God Is Up? Replication and Extension Attempts of

Abstract

Meier et al. (2007) examined relations between the concept of God and the vertical dimension of space. They found that God-related concepts were recognized faster if they were associated with words related to a high location (Study 1) or actually presented in a high location (Study 2). Additionally, participants recalled God-like images as appearing higher than they actually did (Study 3) and judged strangers as having a stronger belief in God if the strangers’ images appeared toward the top versus bottom of a computer screen (Studies 4 and 5). Meier et al. (2007) suggested that people use the vertical dimension of space to represent the concept of God. The present investigation attempted to replicate effects from four of these studies using adequately powered designs while also examining potential individual difference moderators. We successfully replicated the effects of Studies 1 and 3, but not Studies 4 or 5. Individual differences related to God beliefs and religiosity did not moderate the findings. The results and a prior successful replication of Meier et al.’s (2007) Study 2 suggest that the concept of God is represented with the vertical dimension of space, but this representation does not appear to impact perceptions of other’s traits.

Full text

Running Head: GOD IS UP? 1
In Press at the Psychology of Religion and Spirituality
©American Psychological Association, [2021]. This paper is not the copy of record and may
not exactly replicate the authoritative document published in the APA journal. Please do
not copy or cite without author's permission. The final article is available, upon
publication, at: [DOI forthcoming]
God is Up?
Replication and Extension Attempts of Meier et al. (2007)
Brian P. Meier1, Adam K. Fetterman2, David J. Hauser3, & Michael D. Robinson4
1Gettysburg College
2University of Houston
3Queen’s University
4North Dakota State University
BP Meier and AK Fetterman acknowledge support for this project from the John Templeton
Foundation (grant number 61592). Data collection plans for the studies were uploaded to the
Open Science Framework website before data collection commenced: https://osf.io/tgb9j/. The
cleaned data files in SPSS format can also be found at this website. Address correspondence to
Brian P. Meier, Gettysburg College, bmeier@gettysburg.edu.

GOD IS UP? 2
Abstract
Meier et al. (2007) examined relations between the concept of God and the vertical dimension of
space. They found that God-related concepts were recognized faster if they were associated with
words related to a high location (Study 1) or actually presented in a high location (Study 2).
Additionally, participants recalled God-like images as appearing higher than they actually did
(Study 3) and judged strangers as having a stronger belief in God if the strangers’ images
appeared toward the top versus bottom of a computer screen (Studies 4 and 5). Meier et al.
(2007) suggested that people use the vertical dimension of space to represent the concept of God.
The present investigation attempted to replicate effects from four of these studies using
adequately powered designs while also examining potential individual difference moderators.
We successfully replicated the effects of Studies 1 and 3, but not Studies 4 or 5. Individual
differences related to God beliefs and religiosity did not moderate the findings. The results and a
prior successful replication of Meier et al.’s (2007) Study 2 suggest that the concept of God is
represented with the vertical dimension of space, but this representation does not appear to
impact perceptions of other’s traits.
KEYWORDS: Metaphor, God, Vertical Space, Embodiment, Person Perception, Replication

GOD IS UP? 3
God is Up?
Replication and Extension Attempts of Meier et al. (2007)
God is often described in metaphoric terms, such as when Christians refer to God as
“the man upstairs” or a “father” (DesCamp & Sweetser, 2005). Such metaphor-related ways of
speaking about God may help speakers convey some of God’s central attributes, such as
wisdom, power, or goodness.
In addition, metaphors may provide key insights into how people think about God,
potentially in ways that get to the core of our representational systems. Lakoff and Johnson
(1980, 1999) developed the conceptual metaphor theory (CMT) which proposed that
metaphors map abstract concepts such as “love” to more concrete domains like “journeys”.
These metaphoric mappings are cognitive associations, often learned through experience,
that allow people to think about and represent abstract concepts (e.g., love) in ways that are
more concrete and readily perceptible in a direct sense (e.g., journeys). Metaphoric mappings
commonly involve associations between abstract concepts and body-based experiences, likely
because we can easily understand concepts we can actually experience with our senses. Ideas
along these lines are also considered in terms of embodiment or embodied cognition (Fincher-
Kiefer, 2019; Meier et al., 2012).
A number of authors have suggested that CMT and embodiment views are well-suited for
examining religion because religious beliefs tend to be abstract in nature and should therefore
benefit from conceptions related to metaphor or bodily experience (Barsalou et al., 2005;
Soliman et al., 2015; Van Capellan et al., 2021). Some supportive research on this connection
has examined religion, CMT, and embodiment in tasks that involve judgment or behavior. For
example, in a two-week experience sampling study, Fetterman (2016) found that people who had

GOD IS UP? 4
a stronger belief in God reported feeling physically cleaner on days on which they reported
higher self-control, consistent with metaphors linking cleanliness to Godliness. Concerning
explicit thoughts and judgments, Ransom and Alicke (2013) found that participants who adopted
a religious posture (e.g., kneeling) reported a higher frequency of religion-related cognitions.
Adopting religious postures, that is, seemed to affect one’s thinking in the religious domain.
Inspired by CMT and embodiment theory, Meier et al. (2007) examined potential
associations linking representations of God to higher locations in vertical space (which
presumably give rise to expressions or thoughts consistent with God watching down on us). In
specific terms, Meier et al. (2007) hypothesized that people somewhat automatically link God to
verticality, which should result in patterns of encoding, retrieval, and judgment that
systematically (and unwittingly) favor God’s higher spatial position. In these studies, participants
categorized God-related stimuli more quickly when they were linked to words describing higher
locations in space (Study 1) or when they were actually shown in higher spatial positions (Study
2). Furthermore, participants recalled that God-like images had been presented vertically higher
than was actually the case (Study 3) and manipulations of vertical position influenced inferences
concerning pictured strangers’ belief in God (Study 4 and Study 5).
Meier et al. (2007) revealed that people seem to think about God in a way that reflects
associations found in common metaphors. Studies have shown similar findings in other
metaphoric domains (Landau, 2017; Landau et al., 2010). Although the Meier et al. (2007) paper
has been cited 360 times (according to Google Scholar as of March, 2021), the sample sizes in
the experiments were small (Ns = 41, 47, 33, 27, 66, & 55) and there have been few conceptual
replications reported in this area (although see Chasteen et al., 2010 & Guan et al., in press, for
two examples). Furthermore, valid concerns have been raised concerning the replicability of

GOD IS UP? 5
CMT and embodied cognition research (Skulmowski & Rey, 2018). For example, there have
been reported failures to replicate metaphor and embodied cognition effects on social judgments
and in studies involving judgments of other’s traits (e.g., Ashton et al., 2014; Chabris et al.,
2019; Siev et al., 2019; but see Hauser & Schwarz, in press, for a discussion of relevant
boundary conditions; also see Giessner & Schubert, 2019, for a successful replication). For
instance, Siev et al. (2019) reported a failure to replicate the “Macbeth Effect” which occurs
when unethical thoughts trigger desires or behaviors related to cleansing oneself. Similarly,
Ashton et al. (2014) reported a failure to replicate findings linking the personality trait of
agreeableness to a preference for sweet foods. These findings relate to replication issues in
psychological research more generally, given that (only) 50% of research results reported in the
field might be replicable (Klein et al., 2018).
Nonetheless, successful replications have occurred using implicit cognitive tasks in
which metaphor-related cues facilitate metaphor-consistent response latencies (e.g., Meier et al.,
2015). However, it is difficult to know whether such apparent discrepancies between social
judgment and response latency studies are real because there have been no attempts to replicate
any particular relationship across both social judgment and cognitive tasks within the same
metaphoric domain. Efforts of this type were central to the present research.
We conducted a series of studies that attempted to replicate findings reported by Meier et
al. (2007) with the exception of their Study 2, which was replicated by Meier and Fetterman (in
press) in a different type of project that examined multiple metaphors used in describing God
(i.e., spatial location, brightness, & huamanness). In the 2007 paper, Studies 1 and 2 involved
categorization latencies and Study 3 used a recall task – i.e., dependent measures that are
cognitive in nature. Studies 4 and 5, on the other hand, involved a person perception task and

GOD IS UP? 6
found that pictured strangers were inferred to believe in God to a greater extent. The latter is a
“social priming” task because it involves judgments or inferences about other people in the
context of a metaphor-related manipulation. Accordingly, the range of studies reported by Meier
et al. (2007) seemed ideal in making conclusions about the scope and power of conceptual
metaphors in representations of God and religiosity.
We also attempted to extend Meier et al.’s (2007) findings by examining potential
moderators related to belief in God, religiosity, and endorsement of the God is up metaphor. Past
work has shown that individual differences in metaphor use (Fetterman et al., 2016) or
concerning the metaphoric relation of interest (Fetterman et al., 2019) moderate effects of the
type discussed here. In the present case, for example, more religious individuals were presumed
to possess stronger associations involving religious concepts, which should facilitate metaphor-
linked processes related to those concepts (Lee, 2016). We should state, however, that our central
focus was on replication rather than any potential moderation across individuals.
Disclosure Statement
We report all studies, data exclusions, manipulations, and measures, and how we
determined sample sizes. Data-collection plans (uploaded before data collection commenced)
and cleaned data files in SPSS format can be found on the Open Science Framework website:
https://osf.io/tgb9j/.
Study 1: Does Verticality Affect Perceptions of a Stranger’s Belief in God (on a six-point
rating scale)?
Our original data-collection plan was to attempt to replicate Study 4 from Meier et al.
(2007) and to determine if differences existed among three data-collection contexts that are often
used in Psychology: an online context using Amazon Mechanical Turk (Mturk), traditional

GOD IS UP? 7
samples collected in the context of Psychology research labs, and a sample collected in the
context of a Psychology class project.
In Meier et al.’s (2007) Study 4, 27 participants rated strangers’ belief in God as the
strangers’ images appeared near the top or bottom of a computer screen. Meier et al. (2007)
hypothesized that strangers shown in higher versus lower vertical positions would be rated as
having a stronger belief in God, which was confirmed in the original study (d = .58). However,
the sample size was unacceptably small by modern standards. We used G*Power to determine
the sample size needed to detect a small effect (d = .20) with 80% power in a paired-samples t-
test. Although we test differences between contexts and examined individual differences, we
were most interested in the replication of the original finding, which used a paired-samples t test.
The sample size estimate was 199 and we attempted to collect data from approximately 200
participants in each of the contexts listed above.
Method
Participants
The initial sample consisted of 621 participants. Fifty-four participants were removed
prior to analyses for the following reasons: the computer program created duplicate participant
numbers for 20 participants, which precluded linking questionnaire data to rating data, 12
participants did not complete the experimental task, and 22 participants gave the same ratings for
all trials. The final sample consisted of 567 participants:
• 197 participants with an average age of 36.49 (SD = 10.68) years from Mturk who were paid
$1.75 (92 females, 102 males, & 3 missing; 80.71% White, 6.09% Black, 5.58%
Asian/Pacific Islander, 5.58% Hispanic, 1.02% Bi-Racial, .51% American Indian/Alaskan
Native, & .51% missing)

GOD IS UP? 8
• 181 participants with an average age of 19.33 (SD = 1.41) years who were recruited as
volunteers by 21 students as part of a research project in a Statistics and Research Methods
class at Gettysburg College (104 females, 74 males, 2 missing, & 1 other; 80.11% White,
6.63% Hispanic, 6.08% Asian/Pacific Islander, 3.31% Black, 1.66% Bi-Racial, 1.10%
American Indian/Alaskan Native, & 1.10% missing)
• 189 participants with an average age of 19.36 (SD = 3.07) years who completed the study in
the research labs of the second, third, and fourth authors in exchange for course credit (114
females, 71 males, 2 other, & 2 missing; 54.50% White, 29.63% Hispanic, 8.99%
Asian/Pacific Islander, 3.17% Black, 1.59% missing, 1.06% Bi-Racial, & 1.06% American
Indian/Alaskan Native)
Materials and Procedure
We used Inquisit software (www.millisecond.com) to present the study over the internet,
following the procedures of Meier et al.’s (2007) Study 4. Stimuli consisted of one hundred (50
male & 50 female) grayscale images from the AR face database (Martinez & Benavente, 1998).
These images, photographed from the shoulder up, depicted individuals who appear to range in
age from 20 to 50 years. As in Meier et al. (2007), instructions indicated that the researchers
were interested in how people make inferences about the extent to which strangers believe in
God. On each trial, participants saw a new, unfamiliar target and their task was to make a rating
of that person’s belief in God using a rating scale that was shown at the vertical center of the
screen. The 100 images randomly appeared near the top or bottom of the computer screen,
subject to the constraint that there needed to be 50 at each position. The rating scale had a
horizontal arrangement with 1 (no belief in God) toward the left and 6 (strong belief in God)

GOD IS UP? 9
toward the right. After each rating, the image disappeared, and there was a 1,500-ms delay
before the next image appeared.
Participants responded to demographic questions (age, sex, ethnicity, & religious
affiliation) after the task as well as the following single-item self-reported measures:
• Belief in God (M = 5.07; SD = 1.89; Where would you place yourself on the following
scale?; 1 = Strong Atheist - I am 100% certain there is no God to 7 = Strong Theist - I am
100% certain that there is a God)
• Religiosity (M = 3.48; SD = 1.97; To what extent are you a religious person?; 1 = not at all
religious to 7 = very religious)
• God is Up agreement (M = 3.98; SD = 2.23; To what extent do you agree with this statement:
“God is up high”?; 1 = not at all to 7 = strongly agree)
Results
We first examined the effect of position on belief in God ratings using a paired-samples t
test. Contrary to Meier et al. (2007), the effect of image position was not significant, t(566) =
.36, p = .72, d = .02 (95% CI: -.07, .10). The present means, as well as the means from Meier et
al.’s (2007) Study 4, are shown in Table 1. A Forest Plot with the effects and confidence
intervals of the original studies and the comparable replication studies from the current project
are shown in Figure 1.

GOD IS UP? 10
Table 1. Original (Meier et al., 2007, Study 4) and Replication (Study 1) Means and Standard
Deviations for the Belief in God Ratings by Image Position
Belief in God ratings
Image Position
Meier et al., 2007, Study 4
Replication, Study 1
Top
3.61 (.34)
3.53 (.51)
Bottom
3.48 (.27)
3.54 (.52)
Figure 1. Forest Plot of the Effect Sizes and Confidence Intervals of the Original Studies and
Associated Replication Studies from the Current Project
We next used a mixed-model ANOVA with image position (top, bottom) as a within-
participant variable and study context (online, lab, class project) as a between-participants
variable in order to examine the effect of context. The main effect of image position was not
significant, F(1, 564) = .10, p = .76, ηp2 < .01 (95% CI: .00, .00), which accords with the results
of the t-test discussed above. There was a main effect for study context, F(2, 564) = 23.71, p <

GOD IS UP? 11
.001, ηp2 = .08 (95% CI: .04, .12), such that higher belief ratings occurred online (M = 3.70, SD =
.48) relative to in the lab (M = 3.54; SD = .47) or with respect to a class project (M = 3.36, SD =
.47). There was no image position by study context interaction, F(2, 564) = 1.67, p = .19, ηp2 =
.01 (95% CI: .00, .02).
Finally, we examined potential moderation effects by the individual difference variables.
In order to simplify these analyses and their interpretation, we created a difference score for the
image rating task by subtracting the mean rating of images presented at bottom from the mean
rating of images presented at top (M = -.01; SD = .34). Positive (negative) scores mean that
participants gave higher belief in God ratings for higher (lower) targets. We computed
correlations between this difference score with the three potential moderators (note that not every
participant completed all three measures). The belief in God, r(561) = .01 (95% CI: -.07, .09), p
= .86, and religiosity, r(561) = .05 (95% CI: -.03, .13), p = .26, measures were not correlated
with verticality effects. The correlation with the item assessing agreement that God is up was
significant, r(560) = .09 (95% CI: .01, .17), p = .04. This correlation reveals that participants
who generally believed God is up were more likely to ascribe stronger God beliefs to targets
presented upwards (versus downwards). This correlation was small, however.
Study 2: Does Verticality Affect Perceptions of a Stranger’s Belief in God (on a continuous
rating scale)?
In the remaining studies, we abandoned the different data-collection contexts and focused
on replication attempts of the remaining studies of Meier et al. (2007). We first attempted to
replicate their Study 5, which involved a similar social perception task, albeit with some
differences. One, the original paradigm randomized individual trials to one of two ratings -
inferred beliefs in God and likability (their Study 5a) or power (their Study 5b), which Meier et

GOD IS UP? 12
al. (2007) added to address alternative mechanisms. Given the results of Study 1 reported above,
we only attempted to replicate the effects involving God-belief ratings (i.e., we ran one study
with only God-belief ratings as the dependent measure using the paradigm described below).
Two, belief-ratings were made for fewer trials. And three, ratings were made using a continuous
rating bar, which might afford greater sensitivity.
We created a second data-collection plan and uploaded it to the Open Science Framework
website before data collection commenced for this study and the remaining studies
(https://osf.io/tgb9j/). In the current study, this plan revealed that we needed 156 participants to
find a small effect with 80% statistical power using a paired-samples t-test. However, we made a
mistake using G*Power and the correct figure was 199.
Method
Participants
The sample was collected at Queen’s University and consisted of 189 participants. Due to
a programming error, potential moderators for this study were not collected and demographics
that were collected consisted of age and gender, but not race.1 Gender and age was missing from
15 participants and, of the remaining 174 participants, 137 were female and 37 were male, with a
mean age of 19.47 (SD = 2.66) years.
Materials and Procedure
The procedures were identical to Studies 5a and 5b from Meier et al. (2007) and involved
48 images (24 of each sex) from the AR face database (Martinez & Benavente, 1998). Half of
the targets of each sex appeared near the bottom of the screen and half of the targets of each sex
appeared near the top. Two counterbalanced programs ensured that all targets were presented in
high vertical positions for some participants and low vertical positions for others. Target images

GOD IS UP? 13
were presented on a white background with a rating bar arranged horizontally in the vertical
center of the screen. This rating bar had “No Belief in God” toward the left and “Strong Belief in
God” toward the right. Participants used the mouse and mouse cursor to select a point along the
rating bar that reflected their impressions of God belief for each pictured stranger, resulting in a -
1 to +1 scale with .001 increments.
Results
We used a paired-samples t-test to determine whether vertical placement of the images
affected God-belief ratings. The effect of position was not significant, t(188) = .88, p = .38, d =
.06 (95% CI: -.08, .20). The means from this study and the corresponding means from Meier et
al. (2007) are shown in Table 2.
Table 2. Original (Meier et al., 2007, Study 5a & Study 5b) and Replication (Study 2) Means and
Standard Deviations of Belief in God Ratings by Image Position
Belief in God ratings
Image Position
Meier et al., 2007,
Study 5a
Meier et al., 2007,
Study 5b
Replication,
Study 2
Top
439.96 (71.39)
438.55 (68.09)
.04 (.12)
Bottom
412.47 (71.95)
425.73 (73.32)
.03 (.13)
Note: The replication Study 2 used a -1 to +1 rating scale with .001 increments for mouse-click
locations while Studies 5a and 5b from Meier et al. (2007) used a 750-point rating scale with an
increment of 1 for mouse-click locations.
Study 3: Are God and Vertical Space Implicitly Related?
We failed to replicate the studies from Meier et al. (2007) that focused on the perception
of others’ belief in God. Nonetheless, we suspected that the studies using more straightforward

GOD IS UP? 14
cognitive procedures would replicate, so we sought to make this comparison. The current Study
3 is an attempt to replicate Meier et al. (2007) Study 1. That study used an implicit association
test (IAT: Greenwald et al., 1998) to investigate whether participants associate God versus Devil
concepts with verbal descriptors indicating a higher versus lower position in space. Meier et al.
(2007) found a strong, hypothesis-consistent effect (ηp2 = .85), but we sought to be conservative
in our sample size estimate. A G*Power analysis indicated that 34 participants would give us
80% power to detect a medium-sized (ηp2 = .06) IAT effect. Regardless, we sought to exceed this
number as much as possible given the data-collection session timeframe and resources available.
We managed to recruit 94 participants.
Method
Participants
The sample was collected at the University of Texas-El Paso and consisted of 94
participants. We removed one participant who had IAT accuracy rates less than 50%, which left
a sample size of 93 participants (48 females, 44 males, 1 missing) with an average age of 21.24
(SD = 3.84) years. Self-reported race was as follows: 46.24% Hispanic, 22.58% Bi-Racial,
21.50% White, 7.53% Black, & 1.08% Asian/Pacific Islander.
Materials and Procedure
Meier et al.’s (2007) Study 1 procedures were used, which required us to use the original
methodological procedures for the IAT (Greenwald et al., 1998). Stimuli consisted of four words
from each category pertinent to the present associations: God (Almighty, Creator, Deity, &
Lord), Devil (Antichrist, Demon, Lucifer, & Satan), up (High, Top, Above, & Ascend), and
down (Low, Bottom, Below, Descend). Participants were told to categorize the words as quickly
and as accurately as possible. Category labels appeared toward the upper left and right of the

GOD IS UP? 15
computer screen. The choice to be made was always between “God” and “Devil” or between
“Up” and “Down.” Participants were told to press the “Q” key if the word belonged to the
relevant category on the left and the “P” key if it belonged to the relevant category on the right.
Incorrect categorizations were followed by the word “INCORRECT” in red font for 1.5 seconds.
Correct categorizations were followed by a 150-ms blank screen. Each stimulus remained on the
screen until a response was made.
The IAT had seven blocks, but only two blocks comprised the critical comparisons: block
four (“God or Up” versus “Devil or Down”: 48 trials) and block seven (“Devil or Up” versus
“God or Down”: 48 trials). We randomly assigned participants to one of two versions of the IAT,
which counterbalanced which of the two critical blocks occurred first.
Demographics (age, sex, ethnicity, & religious affiliation) were collected after the task as
were other individual difference variables. These items varied slightly from Study 1. Also as in
Study 2, a programming error resulted in the possibility of participants skipping these post-
experiment questions, which a number of participants did. The pertinent moderator questions
consisted of the following:
• Belief in God (N = 43; M = 5.07; SD = 2.06; I believe God Exists; 1 = strongly disagree to 9
= strongly agree)
• Religious Beliefs (N = 62; M = 4.63; SD = 2.28; I am a religious person; 1 = strongly
disagree to 9 = strongly agree)
• God is Up belief (N = 89; M = 5.64; SD = 1.80; To what extent do you view or think about
God as being Up or High; 1 = not at all to 7 = to a large extent)

GOD IS UP? 16
Results
We followed procedures outlined by Greenwald et al. (1998), which involved deleting the
first two trials of each block, replacement of trials with response times that were below 300 ms
and above 3,000 ms with these values, deletion of inaccurate trials (3.2%), and a log-
transformation of the raw latencies. For each participant, we then computed two raw means and
two log-transformed means, one set pertaining to the compatible block (God + Up versus Devil +
Down) and the other set pertaining to the incompatible block (God + Down versus Devil + Up).
Of note, analyses were performed on transformed latency means, but patterns are reported in
terms of millisecond units.
Data were analyzed with a mixed-model ANOVA, with block order and block type as
independent variables. Confirming predictions, the main effect of IAT block type was
significant, F(1, 91) = 326.98, p < .001, ηp2 = .78 (95% CI: .70, .83). As shown in Table 3, this
effect replicates Meier et al. (2007) and indicates that participants implicitly associate God with
higher vertical positions, relative to the Devil.
Table 3. Original (Meier et al., 2007, Study 1) and Replication (Study 3) Means and Standard
Deviations for the Reactions Times (ms) for each IAT Block Type
Response latency (in ms)
IAT Block Type
Meier et al. (2007), Study 1
Replication, Study 3
God-Up and Devil-Down
708.71 (83.66)
803.55 (141.93)
God-Down and Devil-Up
976.35 (184.61)
1,071.65 (227.82)
The remaining effects within the initial analysis were not of major interest, but are
reported. The interaction between block order and block type was significant, F(1, 91) = 4.94, p

GOD IS UP? 17
= .03, ηp2 = .05 (95% CI: .00, .16). This interaction revealed that the IAT effect was stronger
when the compatible block occurred first, t(47) = 12.43, p < .001, d = 1.79 (95% CI: 1.33, 2.25),
relative to second, t(45) = 10.91, p < .001, d = 1.61 (95% CI: 1.17, 2.04). The main effect of
block order was not significant, F(1, 91) = 3.18, p = .08.
To examine effects involving the individual difference variables, we created an IAT score
by subtracting the compatible log-transformed mean from the incompatible log-transformed
mean, with higher numbers indicating associations consistent with vertical metaphors (i.e., God
is up). None of the potential moderators significantly correlated with the magnitude of the IAT
effect: belief in God, r(41) = -.03 (95% CI: -.33, .27), p = .87, religiosity, r(60) = -.10 (95% CI: -
.34, .15), p = .46, or agreement with the God is up item, r(87) = .00 (95% CI: -.21, .21), p = .99.
Study 4: Are God-Like Images Falsely Recalled as Appearing Higher in Vertical Space
than Devil-Like Images?
In our final study, we attempted to replicate Meier et al.’s (2007) Study 3 in which 33
participants were presented with God-like, Devil-like, and neutral images one at a time, with the
images placed in different vertical locations. Meier et al. (2007) found that God-like images were
falsely recalled as appearing higher than Devil-like images.
The Meier et al. (2007) effect size was small to medium (d = .38), but we sought to be
conservative with respect to a small (d = .20) effect size in the present context. The data
collection plan specified that we needed 156 participants to have 80% power to detect effects in
this range. However, as before, we made a mistake with G*Power and the correct number was
199. Given the data-collection session timeframe and available resources, we were only able to
collect data from 125 participants. Even so, sensitivity analyses indicated that 125 participants
gave us 80% power to detect an effect size of d = .25.

GOD IS UP? 18
Method
Participants
The sample was collected at North Dakota State University and consisted of 125
participants (70 females, 53 males, 2 missing) with an average age of 19.25 (SD = 1.90) years.
Self-reported race was as follows: 87.20% White, 5.60% Asian/Pacific Islander, 3.20% Black,
1.60% Hispanic, 1.60% Bi-Racial, and .80% missing.
Materials and Procedure
We used procedures identical to Meier et al.’s (2007) Study 3. Instructions indicated that
the researchers were interested in memory and that the participants would see images to be
encoded for a later memory test. They also indicated that the images would appear at different
locations on the computer screen in order to increase participants’ attention to the task. Stimuli to
be encoded consisted of 5 God-like images, 5 Devil-like images, and 20 neutral images that were
identical to those presented in Meier et al. (2007). Each image appeared for 1 second, which was
followed by a 2-second blank screen. The images appeared in one of five vertical locations from
the top to the bottom of the screen and an equal number of images from each category were
presented in each location.
After all images were shown, participants were told that they would now see each image
again, one at a time, in the center of the screen. On the right side of the screen was a white bar
that spanned from top to bottom. Participants were instructed to use the mouse cursor to select a
point on the vertical bar that reflected the vertical position at which the center of the image had
been presented earlier in the task, which resulted in a 1 (top) to 768 (bottom) vertical scale.
Participants were asked to recall the vertical location of all 30 images, which received a new

GOD IS UP? 19
random ordering during the test phase. After each location choice was made, there was a 100 ms
blank-screen delay until the next image was shown.
Participants responded to demographic (age, sex, ethnicity, & religious affiliation)
questions after the task as well as the single-item self-reported measures used in Study 3.
However, as in Study 3, a programming error resulted in a number of participants skipping these
post-experiment questions related to Belief in God (N = 60; M = 4.67; SD = 1.93), Religious
Beliefs (N = 94; M = 4.62; SD = 2.15), and God is Up belief (N = 111; M = 5.19; SD = 1.81).
Results
As in Meier et al. (2007), we subtracted the remembered location of the religious images
from the remembered location of the neutral images presented at the same location, with positive
(negative) numbers reflecting a higher (lower) vertical placement relative to neutral controls. As
shown in Table 4, and replicating the effect of Meier et al. (2007), the God-like images were
recalled as appearing higher than the Devil-like images, t(124) = 4.15, p < .001, d = .37 (95%
CI: .19, .55).
Table 4. Original (Meier et al., 2007, Study 3) and Replication (Study 4) Means and Standard
Deviations of the Original Vertical Placement of God-Like and Devil-Like Images
Vertical Placement Recall
Image Type
Meier et al., 2007, Study 3
Replication, Study 4
God-Like
36.04 (121.14)
8.87 (128.93)
Devil-Like
-21.02 (113.70)
-52.11 (138.19)
Next, we examined potential moderators. As in the prior studies, we created a difference
score by subtracting the average location chosen for Devil-like images from the average location

GOD IS UP? 20
chosen for God-like images, with positive scores indicative of metaphor-consistent biases. We
then correlated this difference score with the potential moderators. Participants who believed in
God to a greater extent exhibited metaphor-consistent memory biases to a greater extent, though
this relationship did not the reach conventional level of significance, r(58) = .25 (95% CI: -.004,
.47), p = .05. Correlations with the religiosity, r(92) = .09 (95% CI: -.11, .29), p = .38, and God
is up, r(109) = .19 (95% CI: .004, .36), p = .21, items were not significant.
General Discussion
We conducted a series of replication and extension attempts based upon the studies from
Meier et al. (2007). Our results revealed successful replications of effects that involved cognitive
tasks, such as reaction-time based categorizations and memories for God-like images that varied
in vertical location. However, effects that involved the perception of others’ belief in God failed
to replicate. Also, individual differences in belief in God, religiosity, or belief that God is up did
not consistently moderate the associations that were found. We discuss the implications of this
collection of findings below.
Replication Implications
The results bear upon the replicability of metaphoric and embodied effects, including the
realms of religious cognition and symbolism. There have been fewer attempted replications of
embodied or conceptual metaphor effects that involve cognitive tasks (e.g., those using reaction
time metrics) relative to perception tasks involving the judgment of others, the latter of which
often invoke the label of “social priming”. The current replications span both domains in the
context of a single phenomenon - God’s purported verticality. That the cognitive tasks displayed
consistent evidence for metaphoric associations and the social judgments tasks did not accords
with the idea that cognitive priming seems to be the more robust phenomenon (O’Donnell et al.,

GOD IS UP? 21
2018). Thus, in assessing the replicability of metaphor-related (and embodied) phenomena, it
may be important to consider methodological details as some procedures and foci are likely to
produce more robust effects than others.
Our failure to replicate social judgment effects could be due multiple reasons: (1) vertical
location might not affect perceptions of another’s belief in God, (2) the designs used might be
unreliable, or (3) there are unknown moderator variables. All of these factors and others are
possible (e.g., Hauser & Schwarz, in press). Yet, other studies have found support for metaphor-
linked social perception effects that are probably reliable. For example, Persich et al. (in press)
examined metaphoric associations between brightness and God-related concepts. In one study,
participants rated individuals wearing white shirts or dresses as more religious than those
wearing black shirts or dresses. Such results are consistent with metaphors linking religiosity to
lightness (rather than darkness) and we suspect that these effects would replicate. This is because
people dress themselves and the colors of clothing that they wear can therefore be ascribed to
personal preferences to some legitimate extent. Under such circumstances, clothing color can
quite reasonably be used as a cue to the beliefs or personality of the person wearing the clothing
(Gosling, 2008). In the present case, by contrast, targets did not choose vertical locations to have
their pictures presented in and vertical locations on a computer screen probably do not reveal
anything about a person. In other words, picture verticality was simply not “applicable” to the
targets (Higgins, 1996). Future work can examine whether such attribution-related explanations
can distinguish social priming effects that are likely to replicate from those that have a more
fragile basis to them.

GOD IS UP? 22
Individual Differences
Individual differences in belief in God, religiosity, and endorsement of the idea that God
is up did not consistently moderate the present effects, regardless of which type of task was
involved. Other work, by contrast, has shown that individual differences sometimes do moderate
metaphoric and embodied effects (Fetterman et al., 2016, 2018; Persich et al., in press). For
example, people who preferred white to black and lighter to darker rooms reported that they
believed in God to a greater extent in several of Persich et al.’s (in press) studies. As another
example, Fetterman et al. (2018) showed that higher feelings of physical warmth in a daily report
study predicted higher levels of agreeable behavior (e.g., being nice, friendly, & caring on a
given day), which implicates personality processes of a within-person type.
Thus, it is clear that individual differences can moderate metaphor effects, even in
religious domains, though they did not generally do so in the current studies. One reasonable
interpretation of these findings is that implicit cognitive associations linking God concepts to
higher levels of vertical space are culturally shared, if not somewhat universal (Guan et al., in
press; Kövecses, 2005), and thus even those with lesser beliefs in God would still exhibit the
relevant patterns or associations. For example, Guan et al. (in press) found that participants from
Tibet linked Buddhist-related words with higher locations in space. An additional possibility,
though, is that there are individual difference moderators that were not assessed. In this
connection, Persich et al. (in press) and Fetterman et al. (2018) created individual difference
measures that were closely associated with the symbolic phenomenon of interest (e.g., brightness
or warmth), but this individual difference strategy was not fully employed in the present studies.
For example, one might tap individual differences in preferences for low versus high vertical
spaces to determine if they moderate effects like the ones found here. Accordingly, we conclude

GOD IS UP? 23
that there are opportunities for advancement in thinking about, and documenting, the manner in
which individual differences are key to embodied phenomena (Fetterman et al., 2021).
Limitations
The current research does have limitations. First, we made errors in our sample size
estimates as well as in some of the programming, which resulted in lost data. Yet, in terms of
statistical power, we had adequate power, in all studies, to find small effects. The Ns for the
analyses with the individual differences varied across studies and items, though, and this could
have affected results involving these variables.
A second limitation is that the replication attempts carried out in the present context
involved three authors who helped carry out the original research (though a fourth author was not
involved in the original research). This allowed us to approach the studies with care and
deliberation, though not as fully independent scientists. To deal with such issues, we tried to be
as transparent as possible in the conduct of the research. This extended to posting data collection
plans as well as analysis plans at Open Science Framework website. It also extended to the
transparent reporting of all data exclusions, manipulations, and measures. Importantly, we
conclude that verticality manipulations do not influence social perceptions, though we had
originally believed findings of this type to be robust.
Conclusion
In four studies, we successfully replicated effects from Studies 1 and 3, but not Studies 4
or 5, of Meier et al. (2007). These results reveal that metaphors for God that involve vertical
space reflect the ways in which people think about God in implicit cognitive terms. Yet, it does
not appear that people use associations of this type when explicitly judging the belief in God of
unknown others. Attempts to investigate and replicate metaphor and embodied effects might

GOD IS UP? 24
consider systematically examining the cognitive versus social judgement distinction highlighted
in the present work.

GOD IS UP? 25
References
Ashton, M. C., Pilkington, A. C., & Lee, K. (2014). Do prosocial people prefer sweet-tasting
foods? An attempted replication of Meier, Moeller, Riemer-Peltz, and Robinson (2012).
Journal of Research in Personality, 52, 42-46.
Barsalou, L. W., A. K. Barbey, W. K. Simmons, & Santos, A. (2005). Embodiment in religious
knowledge. Journal of Cognition and Culture, 5, 14-57.
Chasteen, A. L. Burdzy, D. C. Pratt, J. (2010). Thinking of God moves attention.
Neuropsychologia, 48, 627-630.
Chabris, C. F., Heck, P. R., Mandart, J., Benjamin, D. J., & Simons, D. J. (2019). No evidence
that experiencing physical warmth promotes interpersonal warmth: Two failures to
replicate. Social Psychology, 50, 127-132.
DesCamp, M. T., & Sweetser, E. E. (2005). Metaphors for God: Why and how our choices
matter for humans? The application of contemporary cognitive linguistics research to the
debate on God and metaphor. Pastoral Psychology, 53, 207-238.
Fetterman, A. K. (2016). On God-belief and feeling clean: Daily experiences are related to
feeling clean, particularly for those high in God-belief. Social Psychological and
Personality Science, 7, 552-559.
Fetterman, A. K., Bair, J. L., Werth, M., Landkammer, F., & Robinson, M. D. (2016). The scope
and consequences of metaphoric thinking: Using individual differences in metaphor
usage to understand how metaphor functions. Journal of Personality and Social
Psychology, 110, 458-476.

GOD IS UP? 26
Fetterman, A. K., Meier, B. P., & Robinson, M. D. (2019). Dispositional properties of metaphor:
The predictive power of sweet taste metaphor for trait and daily prosociality. Journal of
Individual Differences, 38, 175-188.
Fetterman, A. K., Robinson, M. D., & Meier, B. P. (2021). Leveraging individual differences to
understand grounded procedures. Behavioral and Brain Sciences, 44, E6. doi:
10.1017/S0140525X20000382.
Fetterman, A. K., Wilkowski, B. M., & Robinson, M. D. (2018). On feeling warm and being
warm: Daily perceptions of physical warmth fluctuate with interpersonal warmth. Social
Psychological and Personality Science, 9, 560-567.
Fincher-Kiefer, R. (2019). How the body shapes knowledge: Empirical support for embodied
cognition. Washington, DC: American Psychological Association.
Gosling, S. (2008). Snoop: What your stuff says about you. New York, NY: Basic Books.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual
differences in implicit cognition: The Implicit Association Test. Journal of Personality
and Social Psychology, 74, 1464-1480.
Guan, F., Xu, Z., Zhang, B., Yan, W., & Peng, K. (in press). Bidirectional associations between
Tibetan Buddhism and vertical space. Psychology of Religion and Spirituality.
Hauser, D. J. & Schwarz, N. (in press). Situated embodiment: When physical weight does and
does not inform judgments of importance. Social Psychological and Personality Science.
Higgins, E. T. (1996). Knowledge activation: Accessibility, applicability, and salience. In E. T.
Higgins & A. W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (p.
133-168). New York, NY: The Guilford Press.

GOD IS UP? 27
Klein, R. A., Vianello, M., Hasselman, F., et al. (2018). Many labs 2: Investigating variation in
replicability across samples and settings. Advances in Methods and Practices in
Psychological Science, 4, 443-490.
Kövecses, Z. (2005). Metaphor in culture: Universality and variation. New York, NY:
Cambridge University Press.
Lakoff, G., & Johnson, M. (1980). Metaphors we live by. Chicago, IL: University of Chicago
Press.
Lakoff, G. & Johnson, M. (1999). Philosophy in the flesh: The embodied mind and its challenges
to western thought. New York, NY: Basic Books.
Landau, M. J. (2017). Conceptual metaphor in social psychology: The poetics of everyday life.
New York, NY: Routledge.
Landau, M. J., Meier, B. P., & Keefer, L. A. (2010). A metaphor-enriched social cognition.
Psychological Bulletin, 136, 1045-1067.
Lee, S. W. S. (2016). Multimodal priming of abstract constructs. Current Opinion in Psychology,
12, 37-44.
Martinez, A. M., & Benavente, R. (1998). The AR face database (CVC Tech. Rep. No. 24):
http://www2.ece.ohio-state.edu/~aleix/ARdatabase.html.
Meier, B. P., & Fetterman, A. K. (in press). Metaphors for God: God is high, light, and human in
implicit tasks. Psychology of Religion and Spirituality.
Meier, B. P., Fetterman, A. K., & Robinson, M. D. (2015). Black and white as valence cues: A
large-scale replication effort of Meier, Robinson, and Clore (2004). Social Psychology,
46, 174-178.

GOD IS UP? 28
Meier, B. P., Hauser, D. J., Robinson, M. D., Friesen, C. K., & Schjeldahl, K. (2007). What’s
“up” with God? Vertical Space as a representation of the divine. Journal of Personality
and Social Psychology, 93, 699-710.
Meier, B. P., Schnall, S., Schwarz, N., & Bargh, J. A. (2012). Embodiment in social psychology.
Topics in Cognitive Science, 4, 705-716.
Meyer, D. E., & Schvaneveldt, R. W. (1976). Meaning, memory structure, and mental processes.
Science, 192, 27-33.
O’Donnell, M., Nelson, L. D., Ackermann, E., Aczel, B., Akhtar, A., Aldrovandi, S., … Zrubka,
M. (2018). Registered replication report: Dijksterhuis and van Knippenberg (1998).
Perspectives on Psychological Science, 13, 268–294.
Persich, M. R., Steinemann, S., Fetterman, A. K., & Robinson, M. D. (in press). Drawn to the
light: Predicting religiosity using “God is light” metaphor. Psychology of Religion and
Spirituality.
Ransom, M. R. & Alicke, M. D. (2013). On bended knee: Embodiment and religious judgments.
Current Research in Social Psychology,
21,https://crisp.org.uiowa.edu/sites/crisp.org.uiowa.edu/files/2020-04/21.9.pdf.
Giessner, S. R., & Schubert, T. W. (2019). Power and vertical positions in an organization chart:
A pre-registered replication report of Study 3a and a modification of Study 1a, Giessner
& Schubert (2007). Collabra: Psychology, 5, 25. https://doi.org/10.1525/collabra.216
Siev, J., Zuckerman, S. E., & Siev, J. J. (2018). The relationship between immorality and
cleansing: A meta-analysis of the Macbeth Effect. Social Psychology, 49, 303-309.

GOD IS UP? 29
Skulmowski, A., & Rey, G. D. (2018). Adjusting sample sizes for different categories of
embodied cognition research. Frontiers in Psychology, 9, doi:
10.3389/fpsyg.2018.02384.
Soliman, T. M., Johnson, K. A., & Song, H. (2015). It’s not “all in your head”: Understanding
religion from an embodied cognition perspective. Perspectives on Psychological Science,
10, 852-864.
Van Cappellen, P., & Edwards, M. E. (2021). The embodiment of worship: Relations among
postural, psychological, and physiological aspects of religious practice. Journal for the
Cognitive Science of Religion, 6, 56–79.

GOD IS UP? 30
Footnotes
1Two data files for Study 2 were uploaded to OSF (https://osf.io/tgb9j/). One file presents
information from the study proper and the second presents age and gender information only. Due
to a programming error, that is, we could not link particular genders or ages to particular
participants who completed the substantive task for the study.