Through the Language Glass: Why the World Looks Different in Other Languages

by Guy Deutscher

  Why do so many languages develop irregular genders? We don’t know much about the infancy of gender systems, because in most languages the origin of gender markers is entirely opaque.* But the few clues we do have make the ubiquitous irrationality of mature gender systems appear particularly peculiar, because all the signs suggest that in their early days genders were perfectly logical. There are a few languages, especially in Africa, in which the feminine gender marker looks rather like a shortened version of the noun “woman” itself, and the inanimate gender marker resembles the noun “thing.” Likewise, the vegetable gender marker in some Australian languages looks rather similar to the noun . . . “vegetable.” It stands to reason, therefore, that gender markers started out in life as generic nouns such as “woman,” “man,” “thing,” or “vegetable.” And if so, it seems plausible that they would have originally been applied only to women, men, things, and vegetables, respectively.

  But with time the gender markers may start being extended to nouns beyond their original remit, and through a series of such extensions a gender system can quickly be brought out of kilter. In Gurr-goni, for example, the vegetable gender came to include the noun “airplane” through a perfectly natural sequence of little steps: the original “vegetable” gender marker must first have been extended to plants more generally, and hence to all kinds of wooden objects. Since canoes are made of wood, another natural step would have included them in the vegetable gender as well. Since canoes happened to be the main means of transport for the speakers of Gurr-goni, the vegetable gender was then widened to include means of transportation more generally. And so, when the borrowed word erriplen entered the language, it was quite naturally assigned to the vegetable gender. Each step in this chain was natural and made perfect sense in its own local context. But the end result seems entirely arbitrary.

  The Indo-European languages may also have started with a transparent gender system. But suppose, for instance, that the moon came to be included in the masculine gender because he was personified as a male god. Later, the word “month” developed from the word “moon,” so it was only natural that if the moon was a “he” a “month” would also be a “he.” But if so, then words for other time units, such as “day,” can also come to be included in the masculine gender. While each step in this chain of extensions may be perfectly natural in itself, after two or three steps the original logic has become opaque, and so masculine and feminine genders can find themselves applied to a range of inanimate objects for no intelligible reason.

  The worst thing about this loss of transparency is that it is a self-propelling process: the less consistent the system becomes, the easier it is to mess it up even further. Once there are enough nouns with arbitrary genders, children struggling to learn the language may stop expecting to find reliable rules based on the real-world properties of objects, so they may start looking for other types of clues. For example, they can start guessing what gender a noun has on the basis of what it sounds like (if X sounds like Y, and Y is feminine, then maybe X is feminine as well). Incorrect guesses by children may initially be perceived as errors, but with time such errors can stick and so before too long any trace of the original logic can be lost.

  Finally, it is ironic that when a language loses one gender out of three the result may actually increase the waywardness of the system rather than decrease it. Spanish, French, and Italian, for instance, lost the original neuter gender of their Latin forebear, when the neuter coalesced with the masculine. But the result only ensured that all inanimate nouns are randomly assigned to the masculine or feminine genders.

  Nevertheless, the syndrome of genus erraticum is not always an incurable illness in a language. As the history of English can attest, when a language manages to lose not just one gender but two, the result can be a radical overhaul that eliminates the erratic system altogether. Until the eleventh century, English had a full-blown three-gender system just like German. English speakers from the eleventh century would not have understood what Mark Twain was bemoaning in his “Tale of the Fishwife and Its Sad Fate,” since for them a wife (wf) was an “it,” a fish (fisc) was a “he,” whereas fate (wyrd) was a “she.” But all this changed during the twelfth century.

  The collapse of the Old English irregular genders had little to do with improving standards of sex education. The reason was rather that the gender system had critically depended on the doomed system of case endings. Originally, English had a complex case system similar to that of Latin, where nouns and adjectives appeared with different endings depending on their role in the sentence. Nouns of different genders had different sets of such case endings, so one could tell from the endings which gender a noun belonged to. But the system of endings rapidly disintegrated in the century after the Norman Conquest, and once the endings had disappeared, the new generation of speakers hardly had any clues left to tell them which gender each noun was supposed to belong to. These new speakers, who grew up into a language that no longer gave them sufficient information to decide whether a carrot, for example, should be addressed as a “he” or a “she,” fell back on a radical and highly innovative idea, and started to call it an “it” instead. So over a period of just a few generations, the original arbitrary gender system was replaced by a new one with transparent rules, whereby (almost) all inanimate objects came to be referred to simply as “it.”

  Still, a few wily nouns, especially feminine ones, managed to escape the mass sterilization. Mark Twain, who was outraged by the bestowal of femininity upon German turnips, would have been surprised to learn that the same custom was still practiced in England only three centuries earlier. A medicinal manual published in London in 1561, The Most Excellent and Perfecte Homish Apothecarye or Homely Physick Booke for all the Grefes and Diseases of the Bodye, offers the following confection against hoarseness: “He that is become hoorse lately, let him roste a rape [turnip] in ashes or upon the fyre till she be all black, then pare her clene and eate her as warm thou canst.”

  In dialectal varieties of English, some gendered nouns survived for much longer, but in the standard language a great tide of neuters flooded the inanimate world, leaving only a few isolated nouns bobbing about in their femininity. The slow but sure iticization of English can be said to have come to its final mooring on March 20, 2002. For the maritime world, that particular Wednesday seemed no more eventful than any other Wednesday. Lloyd’s List, the newspaper of the shipping industry, published its daily pageful of dispatches on casualties, accidents, and acts of piracy at sea. Among others, it mentioned the ferry Baltic Jet en route from Tallinn to Helsinki, which “had a fire in her port side engine room at 0814, local time”; the tanker Hamilton Energy departed from Port Weller Docks in Canada after “repairs were made to damage suffered when she was in contact with a Saltie. The accident snapped the rudder post and drove her propeller shaft through her gearbox and smashed her engine casting off.” Elsewhere in Canada, a shrimp trawler got stuck in pack ice, but the owner said that “there is a possibility she can be started up and steamed under her own power.” A day, in short, like any other.

  The real ocean-shaking news was reported on a different page, stowed away in the editorial column. Kissed by the punning muse, the editor announced under the headline “Her today, gone tomorrow” that “we have taken the simple yet significant decision to change our style from the start of the next month and start referring to ships as neuter rather than female. It brings this paper into line with most other reputable international business titles.” Reactions from the public were stormy, and the paper was overwhelmed by letters to the editor. An irate Greek reader wrote: “Sir, only a bunch of crusty, out of touch, stuck up Englishmen would dream of trying to change the way we’ve spoken of ships for thousands of years as ‘she.’ Get out of there and go tend to your gardens and hunt foxes, you arrogant ass holes. Sincerely yours, Stephen Komianos.” But not even this silver-tongued plea convinced Lloyd’s List to change her course, and in April 2002 “she” fell by
the quayside.

  GENDER AND THOUGHT

  Languages that treat inanimate objects as “he” or “she” force their speakers to talk about such objects with the same grammatical forms that are applied to men and women. This habit of he-ing and she-ing objects means that an association between an inanimate noun and one of the sexes is shoved down the speakers’ ears whenever they hear the name of this object, and the same association is pushed up their throats whenever they have occasion to mention his or her name themselves. And as anyone whose mother tongue has a gender system will tell you, once the habit has taken hold and the masculine or feminine association has been established, it is very difficult to shake it off. When I speak English, I may say about a bed that “it” is too soft, but I actually feel “she” is too soft. She stays feminine all the way from the lungs up to the glottis and is neutered only when she reaches the tip of the tongue.

  As a basis for serious investigation, however, my professed feelings toward beds hardly constitute reliable evidence. It is not just the anecdotal nature of this information that is the problem, but the fact that I have not provided any proof that the “she” feeling is anything more than tongue-deep—a mere grammatical habit. The automatic association between an inanimate noun and a gendered pronoun does not, in itself, show that the grammatical gender has exercised any deeper effect on the speakers’ thoughts. It does not show, in particular, whether speakers of Hebrew or Spanish, which treat beds as feminine, really associate with beds any womanly properties.

  Over the last century, various experiments have been conducted with the aim of testing precisely this question: Can the grammatical gender of inanimate objects influence speakers’ associations? Probably the first such experiment was conducted at the Moscow Psychological Institute in prerevolutionary Russia. In 1915, fifty people were asked to imagine each day of the week as a particular person, then to describe the person they had pictured for each day. It turned out that all participants envisaged Monday, Tuesday, and Thursday as men but Wednesday, Friday, and Saturday as women. Why should this be so? When asked to explain their choice, many of them could not give a satisfactory answer. But the researchers concluded that the answer could not be unrelated to the fact that the names for Monday, Tuesday, and Thursday have a masculine gender in Russian, whereas Wednesday, Friday, and Saturday are feminine.

  In the 1990s, the psychologist Toshi Konishi conducted an experiment comparing the gender associations of speakers of German and of Spanish. There are quite a few inanimate nouns whose genders in the two languages are reversed. The German air is a she (die Luft) but el aire is he in Spanish; die Brücke (bridge) is also feminine in German but el puente is masculine; and the same goes for clocks, apartments, forks, newspapers, pockets, shoulders, stamps, tickets, violins, the sun, the world, and love. On the other hand, der Apfel is masculine for Germans but la manzana is feminine in Spanish, and so are chairs, brooms, butterflies, keys, mountains, stars, tables, wars, rain, and garbage. Konishi presented a list of such nouns with conflicting genders to German and to Spanish speakers and asked the participants for their opinions on the properties of those nouns: whether they were weak or strong, little or big, and so on. On average, the nouns that are masculine in German but feminine in Spanish (chairs and keys, for example) got higher marks for strength from the Germans, whereas bridges and clocks, which are masculine in Spanish but feminine in German, were judged stronger on average by the Spanish speakers.

  The simple conclusion from such an experiment would be that bridges do have more manly connotations for Spanish speakers than for German speakers. However, one possible objection to this inference is that it may not be the bridge itself that carries such connotations—it may only have been hearing the name together with the masculine article el or un. In this interpretation, when Spanish and German speakers simply look at a bridge, their associations may not be affected at all, and it may be only in the moment of speech, only through the act of saying or hearing the gender marker itself, that a fleeting association with manliness or womanliness is created in the speaker’s mind.

  Is it possible, therefore, to get round the problem and check whether womanly or manly associations for inanimate nouns are present even when the gender markers in the relevant language are not explicitly mentioned? The psychologists Lera Boroditsky and Lauren Schmidt tried to do this by repeating a similar experiment with Spanish and German speakers, but this time communicating with the participants in English rather than in their respective mother tongue. Although the experiment was conducted in a language that treats inanimate objects uniformly as “it,” the Spanish and German speakers still showed marked differences in the attributes they chose for the relevant objects. German speakers tended to describe bridges as beautiful, elegant, fragile, peaceful, pretty, and slender; Spanish speakers as big, dangerous, long, strong, sturdy, towering.

  A more radical way of bypassing the problem was designed by the psychologist Maria Sera and her colleagues, who compared the reactions of French and Spanish speakers but used pictures of objects instead of words. As two closely related languages, French and Spanish mostly agree on gender, but there are still sufficiently many nouns that diverge: the fork, for instance, is la fourchette in French but el tenedor in Spanish, and so are cars (la voiture, el carro) and bananas (la banana, el plátano); on the other hand, French beds are masculine (le lit) but Spanish ones are feminine (la cama), and the same goes for clouds (le nuage, la nube) and butterflies (le papillon, la mariposa). The participants in this experiment were asked to help in the preparation of a movie in which some everyday objects come to life. Their task was to choose the appropriate voice for each object in the movie. They were shown a series of pictures, and for each one they were asked to choose between a man’s voice and a woman’s voice. Although the names of the objects were never mentioned, when French speakers saw the picture of a fork, most of them wanted her to speak in a woman’s voice, whereas the Spanish speakers tended to choose a male voice for him instead. With the picture of the bed, the situation was reversed.

  The experiments described above are certainly suggestive. They seem to show that the grammatical gender of an inanimate object affects the properties that speakers associate with this object. Or at least what the experiments demonstrate is that the grammatical gender affects the responses when speakers are actively requested to indulge their imaginations and come up with associations for such an object. But this last point is in fact a serious weakness. All the experiments described so far suffer from one underlying problem, namely that they forced the participants to exercise their imaginations. A skeptic could argue with some justification that the only thing the experiments proved was that grammatical genders affect associations when the participants are coerced unnaturally to dream up properties for various inanimate objects. In the worst case, one could parody what might be going on in a participant’s mind as something like: “Here I am being asked all sorts of ridiculous questions. Now they want me to think up properties for a bridge—goodness me, what’s next? Well, I’d better come up with something, otherwise they’ll never let me go home. So I’ll say X.” Under such circumstances, the first property that comes to a Spanish speaker’s mind is indeed likely to be more manlike than womanlike. In other words, if you force Spanish speakers to be on-the-spot poets, and extract properties of bridges out of them, the gender system will indeed affect their choice of properties. But how can we tell whether the masculine gender has any influence on speakers’ spontaneous conceptions of bridges, even outside such exercises in poetry on demand?

  In the 1960s, the linguist Susan Ervin tried to downplay the element of creativity with an experiment that involved Italian speakers. She relied on the fact that Italian has very diffuse dialects, so even a native speaker would not be at all surprised to encounter entirely unfamiliar words in an unfamiliar dialect. Ervin invented a list of nonsense words that sounded as if they could be the dialectal terms for various objects. Some of these ended in -o (masculine) and
the others in -a (feminine). She wanted to check what associations these words would evoke in Italian speakers but did not want the participants to realize that they were indulging in creative imagination. So she told them they were going to see a list of words from an Italian dialect that they didn’t know, and she pretended that the aim of the experiment was to check whether people could guess correctly the properties of words merely by the way they sound. The participants tended to attribute to the -o words similar properties to those they attributed to men (strong, big, ugly), whereas the -a words tended to be described with properties that were also used for women (weak, little, pretty). Ervin’s experiment showed that associations were affected by the grammatical gender even when the participants did not realize they were indulging in creative imagination and assumed that the question before them had a correct solution. But while this experiment went some way toward overcoming the problem of subjective judgments, it still did not solve the problem completely, since even if the participants were not aware of being coerced to produce associations on demand, in practice this is exactly what they were required to do.

  In fact, it is difficult to imagine how one could design any experiment that would completely bypass the influence of subjective judgments. For the task requires nothing less than having one’s cake and eating it too: how can any experiment measure whether grammatical genders exert an influence on speakers’ associations, without soliciting these speakers for their associations? A few years ago, Lera Boroditsky and Lauren Schmidt found a way to do exactly that. They asked a group of Spanish speakers and a group of German speakers to participate in a memory game (which was conducted wholly in English, in order to avoid any explicit mention of the genders). The participants were given a list of two dozen inanimate objects, and for each of these objects, they had to memorize a person’s name. For example, “apple” had the name Patrick associated with it, and “bridge” had the name Claudia. The participants were given a fixed period of time to memorize the names associated with the objects, then tested on how well they had managed. A statistical analysis of the results showed that they were better at remembering the assigned names when the gender of the object matched the sex of the person, and that they found it more difficult to remember the names when the gender of the object clashed with the sex of the person. For example, Spanish speakers found it easier to remember the name associated with “apple” (la manzana) if it was Patricia rather than Patrick, and they found it easier to remember the name for a bridge (el puento) if it was Claudio rather than Claudia.