954856

research-article2020
PSSXXX10.1177/0956797620954856Landry et al.Difficult Turned Easy

ASSOCIATION FOR
Research Article PSYCHOLOGICAL SCIENCE
Psychological Science

Difficult Turned Easy: Suggestion Renders 2021, Vol. 32(1) 39­–49

© The Author(s) 2020
Article reuse guidelines:
a Challenging Visual Task Simple sagepub.com/journals-permissions
DOI: 10.1177/0956797620954856
https://doi.org/10.1177/0956797620954856
www.psychologicalscience.org/PS

Mathieu Landry1 , Jason Da Silva Castanheira2,

Jérôme Sackur1,3, and Amir Raz4,5,6
1
Laboratoire de Sciences Cognitives et Psycholinguistique, Département d’Études Cognitives, École
Normale Supérieure de Paris (ENS), École des Hautes Études en Sciences Sociales (EHESS), Centre
National de la Recherche Scientifique (CNRS), Université Paris Sciences et Lettres (PSL); 2Integrated
Program in Neuroscience, McGill University; 3Laboratoire Interdisciplinaire de l’X, École Polytechnique;
4
Department of Psychiatry, McGill University; 5Lady Davis Institute for Medical Research, Jewish General
Hospital, Montreal, Quebec, Canada; and 6Interdisciplinary Institute for Brain and Behavioral Sciences,
Chapman University

Abstract
Suggestions can cause some individuals to miss or disregard existing visual stimuli, but can they infuse sensory
input with nonexistent information? Although several prominent theories of hypnotic suggestion propose that mental
imagery can change our perceptual experience, data to support this stance remain sparse. The present study addressed
this lacuna, showing how suggesting the presence of physically absent, yet critical, visual information transforms an
otherwise difficult task into an easy one. Here, we show how adult participants who are highly susceptible to hypnotic
suggestion successfully hallucinated visual occluders on top of moving objects. Our findings support the idea that, at
least in some people, suggestions can add perceptual information to sensory input. This observation adds meaningful
weight to theoretical, clinical, and applied aspects of the brain and psychological sciences.

Keywords
suggestion, hypnosis, perception, visual imagery, perceptual integration, open data

Received 2/18/20; Revision accepted 6/12/20

Suggestions can dramatically alter how individuals pro- intimated that suggestions induce a response bias for
cess perceptual information (e.g., Lifshitz, Aubert Bonn, hallucination-prone individuals in noisy perceptual con-
Fischer, Kashem, & Raz, 2013), including the suppres- texts (Alganami, Varese, Wagstaff, & Bentall, 2017).
sion of visual inputs (Schmidt, Hecht, Naumann, & Accordingly, positive hallucinations may correspond to a
Miltner, 2017). Conversely, evidence remains ambigu- reinterpretation of the sensory experience rather than to
ous as to whether suggestions can reliably infuse novel genuine changes in the perceptual content. Research into
information into the perceptual stream. This ambiguity consciousness deals with a similar conundrum: Reports
contrasts with prominent theories that emphasize the of awareness may sometimes follow from a response bias
ability of hypnosis to generate perceptual experiences (Peters, Lau, & Ro, 2016). Some researchers have attempted
and hallucinations (e.g., Kirsch & Braffman, 2001; Martin to address this particular issue in the context of hypnotic
& Pacherie, 2019; Spiegel, 2003). Findings that support hallucinations by inducing synesthesia-like experiences
such viewpoints have often come with serious limitations, through posthypnotic suggestions and then validating the
however, such as reliance on self-reports that are prone effect with a challenging perceptual task (Anderson, Seth,
to bias and demand characteristics (e.g., Kirsch et al.,
2008), reverse inferences from brain imaging (e.g.,
Corresponding Author:
McGeown et al., 2012), small sample sizes, and anecdotal Amir Raz, Lady Davis Institute for Medical Research, Jewish General
case-study designs (e.g., Kallio & Koivisto, 2013). Further Hospital, Montreal, Quebec, Canada
highlighting these limitations, recent findings have E-mail: amir.raz@mcgill.ca
40 Landry et al.

Dienes, & Ward, 2014; Cohen Kadosh, Henik, Catena,

Walsh, & Fuentes, 2009; Kallio, Koivisto, & Kaakinen, Statement of Relevance
2017)—so far, however, with mixed results (Schwartzman,
Bor, Rothen, & Seth, 2019). Mounting evidence shows that hypnotic suggestion can
Considering these shortcomings, we here examined regulate various kinds of perceptual experiences, such
whether a suggestion to append novel information to as pain. Yet most of these findings involve reducing
perception can transform a difficult perceptual task into or suppressing an experience. In the present research,
an easy one. Our goal was to provide support for the we asked a complementary question: Can a hypnotic
idea that suggestion can instigate perceptual informa- suggestion enhance or increase perceptual experience?
tion endogenously while avoiding the aforementioned To test this question, we identified young adults who
limitations. To this end, we relied on occlusion-related scored especially high or low on a scale of hypnotic
perceptual integration of object motion, in which the pres- suggestibility. We then provided these individuals with
ence of shape stimuli at the apex of moving lines pro- the suggestion that they would be able to perceive
duces the percept of an occluded figure performing a phantom (i.e., nonexistent) geometric shapes on a
circular revolution around a central axis (Fig. 1a; Lorenceau computer screen while completing a visuospatial task.
& Shiffrar, 1992). Critically, this particular percept van- Our experimental approach rested on the idea that
ishes whenever the occluding shape stimuli are removed being able to imagine these geometric shapes on the
from the display, making it nearly impossible to see the screen would benefit participants’ performance on this
geometric figure and the direction of the revolution otherwise difficult task. Our results were consistent with
without the occluders. We accordingly examined this prediction and show that the suggestion improved
whether a suggestion to imagine the occluders would performance of individuals who scored high on the
allow individuals who exhibit greater sensitivity to sug- suggestibility scale while having little effect on those
gestions, referred to here for convenience as “highly who scored low. These findings imply that individuals
suggestible individuals,” to experience perceptual inte- susceptible to hypnotic suggestions are capable of
gration of the line stimuli and thereby perceive the creating novel perceptual experiences.
geometric figure. We compared their performance with
that of individuals who showed less sensitivity to sugges-
tion, referred to here for convenience as “less-suggestible years, SD = 2.27) had normal or corrected-to-normal
individuals,” as well as with several cohorts of control vision and received course credit in exchange for par-
participants who completed the task both online and ticipation. See Figure S1 in the Supplemental Material
within our laboratory. for a diagram describing the groups and the corre-
sponding experimental conditions.
We had little information regarding the effect size of
Method the experimental suggestion on highly suggestible indi-
viduals for this visual task, so we based our sample size
Participants
for the suggestion conditions on a collection of studies
Drawing from a convenience sample, we prescreened from our own group that similarly investigated the influ-
individuals for hypnotic susceptibility from a pool of ence of hypnotic suggestion on perception and cognition
approximately 500 students in psychology classes at (for a review, see Lifshitz et al., 2013). We reasoned that
McGill University, using the Harvard Group Scale of a meaningful effect size should be at least comparable
Hypnotic Susceptibility, Form A (HGSHS:A; Shor & with and easily detectable with a sample of the same size.
Orne, 1962). Our final sample consisted of 16 highly Following this rationale, we pooled data from our pre-
suggestible individuals (i.e., HGSHS:A score > 8) and vious studies and performed simulations to estimate
16 less-suggestible individuals (i.e., HGSHS:A score < the minimal sample size required to achieve a power
4). We recruited additional participants, not screened of .8 for the detection of the effect of suggestion in
for hypnotic suggestibility, who completed the task highly suggestible individuals at an α of .05 (see the
without occluders present and without receiving hyp- Supplemental Material for details). This procedure
notic suggestion—14 completed the task in our labora- revealed a modest effect size of hypnotic suggestion in
tory and 186 online. To ascertain possible learning highly suggestible individuals (i.e., marginal R 2 for this
effects, we invited 49 random participants, who com- hierarchical regression model = .14), and 13 highly
pleted the task online, to a second session in our labo- suggestible individuals were required to attain a power
ratory. Two additional samples performed the task with level of .8 for an α of .05. Our sample size aligned with
occluders present—that is, 46 participants completed these parameters.
the task online, and 17 completed it in the laboratory. When participants performed the task without
All participants (N = 295; 215 women; mean age = 20.81 occluders during controlled conditions, we aimed to
 a
Task Without Hypnotic Suggestion
Diamond Triangle

Task With Hypnotic Suggestion to Imagine the Visual Occluders

Diamond Triangle

b c
∗ Fixed-Effects Coefficient (β)
Less-
.9 Suggestible −1 −0.5 0 0.5 1 1.5 2
Individuals Intercept
.7 Hypnotic Suggestibility
Highly
Suggestible Suggestion
.5 Individuals Shape
Hypnotic Suggestibility × Suggestion
.3

Accuracy Rate (%)

Hypnotic Suggestibility × Shape
.1 Suggestion × Shape
Hypnotic Suggestibility × Suggestion × Shape
t ith on t h n
ou n W esti ou n it tio
ith estio ith estio W ges
W g gg W g g
g Su g Su

41
Su Su
Fig. 1. (continued on next page)
42 Landry et al.

Fig. 1. Experimental design, accuracy for the highly suggestible and less-suggestible groups, and model results. The experimental task (a)
consisted of four or three moving lines that formed the shape of a diamond, a square, a triangle, or an inverted triangle. Diamonds and squares
moved only in a clockwise or counterclockwise fashion, whereas triangles and inverted triangles could move clockwise, counterclockwise, or
with a directionless motion (i.e., neither clockwise nor counterclockwise; arrows in the figure indicate the direction of motion). The task was
presented without any visual occluders (as shown in the top row), but two groups of participants (highly suggestible and less suggestible)
underwent hypnotic suggestion, in which they were instructed to imagine the occluders at the vertices of the moving lines (as indicated by
the dashed areas in the bottom row). Participants indicated the direction of the moving shape (i.e., clockwise, counterclockwise, or direction-
less motion). (Movies of diamond trials with and without occluders are available in the Supplemental Material.) Discrimination accuracy (b) is
shown separately for highly suggestible and less-suggestible individuals, with and without suggestion, across shape trials. Black dots represent
average accuracy rates, and error bars correspond to bootstrapped 95% confidence intervals. Gray dots represent individual performance. The
asterisk indicates a significant difference between suggestion conditions (p < .05). Coefficients (c) are shown from the single-trial hierarchi-
cal logistic regression model for predicting accuracy. These coefficients are from the best-fitting model following chi-square goodness-of-fit
statistics over the deviance and following the Bayesian information criterion. The red rectangle highlights the statistically reliable interaction
between hypnotic suggestibility and suggestion.

recruit as many online participants as possible from 45° to create a square stimulus and flipped the triangle
psychology classes at McGill University. In contrast, the to create an inverted-triangle stimulus. We randomly var-
subset of individuals asked to complete the control task ied the order of these stimuli on the screen across trials
in our laboratory was comparable in size with both our to discourage participants from replacing the occluders
highly suggestible and less-suggestible groups. The with physical objects affixed to the screen (e.g.,
sample size for participants who completed the task stickers).
twice merely followed from the limited potency of the A second version of the task contained fully visible
learning effects, which we had observed in the preced- occluders in white (HCC = ffffff; RGB value = 255, 255,
ing pilot experiments. Here, we aimed to have a large- 255). When participants were sitting approximately 45
enough sample to assess any potential effect, yet the cm away from the screen, the width and height of the
effect size was quite modest (d = 0.28). Last, given that lines for the diamond and square stimuli approximated
the performance was at ceiling in the presence of 5.7° and 1.3° of visual angle, respectively, and the
occluders, we aimed for a sample size akin to that of square occluders measured roughly 6.3°. For the dia-
the suggestion group to ensure a proper comparison. mond and square stimuli, the length of lines approxi-
Both our online and laboratory samples met this crite- mated 7.6°, and the pentagon occluders were estimated
rion. All procedures were approved by the local insti- at 8.8°. All target stimuli were centered, and they rotated
tutional review board. in a circle around the fixation point (see Fig. 1).
Throughout the task, the diamonds and squares
moved only in a clockwise or counterclockwise fashion,
Task and procedure whereas the triangles and inverted triangles could move
We constructed a Web-based Adobe Flash task and sent clockwise, counterclockwise, or with a directionless
e-mail invitations to participants with the URL. We motion (i.e., neither clockwise nor counterclockwise).
designed the task—which we called “MoTraK”—based Note that the directionless motion could therefore occur
on the paradigm of occlusion-related perceptual inte- only for the upward and inverted triangles. For direc-
gration of object motion (Fig. 1a; Lorenceau & Shiffrar, tionless-motion trials, the shape would move around
1992). The task was composed of trials with moving the fixation point without following a specific trajectory
occluded diamonds, squares, triangles, and inverted while repeatedly expanding and then shrinking in size.
triangles. The task accordingly involved 72 outlines of We included the directionless motion for triangles and
each geometric shape in motion—18 trials for dia- inverted triangles as catch trials. Participants were
monds, 18 for squares, 18 for triangles, and 18 for aware of these contingencies.
inverted triangles—with vertices occluded by shapes Our Adobe Flash interface recorded the trials and
that matched the color of the background. Subse- immediately sent the measures to a password-protected
quently, only segments of the geometric outlines were MySQL online database. The program recorded
visible (i.e., four straight line segments on diamond and responses when participants pressed the “F” key, “J”
square trials and three straight segments on triangle key, or space bar to indicate counterclockwise, clock-
and inverted-triangle trials). We relied on homogenous wise, and directionless motion, respectively, for trian-
colors: uniform gray for the lines (hexadecimal color gles and inverted triangles. Participants completed the
code [HCC] = 666666; RGB value = 102, 102, 102) and task in two separate blocks: The first block consisted
black for the background (HCC = 000000; RGB value = only of diamond and square trials, and the second
0, 0, 0), resulting in medium contrast, which creates a block consisted only of triangle and inverted-triangle
low coherence of motion. We rotated the diamond by trials. We opted for this design because we wanted to
Difficult Turned Easy 43

avoid confusion and ensure that participants considered that participants would receive a hypnotic suggestion.
the response option of directionless motion only during The experimenter then escorted participants to a sepa-
triangle and inverted-triangle trials. rate room to meet with one of the authors (A. Raz), a
To ensure that participants understood the task well, researcher with more than 30 years of experience work-
we included two short training sessions in the preassess- ing with hypnosis and a diplomat of the American Board
ment of MoTraK. During the first training session, par- of Psychological Hypnosis. A. Raz administered a hyp-
ticipants went through consecutive 15-s interactive notic induction adapted from the Carleton University
demonstrations, in which they could make the occluders Responsiveness to Suggestion Scale (Spanos, Radtke,
visible or invisible on a pentagon shape that moved first Hodgins, Stam, & Bertrand, 1983). He then suggested
clockwise, then counterclockwise. Using a pentagon for to all participants that they would be able to view the
training prevented exposure to the actual stimuli prior occluders at the vertices of the moving lines while play-
to data collection. Next, participants practiced on a few ing MoTraK and that this hallucination would allow
trials, and they were given feedback stating whether their them to perform the task quickly and easily. (A script
responses were correct or incorrect. These practice trials of the suggestion is available in the Supplemental Mate-
consisted of six pentagon trials—three clockwise and rial.) Induction and suggestion took about 10 min; par-
three counterclockwise (pseudorandomized). After the ticipants then completed the task. When the task was
practice trials, we informed participants that they would over, A. Raz administered a standard hypnotic termina-
no longer receive feedback. The second training block tion. The experimenter then escorted participants out
occurred between the diamond/square and the triangle/ of the room for debriefing. Thus, we tested participants
inverted-triangle blocks, during which participants under two conditions: first at baseline without sugges-
viewed a single interactive demonstration of direction- tion and then with a specific suggestion to perceive
less motion on a pentagon. To ascertain comprehension, phantom occluders covering the otherwise uncovered
we included no demonstrations in the postassessment corners. Note that A. Raz was blind as to participants’
of MoTraK, only three practice trials with feedback. susceptibility to hypnosis.
Instructions emphasized both speed and accuracy. The
entire task lasted about 15 min. Procedure for online participants
We provided all participants with the URL for MoTraK
Procedure for highly suggestible and
and asked them to complete the task online. A written
less-suggestible groups notice in the task asked them to complete it in a calm
First, for testing performance without suggestion, we environment, free from distractions. Participants pro-
sent all potential participants an e-mail providing them vided consent by clicking on the “Accept” button follow-
with the URL of the Web page hosting MoTraK and ing the consent information. We gathered demographic
inviting them to complete the task online in a calm information, student identification numbers, and IP
environment of their choice. The online consent forms addresses to avoid repeated participation.
informed participants that they had the right to with-
draw from the study at any time and that the data Procedure for participants in the
gathered, including response time and accuracy, would
laboratory
be used only for scientific research. We collected demo-
graphic information as well as Internet protocol (IP) In the laboratory, the experimenter greeted participants
addresses, which allowed us to identify and exclude and led them into a quiet room containing a computer.
participants who completed the task more than once. The experimenter sat beside participants to monitor
In addition, MoTraK automatically assigned a random their engagement and ensure that they refrained from
number (a unique completion code) to each partici- utilizing alternative strategies while performing the task
pant. This number was required to complete the postas- (i.e., participants were to remain seated in a stable and
sessment. During the first session without suggestion, appropriate position, looking forward with their eyes
participants were unaware that this research involved open normally and at the target without averting their
hypnotic suggestion. This strategy minimized the poten- gaze, at an approximate distance of 45 cm from the
tial influence of holdback effects. screen).
Approximately a week after their online participation, Participants who completed the task twice received
we sent participants an e-mail inviting them to partici- an automatically generated e-mail inviting them to par-
pate in a second session at our laboratory. On arriving ticipate once again in our study, either online or at our
at the laboratory, participants were greeted by an exper- laboratory. The purpose of this invitation was to control
imenter who obtained informed consent and disclosed for learning effects. Moreover, an additional group of
44 Landry et al.

participants completed the task with white occluders that of individuals who performed the task with occlud-
present. We expected this experiment to yield ceiling ers present. Again, one sample performed the task
effects across participants, because the percept effortlessly online and another in our laboratory. The purpose of
emerges as soon as the occluders become visible. this control condition was to accurately gauge perfor-
mance when participants endogenously hallucinated the
occluders compared with performance when the occlud-
Analysis
ers were actually present. In this way, we contrasted
We removed anticipation (response time < 150 ms) and how visual imagery measured up against the actual per-
timeout (response time > 3 SDs from the mean). Over- ception of the occluders. Note that we further computed
all, anticipation trials corresponded to less than 1% of the Jeffrey-Zellner-Siow ( JZS) Bayes factor (BF) to eval-
total trials, whereas timeout trials represented approxi- uate evidence in favor of the null hypothesis using the
mately 1% of total trials. No additional observations default Cauchy r scaling value of .707 (Rouder, Speck-
were removed from analysis. We gauged overall per- man, Sun, Morey, & Iverson, 2009). Bootstrapped con-
formance using hierarchical single-trial logistic regres- fidence intervals were computed using MATLAB’s
sion predicting accuracy for each trial (i.e., correct vs. Bootfun algorithm.
incorrect discrimination). Hypnotic suggestibility (low
vs. high), suggestion (with vs. without), shape (squares
and diamonds vs. triangles and inverted triangles), and Results
their interactions were included as fixed factors, and Comparison of highly suggestible and
the participants were included as random factors.
MATLAB (Version R2017B; The MathWorks, Natick, MA)
less-suggestible groups
and the fitglme function were used to fit all regression The performance of highly suggestible individuals
models. We opted for the Laplace fitting method and improved significantly, compared with that of less-
selected the best-fitting model using goodness-of-fit suggestible individuals, for whom the suggestion made
chi-square tests over deviance (α = .05) and by evaluat- little difference. We tested the efficiency of the hypnotic
ing the Bayesian information criterion. Post hoc evalu- suggestion to add new perceptual information (i.e.,
ations were performed using pairwise permutation t visualizing the occluders) by evaluating accuracy rates
tests (i.e., 10,000 permutations). across all trials—those involving diamond, square, tri-
We similarly compared task improvements for highly angle, and inverted triangle shapes—through hypnotic
suggestible individuals with performance from other suggestibility and suggestion conditions (Fig. 1b). Here,
cohorts in different control conditions. We first com- we relied on single-trial logistic regression (see Analysis
pared the performance of highly suggestible individuals section). Fixed factors were included in a stepwise
before and after receiving the suggestion with the per- approach. Our results showed that the best-fitting
formance of individuals who completed the same task model included suggestion (β = 0.35, SE = 0.127, 95%
online and in the laboratory. We relied on nonparamet- CI = [0.1, 0.597]), the Hypnotic Suggestibility × Sugges-
ric two-tailed permutation tests (i.e., 10,000 permuta- tion interaction (β = 1.22, SE = 0.184, 95% CI = [0.857,
tions) to compare mean accuracy rates. 1.58]), the Suggestion × Shape interaction (β = −0.471,
Our goal was twofold: first, to validate that perfor- SE = 0.18, 95% CI = [−0.824, −0.118]), and the Hypnotic
mance was no different between highly suggestible Suggestibility × Suggestion × Shape interaction (β =
individuals and a matched-controlled group prior to 0.69, SE = 0.26, 95% CI = [0.18, 1.2]) as reliable predic-
receiving the suggestion and, second, to demonstrate tors (see Fig. 1c, as well as Tables S1 and S2 in the
that the improvement in highly suggestible individuals Supplemental Material, for details). Following the Hyp-
marked a significant departure from baseline perfor- notic Suggestibility × Suggestion interaction, post hoc
mance following suggestions. One group of participants pairwise permutation tests confirmed limited benefits
also performed the task twice, once online and another between conditions with suggestion and without sug-
time in the laboratory, which allowed us to assess learn- gestion for less-suggestible individuals (M = .46, SD =
ing effects and underline how the improvement seen .17 without suggestion; M = .48, SD = .23 with sugges-
for highly suggestible individuals related to that of tion), t(15) = 0.65, p = .53, JZS BF = 3.25, whereas we
learning. Here, we accordingly contrasted the differ- rejected the null hypothesis for highly suggestible
ence in performance between the first and second ses- individuals when comparing performance with and
sion for this control group and the performance with without suggestion (M = .36, SD = .15 with suggestion;
suggestion minus the performance without suggestion M = .72, SD = .22 without suggestion), t(15) = 5.14, p <
for highly suggestible individuals. Last, we compared .001, JZS BF = 239.88. These results are therefore con-
the performance of highly suggestible individuals with sistent with our primary research objective and provide
Difficult Turned Easy 45

evidence for the hypothesis that the experimental sug- suggestion and significantly improved their perfor-
gestion would change how highly suggestible individu- mance with the suggestion, which further highlights
als process perceptual information and subsequently how suggesting the presence of occluded shapes
improve their performance. Note that our analyses fur- improved performance on an otherwise difficult task.
ther confirmed that the Hypnotic Suggestibility × Sug-
gestion interaction was reliable for both square/ Comparison of highly suggestible
diamond and triangle/inverted-triangle trials separately
(see Tables S3 and S4 in the Supplemental Material).
individuals with the control condition
Moreover, we further controlled for conservative strate- for repeated sessions
gies and the tendency to indicate motionless direction We also sought to assess learning effects on the task.
between highly suggestible and less-suggestible indi- Here, we aimed to corroborate that the benefits we
viduals for diamond trials. This analysis showed no observed for highly suggestible individuals followed
difference between both groups. from the suggestion and not from learning. Note that
Less-suggestible individuals served as a control the less-suggestible individuals already provided infor-
group for highly suggestible individuals because they mation to this effect, because they completed the task
performed the exact same experiment. However, we under the same experimental conditions as the highly
wanted to gauge the benefits of suggestions on highly suggestible individuals; however, we aimed for further
suggestible individuals by comparing additional control confirmation with a larger sample. A separate group of
conditions. In particular, we looked at baseline perfor- participants therefore completed the task twice, once
mance when participants completed the task online online and later in our laboratory (Fig. 3a). We first
(n = 186) and in our laboratory (n = 14). This way, we evaluated evidence of improvement for this control
could further certify suggestion-related improvements group with a pairwise permutation t test over accuracy
for highly suggestible individuals against a larger sam- across the first and second sessions (M = .33, SD = .22
ple. We similarly investigated learning effects in a group for the first session; M = .39, SD = .26 for the second
of individuals (n = 49) who performed the task twice, session), t(48) = 1.92, p = .06, JZS BF = 1.51. Thus,
because highly suggestible individuals also completed evidence favored the null hypothesis, showing that this
the task on two occasions. Last, we also compared highly group showed little improvement from the first to the
suggestible individuals who imagined the presence of second session. Comparing the perceptual benefits
the occluders with participants who played MoTraK with from both the highly suggestible individuals (i.e., per-
the occluders physically present (n = 46 online and n = formance with suggestion minus performance without
17 in our laboratory), thereby comparing veridical per- suggestion) and this control group (i.e., performance
ception with suggestion-induced visual imagery. on the second session minus performance on the first)
further corroborated the gain conferred by the sugges-
Comparison of highly suggestible tion, as we observed a greater increase in performance
for highly suggestible individuals (observed mean dif-
individuals with control condition ference = .31, p < .001, d = 1.26; Fig. 3c). These results,
without occluders therefore, imply that highly suggestible individuals’
We evaluated the performance of highly suggestible improvement on the task did not follow from practice
individuals across sessions, with and without sugges- effects.
tion, against the performance of the online and labora-
tory groups who completed the task without occluders Comparison of highly suggestible
(Fig. 2a). Here, we relied on permutation tests over the individuals with control condition
mean accuracy rate of each group. Without suggestion,
highly suggestible individuals performed similarly to
with occluders
both the online (observed mean difference = −.0099; Last, we wanted to evaluate how visual imagery of the
p = .86; d = −0.06; Fig. 2b) and laboratory groups occluders induced by the suggestion in highly suggest-
(observed mean difference = −.032; p = .67; d = −0.16; ible individuals fared against the actual presence of the
Fig. 2b). Conversely, following suggestion, highly sug- occluding stimuli. One group of participants completed
gestible individuals performed better than the online the task online and another in the laboratory with
group (observed mean difference = .35; p < .001; d = occluding stimuli located at the vertices of the moving
1.67; Fig. 2c) and the laboratory group (observed mean lines. The presence of the occluding stimuli yielded
difference = .34; p < .001; d = 1.45; Fig. 2c). Together, ceiling effects for discrimination accuracy rates (Fig.
both analyses convey that highly suggestible individuals 3b). Thus, as one would expect, the comparison between
performed similarly to the baseline groups without the visual imagery and veridical perception of the occluders
46 Landry et al.

a b
Control Condition Accuracy for Highly Suggestible Individuals
Without Occluders Without Suggestion vs. Controls Without Occluders
Controls Online Controls in the Lab

.9
400

.7
Accuracy Rate (%)

.5

Count
200

.3

.1
0
Online Lab −.2 0 .2 −.2 0 .2
Random Mean Difference

c
Accuracy for Highly Suggestible Individuals
With Suggestion vs. Controls Without Occluders
Controls Online Controls in the Lab

400
Count

200

0
−.4 0 .4 −.2 0 .2 .4
Random Mean Difference
Fig. 2. Performance of highly suggestible individuals who received and did not receive hypnotic suggestion, compared with per-
formance in the baseline control condition. Discrimination accuracy (a) is shown for the baseline control condition in which visual
occluders were absent, separately for participants who completed the task online and in the laboratory. Gray dots represent average
individual accuracy, black dots represent average group accuracy, and error bars correspond to 95% bootstrapped confidence intervals.
Null distributions of random permutations are shown for mean comparisons of accuracy rates between control participants and (b)
highly suggestible individuals who did not receive hypnotic suggestion and (c) highly suggestible individuals who received hypnotic
suggestion. In (b) and (c), distributions are shown separately for comparisons in which the control group completed the task online
and in the laboratory. Red bars indicate observed differences.

revealed that, despite the significant performance improve- the display (observed mean difference with online
ment of highly suggestible individuals following sugges- group = −.24, p < .001, d = −1.69; observed mean differ-
tion, this benefit remained lower than that seen in both ence with laboratory group = −.23, p < .001, d = −1.64;
groups who completed the task with occluding stimuli in see Fig. 3d). Evidence therefore supports the notion that
Difficult Turned Easy 47

a c
Control Condition Accuracy for Highly Suggestible Individuals
Without Occluders (Repeat Sessions) Across Sessions vs. Repeated Sessions in
Controls Without Occluders
.9
400
Accuracy Rate (%)

.7

.5

Count
200
.3

.1
0
Online Lab −.2 0 .2 .4
Random Mean Difference

b d
Control Condition Accuracy for Highly Suggestible Individuals
With Occluders With Suggestion vs. Controls With Occluders
Controls Online Controls in the Lab
.9

300
Accuracy Rate (%)

.7

.5 200
Count

.3
100

.1
0
Online Lab −.2 −.1 0 .1 −.2 −.1 0 .1 .2
Random Mean Difference
Fig. 3. Performance of highly suggestible individuals who received hypnotic suggestion, compared with performance in the control
condition across repeated sessions and when visual occluders were present. Discrimination accuracy is shown for control conditions
in which individuals completed the task (a) when visual occluders were absent and (b) when visual occluders were present. Partici-
pants who completed the task when occluders were absent did so twice, first online and then in the laboratory, whereas those who
completed the task when occluders were present did so only once, either online or in the lab. Gray dots represent average individual
accuracy, black dots represent average group accuracy, and error bars correspond to 95% bootstrapped confidence intervals. Null
distributions of random permutations are shown for mean comparisons of accuracy rates (c) between highly suggestible individuals
across sessions (i.e., performance with suggestion minus performance without suggestion) and control participants who completed
the task twice without receiving hypnotic suggestion (performance on the second session minus performance on the first session)
and (d) between highly suggestible individuals who received hypnotic suggestion and control participants who completed the task
with occluders only once. In (d), distributions are shown separately for comparisons in which the control group completed the task
online and in the laboratory. Red bars indicate observed differences.

the suggestion conveyed reliable perceptual benefits, Discussion

albeit the subjective experience of visualizing the
occluders with suggestion remains substantively dif- Here, we showed that a hypnotic suggestion to see
ferent from actually seeing them. nonexistent occluders improves the performance of
48 Landry et al.

highly suggestible individuals on a challenging visual Open Practices

task. Our findings intimate that the suggestion afforded All data and codes for analyses have been made publicly
highly suggestible individuals with the capacity to expe- available via OSF and can be accessed at https://osf
rience perceptual integration by conjuring the presence .io/25rpx/. The design and analysis plans for the experi-
ment were not preregistered. This article has received the
of the occluders via endogenous means. These influ-
badge for Open Data. More information about the Open
ences, fueled by a suggestion to add visual information
Practices badges can be found at http://www.psychological
to the perceptual stream, yoked top-down processes science.org/publications/badges.
driven by expectation and mindset with bottom-up pro-
cessing mostly driven by sensory inputs. The improve-
ment of highly suggestible relative to less-suggestible
individuals, alongside data from multiple control condi-
ORCID iD
tions, supports this idea—and yet suggestion-based per-
formance hardly reached that measured when occluders Mathieu Landry https://orcid.org/0000-0002-2235-9060
were present. Imagery therefore appears weaker than
actual perception. Acknowledgments
Although generalization to other perceptual pro- We thank Noémie Aubert-Bonn and Alexandra Fischer for
cesses goes beyond the present data, our findings com- their valuable contributions in coding the task, collecting the
plement other reports that document how expectation data, and performing a primary exploration of the results.
and cognition can govern stimulus-driven processes
(Szechtman, Woody, Bowers, & Nahmias, 1998). Our Supplemental Material
results accordingly confirm the reliability of this frame-
Additional supporting information can be found at http://
work to shed light on mental imagery and perceptual journals.sagepub.com/doi/suppl/10.1177/0956797620954856
hallucinations. However, it remains uncertain whether
the current experimental context applies to other forms
References
of atypical perception, such as those observed in clini-
cal disorders. Still, our work paves the way for a more Alganami, F., Varese, F., Wagstaff, G. F., & Bentall, R. P.
scientific understanding of suggestion to elucidate (2017). Suggestibility and signal detection performance in
hallucination-prone students. Cognitive Neuropsychiatry,
mind–body phenomena, including the mechanisms
22, 159–174.
underlying the influence of placebos, symbolic think-
Anderson, H. P., Seth, A. K., Dienes, Z., & Ward, J. (2014).
ing, and expectancy. Can grapheme-color synesthesia be induced by hypno-
sis? Frontiers in Human Neuroscience, 8, Article 220.
Transparency doi:10.3389/fnhum.2014.00220
Action Editor: Karen Rodrigue Cohen Kadosh, R., Henik, A., Catena, A., Walsh, V., & Fuentes,
Editor: Patricia J. Bauer L. J. (2009). Induced cross-modal synaesthetic experience
Author Contributions without abnormal neuronal connections. Psychological
A. Raz designed the experiment, collected the data, and Science, 20, 258–265.
performed an initial exploration of the results with help Kallio, S., & Koivisto, M. (2013). Posthypnotic suggestion
from two students in the lab (see Acknowledgments). M. alters conscious color perception in an automatic manner.
Landry analyzed the data in depth and wrote a draft of the International Journal of Clinical & Experimental Hypnosis,
manuscript. M. Landry, J. Da Silva Castanheira, J. Sackur, 61, 371–387. doi:10.1080/00207144.2013.810446
and A. Raz wrote the final version of the manuscript and Kallio, S., Koivisto, M., & Kaakinen, J. K. (2017). Synaesthesia-
approved it for submission. type associations and perceptual changes induced by
Declaration of Conflicting Interests hypnotic suggestion. Scientific Reports, 7, Article 17310.
The author(s) declared that there were no conflicts of doi:10.1038/s41598-017-16174-y
interest with respect to the authorship or the publication Kirsch, I., & Braffman, W. (2001). Imaginative suggestibility
of this article. and hypnotizability. Current Directions in Psychological
Funding Science, 10, 57–61.
This work was supported by a grant from the Canadian Kirsch, I., Mazzoni, G., Roberts, K., Dienes, Z., Hallquist,
Institute for Health Research, the Bial Foundation, and the M. N., Williams, J., & Lynn, S. J. (2008). Slipping into
Canada Research Chair program to A. Raz. M. Landry was trance. Contemporary Hypnosis, 25, 202–209.
supported by a fellowship from Fonds de Recherche du Lifshitz, M., Aubert Bonn, N., Fischer, A., Kashem, I. F., & Raz, A.
Québec – Nature et Technologies and by the Bial Founda- (2013). Using suggestion to modulate automatic processes:
tion. A. Raz and M. Landry secured funding for this From Stroop to McGurk and beyond. Cortex, 49, 463–473.
research project. J. Da Silva Castanheira was supported by Lorenceau, J., & Shiffrar, M. (1992). The influence of termina-
a fellowship from the Natural Sciences and Engineering tors on motion integration across space. Vision Research,
Research Council of Canada (NSERC). 32, 263–273.
Difficult Turned Easy 49

Martin, J.-R., & Pacherie, E. (2019). Alterations of agency in Schwartzman, D. J., Bor, D., Rothen, N., & Seth, A. K. (2019).
hypnosis: A new predictive coding model. Psychological Neurophenomenology of induced and natural synaes-
Review, 126, 133–152. doi:10.1037/rev0000134 thesia. Philosophical Transactions of the Royal Society B,
McGeown, W. J., Venneri, A., Kirsch, I., Nocetti, L., Roberts, K., 374(1787), Article 20190030. doi:10.1098/rstb.2019.0030
Foan, L., & Mazzoni, G. (2012). Suggested visual halluci- Shor, R. E., & Orne, E. C. (1962). Harvard Group Scale of
nation without hypnosis enhances activity in visual areas Hypnotic Susceptibility, Form: A. Palo Alto, CA: Consulting
of the brain. Consciousness and Cognition, 21, 100–116. Psychologists Press.
Peters, M. A. K., Lau, H., & Ro, T. (2016). Who’s afraid of Spanos, N. P., Radtke, H. L., Hodgins, D. C., Stam, H. J., & Bertrand,
response bias? Neuroscience of Consciousness, 2016(1), L. D. (1983). The Carleton University Responsiveness to
Article niw001. doi:10.1093/nc/niw001 Suggestion Scale: Normative data and psychometric proper-
Rouder, J. N., Speckman, P. L., Sun, D., Morey, R. D., & ties. Psychological Reports, 53, 523–535.
Iverson, G. (2009). Bayesian t tests for accepting and Spiegel, D. (2003). Negative and positive visual hypnotic
rejecting the null hypothesis. Psychonomic Bulletin & hallucinations: Attending inside and out. International
Review, 16, 225–237. Journal of Clinical & Experimental Hypnosis, 51, 130–146.
Schmidt, B., Hecht, H., Naumann, E., & Miltner, W. H. R. Szechtman, H., Woody, E. Z., Bowers, K. S., & Nahmias, C.
(2017). The power of mind: Blocking visual perception by (1998). Where the imaginal appears real: A positron emission
hypnosis. Scientific Reports, 7, Article 4889. doi:10.1038/ tomography study of auditory hallucinations. Proceedings
s41598-017-05195-2 of the National Academy of Sciences, USA, 95, 1956–1960.