University of Pennsylvania
ScholarlyCommons
IRCS Technical Reports Series Institute for Research in Cognitive Science
January 2000
Turning the Tables: Language and Spatial
Reasoning*
Peggy Li
University of Pennsylvania
Lila Gleitman
University of Pennsylvania, gleitman@cattel.psych.upenn.edu
Follow this and additional works at: http://repository.upenn.edu/ircs_reports
University of Pennsylvania Institute for Research in Cognitive Science Technical Report No. IRCS-00-03.
This paper is posted at ScholarlyCommons. http://repository.upenn.edu/ircs_reports/34
For more information, please contact libraryrepository@pobox.upenn.edu.
Li, Peggy and Gleitman, Lila, "Turning the Tables: Language and Spatial Reasoning*" (2000). IRCS Technical Reports Series. 34.
http://repository.upenn.edu/ircs_reports/34
Turning the Tables: Language and Spatial Reasoning*
Abstract
This paper investigates possible influences of the lexical resources of individual languages on the conceptual
organization and reasoning processes of their users. That there are such powerful and pervasive influences of
language on thought is the thesis of the Whorf-Sapir linguistic relativity hypothesis which, after a lengthy
period in intellectual limbo, has recently returned to prominence in the anthropological, linguistic, and
psycholinguistic literatures. Our point of departure is an influential group of cross-linguistic studies that
appear to show that spatial reasoning is strongly affected by the spatial lexicon in everyday use in a community
(Brown and Levinson, 1993b; Pederson et al., 1998). Specifically, certain groups use an absolute spatialcoordinate
system to refer to directions and positions even within small and nearby regions ("to the north of
that coconut tree") whereas English uses a relative, body-oriented system ("to the left of that tree"). The prior
findings have been that users of these two types of spatial systems solve rotation problems in different ways,
ways predicted by the language-particular lexicons. The present studies reproduce these different problemsolving
strategies in monolingual speakers of English by manipulating landmark cues, suggesting that the prior
results were not language effects at all. The results are discussed as buttressing the view that linguistic
idiosyncracies do not materially restrict the thought processes of their users.
Comments
University of Pennsylvania Institute for Research in Cognitive Science Technical Report No. IRCS-00-03.
This technical report is available at ScholarlyCommons: http://repository.upenn.edu/ircs_reports/34
1
Turning the Tables: Language and Spatial Reasoning*
Peggy Li and Lila Gleitman
University of Pennsylvania
(Submitted December 1999)
Running head: LANGUAGE AND SPATIAL REASONING
_____________
*We thank Kathleen Hoffman, Leslie Koo, Aharon Charnov, Lorraine Daley, Jill Kleczko, and
Aurit Lazerus for helping in data collection. We especially thank Kathleen Hoffman for getting us
started on the project. We also thank Henry Gleitman, Robin Clark, and the members of the
CHEESE seminar for important theoretical and methodological contributions. The research was
supported by a predoctoral Fellowship to the first author from the Institute for Research in
Cognitive Science, University of Pennsylvania, by NSF Grant # SBR8920230, and by NIH Grant #
1R01HD375071.
Requests for reprints should be sent to P. Li or L. R. Gleitman, Institute for Research in Cognitive
Science, University of Pennsylvania, 3401 Walnut St. Suite 400A, Philadelphia, PA. 19104. The email
addresses for the authors are pegs@psych.upenn.edu and gleitman@psych.upenn.edu.
2
Abstract
This paper investigates possible influences of the lexical resources of individual languages on the
conceptual organization and reasoning processes of their users. That there are such powerful and
pervasive influences of language on thought is the thesis of the Whorf-Sapir linguistic relativity
hypothesis which, after a lengthy period in intellectual limbo, has recently returned to prominence
in the anthropological, linguistic, and psycholinguistic literatures. Our point of departure is an
influential group of cross-linguistic studies that appear to show that spatial reasoning is strongly
affected by the spatial lexicon in everyday use in a community (Brown and Levinson, 1993b;
Pederson et al., 1998). Specifically, certain groups use an absolute spatial-coordinate system to
refer to directions and positions even within small and nearby regions ("to the north of that coconut
tree") whereas English uses a relative, body-oriented system ("to the left of that tree"). The prior
findings have been that users of these two types of spatial systems solve rotation problems in
different ways, ways predicted by the language-particular lexicons. The present studies reproduce
these different problem-solving strategies in monolingual speakers of English by manipulating
landmark cues, suggesting that the prior results were not language effects at all. The results are
discussed as buttressing the view that linguistic idiosyncracies do not materially restrict the thought
processes of their users.
3
I. Introduction
Language has means for making reference to the objects, relations, properties, and events
that populate our everyday world. Commonsensically, the relevant linguistic categories and
structures are more-or-less straightforward mappings from a preexisting conceptual space,
programmed into our biological nature. This perspective would begin to account for the fact that
the grammars and lexicons of all languages are broadly similar, despite historical isolation and
cultural disparities among them; moreover, that the language learning functions for young speciesmembers
look about the same across languages.
But having assigned the language identities to underlying conceptual identities among their
users, what are we to make of the — less pervasive, but also real — linguistic differences among
languages? Could it be that the situation is symmetrical? That insofar as languages do differ from
each other, there are corresponding differences in the modes of thought of their users? More
marvelously by far, could the linguistic differences be the original causes of distinctions in the way
peoples categorize and reason? Benjamin Whorf and Edward Sapir offer a positive answer to such
questions. In Sapir’s words:
Human beings do not live in the objective world alone, nor alone in the world of social
activity as ordinarily understood, but are very much at the mercy of the particular language
which has become the medium of expression...the “real world” is to a large extent
unconsciously built up on the language habits of the group (E. Sapir as quoted by Whorf
1956; p. 134).
The popularity of this position has diminished considerably in academic favor in the last half
of this century, for two main reasons. First, the universalist position of Chomskian linguistics, with
its potential for explaining the similarities of language learning in children all over the world,
captured the imagination of a generation of scholars. Second, a series of experimental studies
4
documenting the independence of hue and brightness perception from linguistic color-naming
practices seemed to settle the controversy in favor of the universalists (see particularly Berlin and
Kay, 1969; Heider, 1972; Heider and Oliver, 1972; Jameson and Hurvich, 1978).
Recently, however, a number of discussions and experimental studies have reawakened
interest in the question of how language may influence and shape thought. Rightly, the studies of
color vision and naming, elegant and compelling as they were, have been judged too topically
narrow as the basis for writing off the position as a whole. Even more important, effects of
language use, or indeed any learning effects at all, would be least likely in peripheral, low-level
perceptual processes such as hue discrimination. Therefore they do not constitute a fair test (Lucy,
1996).
Many recent studies have focused on a more central perceptual domain: commonalities and
differences in how languages treat spatial properties and relations (Jackendoff, 1996; Talmy, 1978).
To be sure, ultimately linguistic-spatial categories must be built upon a universal perceptual base
originating in brain structure shared by all nonpathological humans (Landau and Jackendoff, 1993).
But within these constraints there is room for languages to differ. This possibility is instantiated in
several crosslinguistic differences in spatial encoding (Brown and Levinson, 1992, 1993a, 1993b;
Bowerman, 1996a, 1996b). Children appear to find it easy to acquire whatever spatial-linguistic
categories their language makes available in its simplest vocabulary and phraseology (Bowerman,
1996a, 1996b; Choi and Bowerman, 1991) and to tailor their speech to accommodate to the domains
of applicability of the language-specific terms (Slobin, 1991).
We pursue here the further question of whether such linguistic differences in the mapping of
space onto language impact the ways that members of a speech community come to conceptualize
the world, as Whorf and Sapir would have it. Do the differences in how people talk result in
5
differences in how they think? Specifically, could the cross-linguistically observed differences in
spatial categorization influence nonlinguistic spatial categorization? Recently, several
commentators have posited that language differences in the spatial domain actually do have such
nonlinguistic effects on category formation and category deployment in tasks requiring spatial
reasoning.
Our point of departure for the current studies comes from a major comparative crosslinguistic
research inquiry focusing on the relationship between linguistic patterning and spatial
reasoning (Brown and Levinson, 1993b; Pederson, Danziger, Wilkins, Levinson, Kita, and Senet
1998). There are in general three ways in which languages express spatial regions and orientations:
in terms of the inherent properties of the objects themselves (“the front of the house,” “ the nose of
the plane”), their cardinal positions (“west of Cleveland”), or their positions relative to the
orientation of the speaker or listener (“on your left,” “to the right of the toolshed”). Most languages
have the formal resources to make reference to spatial arrays in all of these ways. But as Pederson
et al. have documented, in most cases a given speech community favors one of them, at least for
small-scale description. Though in English we could say “Give me the spoon that’s northeast of
your teacup,” this sounds pretty ludicrous. We are strongly inclined to use relative terminology
(“...to the left of your teacup”) instead.
Conceptually, the first step in the cross-linguistic project has been to document the
vocabulary by which speakers typically describe the positions and movements of objects in space.
Standardized procedures were established for eliciting these spatial descriptions from the native
speakers of a broad range of languages, living in both small-scale traditional (e.g., Mayans,
Austronesians) and large-scale (Japanese, Dutch) cultural communities: see Pederson et al., 1998,
for a description of these procedures. The second step was to devise experimental procedures
6
which required subjects of differing linguistic background to observe spatial arrays, and then to
identify or reconstruct “the same” array after being reoriented, usually by 180 degrees. The
question under experimental review is whether incommensurabilities in the spatial vocabularies of
the subject populations predict differences in their performance on the identification and
reconstruction tasks after reorientation.
Absolute versus Relative spatial encoding systems
The major spatio-linguistic distinction studied by Pederson et al. is between absolute versus
relative encoding of spatial directions and locations. For example, Los Angeles is always to the
west of New York, independent of the position of an observer; hence west is an absolute spatial
term. In contrast, for an observer in Cincinnati facing north, Los Angeles is to her left and New
York is to her right; but if she turns to face south, Los Angeles is to her right. Hence left is a term
describing location relative to the body-orientation of the observer herself.
As Pederson et al. have described, most languages with which we are familiar favor relative
spatial terminology for describing small-scale spatial layouts. But for many other language
communities, the facts about everyday speech conventions are otherwise. An example described in
detail in Brown and Levinson (1992) is Tzeltal, a language spoken by about 15,000 Mayans in the
area of Municipo Tenejapa, in Chiapas, Mexico. Their village is on a hill. In Brown and
Levinson’s words,
...there is a system of ‘uphill’/’downhill’ orientation that is fundamental to the spatial
system...based on the overall inclination of the terrain of Tenejapa from high South to low
North, so that [the term for] ‘uphill’ (and correspondingly, ‘downhill’...) [make] primary
reference to the actual inclination of the land...the terms may be used on the flat to refer to
cardinal orientations, or prototypical ‘uphill’ direction. This system then replaces our use of
left/right in many contexts: when there are two objects oriented such that one is to the South
of the other, it can be referred to as the ‘uphill’ object...Now, curiously, this system of
North/South alignment is not complemented by a similar differentiation of the orthogonal.
There is a named orthogonal...but the term is indifferent as to whether it refers to East or
West; what it really means is ‘transverse to the incline.’ So there is a three-way distinction.
7
(p. 596, 1992).
An experimental paradigm: Lining up the animals
Pederson et al. studied spatial reasoning cross-linguistically in several ways, all variations
on a single procedural theme: The subjects memorize the positions of items in an array shown to
them. The array is then removed. After a brief delay, the subjects are turned around (usually, 180
degrees) and asked to recall the original array so as to calculate a response. For example, in the
variant we will use in the experiment presented below, three left-right symmetrical toy animals are
lined up facing the same way on a table top (for a schematic depiction of the procedure, see Figure
1, adapted from Brown and Levinson, 1993b, Pederson et al., 1998). The subject memorizes this
array and then, after a brief delay interval, is moved to another table. This second table is oriented
180 degrees from the original. The subject is now handed the three original animals in random
order, and asked to position them in “the same way as before.”
FIGURE 1 ABOUT HERE
Cross-linguistic outcomes of these studies for Tenejapan and Dutch subject groups are
graphically shown in Figure 2, adapted from Brown and Levinson, 1995. The Figure plots the
distribution of absolute responses by subjects from the two languages.1
These results can be
summarized as follows: The overwhelming majority of speakers of languages favoring absolute
terms consistently rearranged the animals such that if they had on the first table been going north,
they went north after rotation as well. Whereas speakers of languages favoring relative terms
overwhelmingly often rearranged the animals such that if they had been going left on the first table,
they went left after rotation as well. The difference is that what is north does not vary under
rotation, while what is left certainly does.
FIGURE 2 ABOUT HERE
8
It looks, then, as though a language distinction (absolute versus relative spatial terminology)
is influencing reasoning in a very dramatic and straightforward way. However, it is possible that
some third variable that differs between the subject populations is responsible both for the linguistic
difference between them, and for the way they habitually go about solving spatial tasks.
Specifically, each language population in the Brown et al. experiments was tested in its own
community under the social and geographical frame-of-reference context that commonly obtain
there. For example, the Tenejapan population was tested on its hill, out of doors, near a largish
rectangular house. The Dutch population, presumably more used to a school-like situation, was
tested indoors in a laboratory room. Reasonable enough. But this means that only part of the
required experimentation were done, for two factors were varying at once – the language and the
frame of reference in which the spatial task was to be solved. Which one caused the
characteristically different behavior of these groups? To find out, it is necessary to determine
whether a single linguistic group would change its reasoning style if the spatial-contextual
conditions of experimentation were changed. We report on such a set of manipulations below.
Because the results do reveal powerful effects of the conditions of test, we will end by suggesting
nonlinguistic (or “non-Whorfian”) interpretations for them.
II. Spatial reasoning in varying frames of reference: An experimental review
Subjects for all the experiments that we now report were drawn from a single cultural and
linguistic subgroup: monolingual native-English speaking undergraduates at the University of
Pennsylvania. The experimental question was whether we could induce Tenejapan-like and Dutchlike
spatial reasoning behavior in this single population by appropriate changes of the spatial
contexts in which they are tested. For comparability, each of the five experiments employed the
line-up-the-animals task of Pederson et al., 1998.
9
The rotation paradigm: Line up the animals
As just noted, we adopted the ingenious method designed by the Pederson et al. group, here
described in further detail. The materials and procedure are shown in Figures 1 and 3 (adapted
from Pederson et al, 1998). As Figure 1 indicates, the subject was first seated on a swivel chair at a
table (the “Stimulus table”) and asked to study an array of three toy animals (out of 5 animals in the
total test-set of animals) that had been placed there in advance. Subjects were also asked to name
each animal to assure that there was a consensual name for each. Because these were a toy
dinosaur, lion, dog, rabbit, and elephant, there was no disagreement about the labels. Each toy was
symmetrical along its longitudinal axis and was approximately 10 cm long. Both in practice and
test trials, the experimenter always set up the animals so that their noses pointed in the same
cardinal direction, either north or south, depending on the trial (equivalently, in the testing
circumstances, either to the left or to the right). The subject was instructed to study the array for as
long as he liked. Now the animals were scooped up by the experimenter. In an initial practice trial,
the subjects were then handed the three animals and asked to set them up again (“make it the
same”), again on the Stimulus Table (all of our subjects always did this correctly).
Now the experiment proper began. The experimenter set up three of the animals on the
Stimulus Table, as the subject watched. The subject studied this new array as long as he liked,
followed by a 30-second delay. The subject was then swiveled on her chair 180 degrees to face the
Recall Table, which was empty, and handed the three animals in random order. She was told to
“make it the same.” This procedure was repeated (of course, changing the particular animals and
their arrangement on each trial) with each subject five times. About 15% of the time subjects asked
for clarification (of what we meant by “the same”). The experimenter blandly responded “Just
make it the same” and, improbably enough, the subject then always said “OK” and carried out the
10
task.2
FIGURE 3 ABOUT HERE
As schematized in Figure 3, there are two correct ways that subjects could reconstruct the
array on the Recall Table: The left-hand column of animals is the relative solution; the right-hand
column is the absolute solution. With vanishingly rare exceptions, each of our subjects always
chose one of them and individual subjects were consistent across the five trials as to their style of
solution. We now describe three frame-of-reference conditions under which (different) subject
groups of monolingual English speakers were tested.
Experiment 1: Landmarks in the reference world (Blinds-Up/Blinds-Down)
This experiment, and those that follow, we altered the context in which subjects carried out
the line-up-the-animals task by adding implicit landmark cues of various kinds. For after all, the
results found for the Dutch and Tenejapan subjects of Pederson et al. might be attributable to the
differential availability of such landmark information in a featureless laboratory room versus
complex landscape.
FIGURES 4 and 5 ABOUT HERE
Subjects, materials, and procedure: Twenty subjects participated, 10 in each of
two landmark conditions. In both these conditions, subjects were tested in a laboratory room which
was essentially featureless except for a floor-to-ceiling window at one side. As shown in Figure 5,
the testing tables were set up in such a way as to be similar to the placement of the house that had
been visible to Brown and Levinson’s Tenejapan subjects (Figure 4). One half of the subjects were
tested in this room with the blinds pulled down, so that they could not see what lay behind the
window. For the other 10 subjects, the blinds were in their raised position. Under this latter
condition, the subjects if they looked toward the window would view the familiar sight of the
11
university library that lay across the street from the testing laboratory. No mention of the state of
the blinds or of landmarks was made to any of the subjects.
Results: The results are shown in Figure 6. The subjects in the Blinds-Down
condition behaved much as the Dutch subjects in Brown and Levinson (see Figure 2 for
comparison). The subjects in the Blinds-Up condition behaved differently, yielding a U-shaped
distribution that lies somewhere between the prior Dutch and Tenejapan results. Even with the few
subjects tested in each condition, the difference between Blinds-Up and Blinds-Down subjects
approaches significance using the same evaluative instrument used in the prior studies by Brown
and Levinson (Mann-Whitney U-test, p = .056).
FIGURE 6 ABOUT HERE
Experiment 2: Strengthening the landmark cues by going Outdoors
Although the experiment just presented in its Blinds-Up condition added a landmark cue
analogous to the house that was visible to the Tenejapans, raising the blinds by no means
reproduced the rich landmark conditions of an outdoor landscape. We therefore now performed an
additional manipulation which was truer to what we can surmise about the Tenejapan testing
conditions.
Subjects, materials, and procedure: Ten new subjects, again undergraduates at the
University of Pennsylvania, were tested. We found an area of the Penn campus that roughly
reproduced surface landmarks of the Tenejapan testing ground, though it was flattish rather than
hilly. This was a large grassy area, with buildings and roads visible, as schematized in Figure 7. It
being inconvenient to set out tables on this bumpy ground, we laid out large towels as substitute
tables. And we used slightly larger animals so that they would be in reasonable scale to these
towels. In all other regards, the experiment was just as in Experiment 1 (and just as in Brown and
12
Levinson, 1993b).
FIGURE 7 ABOUT HERE
Results: The results of this manipulation are shown in Figure 8 (again, see Figure 1
for comparison).3
As inspection of the Figure shows, the subjects now evidenced a bias toward
absolute responses, much like the Tenejapan subjects of Brown and Levinson and Pederson et al.
To evaluate this finding statistically, we compared the effects for the Blinds-Down (or no landmark)
condition against the present one. Again using the Mann-Whitney U-test, the difference between
these conditions was highly reliable (p = .006). Notice, then, that the difference previously obtained
by Brown and Levinson for Dutch versus Tenejapan speakers (Figure 1) is reproduced here between
groups of Americans when they essentially have no landmarks (and thus act like Dutchmen) versus
when they have strong landmark cues (and thus act like Tenejapans).
FIGURE 8 ABOUT HERE
Experiment 3: Controlled landmark cues (Absolute/Relative Ducks)
In the experiment just described, we looked at the effects of ambient landmark cues in the
visible world. And we found that these biased our subjects toward absolute interpretations of the
spatial reasoning task, as compared to the relative bias of other subjects who were tested in the
featureless (Blinds-Down) condition of Experiment 1. Yet we cannot say that the landmarks out of
doors on the Penn campus matched those of the Tenejapan testing situation, or were greater or lesser
in degree of richness or informativeness. Readers with an eye for detail, in fact, upon examining
Figure 8 will have noticed that the Americans in the outdoor condition were not quite so absolute in
their performances as the Tenejapans had been. Can landmark information, if it is salient enough,
completely determine the degree to which a single population solves spatial problems from a bodyoriented
versus geography-oriented perspective? To find out, we now examined the effects of
13
matched absolutely versus relatively placed landmarks.
Subjects, materials, and procedure: Twenty subjects participated in this
experiment, 10 in each of two conditions. Each of these subjects was tested in the original
laboratory room where Experiment 1 had been conducted. Now the blinds were always up. As
shown in Figure 9, a little toy stood on the Stimulus Table, to the right/south side of the subject
him/her self. This toy was placed there before the subject entered the testing room, and it remained
there, unmentioned and unmoved, for all 5 trials. As the Figure shows, it was a pair of kissing
styrofoam ducks on a paper lake (i.e., a longitudinally symmetrical toy, approximately 17 cm long).
As usual, the subjects’ task was to memorize the positions of the three line-up animals (as in Figure
1) placed on the Stimulus Table as before. No line-up animal was closer than 15 cm to the duck
landmark. When subjects swiveled to the Recall table, they always saw an exact replica of the duck
landmark to one side of it, placed there in advance. For half the subjects (the relative biasing, or
Relative Ducks group), the replica was always on the right of the subject. For the other half (the
absolute biasing, or Absolute Ducks group) the replica was always on the south of the Recall table.
FIGURE 9 ABOUT HERE
Results: The results of this manipulation are graphically shown in Figure 10. Now
the results were exactly like those obtained for the Dutch (relative-responding) and Tenejapan
(absolute-responding) subjects of Brown and Levinson, 1993b (compare Figure 10 and Figure 2).
To evaluate this finding statistically, we again performed the Mann-Whitney U-test and found a
highly reliable difference for the absolute-biasing and relative-biasing conditions of the present
experiment (p = .003).
FIGURE 10 ABOUT HERE
14
III. General discussion
Our investigations were designed to provide further evidence relating to the WhorfSapir
hypothesis, or rather to its descendant theorizing in the current literature of linguistic
relativity. Our starting point was a series of particularly striking demonstrations from Brown and
Levinson, 1993b and Pederson et al., 1998, of a correlation between cross-linguistic spatialterminological
differences and the manner of solving simple tasks of spatial reasoning by speakers
of those languages (Figure 2). Their subjects whose everyday spatial terminology, for relatively
small-scale arrays, was body-oriented solved the rotation problem relatively; whereas subjects
whose terminology (at this grain) encoded cardinal direction solved the same problem absolutely.
This work has been justly acclaimed for the enhanced perspective it is providing on languageculture
relations. At minimum it stands as a welcome antidote to much familiar linguistic and
psychological inquiry that assumes by default that all communities are about the same as, say, a
community of Ivy League sophomores residing together in a Philadelphia dormitory. At the same
time – and as this group of investigators has always been careful to emphasize – the findings are
solely of a correlation between language and reasoning, and as such are hard to interpret causally.
The manipulations we have described were all conducted with English speakers, to see if
the absolute/relative spatial reasoning distinction could be reproduced within a single language
community. If so, this would weaken or negate the claim that language differences, and the putative
constraints these place on reasoning, were the underlying cause of the original effects. Moreover,
we so designed the experiments as to expose an alternative explanation of the original results: Our
subjects behaved absolutely or relatively depending on the presence and strength of the landmark
cues made available to them. As landmark cues were strengthened in three steps: Blinds-Up
(Experiment 1), to Outdoors (Experiment 2), and finally to Absolute Ducks (Experiment 3), the
15
English-speaking group responses moved progressively toward the Tenejapan absolute behavior,
becoming completely equivalent to it under the Absolute Ducks condition. Contrastingly, in the
absence of landmarks (Blinds Down, Experiment 1) or with relative-biasing landmarks (Relative
Ducks, Experiment 3), the subjects behaved like Pederson et al.’s Dutch and Japanese subjects.
When do creatures solve spatial problems “absolutely” versus “relatively”?
The present results suggest that it is not the nature, even the learned nature, of an
individual to solve spatial problems in the absolute versus relative way. Rather, individuals from
the same linguistic-cultural population will solve the same spatial problem differently depending on
the cues made available in the environment. This finding is not really a new one at all. Rather, it
reproduces in detail the findings from prior lines of investigation of spatial reasoning by animals
and young children.
During the 1940's and 1950's, the question was asked whether animals are naturally
“response learners” or “place learners” in regard to navigating through space to arrive at a goal (e.g.,
Tolman, Ritchie and Kalish, 1943). This issue, seen as a crucial one for understanding the nature of
learning, was operationalized within experimental contexts that are formally analogous to the
rotation task as designed by Pederson et al., and replicated in the present article.4
Typically, rats
were trained to find food at one leg of a maze. Then either the rat or the training maze itself would
be rotated 90 or 180 degrees. If the animal continued to turn in the trained direction on the maze
even after rotation, he was judged to have learned only a response (say, “Turn left to get food”). If
instead he responded to extralinguistic cues and thus made a different turn on the rotated trials, he
was assumed to have learned something more sensible; namely, to turn toward the place where the
reward was previously found (say, “Turn North to get food”). But in fact, neither characterization
of rat learning was ever shown to be the correct one. As with the experiments we have just
16
reported, the findings were best understood not by considering possible inherent or learned
tendencies to make the same turn or to go to the same place. Rather, the animals were responsive to
the information provided to them (Blodgett and McCutchan, 1947). In a definitive review of this
literature, Restle wrote:
There is nothing in the nature of a rat which makes it a “place” learner or a “response”
learner. A rat in a maze will use all relevant cues, and the importance of any class of cues
depends on the amount of relevant stimulation provided as well as the sensory capacities of
the animal. In place-response experiments, the importance of place cues depends on the
amount of differential extra-maze stimulation. (1946, p. 226).
In short, thousands of rats gave their lives for very little gain, as the place-learner/response-learner
distinction was apparently misdrawn to start with. When provided with sufficiently rich and stable
landmark cues, any self-respecting rodent will use them and thus be classified as a place learner. In
the absence of landmark cues, the disoriented rat is dependent solely upon his kinesthetic sense and
so repeats his prior response. Related results have been reported for young infants. Though they
often seem to repeat a directional response unheeding of reorientation, if the landmark cues
provided are rich and stable enough even 6-month olds have been shown capable of place learning
(for a review, see Bloch and Morange, 1997).
Why do groups fix on relative or absolute terminology?
Granting now that we produced place learning and response learning in our American
subjects by varying the strength and reliability of landmark cues, how could this bear on the general
tendency of human populations to adopt one or the other strategy in devising and using a spatial
terminology? For after all, we must explain more than why our American subjects choose a
particular strategy of reasoning on the fly, depending on features of the immediate context. The
phenomenon that Pederson et al. have exposed is that communities of speakers choose among
potential linguistic resources and regularly (or “habitually”) prefer either “the spoon to the north of
17
your teacup” or “the spoon to the left of your teacup.” What factor or factors could lead to the longterm
preference for a “place oriented” versus “response oriented” terminology? We will first
mention some global distinctions among the cultural communities whose terminology has been
studied in the cross-language project. Thereafter, we will suggest that a single factor – geographical
cohesion in community life – accounts quite naturally for why social groups develop preferences in
their everyday terminology for referring to regions and directions in space.
Cultural effects on experimental findings: Pederson et al., 1998, reported on the line-upthe
animal task in six languages. Three of these languages (Tzeltal, Longgu, Arrandic) are spoken
in traditional, largely unschooled, linguistic-cultural communities and the results for all three
followed the absolute pattern. Two other languages (Dutch and Japanese), both spoken by
technologically advanced and schooled populations, are those that yielded the relative pattern. (The
sixth language studied, Tamil, is a mixed case that we will discuss just below). In the present
experiments, under the comparable testing condition (Blinds-Down) we again achieved the relative
pattern for a schooled, technologically advanced culture, English-speaking Americans. There has
been extensive discussion of the fact that unschooled and schooled populations often behave quite
differently in experimental tasks, no matter how “simple” these seem to be. Indeed Lucy (1996)
speaking from within the anthropological linguistic community provides a convincing review,
concluding that literacy and school performance have considerable effects on both “patterns of
thought” and “language socialization practices for the inculcation of cultural world-view.” (Lucy,
1996, p. 57). Discussions of the relations of language categories and language practices to school
practice go back to Vygotsky (1978) and are the subject matter of a vigorous and continuing
linguistic anthropological tradition (see also Bernstein, 1971, Greenfield, 1972, Scheiffelin and
Ochs, 1986, Scribner and Cole, 1973). A problem in interpreting the cross-linguistic results on
18
spatial reasoning, then, is that in all the cases on which Pederson et al. reported, they have been
confounded with important cultural variables such as schooling. One might well believe that culture
affects language use and even lexicalization patterns without believing the converse: that language
patterns affect culture.
Pederson (1995; Pederson et al., 1998) made an interesting attempt to unconfound the
linguistic and cultural variables that covaried in the studies that we have discussed thus far by
testing individuals from two communities, both of which spoke the same language, Tamil. Tamil
has the formal resources for both the absolute and relative spatial reference terms. All the same, the
two Tamil-speaking communities varied in their habitual use, one community preferring the
absolute and the other preferring the relative terms for describing small-scale regions and directions.
A test of these two populations might exclude the alternative culture-difference, as opposed to
language-difference, explanation that we have just suggested. However, the results of this
manipulation turned out to be quite unclear. The animals-in-a-row experiment yielded no reliable
difference between the two subpopulations of Tamil speakers. Another experiment, informally
reported in Pederson et al., yields a reliable difference in the predicted direction: The Tamil
population that habitually uses relative terms solved a spatial task relatively, and the population that
habitually uses absolute terms solved the same task absolutely. However, this new experiment in
practice suffered from exactly the same confound it was designed to disentangle. The more
urbanized, schooled, Tamil population was the one exhibiting the relative bias both in speech and in
spatial reasoning (i.e., they behaved more like the Dutch and Japanese). The more rural and
traditional Tamil subpopulation showed the absolute bias that characterized the Tenejapans.
Summarizing, the results of all the animal-in-a-row studies, including our own, seem to
cohere as arising in the presence of supralinguistic cultural differences. These differences predict
19
both the favored linguistic terms and the spatial reasoning patterns, a distinctly “non-Whorfian”
generalization. But why should traditional cultures prefer absolute spatial terminologies and
schooled urbanized cultures prefer the relative ones? “Culture” is too vague and undifferentiated an
explanation of the spatial reasoning phenomena. As we next discuss, all these findings are
suggestive of a more specific solution: Shared landmark cues vary in their stability and familiarity
across communities, much as they do in the various conditions of our experiments.
Geography and the absolute-relative distinction: Recall that the absolute-responding
Tenejapan population resides in a small village on the side of a hill, and employs a 3-term spatial
terminology that alludes to the spatial coordinates of this hill, “downhill” (e.g., “tree standing
downhill of man.”), “uphill,” and “across-the-hill” or, as it were, “athwart.” It is interesting that
spontaneous descriptions from speakers of other spatial-absolute languages studied by this research
group again make reference to what must be rich local landmarks, e.g., “man stands in ‘land of soft
sand’” (Haijlom), “tree standing on side towards sea” (Longgu; these quotes from native speech
appear in Pederson et al., 1998, p. 568). As Pederson (1995) points out, it would not be surprising
at all if persons living in a small and geographically coherent layout would rely heavily on mutually
known local landmarks to orient themselves in space.
Pederson attempted to counter this landmark explanation of absolute solutions in these
spatial tasks by removing his Tamil villagers from their homes and immediate work places: “I
always tried to select an environment for testing, such as the front of the headman’s storehouse,
which would be reasonably removed from the subjects’ daily life.”, p. 55). But as we have
demonstrated, mildly unfamiliar landmarks (as in the Outdoor condition of the present experiments,
see Figure 7) are useable cues and induce absolute responses. Moreover, previously unknown
landmarks, even such adventitious ones as kissing styrofoam ducks (Figure 9), markedly affect
20
spatial reasoning. Regardless of the language spoken, it seems, people will use landmark cues when
these are available. It doesn’t even have to be people: Rats too will take advantage of landmarks to
find hidden food pellets if only these cues are made rich and reliable enough.
Summarizing, we can best understand the findings of the cross-linguistic studies (and the
within-language Tamil study) by noting a cross-cutting factor: the geographical and interpersonal
cohesion of a society. There seems to be no consensual “uphill” that can serve as a reference point
in the very large and shifting communities in which linguistically interacting English, Dutch, or
Japanese speakers generally find themselves. “You’ll find the railroad station just north-east of the
Drexel University parking lot” is not too useful a direction to give the British tourist who has just
arrived in Philadelphia. Body-orientation is the obvious alternative for establishing momentary
spatial coordinates. In contrast, as the Pederson et al. studies appear to show, people like the
Tenejapans, who live in a small, mutually familiar, geographical area, typically use its local
landmarks to devise a spatial coordinate system that makes absolute reference to its stable and
salient features (“uphill,” “inland,” etc.). This is so even when the traditional populations have the
formal linguistic resources for encoding both absolute and relative spatial terminology.
Of course the present authors do not know too much about traditional unschooled cultures
that live in faraway places. Large disparities between investigator and investigated make it difficult
to interpret either naming practices or experimental responses across these cultural divides. On the
other hand, one does not have to go all the way to Chiapas to find communities that favor absolute
spatial terminology. One of us is a native of a highly urbanized culture whose members live and
work and play all crammed together on a skinny little island, about 16 miles long. Culturally
diverse as this community is, its residents share a small, stable, geographical landscape, rich in local
cues. Perhaps this is why their terminology is absolute and --like that of the Tenejapans – makes do
21
with only three terms in habitual use: uptown, downtown, cross-town.
IV. Final Thoughts
We began this discussion by raising the general question of how language categories might,
or might not, influence thought. For the special instance of spatial usage and problem solving just
reviewed, clearly there is a correlation between language, culture, and thought, detailed in the
compelling cross-language studies by Pederson et al. and replicated in our investigations of
American English. However, we also found that the variation in how spatial problems are solved
can be found even when one does not vary the culture or the language of the subject population, but
varies landmark cues instead. Indeed the same differences can be induced in laboratory animals,
again as a function of supplying or withholding landmark information. These findings lead us to
reject the explanation offered by Pederson et al. for their findings. They assert: “The linguistic
system is far more than an available pattern for creating internal representations; to learn to speak a
language successfully requires speakers to develop an appropriate mental representation which is
then available for nonlinguistic purposes.” (1998, p. 586). Elegant as this statement is, we do not
find it defensible on the evidence. Rather, it seems to us, one would want to turn it on its head:
Linguistic systems are merely the available formal and expressive medium that enables speakers to
describe their mental representations of the nonlinguistic world. Depending on the local
circumstances in which human beings find themselves, they select accordingly from this
linguistically available pool of resources for describing regions and directions in space.
Speaking more generally, however, it must be acknowledged that the Whorf-Sapir debate
has been joined, since its inception 50 or so years ago, as a series of local skirmishes: Does
language influence thought about hue? Numerosity? Causality? Objecthood? Or, as in the
present studies, spatial organization? No such particularized investigation can really be decisive as
22
to whether interesting effects of language on thought can be found in other, so far unstudied,
domains. All the same, limited findings do have a way of contingently influencing our broader
beliefs. For example, Pederson et al. suggested that their obtained results for language and space
provided “reason for optimism” that related effects of language on thought would be found in other
domains. Turning the tables, we take the present findings of the robustness of human thought to
variation in linguistic usage patterns to be an optimistic one. All languages have the formal and
expressive power to communicate the ideas, beliefs, and desires of their users. From this vast range
of possibilities, human communities select what they want to say and how they want to say it.
This stance is at bottom the same one that explains why the Elizabethans habitually used terms for
falconry and we do not, and why vacationers at Aspen and Vail find it useful to develop many
words for snow. Transparently, thought influences language use.
23
References
Berlin, B., & Kay, P. (1969). Basic color terms. Berkeley: University of California Press.
Bickel, B. (1997). Spatial operations in deixis, cognition, and culture: Where to orient oneself in
Belhare. In J. Nuyts & E. Pederson E. (Eds.), Language and Conceptualization (pp. 46-83).
Cambridge University Press.
Bloch, H., & Morange, F. (1997). Organising gestures in external space: Orienting and reaching. In
N. Foreman & R. Gillett, A handbook of spatial research paradigms and methodologies, Vol. 1:
Spatial cognition in the child and adult (pp. 15-39). Hove, England UK: Psychology
Press/Erlbaum.
Blodgett H. C., & McCutchan K., (1947). Place vs. response learning in the simple T-maze.
Journal of Experimental Psychology. 37: 412-22.
Blodgett H. C., & McCutchan K., (1948). The relative strength of place and response learning in
the T-maze. Journal of Comparative and Physiological Psychology. 41: 17-24.
Bowerman, M. (1996a). Learning how to structure space for language: A crosslinguistic
perspective. In P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space.
Language, speech, and communication. (pp. 385-436), Cambridge, MA, USA: MIT Press.
Bowerman, M. (1996b). The origins of children's spatial semantic categories: Cognitive versus
linguistic determinants. In J. J. Gumperz & S. C. Levinson (Eds), Rethinking linguistic relativity.
Studies in the social and cultural foundations of language, No. 17. (pp. 145-176) Cambridge,
England UK: Cambridge University Press.
Bowerman, M., de Leon, L., & Choi, S. (1995). Verbs, particles, and spatial semantics: Learning to
talk about spatial actions in typologically different langauges. In E. V. Clark (Ed.), Proceedings of
the Twenty-seventh Annual Child Language Research Forum. (p. 101-110). Stanford, CA: Center
for the Study of Language and Information.
Brown, P., & Levinson, S. (1992). ‘Left’ and ‘Right’ in Tenejapa: Investigating a Linguistic
Conceptual Gap. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung
45(6), 590-611.
Brown, P., & Levinson, S. C. (1993a). 'Uphill' and 'downhill' in Tzeltal. Journal of Linguistic
Anthropology, 3, 46-74.
Brown, P., & Levinson, S. C. (1993b). Linguistic and nonlinguistic coding of spatial arrays:
Explorations in Mayan cognition. Cognitive Anthropology Research Group, Max Planck Institute
for Psycholinguistics, Working paper no. 24.
Choi, S., & Bowerman, M. (1991). Learning to express motion events in English and Korean: The
influence of language specific lexicalization patterns. Cognition, 41, 83-121.
24
Cole, M., & Scribner, S. (1974). Culture and thought: A psychological introduction. New York,
N.Y: John Wiley and Sons.
Gumperz, J. J., & Levinson S. C. (Eds.) (1996). Rethinking Linguistic Relativity. Cambridge:
Cambridge University Press.
Heider, E. R. (1971). “Focal” color areas and the development of color names. Developmental.
Psychology. 4(3), 447-455.
Heider, E. R. (1972). Universals in color naming and memory. Journal of Experimental
Psychology. 93(1), 10-20.
Heider, E. R., & Oliver, D. C. (1972). The structure of the color space in naming and memory for
two languages. Cognitive Psychology. 3(2), 337-354.
Jackendoff, R., & Landau, B. (1991). Spatial language and spatial cognition. In D. J. Napoli &
Kegl, J. A. (Eds), Bridges between psychology and linguistics: A Swarthmore festschrift for Lila
Gleitman. (pp. 145-169). Hillsdale, NJ, USA: Lawrence Erlbaum Associates, Inc.
Jackendoff, R. (1996). The architecture of the linguistic-spatial interface. In P. Bloom, M. A.
Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space. Language, speech, and
communication. (pp. 1-30). Cambridge, MA, USA: MIT Press.
Jameson, D., & Hurvich, L. M. (1978). Dichromatic color language: “Reds” and “greens” don't
look alike but their colors do. Sensory Processes. 2(2), 146-155.
Landau, B. & Jackendoff, R. (1993). “What” and “where” in spatial language and spatial cognition.
Behavioral and Brain Sciences. 16(2), 217-265.
Levinson, S. C. (1996). Language and Space. Annual Review of Anthropology. 25, 353-82
Levinson, S. C. (1996). Frames of reference and Molyneux's question: Cross-linguistic evidence. In
P. Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space. Language,
speech, and communication. (pp. 385-436). Cambridge, MA, USA: MIT Press.
Levinson, S. (1997). From outer to inner space: Linguistic categories and non-linguistic thinking.
In J. Nuyts & E. Pederson (Eds.). Language and Conceptualization (pp. 46-83). Cambridge
University Press.
Levinson, S.C. (1998). Studying spatial conceptualization across cultures: Anthropology and
Cognitive Science. In E. Danziger (Ed.), Language, space and culture. Special Issue of Ethos,
26(1), 7-24.
Li, P., Hoffman, K., & Gleitman L. (1999, July). Relatively speaking, turning Americans into
absolute Tenejapans. Paper presented at the meeting of International Cognitive Linguistics
25
Conference. Stockholm, Sweden.
Lucy, J. A. (1985). Whorf’s view of the linguistic mediation of thought. In E. Mertz & R.
Parmentier (Eds.), Semiotic mediation: sociocultural and psychological perspectives, (pp. 73-97).
New York: Academic Press.
Lucy, J. A. (1996). The scope of linguistic relativity: An analysis and review of empirical research.
In J. J. Gumperz & S. C. Levinson (Eds.), Rethinking linguistic relativity. Studies in the social and
cultural foundations of language, No. 17. (pp. 37-69). Cambridge, England UK: Cambridge
University Press.
Lucy, J. A. (1997). Linguistic relativity. Annual Review of Anthropology, 26, 291-312.
Lucy, J. A. (1997). The linguistics of “color”. In C. Hardin, & L. Maffi (Eds.), Color categories in
thought and language (pp. 320-346). Cambridge, England UK: Cambridge University Press.
Ochs E., & Schieffelin, B. (1984). Language acquisition and language socialization: three
developmental stories and their implications. In R. Shweder & R. LeVine (Eds.), Culture theory:
essays on mind, self, and emotion (pp. 276-320). Cambridge University Press.
Pederson, E. (1993). Geographic and manipulable space in two Tamil linguistic systems. In A. U.
Frank & I. Campari (Eds.), Spatial information theory: A Theoretical Basis for GIS (p. 294-311).
Berlin, Springer.
Pederson, E. (1995). Language as context, language as means: Spatial cognition and habitual
language use. Cognitive Linguistics 6, 33-62.
Pederson, E., Danzigier, E., Wilkins, D., Levinson, S., Kita S. & Senft, G. (1998). Semantic
typology and spatial conceptualization. Language, 74 (3), 557-589.
Reiser, J. J. (1979). Spatial orientation of six-months-old infants. Child Development, 50, 1078-
1087.
Restle, F. (1957). Discrimination of cues in mazes:A resolution of the place-vs.response question.
Psychological Review. 64, 217-28.
Sapir, E. (1949). Selected Writings of Edward Sapir in Language, Culture, and Personality, ed.
David G. Mandelbaum. The University of California Press.
Schieffelin, B., & Ochs, E. (1986a). Language socialization. Annual Review of Anthropology, 15,
163-246.
Scheiffelin, B. B., & Ochs, E. (1991). A cultural perspective on the transition from prelinguistic to
linguistic communication. In M. Woodhead, R. Carr, & P. Light (Eds.), Becoming a person. Child
development in social context, Vol. 1. (pp. 211-229). London, England UK: Routledge.
26
Slobin, D. (1991). Learning to think for speaking: Native language, cognition, and rhetorical style.
Pragmatics 1.7-25.
Slobin, D. (1996). From 'thought and language' to 'thinking for speaking'. In J. J. Gumperz & S.
Levinson, Rethinking Linguistic Relativity (pp. 70-96). Cambridge: Cambridge University Press.
Talmy, L. (1978). Figure and ground in complex sentences. In Joseph Greenberg, et al., Universals
of human language (pp. 625-649). Stanford, CA: Stanford University Press.
Talmy. L. (1983). How language structures space. In H. Pick & L. Acredolo (Eds), Spatial
orientation: Theory, research, and application, (pp. 225-82). New York: Plenum Press.
Tolman, E. C., Ritchie, B. F., & Kalish, D. (1946). Studies in spatial learning: II. Place learning
versus response learning. Journal of Experimental Psychology, 35. 221-229.
Whorf, B. L. (1956). Language, thought and reality: selected writings of Benjamin Lee Whorf, ed.
J. B. Carroll. Cambridge, MA: MIT Press.
Vygostky, L. S. (1978). Mind in society: the development of higher psychological processes, eds.
M. Cole, V. John-Steiner, S. Scribner, & E. Souberman. Cambridge, MA: Harvard University
Press.
27
Notes
1
These authors generally plot percent absolute responses against number of subjects. The resultant
graphs can be visually misleading when the number of subjects in various conditions differs by a
wide margin. Therefore we always replot these results by percent rather than number of subjects.
2
Small procedural details differ between the original Brown and Levinson task and the one
reported here: These authors (see also Pederson et al., 1998) had some distance between the tables,
to avoid their subjects’ sometime tendency to look forward to the recall table while they were at the
stimulus table. With our undergraduate subjects, it was sufficient to tell them not to look toward the
other table. So the Brown et al. subjects were walked, rather than swiveled, between stimulus table
and recall table. This walk-time added a few seconds between training and test, so the delay for
Brown et al. subjects might have been slightly longer than for our subjects. And of course, our toys
were different from their toys.
3
One outlier in this experiment performed in a way that Pederson et al., 1998, observed more often
in their subjects, and which they called “monodirectional.” Such subjects always set up their
animals in a canonical way, regardless of how that relates or does not relate to the original facing
direction on the presentation table. Pederson et al., reasonably enough, excluded such subjects in
presenting their results. Because we had only one subject of this kind, and because the results were
so overwhelming anyhow, we did not exclude her. Another subject on one trial placed two animals
facing each other, as though kissing. We included this odd subject in our results too. Had we
excluded these two individuals, the results shown in Figure 7 would have been even more dramatic.
4
We are indebted to Henry Gleitman (personal communication) for pointing out to us the relevance
of the animal literature in the present regards.
28
Figure Captions
Figure 1. The set up for a trial. Subject first faces the Stimulus Table to memorize an array of
animals. Then the subject is rotated 180 degrees and asked to recreate the array again.
Figure 2. Proportion of absolute choices for speakers of two languages: The majority of the
Tenejapans chose the absolute response for all five trials, while the majority of the Dutch chose the
relative response for all five trials.
Figure 3. Absolute (a) and relative (b) response styles: Subjects who saw the stimulus table set up a
in Figure 1 usually responded in one of these two ways.
Figure 4. Adapted from Brown & Levinson 1993. Subjects were tested outdoors next to a house.
Figure 5. The Library is to the south of the tables just as was the house in Figure 4.
Figure 6. Subjects in the Blinds-Down condition predominantly chose the relative response. In the
Blinds-Up condition, most subjects preferred either all absolute responses or all relative responses.
Figure 7. The subjects were tested outdoors where there are various buildings to the south and
southeast.
Figure 8. Contrasting the Outdoor condition with the indoor Blinds Down condition. Subjects in
the Outdoors condition resemble the Tenejapans while subjects in the Blinds Down condition
resemble the Dutch.
Figure 9. Duck toys are placed on the sides of the table before the subject enters the room.
Figure 10. The subjects’ preference for absolute or relative response is predicted by the positioning
of the ducks.
(Bird’s Eye View)
Subject
Stimulus Table Recall Table
Figure 1
Brown &Levinson (1993b)
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5
Number of Absolute Trials
%ofSubjects
Dutch
Tenejapans
Figure 2
Subject
Stimulus Table Recall Table
Subject
Stimulus Table Recall Table
(Bird’s Eye View)
(a) (b)
Figure 3
1 2
House
N (Downhill)
S (Uphill)
Tenejapans
Figure 4
1 2
Window
Walnut St
Library
N
S
(room)
Condition 2: Indoor BLIND-UP
Placing Americans in a setting like the Tenejapans.
Condition 1: Indoor BLINDS-DOWN
Placing Americans in a setting like the Dutch.
Figure 5
Experiment 1
Blinds Up vs. Down
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5
Number of Absolute Trials
%ofSubjects
IRCS Blinds-Down
IRCS Blinds-Up
Figure 6
N
S
Church
Building
Locust Walk
Open
Field
1 2
Building
Figure 7
Experiment 2
Outdoors
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5
Number of Absolute Trials
%ofSubjects IRCS Blinds-Down
Outdoors
Figure 8
Duck Ponds on the sides of tables as “landmarks.”
N
S
Condition 1:
Relative Biasing
1 2 1 2
Condition 2:
Absolute Biasing
Figure 9
0
10
20
30
40
50
60
70
80
90
100
%ofSubjects
0 1 2 3 4 5
Number of Absolute Trials
Relative Bias
Absolute Bias
Experiment 3
Duck Pond on Tables
Figure 10