You are on page 1of 10

691548

research-article2017
PSSXXX10.1177/0956797617691548Cooperrider et al.Ordered Line

Research Article

Psychological Science

Where Does the Ordered Line Come 2017, Vol. 28(5) 599­–608
© The Author(s) 2017
Reprints and permissions:
From? Evidence From a Culture of sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797617691548
https://doi.org/10.1177/0956797617691548

Papua New Guinea www.psychologicalscience.org/PS

Kensy Cooperrider1, Tyler Marghetis2, and Rafael Núñez3


1
Department of Psychology, University of Chicago; 2Department of Psychological and Brain Sciences,
Indiana University–Bloomington; and 3Department of Cognitive Science, University of California, San Diego

Abstract
Number lines, calendars, and measuring sticks all represent order along some dimension (e.g., magnitude) as position
on a line. In high-literacy, industrialized societies, this principle of spatial organization—linear order—is a fixture of
visual culture and everyday cognition. But what are the principle’s origins, and how did it become such a fixture?
Three studies investigated intuitions about linear order in the Yupno, members of a culture of Papua New Guinea
that lacks conventional representations involving ordered lines, and in U.S. undergraduates. Presented with cards
representing differing sizes and numerosities, both groups arranged them using linear order or sometimes spatial
grouping, a competing principle. But whereas the U.S. participants produced ordered lines in all tasks, strongly
favoring a left-to-right format, the Yupno produced them less consistently, and with variable orientations. Conventional
linear representations are thus not necessary to spark the intuition of linear order—which may have other experiential
sources—but they nonetheless regiment when and how the principle is used.

Keywords
spatial concepts, visual culture, order, number, cognitive diversity, open materials

Received 7/26/16; Revision accepted 1/11/17

Two millennia ago, Aristotle proposed a scheme for (De Soto, London, & Handel, 1965). In fact, adults in such
understanding nature: a ladder on which different beings cultures operate with implicitly linear mental representa-
were ordered, with the simplest on the bottom rung and tions of number (e.g., Dehaene, Bossini, & Giraux, 1993;
the most sophisticated on top (Archibald, 2014). This Restle, 1970), time (e.g., Ishihara, Keller, Rossetti, & Prinz,
ancient scheme embodies the principle of linear order, 2008; Núñez & Cooperrider, 2013; Price, 2009), and other
according to which the order of elements with respect to ordered concepts (Gevers, Reynvoet, & Fias, 2003; van
some dimension—such as simplicity, magnitude, or Dijck & Fias, 2011). Implicit notions of linear order are
sequence—is represented by their positions in a linear even evident in children before they can read (McCrink &
arrangement. Today, the principle of linear order is Opfer, 2014; Nuerk et al., 2015).
instantiated in calendars, volume sliders, number lines, Linear order thus appears to be a fixture of conven-
and scientific graphs, and even implicitly in written text tional representational practices and of human cognition.
(Ingold, 2007; Tversky, 2011). It is a taken-for-granted But what are the origins of this principle, and how has it
cornerstone of visual culture, at least in industrialized become such a fixture? One possibility is that representa-
societies. tional practices exhibiting linear order are ubiquitous
The principle of linear order also appears to occupy a because of the centrality of the principle in cognition.
privileged place in human cognition. Within high-literacy,
industrialized societies, the ordered line is deployed spon-
Corresponding Author:
taneously as an organizational strategy (Bergen & Lau, Kensy Cooperrider, University of Chicago, Department of Psychology,
2012; Fuhrman & Boroditsky, 2010; Tversky, Kugelmass, & 5848 S. University Ave., Chicago, IL 60637
Winter, 1991) and serves as a powerful reasoning tool E-mail: kensy@uchicago.edu
600 Cooperrider et al.

Conversely, the principle may be central in cognition between each additional stimulus and those extremes,
because of the ubiquity of representational practices that inviting categorical responding. Second, to the extent
exhibit it. Current evidence hints at a more complex rela- that intuitions about linear order are evident in cultures
tionship that goes beyond these chicken-or-egg extremes. that lack conventional representations involving the prin-
For instance, in cultures whose members share an intu- ciple, further questions remain about when and how the
ition about the orientation of ordered lines, this intuition principle might be used. Would people in such cultures
mirrors representational practices, such as writing (e.g., use linear order to the same extent as people in the
Bergen & Lau, 2012; Shaki, Fischer, & Petrusic, 2009; United States, or would they instead favor spatial group-
Tversky et al., 1991; Zebian, 2005). ing? And if they were to use linear order, would their
The precise relationship between conventional repre- lines take a consistent spatial form?
sentations and intuitions about linear order is hard to To address these questions, we examined spontane-
disentangle in high-literacy, industrialized cultures. In ous spatial representations of abstract relations in the
such societies, it is difficult to rule out influence from Yupno and in U.S. undergraduates. In U.S. society, as in
representational conventions, even in preliterate children all contemporary industrialized societies, conventional
(Nuerk et al., 2015). The scant work that has been done representations exhibiting linear order abound. Yupno
with more isolated indigenous groups has focused on culture, by contrast, lacks calendrical systems, linear
one instantiation of linear order in particular—the num- measurement tools, writing, or other conventional prac-
ber line—and has presented a mixed picture. Dehaene, tices involving ordered lines, and many Yupno adults
Izard, Spelke, and Pica (2008a) had Mundurucu (Brazil) have not had formal schooling (for additional informa-
participants perform a number-line estimation task in tion, see the Supplemental Material available online and
which they were asked to indicate the appropriate loca- Wassmann, 1993). We used an open-ended task in which
tion of numerical stimuli (e.g., sets of dots) on a graphical participants were asked to arrange sets of cards differing
line. Many Mundurucu participants mapped the stimuli in the size or numerosity represented. This task allowed
on the line according to their magnitude, a mapping con- us to see when people recruited linear order, if at all, and
sistent with the principle of linear order. But a significant how they implemented it. In Study 1, we used a basic
subset of participants produced bimodal responses, map- version of the spatial-arrangement task; in Studies 2 and
ping small numerosities to one endpoint and large 3, we altered the procedure to highlight the possibility of
numerosities to the other (Dehaene, Izard, Spelke, & construing the sets categorically. These different versions
Pica, 2008b). Núñez, Cooperrider, and Wassmann (2012) of the task allowed us to probe the strength of intuitions
performed the same task with unschooled Yupno (Papua about linear order in the two groups.2
New Guinea) participants. Strikingly, all responded in a
categorical fashion, using only the endpoints of the line.
Responses by participants in both cultural groups thus Study 1
call to mind another strategy for organizing abstract rela-
tions: spatial grouping. Both linear order and spatial
Method
grouping have been argued to be fundamental spatial Participants.  Yupno adults (n = 13) participated in the
organization principles in both graphical practices and task in exchange for a small gift. They were tested in the
reasoning (De Soto et al., 1965; Tversky, 2011).1 How- village of Gua, in the Yupno valley of Papua New Guinea.
ever, the possibility that grouping might rival or even At the time of data collection, the valley was not con-
eclipse linear order in some cultures has not been nected by road to any urban centers, and it did not have
investigated. electricity, telecommunications services, or tourist infra-
Key questions about the origins of linear order thus structure. Area residents make their living by subsistence
remain open. First, it is unclear whether members of cul- farming (see the Supplemental Material for further ethno-
tures that lack representational practices involving linear graphic details). One Yupno participant was eliminated
order do indeed have intuitions about the principle. The for failing to understand the instructions. Thus, the final
number-line estimation task cannot provide satisfying sample consisted of 12 participants (6 women). The
answers to this question because of its constraints: It majority of these participants (n = 9) had no formal
imposes a particular spatial organizational strategy (i.e., a schooling whatsoever, and all were “unschooled” accord-
line), in a particular orientation (i.e., left to right), and ing to the cutoff used in previous research in this popula-
assesses whether people can make sense of it. Imposing tion (i.e., no education beyond sixth grade; Núñez et al.,
this graphical representation could spark intuitions about 2012).3 Undergraduates (n = 16) at the University of Cali-
the ordered line; on the other hand, first establishing the fornia, San Diego (UCSD), participated in exchange for
extremes of the set may foreground the relationship course credit (12 women).
Ordered Line 601

Materials. The arrangement task was adapted from the Supplemental Material for the card sets and lists
Tversky et al. (1991). In the original task, participants used). Participants’ arrangements were documented with
placed stickers on square pieces of paper. In our version, video or still photographs, taken from a position facing
to avoid providing a rectilinear frame, we instead asked the participants.
participants to arrange circular cards on a large circular Between the two blocks, participants completed a
sheet. Further, our version of the task used cards rather brief task based on one used previously to induce inter-
than stickers, to allow for adjustments and rearranging. val representations (in which exact spatial distance rep-
Each set of cards represented a domain that was either resents conceptual distance; Tversky et al., 1991). The
categorical or orderable. The categorical stimuli were of task did not induce interval arrangements in either group,
two types: animals (a picture of a pig or dog) and shapes however, and was not used in our later studies.
(a blob or squiggly line). The orderable stimuli were also
of two types: size (a circle with a size that varied across
Analysis
stimuli) and numerosity (from 1 to 10 dots).4 Each size
card showed a centered blue circle, which ranged in Before analysis, the images of the final arrangements
diameter from 1 cm to 5.5 cm, in increments of 0.5 cm. were cropped so that the participants were not visible.
Each numerosity card showed a pseudorandom arrange- These cropped images were then coded in a random
ment of black dots, each 1 cm in diameter. Half of the order to determine the organizational strategy used (Fig.
sets consisted of four cards, and half consisted of five 1); for linear arrangements, the orientation of the line
cards. For the categorical stimuli, four-card sets always was also coded. An arrangement was considered to
involved two tokens of each category (e.g., two pigs, two exhibit spatial grouping if the cards were put into distinct
dogs), and five-card sets involved two tokens of one piles or if adjacency relations suggested category bound-
category and three of the other. The orderable stimuli aries (e.g., a row with the three pigs on the left and the
consisted of both consecutive sets (e.g., a four-card two dogs on the right). An arrangement was considered
numerosity set with cards depicting 1, 2, 3, and 4 dots) to exhibit linear order if the cards were arranged in a
and nonconsecutive sets (e.g., cards depicting 1, 2, 5, and path that preserved the order of the elements perfectly
9 dots). All stimuli were printed on circular cards (8 cm (or if two adjacent cards were in transposed order, which
in diameter) and laminated. occurred rarely). An arrangement that met none of these
criteria was coded as “other.” Finally, arrangements exhib-
Procedure. Participants were tested individually. They iting linear order were coded for their orientation relative
sat in front of a large blue circular sheet (105 cm in diam- to the participant’s body. The eight possible orientations
eter) and were instructed: “I am going to give you a were rightward, rightward-away, away, leftward-away,
mixed-up group of cards. Look them over and lay them leftward, leftward-toward, toward, and rightward-toward.
out in an organized way.” The instructions were devel- For example, a line increasing (larger size, greater numer-
oped in collaboration with an academic linguist with osity) to the participant’s right was coded as “rightward,”
good Yupno proficiency and a native speaker of Yupno a line increasing in a direction both away from and to the
with good English proficiency; they were expressly right of the participant was coded as “rightward-away,”
designed not to bias participants toward any particular and a line increasing toward the participant’s body was
spatial arrangement. For the Yupno participants, prere- coded as “toward.”
corded instructions spoken in Yupno by a native speaker The first author (K. C.) coded all the arrangements.
were played from a laptop, and a native speaker of Reliability was assessed by having a second, naive coder
Yupno with good English proficiency was present to analyze the entire data set. Reliability was 95% for orga-
facilitate communication. For the U.S. participants, the nizational strategy and 92% for directionality. Disagree-
instructions were presented orally by the experimenter. ments were adjudicated by the second author (T. M.).
The experimenter, positioned to the left and slightly in The adoption of a particular strategy (e.g., linear order)
front of participants, handed the shuffled sets of cards to was analyzed on a trial-by-trial basis using mixed logit
them, one set at a time. Participants arranged 12 sets of models, with fixed effects of cultural group (Yupno vs.
cards, presented in two blocks of 6 sets each. Each block U.S.), stimulus type (e.g., size vs. numerosity), and their
consisted of two categorical sets (one of animals and one interaction, and random effects of both participants and
of shapes), a size set, a brightness set (see note 4), and items, when appropriate. The different types of categorical
two numerosity sets. The sets were presented either in stimuli were collapsed in all analyses. All binary predictors
that order or with the positions of the categorical and (e.g., cultural group) were centered. For models that ana-
numerosity sets switched. Whether participants began lyzed all three levels of stimulus type (i.e., categorical, size,
with categorical sets and ended with numerosity sets or and numerosity), size trials acted as a baseline with which
vice versa was counterbalanced across participants (see the two other stimulus types were compared.
602 Cooperrider et al.

Categorical Stimuli Orderable Stimuli


Animals Size Numerosity

Spatial Grouping
Linear Order

Not
Applicable

Fig. 1.  Representative arrangements illustrating the two spatial organization strategies: spatial group-
ing (top) and linear order (bottom). Categorical stimuli (left) can be grouped but not ordered, whereas
orderable stimuli (right) can be either grouped or ordered. Each example shown here was produced by
a different participant in Study 2. The examples of spatial grouping were produced by Yupno partici-
pants, and the examples of linear order were produced by U.S. participants.

All models used the maximal converging random- Thus, although both groups largely reserved the group-
effects structure justified by the experimental design ing strategy for categorical stimuli, this selective use was
(Barr, Levy, Scheepers, & Tily, 2013; see the Supplemen- stronger among the U.S. participants.
tal Material for model specifications). Models were imple-
mented with the lme4 package (Bates, Maechler, Bolker, Linear order.  We next analyzed the use of the ordered
& Walker, 2015) in the R statistical software environment line, a strategy that could be applied only to the two
(R Core Team, 2015). Reported proportions are model types of orderable stimuli, that is, size and numerosity
estimates, which account for individual- and item-specific stimuli. There were no effects of cultural group (p > .56)
variability. or stimulus type (p > .63). Ordered lines were the domi-
nant response for both stimulus types, for both the Yupno
(size: M = .88, 95% CI = [.76, 1.00]; numerosity: M = .86,
Results 95% CI = [.77, .95]) and the U.S. participants (size: M =
Spatial grouping.  We first analyzed the use of spatial 1.00, 95% CI = [.99, 1.00]; numerosity: M = .99, 95% CI =
grouping, an organizational strategy that may be used for [.97, 1.00]; Fig. 2). Thus, although the U.S. participants
both categorical (animals, shapes) and orderable (size, used linear order numerically more often than did the
numerosity) domains. The Yupno and U.S. participants Yupno, this effect did not reach significance. Analyses of
did not differ overall in their use of grouping (p > .32). the orientations of these ordered lines are presented
This strategy was employed equally often for size stimuli later, after Study 3.
(M = .04, 95% confidence interval, CI = [.02, .06]) and
numerosity stimuli (M = .03, 95% CI = [.02, .04]), p > .34,
Discussion
but significantly more often for categorical stimuli (M =
.68, 95% CI = [.63, .73]), b = 5.12, SEM = 0.95, p < .001. Despite living in a culture that lacks representational
However, a significant interaction, b = −2.76, SEM = 1.37, practices exhibiting linear order, Yupno participants used
p = .044, revealed that this increased use of the grouping this organizational principle spontaneously, to an extent
strategy for categorical stimuli was more pronounced comparable to that of U.S. participants. Such practices,
among the U.S. participants (categorical: M = .77, 95% CI = therefore, are not necessary to spark the intuition that
[.72, .83]; size: M = .01, 95% CI = [.01, .02]; numerosity: order can be represented in a linear arrangement. We
M = .00, 95% CI = [.00, .00]) than among the Yupno (cat- also sought to compare the strength of intuitions about
egorical: M = .56, 95% CI = [.47, .64]; size: M = .08, 95% linear order in the two groups by allowing a competing
CI = [.04, .11]; numerosity: M = .07, 95% CI = [.05, .09]). strategy—grouping—but it was used rarely for the order-
No other effects approached significance (all ps > .3). able stimuli. To create a tension between the strategies,
Ordered Line 603

1.0 procedure. Thus, the final U.S. sample consisted of 17


participants (15 women). None of the participants in
Study 2 had participated in Study 1.
Proportion of Trials Using an Ordered Line

.8 Materials.  As in Study 1, the stimuli were either order-


able or categorical. The orderable stimuli were identical
to those in Study 1 (i.e., size and numerosity stimuli). The
categorical stimuli included the same set of animals from
.6 Study 1; the abstract shapes were replaced with more
U.S. ecologically valid concrete objects (rocks and carrying
Yupno bags). We used eight card sets, each consisting of four
cards. Of the orderable stimuli, half were nonconsecutive
.4
and half were consecutive; the nonconsecutive sets
always preceded the consecutive sets.

.2 Procedure.  The procedure was the same as in Study 1,


except for the addition of a fruit-arranging task at the
beginning. For the fruit task, participants were instructed:
“I am going to give you a mixed-up group of objects.
.0 Look them over and lay them out in an organized way.”
Study 1 Study 2 Study 3
The experimenter then handed the participant the fruits
Fig. 2.  Frequency of use of the linear-order strategy for arranging the in a disorganized handful. Culturally common fruits were
size and numerosity stimuli. Results are shown separately for the two used: oranges, passion fruits, and tree tomatoes for the
cultural groups (U.S., Yupno) in each of the three studies. Error bars Yupno; apples, oranges, and avocados for the U.S. par-
represent ±1 SE (one error bar is too short to be visible). The propor-
tions shown are estimates from a model of all three studies. ticipants. There were two trials, each involving two fruits
each of two types. For example, a Yupno participant
might receive two passion fruits and two tree tomatoes
and thus conduct a better test of whether intuitions about on the first trial and then two oranges and two passion
linear order are stronger in U.S. adults than in the Yupno, fruits on the second.
in Studies 2 and 3 we manipulated the procedure to high- After the fruit-arranging task, participants arranged the
light the possibility of spatial grouping. eight sets of cards, one after another, in a fixed order:
one objects set, one animals set, two size sets, two bright-
ness sets, and two numerosity sets (see the Supplemental
Study 2
Material for the card sets).
In Study 2, participants again arranged sets of cards, but
they began by arranging pieces of fruit. Fruit is commonly Analysis. Coding and analysis were conducted as in
grouped by category in everyday life in both cultures Study 1. Reliability was 94% for organizational strategy
(e.g., in marketplaces), and being asked to arrange it and 95% for directionality.
should call to mind this organizational strategy. We rea-
soned that if the Yupno have a weaker intuition about
linear order than U.S. adults do, they might produce fewer
Results
ordered lines after being primed to group by category. Spatial grouping. There was no effect of cultural
group on frequency of use of the grouping strategy; the
Yupno and U.S. participants used grouping to the same
Method
degree (p > .21). Once again, grouping was used equally
Participants. Yupno adults (n = 19) participated in often for the size stimuli (M = .17, 95% CI = [.13, .21]) and
exchange for a small gift. Five were eliminated for failing the numerosity stimuli (M = .18, 95% CI = [.13, .22]), p >
to understand the instructions. Thus, the final Yupno sam- .94, although for both types of stimuli it was used more
ple consisted of 14 participants (4 women). Most (n = 8) than 4 times as often as in Study 1. As before, grouping
had no exposure to formal schooling whatsoever, and all was used significantly more often for the categorical
were classified as unschooled by the previously used cut- stimuli (M = .78, 95% CI = [.75, .82]), b = 3.50, SEM = 0.65,
off (Núñez et al., 2012). UCSD undergraduates (n = 19) p < .001. Although this increased use for categorical stim-
participated in exchange for course credit. Two were uli was, once again, more pronounced for the U.S. par-
eliminated because of irregularities in the experimental ticipants (categorical: M = .83, 95% CI = [.79, .87]; size:
604 Cooperrider et al.

M = .10, 95% CI = [.07, .14]; numerosity: M = .08, 95% CI = orderable size and numerosity sets). However, there was
[.05, .10]) than for the Yupno participants (categorical: M = just one six-card set for each category, and the orderable
.72, 95% CI = [.66, .78]; size: M = .24, 95% CI = [.17, .31]; sets were nonconsecutive (e.g., the numerosity set
numerosity: M = .30, 95% CI = [.22, .38]), this effect was depicted 1, 2, 3, 5, 6, and 7 dots).
only marginally significant, b = −1.57, SEM = 0.93, p = .09.
No other effects approached significance (all ps > .21). Procedure. Participants began with two trials of the
fruit-arranging task used in Study 2. All participants then
Linear order. There were no effects of cultural group arranged four sets in the following fixed order: objects,
(p > .20) or stimulus type (p > .27) on frequency of use animals, size, and numerosity. For each set, participants
of the linear-order strategy. Ordered lines were the domi- were first given two cards and were instructed, as before,
nant response for both stimulus types, for both the Yupno to “lay them out in an organized way.” They were then
(size: M = .61, 95% CI = [.46, .76]; numerosity: M = .63, given the four remaining cards one at a time. With each
95% CI = [.48, .78]) and the U.S. participants (size: M = additional card, they were instructed: “I am going to give
.82, 95% CI = [.73, .90]; numerosity: M = .90, 95% CI = you another card. Carefully look over the whole group
[.83, .97]; Fig. 2). Thus, although the U.S. participants and lay them out in an organized way.” For the categori-
used linear order numerically more often than the Yupno, cal sets, the first two cards were tokens of the two catego-
this effect did not reach significance. Analyses of the ori- ries; the remaining cards loosely alternated between the
entations of these ordered lines are presented later, after categories. For the orderable sets, the first two cards were
Study 3. the set’s extremes; the remaining cards loosely alternated
between small and large magnitudes (see the Supple-
mental Material for the card sets and the order in which
Study 3
the cards within a set were presented).
In Study 3, we further probed the strength of intuitions
about linear order. In addition to the categorical-priming Analysis.  Participants’ final six-card arrangements were
task from Study 2, we included a procedural change to analyzed in the same way as in Studies 1 and 2. Reliabil-
further highlight the possibility of construing the sets cat- ity was 95% for organizational strategy and 86% for
egorically. In Studies 1 and 2, participants received all the directionality.
cards in a set at once, which invited them to consider the
set’s overall ordinal structure. In Study 3, after first receiv-
Results
ing the extremes of a set, participants received one card
at a time. This procedure foregrounded the relationship Spatial grouping. There was no effect of cultural
of each additional card to the extremes, rather than the group on frequency of use of the grouping strategy (p >
overall structure of the set. In this way it was similar to .21). Once again, spatial grouping was used equally often
the number-line estimation task (e.g., Núñez et al., 2012), for size stimuli (M = .42, 95% CI = [.36, .48]) and numeros-
which begins with an introduction of the line’s endpoints ity stimuli (M = .32, 95% CI = [.26, .38]), p > .37, although
as salient anchors. the overall use of grouping doubled from Study 2. Again,
grouping was used significantly more for categorical
stimuli (M = .76, 95% CI = [.73, .79]), b = 1.59, SEM = 0.57,
Method p = .005. Although this increased use of grouping for
Participants. Yupno adults (n = 13) participated in categorical stimuli was, yet again, more pronounced for
exchange for a small gift. Two were eliminated for failing the U.S. participants (categorical: M = .82, 95% CI = [.80,
to understand the instructions. Thus, the final Yupno .85]; size: M = .32, 95% CI = [.26, .39]; numerosity: M =
sample consisted of 11 participants (5 women). Most .21, 95% CI = [.16, .26]) than for the Yupno participants
(n = 8) had no formal education whatsoever, and all (categorical: M = .68, 95% CI = [.64, .71]; size: M = .55,
were classified as unschooled by the previously used cut- 95% CI = [.49, .61]; numerosity: M = .45, 95% CI = [.39,
off (Núñez et al., 2012). UCSD undergraduates (n = 15) .51]), this effect was only marginally significant, b = −1.84,
participated voluntarily in exchange for course credit. SEM = 0.99, p = .06. No other effects approached signifi-
One was eliminated because of irregularities in the cance (all ps > .21).
experimental procedure. This left a final sample of 14
U.S. participants (8 women). No participants in Study 3 Linear order.  The use of linear order differed between
had participated in either previous study. the cultural groups; the U.S. participants produced sig-
nificantly more linear arrangements (M = .79, 95% CI =
Materials.  The materials were similar to those used in [.63, .95]) than the Yupno participants did (M = .22, 95%
Study 2 (i.e., categorical objects and animals sets and CI = [.04, .41]), b = −21.34, SEM = 6.48, p < .001 (Fig. 2).
Ordered Line 605

Thus, although the Yupno participants produced ordered Linear order


lines on the majority of trials with orderable sets in Stud-
ies 1 and 2, this was no longer the case in Study 3. Use of linear order decreased reliably in later studies
Instead, grouping was the dominant strategy for the (Study 1: M = .94, 95% CI = [.90, .97]; Study 2: M = .75,
Yupno, whereas the ordered line remained the dominant 95% CI = [.68, .81]; Study 3: M = .54, 95% CI = [.40, .68]),
strategy for the U.S. participants. There was also a mar- b = −5.16, SEM = 1.59, p = .001. More important, this
ginal effect of stimulus type, b = 4.22, SEM = 2.30, p = .07; influence of task procedures differed significantly
linear order was used more for the numerosity stimuli between the cultural groups, b = −7.41, SEM = 2.75, p =
(M = .60, 95% CI = [.39, .80]) than for the size stimuli .007. To investigate this interaction, we analyzed the two
(M = .48, 95% CI = [.27, .69]). This, in turn, was driven by cultural groups separately. Among the Yupno, the use of
a marginal interaction between cultural group and stimu- linear order was affected significantly by the procedural
lus type, b = 8.61, SEM = 4.54, p = .06: The Yupno pro- changes across the three studies (Study 1: M = .87, 95%
duced more ordered lines for the numerosity stimuli CI = [.79, .94]; Study 2: M = .61, 95% CI = [.49, .72]; Study
(M = .36, 95% CI = [.02, .69]) than for the size stimuli 3: M = .21, 95% CI = [.05, .37]), b = −7.17, SEM = 2.91,
(M = .09, 95% CI = [.00, .28]), whereas the U.S. participants p = .014. The U.S. participants, by contrast, were largely
produced ordered lines equally often for the size stimuli unaffected by the procedural manipulations, continuing
(M = .79, 95% CI = [.55, 1.00]) and the numerosity stimuli to rely on linear order in every study (Study 1: M = .99,
(M = .79, 95% CI = [.55, 1.00]). We next present analyses 95% CI = [.98, 1.00]; Study 2: M = .86, 95% CI = [.79, .94];
of the orientations of ordered lines in all three studies. Study 3: M = .80, 95% CI = [.64, .95]), p > .36. The context
in which participants were asked to arrange the cards
thus had a selective impact on the Yupno, dampening
Analyses Combining All Studies their use of the ordered line.
To confirm that the two cultural groups were differen-
tially affected by task procedures, we analyzed the data Directionality of ordered lines
from all three studies together. The mixed logit model
was analogous to the models for the individual studies, We next analyzed the direction of increase (i.e., larger
with additional fixed effects for study (1–3) and its inter- size, greater numerosity) in participants’ ordered lines.
actions, and with Study 1 acting as a baseline against This analysis collapsed the data across all the studies
which the other studies were compared. Because the fre- because of the limited number of ordered lines produced
quency of use of linear order did not differ between size by the Yupno in Study 3. Overall, the Yupno produced
and numerosity stimuli in any of the studies, stimulus ordered lines that were highly variable in their orienta-
type was not included as a fixed effect in the analysis of tion, and significantly more variable than those produced
linear order. by the U.S. participants, χ2(1, N = 276) = 92.5, p < .001,
Wallraff rank-sum test of angular distance (Fig. 3). Indeed,
among the U.S. participants, almost all the ordered lines
Spatial grouping (94.9%) were along the left-right axis, and they were
Use of grouping did not differ between the size stimuli tightly concentrated around a rightward orientation, with
(M = .15, 95% CI = [.11, .19]) and the numerosity stimuli a bootstrapped 95% CI spanning only 5° (M = 3.3°, 95%
(M = .11, 95% CI = [.08, .14]), p > .32, but was significantly CI = [6.3°, 1.1°]; circular dispersion: κ = 1.8). By contrast,
greater for the categorical stimuli (M = .73, 95% CI = [.69, nearly half of the Yupno participants’ lines (46.3%) were
.77]), b = 5.04, SEM = 1.04, p < .001. Yet again, this not oriented along the left-right axis. Their lines exhib-
increase was marginally more pronounced for the U.S. ited a broad range of orientations, with a 95% CI span-
participants (categorical: M = .82, 95% CI = [.78, .86]; size: ning 38° (M = 21.4°, 95% CI = [3.4°, 41.3°]; circular
M = .07, 95% CI = [.04, .10]; numerosity: M = .04, 95% dispersion: κ = 0.8). Thus, the U.S. participants produced
CI = [.02, .07]) than for the Yupno participants (categori- ordered lines that were highly consistent in a rightward
cal: M = .62, 95% CI = [.55, .69]; size: M = 0.25, 95% CI = orientation, whereas the Yupno produced ordered lines
[.18, .31]; numerosity: M = .19, 95% CI = [.14, .25]), b = that were oriented more freely around the entire 360°
−2.19, SEM = 1.22, p = .07. There was also a highly signifi- range of possible directions.
cant influence of Study, b = 1.29, SEM = 0.50, p < .01.
grouping was more common in Study 2 (M = .36, 95%
General Discussion
CI = [.31, .42]) than in Study 1 (M = .29, 95% CI = [.24,
.34]), and more common yet again in Study 3 (M = .57, In high-literacy, industrialized societies, the principle of lin-
95% CI = [.50, .63]). No other effects approached signifi- ear order is a cornerstone of both visual culture and every-
cance (ps > .24). day cognition. To illuminate the origins of this principle, we
606 Cooperrider et al.

Yupno Participants U.S. Participants Lifelong experience with graphical practices appears
to regiment intuitions about linear order. By regimenta-
tion, we mean a process in which culture reinforces cer-
tain habits of thinking and behaving, with consequences
for when and how those habits are called upon. For
example, experience with conventions for driving on
one side of the road may encourage broad use of that
principle (e.g., when one is walking on the sidewalk)
and entrench a particular way of implementing it (e.g.,
walking on the right). The same consequences can be
seen in the case of linear order. U.S. participants make
broad use of this principle across contexts; in fact, it may
be so entrenched that it springs to mind whenever order-
able elements are in play, even in implicit tasks (e.g.,
Dehaene et al., 1993). Yupno participants, who have not
Fig. 3. Directionality of the ordered lines produced across all three had a lifetime of exposure to linear order, abandon this
studies. Each circular plot is divided into eight sections, according to principle when categorical construals are highlighted.
the scheme used for coding orientation (rightward, rightward-away,
away, etc.). Within these sections, each spoke represents an actual This pattern sheds light on earlier results regarding
ordered line. The spokes are plotted around the centers of the sections, Yupno participants’ performance on the number-line
with Gaussian noise to avoid overplotting. estimation task (Núñez et al., 2012). That task, much like
the procedure used in Study 3, introduces salient
examined how abstract relations are organized spatially in anchors—the line’s endpoints—and thus invites a cate-
two cultural groups, one with diverse and ubiquitous repre- gorical construal. In both the number-line task and our
sentational practices exhibiting the principle (U.S.) and one card-arrangement task in Study 3, the Yupno overwhelm-
without such practices (Yupno). Members of both groups ingly favored spatial grouping. Graphical practices
organized sets of orderable cards into lines, which suggests also regiment intuitions about linear order by determin-
that familiarity with conventional linear representations is ing how it is implemented. Ordered lines in the visual
not necessary to spark intuitions about linear order. But the culture of the United States—and wherever Latin script
two groups did differ in when they used the principle and predominates—proceed rightward. As has been demon-
how they implemented it: The U.S. participants were stead- strated in prior studies (e.g., Bergen & Lau, 2012; Tversky
fast in their use of linear order, regardless of task proce- et al., 1991), we found that U.S. participants strongly
dures, and their ordered lines were consistently oriented favored this direction. In contrast, the Yupno, without
rightward; the Yupno participants, in contrast, relied on lin- this regimenting influence, produced lines oriented in all
ear order primarily when grouping was not a salient alterna- directions.
tive, and their lines were oriented idiosyncratically. We also found a pattern—weak but present numeri-
If the notion of linear order is not rooted ultimately in cally across all three studies—in which U.S. participants
visual culture, then where does it come from? Although it used spatial grouping for categorical stimuli more than
is tempting to consider the principle an a priori intuition, the Yupno did. A possible interpretation of these results
possible experiential sources abound. Rudiments of linear is that spatial organization strategies in U.S. participants
order are present in natural phenomena, such as animal are regimented generally such that linear order is selec-
tracks or human footprints, which embody an unfolding tively applied to orderable stimuli and spatial grouping is
sequence. Similarly, when groups of people walk on nar- selectively applied to categorical stimuli. In the case of
row paths, they form lines that can be construed as grouping, this regimentation may be due to experience
ordered. It may also have been possible for participants to with Venn diagrams and other graphics that colocate ele-
discover the utility of linear order during the arrangement ments of the same category. Without the regimenting
task itself: Putting a series of cards in a line may facilitate influence of this visual culture, the Yupno appear to be
a series of pairwise comparisons, which are necessary to more flexible in their spatial organization strategies.
arrive at a full ordering of a set of elements. Thus, the fact Together, these findings hint at how the ordered line
that the Yupno and the U.S. participants produced similar has come to occupy a privileged place in both visual
arrangements does not necessarily mean that they drew culture and cognition. Unlike other representational con-
on similar—let alone a priori—mental representations. ventions, such as writing (Gelb, 1963), tree diagrams
Rather, the Yupno may have deployed an ad hoc organi- (Lima, 2014), and number lines proper (Núñez, 2011), the
zational scheme, whereas the U.S. participants may have ordered line likely did not have to be invented. Rudi-
drawn on an internalized convention. ments of linear order in the natural world, as well as
Ordered Line 607

expediencies of organizing information, could have moti- found at http://journals.sagepub.com/doi/suppl/10.1177/0956


vated its initial manifestations. But such early uses would 797617691548. This article has received the badge for Open
likely have been limited and evanescent. Since the Mid- Materials. More information about the Open Practices badges can
dle Ages, however, the uses of linear order have multi- be found at http://www.psychologicalscience.org/publications/
badges.
plied as a result of the growing importance of quantitative
information in accounting, navigation, timekeeping, and
other domains (Crosby, 1997); at the same time, ordered Notes
lines have become more durable and portable, enshrined 1. Linear order is sometimes considered an elaboration of spa-
in books, other artifacts, and now digital forms (e.g., tial grouping. These strategies share the idea that like goes with
Buringh & Van Zanden, 2009). As high-literacy, industri- like, but linear order represents additional information (for dis-
cussion, see Tversky et al., 1991). However, this does not imply
alized cultures have become saturated with an ever-
that linear order is always a better representation.
increasing number of manifestations of linear order, the
2. Following the recommendations of Simmons, Nelson, and
principle has, in turn, become increasingly regimented Simonsohn (2012), we report in this article our method for
as a powerful tool of everyday cognition. determining sample sizes, all data exclusions, all manipulations,
and all measures.
Action Editor 3. In all three studies, “unschooled” Yupno participants with a
Ayse K. Uskul served as action editor for this article. small amount of elementary education did not behave differ-
ently from those with no exposure to education whatsoever
(see the Supplemental Material for analyses). Also in all three
Author Contributions
studies, the Yupno participants were run first; the Yupno sam-
All the authors developed the general research direction and ple size for each study was determined by estimating the total
designed Study 1. K. Cooperrider and R. Núñez designed Studies 2 number of unschooled adults available to participate, and then
and 3. K. Cooperrider and R. Núñez collected the Yupno data, and attempting to distribute this number evenly across the three
all the authors supervised collection of the U.S. data. T. Marghetis studies. The sample size for the U.S. participants was then
conducted the statistical analyses, in consultation with K. Cooper- matched to the Yupno sample size, with a few additional par-
rider. All the authors interpreted the results. K. Cooperrider drafted ticipants in case procedural irregularities warranted exclusions.
the manuscript, and critical revisions were provided by T. Marghe- 4. We also prepared orderable circular stimuli that represented
tis and R. Núñez. All the authors approved the final manuscript. brightness, from near-white to black. However, once we began
testing at the Yupno field site—which was indoors and lacked
Acknowledgments overhead lighting—it became apparent that the fine gradations
We thank the Yupno residents of Gua for their participation and of brightness were difficult to differentiate. We thus removed the
assistance, James Slotta for his help with the Yupno language, and brightness items from Study 3 and excluded them from all analyses
Danielle Jacques and Natalie Allen for their research assistance. (see the Supplemental Material for analyses including these stimuli).
U.S. participants always received the same stimuli as the Yupno.
Declaration of Conflicting Interests
References
The authors declared that they had no conflicts of interest with
Archibald, J. D. (2014). Aristotle’s ladder, Darwin’s tree: The
respect to their authorship or the publication of this article.
evolution of visual metaphors for biological order. New
York, NY: Columbia University Press.
Funding Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random
The National Geographic Society (Grant 9131-12) funded the effects structure for confirmatory hypothesis testing: Keep
fieldwork on which the reported studies are based. The National it maximal. Journal of Memory and Language, 68, 255–278.
Science Foundation’s Spatial Intelligence and Learning Center Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015).
(Grant SBE-0541957) funded the postdoctoral research of K. Fitting linear mixed-effects models using lme4. Journal of
Cooperrider. Statistical Software, 67, 1–48.
Bergen, B. K., & Lau, T. T. C. (2012). Writing direction affects
how people map space onto time. Frontiers in Psychology,
Supplemental Material
3, Article 109. doi:10.3389/fpsyg.2012.00109
Additional supporting information can be found at http:// Buringh, E., & Van Zanden, J. L. (2009). Charting the “rise of the
journals.sagepub.com/doi/suppl/10.1177/0956797617691548 West”: Manuscripts and printed books in Europe, a long-
term perspective from the sixth through eighteenth centu-
Open Practices ries. The Journal of Economic History, 69, 409–445.
Crosby, A. W. (1997). The measure of reality: Quantification
and Western society, 1250–1600. Cambridge, England:
Cambridge University Press.
All materials have been made publicly available via the Open Dehaene, S., Bossini, S., & Giraux, P. (1993). The mental rep-
Science Framework and can be accessed at https://osf.io/8dvsz. resentation of parity and number magnitude. Journal of
The complete Open Practices Disclosure for this article can be Experimental Psychology: General, 122, 371–396.
608 Cooperrider et al.

Dehaene, S., Izard, V., Spelke, E., & Pica, P. (2008a). Log or lin- Núñez, R., & Cooperrider, K. (2013). The tangle of space and
ear? Distinct intuitions of the number scale in Western and time in human cognition. Trends in Cognitive Sciences, 17,
Amazonian indigene cultures. Science, 320, 1217–1220. 220–229.
Dehaene, S., Izard, V., Spelke, E., & Pica, P. (2008b). Log or Núñez, R. E. (2011). No innate number line in the human brain.
linear? Distinct intuitions of the number scale in Western Journal of Cross-Cultural Psychology, 42, 651–668.
and Amazonian indigene cultures [Supplemental material]. Price, M. C. (2009). Spatial forms and mental imagery. Cortex,
Science, 320, 1217–1220. 45, 1229–1245.
De Soto, C. B., London, M., & Handel, S. (1965). Social reason- R Core Team. (2015). R: A language and environment for sta-
ing and spatial paralogic. Journal of Personality and Social tistical computing. Retrieved from https://www.R-project
Psychology, 2, 513–521. .org/
Fuhrman, O., & Boroditsky, L. (2010). Cross-cultural differences Restle, F. (1970). Speed of adding and comparing numbers.
in mental representations of time: Evidence from an implicit Journal of Experimental Psychology, 83, 274–278.
nonlinguistic task. Cognitive Science, 34, 1430–1451. Shaki, S., Fischer, M. H., & Petrusic, W. M. (2009). Reading hab-
Gelb, I. J. (1963). A study of writing. Chicago, IL: University of its for both words and numbers contribute to the SNARC
Chicago Press. effect. Psychonomic Bulletin & Review, 16, 328–331.
Gevers, W., Reynvoet, B., & Fias, W. (2003). The mental rep- Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2012). A 21 word
resentation of ordinal sequences is spatially organized. solution. Retrieved from http://ssrn.com/abstract=2160588
Cognition, 87, 87–95. Tversky, B. (2011). Visualizing thought. Topics in Cognitive
Ingold, T. (2007). Lines: A brief history. New York, NY: Routledge. Science, 3, 499–535.
Ishihara, M., Keller, P. E., Rossetti, Y., & Prinz, W. (2008). Tversky, B., Kugelmass, S., & Winter, A. (1991). Cross-cultural
Horizontal spatial representations of time: Evidence for the and developmental trends in graphic productions. Cognitive
STEARC effect. Cortex, 44, 454–461. Psychology, 23, 515–557.
Lima, M. (2014). The book of trees: Visualizing branches of van Dijck, J.-P., & Fias, W. (2011). A working memory account
knowledge. Princeton, NJ: Princeton Architectural Press. for spatial-numerical associations. Cognition, 119, 114–
McCrink, K., & Opfer, J. E. (2014). Development of spatial- 119.
numerical associations. Current Directions in Psychological Wassmann, J. (1993). Das Ideal des leicht gebeugten Menschen:
Science, 23, 439–445. Eine ethno-kognitive Analyse der Yupno in Papua New
Nuerk, H.-C., Patro, K., Cress, U., Schild, U., Friedrich, C. K., & Göbel, Guinea [The ideal of the slightly bent person: An ethno-
S. M. (2015). How space-number associations may be created cognitive analysis of the Yupno from Papua New Guinea].
in preliterate children: Six distinct mechanisms. Frontiers in Berlin, Germany: Dietrich Reimer.
Psychology, 6, Article 215. doi:10.3389/fpsyg.2015.00215 Zebian, S. (2005). Linkages between number concepts, spatial
Núñez, R., Cooperrider, K., & Wassmann, J. (2012). Number thinking, and directionality of writing: The SNARC effect
concepts without number lines in an indigenous group and the REVERSE SNARC effect in English and Arabic
of Papua New Guinea. PLoS ONE, 7(4), Article e35562. monoliterates, biliterates, and illiterate Arabic speakers.
doi:10.1371/journal.pone.0035662 Journal of Cognition and Culture, 5, 165–190.

You might also like