to Bantu, Atlantic, or any of these exciting languages. I couldn't agree more. Any further questions? No. Well, if you would like to continue the discussion, please join us at drinks after this at the Institute of Education over there. I will drag Roland there. I'll be in front of him. So please come and join us and talk more about the topic and your ideas. So thanks again, Roland. Thank you very much. Thank you so much.