Sunteți pe pagina 1din 41

Cognition 132 (2014) 342–382

Contents lists available at ScienceDirect

Cognition
journal homepage: www.elsevier.com/locate/COGNIT

Mapping spatial frames of reference onto time: A review


of theoretical accounts and empirical findings
Andrea Bender ⇑, Sieghard Beller
Department of Psychosocial Science, University of Bergen, N-5020 Bergen, Norway

a r t i c l e i n f o a b s t r a c t

Article history: When speaking and reasoning about time, people around the world tend to do so with
Received 15 May 2013 vocabulary and concepts borrowed from the domain of space. This raises the question of
Revised 30 November 2013 whether the cross-linguistic variability found for spatial representations, and the principles
Accepted 31 March 2014
on which these are based, may also carry over to the domain of time. Real progress in
addressing this question presupposes a taxonomy for the possible conceptualizations in
one domain and its consistent and comprehensive mapping onto the other—a challenge
Keywords:
that has been taken up only recently and is far from reaching consensus. This article aims
Space
Time
at systematizing the theoretical and empirical advances in this field, with a focus on
Frames of reference accounts that deal with frames of reference (FoRs). It reviews eight such accounts by
Temporal perspectives identifying their conceptual ingredients and principles for space–time mapping, and it
Mental time line explores the potential for their integration. To evaluate their feasibility, data from some
Space–time mapping thirty empirical studies, conducted with speakers of sixteen different languages, are then
scrutinized. This includes a critical assessment of the methods employed, a summary
of the findings for each language group, and a (re-)analysis of the data in view of the
theoretical questions. The discussion relates these findings to research on the mental time
line, and explores the psychological reality of temporal FoRs, the degree of cross-domain
consistency in FoR adoption, the role of deixis, and the sources and extent of space–time
mapping more generally.
Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction 2010); and spatial primes can be used to influence the


experience of duration (DeLong, 1981), visuospatial atten-
When speaking about time, people around the world tion (Torralbo, Santiago, & Lupiáñez, 2006; Weger & Pratt,
tend to do so with vocabulary and concepts borrowed from 2008), or reasoning about time (e.g., Boroditsky & Ramscar,
the domain of space (Alverson, 1994; Clark, 1973; 2002; Gentner, Imai, & Boroditsky, 2002). In fact, time and
Haspelmath, 1997; Traugott, 1978). This link reaches space, together with quantity, appear to be computed by a
beyond the extension of word meaning. For instance, generalized magnitude system of the brain (Walsh, 2003),
co-speech gestures often add a spatial dimension to with temporal relations being mapped onto spatial repre-
temporal expressions (Núñez, Cooperrider, Doan, & sentations, but not vice versa (Casasanto & Boroditsky,
Wassmann, 2012; Núñez & Sweetser, 2006); postural sway 2008; Casasanto, Fotakopoulou, & Boroditsky, 2010), at
is affected by whether people embark on a mental time least in humans (Merritt, Casasanto, & Brannon, 2010).
travel into the future or the past (Miles, Nind, & Macrae, In parallel, evidence has accumulated that different
groups of people conceptualize space in different ways
⇑ Corresponding author. (Bennardo, 2002; Haun, Rapold, Call, Janzen, & Levinson,
E-mail address: Andrea.Bender@psysp.uib.no (A. Bender).
2006; Haun, Rapold, Janzen, & Levinson, 2011; Levinson,

http://dx.doi.org/10.1016/j.cognition.2014.03.016
0010-0277/Ó 2014 Elsevier B.V. All rights reserved.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 343

2003; Levinson & Wilkins, 2006; Majid, Bowerman, Kita, discussed with regard to their theoretical implications
Haun, & Levinson, 2004; Senft, 1997), and this raises (Section 7). The potential for integration is outlined in
important questions. If, for instance, the link between the conclusion (Section 8).
space and time is indeed universal, should we then expect Before doing so, two constraints need to be explicated
that the cross-cultural variability found for spatial repre- and one clarification should be made. First, this review
sentations will carry over to the domain of time? And to does not presuppose that all temporal conceptualizations
what extent might culture-specific ways of talking about are derived from space. In fact, some properties of time
space also structure talking about time? Real progress in and temporal entities cannot be spatialized (Galton,
addressing these questions, we argue, presupposes a tax- 2011), and some linguistic groups appear to generalize this
onomy for the possible conceptualizations in one domain to the whole domain of time (e.g., Sinha, Da Silva Sinha,
and its concise and comprehensive mapping onto the Zinken, & Sampaio, 2011, and see Section 6.9 below). How-
other. Such a harmonizing of terminology would facilitate ever, as the main thrust of this paper is to provide an over-
not only cross-domain comparisons in general, but also the view of temporal taxonomies based on the spatialization of
assessment of the influence of spatial representations on time, it will focus on temporal conceptualizations derived
temporal ones in particular, and is therefore currently con- from space. Second, this review will be restricted to theo-
sidered to be one of the important desiderata in this field of retical accounts that are based on, or at least related to,
research (Bender, Rothe-Wulf, Hüther, & Beller, 2012; some type of frames of reference (FoRs) taxonomy. While
Tenbrink, 2011). there may be other taxonomies of spatial conceptualiza-
However, while research in the two domains and the tions (and more options for mapping them onto time), this
acknowledgement of cross-domain transfers do have a restriction is justified by the fact that most of the accounts
long tradition in several disciplines (reviewed in Núñez & that have been proposed recently and that are of relevance
Cooperrider, 2013), the challenge of mapping a taxonomy for this review have chosen this approach. And finally, our
of spatial representations onto the domain of time has usage of the terms ‘‘cultural’’ and ‘‘linguistic’’ requires
been taken up only recently. During the last decade, some a priori clarification. Although we basically utilize
respective attempts have mushroomed, but although sev- cross-linguistic data (i.e., data collected in groups speaking
eral of them even sail under the same flag as ‘‘temporal different languages), we will adopt the term ‘‘cultural’’
frames of reference’’, they differ considerably in terms of whenever preferences for some kind of FoR are referred
theoretical conceptualization and subsequent interpreta- to. The rationale for this is that preferences for FoRs within
tion of data—to the extent of being incompatible with each a speech community are not inherent in the meaning of
other. All too often it has been left to the reader to figure words, or in any language-specific feature for that matter,
out how these accounts are related to each other, to spatial but are a result of agreements or conventions within a
taxonomies, and to the empirical data accumulated during speech community—which we take to be a cultural
recent years. Núñez and Cooperrider therefore conclude phenomenon.
that ‘‘despite intuitive appeal and promise of parsimony, Throughout this paper, some abbreviations will be used
a definitive taxonomy of ‘temporal frames of reference’ as a compromise between conciseness and readability
remains elusive’’ (2013, p. 221). With our review, we (explained in Table 1, upper part). We also attempt to
attempt to systematize the theoretical and empirical use the same terms throughout the paper when referring
advances in this field, by sorting the temporal accounts to the same referents; in cases where being faithful to
proposed so far according to their similarities and differ- alternative accounts requires deviations from this termi-
ences, by comparing the principles according to which they nology (for an overview, see Table 1, lower part), we will
map spatial taxonomies onto time, and by scrutinizing the add the labels preferred in this review in square brackets.
available data with regard to how they would be inter-
preted in the light of each of these accounts.
More specifically, we begin (in Section 2) by describing 2. The domain of time and its relation to space
the theoretical and conceptual ingredients on which most
of the accounts are based, including a brief outline of the Time is an abstract domain in the sense that it is intan-
properties and variants of the concept TIME and of the con- gible and ephemeral, and that we lack sensory organs to
ceptual sources for the construal of temporal taxonomies. perceive it directly. The ability to process temporal infor-
In Section 3, we provide an overview of the different mation is based on two distinct computational mecha-
taxonomies, followed, in Section 4, by their systematic nisms (Pöppel, 1997; Pöppel & Wittmann, 1999), and the
comparison according to the relations they establish awareness of the passing of time is linked to memory pro-
between conceptual sources, the principles they adopt for cesses (Lewis & Miall, 2006). But our attempts to capture
construing frames of reference and for assigning FRONT, time conceptually seem to hinge to a considerable degree
and the reference patterns they distinguish. on metaphorical extension (Lakoff & Johnson, 1980,
The second part of this review is then devoted to a 1999). While not the only possible candidate (Evans,
(re-)analysis of the available empirical data, collected 2003), one domain which suggests itself as a source for
partly as evidence for conceptual innovations of specific such a metaphorical extension is space. Not only is space
accounts and partly with the goal of assessing cross- more concrete than time; it is also linked in various ways
cultural variability. Based on an overview of the methods to the latter, for instance in all processes of change or
employed (Section 5), findings are first presented sepa- motion. But before we turn to the various ways in which
rately for each speech community (Section 6), and are then space is fundamental to time, thus allowing for the transfer
344 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Table 1
Abbreviations and variations of technical terms as used in this paper.

Abbreviations
FoR frame of reference
s-FoR spatial frame of reference
t-FoR temporal frame of reference
F Figure; also: ‘‘referent R’’ (see below)
G Ground; also: ‘‘relatum’’ (Tenbrink, 2011) or ‘‘reference point RP’’ (see below)
V observer’s viewpoint (in temporal contexts often equated with Ego)
X origin (origo) of the coordinate system in establishing a FoR; in ternary relations (relative FoR), X is transferred from V into G, which
then becomes the origo of the secondary coordinate system (X2)
RP reference point (as used by Núñez & Sweetser, 2006, and Yu, 2012) [= G]
R referent (as used by Yu, 2012) [= F]
ME moving Ego (perspective); also: ‘‘Ego-moving’’ (e.g., Boroditsky, 2000; Gentner et al., 2002)
MT moving time (perspective); also: ‘‘time-moving’’ (e.g., Boroditsky, 2000; Gentner et al., 2002)
Technical terms preferred for this paper, and some of their variants
absolute FoR also referred to as: ‘‘field-based’’ (Talmy, 2000; and see Moore, 2004, 2011) or ‘‘extrinsic’’ (Kranjec, 2006)
intrinsic FoR also referred to as: ‘‘ground-based’’ (Talmy, 2000)
relative FoR also referred to as: ‘‘projector-based’’ (Talmy, 2000), ‘‘deictic’’ (e.g., Kranjec, 2006), ‘‘egocentric’’, or ‘‘viewer-centered’’ (elsewhere in
the literature)

Note. Some terms are used in capital letters when they denote concepts of things (as the concept of TIME) or, in the specific case of FRONT, do distinguish FRONT
assigned to entities or fields from ‘‘front’’ in the ‘normal’ sense.

of spatial conceptualizations to time in the first place and to compare these taxonomies in a constructive man-
(Moore, 2011), we need to explicate the ways in which ner, the following aspects need to be explicated: (i) the
space and time differ. extent to which properties of time itself can (or cannot)
be mapped onto the spatial dimension, (ii) the extent to
which concepts of time may vary, and (iii) the ways in
2.1. Properties and concepts of time
which the directionality of time can be assessed.

Attempts to describe and analyze time are plentiful (for


overviews see, e.g., Evans, 2003; Friedmann, 1990; Le 2.1.1. Properties of time
Poidevin, 2003; Newton-Smith, 1980), and we will not According to Galton’s (2011) commendable analysis,
repeat these here. But in order to identify the conceptual time has four distinct properties (or attributes), which it
ingredients of space-based taxonomies of temporal FoRs shares—to varying degrees—with space: extension, linear-

Table 2
Properties of time and how they can (or cannot) be mapped onto the spatial dimension (for more details, see Galton, 2011); further
explanation is given in the text.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 345

Table 3
Concepts of time.

ity, directionality, and transience. Extension implies that ing, musical notation, the time axis of graphs, or the design
time has separate parts (such as distinct moments); linear- of clocks). Directionality allows for the mapping of asym-
ity implies that of any three distinct moments, one has to metric spatial axes onto time, as in the mapping of past
be between the other two; directionality implies an asym- and future to the TOP/DOWN axis in Mandarin (Scott, 1989)
metry between past and future; and transience implies or as in the Moving Ego (ME) and Moving Time (MT) met-
the fleetingness of every single moment. These four prop- aphors to be described below (Clark, 1973; Fillmore, 1971).
erties constitute an ordered series, with extension being As the ME/MT metaphors also involve motion, they make
foundational to linearity, which itself is foundational to additional use of the property of transience (Galton, 2011).
directionality, which is foundational to transience (see Of these properties, the latter three will be of relevance for
Table 2). the remainder of this paper: linearity for the concept of cycli-
Parallel to this increase in hierarchical order, the extent cal time (to be discussed in the next section); directionality
to which these properties can be mapped onto space for the asymmetric properties of time and temporal concepts,
decreases. While mapping is straightforward for the prop- which is essential for the assignment of FRONT and thus for the
erty of extension, it is possible for linearity only if any one of construal of temporal FoRs; and transience for dynamic con-
the three dimensions of space is singled out. Please note texts that involve real or metaphorical motion.
that this may also include closed loops. In addition to the
restriction to one dimension, the mapping of asymmetry
from time to space also requires consideration of specific 2.1.2. Conceptual variants of time
conditions under which this property may emerge: The Three fundamentally different concepts of time will be
asymmetry between UP and DOWN relies on gravitation, considered in this review: linear, cyclical, and radial time
the asymmetry between FRONT and BACK on the anatomy of (see Table 3).
the human body, and the asymmetry between LEFT and The concept of linear time is compatible with the expe-
RIGHT (largely) on cultural values. A less obvious, but none- rience of passing time and the irreversible courses of
theless valid, asymmetry also emerges from the half-axes events. Typically, this concept also implies directionality,
that are radiating out from a central point, namely along aptly illustrated in the metaphor of an ‘‘arrow of time’’
the TOWARDS/AWAY FROM, NEAR/FAR, or INWARDS/OUTWARDS dimen- (Section 2.1.3). Its directionality is reflected, for instance,
sion. Transience, finally, is the only property of time that in diagrammatic depictions that are oriented towards the
space cannot have in and of itself. However, space can future (Galton, 2011), and it structures the canonical way
acquire it by being linked to time, as in motion (Galton, in which we recount history or tell stories.1 Cognitive sci-
2011). When traveling by train, for instance, the places entists tend to consider the linear concept as the prevailing
passed by are transient in a way similar to the elapsed concept of time in human thinking, and conceptualize its
moments. In contrast to changes of place during motion, representation by a mental time line (Walsh, 2003; and
changes of state involve only a negligible spatial compo- see Miles et al., 2010; Weger & Pratt, 2008).
nent and thus might be regarded as almost purely tempo- While the concept of linear time seems to directly
ral. This temporal transience is captured by a concept of spring from the property of linearity, the latter also affords
time not as spatially extended, but simply as a perpetual a concept of cyclical time, except that in this case the line
‘‘getting later’’, as has been described for the Hopi by forms a closed loop (Galton, 2011; and see Le Poidevin,
Whorf (1956; and see Malotki, 1983). 2003; Newton-Smith, 1980). The concept of cyclical time
When time is metaphorically mapped onto space, the
property of extension leads to the description of events as 1
This structure can be violated, of course, for various reasons, and if done
‘‘occupying space’’ or being ‘‘in the middle of’’ something. artistically—as in the movie Memento (2000), directed by Christopher
Nolan—this violation may be experienced as fascinating. However, one of
Linearity affords linguistic metaphors of duration, but also
the very reasons for the positive reviews of Memento was that its nonlinear
graphic and other representational metaphors (as in writ- narrative structure was considered unique and original.
346 A. Bender, S. Beller / Cognition 132 (2014) 342–382

is compatible with our experience of recurring time peri- 2.2. Conceptual sources for taxonomies of temporal frames of
ods that structure our life, such as times of the day, days reference
of the week, seasons, or annual holidays and feasts, and
recurring events, such as the life cycles of plants and Although this review focuses on the spatialization of
animals. It has been described for several non-Western time and in particular on the mapping of taxonomies of
cultures such as Hinduism (Sharma, 1974), the Hopi spatial frames of reference onto temporal relations, two
(Malotki, 1983; Whorf, 1956), or the Maya (Farriss, 1987; other sources for these types of taxonomies also require
León-Portilla, 1990). The linear and the cyclical concept attention as they are frequent ingredients for construals
can be reconciled and integrated into a spiral or helic of temporal taxonomies: the distinction of A-series versus
concept according to which types of events do recur, but B-series, and the equally popular distinction of the Moving
at different points in time and in a non-reversible order. Ego (ME) versus Moving Time (MT) perspective. The heter-
The concept of radial (or ego-centric) time, finally, ogeneity in this field of research basically arises from the
reflects the precision of memory (of the past) and various ways in which these two temporal sources are
anticipation (of the future), which is most pronounced combined with each other and with spatial notions.
for proximal events and decreases with distance from the
subjective now, symmetrically into past and future. Like
2.2.1. Descriptions of time: A-series versus B-series
the aforementioned linear and cyclical concept, it is
One of the oldest research traditions in the domain of
compatible with, and in fact follows directly from, the
time goes back to an essay by the philosopher John
property of directionality when mapped onto space. How-
McTaggart (1908) on ‘‘The Unreality of Time’’, in which
ever, as outlined above, the mapping of the asymmetry
he identifies two different descriptions of temporal order:
property from time to space requires a consideration of
the A-series versus B-series of time (see also Gell, 1992;
specific conditions under which this property may emerge,
Traugott, 1975). Broadly speaking, the A-series description
and one of these conditions involves the half-axes radiat-
of events follows from the way in which these events are
ing out from a central point (Galton, 2011). Illustrative
ordered, as being in the past, present, or future, relative
examples of this radiation are the loudness of sound or
to an observer’s subjective now or deictic center. Being
the brightness of light emitted from a single source, or
part of an A-series also involves a change of status for each
the field of vision or attention from the point of view of
and every event, which was once in the future, then
any given observer. The concept of radial time has received
becomes present, and will finally drift into the past
the least amount of attention (but see Bennardo, 2009),
(Fig. 1a). The B-series description, on the other hand, simply
and although it has been predicted to occur in speech com-
refers to the order of events within a sequence, in which
munities, which prefer a relative (or ego-centric) FoR for
one event precedes (‘‘is earlier than’’) or follows (‘‘is later
temporal descriptions (Bender, Bennardo, & Beller, 2005),
than’’) another one, regardless of the point in time from
it remains unclear whether it only characterizes temporal
which this may be observed (Fig. 1b).
aspects of cognitive processes or may also serve to struc-
This distinction is captured by language in that A-series
ture representations of time (as claimed by Bender,
descriptions typically include tense, which requires
Beller, & Bennardo, 2010).
anchoring in the speaker, and thus are deictic, whereas B-
series descriptions are based on sequencing (expressed
by ‘‘earlier/before’’ and ‘‘later/after’’ relations), which
2.1.3. The arrow of time
requires an anchoring of events with respect to each other,
An implication of Galton’s (2011) third property of time
and are thus non-deictic (Traugott, 1975, 1978).
is the asymmetry between past and future, also labeled
‘‘arrow of time’’. This notion deserves particular attention
here as it will be essential in some accounts for construing 2.2.2. Temporal perspectives: Moving Ego (ME) versus Moving
an absolute FoR. While most physical processes are sym- Time (MT)
metrical with regard to time (i.e., they could be reversed If we think of time as a river, the direction of flow
without violating physical laws), some appear to unfold allows for two different, yet complementary, perspectives
over time from past to future. The increase of entropy, (see Table 4, left column): the Moving Ego (ME) and the
for instance, described by the second law of thermodynam- Moving Time (MT) perspective (e.g., Clark, 1973;
ics, bestows a direction onto the flow of time; and the same Fillmore, 1971). From the Moving Ego perspective, we
holds for the expansion of the universe, for radiation and regard ourselves as moving downstream through station-
radioactive decay, or for the course of biological evolution ary time; we approach future events and leave them
(Gould, 1987; Mackey, 2003). This directedness from past behind, as in (1). From the complementary Moving Time
to future is captured by the ‘‘arrow of time’’ metaphor. perspective, we regard ourselves as stationary and time
Although its status in physics is not entirely uncontrover-
sial (Price, 1996), it is reflected in some of our basic cogni-
(a) A-series (b) B-Series
tive processes: The perception of events precedes memories
of these events, and causes precede effects. Typically, the
arrow of time is seen as pointing into the future, and for
any concept based on this notion, FRONT in time would thus deictic or tensed time non-deictic or sequence time
be assigned to the future (alternative views and assign-
ments will be discussed further below). Fig. 1. The A-series versus B-series of time.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 347

Table 4
Extending the Moving Ego (ME) and Moving Time (MT) perspective beyond deictic examples (A-series) to include non-deictic examples (B-
series).

Note. Non-deictic MT corresponds to the ‘‘Time-RP’’ metaphor (proposed by Núñez et al., 2006); non-deictic ME does not occur in English,
but has been observed in Hausa (Hill, 1978), and in Japanese and Marathi (Shinohara & Pardeshi, 2011); further explanation is given in
the text.

as moving towards us; future events approach us and pass every Wednesday’s meeting’’: The perspective of time as
by, as in (2): moving from the future to the past (MT) would still suggest
a pastwards reading, whereas the inverse perspective of
(1) ‘‘We are approaching Tuesday.’’ [ME]
temporal entities as moving from the past to the future
(2) ‘‘Tuesday is approaching (us).’’ [MT]
(ME) would suggest a futurewards reading, as illustrated
in the right-hand column of Table 4. Accordingly, non-
These two perspectives on time can be regarded as met-
deictic sentences like (3) will also count as MT:
aphoric systems which are accessed during information
processing (for overviews, see Gentner, 2001; Núñez & (3) ‘‘Monday comes before Tuesday.’’ [MT]
Sweetser, 2006). Either of these systems suggests one spe-
cific reading of ambiguous phrases like ‘‘moving a meeting In this case, the metaphorical motion of events—
forward’’ (Miller & Johnson-Laird, 1976), priming either the triggered by their alignment in a sequence, in which ear-
ME or MT perspective shifts these readings more towards lier events are ‘‘in front of’’ later ones—resembles the
the future or the past, and switching from one system to moving time perspective from future to past. This raises
the other incurs cognitive costs, as revealed by reaction the question of whether such (non-deictic) B-series
time in consistent versus inconsistent mapping conditions descriptions could also be compatible with the reverse
(Boroditsky, 2000; Gentner et al., 2002; McGlone & ME perspective. While the label ‘‘Moving Ego’’ may make
Harding, 1998). this sound odd, there is actually no reason why it should
ME and MT are simply two different perspectives on the be impossible or even implausible. The lack of evidence
exact same scene—mirror images or figure/ground reversals from linguistic examples in English alone does not suffice
of each other (Talmy, 2000; Traugott, 1978): In (1), for to assess such a claim. Although seemingly not a domi-
instance, the figure (i.e., the entity to be located) is ‘‘we’’, nant phenomenon across languages, at least one condi-
and the ground (the entity in reference to which the figure tion is conceivable under which B-series descriptions
is located) is Tuesday; in (2), it is the other way around. would indeed be compatible with an ME perspective:
In their original version, the scenes depicted in a phrase When sequences of events are described with later events
like ‘‘moving forward next Wednesday’s meeting’’ involved being ‘‘in front of’’ earlier ones, their alignment reflects
a deictic center or Ego (implied in ‘‘next Wednesday’’), but the futurewards direction, which is otherwise characteris-
the complementary relation of ME and MT also holds for tic of the ME perspective, as in (4), taken from Hill (1978,
non-deictic scenes (Bender et al., 2010). Consider the p. 536; and see Levinson & Majid, 2013; Shinohara &
slightly rephrased non-deictic expression ‘‘moving forward Pardeshi, 2011):
348 A. Bender, S. Beller / Cognition 132 (2014) 342–382

(4) ‘‘Tuesday is before Monday.’’ [MT] for space. Over the years, several different taxonomies
have been proposed, with the taxonomy by Levinson
In any case, the relationship between the ME and MT (2003) being one of the most popular. It will be taken here
perspective on the one hand, and A- and B-series on as the reference point because it has served as the basis for
the other, cannot be straightforward. While ME and MT assessing cross-linguistic variability in spatial references,
are mirror images of each other (i.e., complementary and because it was employed as one of the conceptual
perspectives on the same scene), A- and B-series are not; sources in most recent attempts of space–time mappings
in fact, they are incommensurable with each other (for alternative accounts, see also Bohnemeyer &
(McTaggart, 1908; Traugott, 1975, 1978). As we will see O’Meara, 2012; Danziger, 2010; Talmy, 2000).
below, one of the reasons why the different accounts differ In general, a frame of reference (FoR) is a coordinate
so much is because they define ME and MT perspectives system required to establish the position of a figure in
differently (as being strictly deictic or not), and because reference to a ground from a given perspective (Talmy,
they map them differently onto A- and B-series 2000); sometimes, albeit not necessarily, this perspective
descriptions. coincides with the viewpoint of an observer. The taxonomy
proposed by Levinson (2003) distinguishes three basic
2.2.3. Spatial frames of reference (s-FoR) types: an absolute, an intrinsic, and three variants of a rela-
The theoretical construct of prime interest for this tive FoR, most of which are depicted in Fig. 2 (for detailed
review is frames of reference, which were first designed descriptions, see Bender et al., 2010, 2012; Levinson, 2003).

Fig. 2. Spatial frames of reference, s-FoR (adapted from Bender et al., 2010).
A. Bender, S. Beller / Cognition 132 (2014) 342–382 349

The characterization of these FoRs requires the following the domain of time. With the reflection variant of the
components: a figure F (the object to be located), the relative FoR (Fig. 2d), FRONT is assigned to a position
ground object G (in reference to which F is located), and between G and V or nearer to V, and BACK is assigned to a
the origo and orientation of the coordinate system X. The position beyond G or further away from V. With the
viewpoint of the observer V is optional and relevant for translation variant of the relative FoR (Fig. 2e), the reverse
the relative FoR only. is true. Please note that the relative FoRs correspond to
The absolute FoR derives its orientation from a superor- the half-axes radiating out from a central point, identified
dinate field outside F and G. As argued above (Table 2; and by Galton (2011) as one possibility for mapping the
see Galton, 2011), space generally lacks directionality and directionality property of time to space. This gives rise to
asymmetry, even on Earth (and especially if one leaves an asymmetry along the TOWARDS/AWAYWARDS dimension
aside the vertical dimension that, by virtue of gravitation, (see Table 2), with the reflection variant favoring the
may be seen as ‘naturally’ directed). Assignment of TOWARDS direction, and the translation variant favoring the
orientation to the superordinate field therefore depends AWAYWARDS direction.
on cultural conventions and includes, for instance, the car- Most of the previous work on spatial FoRs has been con-
dinal points, mountain slopes, prevailing wind directions, cerned with documenting culture-specific preferences for
rivers, or the land-sea axis on small islands (Bennardo, specific FoRs (e.g., Bennardo, 2002; Haun et al., 2006,
2000, 2002; Levinson, 2003; Senft, 1997). Importantly, 2011; Hüther, Bentz, Spada, Bender, & Beller, 2013;
neither the position of a potential observer nor the Levinson, 2003; Levinson & Wilkins, 2006; Majid et al.,
orientation of G is relevant for referencing. The cat 2004; Senft, 1997), and with exploring whether differences
in Fig. 2a would therefore be described as ‘‘east of the in these preferences also entail cognitive implications
car’’.2 (Levinson, Kita, Haun, & Rasch, 2002; versus Li &
The intrinsic FoR derives its orientation from the ground Gleitman, 2002; and see Haun et al., 2011; Li, Abarbanell,
entity G and can thus only be adopted if G is perceived as Papafragou, & Gleitman, 2011). In this review, we will focus
being directed. Although a large number of potential on the question of whether preferences for a specific FoR in
ground entities are asymmetric, this does not necessarily spatial contexts may carry over to the temporal domain. A
imply directionality. Assignment of orientation to such an conclusive answer to this question, however, requires a
entity therefore, again, depends on cultural conventions. conclusive mapping of spatial FoRs onto time. The premises
In the case of animates this is typically based on looking for such a mapping are given: Establishing the position of a
direction; but even artifacts such as cars, computers, or figure in reference to a ground from a given perspective
chairs can be assigned a FRONT, depending, for instance, on requires a frame of reference, in time as much as in space.
how people normally interact with them or into which And although space and time differ in various ways
direction they tend to move (e.g., Bennardo, 2000; Clark, (Galton, 2011), the principles on which Levinson’s (2003)
1973). Crucially, this includes an observer, as long as this taxonomy is based are domain-general, as outlined above,
observer serves as the (primary) ground. The cat in and can therefore be applied equally well to both domains,
Fig. 2b could therefore also be described as ‘‘in front of as we aim to demonstrate in the remainder of this paper.
the car’’. The fact that a particular object may have an
intrinsic orientation, however, only allows for, but does
not determine, the choice of an intrinsic FoR. Again, the 2.2.4. Other conceptual sources: route perspective versus
position of a potential observer, if different from G, is irrel- survey perspective
evant when choosing an intrinsic FoR. For the sake of completeness it should be noted that
The relative FoRs derive their orientation from the view- temporal perspectives have also been related to the litera-
point V of an observer, which needs to be different from the ture on mental maps and spatial navigation. Jamalian and
ground object G to establish the ternary relation between F, Tversky (2012), for instance, consider both the ME and
G, and V that is constitutive of a relative FoR (Fig. 2c). As the MT perspective as analogous to a route (or intrinsic or ego-
position of F is still determined in reference to G, however, centric) perspective in space, taken from an embedded
the primary coordinate system with origo X in V needs to be viewpoint and with Ego as the reference point. In contrast,
transferred into G. This secondary coordinate system a calendar view on time is seen as analogous to a survey (or
(anchored in G = X2) can be established in three different absolute) perspective on space, taken from an external
ways, thus giving rise to the three variants of the relative viewpoint and with dates or events as the reference points.
FoR: by rotation, reflection, or translation. Of these three This account is too recent to have inspired much research
variants, reflection and rotation can only be distinguished yet, but should be kept in mind.
in two-dimensional space; in one-dimensional space (or
time, for that matter), they conflate and will thus be treated
jointly as the reflection variant (Fig. 2d). 3. Accounts for mapping spatial frames of reference
These different construals have several implications, onto time
one of which is crucial for the mapping of spatial FoRs to
For just one decade now, attempts have been under-
2
taken to map a taxonomy of spatial frames of reference
In Talmy’s (2000) account, the alignment of entities (e.g., in a queue)
can provide directness and thus create orientation for an absolute FoR. This
(s-FoR) to the domain of time, and these attempts already
version of an absolute FoR is picked up by some scholars when construing encompass more than half a dozen different variants. In
temporal FoRs. this review, the following accounts will be considered:
350 A. Bender, S. Beller / Cognition 132 (2014) 342–382

 the Ego-based vs. field-based frames of reference Recently, Moore (2011) further specified this account in
account by Moore (2004, 2006, 2011); various ways: by categorizing ME and the Ego-centered MT
 the reference-point (RP) metaphors account by perspectives together as ‘‘Ego-perspective (path-config-
Núñez and Sweetser (2006); ured) frames of reference’’, by now explicitly relating his
 the temporal framework models account by Kranjec terms to the A-series (= Ego-perspective) and B-series (=
(2006); field-based), and by discussing possible relationships with
 the temporal frames of reference account by Zinken the spatial FoRs in Levinson’s (2003) taxonomy (Table 5b).
(2010); On the latter issue of mapping spatial FoRs to the temporal
 the reference frames of space and time account by domain, however, he remains rather critical: While the
Tenbrink (2011); and field-based frame of reference is taken as including the
 the temporal frames of reference (t-FoR) account by absolute FoR, the Ego-perspective frame of reference is
Bender, Bennardo, and Beller (2005; see also claimed to be without counterpart and, more specifically,
Bender et al., 2010). to be explicitly not equivalent either to the relative or the
intrinsic FoR (Moore, 2011). This revised account is also
Two other accounts will be considered merely in discernibly closer to the account described in the next
passing: section.

 the time-referent vs. human-referent distinction by


3.2. Reference-point (RP) metaphors (Núñez)
Yu (2012). and
 the route vs. survey perspective account by Jamalian
A second attempt to go beyond temporal perspectives
and Tversky (2012).
and to integrate them with referencing systems is the ref-
erence-point (RP) metaphors account proposed by Núñez
The review will take note only of those aspects that are
and Sweetser (2006; and see Núñez, Motz, & Teuscher,
relevant to the question under scrutiny, namely how
2006). In contrast to Moore (2004, 2011), who splits the
these accounts make use of, and attempt to integrate, the
MT perspective into two distinct categories, Ego-based ver-
conceptual sources for temporal taxonomies, including
sus field-based, and merges one of these (Ego-based) with
A-/B-series, ME/MT perspectives, and frames of reference
the ME perspective as deictic, Núñez’ account classifies ME
(overview in Table 5). For any other detail of these
and MT as two sub-cases of what they call the ‘‘Ego-Refer-
accounts and for the often elaborate metaphorical
ence-Point (Ego-RP) metaphor’’, because they consider
mappings, the reader is advised to consult the respective
both ME and MT as typically involving an Ego. The Ego-
articles.
RP metaphor is set apart from a ‘‘Time-Reference-Point
(Time-RP) metaphor’’ that does not presuppose Ego
3.1. Ego-based versus field-based frames of reference (Moore)
(Table 5c).
Examples like (5) and (7) are thus categorized as
The account proposed by Moore (2004, 2006) provides
instances of Ego-RP metaphors because they make use of
one of the first attempts to combine temporal perspectives
Ego (or rather its subjective now) as the reference point.
with a frame of reference notion (albeit initially not related
Examples like (8) are categorized as instances of Time-RP
to any of the spatial FoR taxonomies). He takes the dichot-
metaphors because they depict sequences of events and
omy of the ME versus MT perspective as the starting point,
use one of these events (here Tuesday) as the reference
but expands their scope beyond examples of the A-series
point.
by distinguishing the MT perspective further into ‘‘Ego-
centered Moving Time’’ and a non-deictic MT, which are (7) ‘‘June is still ahead [Ego-RP]
captured by an Ego-based versus field-based frame of ref- [of me].’’
erence, respectively (Table 5a). (8) ‘‘Monday comes [= (6)] [Time-RP]
For the Ego-based frame of reference, Ego serves as the before Tuesday.’’
reference point. The Ego-based frame of reference thus
involves a deictic center and in this respect corresponds In this sense, the Ego-RP metaphor directly corresponds
to the A-series of time, as in (5): to an A-series description, and the Time-RP metaphor to a
B-Series description, even though this is not explicitly sta-
(5) ‘‘Tuesday is [= (2)] [Ego-based] ted by Núñez and Sweetser (2006). The main difference
approaching (us).’’ between this account and the one proposed by Moore is
In contrast, the field-based frame of reference does not thus related to how strictly each defines the scope of the
involve a deictic center; Ego and its subjective present is ME and MT perspectives: as being restricted to deictic
optional and irrelevant. Instead, it reflects a SEQUENCE AS POSI- (A-series) descriptions, or as reaching further to include
TION ON A PATH metaphor and corresponds to the B-series
non-deictic (B-series) descriptions (see discussion above
time: What counts for this type of reference is the order and Section 2.2.2).
of events within a sequence, as in (6): Núñez and Sweetser (2006) do not explicitly refer their
account to any taxonomy of spatial FoRs either. What they
(6) ‘‘Monday comes [= (3)] [field-based] do point out, though, is a correspondence between the
before Tuesday.’’ Time-RP metaphor and Evans’ (2003) complex temporal
A. Bender, S. Beller / Cognition 132 (2014) 342–382 351

Table 5
Schematic overview of the temporal accounts using frames of reference (FoR).

Note. The terms characteristic for each account are printed bold-faced; correspondence with relevant terms that are not explicated but can
be inferred are printed grey and in square brackets; metaphors are in capital letters; and correspondence to Levinson’s (2003) FoRs are
shaded. Further explanation is given in the text.
352 A. Bender, S. Beller / Cognition 132 (2014) 342–382

sequence model; in addition, they characterize the temporal a sufficient precondition for ternary relations: If Ego serves
relationships in the Time-RP metaphor as ‘‘intrinsic’’ to the as the ground entity G for the reference, the relation is bin-
sequence (Núñez et al., 2006, Footnote 1), with earlier ary (as explained in Section 2.2.3) and thus only affords an
events being described as ‘‘in front of’’ later events. These intrinsic FoR. A-series descriptions may therefore entail an
two statements encourage us to (tentatively) classify the intrinsic FoR as in (9)
Time-RP metaphor as an instance of the intrinsic FoR. This
(9) ‘‘I have a fun afternoon in front of me.’’ [intrinsic]
would also be compatible with how they derive the intrin-
(binary, with F = afternoon and
sic FRONT of G in the Time-RP metaphor, namely from the
G = me/Ego),
prototypical direction of motion (i.e., pastwards; Fig. 1b).
In contrast, the Ego-RP metaphor, for which they assign or a relative FoR as in (10)
Ego the decisive role, might be seen as corresponding to
a relative FoR. (10) ‘‘the day after tomorrow’’ [relative]
(ternary, with F = the day,
3.3. Temporal framework models (Kranjec) G = tomorrow, and V = today/Ego’s
now).
Following on from this classification, Kranjec (2006)
proposes a temporal framework models account, with which As Ego is constitutive of A-series descriptions, Ego’s
he tries to integrate the reference point metaphors with looking direction is taken as the source for assigning FRONT
the A- and B-series of time and with all three spatial FoRs to the future. This holds both for the intrinsic example in
from Levinson’s (2003) account. His taxonomy encom- (9) and the relative example in (10); however, in the latter
passes a deictic temporal framework (which comes close case, FRONT is constrained to that part of Ego’s future which
to what Levinson labels ‘‘relative’’), an intrinsic, and an is also in G’s past (which is why the day after tomorrow,
extrinsic framework or ‘‘absolute’’ FoR in Levinson’s termi- and thus in G’s future, is also in G’s back).
nology (in more recent work, the intrinsic and extrinsic In B-series time, Ego is irrelevant; therefore, both an
temporal framework were relabeled as ‘‘linked’’ and intrinsic and an absolute FoR are, in principle, possible.
‘‘path’’, respectively; see Kranjec & McDonough, 2011). However, in contrast to binary relations in A-series time,
The deictic temporal framework covers the A-series type where the intrinsic FoR is warranted by Ego and its intrinsic
of events, subsumes the ME and MT perspectives, and is FRONT, its adoption in this context requires an additional
thus equivalent to the Ego-RP metaphor proposed by assumption. To endow a ground event G with an intrinsic
Núñez and Sweetser (2006). The classical non-deictic cases FRONT, Zinken argues, events need to be regarded as one
(B-series), which basically rely on the earlier/later relation ‘‘following’’ the other, as in (11):
within temporal sequences, are captured in his account by
the intrinsic temporal framework, akin to Núñez’ Time-RP (11) ‘‘One day comes after the other.’’ [intrinsic]
metaphor. Finally, the one framework that is newly consid-
ered here, the extrinsic temporal framework, is based on The direction of movement can thus be used to assign
the notion of ‘‘time itself’’ (referred to as the ‘‘matrix sense’’ FRONT—here: to the earlier or anterior event (Zinken,
of time by Evans, 2003). According to this notion, time is ‘‘a 2010; and see Fillmore, 1971).
backdrop, or something understood to move forward, inde- The absolute FoR, on the other hand, requires that the
pendent of particular events embedded within it’’ (Kranjec, origo of the coordinate system is not located in the ground
2006, p. 450), and its direction eventually affords assign- entity G, but in the surrounding field. According to Zinken
ment of FRONT to the future (Table 5d). (2010), temporal intervals (such as days or weeks) can be
understood as such bounded entities or ‘fields’, with the
3.4. Temporal frames of reference (Zinken) beginning of the interval (in the past) corresponding to
the field’s FRONT, and with earlier events as being closer to
Similar to the previous account, Zinken (2010) takes the this FRONT. For this very reason, he classifies (12)—different
A-/B-series conception as his starting point and tries to from (11)—as an example of an absolute FoR:
integrate it with Levinson’s (2003) taxonomy of spatial
FoRs. He equates an A-series classification with Moore’s (12) ‘‘Wednesday is after Tuesday.’’ [absolute]
Ego-based frame and Núñez’ Ego-RP metaphor. Both tem-
poral perspectives, ME and MT, are considered to be sub- In a more recent paper (Sinha et al., 2011), Zinken’s
cases of this category (which reveals the stricter reading account has been somewhat modified. Here, the terms
of the temporal perspectives; Section 2.2.2). Conversely, a absolute, intrinsic, and relative are no longer used; instead,
B-series classification is equated with Moore’s field-based ME and MT are classified as ‘‘Ego-relative temporal motion
frame and Núñez’ time-RP metaphor, except that this does constructions’’, and sequence-based ones as ‘‘positional
not entail the MT perspective (Table 5e). The integration time constructions’’, which is equated with Moore’s
with spatial FoRs is based on whether the relations (2011) field-based temporal FoR and McTaggart’s (1908)
involved are binary or ternary (Fig. 2) and on an analysis B-series (Table 5e0 ). It is unclear whether this shift in ter-
of how FRONT is assigned in each of these cases. minology is intended as a conceptual advancement. How-
A-series time always involves an Ego, and is thus a nec- ever, as Zinken himself argued in his earlier paper, such a
essary precondition for ternary relations (which are a nec- bipartite classification would be less powerful than his ori-
essary precondition for the relative FoR). However, it is not ginal tripartite FoR system.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 353

3.5. Reference frames of space and time (Tenbrink) pastwards reading as resulting from an MT perspective.
From the ME perspective, assignment of FRONT is derived
For her account, which aims to provide a cross-domain from Ego’s moving direction into the future. From the com-
framework, Tenbrink (2011) directly builds on Levinson’s plementary MT perspective, the event is seen as being
(2003) taxonomy and expands it by adding dynamic rela- moved into Ego’s FRONT area. The same reading (pastwards
tions to the static ones, and by transferring the resulting to Monday) could also arise from an absolute FoR, accord-
system from space to time. Her reference frames are ing to which the meeting has been simply moved towards
modeled along the following lines: (a) intrinsic, relative, earlier times, regardless of how it is related to Ego’s now.
and absolute concepts (or none of these), (b) static versus
dynamic situations, and (c) external versus internal rela- 3.6. Temporal frames of reference, t-FoR (Bender, Beller, and
tionships between entities (in space only). Based on these Bennardo)
distinctions, she describes 23 different reference frames
for space and 10 for time. Leaving aside the external/ In contrast to most other accounts, Bender and col-
internal distinction, we will focus in this review on her leagues (Bender et al., 2005, 2010, 2012; Rothe-Wulf
absolute, intrinsic, and relative FoR for static situations, et al., 2014; and see Table 5g) do not take the A-/B-series
but will also, where necessary, pay special attention to distinction as their starting point, but rely entirely on the
dynamic situations. design principles laid out for frames of reference by Levin-
The transfer of Tenbrink’s spatial taxonomy to time son (Section 2.2.3). Based on these principles, they propose
additionally makes use of the A- and B-series and their a set of possible FoRs (not necessarily attested to by exam-
mapping onto ME and MT perspectives. Following ples from English) that can be used to explore the extent of
Galton (2011), she describes time as having an ‘‘inbuilt diversity in temporal references.
asymmetry’’, which, however, can be conceived of in They define an absolute t-FoR as one that derives its ori-
two contrasting ways: as a vector from past to future entation from the superordinate field outside figure,
(associated with the deictic A-series), and as a vector ground, and observer. As space itself is the superordinate
based on anteriority/posteriority in sequences (associated field in the spatial domain, so is time in the temporal
with the non-deictic B-series), which points towards domain; and the asymmetry of time (i.e., its directionality)
earlier times (Table 5f). can be recruited for assigning orientation to this field: FRONT
In line with Moore (2006, 2011), the B-series in her is where the arrow of time is pointing to, namely (typi-
account corresponds to the Ego-free MT and thus to a cally) towards the future. Events ‘‘in front of’’ other events
field-based frame of reference; and the anteriority/posteri- or ‘‘moved forward’’ from their previous position would
ority relation within sequences of events is seen as war- thus be further in the future.
ranting its classification as absolute FoR (it cannot be An intrinsic FoR derives its orientation from the ground
intrinsic as events do not have an intrinsic FRONT according entity G; and the asymmetry of events (with a beginning
to Tenbrink, 2011). Her treatment of the A-series, however, and an end) can be recruited for providing these entities
diverges from Moore (2004, 2011) insofar as she does not with orientation: FRONT is assigned to that part of time per-
explicitly account for Ego-centered MT as instance of refer- tinent to the beginning of event G. Events ‘‘in front of’’
ence frames. Instead, she considers a distinction of static other events or ‘‘moved forward’’ from their previous posi-
versus dynamic situations as more important. tion would thus be in the past of the original date. Impor-
The static A-series situations in her account all have a tantly, in the t-FoR account, this also holds if Ego happens
binary structure, with G = Ego (as in (13)), thus affording to coincide with G, as what counts here as G in time is a
an intrinsic FoR (see also Zinken, 2010). temporal entity such as subjective present, and not Ego’s
looking direction.
(13) ‘‘Good times lie before me.’’ [intrinsic]
A relative FoR, finally, requires a ternary relation
FRONT is assigned in line with Ego’s looking direction and between figure F, ground G, and observer V. How F is local-
thus typically to the future. According to Tenbrink (2011), ized in reference to G depends on V’s subjective viewpoint,
ternary relations do not exist in temporal language for sta- and this viewpoint can change relative to the constellation
tic situations, and therefore no relative FoR exists either. In of F and G (e.g., by simply ‘moving’ through time). Cru-
contrast, the dynamic situations in her account afford more cially, it emerges as either one of two different (and in fact
than one FoR. Respective cases include binary relations as opposed) variants (Fig. 2): In the reflection variant, the pri-
in (14), which are classified as intrinsic (again with mary coordinate system originating in V is transferred into
G = Ego), and ternary relations as in (15). G by reflection and thus leads to the assignment of FRONT to
the time between G and V (i.e., nearer to V), whereas in the
(14) ‘‘I’m going forward in time.’’ [intrinsic]
translation variant, the primary coordinate system is trans-
(15) ‘‘Next Wednesday’s meeting has
ferred into G by translation, thus leading to the assignment
been moved forward two days.’’
of FRONT to the time beyond G (i.e., further away from V). In
At least in English, example (15) constitutes an ambig- either case, events are localized symmetrically in one’s
uous case, in which the meeting may be understood as past and future, and thus with diverging FRONTs and BACKs.
having been moved towards Friday (futurewards) or Mon- Such a relative-reflective (or ‘‘egocentric’’) pattern is nicely
day (pastwards). Both readings are classified by Tenbrink demonstrated, for instance, in the French terms for great-
(2011) as being based on a relative FoR: the futurewards grandchildren and great-grandparents, which are all
reading as resulting from an ME perspective and the suffixed by arrière, ‘‘behind’’ (Radden, 2004).
354 A. Bender, S. Beller / Cognition 132 (2014) 342–382

In the t-FoR account, A-series and B-series descriptions distinction concerns the reference point RP, which in other
may reflect certain FoRs more or less closely (e.g., a B-ser- taxonomies (and in this review) is denominated as ground
ies description of sequenced events invites an intrinsic entity G. Following Moore (2011) and Núñez and Sweetser
FoR), but are not seen as logically related. In particular, (2006), the two categories are labeled Ego-RP and Time-RP
A-series descriptions do not entail a relative FoR per se. A (Yu, 2012). The second—and novel—distinction concerns
genuine relative FoR presupposes that the introduction of the referent R (or figure entity F), with time-referent
the subjective viewpoint (V) of an observer (which has to (Time-R) and human-referent (Human-R) as the two cate-
be distinct from G) makes a difference for referencing; gories relevant for Chinese (Yu, 2012; and see Table 6).
and this is not the case for most A-series descriptions. The latter category consists of human sequences (either
For instance, a sentence like (16) as ‘‘deictic human frame’’ or ‘‘sequential human frame’’),
which are analogous to event sequences and thus map
(16) ‘‘Next Wednesday’s meeting [= (15)]
onto an earlier/later relation, with older people and gener-
has been moved forward two
ations ‘‘in front of’’ the sequence and/or Ego.
days.’’
The A-/B-series distinction is mentioned in passing as
could still be understood according to the intrinsic FoR in being related (and presumably overlapping with the RP
that ‘‘forward’’ is interpreted as being towards earlier distinction). ME and MT are mainly used for assigning
times, or according to the absolute FoR in that ‘‘forward’’ FRONT to temporal entities and/or relations. Yu’s account

is interpreted as being simply towards the future. largely remains silent regarding a possible correspondence
Likewise, the ME and MT perspective in the more gen- with (spatial) FoRs, leaning more towards Talmy’s (2000)
eral reading (Section 2.2.2) largely reflect different tempo- than Levinson’s (2003) taxonomy, specifically when it
ral FoRs, with ME reflecting the futurewards orientation comes to assigning FRONT in queue-like compositions such
aligned to the arrow of time (absolute FoR) and MT reflect- as event sequences. Such compositions are seen as
ing the pastwards orientation, towards the beginning of ‘‘encompassive secondary reference objects’’ (Talmy,
time (intrinsic FoR). However, they do not necessarily 2000; and see Moore, 2011). Their FRONT is derived from
always correlate with them: From both perspectives, alignment and/or moving direction and overrides the (pos-
future is where the arrow of time is pointing to (absolute), sible) orientation of single entities. In contrast to many
the beginning of events occurs earlier than their endings other scholars, but in line with Bender and colleagues
(intrinsic), and Ego and future are approaching each other and partly Zinken (2010), this account also grants an
(relative). The assignment of FRONT, arguably the hallmark intrinsic FRONT to times and events (Yu, 2012).
of a frame of reference, is not fully affected by the adoption
of a temporal perspective (except for ambiguous cases like 3.7.2. Route perspective versus survey perspective (Jamalian
(16), the reading of which may indeed be primed by move- and Tversky)
ment; see Boroditsky & Ramscar, 2002). In contrast to the aforementioned account that relates
to spatial FoRs only in passing, the account proposed by
3.7. Further accounts Jamalian and Tversky (2012) is not as elaborate with
regard to temporal concepts as the previous ones. Coming
For the sake of completeness, two more accounts will be from the research tradition in mental maps and spatial
briefly presented despite the fact that they are not yet navigation, these authors focus on the distinction between
widely disseminated: the temporal reference frames route perspective versus survey perspective, which they
account by Yu (2012), and the route versus survey perspec- equate with an intrinsic or egocentric FoR and an absolute
tives account by Jamalian and Tversky (2012). FoR, respectively. The route perspective in space is consid-
ered to be analogous to both the ME and MT perspective in
3.7.1. Temporal reference frames (Yu) time, taken from an embedded viewpoint and with Ego as
In order to resolve the controversy regarding whether, the reference point. In contrast, the survey perspective on
in Chinese, Ego is conceived of as facing toward the future space is considered to be analogous to a calendar view
or the past, Yu (2012) proposes two distinctions. The first on time, taken from an external viewpoint and with dates

Table 6
The four temporal reference frames distinguished by Yu (2012).

Note. A-series and B-series labels were added here for easier comparison.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 355

or events as the reference points (see also Núñez & the exact same scene. As detailed in Section 2.2.2, this sim-
Sweetser, 2006; Zinken, 2010). ple fact implies that the two distinctions address different
dimensions of temporal descriptions and, as a consequence,
4. Comparison of accounts cannot be mapped onto each other in any straightforward
manner.
As we have seen above, most accounts make use of the Some accounts (e.g., Kranjec, 2006; Núñez & Sweetser,
same set of conceptual components for construing tempo- 2006; Zinken, 2010) conceptualize both ME and MT as
ral FoRs (which is partly obscured by idiosyncratic label- involving a deictic center (Ego), as was intended in the
ing), but combine them in distinct ways. As one original version (Section 2.2.2), and therefore map them
consequence, each of the accounts proposed so far differs onto the A-series (Table 5). Other accounts (notably
in substantial ways from any other account. In this section, Moore, 2004, 2011; and see also Tenbrink, 2011) extend
we attempt to analyze similarities and differences between the notion of MT by also including non-deictic cases,
the accounts by addressing the following questions: (1) which they then map onto the B-series. As argued by
How are the conceptual sources related to each other for Bender et al. (2010), such an extension presupposes a
the construal of temporal FoRs? (2) On which principles more general view on temporal moving perspectives,
is FoR construal based, and how is FRONT assignment real- namely as a perspective on time as moving not only
ized in this process? And (3) what types of relations or ref- towards Ego but towards temporal entities and events
erencing patterns are predicted from each of these more generally. Since ME and MT are figure/ground rever-
taxonomies? In closing this section, we also explore (4) sals of each other, this must entail the possibility of the
the potential for integrating these accounts. reversed perspective: a non-deictic ME perspective
according to which not only Ego, but temporal entities
and events more generally are (figuratively) moving
4.1. Relations between conceptual sources
through time (in this case, of course, the perspective
should be labeled ‘‘moving event’’ or ‘‘moving entity’’,
The conceptual sources to be considered here encom-
rather than ‘‘moving Ego’’). In (16), for instance, the meet-
pass A- and B-series, the Moving Ego (ME) and Moving
ing moves towards the future; the same is possible (in
Time (MT) perspective, and up to four frames of reference
both directions) for other events such as floating holidays
(FoRs).
and vacations. As proposed in the account by Bender,
Beller, and Bennardo (2010; and see Table 4 above), the
4.1.1. Relations between A-/B-series and temporal FoRs ME/MT distinction would thus fully intersect the A-/B-ser-
Almost all of the accounts presented here (except for the ies distinction, at least in theory. Whether this theoretical
one by Bender et al., 2010) take deixis, in any one of its con- claim can be corroborated by empirical evidence remains
ceptual variants (e.g., A-/B-series, Ego-/field-based, Ego/ to be seen.
time reference point), as their starting point. In other words,
the basic distinction from which most taxonomies unfold is 4.1.3. Relations between ME/MT perspectives and temporal
the distinction between linguistic expressions that do or do FoRs
not entail a deictic center: Ego, or rather Ego’s present. This Those accounts that conceptualize ME and MT in the
is certainly attributable to the fact that most of these stricter sense as involving a deictic center also tend to
accounts originate from linguistics, and it accommodates equate both perspectives with a relative FoR (Kranjec,
the relevance of deixis, particularly for the relative FoR, 2006; and by inference perhaps Núñez & Sweetser,
which does indeed require the viewpoint V of an observer. 2006). The same holds for Zinken’s (2010) account, albeit
Taking deictic versus non-deictic expressions as the prime with the relative FoR being restricted to ternary relations.
distinction, however, may obscure the fact that the relations Accounts that also admit non-deictic MT cases (Moore,
between the A-/B-series classification on the one hand and a 2004, 2011; Tenbrink, 2011) match them to an absolute
FoR-based classification on the other are more complex FoR (Table 5). The account by Bender et al. (2010) diverges
(Section 2.2.2; and see Levinson, 2003; Moore, 2011). Binary from all of these accounts in various ways. Due to diverg-
relations in which Ego (or V) serves as the primary reference ing design principles, it proposes cases that do or do not
point (= ground G), such as (9) or (13) above, attest to this involve Ego for both the ME and MT perspective, and it
complexity as much as a ternary example like (15). The bin- refuses to conflate the two complementary perspectives
ary relations with G = V do not qualify as examples of a rel- under one FoR. As a consequence, non-deictic MT is seen
ative FoR (Tenbrink, 2011; Zinken, 2010), whereas a ternary as being compatible with an intrinsic FoR, and non-deictic
example like (15), despite clearly qualifying as an A-series ME as being compatible with an absolute FoR. In contrast,
description, could also be interpreted in line with an abso- the two variants of the relative FoR do not comply, in this
lute FoR (Bender et al., 2010; Tenbrink, 2011). Undoubtedly, account, with any general moving perspective.
however, the A-/B-series distinction helps to avoid confu-
sion between past and future on the one hand and earlier 4.1.4. Distinction between static and dynamic situations
and later on the other. Tenbrink (2011) argues that a comprehensive taxonomy
of temporal FoRs should include the distinction between
4.1.2. Relations between ME/MT perspectives and A-/B-series static and dynamic situations (Section 3.5), and as we will
Whereas A- and B-series are incommensurable with each see from the empirical data compiled below, this may
other, ME and MT provide complementary perspectives on indeed be important in some cases. With regard to her clas-
356 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Table 7
Assignment of FRONT depending on the adopted t-FoR, together with the principles on which this is based, in each of the accounts; further explanation is given in
the text.

sification of FoRs, however, it only has a bearing on her con- undisputed, there is no consensus on where this orientation
ceptualization of relative FoRs, which are claimed to exist in may come from. In the accounts reviewed here, at least three
dynamic but not static situations. Why this should be valid different principles can be identified for how orientation of
is an interesting question in itself (for counter-examples, the temporal field is conceptualized (see Table 7).
see Levinson & Majid, 2013; Radden, 2004; Shinohara & In line with causal relations and the change from per-
Pardeshi, 2011; Zinken, 2010), but in either case, this ception to memory, the arrow of time as conceptualized
restriction entitles us to collapse our review across static by most physicists and psychologists points towards the
and dynamic situations for all other FoRs. future (Section 2.1.3). This notion is enlisted by Kranjec
(2006) and Bender and colleagues (2005, 2010) to identify
4.2. Principles of FoR construal and FRONT assignment the orientation of the field as futurewards. In at least two
other accounts (Tenbrink, 2011; and, albeit with reserva-
As explained in Section 2.2.3, a frame of reference (FoR) tions, Moore, 2011), this same arrow of time is regarded
is a coordinate system required to establish the position of as emerging from the viewpoint of Ego, who happens to
a figure F in reference to a ground G from a given perspec- be aligned to the future, thus connecting it to an intrinsic
tive (which may or may not be an observer V). The FoRs FoR. For construing an absolute FoR, these accounts take
differ essentially in terms of where this coordinate system instead the earlier/later (anteriority/posteriority) relation,
originates and how it is oriented. The system’s origo may inherent in the sequence of events (akin to people in a
be in the superordinate field (in the case of the absolute queue), as the source from which they derive the orienta-
FoR), in G (intrinsic), or in V (relative) and is used to estab- tion of the field. According to these accounts, the absolute
lish its orientation. With regard to these design principles, field is therefore oriented pastwards or, more precisely,
spatial frames of reference (s-FoRs) do not differ from tem- towards earlier events.4 Finally, Zinken (2010) also consid-
poral frames of reference (t-FoRs).3 ers the absolute field as being oriented towards earlier

4.2.1. Absolute t-FoR 4


The same seems to hold for Jamalian and Tversky (2012), who equate
The construal of an absolute FoR requires an oriented the absolute FoR with a calendar view on time, with dates or events as the
superordinate field outside F, G, and V. In the spatial domain, reference points and ‘‘earlier/later’’ as the terms of reference. However, this
this field is space, and in the temporal domain, by analogy, it account leaves open whether the field is considered to be oriented towards
is time. But while the relevance of such an oriented field is earlier or later events. Yu (2012) does not specify his account with regard to
a FoR taxonomy, but as he leans on Talmy’s (2000) treatment of queue-like
compositions in a way similar to Moore (2011), we categorize him into this
3
It should be noted, though, that not all accounts reviewed here are group by analogy. Crucially, in his account, not only sequences of events are
based on these principles (see specifically Moore, 2011). referenced absolutely, but also sequences of human generations.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 357

times, but for a different reason. In his account, the field, in with a distinct and unambiguous assignment pattern:
which single events are embedded, consists of larger tem- nearer to Ego for the reflection, and further away from
poral entities—intervals like days, weeks, months, or years. Ego for the translation variant. This conception is in line
The beginning of these intervals is assumed to be their with Galton’s (2011) notion of radial half-axes, and it also
FRONT, and this has always taken place in the past of the conforms with early theoretical considerations put forward
embedded events. by Traugott (1975, 1978), who defined tense, in its sim-
plest form, as the distinction between proximal and distal
4.2.2. Intrinsic t-FoR relations, and thus as symmetrical to the deictic center.
The intrinsic FoR usually attracts little dispute, even
with regard to denomination. For it to be enacted, the 4.3. Referencing patterns
ground object G (or reference point) that serves as origo
of the system needs to have an orientation assigned to it. As a consequence of the different conceptualizations
Amazingly, however, in the temporal domain, little con- and principles for assigning FRONT, the accounts reviewed
sensus can be observed across the different accounts here also make qualitatively distinct predictions with
regarding what would count as an intrinsic orientation— regard to the number and type of the resultant referencing
or whether G, if it is an event, may have one at all (see patterns (see Table 8).
Table 7).
Tenbrink (2011), for instance, claims that this is impos-
4.3.1. Two patterns
sible (and see Zinken, 2010). For this very reason, the two
The early accounts of A-series versus B-series (following
accounts only consider cases in which the observer V coin-
McTaggart, 1908) as well as the approaches that build on
cides with the ground (G = Ego) as examples of an intrinsic
the ME versus MT perspective on time (Clark, 1973;
FoR. Derived from the assumed looking or moving direc-
Fillmore, 1971; and see, e.g., Boroditsky & Ramscar, 2002;
tion of Ego (prevailingly futurewards), FRONT is assigned to
Gentner et al., 2002; McGlone & Harding, 1998) basically
the future5 (in Zinken’s account only for A-series descrip-
distinguish two patterns of assigning FRONT, namely either
tions). In contrast, three accounts propose the opposite
futurewards (A-series, ME perspective) or pastwards/
assignment of FRONT to earlier times, although for different
towards earlier events (B-series, MT perspective). This pat-
reasons: Zinken (2010, for B-series phrases) and Kranjec
tern is reflected in the accounts of Moore (2004, 2011),
(2006)—as well as, perhaps, Núñez and Sweetser (2006)—
Núñez and Sweetser (2006), and Kranjec (2006)—and this
due to the metaphorical moving direction of G, derived from
despite the fact that Moore splits MT into an ego-centered
the earlier/later relation in event sequences; and Bender and
and a field-based version, and Kranjec even distinguishes
colleagues (2005, 2010) due to what they hold to be an
four different cases (extrinsic, intrinsic, and deictic in an
intrinsic FRONT of events (the latter assumption is also shared
ME versus MT version).
by Yu, 2012).

4.3.2. Three patterns


4.2.3. Relative t-FoR
Zinken (2010) and Tenbrink (2011) predict three differ-
To an even greater extent than the intrinsic FoR, the rel-
ent patterns: a simple pattern each for a futurewards
ative FoR seems to open up a wide field for diverging per-
assignment (for A-series, with Ego = G) and a pastwards
spectives and positions. Relative FoRs require a ternary
assignment (for B-series) as well as a more complex
relation, in which Ego (or V) cannot serve as the primary
pattern of assigning FRONT to the space between Ego and a
reference point (or G). But whether they exist in the
ground entity (which is in the future of Ego). In Zinken’s
domain of time, and, if so, in how many variants, remains
(2010) account, this occurs for all cases of A-series, in
disputed (see Table 7).
which Ego is not conflated with G. In Tenbrink’s (2011)
A relative temporal FoR is claimed not to exist at all by
account, it occurs for dynamic situations only, and only
Moore (2011) and not to exist for static situations by
according to one reading of her example (16), in which
Tenbrink (2011). Other accounts do admit its existence.
the entity to be moved is perceived as approaching. While
Zinken (2010), for instance, defines a relative FoR as entail-
these accounts offer a third pattern insofar as they qualify
ing the assignment of FRONT to the future of Ego—due to its
the pastwards movement by fixing it to the future of Ego,
looking direction—but past of or earlier than G (i.e., the
they remain silent on the question of what happens to this
interval between Ego and G). Kranjec (2006) and even
pattern in Ego’s past. It is thus not clear whether assigning
Tenbrink (2011, for dynamic situations) also propose a rel-
FRONT to the past of G (but the future of Ego) is qualitatively
ative FoR, but one that implies two diverging assignments
different from a more general assignment of FRONT to the
of FRONT: towards the future or later times from an ME per-
past and thus from a broader MT perspective, regardless
spective, and towards the past or earlier times from an MT
of Ego’s position.
perspective (see also Jamalian & Tversky, 2012; Núñez &
Sweetser, 2006).
Bender and colleagues (2005, 2010), finally, propose 4.3.3. Four patterns
two relative FoRs, a reflection and a translation one—each In contrast, the account by Bender and colleagues
(2005, 2010) explicitly distinguishes the set of general
5
Arguably, this may also be how one can categorize Yu’s (2012) deictic
movements (pastwards vs. futurewards) from a set of
time frame, which combines Ego-RP with Time-R (thus corresponding to the radial movements (towards Ego/now vs. awaywards from
classical A-series) and is oriented towards the future. Ego/now, both in past and future). This subsumes the
358 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Table 8
Patterns of FRONT assignment and/or forward movement; explanation is given in the text.

half-linear or half-radial patterns acknowledged by Zinken ceived of as being in one’s back, and the past as in front.
(2010) and Tenbrink (2011). However, this apparent arbitrariness in how such a crucial
question could be addressed in fact helps to highlight a
4.4. Potential for integration need for further conceptual clarifications related to a spa-
tio-temporal taxonomy of FoRs.
As is apparent from the above overview, conceptualiza- In order to settle the theoretical disputes, we propose a
tions differ dramatically for all of the FoRs, and for none is taxonomy entirely based on abstract design principles. FoR
there convergence among more than half of the accounts. construal according to this taxonomy depends on the num-
None of the accounts agrees completely with any of the ber of entities for which the relation has to be established
other accounts across all FoRs. And even where two (binary or ternary) and on the origo of the coordinate sys-
accounts propose the same type of FRONT assignment, this tem (in the superordinate field, the ground entity G, or the
may be for quite different reasons. Does this situation leave observer’s viewpoint V). While this is largely consensual
any leeway for us to decide objectively, which of the prin- across the accounts reviewed here, the current proposal
ciples adopted for the construal of the t-FoRs is more sen- differs from these accounts in how it treats linguistic
sible than the others? examples: For an intrinsic FoR to be diagnosed, it cannot
At first glance, this seems to be unlikely. For instance, be decisive whether, for instance, Wednesday’s meeting
the orientation of the temporal field required for the abso- is moved forward into the past, but whether the alignment
lute FoR can be derived from the arrow of time, which of FORWARD with past is derived from the ground (Wednes-
points towards the future, but it can also be derived from day), rather than from the field (time in general) or the
an extensive temporal interval that began in the past, in observer’s viewpoint (now).
‘‘the beginning of all time’’. This is further complicated by Like its counterpart in space, this taxonomy constitutes
the fact that, for some speech communities like Aymara a tool for assessing a broad range of possible references in
(Núñez & Sweetser, 2006), the future is apparently con- time. However, it remains abstract insofar as it contains at
A. Bender, S. Beller / Cognition 132 (2014) 342–382 359

least one free parameter, namely the principle according to Núñez and Sweetser (2006), Kranjec (2006), and Tenbrink
which orientation is assigned to the entity that serves as (2011), with the additional specification that FRONT is
origo for the coordinate system—in each of the FoRs under assigned not simply to the past, but to the past of G within
scrutiny here, albeit to varying degrees. How orientation is the future of Ego (i.e., to the interval between Ego/V and
concretely assigned depends almost entirely on cultural G); and the translation variant could encompass the ME
conventions and may thus vary across and perhaps even cases in these same accounts, with the additional specifica-
within speech communities. If two speakers diverge in tion that FRONT is assigned not simply to the future, but to
how they assign orientation to origo, an apparently identi- the future of both G and Ego.
cal expression would be generated by diverging FoRs, A second source for blurring is the question of whether
whereas adoption of the same FoR would generate diverg- Ego’s assumed looking or moving direction (into the
ing expressions. More specifically, the alignment of FORWARD future) may serve to assign FRONT in any of the FoRs. As
with past could be indicative of an absolute FoR (if time is detailed in Section 4.2.2, some accounts (Tenbrink, 2011;
seen as having its beginning in the past), or of an intrinsic Zinken, 2010) consider binary relations with Ego (V) = G
FoR (if events are seen as being oriented by their begin- as the only cases in which an intrinsic FoR can be adopted.
ning, or if events are seen as moving into the past). In these cases, Ego’s looking direction is seen as providing
In not presupposing a specific alignment, this account the basis for FRONT assignment. ‘‘Ego’’ in time, however, is
differs substantially from all accounts reviewed above, only used figuratively as person; in actual fact it is defined
and even from the initial t-FoR taxonomy. While Bender as an event or time point (i.e., Ego’s subjective present), for
and colleagues (2005, 2010) acknowledged such a free which looking direction does not make sense. Rather, Ego’s
parameter to be fixed by cultural convention in the case looking direction reflects the general view of where time
of the absolute FoR, they did not acknowledge it for the ‘flows’ to, and may thus be used as an indicator for how
intrinsic FoR, which they simply equated with pastward orientation of the absolute field is defined in this speech
alignment.6 Consequently, the original t-FoR taxonomy fell community (see also Lakoff & Johnson, 1999; Yu, 2012).
short of achieving a clear conceptual separation of the This concern is related to the third source for blurring,
design principles on the one hand and their (culture- namely the debate on whether temporal entities and/or
specific) instantiation on the other. the superordinate field can be regarded as oriented at all.
This conceptual separation, however, is crucial for As mentioned above, some accounts (including Zinken,
tighter integration of the accounts reviewed above. As 2010) claim that events do not have an intrinsic orienta-
indicated in Table 7, most of these accounts already agree tion. If, however, time in general, as well as time intervals
on the relevance of the design principles for FoR construal. like weeks or days more specifically, may be seen as direc-
This convergence, however, is partly blurred by four ted, thus providing orientation to the field in an absolute
sources of confusion, three of which can be resolved on sense (Zinken, 2010; and see Yu, 2012), any event serving
theoretical grounds. as ground G for the intrinsic FoR—by virtue of having a
The first source for blurring is the conceptualization of beginning in the exact same way—should also be seen as
the relative FoR, which has given rise to the most contro- directed. This is what actually allows the sequence of
versy—from claims that it does not exist at all, to sugges- events in B-series descriptions to be depicted by a vector
tions of one FoR or even two variants of it. This pointing towards earlier times.
controversy fizzles out if some of the somewhat under- The final source for blurring is the disagreement among
specified assumptions regarding the relative FoR are elab- accounts on what provides the basis for FRONT assignment
orated further. Three of the four accounts that do propose in binary temporal relations, with some scholars recruiting
relative FoRs distinguish two different cases that entail the same principle (e.g., the sequence of events) for justify-
diverging FRONT assignments, namely both to future and ing an absolute FoR that others use to justify an intrinsic
past (Table 7). This in itself is unsatisfactory: The purpose FoR. This disagreement cannot be resolved on theoretical
of adopting a FoR is to locate F in reference to G. Conse- grounds. As we argued above, the principle according to
quently, each FoR should give a non-ambiguous search which orientation is assigned to the origo for the coordi-
space for F. While this does not imply that relations can nate system is largely based on conventions. Whether
be described by only one FoR, it does imply that each FoR and how speakers assign FRONT to time and temporal enti-
should give only one relation of F to G. Assuming two dis- ties such as events can only be assessed empirically. The
tinct variants of the relative t-FoR (akin to the variants of second part of this review therefore scrutinizes the avail-
the relative FoR in space, as described in Section 2.2.3) able empirical data. Based on this analysis, the potential
solves this problem: The reflection variant of the relative for a more coherent theoretical integration will then be
FoR could encompass the MT cases in the accounts of discussed anew.
To summarize, while the ways in which each account
6 assigns FRONT in each of the temporal FoRs currently paint
In the case of the relative FoR, the free parameter to be fixed by cultural
conventions defines how the primary coordinate system with origo X in V is a rather confusing picture, it would indeed be possible to
transferred into G. As described in Section 2.2.3, this can be done by achieve more consensus and coherence: by taking the
rotation, reflection, or translation, thus giving rise to three variants, which design principles for FoR construal as the starting point,
differ in how the secondary coordinate system (anchored in G = X2) is by acknowledging the radial patterns involved in the rela-
oriented. The degree to which speakers may differ in this regard, has been
documented especially for the spatial domain (e.g., Beller et al., 2014;
tive FoR, by using Ego’s alignment with time (its ‘‘looking
Bender et al., 2012; Bennardo, 2000; Bennardo, 2009; Hill, 1982) and will direction’’) as an indicator of how field orientation is
be discussed for the temporal domain in Section 7.3.1. perceived, and by establishing the (culture-specific) princi-
360 A. Bender, S. Beller / Cognition 132 (2014) 342–382

ples for FRONT assignment in the intrinsic and absolute FoR for another person, or by selecting from a set of given
on empirical rather than theoretical grounds. options. The descriptions are then analyzed in terms of
One of the main differences between accounts simply which FoRs they are based on. This type of task can easily
arises from their disciplinary background and traditions. be transferred to the temporal domain.
Most of them, originating in linguistics, present natural
language examples that are then analyzed and sorted into
5.1.1. Elicitation: the Wednesday’s meeting task
categories. To various degrees, these categories are
The classical elicitation paradigm in the temporal
modeled after the A-/B-series distinction, the ME/MT-per-
domain revolves around moving an event (McGlone &
spectives on time, some reference point metaphors, or
Harding, 1998): Participants are informed that ‘‘The meet-
other related conceptual sources. Another approach, which
ing originally scheduled for next Wednesday has been
is more closely related to representational accounts in
moved forward two days.’’ When asked for the day on
anthropology, psychology or cognitive science, begins by
which the meeting will now take place, two answer
covering the space of conceivable reference patterns and
options are provided (Monday and Friday), from which
derives from those the design principles for frames of ref-
participants can choose. The question is ambiguous in that
erence, with the purpose of being broad enough to capture
roughly half of US participants tend to choose Monday and
a potentially large degree of cross-linguistic variability. The
the other half Friday. The former choice is considered to be
resultant grid overlaps substantially with the categories
consistent with an MT perspective and the latter with an
arising from the linguistic approach, but (presently) also
ME perspective.
contains empty cells. For instance, apart from the genera-
If only two perspectives are to be distinguished, one
tional terms reported by Radden (2004), no conclusive evi-
question is sufficient. If, however, more possible perspec-
dence has yet been provided for a relative-reflective
tives are at stake, the original question with the event
reference pattern in the temporal domain (contrary to its
located in the future needs to be supplemented by a ver-
pervasiveness in space). The crucial question, then, is
sion in which the event is located in the past (Bender
whether languages can be found that may provide exam-
et al., 2005; Núñez et al., 2006). According to the refer-
ples for the empty cells. Used in this way, such a grid
ence-point (RP) metaphors account (Núñez et al., 2006),
equips us with a powerful tool to discover new and inter-
the past version forces participants to choose between an
esting phenomena. If, however, these cells remain empty,
interpretation either relative to Ego’s front (leading to a
we will still have gained valuable information with regard
futurewards movement) or relative to the front of the
to cognitive constraints on temporal referencing.
sequence (leading to an earlier-movement). According to
the t-FoR account (Bender et al., 2005, 2010, 2012; Rothe-
5. Empirical investigations: methods and tasks Wulf et al., 2014), only the simultaneous consideration of
a past and a future reference discriminates between all
Regardless of how thoroughly scholars may have possible FoRs.
designed their cross-domain taxonomies for frames of In either case, however, we have no way of detecting
reference, the question of whether and how people really mistakes which people may make in responding: Every
do transfer spatial conceptualizations into the domain of single combination of responses produces a sensible pat-
time—or are otherwise affected by spatial cues when engag- tern. This is generally true regarding all previous research
ing in temporal reasoning—can only be answered through in this field, but it becomes more salient when the correct
empirical investigations. For the following overview, some diagnosis of specific FoRs is at stake. To address this con-
thirty studies were scrutinized, which differ with regard cern, more than one pair of questions should be used,
to the theoretical stance they took, the methods they which enables one to assess intra-individual consistency
adopted, and the findings they obtained. To facilitate com- in addition to FoR choice. While detecting inconsistent
parison, we will first describe in necessary detail their main answers will not reveal whether participants made a mis-
methods, before presenting and discussing their findings take or simply changed their minds, consistent answers
(Sections 6 and Section 7). The range of widely different may be interpreted as support for the assumption that a
tasks can be sorted into the following types: (1) language specific FoR has indeed be adopted.
elicitation, typically combined with some sort of priming,
(2) analysis of bodily expressions such as gesture and pos-
5.1.2. Linguistic priming (of the ME and MT perspective)
tural sway, (3) elicitation of spatial layouts on which tem-
Typically, the Wednesday’s meeting question is embed-
poral relations are mapped, and (4) implicit tasks that use
ded in some sort of experimental context. In the original
a reaction time paradigm with congruency priming.
study by McGlone and Harding (1998; and see Gentner
et al., 2002), this consisted of a list of similarly structured,
5.1. Language elicitation tasks but non-ambiguous temporal sentences like: ‘‘We passed
the deadline two days ago’’. When the context consistently
Generally, elicitation tasks present some kind of (lin- suggested an ME perspective, participants preferred the
guistic or non-linguistic) stimuli and ask participants to Friday response twice as often as the Monday response,
provide a linguistic description or response. This is the case whereas this pattern switched when the context suggested
in the classic reference tasks where arrangements of the MT perspective. The results are generally taken as evi-
objects have to be described (e.g., Pederson et al., 1998; dence for the psychological reality of the ME and MT per-
Senft, 1995), either by way of free answers or instructions spective: They prove that taking either of these temporal
A. Bender, S. Beller / Cognition 132 (2014) 342–382 361

moving perspectives affects the way in which people inter- and these gestures may convey meaning that is not only
pret an ambiguous temporal phrase. relevant to their verbal expressions but also completes
Moving on from there, Boroditsky and her colleagues them (Goldin-Meadow, 2003; Le Guen, 2011a). As gestures
crossed the ‘space–time barrier’ demonstrating that prim- naturally unfold in space, they also provide ecologically
ing spatial perspectives may have similar effects. The valid information about how abstract concepts are spa-
primes they used consist of spatial scenarios including a tially structured by the speaker (Cooperrider & Núñez,
picture and a sentence description, with the scenarios 2009; Núñez et al., 2012).
either depicting a moving observer or a moving object The task assigned to elicit gestural data is similar to the
(Boroditsky, 2000), or involve imagined and real motion previously described language elicitation task in that its
(Boroditsky & Ramscar, 2002). main purpose is to induce people to talk about a certain
domain, for instance by asking them to explain temporal
5.1.3. Visual priming (of additional perspectives and/or frames expressions (Núñez et al., 2012), to re-tell a brief story they
of reference) have just studied (Casasanto & Jasmin, 2012), or to tell the
In a similar vein, the Wednesday’s meeting question can (depicted) history of the universe (Cooperrider & Núñez,
be used to explore the psychological reality of other tempo- 2009). To capture co-speech gesture, people are videotaped
ral concepts or metaphors which are claimed to be also used while talking, for later transcription and analysis. For the
to process temporal information. Núñez et al. (2006), for investigation of spatial and temporal representations, cod-
instance, sought to provide evidence for the relevance of ing the position and orientation of the speaker (with regard
what they term the ‘‘Time-RP’’ metaphor (i.e., an Ego-free to the fixed bearings of a potential absolute frame of refer-
MT perspective; Section 3.2). In order to foreground the ence) is crucial. Ideally, this orientation is rotated midway
anterior/posterior relation, which is also inherent in non- through the interview to disentangle overlapping FoRs (Le
deictic event sequences (B-series), they used visual primes Guen, 2011a). It may also be revealing to contrast indoor
that consist of a graphical array of objects sliding horizon- and outdoor settings, as blocking visual contact with land-
tally across a screen. Likewise, in order to demonstrate the marks may suspend a possible adoption of the absolute
relevance of an ‘‘extrinsic’’ (= absolute) FoR, Kranjec (2006) FoR (Núñez et al., 2012). For data analysis, references to
attempted to foreground the superordinate field, with two deictic time versus non-deictic or sequence time need to
types of stimuli: a single entity moving over some kind of be distinguished, as these two are assumed to recruit dis-
ground, and the picture of a river coming down from a tinct axes (Casasanto & Jasmin, 2012; Emmorey, 2002;
mountain, for which the path of motion had to be indicated. Núñez et al., 2012).
Finally, Rothe-Wulf and colleagues (2014) intended to To compare spontaneous with deliberate gesture
directly investigate the relation between four distinct spa- production, Casasanto and Jasmin (2012) also asked their
tial FoRs and their temporal counterparts, as proposed by participants to make gesture demonstrations that they
Bender and colleagues (2010). In order to activate a specific believe would most naturally accompany speech about
(spatial) FoR, they used as primes three sets of pictures, all earlier and later times.
of which show a superordinate entity in motion (i.e., a con- A reversed strategy was employed by Jamalian and
veyor belt, train, or river) as well as smaller entities located Tversky (2012) who provided gestures together with the
in the superordinate entity; when required for the relative verbal description of a cyclical temporal sequence (e.g.,
FoR, an observer (outside this superordinate entity) was from seed to flower) or with the Wednesday’s meeting
added to the picture. The picture was accompanied by a task. Participants were then asked to depict the sequence
few sentences that describe the scene in a way compatible in schematic diagrams or indicate the next step after its
with one of the four FoRs to be foregrounded in the respec- completion, or to identify the date to which the meeting
tive condition. A final test question on the relation has been moved, respectively. Responses were then coded
between two entities ensured that participants indeed according to whether they are in line with the accompany-
adopted the primed FoR. ing circular versus linear gesture (in sequence description)
or forward versus backward gesture (the Wednesday’s
5.2. Analysis of bodily expressions meeting task).

Bodily expressions accompany most of what we try to 5.2.2. Postural sway


communicate, ranging from the emotions we are People do not only speak with their hands; often,
experiencing (Darwin, 1872; Meeren, van Heijnsbergen, & their whole body is involved. The fact that this type of
de Gelder, 2005; Sauter, Eisner, Ekman, & Scott, 2010) to information may reveal underlying conceptions of time
complex cognitive ideas (Goldin-Meadow, 2003; McNeill, has been shown by Miles et al. (2010). They requested
1992). The two types of bodily expressions that have been participants to embark on a mental time travel into the
used to date for investigating space–time mapping include future (i.e., to imagine what their everyday life circum-
co-speech gesture and postural sway. stances might be like four years in the future = prospec-
tion) or the past (to recall such circumstances four years
5.2.1. Co-speech gesture previously = retrospection) and to envisage the events of
People spontaneously and mostly unconsciously use a typical day. A magnetic motion-tracking system was
gesture in conjunction with words (Iverson & Goldin- used to measure movement (postural sway) in the hori-
Meadow, 1998; McNeill, 1992; Núñez & Sweetser, 2006), zontal plane.
362 A. Bender, S. Beller / Cognition 132 (2014) 342–382

5.3. Mapping temporal relations on spatial layouts the two calendar installations, participants are provided
with a larger number of tokens to be used as representa-
The third type of task, like the previous one, is largely tions of conventional time intervals (i.e., seasons and their
non-linguistic and recruits space as the medium for tem- sub-intervals or constituents, and times of the day). Partic-
poral representations, but does so in a less spontaneous ipants are then asked to ‘‘make a map of the year (or day)’’,
way. Over the years, the following variants have been in which each token should represent one interval of time.
developed, which differ in sometimes small, but important The design of this task differs from the previous one, in that
aspects, and are therefore described sequentially. it (i) more strongly emphasizes the cyclical character of the
temporal entities under scrutiny by using terms for the
5.3.1. Graphic productions annual and diurnal cycle and by illustrating the task with
The task was designed by Tversky, Kugelmass, and a circular diagram, (ii) provides a substantially larger num-
Winter (1991) to explore how space comes to be used for ber of tokens which, in principle, allows for a broader vari-
representing non-spatial relations. Five sub-tasks are used, ety of arrangements, and (iii) does not prompt participants
with one each devoted to spatial, imagined spatial, tempo- with a reference point (i.e., the central stage is not placed
ral, quantitative, and preference concepts (each including for them). The third part of the task, the time landscape
two or three parts). In the temporal sub-task, participants game, involves tokens to be used as representations of time
are provided with a reference point in space: A sticker, intervals. Two of these tokens (or a token and a doll) are
which represents the middle event of a tripartite sequence, placed in line, perpendicular to the participant’s gaze.
is placed on a blank sheet of paper. Sequences consist of One token is then moved along the imaginary line so that
meals, activities, and times of the day. The instructions it reverses its position in relation to the other token (or
run as in the following example (for children): the doll, respectively). Participants are asked to describe
what they saw.
‘‘Now I want you to think about the times of day that
we eat meals, breakfast, lunch, and dinner. I will put a
5.3.3. Picture arrangement task
sticker down for lunch time, and I want you to put a
Chan and Bergen (2005; Bergen & Chan Lau, 2012) devel-
sticker for dinner time and a sticker for breakfast time.
oped Tversky’s design into a card arrangement task, in which
Here’s where I’m putting a sticker for lunch time. Now
the tokens are not blank but contain black and white images
you put a sticker for dinner time (pause), and another
of the entities in question. The five sets of three cards each
sticker for breakfast time.’’ (Tversky et al., 1991, p. 526)
depict growing stages of a living being (i.e., tree, chicken,
butterfly, frog, and woman). As in the original study
The task produces two general types of data: the type of (Tversky et al., 1991), the cards have to be arranged on a
the representation (i.e., nonlinear, ordered, or interval), white sheet of paper ‘‘in sequence from the earliest to the
and its direction. Of interest here is the direction of the latest stage’’ (Bergen & Chan Lau, 2012, p. 3), but partici-
ordered representations. To be categorized as ‘‘ordered’’, pants are not prompted with a reference point.
stickers had to be separate, properly ordered (in any direc-
tion), and more or less in one line7; direction was scored as 5.3.4. Time arrangement tasks
left-to-right (LR), right-to-left (RL), top-to-bottom (TB), and The question that is most relevant for this review is
bottom-to-top (BT). Please note that, in this case, ‘‘top’’ whether the spatial FoRs that are preferred in a given
and ‘‘bottom’’ refer to the sheet of paper used as pad and speech community may have an impact on how its mem-
thus are located in the horizontal plane. In other studies bers conceptualize the passage of time. To address this
(e.g., Brown, 2012) and the remainder of this review, these question, the layout tasks had to be further modified: A
responses are categorized instead as far-to-near (FN) and rotation of participants halfway through the sitting is
near-to-far (NF), respectively. essential in order to disambiguate responses with respect
As described above (in Section 5.2.1), Jamalian and to whether they are based on an absolute s-FoR (oriented,
Tversky (2012, Exp. 1) asked participants to construct ‘‘a e.g., in line with cardinal directions) or a relative s-FoR
simple schematic diagram’’ to convey cyclical temporal (from the participant’s own perspective). The version of
sequences that were described to them verbally (accompa- the task designed specifically for this purpose consists of
nied by gestural priming). In this case, diagrams are coded two parts: a card arranging task and a time-points task
as circular (with the last event being connected back to the (Boroditsky, Gaby, & Levinson, 2008).
first) or linear (without any such connection). For the card arranging task, eight sets of four cards each
are used. The cards in each set contain photographs that
5.3.2. Temporal landscapes depict a temporal progression, such as stages in a life cycle
A similar principle is used by Zinken, Sampaio, da Silva or an event developing through time. Participants are asked
Sinha, and Sinha (2005; and see Sinha et al., 2011) for their to set them down ‘‘so that they are in the correct order’’.
temporal landscapes task, which consists of three parts. In For the time-points task, participants are provided with
a reference point—for instance, by pointing abstractly to
7
Examples of nonlinear responses included stickers forming a triangle a central spot in the air (affording a 3D representation)
rather than a line or representing events out of order (e.g., breakfast— or by placing a token on the ground directly in front of
dinner—lunch). Unfortunately, this category was excluded from the anal-
ysis, which eliminated not only 15–40% of the data for the youngest
the participant (for a 2D representation)—and are then
children and 6–25% for the older children, but also, and more importantly, asked to locate related temporal expressions in reference
potential information on nonlinear time concepts. to the first mark. More specifically, the instructions are
A. Bender, S. Beller / Cognition 132 (2014) 342–382 363

shaped as follows: ‘‘If I tell you that this here is today, rently popular actors) and non-lateralized (presented cen-
where would you put yesterday? And where would you ter-screen). Stimulus category (‘‘before’’ or ‘‘after’’
put tomorrow?’’ Altogether, twelve sets of three tempo- participants were born) is indicated by pressing one of
rally ordered terms each are used. two laterally displaced response keys.
The sessions are videotaped and responses coded for In a second experiment, verbal stimuli are presented
later analysis. The tasks are described in more detail in center-screen, followed by a white circle that may appear
the MPI Field Manual (Boroditsky et al., 2008; also available on the left or right side of the screen. Circle location is indi-
online) and are part of a larger cross-linguistic survey on cated by pressing one of two laterally displaced response
temporal representations (for respective studies, see keys. Whereas the question pursued in Experiment 1 (as
Boroditsky & Gaby, 2010; Brown, 2012; de Sousa, 2012; in previous studies) is whether congruent space–time
Fedden & Boroditsky, 2012; Fuhrman & Boroditsky, 2010; mapping facilitates temporal judgments, Experiment 2
Gaby, 2012; Le Guen & Pool Balam, 2012; Levinson & investigates whether temporal cues may affect visuospa-
Majid, 2013). tial attention and facilitate spatial responses.

5.4. Implicit tasks: reaction time paradigm based on 5.4.3. Speech balloon with tensed words, twofold displaced
congruency priming In addition to the stimuli used in the first task
(Santiago et al., 2007), the main experiment in Torralbo,
The final type of task is also largely non-linguistic, but Santiago, and Lupiáñez’ (2006) study introduces an exter-
takes an indirect approach by contrasting experimental nal perspective: Participants are presented center-screen
conditions that are either congruent or incongruent with with a side-view silhouette of a person’s head, looking
assumed spatio-temporal links. The task labels used here either left- or rightwards. The verbal stimuli are pre-
are invented on the spot (derived from the stimuli used) sented in a speech balloon and appear either in the back
to aid discrimination of and reference to these tasks in or front of the silhouette, either on the left or right side of
the remainder of the paper. the screen. In one version, stimulus category (past/future)
Stimuli typically belong to either a past or future cate- is indicated verbally, while in another version, it is indi-
gory (e.g., words in the past vs. future tense, or pictures cated by pressing one of two laterally displaced response
of persons or buildings from the past vs. present/future, keys. Trials can be congruent along the lateral (left/right)
of persons at a younger vs. older age, or events at an earlier axis and/or along the sagittal (back/front) axis (in the sec-
vs. later stage). Stimuli are presented on a computer screen ond version this is further crossed with response
and have to be categorized, typically by pressing one of congruency).
two keys. Mapping each stimulus category to one of the
two locations of the keys creates congruent and incongru- 5.4.4. Temporal entities of two stages, non-lateralized
ent conditions. For instance, mapping past onto the left key A similar goal is pursued by Boroditsky, Fuhrman, and
and future onto the right constitutes one condition (pre- McCormick (2011), but in a different version: Instead of
sumably the congruent one), while the reversed mapping crossing a lateral condition with a sagittal one, they cross
constitutes the other condition. Participants’ reaction time a lateral (horizontal) condition with a vertical one. To this
and accuracy in making their decision are recorded. If map- end, participants are presented with pairs of images, one
ping conditions produce a main effect, the faster (and/or after the other, all appearing center-screen. Each pair of
more accurate) condition is identified as the congruent con- images depicts the same entity at two different stages
dition, thus revealing the existence and direction of the (e.g., a picture of a person at a younger and an older age).
underlying cognitive space–time mapping. The experimen- Participants are asked whether the second image shows a
tal designs described in the following modify this general conceptually earlier or later time-point than the first
idea in various ways, some of which are crucial for data image. Responses are given by pressing one of two adja-
interpretation. cent keys. For half of the participants, keys are arranged
horizontally (on the left/right axis), while for the other half
5.4.1. Tensed words, laterally displaced they are arranged vertically (perpendicular to the tabletop,
Participants are presented with verbal stimuli, one after on the up/down axis).
another, which may appear either on the left or right side
of the screen. The stimuli consist of tensed verbs and tem- 5.4.5. Buildings of two ages, non-lateralized
poral adverbs. Stimulus category (past/future) is indicated A similar principle is adopted by Miles, Tan, Noble,
by pressing one of two laterally displaced response keys Lumsden, and Macrae (2011), with the exception that the
(Santiago, Lupiáñez, Pérez, & Funes, 2007; for different vertical response is not defined in 3-dimensional, but in
versions of this task, see also Ouellet, Santiago, Funes, & 2-dimensional space. Participants are presented with
Lupiáñez, 2010a; Ouellet, Santiago, Israeli, & Gabay, images, one after another, all appearing center-screen.
2010b). The images depict buildings and cities, representing the
past or a (science fiction) future. Stimulus category (past/
5.4.2. Actors of two ages, non-lateralized future) is indicated by pressing one of four spatially dis-
A similar design is used in Experiment 1 of Weger and placed response keys, to which participants have to move
Pratt’s (2008) study, but with two crucial modifications: their finger right versus left (horizontal condition) or up
Their stimuli are non-ordered (pictures of formerly or cur- versus down (vertical condition). Please note that, although
364 A. Bender, S. Beller / Cognition 132 (2014) 342–382

for vertical responses the same keypad was used, it was English (the same is true for ‘‘advancing X’’ and ‘‘pushing
mounted on an incline to ensure up/down movements to X back’’; see McGlone & Harding, 1998). To the best of
some extent. our knowledge, these three phrases provide the only
instance of such an ambiguity and thus have to shoulder
5.4.6. Temporal sequences, non-lateralized the entire weight of empirical testing in this field. The fact
A version that extends stimuli presentation to larger that in many other languages, these phrases are not dis-
sequences was designed by Santiago, Román, Ouellet, cernibly ambiguous—in the sense of producing equally split
Rodríguez, and Pérez-Azor (2010) to investigate not only readings—renders cross-linguistic comparisons a contro-
congruency effects for space–time mapping, but also a pos- versial endeavor.
sible distance effect for temporal entities that are more or In contrast to this linguistic task, the documentation of
less removed from a reference point. To this end, partici- bodily expressions (Section 5.2), the spatial layout tasks
pants are presented with an unfolding sequence of every- (Section 5.3), and the implicit tasks based on congruency
day activities (in one version by way of video clips, in a priming (Section 5.4) do not take verbal output as their
second version by way of pictures) at full screen size. This prime data, but recruit space as the (implicit) medium
is followed by an order judgment task: Whether a given for expressing temporal relations. As participants are typi-
scene comes ‘‘before’’ or ‘‘after’’ the reference scene is indi- cally not aware of the purpose of the tasks, they are likely
cated by pressing one of two laterally displaced response to respond spontaneously, thus revealing implicit repre-
keys. sentations. In stark contrast to the implicit congruency
tasks, the other two types (bodily expressions and spatial
5.5. Discussion of methods and tasks layout tasks) entail a largely open format, which makes
them highly suitable for exploratory purposes. But even
The four types of tasks described above differ not only these tasks are not entirely impartial: Postural sway and
in the details of how they are designed and conducted, most versions of the layout tasks (except for the 3D version
but also in the questions they allow us to address. In the of the time-points task; Section 5.3.4) are restricted to 2-
following, we briefly discuss their strengths and weak- dimensional responses in the horizontal plane, thus pre-
nesses, before turning to the data obtained from them. empting vertical representations. And in some cases (such
For identifying temporal frames of reference and as possibly in the temporal landscape task described in Sec-
assessing the degree to which they reflect space–time tion 5.3.2), the tokens used may be too large in size and
mapping, a language elicitation task provides a necessary number to afford unrestricted arrangement in the available
first step. After all, temporal referencing is a linguistic working space (Sinha et al., 2011, p. 151). Even the gestural
activity, and systematic data collection on how people data (including the 3D pointing task) is constrained to
do this provides the foundation for any subsequent anal- some extent by the body, which blocks genuinely back-
ysis. In space, this can be accomplished in a straightfor- wards pointing.
ward manner, and large-scale surveys have bestowed on Furthermore, the basic question of whether space–time
us a cornucopia of data on cross-cultural variability in mappings are considered sensible in a speech community
spatial FoR use (e.g., Beller et al., 2014; Bennardo, 2002; cannot be answered unambiguously with these tasks. Both
Haun et al., 2006, 2011; Hüther et al., 2013; Levinson, bodily expressions and spatial layouts have an inherent
2003; Levinson & Wilkins, 2006; Majid et al., 2004; spatial dimension. Co-speech gestures and postural sway
Senft, 1997). The same is not true for the temporal unfold in space, regardless of the domain for which they
domain—partly due to its non-tangible nature, and partly may be emblematic. Likewise, abstract pointing and
due to the ongoing debate on what might count as an arrangement of tokens is also fundamentally spatial in
act of referencing in contrast to simply reading off the nature. In all of these cases, any possibly observed cross-
meaning from the words used. domain consistency could be attributed to this shared spa-
The Wednesday’s meeting task (see Section 5.1.1) pro- tial dimension, whereas unsystematic response patterns
vides an exception in that it allows for more than one read- may have multiple causes (Bender et al., 2012).
ing, which opens the door for experimental manipulations The implicit tasks based on congruency priming, on the
aimed at testing potentially influential factors. However, other hand, do allow the extent of such space–time map-
the task is peculiar in at least two ways. First, it employs ping to be assessed. Most often, however, they are fairly
a dynamic situation (the movement of an event), whereas restrictive with regard to the dimension for which congru-
in the spatial domain, static settings are more typically ency is assumed and/or scrutinized (in most examples for
used. As mentioned above, there is some controversy as this type of task, this is the lateral left/right axis). More-
to whether static and dynamic settings are fundamentally over, task-specific characteristics appear to affect partici-
different (Tenbrink, 2011) or may be covered by the same, pants’ responses (Torralbo et al., 2006).
slightly modified typology of FoRs (Bender et al., 2010, To sum up, the Wednesday’s meeting task is useful for
2012; Levinson, 2003). The second peculiarity is related discriminating several t-FoRs (at least to a certain extent)
to the previous concern in that this whole line of research and for identifying factors influencing their adoption. Its
hinges on one type of phrase and its possible readings: weaknesses are its singularity, the fact that it predomi-
‘‘Moving X forward’’ is not only inherently under-specified nantly produces linguistic data, and that these data are
in theory (Rothe-Wulf et al., 2014), but also happens to not unambiguous. Bodily expressions and spatial layouts,
produce ambiguity in practice, at least among speakers of on the other hand, are non-verbal and more implicit, and
A. Bender, S. Beller / Cognition 132 (2014) 342–382 365

provide an efficient way to assess spatial representations and Pratt (2008) collected their data in Toronto/Canada,
of time. However, the evidence they obtain is at least partly and Miles and colleagues (2010, 2011) in Aberdeen/
confounded with the inherent spatial nature of the task. Scotland.
Implicit congruency tasks, finally, yield more valid data A wealth of data is available on patterns of reference
on the true extent of space–time mapping, but cannot be and spatial representations (overview in Levinson, 2003),
used for exploring the full space of mapping options, and on spatial gesture (e.g., Goldin-Meadow, 2003; McNeill,
cannot be informative of referencing preferences, as they 1992), and on metaphorical space–time mappings in lan-
typically pre-define response patterns (e.g., along the left/ guage (e.g., Bennett, 1975; Clark, 1973; Evans, 2003;
right dimension). Since no single task is adequate to Haspelmath, 1997; Lakoff & Johnson, 1980, 1999; Miller
answer the full set of questions in a satisfactory manner, & Johnson-Laird, 1976; Radden, 2004; Traugott, 1978). In
the best advice that can be derived from this overview is a nutshell, these studies indicate that speakers of English
to combine different methods and exploit a variety of data make use of all three spatial FoRs for describing locations
sources. and movements in space (with a preference for intrinsic
and relative in small-scale space, and absolute in large-
scale space), that spatial vocabulary is mapped onto time
6. Empirical evidence
(which is what triggered this line of research in the first
place), and that time in co-speech gesture is also spatially
The following overview comprises some thirty empiri-
construed (Cienki, 1998; McNeill, 1992).
cal studies on temporal representations and space–time
The metaphorical space–time mapping adopted in lan-
mapping. With few exceptions, none of these studies
guage suggests preferential recruitment of the sagittal axis,
explicitly addresses temporal FoRs; however, their findings
with past and future mapped on BACK and FRONT, respectively
are relevant and instructive for assessing the theoretical
(e.g., the future ‘‘ahead’’, and olden days ‘‘passed by’’). As
accounts presented in the first part of this review. The
we have seen above, however, the reverse direction is also
studies compiled here include, in total, speakers of sixteen
possible, attested to in language elicitation tasks (Sec-
different languages, which will be pooled into nine clusters
tion 5.1), in which only half of the speakers tend to move
based on relatedness. Although some of the studies were
Wednesday’s meeting ‘‘forward’’ to a future date, while
designed as cross-linguistic comparisons, we first present
the other half moves it pastwards.
data for each language cluster separately (current section;
Interestingly, however, this linguistic preference was
overview in Table 9); this allows us to provide some
confirmed on a behavioral level only in a minority of
descriptive information on each of the clusters as a back-
empirical studies. In a spatial layout task, Fuhrman and
ground for the empirical findings—especially for those
Boroditsky (2010, Exp. 1b [3D pointing]) found a small per-
cases for which strong claims on variability have been
centage of responses along the sagittal axis. Postural sway
put forward. The relevance of these findings for the theo-
is also compatible with a BF9 time line (Miles et al., 2010),
retical questions posed above is then discussed in
and data on deliberate gestures reveal that people do concep-
Section 7.
tualize time along this time line (Casasanto & Jasmin, 2012,
Exp. 1). Data on spontaneous gestures, however, indicates
6.1. Indo-European languages adoption of the sagittal axis for deictic time only and with
considerable flexibility (i.e., no marked preference for BF
Of the languages belonging to the Indo-European fam- over FB; see Casasanto & Jasmin, 2012, Exp. 2). More impor-
ily, four have been investigated to a considerable extent tantly, these data attest to a preference of the lateral axis
in this field of research: English, Swedish, German, and over the sagittal axis, specifically for sequence time
Spanish. We take English as the starting point—for the sole (Casasanto & Jasmin, 2012). Adoption of the sagittal axis
reason that most studies were conducted with speakers of for deictic and of the lateral axis for non-deictic expres-
this language and that respective findings are occasionally sions—with future in front or to the right, and past in the
(even if implicitly) taken as a reference point or even as back or to the left—has also been documented for American
indicative of temporal representations and processing in Sign Language ASL (Emmorey, 2002) and French (Calbris,
general. 1990).
The remainder of the findings, including some gestural
6.1.1. English data, most spatial layout data, and all congruency priming
English belongs to the Germanic branch of the Indo- data, coherently indicate a preference for the lateral axis in
European language family. According to Ethnologue8, it is LR direction (see Table 9). Although the design of most of
the native language to 335 million people around the these studies restricted the range of possible responses to
world, and an official or national language in more than one or two dimensions, this constraint does not hold gen-
50 countries. With two exceptions, the findings reported erally and does not invalidate the general trend (see espe-
here all originate from studies conducted in the US cially Fuhrman et al., 2011).
(especially in California, with some additional data from The empirical studies across a broad range of methods
Illinois, Minnesota, Pennsylvania, and Hawai’i); Weger thus suggest that native speakers of English recruit two
distinct axes: (i) the sagittal axis, presumably in both
8
Where available, information on numbers of speakers was taken from
Ethnologue (http://www.ethnologue.com/; see also Lewis, Simons, & Fennig,
9
2013). Abbreviations of directions are explained in footnote to Table 9.
366 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Table 9
Preferences for dimension and direction of the time line axes (direction in brackets were given with still substantial, but lower frequency than main responses).

Language Main findings Type of task Design n References


English BF and FB Moving meeting BF design (ling.) 279 Bender et al. (2010), Rothe-Wulf et al.
(2014)a
LR Gesture Open (LR stimuli) 50 Cooperrider and Núñez (2009)
BF (and LR) Gesture (delib.) Open 32 Casasanto and Jasmin (2012, Exp. 1)
LR (and BF/FB) Gesture (spont.) Open 28 Casasanto and Jasmin (2012, Exp. 2)
BF Postural sway Open 20 Miles et al. (2010)
LR Spatial layout On plane 520 Tversky et al. (1991)
LR Spatial layout Open 13 Boroditsky (2008)
LR (and BF ¼ NF) Spatial layout On plane and open 24 Fuhrman and Boroditsky (2010, Exp. 1)
LR Spatial layout On plane 10 Boroditsky and Gaby (2010)
LR Spatial layout On plane 10 Bergen and Chan Lau (2012)
LR Congr. priming LR design 50 Weger and Pratt (2008)
LR Congr. priming LR design 85 Fuhrman and Boroditsky (2010, Exp. 2+3)
LR Congr. priming Horiz. vs. vertical 19 Miles et al. (2011, Exp. 1)
(inclined)
LR Congr. priming Horiz. vs. vertical 118 Boroditsky et al. (2011)
LR Congr. priming 3D 59 Fuhrman et al. (2011)
LR Congr. priming LR vs: TB ½¼ FN 32 Chen and O’Seaghdha (2013)

Swedish BF Moving meeting BF design (ling.) 284 Rothe-Wulf et al. (2014)

German FB Moving meeting BF design (ling.) 279 Bender et al. (2010),


Rothe-Wulf et al. (2014)
LR Congr. priming LR design 30 Ulrich and Maienborn (2010; Exp. 1)
BF Congr. priming BF design 60 Ulrich et al. (2012, Exp. 1)

Spanish BF Congr. priming BF þ LR 57 Torralbo et al. (2006)


LR LR design
LR Congr. priming LR design 32 Santiago et al. (2007)
LR Congr. priming LR design 96 Santiago et al. (2010)
LR Congr. priming LR design 93 Ouellet et al. (2010a)
LR Congr. priming LR design 20 Ouellet et al. (2010b)

Hebrew More RL than LR Spatial layout On plane 366 Tversky et al. (1991)
RL (and BF ¼ NF) Spatial layout On plane and open 24 Fuhrman and Boroditsky (2010, Exp. 1)
RL Congr. priming LR design 82 Fuhrman and Boroditsky (2010, Exp. 2 + 3)
RL Congr. priming LR design 28 Ouellet et al. (2010b)

Arabic RL (and TB [= FN]) Spatial layout On plane 283 Tversky et al. (1991)
Mandarin (US) LR (and UD) Spatial layout Open 42 Boroditsky (2008)
LR and UD Congr. priming Horiz. vs. vertical 63 Boroditsky et al. (2011)
Mandarin (P.R. China) FB Moving meeting BF design (ling.) 163 Bender et al. (2010)
LR (and TB [= FN]) Spatial layout On plane 33 Bergen and Chan Lau (2012)
UD (and LR) Congr. priming 3D 75 Fuhrman et al. (2011)
LR Congr. priming LR vs: TB ½¼ FN 40 Chen and O’Seaghdha (2013)

Mandarin (Taiwan) LR and TB [= FN] (and RL) Spatial layout On plane 38 Bergen and Chan Lau (2012)
LR (and UD) Spatial layout Open 15 Boroditsky (2008)
TB [= FN] Congr. priming LR vs: TB ½¼ FN 32 Chen and O’Seaghdha (2013)

Mandarin (Singapore) LR and TB [= FB/UD] Spatial layout On plane 32 Miles et al. (2011, Exp. 2)
LR and TB [= FB/UD] Congr. priming Horiz. vs. vertical 25 Miles et al. (2011, Exp. 1)
(inclined)
Cantonese LR (and RL) Spatial layout On plane and open 10 de Sousa (2012)
Tongan BF + FB Moving meeting BF design (ling.) 120 Bender et al. (2010)

Yupno down-/uphill Gesture Open 27 Núñez et al. (2012)


Yélî Dnye LR > NF/FN > East/West Spatial layout On plane 10 Levinson and Majid (2013)
Kuuk Thaayorre East/West Spatial layout On plane 14 Boroditsky and Gaby (2010)
East/West Spatial layout On plane 6 Gaby (2012)
Aymara FB (and LR) Gesture Open 30 Núñez and Sweetser (2006)
Tzeltal Maya LR > NF > Spatial layout On plane 12 Brown (2012)
East/West > DU >
down-/uphill
A. Bender, S. Beller / Cognition 132 (2014) 342–382 367

Table 9 (continued)

Language Main findings Type of task Design n References

Yucatec Maya cyclical (radial) Gesture Open 10 Le Guen and Pool Balam (2012)
LR > DU > RL > cycle > FN > NF Spatial layout On plane and open 26 Le Guen and Pool Balam (2012)
Amondawa LR and RL Spatial layout On plane 4 Sinha et al. (2011)

Note. The following abbreviations are used (here and in the remainder of this paper):
LR = lateral axis (left/right, without indication of direction), LR = left-to-right, RL = right-to-left;
BF = sagittal axis (back/front, without indication of direction), BF = back-to-front, FB = front-to-back;
UD = vertical axis (up/down, without indication of direction), UD = up-to-down, DU = down-to-up (observed as pile-up solution in the card task only); in
contrast to TB in real space;
TB = top–bottom (top/bottom, without indication of direction), TB = top-to-bottom; BT = bottom-to-top; please note that this orientation is specified in the
reference frame of a sheet of paper or the computer keyboard, with ‘‘top’’ referring to the top (further away side) of the page or keyboard, respectively (as
used, e.g., by Bergen & Chan Lau, 2012; Tversky et al., 1991); more accurately, it would be classified as FN = far-to-near or NF = near-to-far, respectively (in
the case of Miles et al., 2011, reclassification as FB or even UD is also acceptable).
a
Converging data on English with language elicitation tasks was also collected, among others, by McGlone and Harding (1998), Boroditsky (2000),
Boroditsky and Ramscar (2002), Kranjec (2006), and Núñez et al. (2006). Although not reported here, these are referred to in the discussion of findings in the
text.

directions (preferably BF when linguistically referring to speakers of the two languages differ fundamentally in
deictic time, with past events as in their back and future how they respond to the Wednesday’s meeting task: The
events as ahead, but also reversed in some cases), and (ii) vast majority of Swedish speakers move the meeting
the lateral axis exclusively in LR direction, when depicting ‘‘forward’’ to a later date, German speakers to an earlier date
the unfolding of events, for instance in writing, graphs, or (Bender et al., 2010; Rothe-Wulf, Beller, & Bender, 2014).
signed language. Both axes are not only cognitively avail- Recruitment of the sagittal axis BF in speakers of
able for information processing, but also appear to be German was also observed in a congruency priming task,
embodied to a certain extent, as emerging in co-speech in which responses to past- versus future-related sen-
gesture and postural sway. tences had to be made by moving a (sagittal) slider back
and forth (Ulrich et al., 2012). However, when left- versus
right-hand responses were required instead, the obtained
6.1.2. Swedish and German
congruency effect reflected a time line along the lateral
Swedish and German, like English, belong to the Ger-
axis LR (Ulrich & Maienborn, 2010). As these two studies
manic branch of the Indo-European language family.
used the exact same material (i.e., tensed sentences that
Swedish is spoken as the native language by roughly 8 mil-
either did or did not have meaning), it appears unlikely
lion people in Sweden (with some additional distribution
that recruitment of either axis (exclusively) depends on
in Denmark and Finland). German is spoken as the native
the deictic or non-deictic nature of the stimulus.
language by roughly 84 million people, mostly in Central
Europe (with some additional major settlements in
Kazakhstan, Namibia, and Paraguay), and is an official lan- 6.1.3. Spanish
guage in five states. The findings reported here originate Spanish belongs to the Romance branch of the Indo-
from studies conducted in Sweden (Göteborg) and Ger- European language family. It is spoken as the native lan-
many (Freiburg and Tübingen), respectively. guage by roughly 400 million people around the world,
Compared to English, considerably less research has and is an official or national language in more than 20
been conducted in these two languages on patterns of ref- countries. The findings reported here all originate from
erence and spatial representations (and none, to the best of studies conducted in Spain (largely from Granada).
our knowledge, on temporal co-speech gesture; for excep- The majority of studies so far predominantly explored
tions, see Ladewig, 2011; Müller, 1998, 2008). Yet, speak- space–time mappings by way of congruency priming tasks
ers of German are known to make use of all three spatial and report a strong preference for the lateral axis in LR
FoRs for describing locations and movements in space, direction (Ouellet et al., 2010a, 2010b; Santiago et al.,
with a preference for the intrinsic and relative FoR (and 2007, 2010; Torralbo et al., 2006). By requesting a left/right
here for the reflective variant) in small-scale space, and mapping on response keys, however, the task specifics may
the absolute FoR in large-scale space (Beller et al., 2014; have suggested this kind of mapping. The fact that
Grabowski & Weiß, 1996a, 1996b; Herrmann & mapping may switch depending on such context factors
Grabowski, 1998; Herrmann & Schweizer, 1998), and most has been demonstrated by Torralbo et al. (2006). In one
likely the same is true for speakers of Swedish (Grabowski version of the speech balloon task (Section 5.4.3),
& Weiß, 1996a). The existing literature also reports exten- responses had to be given by pressing one of two keys,
sive metaphorical space–time mappings in language (e.g., and participants exhibited a main effect of LR congruency.
Grabowski & Miller, 2000; Grabowski & Weiß, 1996a, However, in the version in which responses were given
1996b; Hellberg, 2007; Radden, 2004). vocally, the LR congruency effect gave way to a BF congru-
As in English, the linguistic space–time mapping ency effect. These findings indicate a preference for the
adopted in Swedish and German suggests preferential sagittal mapping for time, but sufficient flexibility in this
recruitment of the sagittal axis, with past and future mapped preference to be overridden in tasks that emphasize the
on BACK and FRONT, respectively. Interestingly, however, lateral axis (Torralbo et al., 2006).
368 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Two aspects are additionally noteworthy. First, the of all responses, aggregated across English and Hebrew
adopted temporal perspective—even for deictic stimuli— speakers, are reported to ‘‘put later events further in front
need not be egocentric, but may be allocentric. The point of the body 93% of the time’’).
of view to be taken in the speech balloon task is anchored
in the face silhouette on the screen, and aligned to its per- 6.3. Chinese
spective. In a similar fashion, the congruency priming task
by Santiago and colleagues (2010), described in Sec- China is home to speakers of roughly 300 different
tion 5.4.6, asked participants to adopt a specific scene languages, which belong to eight major language families.
within a sequence as the reference point and to decide Chinese is the name for a cluster of languages that together
whether another scene comes ‘‘before’’ or ‘‘after’’ it make up the Sinitic branch of the Sino-Tibetan language
(whereas from the participants’ point of view, all scenes family. Many of the languages in the Chinese cluster are
had come before now). In both studies (combined not mutually intelligible, and some of these are comprised
n = 126), participants exhibited congruency effects relative of varieties that are not mutually intelligible either. Two of
to this requested point of view. the largest languages in this cluster are Mandarin and Yue
And second, the study by Santiago and colleagues (Cantonese). Mandarin is spoken as the native language by
(2010) also obtained a distance effect: Responses were fas- roughly 850 million people, basically in Northern and
ter, when the target scene was more distant from the ref- Southwestern China, and is an official language in the
erence scene. In other words, proximal scenes are harder People’s Republic of China (P. R. China), Taiwan, and Singa-
to discriminate from a reference scene than distal pore. Cantonese is spoken by roughly 62 million people in
scenes—in either direction, past and future. P. R. China, Singapore and Malaysia, and is the official lan-
guage of Hong Kong and Macau.
6.2. Hebrew and Arabic The findings reported here originate from studies con-
ducted with speakers of Mandarin living in the US, in P.
Hebrew and Arabic belong to the Semitic branch of the R. China (especially Guangzhou, Shanghai, and Shijiazhu-
Afro-Asiatic language family. Hebrew is spoken by 5 mil- ang), in Taiwan, and in Singapore, and with speakers of
lion people, basically in Israel (with some additional distri- Cantonese living in Hong Kong and Macau. Notably, with
bution elsewhere, specifically as the liturgical language); just one exception, none of these studies was conducted
Arabic is spoken by roughly 225 million people in Northern in an area where Mandarin is the prevailing language (in
Africa and the Middle East, and is an official language in 26 Guangzhou this would be Cantonese, in Shanghai Wu, in
states. All findings reported here (including those on Ara- Taiwan Min or Hakka, and in Singapore Hoklo). However,
bic speakers) originate from studies conducted in Israel. despite its (linguistic) status as one of many Chinese vari-
The variable that has attracted most interest is the—com- eties, Mandarin is also an official language and often the
pared to English—reversed writing direction. In their semi- lingua franca in most Chinese-speaking states; it is the vari-
nal study, Tversky and colleagues (1991) compared ety on which most writing is based; and the people who
speakers of three languages with different writing systems: participated in most of the following studies are reported
English, which is written left to right (LR), Arabic, which is to be highly fluent in Mandarin.
written right to left (RL), and Hebrew in which letters are With regard to spatial frames of reference, speakers of
written RL, but most characters are formed, numbers writ- (Mandarin) Chinese make use of all three major FoRs,
ten, and arithmetic operations performed in the opposite albeit with considerable regional variation (e.g., the prefer-
direction, LR (Tversky et al., 1991; and see Shaki, Fischer, & ence for the absolute FoR is stronger in Northern parts than
Petrusic, 2009). As detailed in Section 5.3.1, the study com- in the South; see Li & Zhang, 2009), and they prefer the
prised five tasks, and the main finding for the temporal task translational variant of the relative FoR over the reflective
was a strong effect of language: a clear LR preference among one (Beller et al., 2014).
English speakers, a strong RL and secondary TB preference For most of their history, the two Chinese languages
among Arabic speakers, and more RL than LR among Hebrew under scrutiny here were written in vertical columns
speakers (please recall that, what in the classification of top-to-bottom (TB), ordered from right-to-left (RL). This
Tversky et al., 1991, is labeled TB, is actually performed has gradually changed in recent decades and to a diverging
along the sagittal axis and thus corresponds to FN). These extent: Whereas in Taiwan, the traditional direction is still
findings were taken as evidence for the assumption that also used, writing in mainland China nowadays almost
temporal relations are indeed represented by spatial means, exclusively follows the Western pattern LR and TB
and that the direction of writing and reading affects the (Bergen & Chan Lau, 2012; Chen & O’Seaghdha, 2013),
direction of the temporal representation. and the same is true for Singapore (Miles et al., 2011). In
A preference for the lateral axis (RL) in Hebrew has also contrast to Mandarin, Cantonese has no conventionalized
been found in subsequent studies and with modified tech- writing and is largely used for informal communication.
niques (Fuhrman & Boroditsky, 2010; Ouellet et al., 2010b), Direction of writing in Hong Kong and Macau is domi-
some of which, however, had a design that constrained nantly LR, followed by TB, whereas RL (which was domi-
responses to the lateral axis. The only other spatial layout nant between the 1920s and 1950s), is nowadays
task (Fuhrman & Boroditsky, 2010, Exp. 1b [3D pointing]) exceedingly rare (de Sousa, 2012).
also produced a small percentage of responses along the In addition to the vertical direction of writing and read-
sagittal axis, but unfortunately, the exact proportion of ing, Mandarin also makes systematic and frequent use of
responses and direction of the axis remain unclear (18.2% vertical metaphors for space–time mapping. In line with
A. Bender, S. Beller / Cognition 132 (2014) 342–382 369

the ‘river model’ of time, earlier times and the past are said section of the current article is confined to findings for Chi-
to be ‘‘up’’, whereas later times and the future are ‘‘down’’ nese and pays specific attention to those studies that are
(Scott, 1989; Yu, 1998). informative on both dimension and direction.
Which of these two aspects affects metaphorical Findings with Mandarin speakers in the US (whose
construals of time, if indeed either of them does so, is a country of origin and native language is unclear) indicate
matter of ongoing controversy. The point of origin for this a preference for the lateral axis LR, accompanied by a sec-
controversy was a publication by Boroditsky (2001), who ondary preference for the vertical axis UD in an open
argued that the availability of vertical metaphors in design (Boroditsky, 2008), and activation of both axes in
Mandarin (in contrast to English, which almost exclusively a congruency priming task (Boroditsky et al., 2011). Similar
uses horizontal spatial metaphors) affects temporal rea- effects, by and large, are reported for Mandarin speakers in
soning. In a priming paradigm, participants had to work Singapore (Miles et al., 2011) and Taiwan (Bergen & Chan
on two spatial judgment tasks, followed by a temporal Lau, 2012; Boroditsky, 2008; Chen & O’Seaghdha, 2013,
judgment task. The former consisted of questions on the although the latter found no evidence of an LR congruency
relation between two objects, arranged either horizontally effect).
or vertically, whereas the latter consisted of questions on In contrast, findings for Mandarin speakers from P. R.
the spatial relationship between two temporal entities China remain mixed: indicating an LR preference over TB
(such as ‘‘March comes earlier than April: true or false?’’). [= FB] in an open design (Bergen & Chan Lau, 2012), UD
The spatiotemporal condition used the terms ‘‘before’’ (over LR) in one congruency priming task (Fuhrman et al.,
and ‘‘after’’, whereas the purely temporal condition used 2011) and exclusively LR in another congruency priming
‘‘earlier’’ and ‘‘later’’. In the spatiotemporal condition, both task (Chen & O’Seaghdha, 2013).
English- and Mandarin-speaking participants were faster A mix of reasons may contribute to this mix of findings,
after the horizontal than after the vertical spatial primes. one clearly being regional variations. No two studies were
The same was true in the temporal condition for the Eng- conducted with the same language community, as indi-
lish speakers, whereas the Mandarin speakers showed cated in the introduction to this section: Some studies
the reversed pattern, being faster after the vertical primes. included mixed samples of Chinese emigrants to the US
This latter finding was taken as support for the claim that (Bergen & Chan Lau, 2012; Boroditsky, 2008; Boroditsky
Mandarin speakers conceptualize time vertically, whereas et al., 2011); others were conducted in Shijiazhuang in
English speakers do so horizontally. This is only partly Hebei province (Bender et al., 2010), in Guangzhou, where
motivated, however, by the usage of vertical metaphors Cantonese is dominant (Chen & O’Seaghdha, 2013), in
in addition to, but not instead of, horizontal metaphors in Shanghai, where Wu prevails (Fuhrman et al., 2011), in Tai-
Mandarin. Mandarin speakers should therefore be more wan (Boroditsky, 2008; Chen & O’Seaghdha, 2013), and in
sensitive to vertical priming than English speakers (but Singapore (Miles et al., 2011), respectively. As noted above,
about as likely to respond to vertical as to horizontal preferences for frames of reference may vary within the
priming). same country (Li & Zhang, 2009), and different traditions
Controversy arose when subsequent studies reported in writing and reading in some of these places may also
difficulties in replicating these findings (Chen, 2007; affect the way in which people represent temporal rela-
Chen & O’Seaghdha, 2013; January & Kako, 2007; Tse & tions (Bergen & Chan Lau, 2012; Boroditsky et al., 2011;
Altarriba, 2008; for successful replications and extensions, Chen & O’Seaghdha, 2013; de Sousa, 2012).
see Boroditsky, 2008; Boroditsky et al., 2011; Fuhrman Matters may be complicated further by the fact that,
et al., 2011; Miles et al., 2011; for a recent review on the what is labeled ‘‘horizontal’’ and ‘‘vertical’’ in this discus-
debate, see Chen & O’Seaghdha, 2013; an overview of the sion, often conflates two dimensions: horizontal the sag-
main findings is provided in Table 10). The first generation ittal and lateral dimension, especially in the stimuli used
of these studies was concerned with the cross-linguistic in the earlier studies (e.g., Boroditsky, 2001), and vertical
comparison and with questions pertaining to the degree the sagittal and genuinely vertical dimension, especially
of linguistic relativity involved. They were therefore in the response modes in some of the later studies
focused on the dimension of space–time mapping and (e.g., Bergen & Chan Lau, 2012; Chen & O’Seaghdha,
remained oblivious to questions of its direction. This 2013). In these latter cases, ‘‘vertical’’ does not refer to

Table 10
Relation of vertical (v) to horizontal (h) conceptions of time in English and Mandarin.

Type of task
Language elicitation with priming 3D Implicit task with congruency priming
pointing
Participants’ Boroditsky January and Chen Tse and Boroditsky Boroditsky Fuhrman Miles Chen and
language (2001) Kako (2007) (2007) Altarriba (2008) et al. (2011) et al. (2011) et al. O’Seaghdha
(2008) (2011) (2013)
English v<h v=h (v = h) v>h v<h v<h v<h v<h v<h
Mandarin v>h – v=h v>h vh v=h v=h v=h v < h (Chin.)
v > h (Taiw.)

Note. The data summarized here show whether effects were stronger or weaker for vertical (v) than for horizontal (h) primes or arrangements, or whether
they were basically equal.
370 A. Bender, S. Beller / Cognition 132 (2014) 342–382

the upright orientation people typically adopt, but to the meeting task (see Section 5.1.1) in a variety of ways:
plane in which sheets of paper or computer keyboards 25.0% of responses reflect an absolute t-FoR, 36.7% an
are placed. The resulting ‘‘top/bottom’’ (TB) classification intrinsic t-FoR, and 30.8% the translation subtype of the
therefore usually translates into ‘‘front/back’’ (FB) or even relative t-FoR (Bender et al., 2010). Although the latter
only ‘‘further/nearer’’ (FN) along the sagittal axis in ‘real was predominantly chosen by the older generation and
space’, which may be more relevant than previously may thus be moribund, it attests to the possibility of such
assumed: Data collection with the Wednesday’s meeting relative FoRs.
question reveals a preference for the FB direction along
the sagittal axis, with ‘‘forward’’ being mapped onto a 6.5. Yupno and Yélî Dnye
pastwards movement (Bender et al., 2010). Findings from
spatial layout tasks are at least compatible with, if not Yupno and Yélî Dnye are both categorized as Papuan
indicative of, the same FB (rather: FN) direction languages, but are not related to each other. Yupno is spo-
(Bergen & Chan Lau, 2012; Miles et al., 2011), and the ken by 5000 people in the Finisterre Range in northeastern
same is true for some of the congruency priming tasks Papua New Guinea, whereas Yélî Dnye is an isolated
(Chen & O’Seaghdha, 2013; Miles et al., 2011). Not com- language spoken by the 5000 inhabitants of Rossel Island,
patible with a sagittal interpretation are the data off the southeastern tip of Papua New Guinea.
reported in Boroditsky (2008) and Fuhrman et al. Speakers of Yupno prefer an absolute FoR for spatial
(2011), who used open formats or offered response descriptions on different scales, based on the topographic
modes in 3D (Boroditsky et al., 2011, had no sagittal contrasts uphill/downhill, co-located with the source and
option available). As long as only writing direction is mouth of the Yupno river (Wassmann, 1994). Yupno has
assumed to affect temporal representations (as in the isolated expressions for mapping spatial conceptualiza-
case of Bergen & Chan Lau, 2012), this conflation of ver- tions on time, but not in a systematic manner (Núñez
tical on paper and vertical in real space does not hamper et al., 2012). Despite this low degree of linguistic mapping,
consistent argumentation. If, however, genuinely vertical data on co-speech gesture (Section 5.2) indicate a transfer
(‘‘up/down’’) space–time metaphors are a possible candi- of referencing preferences. Gestural production while
date for affecting temporal representations, these two explaining temporal expressions reveal systematic spatial-
dimensions need to be distinguished. ization of time: towards the ground indicating the present
The influence of writing direction itself also remains (or deictic center, which is spatially co-located with the
disputed. Four studies explicitly addressing this question speaker), downhill indicating the past, and uphill the future.
report convincing evidence in favor of such an influence: This pattern, albeit deictically anchored in the speaker and
de Sousa (2012) found generational differences in tempo- his or her subjective present, is claimed to be largely allo-
ral representations as a function of changes in the direction centric and thus indicative of the same absolute FoR that is
of Chinese writing in past decades; Bergen and Chan Lau preferentially used in space (Núñez et al., 2012).
(2012) found spatial layouts in line with the prevailing Speakers of Yélî Dnye prefer an absolute (or geocentric)
(and distinct) writing directions in China versus Taiwan; FoR for spatial descriptions, based on a combination of the
and Boroditsky et al. (2011) as well as Chen and mountain/sea axis with an East/West axis. They also make
O’Seaghdha (2013) found the same pattern for congruency use of a body-based intrinsic FoR and to some extent even
priming. In contrast, two studies report strong vertical pat- of a relative FoR (Levinson, 2006). Co-speech gesture is
terns also for speech communities where Chinese is writ- geocentrically anchored and geographically accurate
ten LR (Fuhrman et al., 2011, Exp. 1; Miles et al., 2011). (Levinson & Majid, 2013), and this partly constrains ges-
Apparently, writing direction is a powerful factor, but most tures on time. Spatial locations are recruited only when
likely not the only one to have an impact on temporal referring to the position of the sun or moon (other repre-
representations. sentations include gestures along the East/West axis and
along a vertical axis with downwards denoting here and
6.4. Tongan now, in contrast to upwards for a distant point in time).
The basic concept of time is described as cyclical with-
Tongan is an Austronesian language spoken in the Poly- out calendrical fixed points, and the language possesses a
nesian Kingdom of Tonga, a small island group in the rich system for the grammaticalization of time. Linguistic
Southwest Pacific. Its (approx. 140,000) speakers make space–time mapping, however, is limited, and data collec-
use of all three basic FoRs for spatial descriptions: For tion with the time arrangement tasks (see Section 5.3.4)
small-scale (and static as well as dynamic) settings, they did not yield systematic patterns, either in the card arrang-
prefer the intrinsic FoR and the relative FoR in its transla- ing task or in the more abstract (2D) time-points task. The
tional variant (Beller et al., 2014; Bender et al., 2012; slight majority for relative responses (prevailingly left-to-
Bennardo, 2000, 2009); for large-scale settings, they prefer right) is most likely due to the fact that those who partic-
the absolute FoR, which, depending on context, may be ipated were at least partly literate.
based on a radial land/sea axis or on cardinal directions
(Bennardo, 2000, 2009). 6.6. Kuuk Thaayorre (Pormpuraaw settlement)
Linguistic overlap for space and time is considerable
(Bender et al., 2005). In line with the range of FoRs pre- Kuuk Thaayorre is a small Paman language spoken by
ferred for space, Tongan speakers are also reported to Australian Aborigines in the settlement Pormpuraaw (Cape
respond in linguistic elicitations with the Wednesday’s York Peninsula, Queensland; approx. 250 speakers). For
A. Bender, S. Beller / Cognition 132 (2014) 342–382 371

describing locations and movements in space, Kuuk Tha- downhill/uphill slope of the land, but they also make use
ayorre speakers, like the Yupno, prefer the absolute FoR, of a body-based intrinsic FoR and its projections (Brown,
invoked by one of six absolute directional root forms that 2012; Brown & Levinson, 1992; Levinson & Brown, 1994).
refer to the four cardinal directions and the north and south Yucatec speakers, on the other hand, appear to prefer
banks of a nearby river. Although systematic data on co- intrinsic references over absolute ones, based on cardinal
speech gesture are still lacking, some evidence is reported directions (Bohnemeyer & Stolz, 2006; Le Guen, 2011a).
for pointing towards the position of the sun for indicating Despite these diverging preferences, however, co-speech
time, and for pointing eastwards for indicating a distant gesture is geocentrically anchored and geographically
past. The very few expressions applicable across domains accurate in both groups (Brown, 2012; Le Guen, 2011a,
consist of one term for both ‘‘place’’ and ‘‘time’’ and one term 2011b; and see Haviland, 2003, 2005). More importantly,
encoding both the intrinsic relation ‘‘in front of’’ and tempo- the two groups differ with regard to space–time mapping.
ral priority (Gaby, 2012). Despite this paucity of linguistic Tzeltal provides a wide range of spatial expressions that
space–time mappings, Kuuk Thaayorre speakers respond can be mapped onto time, thus giving rise to a wide range
to the time arrangement tasks (Section 5.3) in line with an of temporal representations, which include the following
absolute FoR: mapping past to the East and future to the concepts: (1) deictically anchored (ego-centered) time vec-
West (Boroditsky & Gaby, 2010; Gaby, 2012). As the Tha- tors, (2) deictically anchored static sequences of time peri-
ayorre do not speak of future as westwards, this might be ods, (3) time as change of state or of location along a
regarded as another example of how the idea of absolute ref- unidirectional time line, (4) time as a unidirectional vector
erences is transferred from space to time. oriented uphill-wards, and (5) cyclic time. The relationship
between the two domains is therefore highly variable, and
6.7. Aymara only occasionally is the future located according to an
absolute FoR, namely uphill (Brown, 2012).
Aymara is one of the two languages that make up the Data collection with the time arrangement tasks (see
Aru (or Jaqi) family, which—besides Quechua—is one of Section 5.3.4) did not yield systematic patterns, either in
the two dominant language families of the central Andes the card arranging task or in the (2D) abstract time-points
in South America. Aymara is spoken by approximately task. A slight majority in the latter for relative responses
2.8 million people in Bolivia, Peru, and Chile. For spatial (prevailingly left-to-right, followed by near-to-far) rather
descriptions on different scales, Aymara speakers make indicates a preference for temporal representations, which
preferential use of an absolute FoR, which is based on the is at odds with preferences for space. With regard to ges-
cardinal directions. Interestingly, aligned with this coordi- ture, convergence across domains appears to be stronger:
nate system is the intrinsic coordinate system emanating In line with the absolute FoR underlying spatial gestures,
from the body, so that East is mapped onto ‘‘front’’ and Tzeltal speakers routinely point to different positions in
West mapped onto ‘‘back’’. More generally, people, objects, the sky to indicate where the sun would be. Occasionally,
and land are conceived of as having a canonical orientation they also point back over the head or shoulder to indicate
Eastwards, facing the sunrise (Núñez & Cornejo, 2012). the past (Brown, 2012).
When talking about time, Aymara speakers have been In contrast, speakers of Yucatec Maya, who make more
observed to use both lateral and sagittal gesture: the for- frequent use of an intrinsic FoR in speaking about space
mer for depicting non-deictic time (i.e., earlier/later rela- and of an absolute FoR in their co-speech gestures, avoid
tions), and the latter for deictic time. In deictic gestures, mappings of temporal entities onto any distinct space
FRONT is assigned to the past, and BACK to the future. In and thus tend to point towards the ground for the here
previous work, this was related to the VISION-IS-KNOWLEDGE and now, and upwards for distant past or future events.
metaphor and a focus on evidentiality, with what one This resistance to cross-domain mappings is also observed
can know (because of personal experience) as providing in the time arrangement tasks, in which a substantial num-
the basic motivation for the FRONT-to-past mapping ber of participants came up with a piling strategy: bottom-
(Núñez & Sweetser, 2006). In the light of the new findings up, so that the most recent one covers all of the older ones
on preferred spatial orientation (Núñez & Cornejo, 2012), (Le Guen & Pool Balam, 2012).
however, one might also argue that the canonical absolute
orientation towards the sunrise provides the basis for the 6.9. Amondawa
FRONT-to-past mapping (as both are located in the East).
Amondawa, finally, is a Kawahib language that belongs
6.8. Tzeltal and Yucatec Maya to the Tupí-Guaraní branch of the Tupí family. It is spoken
by approximately 100 people living in the Uru-eu-wau-
Tzeltal and Yucatec belong to two different branches of wau reservation in Rondônia in Greater Amazonia (Brazil).
the Maya language family. Tzeltal is basically spoken in the Besides documentation of the diverse lexical and construc-
highland of Chiapas (Mexico), Yucatec in the lowlands of tional repertoire for the conceptualization and expression
the Yucatán Peninsula of Mexico and northern Belize of location and spatial motion (Sinha et al., 2011), data
(approx. 370,000 and 770,000 speakers, respectively). on frames of reference or co-speech gesture were not
For describing locations and movements in space, collected.
neither of these languages makes use of a full-fledged rel- Linguistic analyses show that none of the spatial
ative FoR. Tzeltal speakers prefer the absolute (geocentric) expressions is used across domains to situate an event in
FoR, which derives its orientation from the overall relation to a temporal reference point, and no term for an
372 A. Bender, S. Beller / Cognition 132 (2014) 342–382

abstract notion of time exists. Instead, temporal intervals is most obvious for extension: Every spatial representation
such as those belonging to the seasonal and diurnal sys- extends along at least one dimension. And for all but three
tems are expressed by sun-based terms. In three spatial of the examples mustered for this review, temporal data
layout tasks, this lack of space–time mapping at the con- patterns do indeed reflect this property. Almost all of these
structional level was further explored. The two calendar also preserve the properties of linearity and directionality:
installation tasks (Section 5.3.2) exhibited curvilinear rep- Most representations seem to take the shape of a line
resentations, likely fitted into working space, which and are oriented in one direction (with circles being looped
revealed no clear preference for either left-to-right (LR) lines, directed clockwise or counterclockwise). Direction,
or right-to-left (RL) orientation. The time landscape game in turn, provides the foundation for assigning FRONT and is
produced expressions like ‘‘The sun/dry season goes’’, thus essential for construing temporal FoRs, as discussed
‘‘. . . has passed across’’, or ‘‘. . . is coming’’, or ‘‘The night below (Section 7.3).
is coming behind [the sun]’’ (Sinha et al., 2011, p. 159). One of the cases that provide an exception to the pres-
These expressions may be regarded as instantiations of lin- ervation of spatio-temporal properties is Amondawa. Its
guistic space–time mappings. However, the authors con- speakers allegedly do not employ expressions for spatial
sider this an unwarranted over-interpretation, as all of location and/or motion when constructing temporal
these expressions were elicited in situations that involve expressions (Sinha et al., 2011). The empirical data seem
spatial motion (and non-deictic sequences). They conclude, to contradict this claim, as they appear to reveal adoption
instead, that Amondawa speakers do not conceptualize of the lateral axis for temporal representations, but due to
events as occurring in Time as Such, and that they do not some of the specifics of the task and the small sample size,
employ expressions for spatial location and motion to con- no solid inferences can be drawn. The other exceptions are
struct temporal ones (Sinha et al., 2011). Yucatec Maya, whose speakers avoid spatial mappings in
gesture and who do not exhibit consistent patterning in
spatial layout tasks (Le Guen & Pool Balam, 2012; see
7. Theoretical implications
Brown, 2012, for similar observations among Tzeltal
Maya), and Yélî Dnye, for which space–time mapping is
One of the main purposes of this review is to scrutinize
patchy in language, gesture, and behavioral tasks
the accounts of temporal frames of reference (FoRs) pro-
(Levinson & Majid, 2013). A final exception, which is not
posed to date in the light of the available empirical data.
listed above but is recurrent in discussions on this topic,
More precisely, we intend to investigate whether and
is Whorf’s claim that the Hopi do not conceptualize time
how these data would be interpreted by each of these
as anything spatially extended, but as purely temporal:
accounts, and whether and to what extent they are com-
‘‘Nothing is suggested about time except the perpetual
patible with their theoretical predictions. To this end, we
‘getting later’ of it’’ (1956, p. 143).
will now look at the empirical findings with the following
Each of these exceptions may serve as instance of a cul-
questions in mind: Which properties and concepts of time
tural focus on the fourth property of time, its transience,
are reflected in the observed cultural patterns (Sec-
which implies a change of state and the fleetingness of
tion 7.1)? How do they relate to the construct of a mental
each single moment, meaning that it cannot be expressed
time line (Section 7.2)? What do the observed concepts of
spatially in any straightforward way. As most tasks involve
time and the culture-specific mental time lines reveal
a spatial dimension (especially those on bodily expres-
about the principles of FoR construal and the psychological
sions, spatial layout, and congruency priming), empirical
reality of the proposed temporal FoRs (Section 7.3)? And to
support for the absence of space–time mappings is hard
what extent, and based on which grounds, are spatial FoRs
to obtain—unless speakers shoulder the effort to avoid such
really mapped onto time (Section 7.4)?
mapping, as the Yucatec Maya appear to do (Le Guen &
Pool Balam, 2012). And yet, even some of the spatial data
7.1. Properties and concepts of time indicate transience more than any other property: It is
reflected in the piling-up strategy during card arrange-
In Section 2.1, we discussed the properties of time and ment, especially by Mayan speakers (Brown, 2012; Le
how they may or may not be mapped onto space, and we Guen & Pool Balam, 2012), and in gestures that indicate
identified possible concepts of time. Which of these proper- time by pointing to the corresponding position of the sun
ties are preserved, and which concepts are reflected in the in the sky (Brown, 2012; Gaby, 2012; Levinson & Majid,
space–time mappings reported for the sixteen languages 2013; Sinha et al., 2011).
described above? While these questions lie at the heart
of our endeavor, they were addressed only indirectly by 7.1.2. Concepts of time
the studies reviewed here. Central to most of them is a con- The concepts of time considered here include the linear,
cern with the mental time line, but, as we try to demon- cyclical, and radial concept. As described in Section 2.1.2,
strate, respective findings can be used to shed light on all preserve—at least to some extent—the properties of
other questions as well. extension, linearity, and directionality.
The linear concept is typically considered to be the pre-
7.1.1. Spatial and non-spatial properties of time vailing concept of time in human thinking, and the data
The properties of time comprise extension, linearity, compiled in Table 9 support this assumption: Spatial lay-
directionality, and transience (Galton, 2011). Three of out data reveal, in almost all cases, fairly straight, linear
these are largely preserved when time is spatialized. This arrangements; to locate past and future, linguistic phrases
A. Bender, S. Beller / Cognition 132 (2014) 342–382 373

employ dichotomous contrasts along a one-dimensional (Bender et al., 2010). Such an ‘‘anthropocentric’’ view of
continuum; gesture and postural sway do the same; and time also shines through in English, with the observer
congruency priming tasks found congruency effects along ‘towering’ above both past and future (Radden, 2004).
the same continua. And the distance effect observed in congruency priming
The preponderance of the linear concept, however, does tasks by Santiago and colleagues (2010; see Section 5.4.6)
not preclude the existence of other concepts. For instance, demonstrates that proximal scenes are harder to discrimi-
our experience of recurring time periods and cyclical nate from a reference scene than distal scenes—in either
sequences may also motivate a cyclical concept of time. direction, regardless of past or future.
In the past, the cyclical concept has typically been associ- It should also be noted that some of the data reported as
ated with non-Western cultures such as Hinduism evidence for a linear concept do not really allow a decision to
(Sharma, 1974), the Hopi (Malotki, 1983; Whorf, 1956), be made between a linear and a radial concept, as the
Maya (Farriss, 1987; León-Portilla, 1990) or South-Ameri- adopted axis begins or ends in Ego. This holds especially
can Toba (Klein, 1987)—and has given rise to a fair amount for data on the vertical (top/bottom) axis that should be
of controversy. Of the methods reviewed here, only the more accurately coded as nearer/further (Bergen & Chan
documentation of co-speech gesture and some of the spa- Lau, 2012; Chen & O’Seaghdha, 2013; Tversky et al., 1991).
tial layout tasks are suitable for detecting such a cyclical The nearer/further axis is also employed in Tzeltal and Yuca-
concept in the first place. Respective data suggest that, tec Mayan (Brown, 2012; Le Guen & Pool Balam, 2012).
although this concept may not be pervasive, it does exist;
Tzeltal Maya (Brown, 2012) speak of time as cyclical; 7.2. The mental time line
Bergen and Chan Lau (2012) report small proportions of
cyclical arrangement of cards in P.R. China (3%) and Taiwan Recent years have seen accumulating evidence for the
(8%); and Le Guen and Pool Balam (2012) found instantia- proposition that people process a variety of abstract infor-
tions of the cyclical concept among the Yucatec Maya, both mation such as number, size, speed, or time by mapping it
in spatial layout tasks (9%) and in gesture (46%). onto a linear spatial representation (Walsh, 2003). The
Notably, findings like these are not restricted to non- mental time line, a construct widely discussed in this
Western cultures. For instance, Tversky, Kugelmass, and domain, is assumed to extend in a more or less spatial
Winter (1991; see Section 5.3.1) report that 15–40% of manner, along one dimension, in one direction, and poten-
the data for the youngest children and 6–25% for the tially ad infinitum. This notion reflects three of the four
older children consisted of nonlinear responses, including properties of time, namely extension, linearity, and direc-
triangular (i.e., circular) instead of linear arrangements. tionality, and is compatible with a linear concept of time.
And Jamalian and Tversky (2012) obtained 24% circular As detailed in the previous subsection, however, some
diagrams (in contrast to 67% linear diagrams) from Eng- of the observed representations are at odds with the time
lish-speaking adults in a control condition, and even 67% line construct and indicate that it may not be universal
circular diagrams in the condition with circular gesture. (the same has been claimed for the related mental number
The radial (or ego-centric) concept, finally, stresses line; see Bender & Beller, 2011; Núñez, 2008, 2011; Núñez,
asymmetry between proximal and distal events more than Doan, & Nikoulina, 2011). The cyclical concept of time pre-
asymmetry between past and future, thus emphasizing— serves the properties of extension, linearity, and direction
and in fact presupposing—a deictic center. Radial patterns only to some extent, namely if transitivity is suspended
in space are far from unusual: They are characteristic of (Galton, 2011). And the radial concept entirely precludes
the relative FoR, with assignment of FRONT either nearer to the existence of a single mental time line, as its half-axes
Ego (reflection) or further away from Ego (translation), radiate out from the deictic center, thus pointing in dia-
and even some absolute FoRs are based on radiality, espe- metrically opposed directions. Even those data, however,
cially those used on small islands (e.g., Bennardo, 2000, that do support the mental time line construct exhibit
2009). remarkable variation, and they do so with regard to (i)
But how likely is a radial concept to emerge in the the number of available different time lines, (ii) their
domain of time? Several pieces of the currently available dimension and direction, and (iii) their anchoring.
data hint at such a possibility, each of which may appear
weak in isolation but gain weight when considered jointly. 7.2.1. Number of time lines
Acknowledgement of its feasibility goes back at least to Apparently, most people do not possess exactly one
Traugott (1975, 1978), who defined tense in its simplest time line. Some, like the Yucatec Maya or Amondawa, are
form as the distinction between proximal and distal rela- claimed to possess none at all (Le Guen & Pool Balam,
tions and thus as symmetrical to the deictic center. The 2012; Sinha et al., 2011), while others appear to have at
distinction between past and future requires an additional least two (Miles et al., 2011) or even three, if one takes
step—which is not taken by all languages—namely map- the data from Mandarin speakers (compiled in Table 9)
ping the tense relations onto a time line. Conflation of dis- as evidence for alternative recruitment of the sagittal, lat-
tant past and future has also been described for Toba eral, and vertical axis.
(Klein, 1987) and Yucatec Maya (Le Guen & Pool Balam,
2012). When confronted with Wednesday’s meeting task, 7.2.2. Dimension and direction of the time line
the majority of (older) speakers of Tongan tend to read Variability is also profound with regard to the preferred
‘‘moved forward’’ as pastwards in the case of a past meet- axis and direction of the time line. One of the first investi-
ing, and as futurewards in the case of a future meeting gations into space–time mappings found an LR preference
374 A. Bender, S. Beller / Cognition 132 (2014) 342–382

in English, but the reverse in Arabic and Hebrew speakers A related, yet broader distinction based on grounding
(Tversky et al., 1991). Ever since then it has been assumed principles is proposed by Núñez and colleagues (2012).
by psychologists—and recurrently demonstrated for a num- They classify a time line as egocentric if it is grounded in
ber of languages—that the mental time line basically the asymmetries of the conceptualizing Ego (such as
unfolds along the lateral axis (e.g., Ouellet et al., 2010a, back/front, up/down, or left/right), whereas an allocentric
2010b; Santiago et al., 2007, 2010; Ulrich & Maienborn, time line is grounded in the superordinate field.
2010; Weger & Pratt, 2008). This assumption contrasted Interestingly, only very few data patterns unequivocally
sharply with the linguistic emphasis of the sagittal axis attest to a time line that extends literally through the
(again across a range of languages), supported by data on speaker: linguistic mapping on front/back vocabulary (e.g.,
bodily expressions (Miles et al., 2010) and from congruency McGlone & Harding, 1998), some gestures that explicate
priming tasks that permitted sagittal responses (e.g., backward as over the head or shoulder (Brown, 2012;
Torralbo et al., 2006; Ulrich et al., 2012). Evidence for the Casasanto & Jasmin, 2012; Núñez & Sweetser, 2006), and
third dimension along the vertical axis comes from speak- forward/backward postural sway (Miles et al., 2010). As in
ers of Chinese languages (Tables 9 and 10). most of these cases representations were linked to partici-
Interestingly, the direction of time lines is not conven- pants’ subjective past or future, the egocentric nature of
tionalized to the same extent for all three dimensions. The these representations is beyond question. The only other
direction of the lateral axis appears to be non-controversial, type of representations for which this is the case is that
as it basically corresponds to the prevailing writing direc- based on the radial concept of time, which is egocentric by
tion (Fuhrman & Boroditsky, 2010; Ouellet et al., 2010b; definition.
Tversky et al., 1991). The vertical axis appears to be pointing On the other end of the spectrum, Yupno and Kuuk Tha-
uniformly downwards (UD). The two instances of upwards ayorre (as well as, to some extent, Yélî Dnye, Tzeltal and
direction (DU) reported in Table 9 (i.e., Tzeltal and Yucatec Aymara) provide examples for temporal construals that
Maya) are convincingly explained as representations of are undoubtedly allocentric. Regardless of how the speaker
temporal transience rather than spatial orientation is oriented, his or her time line is oriented towards a dis-
(Brown, 2012; Le Guen & Pool Balam, 2012). Only the sagit- tinct, external coordinate such as East/West or down-
tal axis may be oriented both ways (FB or BF), within and stream/upstream (Boroditsky & Gaby, 2010; Núñez et al.,
across speech communities, as attested to in linguistic tasks 2012; and see Fedden & Boroditsky, 2012). It happens to
(e.g., Clark, 1973; McGlone & Harding, 1998; Rothe-Wulf, pass through Ego, but is not emanating from it.
Beller, & Bender, 2014) and co-speech gesture (Casasanto For the bulk of data, however, this classification is not
& Jasmin, 2012)—likely a reflection of the two complemen- easy to make. As time is one-dimensional, thus comprising
tary perspectives on time (ME and MT). the speaker in each stream of events, even time lines that
In at least four of the cases compiled in Table 9, how- are not egocentric likely appear to be so. Most studies com-
ever, none of these dimensions appears to be relevant for piled in this review take the participant as point of origin.
the time line. Instead, speakers recruit geographical coor- It may thus be a mere artifact that time line representa-
dinates for its construals: the inclination of the landscape, tions appear to be egocentric at all. The question of how
with uphill/downhill contrasts as in Tzeltal Maya (Brown, accurate it is to classify a mental time line as egocen-
2012), the flowing direction of a prominent river as in tric—simply because it comprises the speaker—is related
Yupno (Núñez et al., 2012) and other Papuan languages to the question discussed above (in Sections 2.2.2 and
(Fedden & Boroditsky, 2012), or the course of the sun from 4.1) of whether the ME and MT perspective are indeed nec-
East to West as in Thaayorre (Boroditsky & Gaby, 2010; essarily ego-centered or simply happen to include Ego, and
Gaby, 2012), Yélî Dnye (Levinson & Majid, 2013), and per- thus has a direct bearing on FoR assignment as will be
haps Aymara (Núñez & Cornejo, 2012; Núñez & Sweetser, detailed in the next section.
2006). This raises the question of where the time line is
anchored, or grounded. 7.3. Temporal frames of reference

Section 2.2.3 introduced the notion of frames of refer-


7.2.3. Anchoring and grounding: Egocentric versus allocentric ence (FoRs) as one possibility for structuring spatio-tempo-
time lines ral representations, and Section 3 presented a range of
A time line may be classified as egocentric if it is anchored diverging theoretical accounts that are based on this
in Ego (i.e., it takes the speaker as the reference point or deic- notion. This current section is devoted to the question of
tic center), whereas an allocentric time line has its origin in how the empirical data may inform the evaluation of these
an entity distinct from Ego. In this sense, postural sway dur- accounts, by discussing (i) what the observed concepts of
ing a mental time travel into one’s own past or future clearly time and the culture-specific mental time lines reveal
reveals an egocentric time line (Miles et al., 2010). The about the principles of FoR construal and assignment of
speech balloon task with a person’s head as the reference orientation, (ii) how they speak to the psychological reality
point, in contrast, reveals an allocentric time line (Torralbo of the proposed temporal FoRs, and (iii) whether they sup-
et al., 2006). In a similar fashion, some of the spatial layout port a fundamental role for deixis.
tasks provide a reference point outside the speaker
(Boroditsky et al., 2008; Tversky et al., 1991), and so do some 7.3.1. Principles of FoR construal and FRONT assignment
of the congruency priming tasks (Miles et al., 2011; Torralbo As detailed above, the three basic FoRs—absolute, intrin-
et al., 2006; Ulrich et al., 2012). sic, and relative—are construed (and can be distinguished)
A. Bender, S. Beller / Cognition 132 (2014) 342–382 375

largely on the basis of the following principles: The abso- cases compiled in Table 9 are obvious candidates for this:
lute and intrinsic FoR describe binary relations (between the mapping of past events downhill (or rather down-
figure F and ground G), the relative FoR a ternary relation stream) and future events uphill (upstream) in Yupno
(between F, G, and observer’s viewpoint V); and orientation (Núñez et al., 2012) and, to some extent, Tzeltal (Brown,
is derived in the case of the absolute FoR from the superor- 2012), and the mapping of past events to East and future
dinate field (outside F and G), in the case of the intrinsic FoR events to West in Kuuk Thaayorre (Boroditsky & Gaby,
from G, and in the case of the relative FoR from V. As a con- 2010; Gaby, 2012) and perhaps Aymara (Núñez &
sequence, the relative FoR tends to be egocentric, and the Cornejo, 2012; Núñez & Sweetser, 2006).
other two tend to be allocentric. However, even speakers who do not invoke geographi-
The possible concepts of time relate to these principles cal coordinates may adopt an absolute FoR, for instance
in two ways. First, both the linear and the cyclical concept when aligning their mental time line with the arrow of
are one-dimensional and unidirectional representations of time (Lakoff & Johnson, 1999; Yu, 2012). If time is assumed
time that, from a mathematical point of view, require only to ‘flow’ or ‘fly’ from the past into the future, and speakers
two points for precise definition. In other words, they are assumed to be oriented in line with this direction of
afford construal of binary relations and thus an absolute time, then assignment of FRONT to the future would count
or intrinsic FoR (and are therefore unlikely to be Ego-cen- as an instance of the absolute FoR (as proposed by
tered). The radial concept, on the other hand, not only Bender et al., 2005, 2010; Kranjec, 2006). Likewise,
introduces a deictic center, but also presupposes its con- arrangement of tokens along the lateral axis from left to
sideration for establishing the ternary relation between right would count as an instance of the absolute FoR if
three points (whether F is in front of G depends on where the rationale for the arrangement is that the past is located
exactly V is in relation to F and G). In this sense, only the left on the time axis, and the future towards the right (in
radial concept affords a relative FoR. Second, the concepts marked contrast to arrangements based on sequential,
of time also provide orientation for the construal of FoRs: anterior/posterior relations, according to which earlier
Whereas the radial concept defines Ego as the deictic cen- times should be mapped to the left side). In this sense,
ter, thereby establishing the subjective perspective almost all instances of mental time lines would, in fact,
required for a relative FoR, the direction assigned to the qualify as representations of the same temporal orienta-
linear (and, to some extent, the cyclical) concept reflects tion, with almost all of the cases compiled in Table 9 being
the conceptualization of where time is generally flowing then classified as instantiations of an absolute FoR.
to, thus bestowing an orientation upon the superordinate A third possibility put forward by Zinken (2010) is to
field required for an absolute FoR (this argument corre- consider the superordinate field as being oriented towards
sponds to the proposition above that neither the ME nor the beginning of temporal intervals and, ultimately, the
MT perspective presuppose a deictic center, but reflect beginning of time itself. Adoption of this view appears to
more general perspectives on time, and thus do not qualify be reflected in FB gestures (as reported by Casasanto &
as instances of a relative FoR). Jasmin, 2012) and linguistic mapping along the FB axis
Accordingly, orientation of the relative FoR is easy to (as obtained with the Wednesday’s meeting task).
establish, as it simply derives from the deictic center of The intrinsic FoR, finally, is anchored in the ground entity
the radial concept (typically the speaker). In line with the G and can only be adopted if G is regarded as being oriented.
arguments put forward for the existence of a radial concept Crucially, a potential observer may also serve as ground,
of time (Section 7.1.2), the following cases could be which requires the distinction of two cases: anchoring of
enlisted as examples of a relative FoR: languages with the FoR in event G, and anchoring of the FoR in Ego (= G).
proximal/distal distinction only (Traugott, 1975, 1978), Unfortunately, no consensus has been reached as to
the conflation of (distant) past and future in Toba and whether temporal entities may or may not be regarded as
Yucatec Maya (Klein, 1987; Le Guen & Pool Balam, 2012), oriented. This controversy is most explicit for events, but
indications of an anthropocentric perspective on time in can also be generalized to the apparently simpler case of
English and in French generational terms (Radden, 2004), Ego = G. As detailed in Section 4.2.2, some scholars (e.g.,
and the symmetrical forward movement of dates away Tenbrink, 2011; Zinken, 2010) assume that the intrinsic
from now in Tongan (Bender et al., 2010). All of these cases FRONT in this case is derived from Ego’s looking direction,
attest to the possibility of relative FoRs in the domain of which in most cases is futurewards. However, as we argued
time—contrary to claims by Moore (2011) and Tenbrink above (Section 4.4), what would serve as G in temporal
(2011), but in line with the accounts of Kranjec (2006) contexts is not Ego as a person, but his or her subjective
and Zinken (2010). These cases also afford, at least in prin- present, and whether this specific moment in time can be
ciple, the two variants reflection and translation as pro- assigned an intrinsic FRONT—and, if so, towards which direc-
posed by Bender and colleagues (2005, 2010, 2012): The tion—depends on the more general question of whether
French generational terms both for great-grandchildren temporal entities may have an intrinsic FRONT in the first
and great-grandparents as ‘‘behind’’ (Radden, 2004) place. Whereas Tenbrink (2011) claims this to be impossi-
instantiate the reflection variant, and the Tongan moving ble, Zinken (2010) is prepared to assign an intrinsic FRONT
pattern away from Ego the translation variant. at least to temporal intervals, and Bender and colleagues
Establishing orientation of the absolute and intrinsic (2005, 2010) as well as Yu (2012) assign an intrinsic FRONT
FoR is less straightforward, and more controversial. (= their ‘‘beginning’’) to all temporal entities. Combinations
In order to be diagnosed as absolute, external coordi- of a linear concept with assignment of FRONT to earlier
nates need to be recruited for FoR construal. Some of the times—reflected in all cases recruiting the FB axis—thus
376 A. Bender, S. Beller / Cognition 132 (2014) 342–382

appear to be a likely candidate for the diagnosis of an intrin- movements as an indication of an absolute FoR, and past-
sic FoR. ward movements as an indication of an intrinsic FoR (i.e.,
In any case, the problem remains that different Bender et al., 2010; Kranjec, 2006; Núñez et al., 2006; see
accounts classify the same patterns as instances of diverg- Table 7). This classification also entails that, of the time
ing FoRs. The classification of absolute and intrinsic FoRs lines compiled in Table 9, those unfolding in LR, BF, and
by Tenbrink (2011) and see Moore, 2011, for instance, is UD direction are likely to favor the absolute FoR.
diametrically opposed to that by Kranjec (2006) or
Bender and colleagues (2005, 2010). However, whether
time lines oriented pastwards really reflect adoption of 7.3.3. The role of deixis
an absolute FoR and time lines oriented futurewards reflect The preponderance of the absolute FoR for temporal
adoption of an intrinsic FoR cannot be decided on theoret- representations across tasks is at odds with the claim that
ical grounds, as different speech communities may differ in deictic representations differ fundamentally from non-
the principles on which they base their assignment of ori- deictic representations. As described above, almost all
entation. Fortunately, some priming studies explicitly accounts of temporal FoRs take deixis as their point of
aimed to test the psychological reality of these FoRs and departure. They construe their taxonomies from the basic
may thus shed light on this infelicitous situation. distinction between expressions that do (A-series) or do
not (B-series) entail Ego as a deictic center. Quite fre-
7.3.2. The psychological reality of temporal FoRs quently, deictic A-series expressions are equated with a
As mentioned earlier, the Wednesday’s meeting ques- relative temporal FoR, and with the ME perspective—as
tion was primarily used to explore the psychological real- this appears to also presuppose a deictic center, the (mov-
ity of specific perspectives that people are claimed to ing) Ego—and, in some accounts, an Ego-centered MT per-
take while processing ambiguous temporal phrases. The spective. Conversely, non-deictic B-series expressions are
perspective under scrutiny is typically primed before par- equated with an Ego-free MT perspective, and either with
ticipants are asked to answer the target question. Of the an intrinsic or absolute temporal FoR. As argued in Sec-
accounts described in Section 3, only three attempted to tion 4.1, however, this focus on deixis may obscure the
empirically test the space–time mapping they claim and more complex relations between the A-/B-series classifica-
the psychological reality of any additional perspective or tion and a FoR-based classification, and may incite hasty
FoR predicted by them (details in Section 5.1.3). Two of mappings. The crucial role for the unfolding of the time
them are relevant here. line and for preferences in temporal references that has
The reference-point (RP) metaphors account (Núñez been ascribed to deixis is challenged both on theoretical
et al., 2006; see Section 3.2) predicts an Ego-free MT [= and empirical grounds.
intrinsic] perspective (indicative of the ‘‘Time-RP’’ meta- Theoretically, deixis reveals only little about the
phor) in contradistinction to the ‘‘Ego-RP’’ metaphor, adopted FoR. Whereas adoption of a relative FoR does pre-
which is regarded as consisting of the ME and an Ego-cen- suppose a subjective point of view (typically, although not
tered MT perspective. To enforce such an Ego-free MT per- necessarily, the speaker), adoption of a non-relative FoR
spective, visual primes were used that foregrounded the does not preclude the presence of the speaker. In other
anterior/posterior relation between two events. Compared words, descriptions can be non-relative, regardless of
to a control condition, this priming significantly shifted whether Ego is or is not part of the scene. The non-relative
participants’ responses to the earlier date (52% [n = 50] vs. relationship between Monday and Tuesday (as depicted in
71% [n = 45 speakers of US English] in Exp. 2; in Exp. 1 Fig. 1 and Table 4), holds regardless of whether one is
without control condition, Monday was chosen by 64%, aware of or oblivious to the fact that Monday coincides
n = 66 speakers of US English). with Ego’s today. Especially in one-dimensional ‘spaces’
The temporal framework models account (Kranjec, 2006; like time, Ego’s alignment with a sequence of events is dif-
see Section 3.3) predicts an additional ‘‘extrinsic’’ [= abso- ficult to elude. This is nicely illustrated by co-speech ges-
lute] FoR. To enforce this FoR, visual primes were used that tures, produced by speakers of Yupno, which appear to
foregrounded the superordinate field: In two experiments, be speaker-centered and yet depict a general, non-relative
the priming shifted responses to the later date (Friday: 73%, perspective on time as moving upstream into the future
aggregated across experiments, combined n = 44 speakers (Núñez et al., 2012). What counts in terms of the classifica-
of US English). In contrast, priming them with intrinsic tion of a FoR as relative is not whether it reveals the speak-
stimuli (similar to the ones by Núñez et al., 2006) reversed er’s position, or passes through it, but whether it is
the pattern (Monday: 61%, n.s.; n = 36 speakers of US anchored in and emanating from this position.
English). As a consequence, both the ME and MT perspective on
These findings provide empirical support for two time may also afford a more general and inherently non-
important propositions: First, they do attest to effects of deictic view on temporal moving: MT reflects a perspective
spatial priming on temporal reasoning and thus to space– on time as moving not only towards Ego, but towards tem-
time mapping in the context of references. Second, and poral entities and events more generally (in most cases past-
more importantly, they demonstrate that an absolute (spa- wards); and ME reflects a perspective according to which
tial) FoR leads to assigning FRONT futurewards, and an not only Ego, but temporal entities and events more gener-
intrinsic FoR leads to assigning FRONT pastwards (in the ally are moving through time (in most cases futurewards).
direction of earlier events). In other words: The empirical For the time being, this account of the temporal perspectives
data support those accounts that classify futureward is afforded by the empirical data on temporal references and
A. Bender, S. Beller / Cognition 132 (2014) 342–382 377

space–time mappings, but whether it will stand up to more 7.4.1. Sources for space–time mapping
conclusive scrutiny remains an open question. At least four possible sources for space–time mapping
Empirical data that could illuminate the role of deixis have been discussed in the literature: cultural artifacts
are scarce and ambiguous. Some studies seem to confirm such as writing direction, linguistic metaphors, embodied
that mental representations may indeed differ substan- representations of motion, and principles of FoR construal.
tially depending on whether they refer to deictic or non- Initial research on how spatial representations affect
deictic time. Specifically in co-speech gestures, people tend temporal ones unearthed the importance of cultural arti-
to recruit the sagittal axis for deictic, and the lateral axis facts: People’s mental time line turned out to be correlated
for non-deictic expressions (Casasanto & Jasmin, 2012; with the prevailing writing direction (Tversky et al., 1991).
Emmorey, 2002; Núñez et al., 2012). However, the story Subsequent studies confirmed this impact of writing direc-
must be more complicated, as the very same (deictic) stim- tion (Bergen & Chan Lau, 2012; Casasanto & Bottini, 2010;
uli may trigger diverging axis preferences, depending on Chen & O’Seaghdha, 2013; de Sousa, 2012; Fuhrman &
other context factors (Torralbo et al., 2006). And con- Boroditsky, 2010; Ouellet et al., 2010b), and hinted at
versely, both types of stimuli (non-deictic as in the card related factors such as musical notation, the time axis of
arranging task vs. deictic as in the time-points task, Sec- graphs, or the design of calendars and clocks10 (Galton,
tion 5.3.4) may yield the same type of results (e.g., 2011; Jamalian & Tversky, 2012).
Boroditsky & Gaby, 2010). It is also worth noting that, That language itself may serve as a principal source for
despite the preference for the lateral axis to express space–time mappings is as obvious as it is wrong, at least
sequence time, this does not involve FRONT-to-past mapping as a general rule. Adoption of the sagittal axis BF is in line
(as in ‘‘earlier times’’ before ‘‘later times’’), but the opposite with respective linguistic metaphors that put the past
direction (with FRONT in the future direction). In other behind the speaker and the future ahead. Likewise, an
words, both sagittal and lateral axis reveal an absolute FoR. additional vertical time line in speakers of Mandarin may
A more comprehensive explanation for the observed (also) be derived from vertical expressions and may be in
variability across contexts may thus be that speakers of place even in the absence of vertical writing (Fuhrman
most languages have more than one axis readily available et al., 2011; Miles et al., 2011). And speakers of Tzeltal
for temporal representation and reference, and simply pre- Maya occasionally point to the future as uphill in a way
fer the one that is most compatible with their mindset, akin to how they may refer to it verbally (Brown, 2012).
with task specifics, and with contextual factors, which And yet, most non-linguistic data on temporal representa-
may—but need not—involve consideration of deictic versus tions, at least in the Western world, indicate a lateral axis
non-deictic aspects. LR for which there is no linguistic equivalent (Casasanto
& Jasmin, 2012; Cooperrider & Núñez, 2009). Likewise,
speakers of Kuuk Thaayorre do not speak of the future as
7.4. Extent of space–time mapping ‘‘westwards’’ (Gaby, 2012), and speakers of Yupno do not
really speak of it as ‘‘upstream’’ (Núñez et al., 2012).
If our analysis of FoR construals is correct, then one sur- An alternative explanation of this preference for the
prising inference is that the majority of cases compiled in sagittal axis focuses on embodied representations of
Table 9 favor adoption of an absolute FoR. Exceptions motion and/or vision. When walking, places to be reached
include a few cases of refusal to spatialize time at all (i.e., in the future are lying ahead, and previously visited ones
Amondawa and Yucatec Maya), a few cases of adopting are behind, thus providing strong associations of future
the radial concept and concomitantly a genuinely relative with FRONT, and past with BACK (Casasanto & Jasmin, 2012;
FoR (like Toba and, to some extent, perhaps Yucatec Maya Núñez & Sweetser, 2006; and see Miles et al., 2010). Nota-
and older speakers of Tongan), and a few cases of adopting bly, this imagined motion may also give rise to the
the intrinsic FoR (as indicated by the occasional occurrence reversed pattern (of mapping FRONT to the past). And in
of the FB time line). Based on this classification, even the Aymara, the (known) past is indeed regarded as in front
apparently inconsistent responses of Tzeltal Maya and the unknown as in back of Ego (Núñez & Sweetser,
(Brown, 2012) and Yélî Dnye (Levinson & Majid, 2013) 2006; for claims of similar FRONT-to-past mappings in Mal-
would be considered coherently absolute (except for the agasy, Toba, and Maori, see Dahl, 1995; Klein, 1987;
piling-up strategy): LR and NF reflect a lateral and sagittal Thornton, 1987, respectively).
line, respectively, aligned to the flow of time, while East/ A fourth possible source for space–time mappings are
West and downhill/uphill recruit external coordinates, the principles of FoR construals themselves (Bender et al.,
but for the same purpose. Instead of inconsistency in pref-
erences for temporal FoRs, the data thus reveal inconsis-
10
tency in how these temporal FoRs are represented in If cyclical representations are simply triggered by a cyclical concept of
time, there is no reason to assume a specific direction (clockwise or
space—which may not be as surprising, given the lack of
counterclockwise); if, however, they are (also) affected by the canonical
cultural conventions for space–time mappings. direction of clocks, they should be oriented clockwise only. Unfortunately,
Despite converging on the absolute FoR, however, a not all cyclical concepts found in previous research were systematically
great deal of diversity is discerned in how these absolute analyzed (e.g., Tversky et al., 1991). The available data, however, support
FoRs are instantiated, and this raises two important ques- the assumption: Cyclical representations produced by Chinese participants
are oriented exclusively clockwise (Bergen & Chan Lau, 2012), whereas
tions: What motivates the specifics of each of these FoRs, those produced by (non-literate) Yucatec Maya exhibit an almost equal
and to what extent do preferences for a specific FoR in number of clockwise and counterclockwise directions (Le Guen & Pool
the domain of time reflect preferences in space? Balam, 2012).
378 A. Bender, S. Beller / Cognition 132 (2014) 342–382

2010), which appear to highlight the exact same external flexibility of temporal representations depending on task
coordinates for the absolute references among speakers design, material, and other context factors—together with
of Kuuk Thaayorre (Boroditsky & Gaby, 2010; Gaby, a similar flexibility of FoR adoption in spatial tasks—forbids
2012), Yupno (Núñez et al., 2012), or—albeit to a lesser simple comparisons across diverse tasks.
extent—Tzeltal Maya (Brown, 2012). An explicit empirical assessment of cross-domain con-
Alternative availability of these different sources for sistency in FoR adoption has only been attempted by a
space–time mappings may be what allows people to adopt small number of studies to date. The authors of the t-FoR
more than one such representation in parallel (e.g., Miles account (Bender et al., 2005, 2010; see Section 3.6), which
et al., 2011; Torralbo et al., 2006). It also explains why distinguishes four different FoRs, assumed that—specifi-
the direction of time lines is not conventionalized to the cally in the case of the two relative FoR variants—prefer-
same extent for all three dimensions, with the lateral and ences should carry over from space to time. Comparing
vertical axis being significantly more uniform than the sag- speakers of US English, German, Mandarin, and Tongan in
ittal axis (which may be oriented both ways, FB or BF, a language elicitation study with the Wednesday’s meeting
within and across speech communities). task, they found a substantial proportion of relative FoR
Importantly, most of these sources favor an absolute adoption in only one of the languages under scrutiny (i.e.,
FoR: The mental time line fostered by the prevailing writ- among older speakers of Tongan), and the variant used in
ing direction, a substantial number of linguistic metaphors, time (i.e., translation) was identical to the one preferred
and people’s looking direction and imagined motion all in space (Bennardo, 2000). For the other three languages,
more or less reflect (or are aligned to) the general flow of the observed temporal FoRs were not incompatible with
time, thus foregrounding the orientation of the superordi- those available for space, but did not reflect the preferred
nate field (Table 11). The only source that does not seem to ones either (Bender et al., 2010). A follow-up study
favor one particular FoR over any other are the principles attempted to bring the spatial condition more in line with
of FoR construal. These would predict that, whatever FoR the temporal one by replacing the static task with a task on
people prefer in one domain should also be preferred in object movements. Still, responses from speakers of the
the other (Bender et al., 2010). Is this prediction confirmed same four languages did not indicate cross-domain consis-
by the data? tency in FoRs. Notably, correlation between spatial and
temporal FoRs among the US participants, who had worked
on both tasks sequentially, was close to zero (Bender et al.,
7.4.2. Degree of space–time mapping
2012). And finally, Rothe-Wulf and colleagues (2014; see
Based on the assumption that (unidirectional) mental
Section 5.1.3) tested with speakers of US English, Swedish,
time lines attest to an absolute FoR, we re-coded the data
and German whether adoption of a temporal FoR could be
from Table 9, and contrasted them with preferences for
primed by spatial FoRs. They found that only a minor pro-
spatial FoRs where available. Results are summarized in
portion of speakers of only one of the three languages
Table 11.
under scrutiny seemed to be susceptible to this priming
The mere fact that absolute FoRs appear to prevail in the
in the predicted direction.
languages under scrutiny already hints at the conclusion
These findings are compatible with the observation that
that FoR preferences may not carry over from space to time.
speakers do not necessarily base their mapping on the
In only very few of these cases (including Yupno, Yélî Dnye,
same principles: While speakers of Kuuk Thaayorre or
Thaayorre, and to some extent the two Maya groups) do
Yupno appear to take principles of FoR construal as a ratio-
speech communities favor the absolute FoR in small-scale
nale for mapping an absolute FoR from space onto time,
space; more often, they prefer the relative and/or intrinsic
this is obviously not the case with speakers of languages
FoR. Given the somewhat tentative classification of tempo-
such as English or German, who prefer a relative FoR in
ral representations as instantiations of one or another FoR,
space, but the absolute FoR in time.
however, this inference may be too hasty. Furthermore, the

Table 11
Evidence for preferred FoRs in space (s-FoR) and time (t-FoR), based on findings as reported in Section 6 and Table 9 (for classification of patterns into FoRs, see
previous subsection).

Note. The following abbreviations are used: A = absolute, I = intrinsic, and R = relative FoR; subscript letters indicate the type of representation, for which
this FoR is preferred (G = gesture, L = linguistic, nd = non-deictic, d = deictic). Preferred FoRs are printed bold-faced; FoRs in brackets are possible, but rarely
used; entries in gray indicate likely, but unconfirmed preferences; shaded cells indicate lacking data.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 379

8. Conclusion be achieved—not only across speech communities, but also


within them, and even within individual speakers. The
The conceptual relations of space and time are manifold actual shape of the time line (if one is adopted at all) seems
and complex. Time and space share four properties to vary- to depend on contextual factors, task specifics, and mind-
ing extents. Some of the speech communities included in set. This flexibility is less surprising, if one considers that
this review are reported to emphasize the least shared time is not only one-dimensional and unidirectional, but
property transience and thus to avoid space–time also orthogonal to all spatial dimensions. Which spatial
mappings entirely. Most others, however, use spatialized dimension is actually adopted to convey the flow of time
representations of time that exhibit the shared properties therefore does not really matter, as long as its basic prop-
extension, linearity, and direction. These representations erties are preserved. The fact that this may request compli-
reflect three different spatial concepts of time (most fre- ance to cultural conventions, in order to be understood in
quently the linear concept, but occasionally also the cycli- the intended sense, explains why explicit tendencies are
cal or radial concept), which favors the selection of a found in some speech communities and for some dimen-
specific temporal FoR by defining a coordinate system, its sions, but not others.
anchoring, and its orientation. The preponderance of the absolute FoR in the temporal
Unfortunately, theoretical accounts on temporal FoRs domain (at least in the data compiled here) contrasts shar-
diverge broadly in how they conceptualize and distinguish ply with a more diverse pattern in the spatial domain. Only
temporal FoRs—depending, for instance, on what may a small number of studies found cross-domain consistency
count as the superordinate field, whether events may have in FoR adoption, but this is typically the case in those
an intrinsic FRONT, or how a deictic center may affect refer- speech communities with an almost exclusive preference
encing patterns. Of the thirty studies scrutinized for con- for the absolute spatial FoR. Does this devalue our efforts
firming or disconfirming evidence, only very few address to systematize the mapping of spatial frames of reference
this issue directly. They suggest that the absolute FoR does onto time, which motivated this review? We are convinced
derive its orientation from concepts such as the arrow of that, on the contrary, the conclusions of our review render
time, to which the mental time line is aligned in almost such an endeavor even more essential. Not only do infor-
all cases. The intrinsic FoR, in contrast, is anchored in a mative cross-domain comparisons presuppose a consistent
temporal interval or event serving as ground that derives and comprehensive mapping of FoR taxonomies; without
its orientation from assignment of FRONT to the beginning consensus on what exactly may count as a frame of refer-
of temporal intervals and events or from the anteriority/ ence, according to which principles they should be defined
posteriority relation inherent in their sequence. While and construed, and how respective data should be inter-
these two FoRs are fostered by the linear or cyclical con- preted, this field of research will not advance, and will be
cept, the relative FoR is compatible with a radial concept unable to contribute to fundamental theoretical questions
only: It presupposes a subjective viewpoint (or deictic cen- on temporal cognition.
ter) and pertains to ternary relations.
Based on this classification, the empirical data, with few
exceptions, suggest a preponderance of the absolute FoR in Acknowledgements
the temporal domain, but also attest to an amazing variety
in how this FoR is instantiated. To assign orientation to the The writing of this article took place during our stay at
superordinate field, some speech communities recruit the Center for Interdisciplinary Research (ZiF) at the Uni-
external (allocentric) coordinates such as the position of versity of Bielefeld and was supported by a Heisenberg Fel-
the sun or the gradient of landscape or rivers, while others lowship from the Deutsche Forschungsgemeinschaft DFG
use the arrow of time in a more abstract manner. This is (Be 2451/8-1,2) to Andrea Bender and by a grant from the
reflected in a variety of dimensions and directions, along Deutsche Forschungsgemeinschaft DFG for the project on
which the mental time line unfolds: sagittally BF or FB ‘‘Spatial referencing across languages: Cultural preferences
(in line with the direction of looking and moving), laterally and cognitive implications’’ to Andrea Bender and Sieghard
LR or RL (depending on writing direction), vertically UD Beller (Be 2451/13-1, Be 2178/7-1). We are indebted to Sus-
(triggered variably by writing direction and/or perhaps lin- anne Bubser for support with the literature, to Gunter Senft
guistic metaphors), and even cyclically in either of two for valuable comments on a previous version of the paper,
directions (depending on familiarity with clocks). In con- and to Sarah Mannion de Hernandez for proofreading. This
trast to the FoRs based on geographical coordinates, these paper is dedicated to Melissa Bowerman, in appreciation of
time lines might appear to be egocentric as they comprise her scientific support and personal friendship.
the speaker and are construed relative to his or her posi-
tion. As argued above, however, even these time lines
should be considered allocentric in the sense that their ori- References
entation is derived from the direction into which time Alverson, H. (1994). Semantics and experience: Universal metaphors of time
‘flows’ rather than from a subjective perspective of the in English, Mandarin, Hindi, and Sesotho. Baltimore: John Hopkins
speaker. To assess the validity of this proposal, more sys- University.
Beller, S., Singmann, H., Hüther, L., & Bender, A. (2014). How relative is the
tematic and empirical research is imperative.
relative frame of reference? A cross-linguistic study on and an MPT
The diversity in time lines, which reveal the tendency to model of spatial references for frontal and dorsal configurations
spatialize time, reflects the flexibility with which this can (submitted for publication).
380 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Bender, A., & Beller, S. (2011). Cultural variation in numeration systems Chen, J.-Y. (2007). Do Chinese and English speakers think about time
and their mapping onto the mental number line. Journal of Cross- differently? Failure of replicating Boroditsky (2001). Cognition, 104,
Cultural Psychology, 42, 579–597. 427–436.
Bender, A., Beller, S., & Bennardo, G. (2010). Temporal frames of reference: Chen, J.-Y., & O’Seaghdha, P. G. (2013). Do Mandarin and English speakers
Conceptual analysis and empirical evidence from German, English, think about time differently? Review of existing evidence and some
Mandarin Chinese, and Tongan. Journal of Cognition and Culture, 10, new data. Journal of Chinese Linguistics, 41, 338–358.
283–307. Cienki, A. (1998). Metaphoric gestures and some of their relations to
Bender, A., Bennardo, G., & Beller, S. (2005). Spatial frames of reference for verbal metaphoric expressions. In J.-P. Koenig (Ed.), Discourse and
temporal relations: A conceptual analysis in English, German, and cognition: Bridging the gap (pp. 189–204). Stanford: CSLI Publications.
Tongan. In B. G. Bara, L. Barsalou, & M. Bucciarelli (Eds.), Proceedings of Clark, H. H. (1973). Space, time, semantics, and the child. In T. E. Moore
the 27th annual conference of the Cognitive Science Society (Ed.), Cognitive development and the acquisition of language
(pp. 220–225). Mahwah, NJ: Lawrence Erlbaum. (pp. 27–63). New York: Academic Press.
Bender, A., Rothe-Wulf, A., Hüther, L., & Beller, S. (2012). Moving forward Cooperrider, K., & Núñez, R. (2009). Across time, across the body:
in space and time: How strong is the conceptual link between spatial Transversal temporal gestures. Gesture, 9, 181–206.
and temporal frames of reference? Frontiers in Psychology: Cultural Dahl, Ø. (1995). When the future comes from behind: Malagasy and other
Psychology, 3. 486, 1–11. time concepts and some consequences for communication.
Bennardo, G. (2000). Language and space in Tonga: ‘‘The front of the International Journal for Intercultural Relations, 19, 197–209.
house is where the chief sits’’. Anthropological Linguistics, 42, Danziger, E. (2010). Deixis, gesture, and cognition in spatial frame of
499–544. reference typology. Studies in Language, 34, 167–185.
Bennardo, G. (Ed.). (2002). Representing space in Oceania. Canberra: Pacific Darwin, C. (1872). The expression of the emotions in man and animals.
Linguistics. Chicago: University of Chicago.
Bennardo, G. (2009). Language, space and social relationships: A de Sousa, H. (2012). Generational differences in the orientation of time in
foundational cultural model in Polynesia. Cambridge: Cambridge Cantonese speakers as a function of changes in the direction of
University Press. Chinese writing. Frontiers in Psychology: Cultural Psychology, 3, 255.
Bennett, D. C. (1975). Spatial and temporal uses of English prepositions. DeLong, A. (1981). Phenomenological space–time: Toward an experiential
London: Longman. relativity. Science, 213, 681–683.
Bergen, B. K., & Chan Lau, T. T. (2012). Writing direction affects how Emmorey, K. (2002). Language, cognition, and the brain. Mahwah, NJ:
people map space onto time. Frontiers in Psychology: Cultural Lawrence Erlbaum.
Psychology, 3, 109. Evans, V. (2003). The structure of time: Language, meaning and temporal
Bohnemeyer, J., & O’Meara, C. (2012). Vectors and frames of reference: cognition. Amsterdam: John Benjamins.
Evidence from Seri and Yucatec. In L. Filipovic & K. M. Jaszczolt (Eds.), Farriss, N. M. (1987). Remembering the future, anticipating the past:
Space and time across languages and cultures (pp. 217–249). History, time, and cosmology among the Maya of Yucatan.
Amsterdam: John Benjamins. Comparative Studies in Society and History, 29, 566–593.
Bohnemeyer, J., & Stolz, C. (2006). Spatial reference in Yukatek Maya: A Fedden, S., & Boroditsky, L. (2012). Spatialization of time in Mian. Frontiers
survey. In S. C. Levinson & D. P. Wilkins (Eds.), The grammar of space in Psychology: Cultural Psychology, 3, 485.
(pp. 273–310). Cambridge: Cambridge University Press. Fillmore, C. (1971). Lectures on deixis. Stanford, CA: CSLI Publications.
Boroditsky, L. (2000). Metaphoric structuring: Understanding time Friedmann, W. (1990). About time. Cambridge: MIT Press.
through spatial metaphors. Cognition, 75, 1–28. Fuhrman, O., & Boroditsky, L. (2010). Cross-cultural differences in mental
Boroditsky, L. (2001). Does language shape thought? English and representations of time: Evidence from an implicit nonlinguistic task.
Mandarin speakers’ conceptions of time. Cognitive Psychology, 43, Cognitive Science, 34, 1430–1451.
1–22. Fuhrman, O., McCormick, K., Chen, E., Jiang, H., Shu, D., Mao, S., et al.
Boroditsky, L. (2008). Do English and Mandarin speakers think differently (2011). How linguistic and cultural forces shape conceptions of time:
about time? In B. C. Love, K. McRae, & V. M. Sloutsky (Eds.), English and Mandarin time in 3D. Cognitive Science, 35, 1305–1328.
Proceedings of the 30th annual conference of the Cognitive Science Gaby, A. (2012). The Thaayorre think of time like they talk of space.
Society (pp. 427–431). Austin, TX: Cognitive Science Society. Frontiers in Psychology: Cultural Psychology, 3. 300, 1–8.
Boroditsky, L., Fuhrman, O., & McCormick, K. (2011). Do English and Galton, A. (2011). Time flies but space does not: Limits to the
Mandarin speakers think about time differently? Cognition, 118, spatialisation of time. Journal of Pragmatics, 43, 695–703.
123–129. Gell, A. (1992). The anthropology of time: Cultural constructions of temporal
Boroditsky, L., & Gaby, A. (2010). Remembrances of times East: Absolute maps and images. Oxford: Berg.
spatial representations of time in an Australian Aboriginal Gentner, D. (2001). Spatial metaphors in temporal reasoning. In M. Gattis
community. Psychological Science, 21, 1635–1639. (Ed.), Spatial schemas and abstract thought (pp. 203–222). Cambridge:
Boroditsky, L., Gaby, A., & Levinson, S. C. (2008). Time in space. In A. Majid MIT Press.
(Ed.). Field manual (Vol. 11, pp. 52–76). Nijmegen: Max Planck Gentner, D., Imai, M., & Boroditsky, L. (2002). As time goes by: Evidence
Institute for Psycholinguistics. for two systems in processing space time metaphors. Language and
Boroditsky, L., & Ramscar, M. (2002). The roles of mind and body in Cognitive Processes, 17, 537–565.
abstract thought. Psychological Science, 13, 185–188. Goldin-Meadow, S. (2003). Hearing gesture: How our hands help us think.
Brown, P. (2012). Time and space in Tzeltal: Is the future uphill? Frontiers Cambridge, MA: Harvard University Press.
in Psychology: Cultural Psychology, 3, 212. Gould, S. J. (1987). Time’s arrow, time’s cycle: Myth and metaphor in the
Brown, P., & Levinson, S. (1992). ‘‘Left’’ and ‘‘right’’ in Tenejapa: discovery of geological time. Cambridge: Harvard University Press.
Investigating a linguistic and conceptual gap. Zeitschrift für Phonetik, Grabowski, J., & Miller, G. A. (2000). Factors affecting the use of
Sprachwissenschaft und Kommunikationsforschung, 45, 590–611. dimensional prepositions in German and American English: Object
Calbris, G. (1990). The semiotics of French gestures. Bloomington, IN: orientation, social context, and prepositional pattern. Journal of
Indiana University Press. Psycholinguistic Research, 29, 517–553.
Casasanto, D., & Boroditsky, L. (2008). Time in the mind: Using space to Grabowski, J., & Weiß, P. (1996a). Determinanten der Interpretation
think about time. Cognition, 106, 579–593. dimensionaler Lokalisationsäußerungen: Experimente in fünf
Casasanto, D., & Bottini, R. (2010). Can mirror-reading reverse the flow of Sprachen. Sprache & Kognition, 15, 234–250.
time? In C. Hölscher, T. F. Shipley, M. Olivetti Belardinelli, J. A. Grabowski, J., & Weiß, P. (1996b). The prepositional inventory of
Bateman, & N. S. Newcombe (Eds.), Spatial cognition VII (pp. 335–345). languages: A factor that affects comprehension and spatial
Berlin, Heidelberg: Springer. prepositions. Language Sciences, 18, 19–35.
Casasanto, D., Fotakopoulou, O., & Boroditsky, L. (2010). Space and time in Haspelmath, M. (1997). From space to time: Temporal adverbials in the
the child’s mind: Evidence for a cross-dimensional asymmetry. world’s languages. Munich, Newcastle: Lincom Europa.
Cognitive Science, 34, 387–405. Haun, D. B. M., Rapold, C. J., Call, J., Janzen, G., & Levinson, S. C. (2006).
Casasanto, D., & Jasmin, K. (2012). The hands of time: Temporal gestures Cognitive cladistics and cultural override in Hominid spatial
in English speakers. Cognitive Linguistics, 23, 643–674. cognition. Proceedings of the National Academy of Sciences, 103,
Chan, T. T., & Bergen, B. (2005). Writing direction influences spatial 17568–17573.
cognition. In B. Bara, L. W. Barsalou, & M. Bucciarelli (Eds.), Haun, D. B. M., Rapold, C. J., Janzen, G., & Levinson, S. C. (2011). Plasticity
Proceedings of the 27th annual conference of the Cognitive Science of human spatial cognition: Spatial language and cognition covary
Society (pp. 412–417). Mahwah, NJ: Lawrence Erlbaum. across cultures. Cognition, 119, 70–80.
A. Bender, S. Beller / Cognition 132 (2014) 342–382 381

Haviland, J. B. (2003). How to point in Zinacantán. In S. Kita (Ed.), Pointing: Li, J., & Zhang, K. (2009). Regional differences in spatial frame of reference
Where language, culture, and cognition meet (pp. 139–170). Mahwah, systems for people in different areas of China. Perceptual and Motor
NJ: Lawrence Erlbaum. Skills, 108, 587–596.
Haviland, J. B. (2005). Directional precision in Zinacantec deictic gestures: Li, P., Abarbanell, L., Papafragou, A., & Gleitman, L. (2011). Spatial
(Cognitive?) preconditions of talk about space. Intellectica, 2–3(41– reasoning in Tenejapan Mayans. Cognition, 120, 33–53.
42), 25–54. Li, P., & Gleitman, L. (2002). Turning the tables: Language and spatial
Hellberg, S. (2007). The polysemy across image schemas: Swedish FRAM⁄. reasoning. Cognition, 83, 265–294.
Studia Linguistica, 61, 20–58. Mackey, M. C. (2003). Time’s arrow: The origins of thermodynamic behavior.
Herrmann, T., & Grabowski, J. (1998). The dimensional conception of Mineola, NY: Dover.
space and the use of dimensional prepositions in different languages. Majid, A., Bowerman, M., Kita, S., Haun, D. B. M., & Levinson, S. C. (2004).
In D. Hillert (Ed.), Sentence processing: A cross-linguistic perspective Can language restructure cognition? The case for space. Trends in
(pp. 265–291). San Diego: Academic Press. Cognitive Sciences, 8, 108–114.
Herrmann, T., & Schweizer, K. (1998). Sprechen über Raum: Sprachliches Malotki, E. (1983). Hopi time: A linguistic analysis of the temporal concepts
Lokalisieren und seine kognitiven Grundlagen. Bern: Verlag Hans Huber. in the Hopi language. Berlin: Mouton.
Hill, C. A. (1978). Linguistic representation of spatial and temporal McGlone, M. S., & Harding, J. L. (1998). Back (or forward?) to the future:
orientation. Proceedings of the 4th Annual Meeting of the Berkeley The role of perspective in temporal language comprehension. Journal
Linguistics Society, 4, 524–538. of Experimental Psychology: Learning, Memory, and Cognition, 24,
Hill, C. A. (1982). Up/down, front/back, left/right. A contrastive study of 1211–1223.
Hausa and English. In J. Weissenborn & W. Klein (Eds.), Here and there: McNeill, D. (1992). Hand and mind. Chicago: University of Chicago Press.
Cross-linguistic studies on deixis and demonstration (pp. 13–42). McTaggart, J. M. E. (1908). The unreality of time. Mind: A Quarterly Review
Amsterdam: Benjamins. of Psychology and Philosophy, 17, 456–473.
Hüther, L., Bentz, A., Spada, H., Bender, A., & Beller, S. (2013). Influences Meeren, H. K. M., van Heijnsbergen, C. C. R. J., & de Gelder, B. (2005). Rapid
beyond language? A comparison of spatial referencing in native perceptual integration of facial expression and emotional body
French speakers from four countries. In M. Knauff, M. Pauen, N. language. Proceedings of the National Academy of Sciences, 102,
Sebanz, & I. Wachsmuth (Eds.), Proceedings of the 35th annual 16518–16523.
conference of the Cognitive Science Society (pp. 2602–2607). Austin, Merritt, D. J., Casasanto, D., & Brannon, E. M. (2010). Do monkeys think in
TX: Cognitive Science Society. metaphors? Representations of space and time in monkeys and
Iverson, J. M., & Goldin-Meadow, S. (1998). Why people gesture when humans. Cognition, 117, 191–202.
they speak. Nature, 396, 228. Miles, L. K., Nind, L. K., & Macrae, C. N. (2010). Moving through time.
Jamalian, A., & Tversky, B. (2012). Gestures alter thinking about time. In N. Psychological Science, 21, 222–223.
Miyake, D. Peebles, & R. P. Cooper (Eds.), Proceedings of the 34th annual Miles, L. K., Tan, L., Noble, G. D., Lumsden, J., & Macrae, C. N. (2011). Can a
conference of the Cognitive Science Society (pp. 503–508). Austin, TX: mind have two time lines? Exploring space–time mapping in
Cognitive Science Society. Mandarin and English speakers. Psychonomic Bulletin & Review, 18,
January, D., & Kako, E. (2007). Re-evaluating evidence for linguistic 598–604.
relativity: Reply to Boroditsky (2001). Cognition, 104, 417–426. Miller, G. A., & Johnson-Laird, P. N. (1976). Language and perception.
Klein, H. E. M. (1987). The future precedes the past: Time in Toba. Word, Cambridge, MA: Harvard University Press.
38, 173–185. Moore, K. E. (2004). Ego-based and field-based frames of reference in
Kranjec, A. (2006). Extending spatial frames of reference to temporal space to time metaphors. In M. Achard & S. Kemmer (Eds.), Language,
concepts. In R. Sun & N. Miyake (Eds.), Proceedings of the 28th annual culture, and mind (pp. 151–165). Stanford, CA: CSLI Publications.
conference of the Cognitive Science Society (pp. 447–452). Mahwah, NJ: Moore, K. E. (2006). Space to time mappings and temporal concepts.
Lawrence Erlbaum. Cognitive Linguistics, 17, 199–244.
Kranjec, A., & McDonough, L. (2011). The implicit and explicit Moore, K. E. (2011). Ego-perspective and field-based frames of reference:
embodiment of time. Journal of Pragmatics, 43, 735–748. Temporal meanings of FRONT in Japanese, Wolof, and Aymara. Journal
Ladewig, S. H. (2011). Putting the cyclic gesture on a cognitive basis. of Pragmatics, 43, 759–776.
CogniTextes, 6. <http://cognitextes.revues.org/406>. Müller, C. (1998). Redebegleitende Gesten: Kulturgeschichte, Theorie,
Lakoff, G., & Johnson, M. (1980). Metaphors we live by. Chicago, IL: Sprachvergleich. Berlin: Arno Spitz.
University of Chicago Press. Müller, C. (2008). Metaphors, dead and alive, sleeping and waking: A
Lakoff, G., & Johnson, M. (1999). Philosophy in the flesh: The embodied mind dynamic view. Chicago: University of Chicago Press.
and its challenge to Western thought. New York: Basic Books. Newton-Smith, W. H. (1980). The structure of time. London: Routledge &
Le Guen, O. (2011a). Speech and gesture in spatial language and cognition Kegan Paul.
among the Yucatec Mayas. Cognitive Science, 35, 905–938. Núñez, R. E. (2008). Reading between the number lines. Science, 321, 1293.
Le Guen, O. (2011b). Modes of pointing to existing spaces and the use of Núñez, R. E. (2011). No innate number line in the human brain. Journal of
frames of reference. Gesture, 11, 271–307. Cross-Cultural Psychology, 42, 651–668.
Le Guen, O., & Pool Balam, L. I. (2012). No metaphorical timeline in Núñez, R. E., & Cooperrider, K. (2013). The tangle of space and time in
gesture and cognition among Yucatec Mayas. Frontiers in Psychology: human cognition. Trends in Cognitive Sciences, 17, 220–229.
Cultural Psychology, 3, 271. Núñez, R. E., Cooperrider, K., Doan, D., & Wassmann, J. (2012). Contours of
Le Poidevin, R. (2003). Travels in four dimensions: The enigmas of space and time: Topographic construals of past, present, and future in the Yupno
time. Oxford: Oxford University Press. valley of Papua New Guinea. Cognition, 124, 25–35.
León-Portilla, M. (1990). Time and reality in the thought of the Maya. Núñez, R. E., & Cornejo, C. (2012). Facing the sunrise: Cultural worldview
Oklahoma: University of Oklahoma Press. underlying intrinsic-based encoding of absolute frames of reference
Levinson, S. C. (2003). Space in language and cognition. Cambridge: in Aymara. Cognitive Science, 36, 965–991.
Cambridge University Press. Núñez, R. E., Doan, D., & Nikoulina, A. (2011). Squeezing, striking, and
Levinson, S. C. (2006). The language of space in Yélî Dnye. In S. C. Levinson vocalizing: Is number representation fundamentally spatial?
& D. P. Wilkins (Eds.), Grammars of space (pp. 157–204). Cambridge: Cognition, 120, 225–235.
Cambridge University Press. Núñez, R. E., Motz, B. A., & Teuscher, U. (2006). Time after time: The
Levinson, S. C., & Brown, P. (1994). Immanuel Kant among the psychological reality of the Ego- and Time-Reference-Point distinction
Tenejapans: Anthropology as empirical philosophy. Ethos, 22, 3–41. in metaphorical construals of time. Metaphor and Symbol, 21, 133–146.
Levinson, S. C., Kita, S., Haun, D. B. M., & Rasch, B. H. (2002). Returning the Núñez, R. E., & Sweetser, E. (2006). With the future behind them.
tables: Language affects spatial reasoning. Cognition, 84, 155–188. Convergent evidence from Aymara language and gesture in the
Levinson, S. C., & Majid, A. (2013). The island of time: Yélî Dnye, the crosslinguistic comparison of spatial construals of time. Cognitive
language of Rossel Island. Frontiers in Psychology: Cultural Psychology, Science, 30, 401–450.
4, 61. Ouellet, M., Santiago, J., Funes, M. J., & Lupiáñez, J. (2010a). Thinking about
Levinson, S. C., & Wilkins, D. P. (Eds.). (2006). Grammars of space. the future moves attention to the right. Journal of Experimental
Cambridge: Cambridge University Press. Psychology: Human Perception and Performance, 36, 17–24.
Lewis, M. P., Simons, G. F., & Fennig, C. D. (Eds.) (2013). Ethnologue: Ouellet, M., Santiago, J., Israeli, Z., & Gabay, S. (2010b). Is the future the
Languages of the world (17th ed.). Dallas: SIL International. <http:// right time? Experimental Psychology, 57, 308–314.
www.ethnologue.com/>. Pederson, E., Danziger, E., Wilkins, D., Levinson, S., Kita, S., & Senft, G.
Lewis, P. A., & Miall, R. C. (2006). Remembering the time: A continuous (1998). Semantic typology and spatial conceptualization. Language,
clock. Trends in Cognitive Sciences, 10, 401–406. 74, 557–589.
382 A. Bender, S. Beller / Cognition 132 (2014) 342–382

Pöppel, E. (1997). A hierarchical model of temporal perception. Trends in Thornton, A. (1987). Maori oral literature as seen by a classicist. Dunedin,
Cognitive Sciences, 1, 56–61. NZ: University of Otago Press.
Pöppel, E., & Wittmann, M. (1999). Time in the mind. In R. A. Wilson & F. Torralbo, A., Santiago, J., & Lupiáñez, J. (2006). Flexible conceptual
C. Keil (Eds.), The MIT Encyclopedia of the cognitive sciences projection of time onto spatial frames of reference. Cognitive Science,
(pp. 841–843). Cambridge, MA: MIT Press. 30, 745–757.
Price, H. (1996). Time’s arrow & Archimedes’ point: New directions for the Traugott, E. C. (1975). Spatial expressions of tense and temporal
physics of time. New York, Oxford: Oxford University Press. sequencing: A contribution to the study of semantic fields.
Radden, G. (2004). The metaphor TIME AS SPACE across languages. In N. Semiotica, 15, 207–230.
Baumgarten et al. (Eds.), Übersetzen, interkulturelle Kommunikation, Traugott, E. C. (1978). On the expression of spatio-temporal relations in
Spracherwerb und Sprachvermittlung (pp. 225–238). Bochum: AKS. language. In J. H. Greenberg (Ed.), Universals of human language. Word
Rothe-Wulf, A., Beller, S., & Bender, A. (2014). Temporal frames of structure (Vol. 3, pp. 369–400). Stanford, CA: Stanford University
reference in three Germanic languages: Individual consistency, inter- Press.
individual consensus, and cross-linguistic variability (submitted for Tse, C.-S., & Altarriba, J. (2008). Evidence against linguistic relativity in
publication). Chinese and English: A case study of spatial and temporal metaphors.
Santiago, J., Lupiáñez, J., Pérez, E., & Funes, M. J. (2007). Time (also) flies Journal of Cognition and Culture, 8, 335–357.
from left to right. Psychonomic Bulletin & Review, 14, 512–516. Tversky, B., Kugelmass, S., & Winter, A. (1991). Cross-cultural and
Santiago, J., Román, A., Ouellet, M., Rodríguez, N., & Pérez-Azor, P. (2010). developmental trends in graphic productions. Cognitive Psychology,
In hindsight, life flows from left to right. Psychological Research, 74, 23, 515–557.
59–70. Ulrich, R., Eikmeier, V., de la Vega, I., Ruiz Fernández, S., Alex-Ruf, S., &
Sauter, D. A., Eisner, F., Ekman, P., & Scott, S. K. (2010). Cross-cultural Maienborn, C. (2012). With the past behind and the future ahead:
recognition of basic emotions through nonverbal emotional Back-to-front representation of past and future sentences. Memory &
vocalizations. Proceedings of the National Academy of Sciences, 107, Cognition, 40, 483–495.
2408–2412. Ulrich, R., & Maienborn, C. (2010). Left–right coding of past and future in
Scott, A. (1989). The vertical dimension and time in Mandarin. Australian language: The mental timeline during sentence processing. Cognition,
Journal of Linguistics, 9, 295–314. 117, 126–138.
Senft, G. (1995). Sprache, Kognition und Konzepte des Raums in Walsh, V. (2003). A theory of magnitude: Common cortical metrics of
verschiedenen Kulturen. Kognitionswissenschaft, 4, 166–170. time, space and quantity. Trends in Cognitive Sciences, 7, 483–488.
Senft, G. (Ed.). (1997). Referring to space. Oxford: Clarendon. Wassmann, J. (1994). The Yupno as post-Newtonian scientists: The
Shaki, S., Fischer, M. H., & Petrusic, W. M. (2009). Reading habits for both question of what is ‘natural’ in spatial description. Man, 29, 645–666.
words and numbers contribute to the SNARC effect. Psychonomic Weger, U., & Pratt, J. (2008). Time flies like an arrow: Space–time
Bulletin & Review, 16, 328–331. compatibility effects suggest the use of a mental timeline.
Sharma, A. (1974). The notion of cyclical time in Hinduism. Contributions Psychonomic Bulletin & Review, 15, 426–430.
to Asian Studies, 5, 26–35. Whorf, B. L. (1956). In J. B. Carroll (Ed.), Language, thought and reality.
Shinohara, K., & Pardeshi, P. (2011). The more in front, the later: The role Cambridge, MA: The MIT Press.
of positional terms in time metaphors. Journal of Pragmatics, 43, Yu, N. (1998). The contemporary theory of metaphor: A perspective from
749–758. Chinese. Amsterdam, Phil.: John Benjamins.
Sinha, C., Da Silva Sinha, V., Zinken, J., & Sampaio, W. (2011). When time is Yu, N. (2012). The metaphorical orientation of time in Chinese. Journal of
not space: The social and linguistic construction of time intervals and Pragmatics, 44, 1335–1354.
temporal event relations in an Amazonian culture. Language and Zinken, J. (2010). Temporal frames of reference. In P. Chilton & V. Evans
Cognition, 3, 137–169. (Eds.), Language, cognition, and space: Advances in cognitive linguistics
Talmy, L. (2000). Toward a cognitive semantics (Vol. 1): Conceptual (pp. 479–498). London: Equinox Publishing Ltd..
structuring systems. Cambridge: MIT Press. Zinken, J., Sampaio, W., da Silva Sinha, V., & Sinha, C. (2005). Space, motion
Tenbrink, T. (2011). Reference frames of space and time in language. and time in Amondawa: Field manual 2005–2006. Portsmouth:
Journal of Pragmatics, 43, 704–722. University of Portsmouth.

S-ar putea să vă placă și