As space, along with time, is a fundamental defining our physical world, we use a good chunk of the language we speak each day to declare or reflect upon the relationships we observe within the space around us. These can be relationships we hold ourselves with space, such as “I’m in class” or “I’m by the psych. building,” or relationships we observe between other people and objects within space, such as “she is the one standing on the left of Joe in that picture” or “your keys are behind the laptop.” As natural and instinctive as these relationships seem to us, however, we do not all describe them in the same way. Each language indeed approaches and encodes these spatial relations through a different frame of reference—a coordinate system—that views the situation from a different perspective. Recognizing these differences, scientists have been wondering for decades of their significance: could such differences in linguistic codings of space result in different spatial perceptions crosslinguistically? An answer to this would provide significant support for the Whorfian hypothesis that language shapes thought. If language could influence even the human perception of space—with all its physical, observable characteristics, the possibility that it can mold our perception of less tangible concepts seem very hard to reject. After all, we do not have much else to base our imagination and perception of time, love, freedom, or justice on.
The empirical evidence on spatial relations and frames of reference currently suggests that linguistic frames of reference influence nonlinguistic conceptualization of space to propose this thought-shaping property of language as true. After a brief overview of frames of references and the current progress in categorizing languages under them, these supporting sevidence will be presented, along with a summary of and response to the counter-arguments and counter-evidence for this theory. A concrete plan for future exploratory research on bilinguals will be presented at the end.
Frames of reference: definition and overarching categories
First defined by Gestalt psychologists, a frame of reference (FOR) is a specific coordinate system through which people determine and describe the locations of objects in relation to one another. According to Levinson from the Max Planck Institute for Psycholinguistics in Netherlands, there are three distinct FORs across languages for the depiction of spatial orientation and relations: the relative, the intrinsic, and the absolute FOR. The relative FOR, aptly termed the egocentric or body-based FOR, observes direction and relations from the speaker’s perspective, while the intrinsic and the absolute FORs are both relative to an landmark or cue outside of the observer/speaker. For this reason, they are sometimes classified under one same overarching “allocentric” or “geocentric” FOR. The two differ, however, in that the intrinsic FOR is centered around an external object’s perspective, while the absolute FOR typically employs fixed, cardinal bearings such as north/south/east/west or uphill/downhill (use specifically by Tzeltal speakers). Describing the same spatial relationship of a child approaching a TV, a language coding space under the relative FOR would produce “The child is to the left of the TV” for what would be “The child is to the left of the TV(‘s own front)” in the intrinsic FOR, or “The child is south of the TV” in the absolute FOR. Of the three, the absolute FOR remains the most stable, while the relative FOR reverses as soon as the speaker is rotated 180 degrees (what is formerly “left” now becomes “right”). Some languages, such as Dutch and English, allow for depictions of space through all three FOR (though Dutch and English speakers often speak only in terms of the relative, reserving the absolute for describing large-scale geographical locations). Other languages, however, only possess mechanisms for spatial depiction through one single FOR. For instance, speakers of Guugu Yimithirr—an indigenous Australian language—have available in their language only the absolute FOR, and refer even to body parts as “south leg” (Majid 2004). Of the world’s languages, Levinson and Wilkins have extensively studied and categorized approximately 20 for FOR preferences in representation of small-scale, everyday spatial relationships. Japanese join Dutch and English speakers under the predominantly relative category within this data, while Yukatek Mayan, Tzeltal, Hai||om, and Tenejapan Mayans speakers, among others, have been shown to prefer the absolute FOR.
Possible causes of linguistic FORs’ influence upon spatial conceptualization:
As Asifa Majid and his colleagues point out in a 2004 paper, Whorfian effects “need be neither magical nor radical.” Indeed, it makes perfect sense for language to influence conceptualization in general, and conceptualization of space in particular, if through nothing other than forcing speakers to pay attention to the specific features of their surroundings necessary for producing everyday language. Encoding space through the absolute FOR, for example, a language requires its speakers to constantly keep track of theirs and other objects’ cardinal bearings as they move through space, in order to communicate through language. Because of their language, relative-language and absolute-language speakers then focus on different details within the same space, and are therefore said to “perceive” this space differently. Majid called the general process “perceptual tuning” through attention. Another possible Whorfian cause is “re-representation,” or the recoding of perception to match linguistic patterns. In this case, this would suggest that even if speakers across languages all perceive space through one same FOR intially, they would eventually recode their preceptions into their language’s specific FOR for more efficient retrieval and use as they accumulate more information. As the mind tends also to assign concepts of some similarity into the same category (Bowerman & Choi 2003; Gentner 2003), language can also be perceived as a similarity that leads the brain to group together and perceive different spatial relations from the same nonlinguistic FOR, if it depicts these situations under the same linguistic FOR.
Supporting evidence: spatial strategies and learning abilities
In 1996, Daniel Haun and his colleagues compared Dutch speakers of relative FOR with Akhoe Hai||om speakers of absolute FOR to demonstrate a correlation between a speaker’s linguistic FOR preference with his nonlinguistic FOR preference in spatial observation and learning. The researchers first asked subjects to watch the research “finds” a hidden block under a cup (the “Hiding” cup), then tried to condition each subject to find the “Finding” cup in situations where the Hiding and Finding cups maintained relative, intrinsic, or absolute position. Rotating the sujects between watching and finding to reverse relative perspective, Haun et al. found that Dutch speakers mostly searched the cup that maintained relative postion first (left-most cup, if Hiding cup was left-most), and learned faster under relative finding conditions, while Hai||om speakers did so in absolute conditions (southern-most cup, of Hiding cup was southern). This reveals a nonlinguistic preference of the predominant linguistic FOR in spatial strategies. Observing this correlation in both adults and children of both groups , Haun eliminated age and cognitive development as possible factors shaping nonlinguistic FORs, and reasoned the correlation as a Whorfian effect of language’s FORs upon thought and problem-solving in space.
Comparing speakers from another predominantly absolute language, the Tenejapan Mayans, against Dutch speakers, Brown and Levinson (1993) replicated the same nonlinguistic-linguistic relationship in speakers’ preferred FOR to further improve this hypothesis’ empirical support. When rotated and asked to reproduce an animal arrangements (“Make them the same”) in their corresponding language, Dutch speakers arranged the animals so that their noses (fronts) face the same relative direction as before, while the Tenejapan speakers preserved the cardinal direction the animals face instead (north).
Supporting evidence: spatial memory
Levinson (2003) asserted a similar correlation between linguistic and nonlinguistic preferences in memory of spatial orientation, observing that Dutch speakers memorized the orientation of two distinguishable dots on a card shown to them relatively, while Tzeltal speakers did so absolutely. Levinson also discovered that Dutch and Tzeltal speakers memorized path and directions through their language’s corresponding FOR. When asked to point out in a toy maze the path they have seen a toy man traveled previously, Dutch speakers again overwhelmingly responded with the path that preserved relative coordinates, while Tzeltal speakers mostly answered with the path that preserved the absolute FOR. This suggested that people not only describe space with their language’s FOR linguistically, but also observe space from the same FOR perceptually. Another study by Levinson revealed the same pattern in transitive inferences of spatial relations. Researchers first showed the participants a cube and a cone on a table-top, then a cube and a cylinder, then set a cone on an empty table and asked each subject to place a cylinder next to the cone. Here, Dutch speakers mostly arranged the two solids to preserve relative orientation (if cube was right of cone, and cylinder right of cube: cylinder placed right of cone), while Tzeltal speakers placed the cylinder down to preserve absolute orientation isntead (if cube was south of cone, and cylinder south of cube: cylinder placed south of cone).
Supporting evidence: co-speech hand gestures
The language’s FOR is preserved not also in non-linguistic conceptualization, but also in co-speech gestures. Haviland (1998) found that Tzeltal and Guugu Yimithirr speakers, typically encoding directions from an absolute FOR, represent an object moving west with hand motions towards the west, no matter which direction they are facing. They gesture towards the left, for example, if they are facing north, but gesture right if they are facing south. On the other hand, relative-language speakers of English, Japanese and Dutch observe and represent object movements with hand gestures predominantly along their left/right axis. Speakers of intrinsic languages similarly maintain the linguistic FOR in their gestures, indicating movements’ directions from the mover’s perspective instead of their own. They gesture away from their body, instead of from left to right or from north to south (Haviland 1998).
Refuting arguments:
Of course, the above studies only reliably demonstrate a correlation between the linguistic and the nonlinguistic FOR; they strongly suggest, but do not in any way prove language and the preferred linguistic FOR as the cause for conceptualization of space through this same FOR. Fortunately, they also demonstrate the reversed causal effect—individual perception shaping the language of space—as implausible. As mentioned above, speakers of the same language are shown to prefer the same linguistic FORs for the 20+ languages analyzed. For spatial perception to cause certain linguistic FOR preferences, every single person speaking the same language in a country must think from the same FOR. Except for an overarching cultural influence or a shared biological gene triggering the same FOR, this is too much of a coincidence to be a valid argument.
If the cause for certain linguistic and nonlinguistic FOR preferences is biological, it would also only make sense as a universal cause, not as a gene shared coincidentally by only members of a specific community. As data has revealed that different languages may encode space from a different FOR, the cause therefore is not the same universally, and therefore seems not to be biological in nature.
We now look at the last case, where the linguistic FOR preference is not a cause, but a co-effect of the perceptive FOR by some overarching, community-oriented influence. Researchers from Max Planck Institute summarized previous arguments for such an external, causal factor into three possible categories: the environment, habitual action, and larger cognitive trends for thought. Even within just the small sample of the 20 analyzed languages, however, no significant trends can be spotted to support any of these categories as the causal factor (Majid 2004).
Majid pointed out that one “environmental” theory hypothesizes that rural communities often lack the cardinal directions needed for the absolute FOR, whereas more sophisticated, mobile urban societies would employ the absolute FOR instead for more flexibility. From just the few countries we have observed, emperical evidence demonstrates otherwise. Dutch and English speakers, for examples, are relative speakers, even though their societies arguably remain two of the most advanced and urbanized in the world. The other environmental hypothesis associates the absolute frame with smaller, more “insular” communities (Li & Gleitman 2002). No pattern from the data supports this, though it does suggests the opposite, that larger, more globalized societies often employ the relative FOR. The pattern across the languages analyzed also does not reflect any pattern in habitual action, specifically mode of subsistence, between languages of similar FORs. And though cognitive trends of individualism vs. collectivism seem reasonable as a cause, analysis according to the definitions of Greenfield et al. determines only Dutch and English speakers as individualists out of the 20 languages, and the rest collectivists (Majid 2004). As Japanese and Yukatek Mexican collectivists also make predominant use of the relative FOR, no concrete generalizations could be made about cognitive trends as a emcompassing determinant for FOR preferences in linguistic coding and nonlinguistic perception of space. This, as Majid points out, of course does not disprove any of these factors as a causal factor. Having thoroughly analyzed only a small amount of the world’s thousands of languages, we also do not have enough data to make any reliable generalizations about the existence or degree of influence of such an encompassing cause.
Refuting evidence:
In 2002, Li and Gleitman claimed to have produced empirical evidence revealing the environment as the overarching shaper of both spatial language and spatial thought, and therefore refuting language’s influence on spatial perception. They emphasized indoor vs. outdoor testing locations as of major significance, and asserted that the previously observed correlation only resulted from testing Dutch speakers indoors while testing indigenous speakers outdoors (due to lack of proper facilities in these poor, rural communities). Using only American college students to keep language constant, Li and Gleitman adapted a version of the animal arrangement test with varied testing conditions—indoors with blinds closed (few landmarks), indoors with blinds raised (some landmarks), and outdoors (many landmarks). The two discovered that, though results under the blinds-closed condition agree with those from previous findings, under the blinds-raised and outdoors condition half the subjects opted to rearrange the animals to maintain the allocentric FOR based on a landmark. Subjects also asked for clarifications of what the researchers meant by “same” much more often in the latter two conditions. This indicated the subjects’ awareness of the landmark-oriented rearrangement possibility, even though they speak and arguably think in the relative FOR. Li and Gleitman reasoned then that we have all three FORs readily accessible, and the environment rather than language determines which one we actually employ. Their results seemed also to have invalidated previous evidence for Whorfian effects, on grounds of unequal testing conditions. Levinson et al. (2002) soon refuted this claim with emperical evidence, however. First, they pointed out that Whorfian results previously obtained from Dutch speakers were all collected under blinds-raised conditions, which according to Li and Gleitman themselves are comparable to the outdoors conditions indigenous speakers were tested under. Replicating previous “indoors” experiments outdoors with Dutch speakers, Levinson found that 95% of the subjects still gave more relative than absolute responses in both conditions, and speculated that Li and Gleitman’s subjects simply second-guessed the two’s simple tasks to produce the irregular results. We of course are still unable to conclusively refute or support Whorfian effects by testing so few languages, but as of now, much more evidence exists to demonstrate rather than disprove language’s shaping effects on spatial strategies and memory.
Possible direction for future research:
Experimenting with bilinguals seems to be the best next step in order to determine the extent of Whorfian effects on spatial conceptualization and perception. Research had shown before that simply changing the “language” a person uses to speak about a concept can change that person’s perceptual approach to that concept. Boroditsky (2001) demonstrated this by studying English-Mandarin bilinguals’ perception of time—the passing of which is described with horizontal directions of “forward/back” in English, but with vertical “up/down” direction in Mandarin. Not only did she discovered that English speakers were faster at determining event order when given horizontal language cues, while Mandarin speakers performed better when given vertical cues, Boroditsky found that English speakers quickly started to behave more Mandarin-like in tests after even brief training in vertical encodings of time. This reveals just how powerful language’s influence could be on thought and perception. We can much more convincingly claim Whorfian effects of language on spatial thought if we could reproduce similar results—that, keeping basic, nonlinguistic ideas of space constant (one person), simply changing the way a person speaks about space could change the way they think about space.
Possible theorical experiment:
For such an experiment, it would be most preferable to compare bilinguals who speak both an allocentric and an egocentric language (say, English-Tzeltal bilinguals for relative/absolute, or English-Mopan Mayan speakers for relative/instrinsic). This study deems it best to compare English speakers to Mopan Mayans from Belize, as the Mopan language allows only for allocentric coding of space (Levinson and Wilkins). Since their language does not provide mechanisms for relative coding at all, we can more safely claim any nonlinguistic relative coding we observe in English-Mopan bilinguals as a result of their English knowledge. Along with 10 English monolinguals and 10 Mopan monolinguals, it is also best if we could find 10 English-native bilinguals as well as 10 Mopan-native bilinguals for a subject pool of 40, and analyze each group’s nonlinguistic FOR preference using a variant of the animal arrangement test. The hypothesis then follows that: if we test each of these groups in their native language (Englisn monolinguals and English-native monolinguals in English, Mopan monolinguals and Mopan-native bilinguals in Mopan), the English monolinguals would prefer the relative FOR and the Mopan the allocentric FOR, while the bilinguals would nonlinguistically encode both from egocentric and allocentric FORs, with the English-native behaving more English-like, and the Mopan-native more Mopan-like. If we test the bilinguals in their respective second language, however, English-native will behave more Mopan-like than during the native-language experiment, while the Mopan-native will become more English-like. In order to perform the second-language experiment, it’s best to immerse the bilinguals in their second-language community for at least a few days (possibly under the guise of cultural-exchange programs in order to prevent the subjects from second-guessing the purpose), conditioning bilinguals into their second-language “mind frame” before testing them for most reliable results.
Practical limitations and suggested remedy:
As most allocentric languages “discovered” are spoken by small, relatively unknown, indigenous communities of low literacy rates, however, it is nearly impossible to find any English natives or indigenous natives who are bilingual for the purpose of this experiment. Until a more viable allocentric language is found, then, the only solution is to train relative English or Dutch speakers to become “bilingual” in both egocentric and allocentric FORs instead. Conditioning them into describing spatial relationships relative to a landmark, researchers could put typically relative speakers into a linguistically intrinsic “mind frame” to see if a nonlinguistic change in spatial conceptualization follows. (Method inspiration: Boroditsky’s training of English speakers to think of time in Mandarin-like vertical metaphors.)
Design:
Say we decide to test English speakers for this experiment. In order to eliminate possible indoors/outdoors differences, the experiment will be performed in a room with a large glass window, under blinds-raised conditions. Participants will mostly be American college students whose native language is English; they can be recruited either through flyers or emails, must be 18 years old or order, and must give their consent to participate in writing. In order observe their most candid FOR preference, we must also prevent the participants from guessing the purpose of our experiments, and therefore will disguise the experiment as a study on learning habits and strategies instead. First, we will replicated the Hiding-Finding cup experiment (Haun et al. 1996) with the subjects, observing and recording their preferred FOR as they are learning to find the cups. Then we will condition these same subjects—speakers of the predominantly relative English language—into observing space intrinsically. For this purpose, we will ask each subject to describe to us each of a set of two-object pictures (a toy house by a little girl, a cup by a TV, a man by a truck, etc.), in which one object is the obvious landmark for the others to be described by (larger, more dominant, has a distinguishable “front,” i.e. the little girl, the TV, the truck). It is predicted that the participants will start out describing from relative perspective (“The man is to the right of the truck,” for a front-profile picture of a man standing by a truck’s driver door). We will inform the participant that this description is “incorrect”, tell them the “correct” answer (“The man is to the left of the truck,” from the perspective of the truck’s own front), move on the the next picture, and recycle picture if necessary as we move through the set, until the participant could “correctly” describe each picture in the set. It is also important to record whether a participant immediately start out describing from the intrinsic FOR, and how long each participant takes to adapt the intrinsic FOR each time.
After conditioning the participants, we then go through the Finding-Hiding cup experiment again with the subjects, explaining this to them as a post-test to determine whether the spatial exercise with the pictures (in truth the conditioning process) helped improve their “finding abilities.” The hypothesis follows that they will now try to “find” from the intrinsic FOR first. We again make note of the first cup each participant turn over to see if it maintained position relatively or intrinsically (to an external landmark) with the Hiding cup, then release the participants after asking each if they would like to be informed of the final results and conclusion in the future.
Possible scenarios: A majority of subjects will
1. Starts out finding from the relative FOR, and ends up in the intrinsic FOR: this agrees with our hypothesis, and demonstrates that language indeed has a powerful effect on shaping the way we conceptualize space. If language could change how we think about something as tangible as space, and after only such brief conditioning, there is no imagining the extent of its influence on every other domain of our perception and conceptualization. Even in this case, though, we should be careful to test other languages of other initial FORs first before making any definite generalizations.
2. Starts out and ends up in the relative FOR: this does not agree with our hypothesis, but it does not refute language as a possible influencing factor affecting spatial concepts, either. It simply reveals the limit of such an influence. This limit makes sense, however. After all, space is a tangible and easily observable idea. Humans need not constantly look to language, especially after only such a brief shift in linguistic FOR, as a way to help characterize space.
3. Starts out and ends up in the intrinsic FOR: this is inconsistent with previous analysis about English’s typically relative FOR. If a majority of subjects respond in this way, the problem may be that the subject pool is too small, or too particular. The experiment must be reperformed on another subject pool before any reliable conclusion can be drawn.
4. Starts out intrinsic, and ends up relative: it seems the participants have second-guessed the purpose of the experiment if this is the case. The experiment must be redesigned—embedded within more complex disguise—and reperformed before any reliable conclusions could be drawn.
Also, if a significant number of participants immediately start out describing the pictures from an intrinsic FOR during the conditioning process, it might be that pictures—reducing the speaker’s involvement in a spatial relationship—automatically prompt subjects to approach them from the intrinsic FOR. If this is the case, another conditioning method must be devised and implemented with another subject pool before any conclusions can be made.
Whatever the case, though, it is difficult to deny language’s influence on the the human cognition of space based on the existing evidence. Regardless of nature’s role and extent of importance in determining our cognitive structures and frame of reference, this demonstrates nurture’s role in general, and language’s role in particular, in shaping our thought and conceptualizations about ourselves and the world we live in.
Works Cited:
Boroditsky, L. (2001). Does Language Shape Thought?: Mandarin and English Speakers' Conceptions of Time. Cognitive Psychology, 43(1), 1-22
Bowerman, M. and Choi, S. Space under construction: language-specific spatial categorization in first language acquisition. Language in Mind (Gentner, D. and Goldin-Meadow, S., eds), MIT Press (2003), 387–427.
Gentner, D. Why we're so smart. Language in Mind (Gentner, D. and Goldin-Meadow, S., eds), MIT Press (2003), 195–235.
Haun, D.B., Rapold, C.J., Call, J., Janzen, G., & Levinson, S.C. (2006). Cognitive cladistics and cultural override in Hominid spatial cognition. Proceedings of the National Academy of Sciences of the United States of America, 103(46), 17568-17573.
Haviland, J.B. (1998) Guugu Yimithirr cardinal directions. Ethos, 26, 25–47.
Levinson, S.C. Frames of reference and Molyneux's question: cross-linguistic evidence. In: P. Bloom et al.Language and Space, MIT Press (1996), 109–169.
Levinson, S.C., Kita, S., Haun, D.B., & Rasch, B.H. (2002). Returning the tables: language affects spatial reasoning. Cognition, 84(2), 155-188.
Li, P. & Gleitman, L. (2002). Turning the tables: language and spatial reasoning. Cognition, 83 (3), 265–294.
Majid, A. (2002). Frames of reference and language concepts. Trends in Cognitive Sciences, 6(12), 503-504.
Majid, A., Bowerman, M., Kita, S., Haun, D., & Levinson, S.C. (2004). Can language restructure cognition? The case for space. Trends in Cognitive Sciences, 8(3), 108-114.
